Method Research And System Implementation Of FPC Defect Detection Based On Machine Vision

FPC is a kind of printed circuit board(PCB),which is light,thin and flexible.Because of these characteristics,in the manufacturing process of the finished flexible circuit board,the product is easily damaged,so the quality inspection work has become a key step in the production process.Most of the current quality inspections are observed by the naked eye of the employees,which is not only inefficient,but also the judgment results are easily affected by objective and subjective conditions.This topic is aimed at the FPC provided by the manufacturer,the size of FPC including multiple irregular lines is 250 mm ? 185 mm.It is designed to realize the design of FPC defect detection algorithm based on machine vision and realization of system,realize the purpose of real-time control quality and improve detection efficiency,and has important application value.Based on the existing FPC surface defect types and based on the existing mechanical structure,this paper designs the process flow and overall solution for the overall requirements of the FPC defect detection system,and completes the joint adjustment and communication of the hardware and software components.This topic realizes the application requirements for defect detection.Then,the sample image and the template image are registered based on the cross-correlation of the gray information,the size of the entire FPC picture is large,and the traditional template matching algorithm cannot correctly match the rotated and translated images.For this problem,firstly,each line is extracted by a mask,and image registration is performed for each line,then the matching accuracy and speed are improved,and then local template matching is first performed to find feature points,and then affine transform is performed according to the feature points to achieve registration of sample images and template images.Then,the related image processing algorithm is studied,and the FPC defect recognition method based on reference comparison method is designed for two types of defects: line defect and pad defect.The difference between light and size in the sample and template diagrams,as well as the noise generated during the process of acquisition,transmission,imaging,etc.,will result in the difference caused by the defect and the non-defective part.For this problem,this article uses two filtering methods in sequence,the image segmentation is first used to remove the non-defective portions,and these portions and defects are very different in grayscale,after filtering out most of the non-defective portions,and these defects are smaller in size,these portions are further removed by morphological processing,and the gray level of these portions is not much different from the defect.Finally,statistical analysis of R channel component data of RGB color space of two types of defects,to achieve defect classification algorithm.Finally,according to the needs analysis and the research objectives of the thesis,the composition of the whole defect detection system,the process flow of the system,the defect detection process,and the visual interface are elaborated.The visualization system integrating FPC detection and classification and traceable detection data.By testing,the detection accuracy of the system reaches 0.1mm,the miss detection rate is less than 1%,and the false detection rate is less than 3%,which improves the reliability and efficiency compared with the traditional detection method.