基于优化SVM的虫害图像识别研究  

Research on Insect Pest Image Recognition Based on Optimized SVM

作　　者：马佳佳 陈友鹏[2] 王克强[1] 刘展眉 温艳兰 林钦永 蔡肯[1] Ma Jiajia;Chen Youpeng;Wang Keqiang;Liu Zhanmei;Wen Yanlan;Lin Qinyong;Cai Ken(Zhongkai University of Agriculture and Engineering,Guangzhou 510225;Guangzhou Nanyang Polytechnic College,Guangzhou 510925)

机构地区：[1]仲恺农业工程学院,广州510225 [2]广州南洋理工职业学院,广州510925

出　　处：《中国粮油学报》2022年第5期10-15,共6页Journal of the Chinese Cereals and Oils Association

基　　金：广东省普通高校重点领域专项(2019GZDXM007)。

摘　　要：农作物虫害问题严重威胁我国农业及粮食生产安全。针对传统虫害识别智能化水平低、效率低等问题,提出一种基于优化SVM的虫害图像识别方法。本研究以草地贪夜蛾成虫图像作为实验对象,采用HOG特征描述符提取图像特征信息,通过粒子群优化算法对SVM模型的内部参数进行寻优。模型在经过训练后,对简单背景下虫害图像的识别准确率达100%,对复杂背景下样本的识别准确率达93.89%。模型识别效果明显优于其他对比模型,为机器学习在农作物虫害识别中的应用提供一定参考。The problem of crop pests has seriously threatened the safety of agriculture and food production in China.Aiming at the problems of low intelligence level and low efficiency of traditional pest identification,a pest image identification method based on optimized SVM was proposed.In the present study,the adult image of the grassland moth was taken as the experimental object.The HOG feature descriptor was used to extract the image feature information,and the internal parameters of the SVM model were optimized by the particle swarm optimization algorithm.After training,the identification accuracy of the model for pest images under simple background was 100%,and the identification accuracy for samples under complex background was 93.89%.The identification effect of the model was obviously better than other comparative models,which provides some reference for the application of machine learning in crop pest identification.

关 键 词：虫害识别 机器学习 图像处理 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]