High-resolution SAR image change detection method based on global-local spp Net

Abstract

The invention discloses a high-resolution SAR image change detection method based on global-local SPP Net, comprising the following steps: selecting part of labeled data as training samples from two registered SAR images of the same area with different phases; The training samples are normalized to [0, 1], denoted as X1; m groups of larger-scale image blocks are selected from X1 and sent to the local large-scale SPP Net for region of interest detection training, and the trained interest region is obtained. Area detection model; select n groups of smaller-scale image blocks from X1 and send them to the local small-scale SPP Net for change detection training to obtain a trained change detection model; then arbitrarily select a large image of a×b from the area to be detected, As test data, it is sent to the local large-scale SPP Net region of interest detection network for ROI testing, and the final ROI test result is obtained; then the obtained ROI detection results are sent to the local small-scale SPP Net change detection network for change detection testing. Get the final change detection result map.

Description

【Technical field】 [0001] The invention belongs to the combination of deep learning and remote sensing image processing, and specifically relates to a high-resolution SAR image change detection method based on global-local SPP Net, which realizes the change detection of high-resolution SAR images. 【Background technique】 [0002] In recent years, with the rapid development of aviation and aerospace remote sensing technology, change detection technology has made some progress in all aspects after decades of development. From the perspective of data sources, change detection is no longer limited to the use of a single remote sensing image, but comprehensively utilizes multi-source, multi-platform, multi-resolution remote sensing images, GIS data and some auxiliary data to detect change information; from a technical point of view , the traditional change detection methods are becoming more and more perfect, and new methods are constantly emerging. Among them, the traditional SAR...

AI Technical Summary

Problems solved by technology

However, this method is highly dependent on the difference map, and good detection results can only be obtained on the basis of obtaining a better difference map

At present, there are not many researches on SAR image change detection combined with deep learning. Most of the implemented detection methods are for small-scale images, using DBN or AE methods. For high-resolution images, the accuracy is relatively low

Images