One of the significant issues is resource recycling of outdated objects classification. It can effectively improve the efficiency of resource recycling and further reduce the harm caused by environmental pollution. By the gradual intellectualization of modern industries, traditional image classification algorithms no longer proper the requirements of garbage classification because there are lots of requirements for sorting equipment. This paper proposes to build outdated items' Classification Network “GCNet” based on a convolutional neural network. By constructing a realization mechanism, the model completes local and global feature extraction. It can complete productive feature information obtained. At the same time, through the feature combination mechanism, it’s of different levels and sizes are fused to make more effective use of properties and avoid gradient disappearance. Experimental results prove that “GCNet” has achieved promising results on related outdated items' classification data sets. It has received an improvement in image identification of oldestablished items.