A Rice Quality Detection System Based On Computer Vision

The broken rate, chalky degree and yellow rate are important factors of the rice quality. It also reflects the level of rice planting skill and the rice machining skill. The system designed by this article could effectively detect the rice broken rate, chalky degree and yellow rate. The system could be used to detect rice quality by the testing organizations or to monitor the rice quality by the rice producers real-timely.The system realization algorithm is based on digital image processing and pattern recognition knowledge. At first we introduced the block diagram of this article, then we introduced the description of image acquisition system in details. At last we introduce the algorithms of how to process the rice image, give a discussion about the algorithms and choose the most suitable algorithms for the system. The system realization algorithm includes the preprocessing algorithm and the detection algorithm. First step of image pre-processing algorithm is convert the RGB color image to gray image, and then we discussed the median filtering, Gaussian filtering, Mean filtering, Laplacian sharpening, algorithms, and finally using the Mean filter combined with the Laplacian sharpening to preprocess the image. Then we introduced three ways to detect the broken rice:Aspect ratio detection, Area ratio detection and perimeter detection. After that we give a new comprehensive way to detect the broken rate of the rice. The chalky degree detection is based on threshold segmentation, which we give a suitable threshold for the system. At last we discussed some color spaces for the yellow rice detection, choose the CMY and HSI color space for this system.The system detected the quality of rice in three aspects:The broken rate, chalky degree and yellow rate. The system is through experimental methods, adjusting the algorithm and parameter data regularly, and finally got satisfactory results. At the end of the article, we introduced the software designed based on above algorithms, and at last give a conclusion of this article.