Static Environment Perception Based on High-Resolution Automotive Radars.

High-resolution radar sensors have the capability to perceive the surroundings around the vehicle very exactly by detecting thousands of reflection points per measurement cycle. To model the static environment with these detection points, a novel approach of occupancy grid mapping is developed in this paper. The reflection amplitudes of all data points are compensated, normalized, and then converted to a detection probability value that is based on a predefined radar sensor model. According to the movement of the test vehicle, the a posteriori occupancy probability after several measurement cycles is computed to build the occupancy grid map. Thereafter this occupancy grid map is transformed into a binary grid map, where the grid cells, with an obstacle present, are defined as occupied. Through the Connected-Component Labelling algorithm, these occupied grid cells are then clustered and all the outliers with only a few grid cells are eliminated. Then, the boundaries of the clustered, occupied grid cells are recognized by the Moore-Neighbor Tracing algorithm. Based on these boundaries, the free space of an interval-based model is determined by using the Bresenham’s line algorithm. The occupancy grid map and the free space detection results elaborated in this paper from the recorded radar measurements show a perfect match with the real road scenarios.