基于多视角模板匹配的零件图像检索方法  

Workpiece Image Retrieval Method Based on Multi-view Template Matching

作　　者：苏欣 赖复尧[1] 余容平 李柏林[2] 熊鹰[2] SU Xin;LAI Fuyao;YU Rongping;LI Bailin;XIONG Ying(Southwest Institute of Electronic Technology,Chengdu 610036,China;School of Mechanical Engineering,Southwest Jiaotong University,Chengdu 610031,China)

机构地区：[1]西南电子技术研究所,四川成都610036 [2]西南交通大学机械工程学院,四川成都610031

出　　处：《机械制造与自动化》2023年第1期222-225,229,共5页Machine Building & Automation

摘　　要：针对工业场景下零件的自动化生产需求,提出一种基于多视角模板匹配的零件图像检索方法。提取零件的尺寸信息对零件进行粗分类;对尺寸上难以区分的零件进行动态分块,将分块后的图像进行哈希“指纹”提取,用于计算其相似度,并找出最相似的3个零件。提取上述3个零件侧面信息的哈希“指纹”并计算相似度,将相似度最大值所对应的类别标签作为最终检索结果。实验结果表明:该方法精度高达98.16%,更符合零件的实际生产应用。In light of the demand of automatic production of parts in industrial scene,a part image retrieval method based on multi-view template matching is proposed.The dimension information of the top view of the parts is extracted for rough clarification of the parts.The parts that are difficult to be distinguished in size are dynamically partitioned,and the hash"fingerprint"of the partitioned image is applied to calculate the similarity and find out the three most similar parts.The side information of the above three parts is extracted for hash"fingerprint"and similarity calculation,and the category label corresponding to the maximum similarity is taken as the final retrieval result.The experimental results show that the accuracy of the proposed method is as high as 98.16%,which is more in line with the practical production application of parts.

关 键 词：图像检索 模板匹配 零件 多视角 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]