Publication: Development of AI Vision Inspection System for UAV imagery Surveillance of Transmission Towers
Date
2024
Authors
Hussein S.Y.S.H.
Dziyauddin R.A.
Muhtazaruddin M.N.
Yaghoobi Y.J.
Din N.M.
Journal Title
Journal ISSN
Volume Title
Publisher
Institute of Electrical and Electronics Engineers Inc.
Abstract
High-voltage transmission line networks are essential for electricity delivery and require proactive maintenance to prevent breakdowns due to steady demand and full-capacity operation. Federal and state regulations require annual inspections of the right-of-way (ROW) and transmission infrastructure. Traditionally, helicopters are used for these inspections, but they are costly. Another issue is that power outages and financial losses are still caused by vegetation encroachment in power line corridors. To address this, we propose an automated method combining robotics, photogrammetry, and computer vision. Unmanned aerial vehicles (UAVs) present a viable monitoring alternative because of their speedy and cost-effective high-resolution image capturing. However, segmenting vegetation encroachment in these images is difficult due to the complexity and pixel imbalance. We propose a deep learning-based approach to address these challenges by dividing the dataset into 3 classes: power line, background, and vegetation. Using the UAV- VEPL-NET dataset and a Convolutional Neural Network (CNN) model, initial training will be conducted with the help of Roboflow software. This approach will help the system detect with high accuracy. The final model will then be embedded into the UAV for testing. Overall, the average precision of the demo training in Roboflow was 93%, whereas the accuracy for every class was 93% for the power line, 89% for the background, and 96% for the vegetation. ? 2024 IEEE.
Description
Keywords
Convolutional neural networks , Deep neural networks , Failure analysis , HVDC power transmission , Inspection equipment , Outages , Power distribution lines , Power transmission lines , Risk management , Unmanned aerial vehicles (UAV) , Aerial vehicle , Convolutional neural network , Inspection system , Power lines , Right of way , Rights-of-way , Roboflow , Transmission inspection system , Unmanned aerial vehicle , Vision inspection systems