This study demonstrated the feasibility of using a hyperspectral imaging system with three measurement modes to quantitatively estimate the internal quality of onions. This approach achieved respectable results in comparison to spectroscopy, which has been predominantly used for this purpose in the past. In particular, DM and SSC of onions were predicted more successfully than firmness using the hyperspectral imaging technique. As a nondestructive sampling method, interactance mode performed comparably to or even better than the destructive sampling mode (transmittance) and thus holds potential for use in online quality inspection. Both nondestructive sensing modes can be used to screen onions with high DM content, which has practical value to the onion breeding industry. Further research is needed to study how to implement the interactance mode of the hyperspectral imaging technique in an onion packing line.


Figure 1. Diagram of the line-scan hyperspectral imaging system with interactance mode


Figure 2. (a) Reflectance, (b) interactance, and (c) transmittance modes used for line-scan hyperspectral image acquisition


Figure 3. Flowchart of the data analysis procedures in both the image and spectral domains


Figure 4. Scatter plots of predicted (by partial least square regression) versus actual (measured by destructive methods) values of (a) firmness, (b) SSC, and (c) DM of onions in reflectance, interactance, and transmittance modes. The 45° line was added to indicate the 1:1 ratio

Go back to Project’s page »