Deep Learning for SAR Applications: Port Monitoring, Airbase Monitoring and Land Cover Classification with RADARSAT-2

摘要

Deep learning (DL) has the potential to automate SAR analysis, especially for wide area mapping and monitoring patterns of activity on a regular basis. We apply deep learning to three use cases using data from RADARSAT-2: port monitoring, airbase monitoring and land cover classification. We adapt state-of-the-art DL object detection and semantic segmentation approaches from the computer vision domain for SAR analysis. We focus on CNN (Convolutional Neural Network) DL architectures and how to modify them for SAR’s coarser spatial resolution, large dynamic range, side-looking imaging geometry and the presence of speckle noise.