A Method for Camera Vision Based Parking Spot Detection 2006-01-1290
A main requirement for a driver assisting automated parking system is the identification and localisation of a free parking spot. A sensory system consisting of a reverse aid camera, a processing unit, vehicle track sensors and a link to the vehicles CAN-Bus is able to detect a parking spot, using vision processing algorithms.
This report starts with a case differentiation regarding parking situations. In reality a combination between several attributes and conditions define the initial parking situation and the specific demand on the sensory system to detect the parking spot.
The described method for camera vision based parking spot detection is characterised on specific situations, such as parking spots marked by painted lines and the kerbstone and well lighted rich-contrast situations. The method in general starts with the image processing on the incoming camera images. Based on the detection of attributes hypotheses for the identification of free parking spots are established and centralised in a hypotheses pool. Beneath the detected attributes another evaluation criteria is established, by the tracking of the hypothetical parking spots during the movement of the car. The most presumable parking spot is then displayed in the image of the reverse aid camera.
The report concludes with an evaluation of the achieved results under specific initial parking situations.