Abstract：In order to solve the problem of single target equipment, simple abnormal state, low comprehensive recognition rate and accuracy rate in the abnormal state identification of underground tunnel cable equipment, an improved YOLO target detection architecture is proposed in this paper to locate cable equipment and identify abnormal state. Firstly, the image scaling method is used to adjust the image size to 448×448, and then the features are extracted by convolutional neural network. Each layer adopts the batch normalization method to standardize the model, and finally predicts the target bounding box through the RPN network. Using the image data in Zhuhai underground cable tunnel, the simulation experiment is carried out and compared with YOLO and Faster RCNN algorithm. The experimental results verify the effectiveness of the proposed method, and the algorithm has high recognition accuracy and good robustness which can be effectively used in the inspection robot system of underground cable tunnel.
