机械

针对多固体生活垃圾分类问题，提出一种自动分类机器，对其结构、控制电路、算法进行了设计。利用四个矩阵光纤传感器、两级传送带对垃圾进行分拣；使用YOLOv5网络特征提取模型和迁移学习的垃圾识别方法，有效解决了垃圾数据集较少的问题。基于YOLOv5构建的检测模型，包括依次连接的Backbone模块、Neck模块、Head模块和serial通信模块。该模型在自建的垃圾数据集上进行训练和测试，平均准确率达到0.99。在实际应用中，将训练好的模型部署在自行研制的开发板上，与自主制作的垃圾分类机器配合使用。实验结果表明，该机器能够准确识别垃圾种类并完成分类回收，针对10～15个垃圾，从投入到识别完成用时15～20 s，表现出良好的稳定性和效率。

Aiming at solving the problem of solid household waste classification, our team propose a scheme for an automatic sorting machine and innovatively design the structure, control circuit, and algorithm. Four matrix fiber optic sensors and two-stage conveyor belts are used to sort garbage. By using the YOLOv5 network feature extraction model and transfer learning, the garbage recognition method is employed to effectively solve the problem of limited garbage datasets. The detection model based on YOLOv5 includes Backbone module, Neck module, Head module, and serial communication module connected in sequence. The method was trained and tested on the self-built garbage data set, which has achieved average accuracy of 0.99. In practical application,the trained model is deployed on the self-developed board and used with the self-made garbage sorting device.Experimental results show that the system can accurately identify the types of garbage and complete the classification and recycling. It takes 15 to 20 seconds to identify around 10 to 15 pieces of garbage since being thrown into the machine,and the machines shows good stability and efficiency in use.