실내 무인 자율 발렛파킹을 위한 Way-point 검출 및 경로 예측 방법

초록

In this paper, we propose a way-point detection method for an autonomous valet parking system in an indoor environment where circular LED way-points are installed on the ceiling. As a pre-processing, camera‘s exposure value is automatically adjusted using incremental searching to remove other objects except way-points. We set binarization threshold using a standard deviation from Gaussian function modeled by background intensity distribution. For predicting the path where a vehicle is going to drive, the detected way-points are classified into "straight path class’ and "turning path class’ according to the its’ location, and the least square based line fitting is applied to each class. The experimental result using dataset captured in the indoor parking lot shows good performance of way-point detection and path estimation.