Surface defect detection of lithium battery electrode plates based on YOLOv8n

The cathode and anode sheets are critical components of lithium batteries, and the quality of their coating significantly impacts the battery's performance and lifespan. Defective sheets can pose serious safety risks. To enhance the speed and accuracy of detecting coating defects in lithium battery sheets, this paper presents a YOLOv8n-based defect detection and recognition algorithm. Given the low resolution of lithium battery sheet images and the small size of defect targets, we have improved the C2f module in the backbone network by incorporating conditional parameter convolution technology. This allows dynamic adjustment of convolution kernel parameters, thereby strengthening the model's ability to detect minor defects. Furthermore, the ADown module is employed for down-sampling operations, which reduces the spatial dimensions of feature maps, thus improving the efficiency and accuracy of the object detection task. Experimental results indicate that, compared to the original YOLOv8n model, the improved model achieves an average precision (mAP) of 92.4%, representing a 3.3% improvement. This demonstrates that the model is highly effective for high-speed, high-precision detection of coating defects in lithium battery sheets, better meeting the demands of the lithium battery intelligent manufacturing sector.