一种融合无监督特征增强的立铣刀缺陷检测算法研究OA

As an important tool in the field of precision machining,the surface quality of end mill directly affects the machining safety and the quality of finished products.In the actual production process,due to factors such as manufacturing process fluctuations and manual operation,the surface of the end milling tool will inevitably produce micron-scale defects.Aiming at the problem that the existing supervised detection algorithms rely on a large number of high-quality labeled data and unsupervised detection algorithms are prone to overkill,this paper proposes a supervised end mill defect detection algorithm that integrates unsupervised feature enhancement.Based on the YOLOv11 supervised detection framework,this method introduces the EfficientAD unsupervised anomaly detection branch to extract the high-dimensional anomaly features of the end mill image,and enhances the input feature expression ability through the feature fusion mechanism.The experimental results show that the method can effectively reduce the false detection rate and missed detection rate under the premise of ensuring real-time performance,and improve the detection accuracy of small defects.It has strong robustness and generalization ability,and can meet the practical application requirements of on-line defect detection of end mills.