Remote Sensing Image Change Detection Based On U-Net Model

Remote sensing image change detection is a method to analyze and judge the changed area of remote sensing images in different periods of the same region through image processing methods.This is conducive to the monitoring of land use and land cover change,and the protection and detection of cultivated land and land resources.As an interdisciplinary technology,it is one of the research hotspots of remote sensing.The rapid development of satellite technology improves the resolution of remote sensing image,expands the coverage,and the information contained in remote sensing image is more and more abundant.Because the traditional remote sensing image change detection method has strict requirements on the image preprocessing stage,and many aspects need to be operated manually,it will obtain low accuracy change detection results.So how to use the information efficiently and reasonably has become a new challenge.With the continuous improvement of computer performance and the continuous progress of artificial intelligence technology,the deep learning method has gradually entered the public field of vision.Compared with the traditional method of extracting specific information for specific tasks,the original data can be extracted layer by layer through the deep learning method to obtain deep information,so the ability to describe the problem will be more detailed.Therefore,based on GF-1 remote sensing image,this paper uses deep learning u-net model with strong learning ability to deal with the problem of remote sensing image change detection.The specific research contents are as follows1.The traditional remote sensing image change detection methods are divided into direct comparison method and post classification comparison method,and several common methods are introduced.The systematic explanation uses deep learning to process remote sensing image change detection method,and explains the main ideas and architecture of convolution neural network and full convolution neural network in detail.At the same time,the u-net model used in this paper is explained in detail,and the u-ne model is analyzed The model structure and parameter setting of T model are analyzed.2.This paper discusses the process of remote sensing image change detection,including image preprocessing,classification system development,image and label sample set construction,model parameter selection,model configuration and optimization,change detection result map and accuracy evaluation.3.In order to test the feasibility of u-net model in remote sensing image change detection and its advantages over traditional remote sensing image change detection methods,Shuangcheng District of Harbin city is selected as the research area,and GF-1remote sensing images in 2013,2016 and 2019 are selected as the research data.By constructing u-net model,adjusting model parameters and optimizing the network,automatic change extraction is realized The change detection results are obtained.By comparing and analyzing the traditional and u-net model of remote sensing image change detection methods,it can be concluded that the accuracy of u-net model is higher than the traditional remote sensing image change detection methods,and the change detection effect is better.The results show that,as a deep learning model,u-net model can deeply mine the information contained and express deep-seated features.Through the combination of GF-1 remote sensing image and u-net model to detect the change of remote sensing image,it can achieve higher detection accuracy and operation efficiency.At the same time,it also shows the effectiveness of u-net model in remote sensing image change detection.