Research On Algorithms Of Multi-sensor Images Edge Extraction And Matching

Multi-sensor image matching algorithm is an knotty research direction in the field of computer vision,it is widely used in exact control and guide?video scout and target identify.It is of great theoretics meaning and pratice importance.This thesis aims at infrared image and visible image,studies the multi-sensor images edge extraction arithmetics and the robust matching algorithm.To gather up,the main contributions of this thesis are as follow:(1)We analyse and conclude the key technologies of image edge extraction and matching algorithm.First,we analyse the otherness and commonness of the multi-sensor images,and then detect the edges by hackneyed edge extraction algorithm and sum up the advantages and disadvantages of them.Second,we use ESD distance and the Hausdorff distance to calculate the similarity of the edge images,in addition,we estimate the effect of the image matching algorithm by the correlative performance index(MAX?PSR?PCE and RP).(2)A multi-sensor image matching algorithm based on wavelet transform edge detection algorithm is proposed.Wavelet transform edge detection algorithm has some shortages,we propose the reformative algorithm.First,we choose the dual thresholds and the Half-Neighborhood Algorithm to improve the fitness ability of the algorithm.And then introduce the Facet model to improve the positioning accuracy of edge detection,and wipe off the useless information.Then we use ESD distance and the Hausdorff distance to calculate the similarity of the edge images.The results show that our method improve the extraction effect on a certain extent.It wipes off many trashy edges and it is more robust and accurate.(3)A multi-sensor image matching algorithm based on the reformative CV model is proposed.The conventional CV model's capability will decline if the surroundings is complexity.So we combine local information with full-scale information to play down the influence of gray asymmetry,and introduce restriction and neumann boundary condition to make the edges detection algorithm more robust.The results show that the algorithm we proposed can accomplish matching under the complexity surroundings and it is more robust,fast and adaptive than the conventional CV model.