A 3D Coarse-to-fine Localization Method of Charging Port for Electric Vehicle Automatic Charging System

The localization of the charging port for the electric vehicle automatic charging system is a long-standing problem because of the texture-poor charging port and the high precision requirement in this task. In this paper, we propose an coarse-to-fine method for the localization of charging ports on electric vehicles, aiming to enhance the automation and efficiency of the charging process. This approach utilizes a data-driven object detection model to detect the charging port in the pictures captured by a 3D industrial camera and gets the point cloud of charging port by back-projecting the detected region of aligned depth images. Subsequently, a PCA-based point cloud processing technique is applied to provide a coarse 6D localization for the robot arm. In the fine localization stage, a hierarchical template matching approach is utilized to get the precise position and orientation information of the charging port, enabling the robot arm to perform the automatic insertion of the charging gun under various environmental conditions. Extensive experiments are conducted on real electric vehicles to validate our method. The experimental results demonstrate that our approach achieves high localization accuracy with a translation error of less than 0.5mm and a rotation error of less than 1.0° and a high success rate of 96% when deployed in the robot arm for automated charging. Furthermore, the method exhibits robustness and stability across different environments.