基于色差聚类的原木图像端面检测与统计  

Logs End Detection and Statistics by Color Difference Clustering

作　　者：唐浩 王克俭[1] 李晓烨 剪文灏[2] 谷建才[3] TANG Hao;WANG Ke-jian;LI Xiao-ye;JIAN Wen-hao;GU Jian-cai(College of Information Science and Technology,Agriculture University of Hebei,Baoding,Hebei 071000,China;Mulan Weichang State-owned Forest Farm Administration of Hebei Province,Chengde,Hebei 067000,China;College of Forestry,Agriculture University of Hebei,Baoding,Hebei 071000,China)

机构地区：[1]河北农业大学信息科学与技术学院,河北保定071000 [2]河北省木兰围场国有林场管理局,河北承德067000 [3]河北农业大学林学院,河北保定071000

出　　处：《计量学报》2020年第6期682-688,共7页Acta Metrologica Sinica

基　　金：河北省高等学校技术研究项目(ZD2016158)。

摘　　要：针对自然环境下外界因素对原木截面检测的干扰,使用图像处理技术,以自然环境中原木堆放存储时图像为处理对象,设计了一种原木截面识别方法。通过色差值聚类将原木图像分割为原木截面、孔隙及背景,去除背景干扰提取原木端面;采用逐级开运算与改进分水岭算法,对端面进行分割计数。实验结果表明:在自然环境下的正检率91.88%,错检率5.08%,漏检率8.12%,满足了原木截面识别计数的需求。Logs end detection is disturbed by the natural environment.An end detection method for logs accumulation state in natural environment is introduced based on image processing technology.Segmentation of logs images into logs sections,pores and backgrounds by color difference clustering to remove background interference and extracting logs end,segmentation counting is performed using hierarchical opening operation and an improved watershed algorithm.The results show that the correct detection rate is 91.88%,the false detection rate is 5.08%,and the missed detection rate is 8.12%under the high interference environment of natural environment.The method satisfies the requirement for the identification and counting of logs.

关 键 词：计量学 图像处理 原木端面检测 色差聚类 逐级开运算 分水岭算法 

分 类 号：TP96[自动化与计算机技术]