Camera Calibration And Position And Pose Detecting On Vision Measurement System Of PCB

With the quick development of the semiconductor manufacture and the popularization of the machine vision in China, it has become a necessary method to test products with using machine visions in product line. The purpose of this paper is to provide machine vision support for PCB automatical alignment system. We construct a machine vision system from three aspects: feature point detection, camera calibration, position and pose detecting system base on plane vision.1 The purpose of feature point detection is to provide image control points for camera calibration and relative position and orientation detection system. We use corner detection for the camera calibration and calculate centroid of PCB through-hole for the position and pose detecting system. We found the defect of SUSAN corner detection algorithm, that is, the stability became lower when the algorithm detecting symmetry complex corner. This paper analysis the SUSAN detection principle and propose"symmetry-point univalue segment assimilating nuleus"principle. The experiment had proved that the new principle can detect various corners like USAN principle, and the new principle is better than the USAN principle when detecting symmetry complex corner.2 Based on the analysis of various types of camera calibration algorithms and system application environment, took into account the system accuracy and cost requirements, this paper puts out a linearity imaging model including radial distortion, and use Zhang ZY camera calibration algorithm. Finally we verify the calibration system accuracy with reverse projection and 3D reconstruction base on plane vision.3 The machine vision system of this paper is to measure the relative position and pose both the PCB and the film. To consider the application conditions and the objects features to be detected, this paper selects the specific centroid of through-holes of PCB image and film image as control points. Finally we calculate the plane motion that correspond graphic geometric transform through 3D reconstruction base on plane vision.4 Finally we verify effective and feasible of every module in this project by design experiments scientifically. And we verify the pricision of every module in this paper by large numbers of simulation experiments and actual image experiments.The paper constructs machine vision for the PCB automatical match system, at the same time improving the corner detection algorithm. Experiments proved that the corner detection algorithm and the sets of machine vision achieved good results.