Mobile System to Aid in the Identification and Classification of Electrical Assets using Convolutional Neural Network

The electrical assets management is usually done through teams of specialized professionals who go to the assets of the network to obtain the images manually. The images obtained must then be used to classify them. This procedure, however, requires a high financial cost. Therefore, this work proposes a new mobile asset identification and asset classification system using an imaging approach with Convolutional Neural Network (CNN). We evaluated the system in two stages with classical techniques of feature extraction and classification, as well as two approaches using CNN. The best combination was YOLO with a 96.63% accuracy and a classification time of 0.2s with Jetson TX2. According to the results, the proposed system automates network assets identification.