Copper and Aluminum Scrap Detection Model Based on Improved YOLOv11n

With the development of global industry, the recycling of waste copper and aluminium metals from end-of-life vehicles holds significant economic and environmental value. Machine vision technology plays a crucial role in scrap metal recycling, where accurate identification of waste metals has consistently been a key focus in this field. To address issues such as missed detections, false positives, and low detection accuracy caused by complex backgrounds in copper and aluminium scrap detection, along with the need for lightweight deployment, an improved YOLOv11n-based detection model for copper and aluminium scraps is proposed. The C3k2_Star_CAA module was designed to replace the original C3K2 module, improving feature extraction and target recognition in complex environments. The FPNSC module was introduced in place of the original SPPF module to capture finer features while keeping the model lightweight. A Feature Focus Diffusion Pyramid Network (FDPN) was adopted to substitute the original feature pyramid network, enhancing multi-scale feature fusion through FocusFeature modules and a diffusion mechanism. Additionally, a Dynamic Detection Head (DyHead) was incorporated to replace the original detection head, integrating multi-scale features to improve detection accuracy for objects of different sizes. Experimental results demonstrate that the improved model achieved a 9% increase in [email protected], a 6.5% improvement in [email protected]:0.95, along with 3.9% higher Precision, 12.4% better Recall, and 8.2% greater F1-Score on the dataset, while also showing significant improvements in both parameter count and detection speed, making it more suitable for copper and aluminium scrap detection tasks. Furthermore, generalisation experiments using public datasets confirmed that the model exhibits strong generalisation capabilities.