Automatic parking slot detection based on around view monitor (AVM) systems

Abstract

Automatic parking systems have significant effects on intelligent transport systems (ITS) and have been extensively researched. However, most existing vision-based automatic parking slot detection methods cannot obtain the desired results due to variation in light intensity or complex obstacle conditions. Besides, most previous parking slot detection methods only consider the target position occupied by vehicles but ignore the existence of small objects such as traffic cones and parking lock inside the parking marking. In order to overcome these insufficiency, this paper proposes a novel vision-based automatic parking slot detection method to detect various parking slot markings by Around View Monitor (AVM) system which consists of four fisheye cameras around the vehicle. The process utilizes a Line Segment Detector (LSD) based on edge information to detect parking slot markings which have a pair of parallel lines with the fixed distance in AVM image, then use the image segmentation algorithm and the stereo vision algorithm to calculate the height of small obstacles inside the parking marking. Experimental results show the proposed methods perform better than method based on Hough transforms in continuity and completeness. Furthermore, the proposed method can effectively reduce the frequency of false detections and missing detections.