Rapid and non-destructive characterization of post-harvest quality of seeds, fruits, and vegetables is paramount to increasing crop profitability for growers and nutritious value and product quality for customers. Furthermore, simulation techniques with measurements of tissue characteristics can establish reference datasets providing scientists useful information for understanding of tissue development at the fundamental level. Our lab has developed non-destructive sensing techniques (such as: hyperspectral imaging, electronic nose, and thermal imaging) for various crop postharvest quality sensing such as blueberry internal bruising, onion postharvest diseases, and cotton fiber foreign matter.
Awards
2014, Best Paper Award from the Information and Electrical Technologies (IET) division of the American Society of Agricultural and Biological Engineers (ASABE)
Hyperspectral imaging
2020
Jiang, Y.; Snider, J. L.; Li, C.; Rains, G. C.; Paterson, A. H.
Ground based hyperspectral imaging to characterize canopy-level photosynthetic activities Journal Article
In: Remote Sensing, vol. 12, no. 2, pp. 315, 2020.
@article{Jiang2020,
title = {Ground based hyperspectral imaging to characterize canopy-level photosynthetic activities},
author = {Jiang, Y. and Snider, J. L. and Li, C. and Rains, G. C. and Paterson, A. H. },
year = {2020},
date = {2020-02-01},
journal = {Remote Sensing},
volume = {12},
number = {2},
pages = {315},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
Zhang, M.; Jiang, Y.; Li, C.; Yang, F.
Fully convolutional networks for blueberry bruising and calyx segmentation using hyperspectral transmittance imaging Journal Article
In: Biosystems Engineering, vol. 192, pp. 159-175, 2020.
@article{Zhang2019b,
title = {Fully convolutional networks for blueberry bruising and calyx segmentation using hyperspectral transmittance imaging },
author = {M. Zhang and Y. Jiang and C. Li and F. Yang},
url = {https://www.sciencedirect.com/science/article/pii/S1537511020300301?dgcid=author},
doi = {https://doi.org/10.1016/j.biosystemseng.2020.01.018},
year = {2020},
date = {2020-01-27},
urldate = {2020-01-27},
journal = {Biosystems Engineering},
volume = {192},
pages = {159-175},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
2018
Fan, S.; Li, C.; Huang, W.; Chen, L.
Data fusion of two hyperspectral imaging systems with complementary spectral sensing ranges for blueberry bruising detection Journal Article
In: Sensors, vol. 18, no. 12, pp. 4463, 2018.
@article{Fan201801,
title = {Data fusion of two hyperspectral imaging systems with complementary spectral sensing ranges for blueberry bruising detection},
author = {S. Fan and C. Li and W. Huang and L. Chen},
url = {http://sensinglab.engr.uga.edu//srv/htdocs/wp-content/uploads/2019/11/Data-fusion-of-two-hyperspectral-imaging-systems-with-complementary-spectral-sensing-ranges-for-blueberry-bruising-detection.pdf},
doi = {https://doi.org/10.3390/s18124463},
year = {2018},
date = {2018-12-17},
urldate = {2018-12-17},
journal = {Sensors},
volume = {18},
number = {12},
pages = {4463},
abstract = {Fan, S., Li, C., Huang, W., & Chen, L. (2018). Data fusion of two hyperspectral imaging systems with complementary spectral sensing ranges for blueberry bruising detection. Sensors, 18(12), 4463.
Currently, the detection of blueberry internal bruising focuses mostly on single hyperspectral imaging (HSI) systems. Attempts to fuse different HSI systems with complementary spectral ranges are still lacking. A push broom based HSI system and a liquid crystal tunable filter (LCTF) based HSI system with different sensing ranges and detectors were investigated to jointly detect blueberry internal bruising in the lab. The mean reflectance spectrum of each berry sample was extracted from the data obtained by two HSI systems respectively. The spectral data from the two spectroscopic techniques were analyzed separately using feature selection method, partial least squares-discriminant analysis (PLS-DA), and support vector machine (SVM), and then fused with three data fusion strategies at the data level, feature level, and decision level. The three data fusion strategies achieved better classification results than using each HSI system alone. The decision level fusion integrating classification results from the two instruments with selected relevant features achieved more promising results, suggesting that the two HSI systems with complementary spectral ranges, combined with feature selection and data fusion strategies, could be used synergistically to improve blueberry internal bruising detection. This study was the first step in demonstrating the feasibility of the fusion of two HSI systems with complementary spectral ranges for detecting blueberry bruising, which could lead to a multispectral imaging system with a few selected wavelengths and an appropriate detector for bruising detection on the packing line.},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
Fan, S., Li, C., Huang, W., & Chen, L. (2018). Data fusion of two hyperspectral imaging systems with complementary spectral sensing ranges for blueberry bruising detection. Sensors, 18(12), 4463.
Currently, the detection of blueberry internal bruising focuses mostly on single hyperspectral imaging (HSI) systems. Attempts to fuse different HSI systems with complementary spectral ranges are still lacking. A push broom based HSI system and a liquid crystal tunable filter (LCTF) based HSI system with different sensing ranges and detectors were investigated to jointly detect blueberry internal bruising in the lab. The mean reflectance spectrum of each berry sample was extracted from the data obtained by two HSI systems respectively. The spectral data from the two spectroscopic techniques were analyzed separately using feature selection method, partial least squares-discriminant analysis (PLS-DA), and support vector machine (SVM), and then fused with three data fusion strategies at the data level, feature level, and decision level. The three data fusion strategies achieved better classification results than using each HSI system alone. The decision level fusion integrating classification results from the two instruments with selected relevant features achieved more promising results, suggesting that the two HSI systems with complementary spectral ranges, combined with feature selection and data fusion strategies, could be used synergistically to improve blueberry internal bruising detection. This study was the first step in demonstrating the feasibility of the fusion of two HSI systems with complementary spectral ranges for detecting blueberry bruising, which could lead to a multispectral imaging system with a few selected wavelengths and an appropriate detector for bruising detection on the packing line.
Currently, the detection of blueberry internal bruising focuses mostly on single hyperspectral imaging (HSI) systems. Attempts to fuse different HSI systems with complementary spectral ranges are still lacking. A push broom based HSI system and a liquid crystal tunable filter (LCTF) based HSI system with different sensing ranges and detectors were investigated to jointly detect blueberry internal bruising in the lab. The mean reflectance spectrum of each berry sample was extracted from the data obtained by two HSI systems respectively. The spectral data from the two spectroscopic techniques were analyzed separately using feature selection method, partial least squares-discriminant analysis (PLS-DA), and support vector machine (SVM), and then fused with three data fusion strategies at the data level, feature level, and decision level. The three data fusion strategies achieved better classification results than using each HSI system alone. The decision level fusion integrating classification results from the two instruments with selected relevant features achieved more promising results, suggesting that the two HSI systems with complementary spectral ranges, combined with feature selection and data fusion strategies, could be used synergistically to improve blueberry internal bruising detection. This study was the first step in demonstrating the feasibility of the fusion of two HSI systems with complementary spectral ranges for detecting blueberry bruising, which could lead to a multispectral imaging system with a few selected wavelengths and an appropriate detector for bruising detection on the packing line.
2017
Zhang, M.; Li, C.; Takeda, F.; Yang, F.
Detection of internally bruised blueberries using hyperspectral transmittance imaging Journal Article
In: Transactions of ASABE, 60(5), 1489-1502, 2017.
@article{Zhang2017,
title = {Detection of internally bruised blueberries using hyperspectral transmittance imaging},
author = {M. Zhang and C. Li and F. Takeda and F. Yang},
url = {http://sensinglab.engr.uga.edu//srv/htdocs/wp-content/uploads/2019/11/Detection-of-internally-bruised-blueberries-using-hyperspectral-transmittance-imaging.pdf},
doi = {10.13031/trans.12197},
year = {2017},
date = {2017-08-17},
urldate = {2017-08-17},
journal = {Transactions of ASABE, 60(5), 1489-1502},
abstract = {Zhang, M., Li, C., Takeda, F., & Yang, F. (2017). Detection of internally bruised blueberries using hyperspectral transmittance imaging. Transactions of the ASABE, 60(5), 1489-1502.
Internal bruise damage that occurs in blueberry fruit during harvest operations and postharvest handling lowers the overall quality and causes significant economic losses. The main goal of this study was to nondestructively detect internal bruises in blueberries after mechanical damage using hyperspectral transmittance imaging. A total of 600 hand-harvested blueberries were divided into 20 groups of four storage times (30 min, 3 h, 12 h, and 24 h), two storage temperatures (22°C and 4°C), and three treatments (stem bruise, equator bruise, and control). A near-infrared hyperspectral imaging system was used to acquire transmittance images from 970 to 1400 nm with 5 nm bandwidth. Images were acquired from three orientations (calyx-up, stem-up, and equator-up) for fruit in the control and stem bruise groups and from four orientations (calyx-up, stem-up, equator-up, and equator-down) in the equator bruise groups. Immediately after imaging, the fruit samples were sliced, and the sliced surfaces were photographed. The color images of sliced fruit were used as references. By comparing with the reference color images, the profiles of spatial and spectral intensities were evaluated to observe the effect of orientation and help extract regions of interest (ROIs) of bruised and healthy tissues. A support vector machine (SVM) classifier was trained and tested to classify pixels of bruised and healthy tissues. Classification maps were produced, and the bruise ratio was calculated to identify bruised blueberries (bruise ratio >25%). The average accuracy of blueberry identification was 94.5% with the stem-up orientation. The results indicate that detecting bruised blueberries as soon as 30 min after mechanical damage is feasible using hyperspectral transmittance imaging.},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
Zhang, M., Li, C., Takeda, F., & Yang, F. (2017). Detection of internally bruised blueberries using hyperspectral transmittance imaging. Transactions of the ASABE, 60(5), 1489-1502.
Internal bruise damage that occurs in blueberry fruit during harvest operations and postharvest handling lowers the overall quality and causes significant economic losses. The main goal of this study was to nondestructively detect internal bruises in blueberries after mechanical damage using hyperspectral transmittance imaging. A total of 600 hand-harvested blueberries were divided into 20 groups of four storage times (30 min, 3 h, 12 h, and 24 h), two storage temperatures (22°C and 4°C), and three treatments (stem bruise, equator bruise, and control). A near-infrared hyperspectral imaging system was used to acquire transmittance images from 970 to 1400 nm with 5 nm bandwidth. Images were acquired from three orientations (calyx-up, stem-up, and equator-up) for fruit in the control and stem bruise groups and from four orientations (calyx-up, stem-up, equator-up, and equator-down) in the equator bruise groups. Immediately after imaging, the fruit samples were sliced, and the sliced surfaces were photographed. The color images of sliced fruit were used as references. By comparing with the reference color images, the profiles of spatial and spectral intensities were evaluated to observe the effect of orientation and help extract regions of interest (ROIs) of bruised and healthy tissues. A support vector machine (SVM) classifier was trained and tested to classify pixels of bruised and healthy tissues. Classification maps were produced, and the bruise ratio was calculated to identify bruised blueberries (bruise ratio >25%). The average accuracy of blueberry identification was 94.5% with the stem-up orientation. The results indicate that detecting bruised blueberries as soon as 30 min after mechanical damage is feasible using hyperspectral transmittance imaging.
Internal bruise damage that occurs in blueberry fruit during harvest operations and postharvest handling lowers the overall quality and causes significant economic losses. The main goal of this study was to nondestructively detect internal bruises in blueberries after mechanical damage using hyperspectral transmittance imaging. A total of 600 hand-harvested blueberries were divided into 20 groups of four storage times (30 min, 3 h, 12 h, and 24 h), two storage temperatures (22°C and 4°C), and three treatments (stem bruise, equator bruise, and control). A near-infrared hyperspectral imaging system was used to acquire transmittance images from 970 to 1400 nm with 5 nm bandwidth. Images were acquired from three orientations (calyx-up, stem-up, and equator-up) for fruit in the control and stem bruise groups and from four orientations (calyx-up, stem-up, equator-up, and equator-down) in the equator bruise groups. Immediately after imaging, the fruit samples were sliced, and the sliced surfaces were photographed. The color images of sliced fruit were used as references. By comparing with the reference color images, the profiles of spatial and spectral intensities were evaluated to observe the effect of orientation and help extract regions of interest (ROIs) of bruised and healthy tissues. A support vector machine (SVM) classifier was trained and tested to classify pixels of bruised and healthy tissues. Classification maps were produced, and the bruise ratio was calculated to identify bruised blueberries (bruise ratio >25%). The average accuracy of blueberry identification was 94.5% with the stem-up orientation. The results indicate that detecting bruised blueberries as soon as 30 min after mechanical damage is feasible using hyperspectral transmittance imaging.
Fan, S.; Li, C.; Huang, W.; Chen, L.
Detection of blueberry internal bruising over time using NIR hyperspectral reflectance imaging with optimum wavelengths Journal Article
In: Postharvest Biology and Technology, 134, 55-66, 2017.
@article{Fan2017,
title = {Detection of blueberry internal bruising over time using NIR hyperspectral reflectance imaging with optimum wavelengths},
author = {S. Fan and C. Li and W. Huang and L. Chen},
url = {http://sensinglab.engr.uga.edu//srv/htdocs/wp-content/uploads/2019/11/Detection-of-blueberry-internal-bruising-over-time-using-NIR-hyperspectral-reflectance-imaging-with-optimum-wavelengths.pdf},
doi = {10.1016/j.postharvbio.2017.08.012},
year = {2017},
date = {2017-08-13},
urldate = {2017-08-13},
journal = {Postharvest Biology and Technology, 134, 55-66},
abstract = {Fan, S., Li, C., Huang, W., & Chen, L. (2017). Detection of blueberry internal bruising over time using NIR hyperspectral reflectance imaging with optimum wavelengths. Postharvest biology and technology, 134, 55-66.
Early detection of internal bruising is one of the major challenges in blueberry postharvest quality sorting processes. The potential of using near infrared (NIR) hyperspectral reflectance imaging (950–1650 nm) with reduced spectral features was investigated for blueberry internal bruising detection 30 min to 12 h after mechanical impact. A least squares support vector machine (LS-SVM) was used to develop classification models to compute the spatial distribution of bruising based on the spectra extracted from regions of interest (ROIs) at four measurement times (30 min, 2 h, 6 h, and 12 h after mechanical impact). Three feature selection methods were used to select optimum wavelengths or band ratio images for bruising detection. The classification model, developed using optimum wavelengths selected by competitive adaptive reweighted sampling (CARS) (CARS-LS-SVM model) and full spectra (full spectra-LS-SVM), had similar performance in the identification of bruised blueberries. Band ratio images (1235 nm/1035 nm) achieved a comparable accuracy with the CARS-LS-SVM model at 6 h, and higher accuracy than CARS-LS-SVM and full spectra-LS-SVM models at 12 h. The overall classification accuracies of 77.5%, 83.8%, 92.5%, and 95.0% were obtained by band ratio images for blueberries 30 min, 2 h, 6 h, and 12 h after impact, respectively. In order to evaluate the performance of the proposed methods, additional validation samples were processed by the detection algorithm. The overall discrimination accuracies for healthy and bruised blueberries in the validation set were 93.3% and 98.0%, respectively, for CARS-LS-SVM model, and 93.3% and 95.9%, respectively, for band ratio images. The overall results indicated that NIR reflectance imaging can detect blueberry internal bruising as early as 30 min after mechanical impact, and band ratio images computed from two wavelengths showed great potential to detect blueberry internal bruising on the packing line.},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
Fan, S., Li, C., Huang, W., & Chen, L. (2017). Detection of blueberry internal bruising over time using NIR hyperspectral reflectance imaging with optimum wavelengths. Postharvest biology and technology, 134, 55-66.
Early detection of internal bruising is one of the major challenges in blueberry postharvest quality sorting processes. The potential of using near infrared (NIR) hyperspectral reflectance imaging (950–1650 nm) with reduced spectral features was investigated for blueberry internal bruising detection 30 min to 12 h after mechanical impact. A least squares support vector machine (LS-SVM) was used to develop classification models to compute the spatial distribution of bruising based on the spectra extracted from regions of interest (ROIs) at four measurement times (30 min, 2 h, 6 h, and 12 h after mechanical impact). Three feature selection methods were used to select optimum wavelengths or band ratio images for bruising detection. The classification model, developed using optimum wavelengths selected by competitive adaptive reweighted sampling (CARS) (CARS-LS-SVM model) and full spectra (full spectra-LS-SVM), had similar performance in the identification of bruised blueberries. Band ratio images (1235 nm/1035 nm) achieved a comparable accuracy with the CARS-LS-SVM model at 6 h, and higher accuracy than CARS-LS-SVM and full spectra-LS-SVM models at 12 h. The overall classification accuracies of 77.5%, 83.8%, 92.5%, and 95.0% were obtained by band ratio images for blueberries 30 min, 2 h, 6 h, and 12 h after impact, respectively. In order to evaluate the performance of the proposed methods, additional validation samples were processed by the detection algorithm. The overall discrimination accuracies for healthy and bruised blueberries in the validation set were 93.3% and 98.0%, respectively, for CARS-LS-SVM model, and 93.3% and 95.9%, respectively, for band ratio images. The overall results indicated that NIR reflectance imaging can detect blueberry internal bruising as early as 30 min after mechanical impact, and band ratio images computed from two wavelengths showed great potential to detect blueberry internal bruising on the packing line.
Early detection of internal bruising is one of the major challenges in blueberry postharvest quality sorting processes. The potential of using near infrared (NIR) hyperspectral reflectance imaging (950–1650 nm) with reduced spectral features was investigated for blueberry internal bruising detection 30 min to 12 h after mechanical impact. A least squares support vector machine (LS-SVM) was used to develop classification models to compute the spatial distribution of bruising based on the spectra extracted from regions of interest (ROIs) at four measurement times (30 min, 2 h, 6 h, and 12 h after mechanical impact). Three feature selection methods were used to select optimum wavelengths or band ratio images for bruising detection. The classification model, developed using optimum wavelengths selected by competitive adaptive reweighted sampling (CARS) (CARS-LS-SVM model) and full spectra (full spectra-LS-SVM), had similar performance in the identification of bruised blueberries. Band ratio images (1235 nm/1035 nm) achieved a comparable accuracy with the CARS-LS-SVM model at 6 h, and higher accuracy than CARS-LS-SVM and full spectra-LS-SVM models at 12 h. The overall classification accuracies of 77.5%, 83.8%, 92.5%, and 95.0% were obtained by band ratio images for blueberries 30 min, 2 h, 6 h, and 12 h after impact, respectively. In order to evaluate the performance of the proposed methods, additional validation samples were processed by the detection algorithm. The overall discrimination accuracies for healthy and bruised blueberries in the validation set were 93.3% and 98.0%, respectively, for CARS-LS-SVM model, and 93.3% and 95.9%, respectively, for band ratio images. The overall results indicated that NIR reflectance imaging can detect blueberry internal bruising as early as 30 min after mechanical impact, and band ratio images computed from two wavelengths showed great potential to detect blueberry internal bruising on the packing line.
Zhang, M.; Li, C.; Yang, F.
Classification of Foreign Matter Embedded inside Cotton Lint using Short Wave Infrared (SWIR) Hyperspectral Transmittance Imaging Journal Article
In: Computers and Electronics in Agriculture, 139, 75-90, 2017.
@article{Zhang2017b,
title = {Classification of Foreign Matter Embedded inside Cotton Lint using Short Wave Infrared (SWIR) Hyperspectral Transmittance Imaging},
author = {M. Zhang and C. Li and F. Yang},
url = {http://sensinglab.engr.uga.edu//srv/htdocs/wp-content/uploads/2019/11/Classification-of-Foreign-Matter-Embedded-inside-Cotton-Lint-using-Short-Wave-Infrared-SWIR-Hyperspectral-Transmittance-Imaging.pdf},
doi = {10.1016/j.compag.2017.05.005},
year = {2017},
date = {2017-06-18},
urldate = {2017-06-18},
journal = {Computers and Electronics in Agriculture, 139, 75-90},
abstract = {Zhang, M., Li, C., & Yang, F. (2017). Classification of foreign matter embedded inside cotton lint using short wave infrared (SWIR) hyperspectral transmittance imaging. Computers and Electronics in Agriculture, 139, 75-90.
Cotton is an important source of natural fiber around the world. Cotton lint, however, could be contaminated by various types of foreign matter (FM) during harvesting and processing, leading to reduced quality and potentially even defective textile products. Current sensing methods can detect the presence of foreign matter on the surface of cotton lint, but they are not able to efficiently detect and classify foreign matter that is mixed with or embedded inside cotton lint. This study focused on the detection and classification of common types of foreign matter hidden within the cotton lint by a short wave near infrared hyperspectral imaging (HSI) system using the transmittance mode. Fourteen common categories of foreign matter and cotton lint were collected from the field and the foreign matter particles were sandwiched between two thin cotton lint webs. Operation parameters were optimized through a series of experiments for the best performance of the transmittance mode. After acquiring transmittance images of the cotton lint and foreign matter mixture, minimum noise fraction (MNF) rotation was utilized to obtain component images to assist visual detection and mean spectra extraction from a total of 141 wavelength bands. The optimal spectral bands were identified by using the minimal-redundancy-maximal-relevance (mRMR)-based feature selection method. Linear discriminant analysis (LDA) and a support vector machine (SVM) were employed to classify foreign matter at the spectral and pixel level, respectively. Over 95% classification accuracies for the spectra and the images were achieved using the selected optimal wavelengths. This study indicated that it was feasible to detect botanical (e.g. seed coat, seed meat, stem, and leaf) and non-botanical (e.g. paper, and plastic package) types of foreign matter that were embedded inside cotton lint using short wave infrared hyperspectral transmittance imaging.},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
Zhang, M., Li, C., & Yang, F. (2017). Classification of foreign matter embedded inside cotton lint using short wave infrared (SWIR) hyperspectral transmittance imaging. Computers and Electronics in Agriculture, 139, 75-90.
Cotton is an important source of natural fiber around the world. Cotton lint, however, could be contaminated by various types of foreign matter (FM) during harvesting and processing, leading to reduced quality and potentially even defective textile products. Current sensing methods can detect the presence of foreign matter on the surface of cotton lint, but they are not able to efficiently detect and classify foreign matter that is mixed with or embedded inside cotton lint. This study focused on the detection and classification of common types of foreign matter hidden within the cotton lint by a short wave near infrared hyperspectral imaging (HSI) system using the transmittance mode. Fourteen common categories of foreign matter and cotton lint were collected from the field and the foreign matter particles were sandwiched between two thin cotton lint webs. Operation parameters were optimized through a series of experiments for the best performance of the transmittance mode. After acquiring transmittance images of the cotton lint and foreign matter mixture, minimum noise fraction (MNF) rotation was utilized to obtain component images to assist visual detection and mean spectra extraction from a total of 141 wavelength bands. The optimal spectral bands were identified by using the minimal-redundancy-maximal-relevance (mRMR)-based feature selection method. Linear discriminant analysis (LDA) and a support vector machine (SVM) were employed to classify foreign matter at the spectral and pixel level, respectively. Over 95% classification accuracies for the spectra and the images were achieved using the selected optimal wavelengths. This study indicated that it was feasible to detect botanical (e.g. seed coat, seed meat, stem, and leaf) and non-botanical (e.g. paper, and plastic package) types of foreign matter that were embedded inside cotton lint using short wave infrared hyperspectral transmittance imaging.
Cotton is an important source of natural fiber around the world. Cotton lint, however, could be contaminated by various types of foreign matter (FM) during harvesting and processing, leading to reduced quality and potentially even defective textile products. Current sensing methods can detect the presence of foreign matter on the surface of cotton lint, but they are not able to efficiently detect and classify foreign matter that is mixed with or embedded inside cotton lint. This study focused on the detection and classification of common types of foreign matter hidden within the cotton lint by a short wave near infrared hyperspectral imaging (HSI) system using the transmittance mode. Fourteen common categories of foreign matter and cotton lint were collected from the field and the foreign matter particles were sandwiched between two thin cotton lint webs. Operation parameters were optimized through a series of experiments for the best performance of the transmittance mode. After acquiring transmittance images of the cotton lint and foreign matter mixture, minimum noise fraction (MNF) rotation was utilized to obtain component images to assist visual detection and mean spectra extraction from a total of 141 wavelength bands. The optimal spectral bands were identified by using the minimal-redundancy-maximal-relevance (mRMR)-based feature selection method. Linear discriminant analysis (LDA) and a support vector machine (SVM) were employed to classify foreign matter at the spectral and pixel level, respectively. Over 95% classification accuracies for the spectra and the images were achieved using the selected optimal wavelengths. This study indicated that it was feasible to detect botanical (e.g. seed coat, seed meat, stem, and leaf) and non-botanical (e.g. paper, and plastic package) types of foreign matter that were embedded inside cotton lint using short wave infrared hyperspectral transmittance imaging.
2016
Jiang, Y.; Li, C.; Takeda, F.
Nondestructive detection and quantification of blueberry bruising using near-infrared (NIR) hyperspectral reflectance imaging Journal Article
In: Scientific Reports, 6, 35679, 2016.
@article{Jiang2016,
title = {Nondestructive detection and quantification of blueberry bruising using near-infrared (NIR) hyperspectral reflectance imaging},
author = {Y. Jiang and C. Li and F. Takeda },
url = {http://sensinglab.engr.uga.edu//srv/htdocs/wp-content/uploads/2019/11/Nondestructive-detection-and-quantification-of-blueberry-bruising-using-near-infrared-NIR-hyperspectral-reflectance-imaging.pdf},
doi = {10.1038/srep35679},
year = {2016},
date = {2016-10-04},
urldate = {2016-10-04},
journal = {Scientific Reports, 6, 35679},
abstract = {Jiang, Y., Li, C., & Takeda, F. (2016). Nondestructive detection and quantification of blueberry bruising using near-infrared (NIR) hyperspectral reflectance imaging. Scientific reports, 6, 35679.
Currently, blueberry bruising is evaluated by either human visual/tactile inspection or firmness measurement instruments. These methods are destructive, time-consuming, and subjective. The goal of this paper was to develop a non-destructive approach for blueberry bruising detection and quantification. Experiments were conducted on 300 samples of southern highbush blueberry (Camellia, Rebel, and Star) and on 1500 samples of northern highbush blueberry (Bluecrop, Jersey, and Liberty) for hyperspectral imaging analysis, firmness measurement, and human evaluation. An algorithm was developed to automatically calculate a bruise ratio index (ratio of bruised to whole fruit area) for bruise quantification. The spectra of bruised and healthy tissues were statistically separated and the separation was independent of cultivars. Support vector machine (SVM) classification of the spectra from the regions of interest (ROIs) achieved over 94%, 92%, and 96% accuracy on the training set, independent testing set, and combined set, respectively. The statistical results showed that the bruise ratio index was equivalent to the measured firmness but better than the predicted firmness in regard to effectiveness of bruise quantification, and the bruise ratio index had a strong correlation with human assessment (R2 = 0.78 − 0.83). Therefore, the proposed approach and the bruise ratio index are effective to non-destructively detect and quantify blueberry bruising.},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
Jiang, Y., Li, C., & Takeda, F. (2016). Nondestructive detection and quantification of blueberry bruising using near-infrared (NIR) hyperspectral reflectance imaging. Scientific reports, 6, 35679.
Currently, blueberry bruising is evaluated by either human visual/tactile inspection or firmness measurement instruments. These methods are destructive, time-consuming, and subjective. The goal of this paper was to develop a non-destructive approach for blueberry bruising detection and quantification. Experiments were conducted on 300 samples of southern highbush blueberry (Camellia, Rebel, and Star) and on 1500 samples of northern highbush blueberry (Bluecrop, Jersey, and Liberty) for hyperspectral imaging analysis, firmness measurement, and human evaluation. An algorithm was developed to automatically calculate a bruise ratio index (ratio of bruised to whole fruit area) for bruise quantification. The spectra of bruised and healthy tissues were statistically separated and the separation was independent of cultivars. Support vector machine (SVM) classification of the spectra from the regions of interest (ROIs) achieved over 94%, 92%, and 96% accuracy on the training set, independent testing set, and combined set, respectively. The statistical results showed that the bruise ratio index was equivalent to the measured firmness but better than the predicted firmness in regard to effectiveness of bruise quantification, and the bruise ratio index had a strong correlation with human assessment (R2 = 0.78 − 0.83). Therefore, the proposed approach and the bruise ratio index are effective to non-destructively detect and quantify blueberry bruising.
Currently, blueberry bruising is evaluated by either human visual/tactile inspection or firmness measurement instruments. These methods are destructive, time-consuming, and subjective. The goal of this paper was to develop a non-destructive approach for blueberry bruising detection and quantification. Experiments were conducted on 300 samples of southern highbush blueberry (Camellia, Rebel, and Star) and on 1500 samples of northern highbush blueberry (Bluecrop, Jersey, and Liberty) for hyperspectral imaging analysis, firmness measurement, and human evaluation. An algorithm was developed to automatically calculate a bruise ratio index (ratio of bruised to whole fruit area) for bruise quantification. The spectra of bruised and healthy tissues were statistically separated and the separation was independent of cultivars. Support vector machine (SVM) classification of the spectra from the regions of interest (ROIs) achieved over 94%, 92%, and 96% accuracy on the training set, independent testing set, and combined set, respectively. The statistical results showed that the bruise ratio index was equivalent to the measured firmness but better than the predicted firmness in regard to effectiveness of bruise quantification, and the bruise ratio index had a strong correlation with human assessment (R2 = 0.78 − 0.83). Therefore, the proposed approach and the bruise ratio index are effective to non-destructively detect and quantify blueberry bruising.
Zhang, R.; Li, C.; Zhang, M.; Rodgers, J.
Shortwave infrared hyperspectral reflectance imaging for cotton foreign matter classification Journal Article
In: Computers and Electronics in Agriculture, 127, 260-270, 2016.
@article{Zhang2016,
title = {Shortwave infrared hyperspectral reflectance imaging for cotton foreign matter classification},
author = {R. Zhang and C. Li and M. Zhang and J. Rodgers},
url = {http://sensinglab.engr.uga.edu//srv/htdocs/wp-content/uploads/2019/11/Shortwave-infrared-hyperspectral-reflectance-imaging-for-cotton-foreign-matter-classification.pdf},
doi = {10.1016/j.compag.2016.06.023},
year = {2016},
date = {2016-06-20},
urldate = {2016-06-20},
journal = {Computers and Electronics in Agriculture, 127, 260-270},
abstract = {Zhang, R., Li, C., Zhang, M., & Rodgers, J. (2016). Shortwave infrared hyperspectral reflectance imaging for cotton foreign matter classification. Computers and Electronics in Agriculture, 127, 260-270.
Cotton contaminants seriously reduce the commercial value of cotton lint and further degrade the quality of textile products. This research aims to investigate the potential of a non-contact technique, i.e., liquid crystal tunable filter (LCTF) hyperspectral imaging, to inspect foreign matter on the surface of cotton lint. The foreign matter samples used in this study included 11 types of botanical foreign matter and 5 types of non-botanical foreign matter. Hyperspectral images of the foreign matter were acquired using a LCTF hyperspectral imaging system with a spectral range from 900 to 1700 nm. The mean spectra of the foreign matter and lint samples were extracted manually from the images. Linear discriminant analysis was applied to classify different types of foreign matter and cotton lint according to their spectral features. Classification accuracies of 96.5% and 95.1% were achieved with leave-one-out and four fold cross-validation, respectively. For pixel-level image classification, a majority of the pixels for different types of foreign matter were classified correctly by a support vector machine, using the top features of the minimum noise fraction transformation. The results demonstrate that non-contact liquid crystal tunable filter hyperspectral imaging is a promising method to discriminate foreign matter materials from cotton lint.},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
Zhang, R., Li, C., Zhang, M., & Rodgers, J. (2016). Shortwave infrared hyperspectral reflectance imaging for cotton foreign matter classification. Computers and Electronics in Agriculture, 127, 260-270.
Cotton contaminants seriously reduce the commercial value of cotton lint and further degrade the quality of textile products. This research aims to investigate the potential of a non-contact technique, i.e., liquid crystal tunable filter (LCTF) hyperspectral imaging, to inspect foreign matter on the surface of cotton lint. The foreign matter samples used in this study included 11 types of botanical foreign matter and 5 types of non-botanical foreign matter. Hyperspectral images of the foreign matter were acquired using a LCTF hyperspectral imaging system with a spectral range from 900 to 1700 nm. The mean spectra of the foreign matter and lint samples were extracted manually from the images. Linear discriminant analysis was applied to classify different types of foreign matter and cotton lint according to their spectral features. Classification accuracies of 96.5% and 95.1% were achieved with leave-one-out and four fold cross-validation, respectively. For pixel-level image classification, a majority of the pixels for different types of foreign matter were classified correctly by a support vector machine, using the top features of the minimum noise fraction transformation. The results demonstrate that non-contact liquid crystal tunable filter hyperspectral imaging is a promising method to discriminate foreign matter materials from cotton lint.
Cotton contaminants seriously reduce the commercial value of cotton lint and further degrade the quality of textile products. This research aims to investigate the potential of a non-contact technique, i.e., liquid crystal tunable filter (LCTF) hyperspectral imaging, to inspect foreign matter on the surface of cotton lint. The foreign matter samples used in this study included 11 types of botanical foreign matter and 5 types of non-botanical foreign matter. Hyperspectral images of the foreign matter were acquired using a LCTF hyperspectral imaging system with a spectral range from 900 to 1700 nm. The mean spectra of the foreign matter and lint samples were extracted manually from the images. Linear discriminant analysis was applied to classify different types of foreign matter and cotton lint according to their spectral features. Classification accuracies of 96.5% and 95.1% were achieved with leave-one-out and four fold cross-validation, respectively. For pixel-level image classification, a majority of the pixels for different types of foreign matter were classified correctly by a support vector machine, using the top features of the minimum noise fraction transformation. The results demonstrate that non-contact liquid crystal tunable filter hyperspectral imaging is a promising method to discriminate foreign matter materials from cotton lint.
Mustafic, A.; Jiang, Y.; Li, C.
Cotton contamination detection and classification using hyperspectral fluorescence imaging Journal Article
In: Textile Research Journal, 86(15), 1574-1584, 2016.
@article{Mustafic2016,
title = {Cotton contamination detection and classification using hyperspectral fluorescence imaging},
author = {A. Mustafic and Y. Jiang and C. Li},
url = {http://sensinglab.engr.uga.edu//srv/htdocs/wp-content/uploads/2019/11/Cotton-contamination-detection-and-classification-using-hyperspectral-fluorescence-imaging.pdf},
doi = {10.1177/0040517515590416},
year = {2016},
date = {2016-03-03},
urldate = {2016-03-03},
journal = {Textile Research Journal, 86(15), 1574-1584},
abstract = {Mustafic, A., Jiang, Y., & Li, C. (2016). Cotton contamination detection and classification using hyperspectral fluorescence imaging. Textile Research Journal, 86(15), 1574-1584.
The presence of foreign matter in ginned cotton lowers the quality and ultimately the monetary value of cotton. Previous studies have shown benefits of using ultraviolet excited fluorescence to detect certain cotton contamination that is difficult to detect using other methods. The overall goal of this study was to explore the feasibility of using hyperspectral fluorescence imaging as a complementary tool for foreign matter differentiation. The mean spectra of lint and seven types of foreign matter were extracted from the hyperspectral fluorescence images using a region-of-interest-based approach. The principal component analysis was applied to select the optimal features from a total of 113 wavelengths covering the spectral range of 425–700 nm. The linear discriminant analysis with the selected wavelengths achieved an average classification rate of 90% for all samples. Therefore, this imaging method could be used as a complementary sensing modality to current instruments that are employed for cotton quality assessment in the textile industry.},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
Mustafic, A., Jiang, Y., & Li, C. (2016). Cotton contamination detection and classification using hyperspectral fluorescence imaging. Textile Research Journal, 86(15), 1574-1584.
The presence of foreign matter in ginned cotton lowers the quality and ultimately the monetary value of cotton. Previous studies have shown benefits of using ultraviolet excited fluorescence to detect certain cotton contamination that is difficult to detect using other methods. The overall goal of this study was to explore the feasibility of using hyperspectral fluorescence imaging as a complementary tool for foreign matter differentiation. The mean spectra of lint and seven types of foreign matter were extracted from the hyperspectral fluorescence images using a region-of-interest-based approach. The principal component analysis was applied to select the optimal features from a total of 113 wavelengths covering the spectral range of 425–700 nm. The linear discriminant analysis with the selected wavelengths achieved an average classification rate of 90% for all samples. Therefore, this imaging method could be used as a complementary sensing modality to current instruments that are employed for cotton quality assessment in the textile industry.
The presence of foreign matter in ginned cotton lowers the quality and ultimately the monetary value of cotton. Previous studies have shown benefits of using ultraviolet excited fluorescence to detect certain cotton contamination that is difficult to detect using other methods. The overall goal of this study was to explore the feasibility of using hyperspectral fluorescence imaging as a complementary tool for foreign matter differentiation. The mean spectra of lint and seven types of foreign matter were extracted from the hyperspectral fluorescence images using a region-of-interest-based approach. The principal component analysis was applied to select the optimal features from a total of 113 wavelengths covering the spectral range of 425–700 nm. The linear discriminant analysis with the selected wavelengths achieved an average classification rate of 90% for all samples. Therefore, this imaging method could be used as a complementary sensing modality to current instruments that are employed for cotton quality assessment in the textile industry.
2015
Jiang, Y.; Li, C.
mRMR-based feature selection for classification of cotton foreign matter using hyperspectral imaging Journal Article
In: Computers and Electronics in Agriculture, 119, 191-200, 2015.
@article{Jiang2015,
title = {mRMR-based feature selection for classification of cotton foreign matter using hyperspectral imaging},
author = {Y. Jiang and C. Li},
url = {http://sensinglab.engr.uga.edu//srv/htdocs/wp-content/uploads/2019/11/mRMR-based-feature-selection-for-classification-of-cotton-foreign-matter-using-hyperspectral-imaging-.pdf},
doi = {10.1016/j.compag.2015.10.017},
year = {2015},
date = {2015-10-23},
urldate = {2015-10-23},
journal = {Computers and Electronics in Agriculture, 119, 191-200},
abstract = {Jiang, Y., & Li, C. (2015). mRMR-based feature selection for classification of cotton foreign matter using hyperspectral imaging. Computers and Electronics in Agriculture, 119, 191-200.
Different cotton foreign matter causes various levels of damage to textile products and decreases the monetary value of cotton. Hyperspectral imaging technique has shown the capability of discriminating the foreign matter, but its large amount of information which is mostly correlated and redundant limits the classification accuracy and processing speed. The goal of this study was to explore a new method of feature selection (minimum Redundancy Maximum Relevance algorithm) to select optimal wavelengths from the visible to near infrared spectra of the hyperspectral imaging data for cotton foreign matter classification. A spectral dataset containing 480 samples was collected from hyperspectral reflectance images of cotton lint and 15 types of foreign matter. Each sample was represented by a mean spectrum containing 256 wavelengths ranging from 400 nm to 1000 nm. The dataset was pre-processed by removing the noise, and the number of wavelengths was reduced from 256 to 223 by removing those with a signal to noise ratio lower than 10 dB. The optimal wavelengths were selected from the pre-processed dataset by a two-stage approach. The first step was to rank the features using the minimum Redundancy Maximum Relevance algorithm and to provide only the top ranked features for the following feature selection. In the second step, the sequential backward elimination was applied to the top ranked wavelengths to select the optimal wavelengths for foreign matter classification. The generality of the selected wavelengths was evaluated by comparing the classification performance using the Linear Discriminant Analysis (LDA), Support Vector Machine (SVM), and Artificial Neural Networks (ANNs). A total of 12 wavelengths were selected as the optimal feature set for foreign matter classification. Eight wavelengths from the visible range were related to the natural or artificial pigments of foreign matter, and the other four from the near-infrared range were related to the proteins or nutrients in foreign matter. The selected wavelengths achieved average classification rates of 91.25%, 86.67%, and 86.67% for the LDA, SVM, and ANNs, respectively, indicating the generality of the selected features. This study explored a new method for hyperspectral imaging optimal wavelength selection and the selected wavelengths can be used with different classifiers for cotton foreign matter classification.},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
Jiang, Y., & Li, C. (2015). mRMR-based feature selection for classification of cotton foreign matter using hyperspectral imaging. Computers and Electronics in Agriculture, 119, 191-200.
Different cotton foreign matter causes various levels of damage to textile products and decreases the monetary value of cotton. Hyperspectral imaging technique has shown the capability of discriminating the foreign matter, but its large amount of information which is mostly correlated and redundant limits the classification accuracy and processing speed. The goal of this study was to explore a new method of feature selection (minimum Redundancy Maximum Relevance algorithm) to select optimal wavelengths from the visible to near infrared spectra of the hyperspectral imaging data for cotton foreign matter classification. A spectral dataset containing 480 samples was collected from hyperspectral reflectance images of cotton lint and 15 types of foreign matter. Each sample was represented by a mean spectrum containing 256 wavelengths ranging from 400 nm to 1000 nm. The dataset was pre-processed by removing the noise, and the number of wavelengths was reduced from 256 to 223 by removing those with a signal to noise ratio lower than 10 dB. The optimal wavelengths were selected from the pre-processed dataset by a two-stage approach. The first step was to rank the features using the minimum Redundancy Maximum Relevance algorithm and to provide only the top ranked features for the following feature selection. In the second step, the sequential backward elimination was applied to the top ranked wavelengths to select the optimal wavelengths for foreign matter classification. The generality of the selected wavelengths was evaluated by comparing the classification performance using the Linear Discriminant Analysis (LDA), Support Vector Machine (SVM), and Artificial Neural Networks (ANNs). A total of 12 wavelengths were selected as the optimal feature set for foreign matter classification. Eight wavelengths from the visible range were related to the natural or artificial pigments of foreign matter, and the other four from the near-infrared range were related to the proteins or nutrients in foreign matter. The selected wavelengths achieved average classification rates of 91.25%, 86.67%, and 86.67% for the LDA, SVM, and ANNs, respectively, indicating the generality of the selected features. This study explored a new method for hyperspectral imaging optimal wavelength selection and the selected wavelengths can be used with different classifiers for cotton foreign matter classification.
Different cotton foreign matter causes various levels of damage to textile products and decreases the monetary value of cotton. Hyperspectral imaging technique has shown the capability of discriminating the foreign matter, but its large amount of information which is mostly correlated and redundant limits the classification accuracy and processing speed. The goal of this study was to explore a new method of feature selection (minimum Redundancy Maximum Relevance algorithm) to select optimal wavelengths from the visible to near infrared spectra of the hyperspectral imaging data for cotton foreign matter classification. A spectral dataset containing 480 samples was collected from hyperspectral reflectance images of cotton lint and 15 types of foreign matter. Each sample was represented by a mean spectrum containing 256 wavelengths ranging from 400 nm to 1000 nm. The dataset was pre-processed by removing the noise, and the number of wavelengths was reduced from 256 to 223 by removing those with a signal to noise ratio lower than 10 dB. The optimal wavelengths were selected from the pre-processed dataset by a two-stage approach. The first step was to rank the features using the minimum Redundancy Maximum Relevance algorithm and to provide only the top ranked features for the following feature selection. In the second step, the sequential backward elimination was applied to the top ranked wavelengths to select the optimal wavelengths for foreign matter classification. The generality of the selected wavelengths was evaluated by comparing the classification performance using the Linear Discriminant Analysis (LDA), Support Vector Machine (SVM), and Artificial Neural Networks (ANNs). A total of 12 wavelengths were selected as the optimal feature set for foreign matter classification. Eight wavelengths from the visible range were related to the natural or artificial pigments of foreign matter, and the other four from the near-infrared range were related to the proteins or nutrients in foreign matter. The selected wavelengths achieved average classification rates of 91.25%, 86.67%, and 86.67% for the LDA, SVM, and ANNs, respectively, indicating the generality of the selected features. This study explored a new method for hyperspectral imaging optimal wavelength selection and the selected wavelengths can be used with different classifiers for cotton foreign matter classification.
Chugunov, S.; Li, C.
Monte Carlo simulation of light propagation in healthy and diseased onion bulbs with multiple layers Journal Article
In: Computers and Electronics in Agriculture, 117, 91-101, 2015.
@article{Chugunov2015,
title = {Monte Carlo simulation of light propagation in healthy and diseased onion bulbs with multiple layers},
author = {S. Chugunov and C. Li},
url = {http://sensinglab.engr.uga.edu//srv/htdocs/wp-content/uploads/2019/11/Monte-Carlo-simulation-of-light-propagation-in-healthy-and-diseased-onion-bulbs-with-multiple-layers.pdf},
doi = {10.1016/j.compag.2015.07.015},
year = {2015},
date = {2015-07-24},
urldate = {2015-07-24},
journal = {Computers and Electronics in Agriculture, 117, 91-101},
abstract = {Chugunov, S., & Li, C. (2015). Monte Carlo simulation of light propagation in healthy and diseased onion bulbs with multiple layers. Computers and Electronics in Agriculture, 117, 91-101.
It remains unanswered how the light interacts with healthy and pathogen-infected onion tissues in a multi-layer structure. The overall goal of this study was to simulate light propagation (including scattering and absorption) in healthy and pathogen-infected onion bulbs in the visible and near infrared (NIR) range using Monte Carlo simulations. Healthy onions and bulbs infected with two major onion post-harvest diseases, Botritis Allii (Neck Rot) and Burkholderia Cepacia (Sour Skin), were considered as the subjects of the simulation. Multi-layered models (18 layers in total) of healthy and infected onion bulbs were developed representing onion structure in the form of parallel slabs. Variance of optical properties was introduced into the models using median and quartile values computed from the experimental data. Monte Carlo-simulations were performed for the developed models to generate optical responses of 33 cases of healthy and infected onions representing different stages of disease propagation in the spectral range 550–1650 nm. Optical responses of all the cases were assessed with statistical tests. Study of spatially-resolved scattering reflectance was conducted to identify patterns typical for infected onions. Optical responses were measured experimentally to validate the simulation results for healthy onions. A total of 18 configurations (out of 33) of infected onions showed significant difference from healthy bulbs and demonstrated great potential for nondestructive detection. Confident detection was determined for onions with infection as deep as in the 3rd scale. The proposed optimal window for disease detection was 670–870 nm. The greatest discrepancy between optical response of infected and healthy onions was found at 800 nm. Spatially-resolved reflectance of the Neck Rot-infected onions showed consistent lower intensity than that of healthy onions over the entire studied radial range, whereas the Sour Skin-infected onions exhibited differences in a limited radial range. Light penetration simulation revealed that photons can reach 5–6 mm deep in the bulb in the case of one dry skin in the wavelength of around 800 nm and 1100 nm. Validation results suggested that although the overall pattern of the simulated results and experimental measurements was similar, the systematic error was likely caused by the curvature of the onion bulb and the measurement instrument. This study was the first attempt to use Monte Carlo simulations in the field of post-harvest research to model complex tissues of vegetables using more than 2 layers. The results of the simulation could be useful in developing non-destructive optical sensing methods for onions.},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
Chugunov, S., & Li, C. (2015). Monte Carlo simulation of light propagation in healthy and diseased onion bulbs with multiple layers. Computers and Electronics in Agriculture, 117, 91-101.
It remains unanswered how the light interacts with healthy and pathogen-infected onion tissues in a multi-layer structure. The overall goal of this study was to simulate light propagation (including scattering and absorption) in healthy and pathogen-infected onion bulbs in the visible and near infrared (NIR) range using Monte Carlo simulations. Healthy onions and bulbs infected with two major onion post-harvest diseases, Botritis Allii (Neck Rot) and Burkholderia Cepacia (Sour Skin), were considered as the subjects of the simulation. Multi-layered models (18 layers in total) of healthy and infected onion bulbs were developed representing onion structure in the form of parallel slabs. Variance of optical properties was introduced into the models using median and quartile values computed from the experimental data. Monte Carlo-simulations were performed for the developed models to generate optical responses of 33 cases of healthy and infected onions representing different stages of disease propagation in the spectral range 550–1650 nm. Optical responses of all the cases were assessed with statistical tests. Study of spatially-resolved scattering reflectance was conducted to identify patterns typical for infected onions. Optical responses were measured experimentally to validate the simulation results for healthy onions. A total of 18 configurations (out of 33) of infected onions showed significant difference from healthy bulbs and demonstrated great potential for nondestructive detection. Confident detection was determined for onions with infection as deep as in the 3rd scale. The proposed optimal window for disease detection was 670–870 nm. The greatest discrepancy between optical response of infected and healthy onions was found at 800 nm. Spatially-resolved reflectance of the Neck Rot-infected onions showed consistent lower intensity than that of healthy onions over the entire studied radial range, whereas the Sour Skin-infected onions exhibited differences in a limited radial range. Light penetration simulation revealed that photons can reach 5–6 mm deep in the bulb in the case of one dry skin in the wavelength of around 800 nm and 1100 nm. Validation results suggested that although the overall pattern of the simulated results and experimental measurements was similar, the systematic error was likely caused by the curvature of the onion bulb and the measurement instrument. This study was the first attempt to use Monte Carlo simulations in the field of post-harvest research to model complex tissues of vegetables using more than 2 layers. The results of the simulation could be useful in developing non-destructive optical sensing methods for onions.
It remains unanswered how the light interacts with healthy and pathogen-infected onion tissues in a multi-layer structure. The overall goal of this study was to simulate light propagation (including scattering and absorption) in healthy and pathogen-infected onion bulbs in the visible and near infrared (NIR) range using Monte Carlo simulations. Healthy onions and bulbs infected with two major onion post-harvest diseases, Botritis Allii (Neck Rot) and Burkholderia Cepacia (Sour Skin), were considered as the subjects of the simulation. Multi-layered models (18 layers in total) of healthy and infected onion bulbs were developed representing onion structure in the form of parallel slabs. Variance of optical properties was introduced into the models using median and quartile values computed from the experimental data. Monte Carlo-simulations were performed for the developed models to generate optical responses of 33 cases of healthy and infected onions representing different stages of disease propagation in the spectral range 550–1650 nm. Optical responses of all the cases were assessed with statistical tests. Study of spatially-resolved scattering reflectance was conducted to identify patterns typical for infected onions. Optical responses were measured experimentally to validate the simulation results for healthy onions. A total of 18 configurations (out of 33) of infected onions showed significant difference from healthy bulbs and demonstrated great potential for nondestructive detection. Confident detection was determined for onions with infection as deep as in the 3rd scale. The proposed optimal window for disease detection was 670–870 nm. The greatest discrepancy between optical response of infected and healthy onions was found at 800 nm. Spatially-resolved reflectance of the Neck Rot-infected onions showed consistent lower intensity than that of healthy onions over the entire studied radial range, whereas the Sour Skin-infected onions exhibited differences in a limited radial range. Light penetration simulation revealed that photons can reach 5–6 mm deep in the bulb in the case of one dry skin in the wavelength of around 800 nm and 1100 nm. Validation results suggested that although the overall pattern of the simulated results and experimental measurements was similar, the systematic error was likely caused by the curvature of the onion bulb and the measurement instrument. This study was the first attempt to use Monte Carlo simulations in the field of post-harvest research to model complex tissues of vegetables using more than 2 layers. The results of the simulation could be useful in developing non-destructive optical sensing methods for onions.
Wang, W.; Li, C.
A multimodal machine vision system for quality inspection of onions Journal Article
In: Journal of Food Engineering, 166, 291-301, 2015.
@article{Wang2015,
title = {A multimodal machine vision system for quality inspection of onions},
author = {W. Wang and C. Li},
url = {http://sensinglab.engr.uga.edu//srv/htdocs/wp-content/uploads/2019/11/A-multimodal-machine-vision-system-for-quality-inspection-of-onions-.pdf},
doi = {10.1016/j.jfoodeng.2015.06.027},
year = {2015},
date = {2015-06-18},
urldate = {2015-06-18},
journal = {Journal of Food Engineering, 166, 291-301},
abstract = {Wang, W., & Li, C. (2015). A multimodal machine vision system for quality inspection of onions. Journal of Food Engineering, 166, 291-301.
A multimodal machine vision system was developed to evaluate quality factors of onions holistically and nondestructively. The system integrated hyperspectral, 3D, and X-ray imaging sensors. A LabVIEW program was developed to acquire color images, spectral images, depth images, X-ray images of onions, and measure the weight of onions. With the multimodal data collected, algorithms were developed to calculate the maximum diameter, volume, density, and detect latent defects of onions. Three groups of sweet onions (regular, inoculated with Burkholderia cepacia, and inoculated with Pseudomonas viridiflava) were tested. Results showed that the system accurately measured the weight (RMSE = 3.6 g), diameter (RMSE = 1.7 mm), volume (RMSE = 16.5 cm3), and density (RMSE = 0.03 g/cm3) of onions, and correctly classified 88.9% healthy and defective onions. This work demonstrated a promising approach to evaluate both external and internal quality parameters of onions, which is applicable to onion packinghouses. The proposed system and methods are also potentially applicable to quality inspection of other agricultural products.},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
Wang, W., & Li, C. (2015). A multimodal machine vision system for quality inspection of onions. Journal of Food Engineering, 166, 291-301.
A multimodal machine vision system was developed to evaluate quality factors of onions holistically and nondestructively. The system integrated hyperspectral, 3D, and X-ray imaging sensors. A LabVIEW program was developed to acquire color images, spectral images, depth images, X-ray images of onions, and measure the weight of onions. With the multimodal data collected, algorithms were developed to calculate the maximum diameter, volume, density, and detect latent defects of onions. Three groups of sweet onions (regular, inoculated with Burkholderia cepacia, and inoculated with Pseudomonas viridiflava) were tested. Results showed that the system accurately measured the weight (RMSE = 3.6 g), diameter (RMSE = 1.7 mm), volume (RMSE = 16.5 cm3), and density (RMSE = 0.03 g/cm3) of onions, and correctly classified 88.9% healthy and defective onions. This work demonstrated a promising approach to evaluate both external and internal quality parameters of onions, which is applicable to onion packinghouses. The proposed system and methods are also potentially applicable to quality inspection of other agricultural products.
A multimodal machine vision system was developed to evaluate quality factors of onions holistically and nondestructively. The system integrated hyperspectral, 3D, and X-ray imaging sensors. A LabVIEW program was developed to acquire color images, spectral images, depth images, X-ray images of onions, and measure the weight of onions. With the multimodal data collected, algorithms were developed to calculate the maximum diameter, volume, density, and detect latent defects of onions. Three groups of sweet onions (regular, inoculated with Burkholderia cepacia, and inoculated with Pseudomonas viridiflava) were tested. Results showed that the system accurately measured the weight (RMSE = 3.6 g), diameter (RMSE = 1.7 mm), volume (RMSE = 16.5 cm3), and density (RMSE = 0.03 g/cm3) of onions, and correctly classified 88.9% healthy and defective onions. This work demonstrated a promising approach to evaluate both external and internal quality parameters of onions, which is applicable to onion packinghouses. The proposed system and methods are also potentially applicable to quality inspection of other agricultural products.
Chugunov, S.; Li, C.
In: Computer Physics Communications, 194, 64-75, 2015.
@article{Chugunov2015b,
title = {Parallel implementation of inverse adding-doubling and Monte Carlo multi-layered programs for high performance computing systems with shared and distributed memory},
author = {S. Chugunov and C. Li},
doi = {10.1016/j.cpc.2015.02.029},
year = {2015},
date = {2015-02-21},
urldate = {2015-02-21},
journal = {Computer Physics Communications, 194, 64-75},
abstract = {Parallel implementation of two numerical tools popular in optical studies of biological materials–Inverse Adding-Doubling (IAD) program and Monte Carlo Multi-Layered (MCML) program–was developed and tested in this study. The implementation was based on Message Passing Interface (MPI) and standard C-language. Parallel versions of IAD and MCML programs were compared to their sequential counterparts in validation and performance tests. Additionally, the portability of the programs was tested using a local high performance computing (HPC) cluster, Penguin-On-Demand HPC cluster, and Amazon EC2 cluster. Parallel IAD was tested with up to 150 parallel cores using 1223 input datasets. It demonstrated linear scalability and the speedup was proportional to the number of parallel cores (up to 150x). Parallel MCML was tested with up to 1001 parallel cores using problem sizes 104–109 photon packets. It demonstrated classical performance curves featuring communication overhead and performance saturation point. Optimal performance curve was derived for parallel MCML as a function of problem size. Typical speedup achieved for parallel MCML (up to 326x) demonstrated linear increase with problem size. Precision of MCML results was estimated in a series of tests - problem size of 106 photon packets was found optimal for calculations of total optical response and 108 photon packets for spatially-resolved results. The presented parallel versions of MCML and IAD programs are portable on multiple computing platforms. The parallel programs could significantly speed up the simulation for scientists and be utilized to their full potential in computing systems that are readily available without additional costs. },
keywords = {},
pubstate = {published},
tppubtype = {article}
}
Parallel implementation of two numerical tools popular in optical studies of biological materials–Inverse Adding-Doubling (IAD) program and Monte Carlo Multi-Layered (MCML) program–was developed and tested in this study. The implementation was based on Message Passing Interface (MPI) and standard C-language. Parallel versions of IAD and MCML programs were compared to their sequential counterparts in validation and performance tests. Additionally, the portability of the programs was tested using a local high performance computing (HPC) cluster, Penguin-On-Demand HPC cluster, and Amazon EC2 cluster. Parallel IAD was tested with up to 150 parallel cores using 1223 input datasets. It demonstrated linear scalability and the speedup was proportional to the number of parallel cores (up to 150x). Parallel MCML was tested with up to 1001 parallel cores using problem sizes 104–109 photon packets. It demonstrated classical performance curves featuring communication overhead and performance saturation point. Optimal performance curve was derived for parallel MCML as a function of problem size. Typical speedup achieved for parallel MCML (up to 326x) demonstrated linear increase with problem size. Precision of MCML results was estimated in a series of tests - problem size of 106 photon packets was found optimal for calculations of total optical response and 108 photon packets for spatially-resolved results. The presented parallel versions of MCML and IAD programs are portable on multiple computing platforms. The parallel programs could significantly speed up the simulation for scientists and be utilized to their full potential in computing systems that are readily available without additional costs.
Jiang, Y.; Li, C.
Detection and Discrimination of Cotton Foreign Matter Using Push-Broom Based Hyperspectral Imaging: System Design and Capability Journal Article
In: PLoS ONE , 10(3), e0121969, 2015.
@article{Y2015,
title = {Detection and Discrimination of Cotton Foreign Matter Using Push-Broom Based Hyperspectral Imaging: System Design and Capability},
author = {Y. Jiang and C. Li},
doi = {10.1371/journal.pone.0121969},
year = {2015},
date = {2015-02-06},
urldate = {2015-02-06},
journal = {PLoS ONE , 10(3), e0121969},
abstract = {Cotton quality, a major factor determining both cotton profitability and marketability, is affected by not only the overall quantity of but also the type of the foreign matter. Although current commercial instruments can measure the overall amount of the foreign matter, no instrument can differentiate various types of foreign matter. The goal of this study was to develop a hyperspectral imaging system to discriminate major types of foreign matter in cotton lint. A push-broom based hyperspectral imaging system with a custom-built multi-thread software was developed to acquire hyperspectral images of cotton fiber with 15 types of foreign matter commonly found in the U.S. cotton lint. A total of 450 (30 replicates for each foreign matter) foreign matter samples were cut into 1 by 1 cm2 pieces and imaged on the lint surface using reflectance mode in the spectral range from 400-1000 nm. The mean spectra of the foreign matter and lint were extracted from the user-defined region-of-interests in the hyperspectral images. The principal component analysis was performed on the mean spectra to reduce the feature dimension from the original 256 bands to the top 3 principal components. The score plots of the 3 principal components were used to examine clusterization patterns for classifying the foreign matter. These patterns were further validated by statistical tests. The experimental results showed that the mean spectra of all 15 types of cotton foreign matter were different from that of the lint. Nine types of cotton foreign matter formed distinct clusters in the score plots. Additionally, all of them were significantly different from each other at the significance level of 0.05 except brown leaf and bract. The developed hyperspectral imaging system is effective to detect and classify cotton foreign matter on the lint surface and has the potential to be implemented in commercial cotton classing offices.},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
Cotton quality, a major factor determining both cotton profitability and marketability, is affected by not only the overall quantity of but also the type of the foreign matter. Although current commercial instruments can measure the overall amount of the foreign matter, no instrument can differentiate various types of foreign matter. The goal of this study was to develop a hyperspectral imaging system to discriminate major types of foreign matter in cotton lint. A push-broom based hyperspectral imaging system with a custom-built multi-thread software was developed to acquire hyperspectral images of cotton fiber with 15 types of foreign matter commonly found in the U.S. cotton lint. A total of 450 (30 replicates for each foreign matter) foreign matter samples were cut into 1 by 1 cm2 pieces and imaged on the lint surface using reflectance mode in the spectral range from 400-1000 nm. The mean spectra of the foreign matter and lint were extracted from the user-defined region-of-interests in the hyperspectral images. The principal component analysis was performed on the mean spectra to reduce the feature dimension from the original 256 bands to the top 3 principal components. The score plots of the 3 principal components were used to examine clusterization patterns for classifying the foreign matter. These patterns were further validated by statistical tests. The experimental results showed that the mean spectra of all 15 types of cotton foreign matter were different from that of the lint. Nine types of cotton foreign matter formed distinct clusters in the score plots. Additionally, all of them were significantly different from each other at the significance level of 0.05 except brown leaf and bract. The developed hyperspectral imaging system is effective to detect and classify cotton foreign matter on the lint surface and has the potential to be implemented in commercial cotton classing offices.
Berry impact record device (BIRD)
Please visit the BIRD project for details.
Electronic nose (e-nose)
2015
Konduru, T.; Rains, G.; Li, C.
Detecting sour skin infected onions using a customized gas sensor array Journal Article
In: Journal of Food Engineering, 160, 19-27, 2015.
@article{Konduru2015,
title = {Detecting sour skin infected onions using a customized gas sensor array},
author = {T. Konduru and G. Rains and C. Li},
doi = {10.1016/j.jfoodeng.2015.03.025},
year = {2015},
date = {2015-03-11},
urldate = {2015-03-11},
journal = {Journal of Food Engineering, 160, 19-27},
abstract = {The overall goal of this study was to test a customized gas sensor array in its ability to detect an important postharvest disease (sour skin) in onions. The sensor array consists of seven metal oxide semiconductor gas sensors and a microcontroller-based automatic data logging system. Three features were extracted from the sensor responses and three baseline correction methods were employed to correct the sensors’ responses. The gas sensor array was tested in two separate experiments with two treatments (control and sour skin). The multivariate data analysis revealed that the “relative response” feature combined with relative baseline correction method provided the best discrimination of infected onions among healthy ones. The best performance (85%) was achieved by using the support vector machine model when the data collected from an independent experiment were used for validation. The study demonstrated the potential of a gas sensor array to detect sour skin-infected onions placed among healthy onions in storage.},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
The overall goal of this study was to test a customized gas sensor array in its ability to detect an important postharvest disease (sour skin) in onions. The sensor array consists of seven metal oxide semiconductor gas sensors and a microcontroller-based automatic data logging system. Three features were extracted from the sensor responses and three baseline correction methods were employed to correct the sensors’ responses. The gas sensor array was tested in two separate experiments with two treatments (control and sour skin). The multivariate data analysis revealed that the “relative response” feature combined with relative baseline correction method provided the best discrimination of infected onions among healthy ones. The best performance (85%) was achieved by using the support vector machine model when the data collected from an independent experiment were used for validation. The study demonstrated the potential of a gas sensor array to detect sour skin-infected onions placed among healthy onions in storage.
Konduru, T.; Rains, G.; Li, C.
A customized metal oxide semiconductor-based gas sensor array for onion quality evaluation: system development and characterization Journal Article
In: Sensors, 15(1), 1252-1273, 2015.
@article{Konduru2015b,
title = {A customized metal oxide semiconductor-based gas sensor array for onion quality evaluation: system development and characterization},
author = {T. Konduru and G. Rains and C. Li},
doi = {10.3390/s150101252},
year = {2015},
date = {2015-01-04},
urldate = {2015-01-04},
journal = {Sensors, 15(1), 1252-1273},
abstract = {A gas sensor array, consisting of seven Metal Oxide Semiconductor (MOS) sensors that are sensitive to a wide range of organic volatile compounds was developed to detect rotten onions during storage. These MOS sensors were enclosed in a specially designed Teflon chamber equipped with a gas delivery system to pump volatiles from the onion samples into the chamber. The electronic circuit mainly comprised a microcontroller, non-volatile memory chip, and trickle-charge real time clock chip, serial communication chip, and parallel LCD panel. User preferences are communicated with the on-board microcontroller through a graphical user interface developed using LabVIEW. The developed gas sensor array was characterized and the discrimination potential was tested by exposing it to three different concentrations of acetone (ketone), acetonitrile (nitrile), ethyl acetate (ester), and ethanol (alcohol). The gas sensor array could differentiate the four chemicals of same concentrations and different concentrations within the chemical with significant difference. Experiment results also showed that the system was able to discriminate two concentrations (196 and 1964 ppm) of methlypropyl sulfide and two concentrations (145 and 1452 ppm) of 2-nonanone, two key volatile compounds emitted by rotten onions. As a proof of concept, the gas sensor array was able to achieve 89% correct classification of sour skin infected onions. The customized low-cost gas sensor array could be a useful tool to detect onion postharvest diseases in storage.},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
A gas sensor array, consisting of seven Metal Oxide Semiconductor (MOS) sensors that are sensitive to a wide range of organic volatile compounds was developed to detect rotten onions during storage. These MOS sensors were enclosed in a specially designed Teflon chamber equipped with a gas delivery system to pump volatiles from the onion samples into the chamber. The electronic circuit mainly comprised a microcontroller, non-volatile memory chip, and trickle-charge real time clock chip, serial communication chip, and parallel LCD panel. User preferences are communicated with the on-board microcontroller through a graphical user interface developed using LabVIEW. The developed gas sensor array was characterized and the discrimination potential was tested by exposing it to three different concentrations of acetone (ketone), acetonitrile (nitrile), ethyl acetate (ester), and ethanol (alcohol). The gas sensor array could differentiate the four chemicals of same concentrations and different concentrations within the chemical with significant difference. Experiment results also showed that the system was able to discriminate two concentrations (196 and 1964 ppm) of methlypropyl sulfide and two concentrations (145 and 1452 ppm) of 2-nonanone, two key volatile compounds emitted by rotten onions. As a proof of concept, the gas sensor array was able to achieve 89% correct classification of sour skin infected onions. The customized low-cost gas sensor array could be a useful tool to detect onion postharvest diseases in storage.
Thermography
2017
Kuzy, J. D.; Jiang, Y.; Li, C.
Blueberry bruise detection by pulsed thermographic imaging Journal Article
In: Postharvest Biology and Technology, 136, 166-177, 2017.
@article{Kuzy2017,
title = {Blueberry bruise detection by pulsed thermographic imaging},
author = {J. D. Kuzy and Y. Jiang and C. Li},
url = {http://sensinglab.engr.uga.edu//srv/htdocs/wp-content/uploads/2019/11/Blueberry-bruise-detection-by-pulsed-thermographic-imaging.pdf},
doi = {10.1016/j.postharvbio.2017.10.011},
year = {2017},
date = {2017-10-24},
urldate = {2017-10-24},
journal = {Postharvest Biology and Technology, 136, 166-177},
abstract = {Kuzy, J., Jiang, Y., & Li, C. (2018). Blueberry bruise detection by pulsed thermographic imaging. Postharvest Biology and Technology, 136, 166-177.
Blueberries are prone to internal bruising damage during harvesting and postharvest handling. Accurate assessment of bruising damage improves profitability by allowing allocation of berries to appropriate product streams. The goal of this study was to develop a pulsed thermographic imaging system and explore its feasibility in non-destructively detecting bruised blueberries. In this paper, the design and construction of a pulsed thermographic imaging system was described. A total of 200 blueberry fruit samples from two southern highbush cultivars (Farthing and Meadowlark) were collected and bruising treatments were applied to half of the samples. Relevant features were extracted and were demonstrated to be significantly different between healthy and bruised fruit. Classification was performed using linear discriminant analysis, support vector machine, random forest, k-nearest-neighbors, and logistic regression classifiers. Accuracies of up to 88% and 79% were obtained for Farthing and Meadowlark berries, respectively. These results demonstrate the feasibility of pulsed thermography to discriminate between bruised and healthy blueberries.},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
Kuzy, J., Jiang, Y., & Li, C. (2018). Blueberry bruise detection by pulsed thermographic imaging. Postharvest Biology and Technology, 136, 166-177.
Blueberries are prone to internal bruising damage during harvesting and postharvest handling. Accurate assessment of bruising damage improves profitability by allowing allocation of berries to appropriate product streams. The goal of this study was to develop a pulsed thermographic imaging system and explore its feasibility in non-destructively detecting bruised blueberries. In this paper, the design and construction of a pulsed thermographic imaging system was described. A total of 200 blueberry fruit samples from two southern highbush cultivars (Farthing and Meadowlark) were collected and bruising treatments were applied to half of the samples. Relevant features were extracted and were demonstrated to be significantly different between healthy and bruised fruit. Classification was performed using linear discriminant analysis, support vector machine, random forest, k-nearest-neighbors, and logistic regression classifiers. Accuracies of up to 88% and 79% were obtained for Farthing and Meadowlark berries, respectively. These results demonstrate the feasibility of pulsed thermography to discriminate between bruised and healthy blueberries.
Blueberries are prone to internal bruising damage during harvesting and postharvest handling. Accurate assessment of bruising damage improves profitability by allowing allocation of berries to appropriate product streams. The goal of this study was to develop a pulsed thermographic imaging system and explore its feasibility in non-destructively detecting bruised blueberries. In this paper, the design and construction of a pulsed thermographic imaging system was described. A total of 200 blueberry fruit samples from two southern highbush cultivars (Farthing and Meadowlark) were collected and bruising treatments were applied to half of the samples. Relevant features were extracted and were demonstrated to be significantly different between healthy and bruised fruit. Classification was performed using linear discriminant analysis, support vector machine, random forest, k-nearest-neighbors, and logistic regression classifiers. Accuracies of up to 88% and 79% were obtained for Farthing and Meadowlark berries, respectively. These results demonstrate the feasibility of pulsed thermography to discriminate between bruised and healthy blueberries.
Kuzy, J. D.; Li, C.
A Pulsed Thermographic Imaging System for Detection and Identification of Cotton Foreign Matter Journal Article
In: Sensors, 17(3), 518, 2017.
@article{Kuzy2017b,
title = {A Pulsed Thermographic Imaging System for Detection and Identification of Cotton Foreign Matter },
author = {J. D. Kuzy and C. Li},
url = {http://sensinglab.engr.uga.edu//srv/htdocs/wp-content/uploads/2019/11/A-Pulsed-Thermographic-Imaging-System-for-Detection-and-Identification-of-Cotton-Foreign-Matter.pdf},
doi = {10.3390/s17030518},
year = {2017},
date = {2017-03-02},
urldate = {2017-03-02},
journal = {Sensors, 17(3), 518},
abstract = {Kuzy, J., & Li, C. (2017). A pulsed thermographic imaging system for detection and identification of cotton foreign matter. Sensors, 17(3), 518.
Detection of foreign matter in cleaned cotton is instrumental to accurately grading cotton quality, which in turn impacts the marketability of the cotton. Current grading systems return estimates of the amount of foreign matter present, but provide no information about the identity of the contaminants. This paper explores the use of pulsed thermographic analysis to detect and identify cotton foreign matter. The design and implementation of a pulsed thermographic analysis system is described. A sample set of 240 foreign matter and cotton lint samples were collected. Hand-crafted waveform features and frequency-domain features were extracted and analyzed for statistical significance. Classification was performed on these features using linear discriminant analysis and support vector machines. Using waveform features and support vector machine classifiers, detection of cotton foreign matter was performed with 99.17% accuracy. Using frequency-domain features and linear discriminant analysis, identification was performed with 90.00% accuracy. These results demonstrate that pulsed thermographic imaging analysis produces data which is of significant utility for the detection and identification of cotton foreign matter.},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
Kuzy, J., & Li, C. (2017). A pulsed thermographic imaging system for detection and identification of cotton foreign matter. Sensors, 17(3), 518.
Detection of foreign matter in cleaned cotton is instrumental to accurately grading cotton quality, which in turn impacts the marketability of the cotton. Current grading systems return estimates of the amount of foreign matter present, but provide no information about the identity of the contaminants. This paper explores the use of pulsed thermographic analysis to detect and identify cotton foreign matter. The design and implementation of a pulsed thermographic analysis system is described. A sample set of 240 foreign matter and cotton lint samples were collected. Hand-crafted waveform features and frequency-domain features were extracted and analyzed for statistical significance. Classification was performed on these features using linear discriminant analysis and support vector machines. Using waveform features and support vector machine classifiers, detection of cotton foreign matter was performed with 99.17% accuracy. Using frequency-domain features and linear discriminant analysis, identification was performed with 90.00% accuracy. These results demonstrate that pulsed thermographic imaging analysis produces data which is of significant utility for the detection and identification of cotton foreign matter.
Detection of foreign matter in cleaned cotton is instrumental to accurately grading cotton quality, which in turn impacts the marketability of the cotton. Current grading systems return estimates of the amount of foreign matter present, but provide no information about the identity of the contaminants. This paper explores the use of pulsed thermographic analysis to detect and identify cotton foreign matter. The design and implementation of a pulsed thermographic analysis system is described. A sample set of 240 foreign matter and cotton lint samples were collected. Hand-crafted waveform features and frequency-domain features were extracted and analyzed for statistical significance. Classification was performed on these features using linear discriminant analysis and support vector machines. Using waveform features and support vector machine classifiers, detection of cotton foreign matter was performed with 99.17% accuracy. Using frequency-domain features and linear discriminant analysis, identification was performed with 90.00% accuracy. These results demonstrate that pulsed thermographic imaging analysis produces data which is of significant utility for the detection and identification of cotton foreign matter.
For more papers please visit the full publication list.