Automatic-Detection Method for Mining Subsidence Basins Based on InSAR and CNN-AFSA-SVM

Mining subsidence disasters are common geological disasters. Accurate and effective identification of their deformation position is significant in preventing controlling monitoring illegal mining. In this study, deep learning, combined with a support vector machine (SVM), has been used to establish an automatic-detection method for mining basins using Sentinel-1A data. The Huainan area was selected as the experimental verify method. interferogram obtained differential radar interferometry (D-InSAR) process data seven landscapes, basin other targets were extracted manually training samples. Subsequently, AlexNet, VGG19, ResNet50 convolutional neural networks (CNNs) extract feature vectors SVM classifier, detected large-area InSAR interferogram. Non-maximum suppression remove repeated search box improve detection accuracy basins; artificial fish swarm algorithm strong optimization ability good global convergence introduced into parameter construct improved ResNet50_SVM model. results show that: (1) three CNN_SVM methods can accurately detect dry-mining automatically large regional interference maps, providing essential scientific basis government monitor activities prevent control areas; (2) approximately 80%, that 83.7%, superior AlexNet_SVM VGG19_SVM; (3) based on AFSA 88.3%, which better than unimproved Resnet50_SVM