An Enhanced Yolov5 Algorithm Based on Distill Model for Biomass Material Detection

In response to the intelligent needs of automatic feeding detection systems in the biomass power generation industry, this paper proposes an enhanced YOLOv5 algorithm based on a distillation model. An image dataset containing various types of biomass materials, such as wood chips, straw, and leaves, was constructed. The model was optimized through data augmentation and transfer learning to enhance its generalization ability and accuracy. By combining high-resolution image acquisition with distillation model optimization, this approach effectively addresses issues like complex environments and image blurriness in biomass power plants, providing a more reliable data foundation for subsequent detection. Using the enhanced YOLOv5 model for object detection, combined with three-dimensional information obtained from depth cameras, precise positioning of biomass materials is achieved. Experimental results show that this method significantly improves performance in biomass material detection tasks, with an average precision (mAP) reaching 92.2%, an improvement of 6.5% compared to original YOLOv5 models. This effectively enhances detection accuracy and stability, providing technical support for the automation upgrade of biomass power plants.