Self-adapting SAR image classification method based on local standard deviation

Abstract

A self-adaptive SAR image categorization method based on local standard deviation is provide, which comprises: (1) according to an original data, calculating a local standard deviation s (i, j) for the neighborhood of certain point (i, j); (2) carrying out self-adaptive selection of a GLCM calculation window by the local standard deviation. The specific procedures are: setting the threshold value of s as sth; judging whether the certain point s (i, j) is larger than the threshold value sth; if the certain point s (i, j) is larger than the threshold value sth, selecting a large size window for the calculation of GLCM; if the certain point s (i, j) is smaller than the threshold value sth, selecting a small size window for the calculation of GLCM; (3) setting other parameters of GLCM and calculating the GLCM of the certain point according to the size of window; (4) extracting a characteristic quantity from GLCM to compose an eigenvector and categorize the characteristic vector by adopting a c mean value clustering algorithm. Therefore, the invention can enhance the categorization accuracy of larger military objects.

Description

technical field [0001] The invention belongs to the field of image processing, and more specifically relates to an adaptive SAR (Synthetic Aperture Radar, abbreviated as SAR) image classification method based on local standard deviation. Background technique [0002] SAR is a high-resolution microwave imaging radar. It has the ability to work all day and all day, and can penetrate natural vegetation, artificial camouflage and surface soil at a certain depth, so it has attracted much attention. With the increasing number of SAR image acquisition systems, the rapid development of SAR image application technology is urgently needed. SAR image classification is a key technology for further understanding and interpretation of SAR images, and it has important application value and practical significance in both military and civilian applications. The task of classification is to select the characteristic parameters by analyzing the characteristics of various targets, divide the f...

AI Technical Summary

Problems solved by technology

Among them, the selection of parameters other than w is relatively easy, and has little impact on classification, but the difference in classification effect caused by different selection of w is the largest.

Although a large-scale window can eliminate the influence of the background and other objects on the target classification, it seriously damages the characteristics of the target. The small size is the opposite. While retaining the details of the target, the interference of other objects is relatively large.

The traditional method of selecting a fixed window cannot solve this contradiction

Images