Marine Vehicles Synthetic SAR Dataset Generation for Automatic Target Recognition
Remote sensing of marine vehicles is involved with object detection, tracking, and classification. This task is a real challenge in the maritime environment. In recent years, the deep learning techniques are demonstrated to be effective classifiers for this purpose, however, they need to be trained properly using rich training datasets that are rarely publicly available. To overcome this limitation, there is a need for a large-scale multi-look dataset of marine vehicles while take account of different sensor ranges, azimuth and elevation observation perspectives, operating contexts, ocean and atmospheric conditions, and marine vehicle type and wakes models. In this study, we used IRIS electromagnetic modeling and simulation system for virtualization of such a maritime environment and test vehicles and created different scenarios signifying the requirements. Through this approach, we initially construct and employ the physics-based CAD models of the test marine vehicles and their corresponding wake formation patterns that realistically represent their operating signature in the marine environment. Next, we apply our specific remote sensing techniques (i.e., EO/IR, SAR, and LIDAR) to generate unique multimodality synthetic imagery of the test marine vehicles. To evaluate and verify the effectiveness of this approach, we compared our generated simulated marine vehicle imagery with those images of the corresponding physical remote sensors. In this paper, we discuss the technical aspects of this work and detail our primarily evaluations of the obtained results.
"Marine Vehicles Synthetic SAR Dataset Generation for Automatic Target Recognition"
ETD Collection for Tennessee State University.