Surveyors tap into machine learning to explore faster ways to map the world’s power lines

Overview

A Durham University KTP has helped NM Group to explore how machine learning can speed up the painstaking process involved in mapping power lines from the air.

The partnership has helped the company make steps towards automating the processing of huge swathes of data gathered by aerial scanners and cameras when assessing the safety and performance of power networks.

NM Group is now at the forefront of its sector’s race to unlock the full potential of emerging machine learning technology – a move that promises to reduce current processing times and revolutionise surveying capabilities.

The challenge

With the worldwide demand for electricity growing day by day, utility companies are under increasing pressure to maintain supply, safety and reliability.

Their need to survey their vast networks of overhead lines is crucial to maintaining statutory distances between vegetation, ground and other potential hazards.

High resolution LiDAR data is one of the best remote sensing methods for identifying power wires and structures.

NM Group has grown a global business using this remote sensing technology to create highly accurate 3D representations of these vast and complex networks. Power companies then use these models to manage the encroachment of vegetation and prioritise resources.

This process currently requires a huge manual effort to identify every individual feature that is plotted on a 3D point cloud graph after being photographed – including roads, trees and power lines.

The timescales involved create a major challenge to the entire industry, especially in areas prone to bush fires in Australia and the US where huge areas of vegetation need to be surveyed before hazardous heat of the summer.

The solution

NM Group identified the opportunity to harness the power of machine learning to automate the processing of data gathered in the surveying of power networks.

Having worked with Durham University academics on two previous and highly successful KTPs, the company turned to the institution again to explore the possibilities of speeding up results.

A research associate with significant knowledge of machine learning and NM Group’s sector was paired with the company to provide a bridge between the academic and commercial worlds that the KTP straddled.

They focused on exploring the use of deep neural networks for 3D point clouds – a cutting edge area of research that to date has been relatively untouched by LiDAR yet presents exciting opportunities to revolutionise the industry.

Significant progress was made towards achieving NM Group’s ambitions, with the company gaining an in-depth understanding of the viability of its ambitions. Foundational tools were also created to automatically model conductors.

Value

This remains an active area of research for NM Group and an exciting area of future development.

The company is now well informed about the possibilities of automation and can monitor the market for other potential partners with similar ambitions. The business has the knowledge to engage confidently with potential collaborators and to evaluate their work.

The team also continues to explore opportunities to continue the project internally.