Radar-based road-traffic monitoring in urban environments

This work presents a novel approach to object detection and tracking in urban environments using images obtained from a radar network, deployed in an urban environment. The proposed system detects, tracks and computes the speed of vehicles and generates alerts when vehicles exceed the predefined road speed limit. The available radar model is a low-cost device oriented to marine environments rather than terrestrial applications. For this reason, we emphasize in the development of a realistic, robust, efficient and effective algorithm which deals with the hardware limitations to provide a suitable overall performance. To reach this objective, we propose dual background subtraction model to detect objects and a tracking method based on the particle filter algorithm. Furthermore, to ensure real time restriction even in HD imagery, our method takes advantage in a natural way of multicore systems and exploits advanced SIMD capabilities available in last multicore processors families. Experimental results demonstrate that the proposed system is able to detect and track multiple objects and to provide speeding alarms when needed. It is also capable to handle target occlusions and disappearances derived from the radar limitations and the noisy urban environment.