黄瓜蚜虫的图像识别与计数方法  

Image Recognition and Counting for Glasshouse Aphis gossypii

作　　者：邱白晶[1] 王天波[1] 李娟娟[1] 李坤[1] 

机构地区：[1]江苏大学现代农业装备与技术省部共建教育部重点实验室,镇江212013

出　　处：《农业机械学报》2010年第8期151-155,共5页Transactions of the Chinese Society for Agricultural Machinery

基　　金：国家自然科学基金资助项目(30571240);国家"863"高技术研究发展计划资助项目(2008AA100905;2006AA10A305-3;2008AA100901)

摘　　要：通过分析蚜虫区域、绿色背景和蚜叶区的G分量特点,建立G分量阈值确定原则,并采用G分量阈值将蚜虫区域和非蚜虫区域分离开。针对蚜虫的粘连重叠问题,利用扩展极小值阈值变换的方法对输入图像进行标记,对标记后的图像进行距离变换和分水岭分割,以去除粘连。试验结果表明:算法能有效地分割粘连重叠的蚜虫,过分割率与欠分割率之和为3.14%。计数准确率达到96.2%,高于直接计数的80.7%。Characteristics of the Aphis gossypii area,background and mixed area （include Aphis gossypii area and background） were analyzed and principle of determining was establish based on the threshold G component. Then Aphis gossypii area and non-Aphis gossypii area were separated using the threshold G component. For the overlapping Aphis gossypii,the input image were marked using the minimum extension transform,then distance transform and watershed algorithm was applied to the marked image,and the overlapping was removed. Experimental results showed that this algorithm could effectively segment the overlapping Aphis gossypii. The sum of over-segmentation rate and under-segmentation rate was 3.14%. The accurate rate was 96.2%,which was higher than the direct counting.

关 键 词：黄瓜蚜虫 温室 图像处理 粘连 检测 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术] S431[自动化与计算机技术—计算机科学与技术]