Research On Image Corner Detection Method

As a feature point,corner points have a very important role in 3D image reconstruction,image matching,motion estimation,target recognition and tracking,and also have an important application in real-time processing system.As an important two-dimensional local feature in image,corner feature represents the local features and shape features of an image,and determines the target contour feature.Corner points has the advantages of less computation,no change with environmental conditions,rotation,translation,scaling invariance,etc.Therefore,corner detection algorithm has very important application and research value in machine vision.In this paper,the detection algorithm is classified according to the difference of the detection ideas,and the principle of the detection method b ased on the gray level and the edge characteristic is summarized.Firstly,this paper analyzes the advantages and disadvantages of CSS algorithm and the improvement direction of ACSS algorithm.Secondly,several detection algorithms based on the change of gray scale size are analyzed,and a fast detection method aimed at gray scale change is proposed to solve the shortcomings of these algorithms.The experimental results show that the improved algorithm is better in real-time performance,but its detection effect has not been improved,and the anti-noise performance is not ideal.In this paper,an improved method based on Harris operator and Minimum Intensity Change is proposed for the existing problems such as poor real-time performance,large amount of computation and weak anti-noise ability.This method improves the corner detection effect by combining the local weighted entropy of the gray image with the MIC operator.Firstly,the candidate corner set is obtained by using the local weighted entropy of the image.Then the height of the gray scale is calculated by using the CRF(Corner Response Function),and the candidate corner points are divided into three categories according to the CRF value.Finally,we use the adaptive template and the threshold of the MIC algorithm for corner detection to get the best match point.The experimental results show that the proposed algorithm can improve the real-time and accuracy of the original algorithm effectively.In the case of noise reduction,the algorithm removes most of the pseudo-corner points while increasing the number of extracted real corners.