Multiple Camera based Lightweight Localization Technology for Indoor Parking lot

Recently, there has been an active development in AVP (Auto Valet Parking) solutions for various last mile services. Despite the efforts, there is a limitation in reaching complete automation by only relying on autonomous vehicle sensors in complex and diverse parking environments. Especially in an indoor parking environment where GPS sensors are not supported, the autonomous vehicles have difficulties in locating themselves. In a such environment, the localization problem can be effectively coped by using smart infrastructure installed in the parking lot to analyze parking environment information. In this paper, we propose a method for mapping multiple camera images into a single grid map, and lightweight localization algorithm by tracking vehicles in parking lots, including environment where perceptive fields of surveillance cameras are not overlapping. By providing localization information of autonomous vehicles, the safety in AVP solution can be improved.