基于二值图像连通域的甘蔗螟虫识别计数方法

Binary image connected domain based sugarcane borer recognition and counting algorithm

针对人工普查方法费时费力的缺陷,提出一种广域环境下甘蔗螟虫图像自动获取及基于二值图像连通域的识别计数方法。使用最大类方差法获取甘蔗螟虫与背景分割的最佳阈值,采用该阈值将田间甘蔗螟虫图像进行二值化,并对二值图像进行形态学膨胀和腐蚀的预处理。将二值图像转化为二维矩阵,采用深度搜索算法,递归计算二维矩阵中连通域的个数,实现对获取图像中甘蔗螟虫的识别计数。真实田间环境的测试表明,本算法的计数准确率可达96.4%。

To address the issue that traditional method of artificial census waste a lot of time and labors,an approach for automatically acquiring images of sugarcane borer in wide-area environment and recognizing-counting methods based on binary image connected domain are put forward.The optimal split threshold of sugarcane borer and background is obtained by using the maximum between-class variance.The threshold is applied to binarize the image of sugarcane borers and the binary images are subjected to morphological expansion and corrosion pretreatment.The binary image is transformed into a two-dimensional matrix.The number of connected domains in the two-dimensional matrix is recursively calculated via depth search algorithm.In this way,the counting of sugarcane borers in images is realized.The test in the real field environment shows that the counting accuracy of this algorithm can reach up to 96.4%.