Research On Multi-sensor Image Matching Algorithm For Scene Matching Aided Navigation System

With the continuous development of aerospace technology,modern warfare has entered the stage of high-tech information,precision strike and guidance weapons are playing an irreplaceable role.Because the scene matching technology has many advantages,such as anti-jamming,autonomy,accuracy and so on,it has been widely concerned.The scene matching as an auxiliary navigation tool can be used to improve the guidance accuracy and enhance the autonomy of the guidance system.Scene matching is to obtain the position of the real-time image in the reference image by matching the pre prepared reference image with the real time image.Because the location of each coordinate on the reference image is known,the correct matching result means that the real-time position of the vehicle is obtained.This position is then converted into navigation information to improve navigation accuracy.This paper focuses on image matching algorithm in scene matching.First of all,this paper analyzes the present situation of the research on the scene matching,and selects the scene matching as the breakthrough point.In this paper,the advantages and disadvantages of two kinds of commonly used algorithms in scene matching are analyzed.Finally,the line segments and the moment invariants are selected as the features of this paper.After analyzing the advantages and disadvantages of each line segment extraction algorithm,the LSD algorithm is selected as the line segment extraction method in this paper.However,the LSD algorithm can segment the whole lines in the image,and a large number of line segments and collinear segments in the image will seriously affect the speed of the matching algorithm.In order to solve the above problems,this paper proposes a multi-step line segment optimization strategy,which can reduce a lot of useless line segments,and retain the common segments between the two images.The matching speed is greatly improved,and the algorithm can overcome the problem of rotation and scale change.In addition,taking into account the use of the algorithm requires the existence of straight lines in the scene,greatly limiting the applicability of the algorithm.Therefore,this paper presents a new algorithm with wider applicability,which combines the image moment invariants and the parameterized template vectors.The algorithm combines the advantages of the former to adapt to the variation of gray level between multi-sensor images and the latter on the advantages of rotation,scale transformation adaptation.At the same time,we improve the original parametric template vector matching algorithm to enhance the robustness of the algorithm.At the same time,this paper uses the hierarchical search strategy of Pyramid,which reduces the time complexity of the algorithm and improves the matching precision of the algorithm.The experimental results show that the two improved algorithms proposed in this paper have higher matching accuracy and matching probability,but the time complexity of the algorithm needs to be improved.