Research On 3D Reconstruction And Defect Recognition Of SMT Components Based On Machine Vision

PCB is the core component of electronic products,which is widely used in various industries in modern society,and the market demand is large.Defect detection of PCB component is the necessary link in PCB production.The trend of miniaturization and high integration of electronic components and the development of SMT make the density of components on PCB larger.Traditional manual inspection method can not meet the requirements of industrial detection precision and speed.AOI is a new detection technology based on digital image processing,which has the advantages of non-contact,high precision and fast.It is gradually applied in PCB defect detection.However,the general 2D visual detection method is difficult to identify the defects related to the height of components,and it is easy to produce leakage detection.3D detection can effectively identify such height defects,and play a positive role in ensuring product quality and reducing scrap rate,so it is of great significance and value to develop a height defect recognition system for SMT components based on machine vision.In this paper,a visual recognition system is designed for the height defect of PCB components after SMT process.It is intended to solve the problems of precise stitching of PCB images,accurate 3D reconstruction of SMT components and height defect recognition of SMT components.The main research contents are as follows:(1)SMT component height defect visual recognition system scheme design.This paper analyzes the requirements of SMT component height defect detection,completes the hardware design of image acquisition module and grating projection module,and builds the system.According to the defect recognition process and functional requirements,the main function modules of the system software and the overall image processing algorithm are designed.Finally,a PCB image location and pose correction scheme based on mark points is designed.(2)PCB image precise stitching method based on improved SURF registration.According to the characteristics of PCB image rich in texture and complex features,a PCB image registration method based on overlap area and compound condition constraints is proposed.In order to improve the speed and robustness of PCB image stitching,a feature extraction method based on overlap regions and a feature matching method based on image block matching are proposed.Aiming at the problem that the similar features of PCB images are prone to feature mismatching,a rough screening method of feature points based on Hamming distance and a fine screening method of feature points based on multi-condition constraints are proposed to effectively remove mismatched points and improve the stitching accuracy of PCB images.(3)Three dimensional reconstruction method of SMT components based on multi grating data fusion.Aiming at the problem of occlusion and shadow of PCB components,a 360 ° multi grating shadowless 3D projection system is designed.Aiming at the height error of 3D model caused by the interference of PCB surface reflection,a height error filtering method based on limited filtering is proposed.In order to further eliminate the shadow error of components and improve the accuracy of 3D reconstruction,a multi grating height data fusion method based on Gaussian weighted average is proposed.(4)SMT component height defect identification method.Aiming at the problem of inaccurate positioning of components with small size and simple texture on PCB,an accurate positioning method of components based on multi feature is proposed.A threshold segmentation method based on gray statistics is proposed to segment PCB component body and pin region for height extraction.Aiming at the height measurement error caused by the bending and tilt of large-size PCB,the surface reconstruction method of PCB based on HSV color and the height error correction method of components are proposed.Finally,different height data processing and judgment methods are designed for four types of height defects.To sum up,according to the technical requirements of SMT component defect recognition,this paper designs a set of SMT component 3D reconstruction and defect recognition system based on machine vision.The system uses high-performance Advantech industrial control computer as the core processor,high-speed Coa XPress industrial camera,high-performance Matrox image acquisition card,industrial lens,RGB three-color light and high-precision UPOLabs grating projector as the image acquisition and grating projection unit,and Siemens PLC as the motion control core to build the hardware platform.Then,the PCB image precise stitching method based on improved SURF registration,SMT component 3D reconstruction method based on multi grating data fusion and SMT component height defect recognition method are proposed.Finally,the system software is designed based on Windows development platform,Visual Studio 2015 and QT software platform,including offline parameter setting function,online detection function and data management function.Through the batch test of PCB workpiece,the accuracy rate of defect recognition of the system designed in this paper is more than 99%,which basically meets the production demand,and the system is stable and reliable.