纹理边界引导的复合材料圆孔检测方法  

Detection method for composite hole guided by texture boundary

作　　者：张太恒 梅标 乔磊 杨浩杰 朱伟东[1] ZHANG Tai-heng;MEI Biao;QIAO Lei;YANG Hao-jie;ZHU Wei-dong(School of Mechanical Engineering,Zhejiang University,Hangzhou 310027,China;Institute of Advanced Technology,Zhejiang University,Hangzhou 310027,China)

机构地区：[1]浙江大学机械工程学院,浙江杭州310027 [2]浙江大学先进技术研究院,浙江杭州310027

出　　处：《浙江大学学报（工学版）》2020年第12期2294-2300,共7页Journal of Zhejiang University：Engineering Science

基　　金：国家自然科学基金资助项目(51675479,51805478)。

摘　　要：针对复合材料纹理背景下圆形基准孔检测困难的问题,提出一种纹理边界引导的圆孔检测方法.该方法将局部三进制模式(LTP)与灰度共生矩阵(GLCM)对比度融合为纹理对比度,通过提取纹理对比度特征实现孔内纹理和孔外纹理的快速分割,得到与圆孔边界近似吻合的纹理边界,利用纹理边界的位置信息去除绝大多数非圆孔边界的边缘点,利用纹理边界的连通信息对剩余边缘点进行分组,使用内嵌圆参数统计机制的随机圆检测算法从每组边缘点各检出1个圆孔目标,进而完成对多个圆孔目标的检测.实验结果表明,在复合材料圆孔检测场景中该方法有94%以上检出率,3%以下检错率和较高的检测速度,并表现出良好的检测鲁棒性.A novel hole detection method guided by texture boundary was proposed,aiming to solve the detection problem of composite circular reference holes.The texture contrast based on local ternary pattern(LTP)and graylevel co-occurrence matrix(GLCM)was extracted first to perform the fast segmentation of intra-hole texture and out-of-hole texture,which could obtain the texture boundaries closely matching the hole boundaries.Then,the position information of texture boundaries was used to remove most of the edge points that belong to non-hole boundaries.The connectivity information of texture boundaries was used to group the remaining edge points.Finally,the randomized circle detection method with circle-parameter statistics mechanism was applied to detect a single circle from each group of edge points,which completed the detection of multiple hole targets.Experimental results showed that this method had a detection rate of more than 94%,an error rate of less than 3%,a high detection speed and good detection robustness in the composite hole detection scene.

关 键 词：局部二进制模式(LBP) 局部三进制模式(LTP) 灰度共生矩阵(GLCM) 纹理分割 圆检测 机器人制孔 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]