Abstract

To monitor cucumbers' external quality, such as color changes or the presence of any damage during storage, a machine vision system was used. Red, Green, Blue (RGB) images were acquired in a "soft box," which provided a highly diffused lighting scene for observing visual changes such as color and appearance in the skin of cucumber. The RGB images were transformed into L*, a*,b*, and HSV spaces. Histograms for each channel in each color space were evaluated for image segmentation, and the blue (B) channel in the RGB color space was found superior in terms of measuring damage progression. Damage progression plots (DPP) were made from accumulated grayscale images in each of the color channels and to observe variation over time, absolute differential damage progression (ADDP) plots were generated. Overall, the order of channel utility was [B], [R, G, V], and [H, S, L*, a*, b*]. To assess which channel, in which colorspace, was most sensitive, i.e., could capture most of the information regarding day-to-day color changes, a principal component analysis (PCA) was performed. The PCA showed that all individual components in the RGB color space were suitable for obtaining information about the external changes of cucumber. Based on the results, the machine vision approach is recommended as a non-destructive technique for monitoring the external quality of stored fresh produce.

This work is licensed under a Creative Commons Attribution 4.0 License.