Probabilistic Occupancy Filter for Parking Slot Marker Detection in an Autonomous Parking System Using AVM

Abstract

Many car makers have examined parking assistance systems and auto parking systems that automatically find free parking spaces and can park a car. The around view monitoring (AVM) can compensate for the disadvantages of distance-sensor-based detection because it can recognize parking spaces based on parking slot markers instead of empty spaces. However, in the case of AVM-based parking marker recognition, false-positive (FP) features can be detected from 3-D objects and shadows, and the parking slot marker is occluded by the vehicle on which the AVM camera is attached. In this paper, we propose a probabilistic occupancy filter to detect parking slot markers. This filter uses a series of AVM images and onboard sensors to improve the occlusion problem and reduce the FPs from other objects. Each pixel of an AVM image has an occupancy probability of parking slot marker features. The occupancy probability compensates for using vehicle motion and updates using the error model of the AVM system and the performance indicator of the feature extractor.