Monocular Image and AdaBoost Learning Based Nighttime Preceding Vehicle Detection for ADAS and Intelligent Headlamp System
Seong-Uk Kim, Jeongmin Park, Joon-Woong Lee · Journal of Institute of Control Robotics and Systems · 2017
We propose a new algorithm to detect preceding vehicles using AdaBoost learning and linear-regression-based tracking in the road traffic scene at night. Many existing algorithms that can detect preceding vehicles at night mainly rely on the threshold values of colors or intensities in exploring the light blobs that correspond to the taillights of the vehicles because the salient cues for nighttime vehicle detection are lamp lights. In addition, the algorithms require many heuristic rules to select the real taillights among the detected light blobs. Hence, these algorithms suffer from difficulties in determining the threshold values and rules that can address various illumination conditions and camera exposure parameters. In contrast, the proposed algorithm, which uses the AdaBoost learning, does not rely on such threshold values and rules. To enhance the robustness of the detection while overcoming difficult traffic environments, the algorithm also uses a tracking scheme based on linear regression. The experimental results of actual road images show the effectiveness of the proposed algorithm.