Detection and Identification of Coating Defects in Lithium Battery Electrodes Based on Improved BT-SVM

Abstract

Aiming to address the problems of uneven brightness and small defects of low contrast on the surface of lithium battery electrode (LBE) coatings, this study proposes a method for detection and identification of coatings defects in LBEs based on an improved Binary Tree Support Vector Machine (BT-SVM). Firstly, adaptive Gamma correction is applied to enhance an image, and an improved Canny algorithm combined with morphological processing is used to accurately detect the defect regions. Secondly, the shape and grayscale features of the defects are extracted from the connected defect areas, and these features are then fused and normalized. Finally, a BT-SVM multi-class classification model is constructed, with the Whale Optimization Algorithm (WOA) employed to assist in hyperparameter tuning. The experimental results show that the proposed method can effectively detect and identify five common types of defects in the coating of LBEs, including scratches, bubbles, metal leakage, particles, and decarbonization, with an average detection accuracy of 94.4% and an average detection time of less than 0.2 s, meeting the real-time detection requirements for online defect inspection. After Whale Optimization, the BT-SVM defect recognition model achieves an average recognition accuracy of 98.7%, significantly enhancing the performance of current defect detection technologies for LBE coatings.