Research On Target Detection And Recognition Of And Bridge In Image

As two kinds of large target in natural scene, the automatic detection and recognition of airport and bridge has very important application value in civil field as well as military field. Although there is a growing interest that many experts and scholars research it, there has been no effective and adaptive ones in the literature to our best acknowledge yet.This paper that adopts model-driving strategy researched and implemented respectively a detection method of water-bridge and a detection method of airport based on SVM combining with prior theories.The main idea of the two arithmetics is similar. Firstly, the image is segmented, in order to find the potential object, the segmentation result still need be proceeded with a series of processing. Secondly, the total15features extracted from potential object and the environment around potential object will be consisted of a eigenvector. Finally, the eigenvector is judged by SVM, and then the true object will be marked in the source image.The special feature that is extracted depends on the different potential object. The features of a potential bridge object include the complicated degree between the two near river area, the sign which presents whether the four vertexs is repeated, and Zernike moments. Except the Zernike moments feature, a potential airport object still extratc the width and the length of the potential runway, the numbers of the road linking the two runway and the texture features of the potential runway area.In the paper, the special process of the two algorithms was clearly described, and some images taked from the true scene were used to test the algorithms. The experiments show the two arithmetics have the higher accuracy rate.