Factors Affecting the Discrimination of Weed from Chinese Kale Seed by Image Processing Technique

Abstract

An image processing technique was adapted in an attempt to discrimination of weed from Brassica alboglaba (Chinese kale) seeds, since manual inspection for recognizing weed from vegetable seed usually consumed long time and prone to error. As well, an image processing technique could detect object rapidly, and it is non-destructive technique. Therefore, factors affecting the discrimination of weed from Chinese Kale via image processing technique were assessed in this study. Brassica alboglaba and 2 kind of weed seeds (Slender amaranth: Amaranthus viridis and Pigweed Purslane: Portulaca oleracea) were photographed (100 images per species, 4 cm. above sample holder, 200x200 pixel resolution). Width, length, and area also intensity values of RGB color of weed seeds were examined via a developed Python-based image processing program. The results showed that the average width, length and area for B. alboglaba were significantly larger than for A. viridis and P. oleracea seeds. The intensity values of color of B. alboglaba seed was also significantly different from that for A. viridis and P. oleracea, while the color values of A. viridis and P. oleracea seeds were not significantly different. This study demonstrated that width, length, area and RGB color enable discrimination of weed from Chinese kale seeds.