Software Implementation of Image Processing Techniques for EV Socket Pose Estimation with Neural Network Support

The increasing adoption of electric vehicles (EVs) highlights the growing need for efficient and safe charging systems. A significant challenge in automating these systems is the accurate detection and alignment of the charging socket, especially in high-power environments where manual handling poses safety risks and reduces operational efficiency. This thesis, developed at California State University Los Angeles in collaboration with InnoTech System LLC, focuses on the CCS Type 1 charging socket, a widely adopted standard in North America, with the aim of automating its detection and pose estimation. The motivation for this work arises from the need to automate the charging process to enhance both safety and efficiency. A pre-trained YOLOv8 neural network was used to detect the socket and provide an initial estimate of its position, but further refinement was required. Detecting both the position and orientation of the socket is essential to ensure proper plug alignment, a critical aspect for fully automated systems. The main objective of this thesis is to develop a system capable of detecting not only the position of the socket but also its orientation. The process begins with the YOLOv8 neural network’s estimate, followed by a series of refinement steps using OpenCV in Python for image processing. Testing under various environmental conditions ensures robustness and reliability in real-world scenarios, where factors such as lighting and occlusions may affect performance. This thesis integrates machine learning with image processing to address these challenges, presenting a reliable and adaptable solution for the automation of EV charging systems. The research is structured across key chapters, including a review of the state of the art, an in-depth discussion of the components used, a detailed explanation of the detection and pose estimation methods, and a comprehensive evaluation of the system’s performance through extensive testing.