基于机器视觉的小麦害虫分割算法研究

农业病虫害的自动识别是精准农业研究方向之一。以小麦蚜虫为例，运用机器视觉技术对非特定场景下害虫的分类和分割算法进行了研究。在分类上，训练了SVM分类器和基于k-均值聚类的分类方法。比较得出，SVM分类器和k-均值聚类算法在处理精度和速度上各有优势；在分割上，运用合并和分裂相结合的区域生长算法分割害虫和叶片，进行自动识别。分析表明，该算法对害虫的分类效果好、分割识别准确率达到90．7％，速度能够满足实时处理的要求，为农业机械精准施药提供了技术上的支持。

Algorithm for segmentation of insect pest images from wheat leaves based on machine vision

Automatic inspection is one of the research directions of precision agriculture.Taking the example of Aphid,the authors studied the algorithms for automatically classifying and segmenting insect pest images in nonspecific environment respectively based on machine vision.For classification, a Support Vector Machine(SVM) sorter was trained and the method of k-means clustering algorithm was developed.Compared with SVM sorter,kmeans has its superiority in rate and SVM has superiority in precision.For segmentation,region-growing algorithm which combined merge and fission was applied to segment the images of insect pest and wheat leaves.The analysis indicate that effect of classification is good and recognition accuracy is 90.7%,the speed can meet the requirement of real-time image processing.These algorithms provided the technology support for precision spraying in agricultural mechanization.