Research on Intelligent Accounting Method for Power Transmission Channel Clearing Engineering Based on Binocular Visual Perception

In order to solve the problem of complex engineering quantity accounting process and low standardization level in the clearance work of transmission channels, this paper proposes an intelligent accounting method for transmission channel clearance engineering based on binocular visual perception. Firstly, the RGB image and depth image features are extracted separately through convolutional networks, and the RGB image features and depth features are fused using cross modal feature fusion module(CMF) and compact pyramid module (CPM) to achieve automatic recognition and contour extraction of tree stumps on the job site; Subsequently, a tree diameter measurement module based on 3D information reconstruction was constructed to achieve automatic calculation of tree diameter. Experiments have shown that the mAP for identifying tree stumps in transmission channel clearing sites reaches 0.972, and the relative error of tree diameter measurement is less than 5%, verifying the effectiveness of the method proposed in this paper and providing technical support for intelligent engineering quantity accounting in transmission channel clearing sites.