Garbage classification can reduce pollution, promote resource circulation, and is the key to sustainable development. An intelligent garbage classification algorithm based on the improved YOLOv9, which is named MC-YOLOv9, is proposed to address the challenges of classification and resource utilization in the current garbage management field. Building upon the original YOLOv9 algorithm, MC-YOLOv9 introduces the Dual Convolution (DualConv) module, Multi-Scale Dilated Attention (MSDA) mechanism, and innovates on the loss function. A multi-class garbage dataset is constructed to train and optimize the MC-YOLOv9 model, aiming to enhance its accuracy and efficiency in garbage classification, and is compared with other common classification methods. Experimental results demonstrate that the improved MC-YOLOv9 model maintains the original accuracy of YOLOv9 while reducing GFLOPs by 9% and increasing recognition speed by 45.9%. Compared with traditional models such as YOLOv7, YOLOv7-Tiny, YOLOv5, and Faster R-CNN, the improved MC-YOLOv9 model achieves speed increases of 10.6%, 1.80%, 13.90%, and 54.9%, respectively. The experimental results indicate that the improved MC-YOLOv9 model can effectively enhance the recognition speed of garbage classification, providing new insights for the application of deep learning technology in the field of garbage management.