摘要
为进一步提高对仓储害虫自动识别的准确度及种类数量,提出一种普适有效的仓储害虫局部特征提取方法,即首先自动判别出图像中仓储害虫的头部和尾部,然后采用基于兴趣区间对偶点分析的方法提取特征。提取出仓储害虫的鞘翅等效长、尾部弧度等7个局部形态学特征,构建了优化的形态学特征空间,采用SAA-SVM分类器对15类常见仓储害虫分类的正确率达到94.8%,解决了仓储害虫分类识别中多种类精确识别的难题。
A universal and effective method for extracting the local features of storage insects was proposed in order to improve the accuracy and the number of species to be recognized further. The method included two steps. Firstly, the head and the tail of the storage insects in the image were determined automatically,and then features were extracted based on the dual point analysis within interesting interval. Seven local morphological features were extracted including coleopteran equivalent length,tail radian of storage insects and so on,and the optimized morphological feature space was constructed. The insects were classified based on simulated annealing algorithm and support vector machine （SAA-SVM）. The results showed that the recognition accuracy was 94.8% for the fifteen species of common storage insects. The precise identification problem of more species was solved in classification of stored-grain pests.
出处
《河南农业科学》
CSCD
北大核心
2014年第2期84-87,共4页
Journal of Henan Agricultural Sciences
基金
国家自然科学基金项目(31101085)
河南省基础与前沿技术研究计划项目(122300410145)
河南省高等学校青年骨干教师资助计划项目(2011GGJS-094)
华北水利水电大学高层次人才科研启动项目(201118)
安阳职业技术学院工程技术类科研项目(AZKYGC-2013A01)
关键词
仓储害虫
形态学特征
特征提取
分类识别
stored-grain pests
morphological features
feature extraction
classification
