注射制品表面缺陷在线检测与自动识别.pdfVIP

模具工业 2013年第39卷第7期 7 注射制品表面缺陷在线检测与 自动识别 杨 威 ，毛 霆，张 云，周华民 (华中科技大学 材料成形与模具技术国家重点实验室，湖北 武汉 430074) 摘要：基于机器视觉技术的注射制品表面缺陷检测与识别可有效解决人工抽样检测问题，为克服现有 缺陷识别算法对于不同的制品需分别进行样本训练、图像质量要求高、可操作性差等难题，在采用图像 处理技术对制品表面缺陷进行检测的同时，提出一种基于缺陷区域轮廓、制品轮廓、区域灰度等特征的 缺陷自动识别算法，避免了大量制品图像样本的训练过程，提高了可操作性。基于该方法，开发 了一套 注射制品表面缺陷在线检测与识别系统，试验表明，对短射、飞边、裂纹3种常见表面缺陷的识别率为 91．8％ 。 关键词：注射制品；缺陷检测；自动识别；机器视觉 中图分类号：TG76；TP391．77 文献标识码：A 文章编号：1001—2168(2013)07—0007—06 on．1inesurfacedefectsdetection and automaticrecognition of injectionmouldingproducts YANG Wei，MAO Ting，ZHANG Yun，ZHOU Hua-min (StateKeyLaboratoryofMaterialProcessingandDie＆MouldTechnology．Huazhong UniversityofScience Technology，Wuhan，Hubei43o074，China) Abstract：The automatic surface defects detection nad recognition system for injection moulding products based on machine vision can effectively solve the problem matartificial samplnig detections were of low automation degree and efficiency．In order t0 overcome thathte existnig defects recognition algorithms needed samplestraining ofr differentprod- ucts respectively，required high qualiyt imagesbuthad poormnaeuverabiliyt，Using image processnig algorithms to detect surface defects，the defects automatic recognition algo- ritmh s based on the features of defect region profiles．productprofiles and regionalgray were proposed to avoid sma ples training na d getgood mna euverability．And also na auto- matic surface defects detection nad recognition system for injection moulding products based on the algorithms was developed．The experimental results showed hte recognition accuracy rateofr shortshot，flash，crack is91．8％． Keywords：injectionmouldingproduct；defects detection；automaticrecognition；machinevi- s