Abstract
This paper describes a robust tracking method that combines a scanning laser radar (SLR) with a camera to measure headway distances and lateral positions more accurately. With this method, a preceding vehicle is first detected on the basis of the digital data output by the SLR unit, and the vehicle is then tracked by both SLR and the camera. A variable shape template matching method is used in processing the images taken with the camera. The image recognition technique can track the preceding vehicle despite frequent changes in the headway distance and the direction. Robust and highly accurate tracking is achieved by combining the wide angle, high-resolution measurement obtained by image recognition with the SLR distance measurement. The proposed method has been validated on a driving simulator using data recorded on Japanese roads. The preceding vehicle was successfully tracked and its position was measured as expected on curves and on straight roads in general. The results also confirmed that this method has some advantages over detection by SLR alone in difficult detection situations.
