Research Of Corner Detection Algorithm Based On Image Edge

Corners are the points of the image gray value changing dramatically, including the points of contour curvature maximum changes, the intersection points of multiple straight line, the isolated points on monotone background etc. Owing number of its high amount of information and the less corners, corner detection have an important significance in image matching, target tracking and recognition and 3D reconstruction and motion estimation etc. Corner detection algorithms are generally divided into two parts:the corner algorithms based on edge and gray.The corner detection algorithms based on edge are mainly researched, the concrete research contents and results are as follows:Gabor transform is windowed Fourier transform, which is a special case when the window function of the short time Fourier transform is the Gauss function. Gabor transform is convolution of the two-dimensional image to extract features components of the image. The Gabor function can extract the related features in different directions and scales. The frequency and direction of Gabor filter is similar to human visual system, so it can be applied to feature extraction, fingerprint recognition, iris recognition, face recognition, texture analysis etc. A multi-channel odd Gabor gradient correlation matrices is proposed based on the analysis of odd Gabor filter excellent edge detection performance. This algorithm have a good improvement in corner detection and localization accuracy than the Harris algorithm, the curvature scale space(CSS) algorithm.Invariant moments theory and the image corners detection algorithm are important research topics in computer vision and pattern recognition. The commonly used moment invariants theory based on region is two-dimensional invariant moment theory based on region, which was first proposed by Hu in 1962. This theory can be used to extract the image morphological characteristics, such as size, rotation. There are quit a few relevant algorithms. Because Hu invariant moments have simple concept, the stable recognition rate and scaling, translation and rotation invariance, it is widely used in the feature extraction of image. A contour corner detection algorithm based on multi-scale local moment invariants based on the multi-scale product and sum algorithm and the local invariant moments algorithm. The method can enhance the corners feature information, suppressing noise and achieving good results.