基于DFD-YOLOv11n的钢材装备表面缺陷检测算法研究OA

To address the issues of fuzzy features,insufficient multi-scale expression and limited detection accuracy in surface defect detection of steel equipment,a lightweight improved algorithm DFD-YOLOv11n based on YOLOv11n architec-ture is proposed.The algorithm achieves performance optimization through triple structure innovation.Firstly,the C3K2 module of dynamic snake convolution is introduced to improve feature extraction,and the capturing ability of elongated and curved features is significantly enhanced by adaptive convolution kernel deformation strategy.Secondly,the feature focusing diffusion pyramid network is designed to improve the utilization of context information through the bidirectional fusion mechanism of multi-scale features.Thirdly,the dynamic task alignment detection head is designed to improve the detection performance through the cooperative optimization strategy of classification and location branches.The experi-mental data show that the mAP of the improved algorithm on the NEU-DET dataset reaches 77.1％,which is 6.5 percent-age points higher than that of the YOLOv11n model,and the detection accuracy is 6.8 percentage points higher.In terms of maintaining lightweight characteristics,the number of model parameters(2.74 × 106)and computational complexity(9.4× 109)are controlled in the category of cutting-edge lightweight models,while achieving a real-time detection speed of 116 frames per second(FPS).The DFD-YOLOv11n achieves an optimal balance between detection accuracy and inference speed,and its comprehensive performance indicators provide a new solution for industrial-grade surface defect detection.