Polarization SAR image change detection method based on deep stack network

Abstract

The invention discloses a polarization SAR image change detection method based on a deep stack network, comprising: 1) obtaining the polarization scattering matrix S of two sample images after registration 1 and the polarization scattering matrix S of the two images to be detected after registration 2 ; 2) Generate polarization covariance matrix C; 3) Preprocess polarization covariance matrix C; 4) Obtain feature matrix F1; 5) Change detection standard map of two sample images after artificial marker registration; 6) Obtain the feature matrix F21 based on the image block; 7) Obtain the feature matrix F22 based on the image block corresponding to the two images to be detected after registration; 8) Construct the feature matrix W1 of the training data set D and the feature matrix W2 of the test data set T ; 9) Construct a polarization SAR image change detection model based on a deep stack network; 10) Train the model; 11) Perform change detection on the feature matrix F22 according to the trained model, this method can realize the change detection of the polarization SAR image , the operation is relatively simple.

Description

technical field [0001] The invention belongs to the technical field of image processing, and relates to a polarization SAR image change detection method based on a deep stack network. Background technique [0002] Synthetic aperture radar (SAR) can detect ground or space targets in real-time and all-weather, and can obtain two-dimensional high-resolution images of ground targets in real time. It is widely used in military and civilian fields. In recent years, the data acquisition methods of polarimetric synthetic aperture radar (PolSAR) have gradually increased. Completely and accurately revealing the scattering mechanism of the target can enhance the ability of clutter suppression and anti-jamming. It provides powerful data support for target decomposition, classification and detection, and also provides a new research direction for change detection technology. [0003] Image change detection is an image processing technology that compares and analyzes two SAR images coll...

AI Technical Summary

Problems solved by technology

[0006] These change detection methods manually extract features, directly use certain characteristics of the image for detection, and do not use deep learning methods, so the method of feature extraction is very cumbersome

Images