A Study On Matching Algorithm Of Irregular Curve

As an important part of image processing and computer vision,curve matching technology is applied to image retrieval,object recognition and visual navigation.How to extract the required feature information from the image and realize the correct matching is the research content of this field.Due to the difference of curve length and shape in the curve extraction phase of feature descriptor matching,it is difficult to meet the demand of accurate matching.At present,the curvature curve matching algorithm can only solve the problem of matching a single curve in the image.When there are multiple curves in the image,the calculation amount is large and it is difficult to meet the affine invariance requirements.Therefore,this paper combines the feature descriptor with the curvature integral matching algorithm and presents the irregular curve matching algorithm.Feature descriptor matching can form the one-to-one mapping relation of multiple curves in the image,the curve fitting of edge feature is carried out by using cubic spline interpolation algorithm,and the feature descriptor is optimized by using curvature integral to achieve accurate matching.The main research contents of this paper include:Firstly,the feature descriptor is used to extract,describe and match the feature curve of the image,and then the edge feature curve is matched.According to the idea of SIFT(Scale-Invariant Feature Transform)feature point scale space,through the local optimization of the feature line extraction algorithm,a feature curve descriptor with rotation,scaling and illumination invariance is formed.According to the improved algorithm of K-D tree nearest neighbor search and Euclidean distance constraint principle,edge feature matching is carried out.Secondly,in order to obtain the curvature information of the curve,the curve fitting of the discrete points of the edge feature curve is carried out by the cubic spline interpolation algorithm.The interpolation curve is obtained by solving the coefficient of the curve equation and constraining the endpoint.Compared with other fitting methods,this algorithm can make the fitted curve pass through any discrete point,which ensures the accuracy of curve fitting and the smoothness of higher order function.Finally,the local matching of irregular curves is accomplished by using curvature integral matching and fast normalized cross correlation algorithm.Based on the results of cubic spline interpolation curve,the curvature integral of the interpolation curve function is solved according to the mathematical characteristics of the curve,and the curvature integral identifier is constructed.Local matching of curves is realized by fast normalized cross correlation algorithm.The experimental simulation results show that the irregular curve matching algorithm proposed in this paper solves the limitation of large search range of curvature matching,improves the matching accuracy and meets the requirement of scale invariance.