Research on an Algorithm for Binocular Stereo Vision Image Matching Based on SIFT Operator
|School||Xi'an University of Electronic Science and Technology|
|Course||Computer System Architecture|
|Keywords||Binocular stereo vision Image matching SIFT Feature point|
Binocular stereo vision is a technique on how to possible understand and perceive the objective world by computer rather than human beings. It can require the three-dimensional (3D) geometry information of objects form two images that shooting from two different angles. This technique has a rapidly development these years, it has been successfully used in many fields of civil and martial.Image stereo matching is the most significant part in binocular stereo vision technique. A perfect matching algorithm is fast and accurately, it should have good practicability. But image matching algorithm strongly depends on image itself, that is, every stereo image stereo matching algorithm is proposed aiming at matching certain type of images, there is no such a stereo matching algorithm can process any type of images nowadays, so it is difficult to solve this problem completely. Sometimes, the matching algorithm even can not reach the requirement of applications, so this issue has always been developed and improved.This paper introduces and analyzes theory of binocular stereo vision technique detailedly, then makes a deep research on image matching algorithm. The focus on this paper is an image feature matching algorithm based on scale invariant features transform (SIFT) operator. The principle of SIFT matching algorithm is researched. In this algorithm, SIFT operator is used to detect the feature points, and each feature point is assigned a feature descriptor. It is accurate and robust towards image distortion and noise. In order to improve the matching rate of SIFT algorithm, an improved SIFT matching algorithm is proposed and implemented. Taking many different types of images as experiment images, with the improved algorithm, the shape character and relative position of objects in the image can be shown well, the method is stable and fast, the matching rate is improved.