Development of an expert vision-based system for inspecting rice quality indices

In this study, a computer vision system comprising of a special rice tray, scanner, and computer-aided processing software was developed to assess rice appearance quality. The applicability of the system was evaluated for assessment of four rice varieties. Rice grains were accurately (>98%) classified into whole and broken kernels regarding their dimensional features estimated precisely with coefficient of determination (R2) of more than 98% and root mean squared error of 0.08. Optimal thresholding on the vertical coefficient of wavelet transform resulted in fissure detection with an accuracy of 96.51%. Red and black spots of the rice kernels were also precisely detected by thresholding on the red colour difference and gray-scale components respectively. Results indicated that very high accuracies (R2 about 99%) were obtained for whiteness and chalkiness measurements. It was concluded that the image processing technique has a significant potential to be applied for appearance quality assessment of rice kernels.