Defect Classification of Medium-Voltage Overhead Line Conductors Using YOLO11

Defects in medium-voltage overhead line conductors may adversely affect the reliability and continuity of electrical distribution networks, highlighting the importance of effective automated inspection methods. This study investigates whether region-of-interest cropping improves the classification of overhead conductor conditions compared with conventional full-image input strategies. A two-branch classification framework based on the YOLO11 architecture was constructed using a publicly available three-class image dataset representing medium-voltage overhead line conductors under normal, surface-defect, and strand-fraying conditions. The dataset was partitioned into 1,440 training, 180 validation, and 180 test images. While the full-image branch classified the original images directly, the ROI branch employed ground-truth bounding-box annotations to extract individual conductor regions, yielding 323 object-level samples in the held-out test split. YOLO11n-cls, YOLO11s-cls, and YOLO11m-cls were evaluated using accuracy together with macro-averaged precision, recall, and F1-score. Full-image classification yielded accuracies of 94.44%, 93.89%, and 93.89%, with corresponding macro F1-scores of 93.91%, 93.92%, and 93.48%, respectively. ROI-based classification increased the corresponding accuracies to 98.76%, 97.21%, and 97.52% and the macro F1-scores to 98.56%, 96.87%, and 97.40%, representing accuracy differences of +4.32, +3.32, and +3.63 percentage points, respectively. YOLO11n-cls achieved the highest ROI-based performance, indicating that object-focused preprocessing can reduce irrelevant background information and enable lightweight architecture to achieve high classification accuracy. Because the two branches operated on different evaluation units whole images and object-level crops the comparison is interpreted as a strategy-level analysis rather than a direct paired evaluation.

Authors

Institutions

Publication Details

Journal
Electrical Engineering and Energy (ELENE)
Published
2026-11-19
DOI
https://doi.org/10.64470/elene.2026.50
Primary Topic
Power Line Inspection Robots
Type
article
Field-Weighted Citation Impact
0.00
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
article

Defect Classification of Medium-Voltage Overhead Line Conductors Using YOLO11

Mücahid Mustafa Sarıtaş, Murat Köklü, Oya Kılcı
Electrical Engineering and Energy (ELENE)
Power Line Inspection Robots
article

Defect Classification of Medium-Voltage Overhead Line Conductors Using YOLO11

Mücahid Mustafa Sarıtaş, Murat Köklü, Oya Kılcı
article en

Abstract

Defects in medium-voltage overhead line conductors may adversely affect the reliability and continuity of electrical distribution networks, highlighting the importance of effective automated inspection methods. This study investigates whether region-of-interest cropping improves the classification of overhead conductor conditions compared with conventional full-image input strategies. A two-branch classification framework based on the YOLO11 architecture was constructed using a publicly available three-class image dataset representing medium-voltage overhead line conductors under normal, surface-defect, and strand-fraying conditions. The dataset was partitioned into 1,440 training, 180 validation, and 180 test images. While the full-image branch classified the original images directly, the ROI branch employed ground-truth bounding-box annotations to extract individual conductor regions, yielding 323 object-level samples in the held-out test split. YOLO11n-cls, YOLO11s-cls, and YOLO11m-cls were evaluated using accuracy together with macro-averaged precision, recall, and F1-score. Full-image classification yielded accuracies of 94.44%, 93.89%, and 93.89%, with corresponding macro F1-scores of 93.91%, 93.92%, and 93.48%, respectively. ROI-based classification increased the corresponding accuracies to 98.76%, 97.21%, and 97.52% and the macro F1-scores to 98.56%, 96.87%, and 97.40%, representing accuracy differences of +4.32, +3.32, and +3.63 percentage points, respectively. YOLO11n-cls achieved the highest ROI-based performance, indicating that object-focused preprocessing can reduce irrelevant background information and enable lightweight architecture to achieve high classification accuracy. Because the two branches operated on different evaluation units whole images and object-level crops the comparison is interpreted as a strategy-level analysis rather than a direct paired evaluation.

Electrical Engineering and Energy (ELENE)Vol. 5(3)
Selçuk University (TR), KTO Karatay University (TR)
Power Line Inspection Robots
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon