Hough transform based ego-velocity estimation in automotive radar system

Abstract

In general, the velocity estimated through the automotive radar is the relative velocity between the ego-vehicle and the target. To estimate the absolute velocity of the ego-vehicle (i.e. ego-velocity), it is necessary to get help from other automotive sensors, such as cameras and lidars. In this letter, therefore, a method for estimating the ego-velocity is proposed using only radar sensor data. To estimate the ego-velocity, stationary targets in the detection result have to be identified because the velocity relative to the stationary target can be directly converted to the ego-velocity. In the proposed method, the radar data is first processed to extract the relative distance from the ego-vehicle to the target, relative velocity between the ego-vehicle and the target, and angle between the ego-vehicle and the target. Then, the detected targets are represented in the angle-velocity domain because moving and stationary targets in this domain exhibit different characteristics. In other words, the moving targets are randomly placed, but the stationary targets exist around a specific line. Finally, the Hough transformation is applied to automatically determine the line, and the ego-velocity is calculated from the transform result. The proposed method shows a maximum error of 2.4% in the velocity range of 30-90 km/h.