Drone-based power line inspection using AI - Vattenfall project

This project was part of Vattenfall’s contribution to Vinnova’s “Drones of the future” programme.

In this demo video we demonstrate the algorithms of powerline and high-risk-vegetation detection. The powerline detection problem is approached by using Qamcom's deep learning neural network platform. The high-risk-vegetation detection problem is approached using a Structure-from-Motion(SfM)-based geometric algorithm. The speed of the algorithms are displayed on the top-left.

The powerline detection algorithm extracts the pixel position of the powerline in the video frame. The high-risk-vegetation detection algorithm detects obstacles above the ground, which includes high trees but also the powerline poles. The latter can be removed from the data given the known GPS coordinates. The intensity of the color-mask on the detected obstacles indicates the geometric height of the object.

By combing the results of powerline detection and high-risk-vegetation detection, it determines the potential contact risk between powerline and obstacles. Once there is contact risk, the box defining the Region of Interest (ROI) changes color to red. Otherwise, it will remain yellow.

The algorithms have been deployed both on GPU server and on Nvidia embedded GPU Jetson Xavier. The intention is to install Xavier on drones (UAVs) for real-time inspection.

__________________________________________

Qamcom is a leading Swedish technology and research company with unique competence within hardware, software, and system development. Qamcom offers value-driven technology solutions, products, and services in the fields of advanced signal processing, artificial intelligence, wireless communications systems, industrial IoT and system safety. Qamcom’s mission is to turn technology into value for society, businesses, and people. Based on insights and focus on end users, Qamcom bridge the gap between technology and application for Automotive, Telecom, Space, MedTech, Heavy Industry and Smart Cities.

► Website: https://www.qamcom.com/

► LinkedIn:   / qamcom-research-and-technology  
► YouTube:    / @qamcom2862  
► Facebook:   / qamcom  
► Twitter:   / qamcom