Target-oriented Characteristics Analysis And Extraction Methods

As an important transport hub, the bridges mostly locate at the intersection of trunk roads and rivers. Due to its special location, the automatic identification and precise positioning of bridges is of great significance in military and in civilian.Based on the high-order correlation coefficient, this paper extracted the feature which has the minimum coefficient, and selected the features according to the correlation coefficient. The characteristics of images has characteristics of small sample, high dimension, great noisy, high redundancy and non-linear, the linear correlation analysis can partly show the image feature data rules, and can show only the simple linear structure of the data, it can’t show the nonlinear nature and complexity. This feature selection method based on high-order correlation coefficient makes up the defects that linear correlation can’t accurately determine the nonlinear structure.This method makes full use of the complementary characteristics of bridges in remote sensing images, mainly recognizes bridges based on the results of panchromatic images, and supplemented by SAR images through fusion on the decision level. The innovation of this paper is a recognition method based on complementary features of multi-source remote sensing images. The targets in multi-source remote sensing images of the same sensor consist of a plurality of pixels, showing the unique color, texture, structural, geometry and edge profile characteristics, the bridge identification method based on the single image characteristics has slower running speed and higher error rate. The experiment results show that the method based on complementary features for multi-source remote sensing images has effectively improved the accuracy of bridge recognition and has less synthesis limits, is adapt to bridge identification of more remote sensing images.