Research On Image Matching Algorithm In Binocular Stereo Vision

Binocular stereo vision is a method of obtaining three-dimensional geometric objects information by the multiple images based on the disparity principle. The method has been widely used in many fields and has been a research hotspot in recent years.This paper researches on stereo matching algorithm, the most important and difficult issue in stereo vision, analyzes four key components of stereo matching algorithm, and proposes a new stereo matching algorithm which combines advantages of the feature-based stereo matching algorithm and the area-based stereo matching algorithm for requirement of dense matching and strong robustness against light transformation, gray transformation and rotation transformation.The algorithm first uses Harris corner detection algorithm to extract the feature points, and then constructs SIFT local feature descriptor based on 3D histogram of gradient location and orientation by exploiting image gradients information, and then reduces the search space of matching features from 2D to 1D according to epipolar constraint, the initial matching result is finally obtained in the light of the nearest neighbor matching based on Euclidean distance of SIFT descriptors. And then based on the initial matching result, the algorithm uses area-based stereo matching algorithm to match the unmatched region, and improve the speed and accuracy by taking advantages of epipolar constraint and disparity range constraint.The experimental results prove that this algorithm wins advatanges of the feature-based stereo matching algorithm and area-based stereo matching algorithm, can get more dense disparity map; and have good robustness against affine transformation, 3D viewpoint transformation and non-linear image intensity transformation.