The detection of recyclable garbage plays an important role in waste-to-energy and carbon neutrality. However, due to the complexity of waste accumulation in waste-to-energy plants, the current garbage grabbing algorithms have limited accuracy. In order to overcome the above problems and avoid waste of resources, a novel recyclable garbage detection algorithm and corresponding system are studied in this paper. The CenterNet model is optimized by feature fusion so that it can better extract the subtle features of garbage. The YOLO model and original CenterNet model are adopted for garbage detection, and the backbone of YOLO model is optimized with VGG and DenseNet. Based on it, a garbage detection system is designed and a recyclable garbage dataset is constructed. The validation results show that the algorithm proposed in this paper is efficient and valid.

可回收垃圾的检测在垃圾发电和碳中和方面发挥着重要作用。然而，由于垃圾焚烧发电厂垃圾堆积的复杂性，目前的垃圾抓取算法准确性有限。为了克服上述问题，避免资源浪费，本文研究了一种新的可回收垃圾检测算法及相应的系统。CenterNet模型经过特征融合优化，可以更好地提取垃圾的细微特征。垃圾检测采用YOLO模型和原始CenterNet模型，YOLO模型的主干采用VGG和DenseNet进行优化。在此基础上，设计了垃圾检测系统，构建了可回收垃圾数据集。