Advanced YOLO-based Real-time Power Line Detection for Vegetation Management

Shuaiang Rong, Lina He, Salih Furkan Atici, Ahmet Enis Cetin · 2025

Read Advanced YOLO-based Real-time Power Line Detection for Vegetation Management by Shuaiang Rong, Lina He, Salih Furkan Atici, Ahmet Enis Cetin (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2503-00044.txt.

Introduction

Shuaiang Rong and three co-authors aim drones at a dull but dangerous problem: plants creeping toward power lines. Their work appeared in twenty twenty-five. Vegetation touching lines sparks outages and fires, and patrol imagery piles up faster than humans can scan.

Their detector builds on the YOLOv8 vision model, famed for speed. Directional filters sharpen the thin, angled look of wires and nearby growth. Output boxes rotate to hug the lines, tightening localization. A post-processing step turns the boxes into a numeric encroachment score for each span. On a standard power line data set, the framework proved accurate enough for live monitoring. Utilities get a ranked list of spans to trim first.

Catalog record

CREATORShuaiang Rong, Lina He, Salih Furkan Atici, Ahmet Enis Cetin
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE49 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2503-00044

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Advanced YOLO-based Real-time Power Line Detection for Vegetation Management Shuaiang Rong, Lina He, Salih Furkan Atici, Ahmet Enis Cetin Abstract Power line infrastructure is a key component of the power system, and it is rapidly expanding to meet growing energy demands. Vegetation encroachment is a significant threat to the safe operation of power lines, requiring reliable and timely management to enhance the resilience and reliability of the power network. Integrating smart grid technology, especially Unmanned Aerial Vehicles (UAVs), provides substantial potential to revolutionize the management of extensive power line networks with advanced imaging techniques. However, processing the vast quantity of images captured by UAV patrols remains a significant challenge. This paper introduces an intelligent real-time monitoring framework for detecting power lines and adjacent vegetation. It is developed based on the deep-learning Convolutional Neural Network (CNN), You Only Look Once (YOLO), renowned for its high-speed object detection capabilities. Unlike existing deep learning-based methods, this framework enhances accuracy by integrating YOLOv8 with directional filters. They can extract directional features and textures of power lines and their vicinity, generating Oriented Bounding Boxes (OBB) for more precise localization. Additionally, a post-processing algorithm is developed to create a vegetation encroachment metric for power lines, allowing for a quantitative assessment of the surrounding vegetation distribution. The effectiveness of the proposed framework is demonstrated using a widely used power line dataset. PubID: pubid: 0000–0000/00$00.00 © 2021 IEEE I Introduction Power lines can produce sparks upon tree contact due to factors such as line sag, tree growth, and severe weather conditions during natural hazards. Tree-contacted power line sparking has been identified as a significant cause of the most destructive wildfires in California in recent years [ 1 ] , as well as the devastating Maui fire in 2023 [ 2 ] . The American power grid has more than 500,000 miles of high-voltage transmission lines and 5 million miles of distribution lines, and it is still rapidly growing with decarbonization initiatives [ 3 ] . Inspecting such a large infrastructure is both labor-intensive and time-consuming, relying on traditional methods like manual patrols and helicopter surveys that have been in use for decades. Ground patrols allow thorough evaluations but are slow and costly, while helicopter surveys cover more areas quickly but with lower detection accuracy. Both approaches rely heavily on human visual observation, increasing the risk of missed or incorrect assessments [ 4 ] . I-A Vegetation Monitoring for Power Line Inspections Satellite imagery has been studied for vegetation monitoring within power line corridors. Sensors like Synthetic Aperture Radars (SAR) [ 5 ] , operating in the microwave region of the electromagnetic spectrum, and optical sensors [ 6 ] , capturing multispectral data in the visible and near-infrared regions, have been widely explored. Some studies [ 7 , 8 , 9 ] have used multispectral stereo pairs of satellite images to identify trees with extracted Normalized Difference Vegetation Index (NDVI) information to evaluate encroachment to power lines; however, capturing stereo images is technically challenging and costly for large-scale monitoring [ 10 ] . Recent studies focused on monocular satellite imagery combined with a machine learning technique, i.e., Adaboost classifier [ 11 ] and Support Vector Machine (SVM) [ 12 ] for extracting, segmenting, and classifying trees. However, these methods are constrained by data availability in low-light or cloudy conditions. Additionally, in these studies, the positions of power lines are typically predetermined, as they struggle to automatically detect power line conductors and bundles. Even the highest-resolution imagery is often too coarse to capture the thin and small features of power lines, particularly distribution lines, which are smaller, positioned lower, and frequently obscured by trees, making them unsuitable for fine-scale detection. Fixed vision sensors [ 13 ] and ground-based vehicles [ 14 ] have been explored for higher-resolution data acquisition, but their limited view angles present challenges. In contrast, airborne Light Detection and Ranging (LiDAR) three-dimensional (3D) modeling with higher flexibility, has gained greater attention in the industry, typically employing fixed-wing aircraft or helicopters. Utilities such as Pacific Gas and Electric (PG&E) and Duke Energy have utilized LiDAR to assess risks near power lines [ 15 , 16 ] . Existing studies [ 17 , 18 , 19 , 20 ] showed that power line clustering, as well as the extraction and classification of trees, can be achieved by analyzing the 3D point cloud data using line detection (e.g., Hough transform) [ 17 ] , intensity analysis [ 18 ] , and advanced Deep Neural Networks (DNNs) (e.g., CA-PointNet++ [ 19 ] , VEPL-Net [ 20 ] , and RandLA-Net [ 21 ] ). Despite its effectiveness, the broader use of airborne LiDAR is limited by heavy payloads, high power consumption, and the need for high point density to map small objects [ 22 ] . For instance, high-voltage transmission lines are easier to map due to elevated positions, whereas lower voltage distribution lines are often obscured by dense forests, making detection more challenging [ 18 ] . I-B Opportunities in UAV-Based Monitoring As an alternative, UAVs equipped with lightweight vision sensors offer a more cost-effective and practical solution. They can operate closer to the ground and capture high-resolution images in both the visible red, green, and blue (RGB) channels and infrared spectrums, providing detailed views of the environment [ 23 ] . This makes it particularly effective for monitoring both transmission and distribution lines. However, analyzing these captured images still relies on human observation, making it prone to errors and oversight. TABLE I: Overview of Different Data Resources and Techniques Data Resource Main Technique Power Line Detection Vegetation Detection Data Resolution Computational Cost Position Accuracy Communication Load Satellite SAR/ Optical Imagery Stereo matching & NDVI [ 7 , 8 , 9 ] No Yes Low Medium Low High Adaboost, SVM [ 11 , 12 ] No Yes Medium High Medium High Airborne LiDAR (fixed-wing/ helicopter) Hough Transform & Intensity analysis [ 16 , 17 ] Yes No Medium Medium Medium High 3D point cloud data classification with DNNs [ 18 , 19 , 20 ] Yes Yes Medium High High High UAV-Based Cameras Image processing (Canny, Hough Transform, etc.) [ 28 , 29 ] Yes No High Low Low High CNN-based classification [ 31 , 32 ] Yes No High Medium Low High UNet-based segmentation [ 33 , 34 , 35 ] Yes No High High Medium Medium Proposed PL-YOLO Yes Yes High Medium High Low Utility companies are beginning to pilot advanced wireless technologies, such as private Long Term Evolution (LTE) networks, to support intelligent sensors [ 24 ] . However, it is still challenging to manage the communication and control of many inspection UAVs to support their transmission of a large volume of streaming image data. In addition, the critical control and non-payload data, including location and speed, demand robust communication that can be hindered by limitations in channel data rates. This can lead to data traffic congestion, potentially jeopardizing the safe navigation and operation of UAVs, which in turn poses risks to power line infrastructure and public safety. Therefore, there is a critical need to develop automatic and real-time detection methods to enhance the efficiency and applicability of UAV inspections. I-C Image-Based Detection of Power Lines Image-based

Read the complete text