A Method of Building Change Detection in Remote Sensing Imagery Based on Convolutional Neural Network

Abstract

The present invention provides a method for detecting building changes in remote sensing images based on convolutional neural networks, comprising the following steps: Step 1, constructing a sample library according to existing images and land cover vector files, the sample library includes building classification data sets, Building change detection simulation data set; step 2, use the building classification data set in the sample library constructed in step 1 to train the building extraction network (Mask R-CNN or multi-scale fully convolutional network) to learn high-resolution The building features in the remote sensing image, and then use the trained network to extract the multi-temporal building binary classification map of the change area; step 3, use the simulated data set in the sample library built in step 1 to train the change detection network Mask CD‑net , and then directly use Mask CD‑net to predict the building change detection dataset and get the change detection results. If it contains real change data sets, it will be used to further train and refine Mask CD‑net to get better change detection results.

Description

technical field [0001] The invention relates to a deep learning method for detecting building changes in high-resolution remote sensing images, and in particular proposes a self-trainable change detection network. This method can be used for building change detection in two-phase remote sensing images. Background technique [0002] Change detection refers to the process of identifying differences in the state of an object or phenomenon by observing the same area at different times. Change detection frameworks use multitemporal datasets to qualitatively analyze the temporal effects of phenomena and quantify changes. In remote sensing applications, changes are viewed as changes in surface composition with different rates. Land cover and land use change information is so important that it has practical uses in various applications such as GIS database updates, urban development trends, natural hazard assessment and forest fires, logging, etc. [0003] Change detection analys...

AI Technical Summary

Problems solved by technology

At the same time, the network trains itself by simulating the parallax changes of buildings, which solves the problem of complete registration of buildings on high-resolution aerial images from different perspectives.

Images