Remote sensing image change detection method based on dense connection and geometric structure constraint

Abstract

The invention provides a remote sensing image change detection method based on dense connection and geometric structure constraint. The method comprises the steps: carrying out the preprocessing of afront-and-back-stage change detection input image, and comprises the calculation of an image normalization parameter, the normalization of a front-and-back-stage image, and the color correction of a back-stage image; dCGC-CD model training and testing are carried out, and change area information extraction is carried out on the remote sensing images in the early and later periods; wherein the DCGC-CD model comprises a front-and-back-stage change feature differential encoding module and a multi-branch geometric structure constraint decoding module, and the front-and-back-stage image change feature differential encoding module comprises a branch for an input front-stage image and a branch for an input back-stage image; the multi-branch geometric structure constraint decoding module comprisesthree branches; wherein the first branch is used for early-stage edge prediction, the second branch is used for change area prediction, the third branch is used for later-stage edge prediction, and the structures of the first branch and the third branch and the structure of the second branch interact through a loss function to effectively constrain a final change detection result through geometric structure information.

Description

technical field [0001] The invention relates to the fields of computer vision and remote sensing, in particular to a convolutional neural network (CNN) remote sensing image change detection and elimination method based on dense connections and structural constraints. Background technique [0002] In recent years, with the large-scale application of technologies such as cloud computing, big data, and deep learning, the intelligent change detection technology of remote sensing images has made great progress. Among them, the high-resolution remote sensing image change detection technology based on convolutional neural network, that is, the technology of discovering the change area through the CNN feature changes on the previous and later images, can be widely used in remote sensing natural resource monitoring, geographic and national conditions data update, earthquake prevention and disaster mitigation and other tasks. It has huge economic and social value. [0003] Traditiona...

AI Technical Summary

Problems solved by technology

[0005] Although the above-mentioned DCNN structure of channel synthesis and dual-branch differential fusion can abstract the changing features of the image layer by layer, it cannot realize the cross-layer reuse of the features in the change detection network structure, and lacks geometric structure information (such as edge structure) abstract description of changes

Images