An Enhanced-RANSAC-Fusion and Kd-Tree-Filtering Pipeline Method for Multiscale 3-D Substation Components Extraction

The construction, operation, and maintenance of smart grids typically rely on high-precision 3-D models of substations, for which 3-D scanning and object extraction based on laser point cloud technology offer a feasible approach to accurate modeling in complex environments. Due to the structural complexity and occlusion in substations, the extraction of centimeter- to meter-scale components from large-scale laser point cloud data remains challenging. Representative components at different scales are selected in this study, with insulators at the centimeter scale and crossarms at the meter scale. An enhanced-RANSAC-fusion and Kd-Tree-filtering pipeline method (ERKF) is proposed to address the limitations of traditional random sample consensus (RANSAC), which often fails to capture fine-scale structures due to geometric mismatch. In the proposed method, pass-through filter and point cloud clipping were adopted to reduce background noise and preserve critical details, and then the processed data with the initial RANSAC output are merged to recover accessory components. Next, a Kd-Tree is used to search for and remove nontarget points. This enables a more complete extraction of substation components from the 3-D point cloud. Experimental validation demonstrates that in substation environments, the accuracy precision of insulator and crossarm extraction improves by approximately 24.5% compared with the traditional RANSAC algorithm, providing substation monitoring personnel with an intuitive method for equipment fault diagnosis and defect identification.