Abstract: With the development of Energy Internet, Jilin electric power communication network is becoming more and more intelligent, which brings new challenges to communication operation and maintenance. The spare parts management is the key link of communication operation and maintenance, facing the lean management of spare parts with various quantity, standard and variety has become the main difficult problem at present. In this paper, a semantic segmentation and location algorithm based on machine learning machine is proposed to construct the semantic sub-library of spare parts, and to realize fast image retrieval based on semantic and content in the location stage, to realize the rapid and accurate location of spare parts, improve the efficiency of equipment inventory, emergency repair, spare parts for the standardization, lean, life-long management to provide basic security.