Research on Real-Time Sensing Method of Three-Dimensional Information of Substation Based on Convolutional Neural Network Stereo Matching

The power system serves as a critical pillar of modern society, with its stable operation being essential for maintaining social order and ensuring the reliability of power supply. As the power system continues to evolve, the operation and maintenance of substations have become increasingly significant, directly impacting the stability and reliability of the overall system. Substation maintenance typically involves proximity to live electrical components, where the environment is complex, and measuring the distance between personnel and live equipment or conductors is challenging. Any negligence by the personnel may result in an insufficient safety margin from live components, potentially leading to accidents.This study proposes a real-time three-dimensional information perception method for substations based on CNN stereo matching, aiming to enhance safety monitoring and risk prevention capabilities in substations. The core of this research lies in the PSMNet stereo matching network, which leverages the robust feature extraction capabilities of CNNs and the spatial perception strengths of stereo vision. By integrating the CBAM attention mechanism, the model's performance is further improved. This approach enables the real-time and accurate acquisition of depth information and spatial structure at substation sites, providing reliable technical support for the assessment and maintenance of safe distances.