Efficient and Accurate HD Map Generation for Unstructured Automated Valet Parking (UAVP) Environments

This paper proposes a new approach to detecting road boundary for generating high-definition (HD) maps of unstructured automated valet parking (UAVP) environments. The proposed framework presents a height moment tile to efficiently represent the height distribution of UAVP environments, a filtering scheme to remove both static and dynamic obstacles, a curb detection algorithm, and a road boundary expansion algorithm. Our approach is evaluated through experiments conducted at an outdoor parking lot in the Pusan National University. The results show that the proposed framework is able to detect accurate and efficient road boundary in UAVP environments.