基于GA优化SVM的干制红枣品种分类方法  

Study on Classification Method of Jujube Varieties Based on GA Optimized SVM

作　　者：苏军 饶元[1] 张敬尧 李绍稳[1] SU Jun;RAO Yuan;ZHANG Jingyao;LI Shaowen(Anhui Agricultural University, Hefei 230036, China)

机构地区：[1]安徽农业大学信息与计算机学院,安徽合肥230036

出　　处：《洛阳理工学院学报（自然科学版）》2018年第4期65-69,93,共6页Journal of Luoyang Institute of Science and Technology：Natural Science Edition

基　　金：农业部引进国际先进科学技术948项目(2016-X34)

摘　　要：基于遗传算法(GA)优化支持向量机(SVM)分类模型能够显著改善传统SVM的分类精度。以山东大枣、新疆灰枣、新郑大枣以及稷山板枣4类品种的干制红枣为研究对象,首先采用简单线性迭代聚类算法(SLIC)对预处理后的红枣图像进行分割处理;接着针对每类红枣,提取了其6个颜色特征和20个不同角度的纹理特征等26个参数;最后将以上参数输入基于GA优化的SVM分类模型(GA-SVM)。实验结果表明:与传统SVM算法相比,GA-SVM算法对红枣的分类准确率提升了20. 00%。The classification accuracy is significantly improved when optimizing the key parameters of traditional Support Vector Machine (SVM) based on Genetic Algorithm (GA). Taking Chinese dried jujube produced in Shandong, Xinjiang provinces and Xinzheng and Sheshan cities as research objects, the pre-processed jujube images are firstly segmented by simple linear iterative clustering algorithm (SLIC). Then, for each variety of dried red jujubes, 26 parameters including 6 color features and 20 different angle texture features are extracted. Finally, the aforementioned parameters are input into GA-SVM classification model. Results show that, compared with traditional SVM algorithm, the GA-SVM algorithm improves the accuracy rate of red jujubes classification by 20.00%.

关 键 词：遗传算法 支持向量机 干制红枣 品种分类 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]