Weld seam detection and extraction method based on S-YOLO

Weld seam recognition and feature point extraction are critical for welding automation and industrial intelligence. While existing deep learning models perform well in weld seam feature extraction, their high computational complexity and insufficient real-time capabilities limit their application in embedded devices and industrial environments. This paper proposes a lightweight weld seam segmentation algorithm, S-YOLO, based on the YOLOv8n-seg model. By employing a lightweight backbone, the model reduces computational complexity while enhancing real-time performance and accuracy. Experimental results show that S-YOLO improves accuracy, reduces floating-point operations by 17.5% and parameters by 27.5%, and meets real-time industrial requirements. This research offers a lightweight, efficient solution for weld seam recognition and feature point extraction, reducing resource consumption and broadening deep learning applications in industrial scenarios.