Research And Application For FPC Detection Based On Machine Vision

FPC is one kind of printed circuit board and has become the new darling of electronic products in recent years with its high realibility and great flexibility.At present,there are still large technicial gap of defect detection for FPC.Due to the lack of advanced detection technology and equipment,product quality inspection has been stuck in traditional artificial vision test phase,not only causes labor intensive but also has high miss rate and false rate,can not meet the needs of businesses and consumers.Detection method based on machine vision provides a way to solve automated detection of FPC surface defects.This article combined flexible detection algorithms and real-time motion control platform to achieve high-speed and high-precision FPC automatic defect detection.Aiming at several common types of surface defects FPC,this paper designed overall program for FPC defect detection system and set up hardware platform for FPC automatic visual detect system,including mechanical structure,lighting collection system and motion control system.Color information characteristics of FPC background and defects was taken pictures by the use of self-designed combined light and color CCD camera,differences in imaging in RGB color space of different color channels can effectively distinguish defects between background.Then,a dual mark point positioning method is proposed for system calibration and FPC image recognition and positioning.Camera calibration is realized with the help of calibration board,vision coordinate and motion coordinate calibration are achieved through the movement of workpiece,FPC patch target position calibration is realized by regulation of double mark points and target location.This method has advantages of high speed and high precision and can realize precise positioning of each patch on FPC in the case of inaccurate positioning FPC.Next,for different types of FPC,using mask algorithm to extract region of interest can achieve the contour extraction of different shapes of patch,makes the system has generality and can detect different types of FPC;For position features of different defects,the defects are divided into inside contour and outside and the parallel detection is performed,multi information color detection algorithm is used to identify defects and machine learning clustering algorithm is used to automaticly classify the defects.Based on the above,this paper developed an automatic visual defect detection and classification system of FPC reinforcement sheet,the conventional detection speed can reach 60-80 chip /min and the detection accuracy is 0.02 mm.Also,this paper tests different defects in different types of FPC board,the results prove that the system has greatly improved in reliability, accuracy and efficiency compared with traditional detection methods.This detection effect is stable and has good robustness,it has important application value and provides a good solution for FPC patch defect detection.