基于C－SVM的大米品种识别研究

摘要

提出了一种基于支持向量机（C－SVM）区分大米品种的方法。首先对大米图像进行阈值分割、平滑处理等预处理，并根据大米的粒型特点，提取米粒的面积、周长等6个形态特征。利用Orange Canvas数据挖掘软件先对linear和RBF核函数进行核参数选择，并在Opencv 3．0环境下，编程实现K－means、linear和RBF的3种大米品种识别方法，对10组混合大米图像进行品种测试。试验结果表明，支持向量机线性核函数对大米品种识别具有较高的预测稳定性，识别分类准确率约为99％。%This paper proposed a method based on support vector machine(C-SVM) to distinguish rice varieties.At first, it did the image threshold segmentation, then proceeded the smooth processing.And according to the characteristics of rice grain shape, extracted area, perim-eter and so on, using Orange Canvas data mining software to select kernel parameters of linear and RBF kernel function, and accomplish rice varieties recognition by programing using K means, linear function in SVM and RBF methods under Opencv 3.0.Ten groups of mixed rice were conducted the recognition test, the results showed that linear function in SVM could identify rice varieties in a superior prediction stability with classification accuracy at about 99%.