Improved Mask Recurrent Convolutional Neural Network for Remote Sensing Landslide Detection

Remote sensing has become an essential tool in detecting landslides, as it enables efficient assessment of huge areas. Insufficient training data, class imbalance, and complexities in extracting relevant features are some of the challenges that existing methods face. A new approach is worked in the proposed method to address these issues by combining data augmentation, semantic segmentation through DeepLabV3+, and detection accelerated by Improved Mask Recurrent Convolutional Neural Network (RCNN). The Landslide detector dataset is applied to make significant advancements in the proposed method, resulting in an accuracy of 97.34%, precision of 96.78%, and recall of 95.59%. Faster RCNN, Digital Elevation Model-CNN (DEM-CNN), Distance Domain Transfer Learning (DDTL), and Automated Landslide Detection Model (ALDM) cannot be best by this technique due to its limitations. The proposed method, of this paper aims to improve the accuracy and reliability of landslide detection in remote sensing applications by enhancing the segmentation and detection process.