Feature Detection Techniques Based On Computer Vision Research

Feature detection problem has always been in the field of computer vision research is an important research direction, while the characteristics of computer vision-based detection technology in the feature match there is a problem. Therefore studying a high accuracy, robustness, performance, stability and applicability of a strong feature matching algorithm in computer vision tasks are still difficult.Thesis research work includes:(1) This article studies the basic theory of computer vision-related research and to achieve the objectives of the measured object feature extraction and matching, and proved to achieve good results.(2) Feature point extraction, the traditional SUSAN algorithm, Harris algorithm in-depth study , through experimental and comparative analysis of the experiment, comparing the classic Harris feature point detection algorithm based on the. In this paper, image enhancement technologies to the use of an improved multi-scale Harris corner detection algorithm. This method has smaller error, false corner less error rate low, improve the accuracy of the more obvious advantages.(3) Camera calibration, in order to determine the specific location of the camera, property and the establishment of the camera imaging model,parameters to determine the spatial coordinates of image points and space points of correspondence between the camera calibration carried out, This in-depth study of traditional and self-calibration method based on the calibration method using two-step calibration, calibration equipment to avoid the traditional high requirements, easy to operate and other shortcomings, but also a more accurate self-calibration method, parameters has more robust.(4) In the aspect of feature matching, comparing the feature matching algorithms, the paper comprehensively analysizes the strong and weak points of traditional matching algorithms; Template matching method adopted to reduce the complexity of the image calculated to improve the accuracy of template matching, experimental verification of the feature matching algorithm in feature matching rate, matching accuracy and robustness in terms of matching obtained good results.