Settlement Area Detection in InSAR Interferogram Using YOLO V5 Model Over Mining Region

The land surface deformation is a common phenomenon in mining areas. Effective and rapid monitoring of it is important for mining management, especially for the detection of illegal mining. Thanks to interferometric Synthetic Aperture Radar (InSAR), which enables surface deformation monitoring with millimeter accuracy, it is possible to find the settlement area over mining region. However, it will be a very time-consuming and labor-intensive task when the monitoring area becomes larger and InSAR data becomes more frequent, such as the Sentinel-1 data, the interferogram is as short as 12 days. With the development of deep learning, it has shown great potential in the field of remote sensing image processing such as segmentation, classification and recognition. In this study, we combine the InSAR technology with deep learning by applying the YOLOv5 model to detect the deformation fringe in the wrapped interferogram which is corresponding to the rapid ground settlement area in mining region. By testing the trained model in Zhalainuoer District, Hulunbeier City and Dongsheng District, Ordos City, the deformation areas caused by mining are automatic detected with 92% precision.