Research on Fault Detection Technology of Inspection Robot Based on Electric Power Big Data Platform and Target Detection Algorithm

In order to improve the accuracy of foreign object recognition of transmission line in inspection robot operation and maintenance, a fault detection method based on improved YOLOv7 is proposed. On the basis of the popular object detection network YOLOv7, this method introduces the CA attention mechanism, and combines it with YOLOv7 by embedding CA attention mechanism into the connection part of feature extraction network and feature fusion network, so as to improve the detection effect of the model. The experimental results show that when the attention mechanism is embedded in the same position, the mPA of YOLOv7 model with CA attention mechanism is improved by 1.67%, 1.69%, 0.26%, and 1.54% compared with the original YOLOv7 model and the model with SEnet, CBAM, Conformer and other attention mechanisms. When embedding the same attention mechanism, YOLOv7 model with attention mechanism embedded the connection between feature extraction network and feature fusion network has a 2.68% higher mPA compared to the model with attention mechanism embedded at the connection position between feature fusion network and final detection part of YOLO head, and its performance has been significantly improved. Compared with classical target detection models such as Faster-RCNN and Mobilnet and popular target detection models such as YOLOv5 and C-YOLOv5, the improved YOLOv7 model proposed in this paper has better detection results in all categories of foreign body detection, and mPA is increased by 10.98%, 20.72%, 3.33% and 2.01%, respectively. It has good detection performance and is worthy of further research.