Detection of tilted electricity poles using image processing and computer vision techniques

Date

2023-9-04

Author

Karaali, Ediz

Metadata

Show full item record

This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License.

Item Usage Stats

376
views

140
downloads

Cite This

Regular maintenance is essential in efficiently transporting electricity from one point to another to ensure an uninterrupted energy supply, and risk assessments are imperative. Since human personnel predominantly carry out such tasks, demanding a considerable workforce and causing errors, autonomous and efficient methods for effectively surveying electricity poles to detect possible anomalies are required. This study presents an autonomous risk classification model to address an anomaly type: insecurely tilted electricity poles. During this study, extensive data, 8775 electricity pole images, was gathered from various regions by Unmanned Aerial Vehicles (UAVs). An object detector, Faster R-CNN, within the proposed model was integrated to precisely identify different types of electricity poles: steel, concrete, and wooden. Subsequently, image processing techniques are deployed to determine pole tilt angles accurately. Based on these angles, electricity poles are classified as "risky" or "not risky," indicating their structural integrity and stability. The test results of the proposed model reveal the following two significant outcomes: (a) Electricity poles can be detected with a mean average precision rate of 94.40%, and (b) Representative lines for the poles and their respective tilt angles can be measured with an accuracy of 95.95%, which demonstrates an exceptional performance. This model holds substantial promise for practically detecting risks associated with tilted electricity poles and can significantly contribute to ensuring the reliability and safety of power distribution systems.

Subject Keywords

Tilted Pole Risk Detection, Deep Learning, Image Processing, Unmanned Aerial Vehicles, Object Detector

URI

https://hdl.handle.net/11511/105553

Collections

Graduate School of Natural and Applied Sciences, Thesis