The Research Of Scence Matching Algorithm Based On Visible Light Image And Infrared Image

The scene matching technology is based on various advanced technologiessuch as the sensor technology, image processing technology and spacetechnology, and it is a new technology which more used for aircraft navigation.This new technology has an extremely important application value in aircraftnavigation guidance process, that’s because cruise guidance must have suchcharacteristics as innovative, strong anti-jamming, high precision and so on, butthese parameter requirements just rely on the inertial navigation and GPS notenough. Scene matching technology as a kind of image matching technologyapplies in cruise terminal guidance, which can improve the guidance precisionand make autonomy strengthen.Scene matching is that using corresponding relationship between thereal-time images intake geomorphic feature and the reference map which hasbeen pretreated and had actual location datum figure to determine the guidanceinformation. In recent years, in order to improve the precision and speed of thecruise terminal guidance, scene matching technology has been widely used, sofast and real-time matching technology is worthy to study.Firstly, based on the summarization principle of scene matching, aimed atthe characteristics of the infrared and visible image, this paper does feasibilityanalysis for the matching algorithm based on the characteristics of differentsource image matching problem, studies many kinds of feature extractionmethods，and analyzes the advantages of various methods.Then this paper focuses on four key elements of the scene matchingalgorithm: the feature space (extracted a variety of characteristics), the similaritymeasure (improved Hausdorff distance), the search space (3-4DT and circularprojection transformation) and the search algorithm (search algorithm by leapsand bounds).Next, this paper reasonably combines the above key elements, and simulatesthe matching algorithm. In the process of simulation results analysis, for edgefeatures direct matching results is not ideal, this paper gets a good matching effect after the edge feature refined. And because refining process and distancetransform consumed a lot of time, this paper proposes a method that theapplication of Hough transformation directly acts on the edge binary image, getskeleton features. Using the skeleton features to matching not only can increasethe matching speed, but also can improve the precision of matching.Finally, this paper presents the predicted results that the skeletoncharacteristics of good matching performance apply to scene matching tosimulate vehicle route.