Fully Automated Valet Parking System Based on Infrastructure Sensing

Abstract

In this paper, we propose a novel automated valet parking system exploiting infrastructure hardware for a more applicable, safe and efficient solution. The infrastructure equipped with camera and computing device-detect automobiles on IPM (Inverse Perspective Mapping) image, generate a path to the empty parking space and control the vehicle to track the path. In order to recognize empty parking spaces and vehicles, a deep learning-based recognition model was used to have a robust recognition rate in various lighting conditions and environments. In addition, a more efficient route was created by creating a parking path considering the vehicle model. Our system was tested in a 1/4-sized experimental environment; it showed a high parking success rate and confirmed that our system works well in various environments. Therefore, from our system proposed in this paper, we can see the possibility of the development of an efficient and highly intelligent free autonomous parking system based on infrastructure sensing. Our source code is publicly available at https://github.com/FVCD2019. © 2021, The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd.