抄録
The objective of this study is to reduce rear-end vehicle collisions caused by human errors. In order to achieve this objective, we studied a technique to detect preceding vehicles with image processing using a monocular in-vehicle camera and a computer. As an image characteristic for vehicle detection, we adopt the shadow under the vehicle. However, the detection of preceding vehicles using only the shadow has low reliability. In this paper, we propose a technique of identifying the region over the shadow as either vehicle or non-vehicle by machine learning. Experimental results show that the proposed method is more effective than the method using only the shadow.
