EIM-YOLO: A Defect Detection Method for Metal-Painted Surfaces on Electrical Sealing Covers

Electrical sealing covers are widely used in various industrial equipment, where the quality of their metal-painted surfaces directly affects product appearance and long-term reliability. Micro-defects such as pores, particles, scratches, and uneven paint coatings can compromise protective performance during manufacturing. In the rapidly growing new energy vehicle (NEV) industry, battery charging-port sealing covers are critical components, requiring precise defect detection due to exposure to harsh environments, like extreme weather and dust-laden conditions. Even minor defects can lead to water ingress or foreign matter accumulation, affecting vehicle performance and user safety. Conventional manual or rule-based inspection methods are inefficient, and the existing deep learning models struggle with detecting minor and subtle defects. To address these challenges, this study proposes EIM-YOLO, an improved object detection framework for the automated detection of metal-painted surface defects on electrical sealing covers. We propose a novel lightweight convolutional module named C3PUltraConv, which reduces model parameters by 3.1% while improving mAP50 by 1% and recall by 3.2%. The backbone integrates RFAConv for enhanced feature perception, and the neck architecture uses an optimized BiFPN-concat structure with adaptive weight learning for better multi-scale feature fusion. Experimental validation on a real-world industrial dataset collected using industrial cameras shows that EIM-YOLO achieves a precision of 71% (an improvement of 3.4%), with mAP50 reaching 64.8% (a growth of 2.6%), and mAP50–95 improving by 1.2%. Maintaining real-time detection capability, EIM-YOLO significantly outperforms the existing baseline models, offering a more accurate solution for automated quality control in NEV manufacturing.