Research On Multi-modality Image Matching Technology Based On Salient Contour

Multi-sensor image matching has a broad application prospect in such fields as visual navigation,precision guided and topographic survey. Multi-sensor image matching algorithm based on salientcontour is researched on the basis of human’s cognition process of the images.First, according to researches on the human’s cognition process and forming mechanism ofsalient edges, the definition of salient edges was presented and the validity of this definition wasverified by experiments.Second, in view of the issue that it is difficult to extract salient edges of real multi-sensor imageswith noises fast and accurately by using common algorithms, according to the forming mechanism ofsalient edges, a novel salient edges detection algorithm and its rapid calculation were proposed basedon possibility fuzzy C-means (PFCM) kernel clustering. The two-dimensional feature space wereconstructed by each pixel’s gray and texture features; The data set was compressed based on the ideaof data reduction; The robustness were improved further by mapping sample space of reduced datainto Gaussian kernel space; Image segmentation was carried out by using fast kernel possibility fuzzyC-means (FKPFCM) algorithm and then salient edges extraction. Experimental results show that thismethod extracts salient edges for real multi-sensor images with noises more accurately than algorithmbased on force field and the FCM algorithm; and the proposed method is on average about56timesfaster than PFCM algorithm in real time and has better robustness.Last, in view of the issue that it is difficult to solve the problems of partial contour matching anddeformation, according to the conclusions that partial contour structure has relative stability and agood description needs to merge the contour’s local and global information, a contour descriptionmethod based on local scale was proposed; According to requirement of the linearity of optimizedpath, a contour matching algorithm was proposed based on improved DTW by scratchable latex pathconstraint conditions. Experimental results show that the proposed method can realize the whole andpartial matching for multi-sensor images, which have geometric transformation (rotation, scaled) anda certain deformation. Its matching precision is92%on average, and improves30%compared withHD method, improves26%compared with traditional DTW.