Occlusion Detection for Dynamic Adaptation

Occlusion is a common issue for object detection and tracking applications using a remote sensor platform, especially in complex urban environments where occlusions from buildings, bridges, and trees are frequent events. While occlusions are unavoidable, the events can be predicted to occur before the object of interest is obscured if there is prior knowledge of the observed environment. To aid in object detection and tracking tasks, we create an environment to map terrain and find obscured regions in the scene which helps with re-detecting objects once they are no longer obscured. We propose a dynamic data driven applications systems (DDDAS) framework for detecting occluded regions in an imaged scene by integrating streams of real data with a physics-based simulation model that updates based on the most recent images.