Multilayer restricted Boltzmann machine-based SAR (synthetic aperture radar) image positive and negative type variation detection method

Abstract

The present invention discloses a multilayer restricted Boltzmann machine-based SAR (synthetic aperture radar) image positive and negative type variation detection method. The method includes the following steps that: step 101, the multilayer restricted Boltzmann machine-based SAR (synthetic aperture radar) image positive and negative type variation detection method is started; step 102, three types of difference images are constructed for two registered SAR images at different time phases in the same area; step 103, fuzzy C-means clustering is performed on the difference images, so that rough three-type variation detection results are obtained; step 104, non-noise points with high possibility are selected as training samples of an improved multilayer restricted Boltzmann machine according to the variation detection results, so that the improved multilayer restricted Boltzmann machine can be trained; and step 105, and an image to be detected is inputted into a trained network, so that a final variation detection image is obtained. According to the method of the invention, variation types are divided into three types, namely, a positive variation type, a negative variation type and a non-variation type, and therefore, the accuracy of variation detection can be improved.

Description

technical field [0001] The invention belongs to the field of computer technology and mainly solves the problem of remote sensing image change detection. The invention obtains three types of difference images from two remote sensing images of different time phases, and then uses a multi-layer restricted Boltzmann machine to classify the three types of difference images. Complete the change detection of remote sensing images. The invention can be applied to the detection of changes in remote sensing images of disaster areas during natural disaster detection and rescue, urban development planning, geological research and other fields, and completes the detection of changes in remote sensing images in specific areas. Background technique [0002] Synthetic Aperture Radar (SAR) has the characteristics of high resolution, all-weather work, effective identification of camouflage and penetration of cover, and has been widely used in military, scientific research and industrial and a...

AI Technical Summary

Problems solved by technology

[0004] The classic process of dealing with change detection problems: (1) preprocessing; (2) generating difference map; (3) analyzing difference map, there are four commonly used analysis methods of difference map, threshold analysis, graph cut analysis and level set analysis , the traditional analysis method has the following disadvantages (1) image classification by optimizing the objective function will often fall into a local optimal solution; (2) iteration based on a fixed and complex formula limits the application of the algorithm; (3) the algorithm does not learn ability, the ability to classify is limited; in machine learning, the change detection problem is a problem of unbalanced classification, which can evolve into an incremental learning problem, and the neural network is the best solution to this kind of problem

Images