Description and Matching Algorithm for Line Matching

In order to resolve the problem of image lines matching,a line matching algorithm based on Haar wavelet response is proposed.Firstly,lines are extracted by Canny algorithm.The statistical linear Haar-wavelet response vector is adopted as the main direction of the straight line,so as to avoid the rotation invariant.Secondly,by constructing six Gaussian scale factors,the neighborhood of straight line is divided into non-overlapping sub-region in the corresponding scale space. Then calculating response characteristics of Haar-wavelet, and using the normalized characteristics of the statistical weighted mean and variance as a linear descriptor.Finally,constructing the cascade feature descriptors is used to achieve wide baseline image matching.Experimental results show that the proposed method is robust to changes in occlusion,rotation,affine and scale.Meanwhile,the method can achieve stable line matching between the IR and visible images.