A Vision‐Based Classification‐Assisted Method for Foreign Object Detection in Wireless Power Transfer Systems

ABSTRACT

Wireless charging technology offers convenience, efficiency and safety, and is widely used in consumer electronics, electric vehicles and medical implant devices. However, metallic foreign objects entering the coupling area can induce eddy‐current heating, reduce transmission efficiency and pose serious safety risks. Therefore, reliable foreign object detection is essential for safe operation. This letter proposes a classification‐assisted foreign object detection method for wireless power transfer systems based on ResNet‐18 and PatchCore. The approach enables coil classification and foreign object detection through image‐based recognition, eliminating blind spots associated with non‐metallic objects and providing a visual foundation for interpretable safety decisions. Experimental results demonstrate a joint accuracy of approximately 95% for coil classification and foreign object detection.