Research On Surface Defect Detection Of CMOS Image Sensor Based On Machine Vision

CMOS image sensor is a key device used to acquire image data on smart phones,and its quality will directly affect the user experience of smart phones.The traditional appearance quality detection of CMOS image sensor mainly relies on manual visual inspection,which has problems such as low inspection efficiency,high false positive rate,and unstable inspection standards.Using machine vision inspection instead of manual visual inspection can effectively address the above problems.At present,there are rare researches on CMOS image sensor surface defect detection in the field of machine vision quality inspection,so as to fill the blank of research in this field this paper carries out research on CMOS Image Sensor defect detection based on machine vision.According to the inspection standard of CMOS image sensor surface defect,a design scheme of optical imaging system is put forward.This paper makes in-depth analysis on CMOS image sensor appearance characteristic,according to the characterisitics the industrial camera,industrial lens and light source can be determined.The experiments show that the location is accurate and robustness.The experiments verify the feasibility of optical imaging system.An algorithm of CMOS image sensor image location and geometric correction is proposed.Applying the template matching algorithm base on Open CV to locate the CMOS image sensor roughly.Considering the limitation of template matching algorithm,CMOS image sensor feature extraction algorithm is propose.Finally,research on Geometric Correction based on the perspective transformation.The experiments show that the location is accurate and robustness.Aiming at the characteristic of CMOS image sensor defect,two kinds of defect detection Algorithm are proposed.Golden ball deviation algorithm is designed on the basis of extracting the feature of golden golden ball accurately,and then analyse the position deviation on PCB pad,Using the mathematical morphology and edge detection algorithm to enhance the feature of golden ball.Golden wire collapse algorithm is designed based on the geometric model of normal golden wire,and then analyse the posture of golden wire feature in geometric model,to recognize the golden wire collapse defect.By designing offline test software,collecting batch sample image data to verify the detection effect,stability and timeliness.Experiments show that the CMOS image sensor surface defect detection method can fulfill the designed objective of inspection system,and it is of great significance to improve the shipping quality of CMOS image sensor.