Parking slot tracking and vehicle localization by fusing AVM and vehicle chassis sensors information based on dynamic Bayesian network

Abstract

During the process of automatic parking, parking slot detection often becomes discontinuous due to visual blind areas and vehicle occlusion, causing the vehicle to lose relative positional information with respect to the parking slot, which may eventually lead to parking failure. To address this issue, this paper proposes a fusion method of AVM and vehicle chassis sensor information based on a dynamic Bayesian network. The proposed method enables simultaneous parking slot tracking and vehicle self-localization, thereby ensuring continuous and accurate detection of both the vehicle’s position and the parking slot’s position. Firstly, the coupling mechanism between parking slot tracking and vehicle localization is modeled based on dynamic Bayesian network. Additionally, A robust extended Kalman filter algorithm for accurate vehicle localization and parking slot tracking is used, which can fuse the information of vision and vehicle motion and reduce the impact of dynamic measurement error. Finally, the proposed algorithm is validated through real car tests with a light pickup truck. The test results show that the proposed algorithm can improve the accuracy of parking slot tracking and vehicle localization and can meet the functional requirements of automatic parking.