Comparison on eggplant fruit grading between nir-color camera and color camera

An eggplant grading system was introduced in a JA (agricultural cooperative association) at Okayama, 2002. 6 CCD color cameras were installed to evaluate fruit appearance (fruit color, size, shape, bruise, and diseases), before and after turning over in a line and connected to 2 PCs through 6 image grabber boards, Physical properties of the fruits such as fruit length, average, maximum, and minimum diameters, area, apparent volume, fruit color, calyx color, fruit shape, degree of fruit bend, bruise number, bruise area, and so on were extracted from the images. The 6 CCD color cameras were divided into 2 units, which consisted of 3 cameras each set, were installed for complete surface inspection. However it is not easy to detect several kinds of defects on eggplant fruit surface because the fruit has very dark purple color and several defects are very similar colors with normal skin. In this paper, a new color camera whose sensitivity ranges from visible region to infrared region was used. Most of the fruits have higher spectral reflectance in infrared region (700-1200nm). The color CCD camera had the same G and B components with usual color camera, but R signal includes infrared region was specially ordered. Although a color balance among R, G, and B is lost, it was considered that slight color changes of defects and discrimination from dark background were easier than the usual color CCD camera. HSI and chromaticity conversions were tested as a preprocessing to detect various defects. With the improvement on the detections of bruises and defects, this will further increase the effectiveness of the current eggplant-grading machine.