Research On Key Technologies Of Satellite-UAV Heterogeneous Remote Sensing Image Change Detection For Emergency Applications

The change detection of remote sensing images is an effective technical to analyze the influence of events in the area,the historical evolution,and development trend of the target area.The current change detection technology is mainly developed for the homologous satellite remote sensing image data,but due to the long period of satellite acquisition of the same geographic image,its timeliness for change detection in emergency scenarios such as rescue and disaster relief and emergency handling is poor.In recent years,with the development of UAV technology,the advantages of timeliness and flexibility of UAV remote sensing technology are becoming prominent.If use UAV images as post-event images and detect changes with satellite pre-event images,it can meet the timeliness requirements of emergency scenarios.However,at present,there are still some technical problems in applying UAV images to change detection.For example,the difficulty of registering a large number of UAV images with a small field of view to a satellite image,and the difficulty of change detection caused by the complex difference between satellite and UAV images.In view of this,this article focuses on the problem of UAV and satellite image change detection in emergency scenarios.It is difficult to register multiple small field of view UAV images to satellite images,and the existing heterogeneous change detection methods are difficult.Effectively deal with the problem of change detection of satellite and UAV images.It is proposed to stitch UAV images to obtain large-format UAV images,and then manually register two large-scale remote sensing images,and finally based on deep learning.The key technologies and innovations studied and solved in this process in this paper are as follows:Aiming at the slow speed of stitching UAV images,this paper proposes an image stitching method based on point feature matching,called FUIS.FUIS uses the pose information of the image to filter a small number of anchor points from the feature points through the grid method,and then uses the domain window method to quickly and accurately match the anchor points,and then stitch the images.Experiments have proved the significant improvement FUIS has achieved in stitching speed.Aiming at the difficulty of using the existing change detection methods to effectively deal with the problem of satellite-UAV image change detection,this paper proposes a change detection method based on a dual-channel full convolutional network,called SUNet.On the basis of the full convolutional network,this paper designs three optimization methods for the satellite-UAV heterogeneous change detection task to improve the detection performance,including a dual-channel structure,edge assistance information,and IoU-WCE loss.This is SUNet.Experiments have proved that compared to the current state-of-the-art(SOTA)method,SUNet has significantly improved the detection accuracy in satellite-UAV heterogeneous change detection tasks.Through the research of these two key technologies,the technical process of satellite-UAV heterogeneous remote sensing image change detection technology for emergency scenarios has been realized,and the requirements of emergency event analysis and processing have been met,laying a technical foundation for the detection and analysis of remote sensing changes in emergency scenarios.