Abstract
Keywords:
stereovision, free space modelling, obstacle detection.
Abstract
This paper presents a new method proposed to detect a free parking slot in a urban environment. Such a function is already proposed on new vehicles equipped with few ultrasonic devices, but these sensors give too poor data in order to cope with all situations. A more generic approach, based on a stereovision sensor mounted on the vehicle, is proposed. While going along a street, a periodic function can be activated at 10Hz, to look for a parking slot. It consists in several tasks executed in parallel using a pipeline organization: image acquisition and rectification, pixel-based stereo matching, 3D reconstruction, construction of a scene model with road limits and perceived obstacles, fusion in an environment model, using an estimation of the vehicle motion, and finally, the detection of a free parking slot. Experimental validations on several testbeds, have shown that the final precision on the free space dimension is about 10cm.