Research On Surface Inspection System For Ceramic Rings Based On Machine Vision

Insulating ceramic ring has been widely used in electronic devicesdue to its excellent performance, and its quality will directly influence thesafety of devices. At present, most of the domestic manufacturers ofceramic ring still use manual inspection for quality assurance, which isinefficient and costs a lot. With the development of machine vision theoryand computer technology, it has become an inevitable tendency to inspectthe surface quality using machine instead of traditional manual work.Based on the analysis of technical requirements of surface inspectionsystem (SIS) for ceramic rings, a practical SIS scheme using machinevision technology is proposed. The system hardware is designed, as wellas data processing in software. The main contents are as follows:1. Design hardware platform. A series of main modules are designedon the system layer. In view of the specific requirements, selection ofcamera, lens and other equipment is introduced.2. Aiming at the problem that it’s hard to acquire both highrecognition rate and real-time property, a specific inspection processingof SIS is given. To meet the requiring detection rate and speed, the systemuses hierarchical processing logic, which is composed of the fastprocessing and precise processing. Based on slide window, a fastprocessing method is setup to judge whether a ceramic ring containsdefect. Precise processing is used only on defected ones to find out whichtype the defect belongs to. In this way, the system can not only guaranteedetection rate, but also get more efficient.3. The images of defected ceramic rings are usually complex and it’shard to extract defect regions. After analyzing several commonsegmentation algorithms and considering image characteristics, atwo-stage image segmentation method is proposed. The firstsegmentation is to separate ceramic ring from background based on threshold; and the second segmentation adopt adaptive Canny operatorand improved mathematical morphology to extract defect region.4. Analyze the characteristics of4types of ceramic ring defect, andextract feature parameters. To realize the recognition of defects, arule-based defects classification method and DAG-SVM-based defectclassification method are proposed separately. Select200ceramic ringswhich contain various types of defects for classification testing. Therecognition rates of rule-based method and DAG-SVM-based method are83.5%and89%respectively.