An automatic method for identifying different variety of rice seeds using machine vision technology

Abstract

An automatic method for identifying different variety of rice seeds using machine vision technology was investigated, and a detection system which was consisted of an automatic inspection machine and an image-processing unit, was also developed. The system could continually present matrix-positioned rice seed to CCD cameras, and singularize each rice seed image from the background. The inspection machine comprised scattering and positioning devices, a photographing station, a parallel discharging device, and a continuous conveyer belt with carrying holes for the rice seed. The rice seeds' image was achieved continuously by single chip controlled device. The line was suspended per second by the device, and the images of seeds were collected by the camera during the intervals. Image analysis was carried out by Visual C++ 6.0. Color features in RGB (red, green, blue) and color spaces were computed. A back-forward neural network was trained to identify rice seeds. Almost all 86.65% rice seeds were correctly identified. The correct classification rates for five rice varieties were: No.5 ‘Xiannong’ of 99.99%, ‘Jinyougui’ of 99.93%,‘You166’ of 98.89%, No. 3 ‘Xiannong’ of 82.82% and ‘Medium you’ 463 of 86.65%, respectively. Based on the results, it was concluded that the system was enough to use for inspection of varieties of different rice seeds based on their appearance characters of seeds.