Time sequence SAR (Synthetic Aperture Radar) image classification method on the basis of distribution difference and incremental learning

Abstract

The present invention relates to a time sequence SAR (Synthetic Aperture Radar) image classification method on the basis of distribution difference and incremental learning. The time sequence SAR image classification method comprises the selection of incremental samples and the incremental learning of the incremental samples. The time sequence SAR image classification method comprises the steps of: firstly, carrying out change detection on time sequence SAR images by a method based on the distributional difference, utilizing an Egdeworth approach principle to estimate the statistical distribution of the SAS images of a source domain and a target domain, calculating a cross entropy difference index of image distribution, obtaining a change detection image according to a detection threshold acquired by a CFAR (Constant False Alarm Rate) method, and using unchanged regions as incremental samples; and then initializing an SVM (Space Vector Modulation) classifier by utilizing a known training set of the images of the source domain, and completing the incremental learning of an integral incremental sample set and the parameter update of an SVM classification model by the iterative process. The distributional difference change detection method is combined with the incremental learning of a support vector machine and domain self-adaptation from the source domain to the target domain can be completed, so that high-precision classification of the images of the target domain is realized.

Description

technical field [0001] The invention belongs to the technical field of image processing, in particular to a time-series SAR image classification method based on distribution difference and incremental learning. Background technique [0002] In recent years, with the emergence of a large number of time-series image resources, a series of new theoretical methods have been used for the interpretation of time-series images. It has been widely used, and it is an important and effective means for people to study and track the evolution of natural history and monitor the dynamic changes of the environment and resources. In the classification of time series images, from different starting points, there are mainly two mainstream directions: [0003] 1) Use an image of a known training set in the time series image to classify another image of an unknown training set, such as a classification method based on domain adaptation. This method uses the time-dependent correlation between t...

AI Technical Summary

Problems solved by technology

[0006] The purpose of the present invention is to propose a new time-series SAR image classification method based on distribu

Images