Unsupervised hyperspectral image change detection method based on adversarial learning

Abstract

The invention discloses an unsupervised hyperspectral image change detection method based on adversarial learning, and mainly solves the problems of high false detection rate and low detection precision caused by insufficient training data of the existing supervised technology. According to the implementation scheme, the method comprises the steps of 1) constructing an unsupervised spectrum mapping network based on adversarial learning, and forming a dual-time-phase spectrum mapping network, 2) inputting dual-time-phase hyperspectral images, and training the dual-time-phase spectrum mapping network; 3) extracting a spectral dimension feature map of the trained double-time-phase hyperspectral image, and performing principal component analysis on the spectral dimension feature map to obtain a double-time-phase one-dimensional spectral dimension principal feature map; 4) sequentially performing spatial feature enhancement and binaryzation on the dual-time-phase main feature graph to obtain a dual-time-phase binary graph; and 5) obtaining an unsupervised hyperspectral image change detection result by calculating the residual error of the double-time-phase binary image. The invention reduces the false detection rate, improves the detection precision, and can be used for land investigation, city research, and disaster detection and evaluation.

Description

technical field [0001] The invention belongs to the technical field of image processing, and further relates to an unsupervised hyperspectral image change detection method in the technical field of image change detection, which can be used for land survey, urban research, disaster detection and evaluation. Background technique [0002] Hyperspectral change detection is the process of identifying surface change areas by analyzing a set of hyperspectral images collected at different times in the same geographical area. The spatial information in the image can reflect information such as the outer contour of the change area, while the spectral information can reflect Hyperspectral images can use these two pieces of information to more accurately detect changes in the outside or inside of an object. The spectral information of hyperspectral images includes multiple bands from visible light to thermal infrared. Due to different substances, the spectral curves reflected in each ba...

AI Technical Summary

Problems solved by technology

Although the concept of superpixels is proposed in this method, and the spectral, texture and spatial information of hyperspectral images are utilized through simple linear iterative clustering, however, since this method only imposes divergence constraints in network learning, it does not consider as a high The particularity of spectral image change detection ignores the constraints on the spectral dimension, so the trained network model has a weak ability to distinguish between changed and unchanged test samples, resulting in low accuracy of change detection

Images