The Key Technology Research On Autofocus And Reconstruction In Machine Vision

The key technology of machine vision has been researched more widely with it’sincreasingly applications. The image-based three-dimensional reconstruction methodhas become a hot research topic because of the low-cost, strong third dimension andhigh automation. So it has a broad application prospect. The3D (three dimensional)reconstruction technique based on image is an important content and the final researchgoal of machine vision. The3D reconstruction method restores geometric structure3Dspace visible surface from2D (two dimensional) image. The3D reconstruction methodis actually the inverse problem of computer graphics and the key technologies areautofocus, the camera calibration, image preprocessing and feature extraction, imagefeature matching,3D information recovery and depth interpolation. This paper studiedimage-based3D reconstruction techniques in-depth and focused on several key issues inthe reconstruction process. The main contents are as follows:Firstly, the paper researched the key technology (auto focusing technology) ofimage acquisition and analyzed the factors that affect high-quality images acquisition.The existing focus evaluation methods exists common faults such as low adaptability ofevaluation function and simple as well as blind and poor generality of evaluationwindow selection. Aiming at these, focus evaluation methods based on wavelettransform and adaptive genetic algorithm (AGA) is proposed in this paper. First, db7wavelet transform is done on criteria images. Then the optimal evaluation region issearched in two-dimensional wavelet coefficients automatically by AGA. Finally, thefocus criteria function based on weighteM2dWToperator is used to evaluate.Then the calibration techniques are studied. The camera calibration is the actualprocess to determine the camera’s internal and external parameters. The internalparameters are the camera’s internal geometry and optical characteristics. The externalparameters are the three-dimensional position and direction of camera coordinatesystem relative to the world coordinate system. The basic calibration theoreticalknowledge and common calibration model are also analyzed in this paper. This paperfully considered the lens radial distortion and tangential distortion in the camera model.The Zhang Zhengyou calibration method was used to realize the pinhole cameracalibration on the basis of OpenCV (Open Computer Vision) and VC++and thecalibration results are output. Finally, the paper introduces the common methods of vision-based3Dreconstruction and further discuses the key technology of3D reconstruction techniquebased on binocular vision: feature extraction, feature matching and depth informationcalculation. The three-dimensional reconstruction system is developed on the OpenCVand VC++platform. The depth information of the object or scene can be obtained bythe binocular stereo match of image pairs and its three-dimensional information isreconstructed.