节点文献

组合光源与图像处理算法在工件表面缺陷检测中的应用

The Application of Combinatorial Lamp-house and Image Processing Algorithm in Workpiece Surface Defect Inspecting

【作者】 孔祥伟

【导师】 曲兴华;

【作者基本信息】 天津大学 , 测试计量技术及仪器, 2007, 硕士

【摘要】 为了不断提高产品质量和生产效率,金属工件表面缺陷在线自动检测技术在生产过程中显得日益重要。本课题的研究对象是金属镀层工件表面上的多种随机缺陷,目的是把有缺陷的工件检测出来,分选出来,从而实现质量检测的目的。针对本课题不同金属镀层工件表面强反射程度不同和工件表面缺陷种类繁多、位置随机等的具体情况,以及工件缺陷检测过程中不能出现缺陷工件漏检的要求,结合表面检测技术发展并在技术上力图创新及结果上力图经济实用的目的,设计了一套能够实现对金属工件表面缺陷进行双面、高速、高分辨率的自动检测系统。检测系统以视觉检测理论为检测原理,软件设计以模块化为设计思想,主要完成以下工作:1.针对本系统中金属镀层工件的强反射特性,充分考虑了照明技术对缺陷检测的影响,研究了多种照明方式的检测效果,最终为检测系统设计了高角度照明和低角度照明的组合方式的照明光源。2.阈值选取模块引入大津阈值分割法,工件阈值和缺陷阈值采用两次分别选取阈值,并针对本系统采集到的工件缺陷图像情况对大津阈值分割法进行了有效的改进。3.缺陷识别模块针对不同缺陷形状的区别,编写算法求出缺陷的一组形状描述特征参数,设计了二叉树分类器,判别规则采用相互独立的不同特征参数的组合,通过大量的实验,得到缺陷识别的特征参数分割阈值,对缺陷类型进行了准确的识别。4.完善了软件系统,对软件系统进行了大量地测试,避免了重复性计算,防止内存泄露,提高了软件的稳定性和高效性。5.缺陷检测软件设计了去光圈的灵敏度系数和缺陷检测的灵敏度系数,软件系统可以针对不同形状,不同镀层的工件表面方便地调整这两个系数以适应系统的技术要求。

【Abstract】 In order to continually raise the speed and efficiency of production, the automatic and on-line inspecting technology about surface defect of metal production becomes more and more important. What this thesis being studied is the stochastic defects on the surface of electroplated workpieces. We expect to make out all the defective workpieces and take them off and let the quality detection come true.There are some difficulties in this case as follows: different metal surface can cause different strong reflection, the kinds of different detect is so many and the position of detect is random, and what is more, any defect production can’t be neglected. In order to innovate surface detect inspection technology and realize economical results, after analyzing various ready inspection methods and considering specific technical requirements, This paper designs the automatic inspection system, which can realize two-side, high speed and high resolution inspection.The system looks on computer vision detection theory as inspecting principle.The software system introduces modular design. What I have done is as followed:1. Considering the strongly reflective production surface, I have studied the influence of lighting engineering to the defect inspection, and the detection results of different illumination system. I design out the illuminating light that is assembled by high-angle illumination and low-angle illumination.2. The threshold modularization introduces Ostu algorithm threshold method. It designs a method to computes the workpiece threshold and flaw threshold by choosing the thresholds separately in two times. I have improved Ostu algorithm threshold method to suit this detection system.3. The flaw pattern recognition module designs the algorithm to compute a set of shape description characteristics of the flaws and adopts multiple classifiers, based on the shape difference of the flaw. The rule of every layer is a set of mutual independent flaw parameters, considering the difference of the flaws. I get the characteristics parameter thresholds through a lot of experiments, and the parameter thresholds can differentiate the flaws clearly.4. I perfect the software system. By making enough software testing, the software system can avoid double counting and avoid EMS memory leaking. The stability and efficiency of the software system have been improved.5. The inspecting software designs the sensitivity coefficient of wiping aperture and the inspection sensitivity coefficient. Basing on different shape and different plating coat of the production surface, it can adjust the sensitivity coefficients to fit the system technique requirement.

  • 【网络出版投稿人】 天津大学
  • 【网络出版年期】2009年 04期
  • 【分类号】TP391.41
  • 【被引频次】73
  • 【下载频次】1271
  • 攻读期成果
节点文献中: 

本文链接的文献网络图示:

本文的引文网络