A Rapid Defect Detection Method for Hard Tools Based on Improved YOLOv9

To address the need for on-site detection of industrial tools, this paper proposes a rapid defect detection algorithm for rigid industrial tools based on an improved YOLOv9 framework. The algorithm enhances the YOLOv9 object detection model by integrating an SCDown module and a small target detection branch. Considering the scarcity of defect data for rigid industrial tools, we first pre-train the model on publicly available industrial datasets of steel materials or railway tracks, followed by transfer learning on a custom-built rigid industrial tool dataset to construct the rapid defect detection model. Experimental validation confirms that the proposed method achieves a detection speed of 61 frames per second and 94.3% defect detection accuracy on edge computing devices, fulfilling the requirements for real-time defect inspection of rigid industrial tools in edge deployment scenarios.