The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences
Download
Publications Copernicus
Download
Citation
Articles | Volume XLVI-4/W3-2021
Int. Arch. Photogramm. Remote Sens. Spatial Inf. Sci., XLVI-4/W3-2021, 239–246, 2022
https://doi.org/10.5194/isprs-archives-XLVI-4-W3-2021-239-2022
Int. Arch. Photogramm. Remote Sens. Spatial Inf. Sci., XLVI-4/W3-2021, 239–246, 2022
https://doi.org/10.5194/isprs-archives-XLVI-4-W3-2021-239-2022

  11 Jan 2022

11 Jan 2022

MACHINE LEARNING APPROACH FOR TENAGA NASIONAL BERHAD (TNB) OVERHEAD POWERLINE AND ELECTRICITY POLE INVENTORY USING MOBILE LASER SCANNING DATA

M. S. A. Mohd Rapheal1, A. Farhana2, M. R. Mohd Salleh2, M. Z. Abd Rahman2, Z. Majid2, I. A. Musliman2, A. F. Abdullah3, and Z. Abd Latif4 M. S. A. Mohd Rapheal et al.
  • 1Project GIS, Distribution Network, Tenaga Nasional Berhad, Universiti Tenaga Nasional, Kajang, Selangor, Malaysia
  • 2Department of Geoinformation, Faculty of Built Environment and Surveying, Universiti Teknologi Malaysia, Johor, Malaysia
  • 3Department of Biological and Agricultural Engineering, Faculty of Engineering, Universiti Putra Malaysia, 43400, Serdang, Malaysia
  • 4Faculty of Architecture, Planning and Surveying, Universiti Teknologi MARA, Selangor, Malaysia

Keywords: Mobile Laser Scanner (MLS), Electricity Assets, Machine Learning, Overhead Powerline, Electricity Poles

Abstract. Electricity assets recognition and inventory is a fundamental task in the geospatial-based electrical power distribution management. In Malaysia, Tenaga Nasional Berhad (TNB) aims to complete their assets inventory throughout the country by 2022. Previous research has shown that a method for assets detection especially for TNB is still at an early stage, which mainly relied on manual extraction of the assets from different data sources including mobile laser scanner (MLS). This research aims at evaluating a geospatial method based on machine learning to classify the TNB assets using high density MLS data. The MLS data was collected using Riegl VMQ-1 HA scanner and supported by the base station and control points for point cloud registration purpose. In the first stage the point clouds were classified into ground and non-ground objects. The non-ground points were further classified into different landcover types i.e. vegetation, building, and other classes. The points classified as other classes were used for overhead powerline and electricity poles classification using random forest-based Machine Learning (ML) approach in LiDAR 360 software. Based on the classified point clouds, detailed characteristics of electricity poles (i.e. number of poles, height, diameter and inclination from ground) and overhead powerlines (number of cable segments) were estimated. This information was validated using field collected reference data. The results show that the detection accuracy for electricity poles and overhead power line are 65% and 63% respectively. The estimation of length, diameter and height of the spun pole from point clouds has produced Root Mean Square Error (RMSE) value of 0.081cm, 0.263 cm and 0.372 cm respectively. Meanwhile for the concrete pole, the length, diameter and height has been successfully estimated with the value of RMSE of 0.034 cm, 0.029 cm and 0.331 cm respectively. The length of overhead powerline was estimated with 59.02 cm RMSE. In conclusion, the MLS data had show promising results for a semi-automatic detection and characterization of TNB overhead powerlines and poles in the sub-urban area. Such outcome can be used to support the inventory and maintenance process of the TNB assets.