Research on Intelligent Monitoring Terminal of Transmission Line Hidden Target Based on Multi-dimensional Optical Information

The construction and maintenance of transmission lines are critical to ensuring national energy security. Current methods for detecting hidden targets in transmission lines involve uploading massive amounts of images from various monitoring points to a centralized video platform and then carrying out the algorithmic recognition, which demands an extremely high computational load on the platform. Therefore, designing edge-intelligent algorithm deployments at the monitoring terminals can effectively distribute the load, reduce the redundancy of uploaded images, and lower the computational burden and power consumption of the central platform. To ensure the safe and stable operation of transmission lines, this paper proposed a novel intelligent monitoring terminal based on multi-dimensional optical information fusion. TensorRT inference optimization engine, CUDA programming model and parameter quantization method are used to achieve edge deployment and optimization of hidden target detection algorithm on the terminal. Results show that the intelligent surveillance terminal effectively balances the contradictory relationship between model accuracy, computational ability and power consumption, and ultimately improves the inference speed while maintaining the accuracy.