Citrus Quality Detection Technology Based On SVM

China is one of the main origin of citrus production. It has more than 4000 years of cultivation history with a variety of citrus and rich resources. The number of annual output of citrus of Hunan is very much, but the citrus post-processing has not been taken enough attention. At present, the citrus quality grading almost depends on manpower and machines, which has low work efficiency and big error. Therefore, The way how to carry out the grading of citrus automatic for reducing labor and improving productivity need to been found is the urgent mission.So, this paper uses image processing technology to classify the citrus. After the preprocess of the citrus images, the shape and size of citrus, which take as the characteristic parameters of citrus, has been computed. And then the kinds of citrus are classified. The work of this paper include:first, the citrus image are collected by a set of simple device, then, this images are classified by people and this is the grading standards. Second, MATLAB is used to preprocess the citrus images, this process include:the RGB image model is changed to HSV model, the S component’s value is used to present gray image, then the gray image has been binarization and median filter is used to realize image denoising, and the boundary information has been got by using Canny. Third, SVM classify algorithm has been studied, and the SVM model has been constructed by the characteristic parameters of citrus, which are perimeter, area and circularity. At last, the test image was predicted by using this SVM model, and the result indicated that the SVM classifier’s precision can reach 95%, this show that this method can be used in automatic identification for citrus.