Automated Inventory of Electrical Distribution Assets Based on Image Recognition and Ground LiDAR

This paper introduces a deep learning-based method for automatically locating and cataloging electrical distribution network assets that combine 2D image analysis with LiDAR data. The proposed method has five steps: neural network identification of poles and assets; pole location estimation via the intersection of lines of bearing; LiDAR cloud extraction and classification; pole presence validation; and electrical grid reconstruction through pole interconnections. The approach enhances asset management accuracy and improves maintenance and network planning. Results show successful mapping of 83% of electrical distribution lines and accurate location of 83% of utility poles in urban, rural, and forest environments combined without requiring human intervention.