Improving Insulators Detection Accuracy via Image Enhancement Techniques: Case of Indigenous Aerial Image Dataset

• Published in: IEEE access Vol. 12; pp. 145582 - 145589
• Main Authors: Muhammad Jiskani, Shafi, Hussain, Tanweer, Ali Sahito, Anwar, Shaikh, Faheemullah, Kumar, Laveet
• Format: Journal Article
• Language: English
• Published: Piscataway IEEE 2024; The Institute of Electrical and Electronics Engineers, Inc. (IEEE)
• Subjects: Accuracy; Algorithms; Autonomous aerial vehicles; Condition monitoring; contrast limited adaptive histogram equalization (CLAHE); Datasets; Feature extraction; Histograms; hue saturation value (HSV) color space; Image color analysis; Image enhancement; Image segmentation; Image transmission; indigenous dataset; insulator detection; Insulators; Lighting; Machine learning; Monitoring; NTDCL Pakistan; Power transmission lines; Transmission lines; Unmanned aerial vehicles; YOLO; you only look once (YOLO)
• DOI: 10.1109/ACCESS.2024.3474255
• ISSN: 2169-3536

The challenging task of insulator monitoring through aerial images is addressed in high voltage transmission network and highlights the limitations of traditional human patrolling with emphasize on utilization of unmanned aerial vehicles UAVs utilizing machine learning algorithms. This research has been accomplished by creating indigenous dataset of 500kV transmission network of National Transmission and Despatch Center Limited (NTDCL). 608 original images were captured in diverse lighting and topographical conditions which were then augmented to 3618 images. To improve the detection accuracy of YOLOv8s algorithm in aerial images, HSV and CLAHE image enhancement techniques were employed to improve the visual feature of insulator with suppressed noise. YOLOv8s algorithm with image enhancement has improved detection accuracy from 88% to 95% demonstrating the effectiveness of integrating image enhancement technique for insulator monitoring, offering promising improvement in maintenance practices and operational reliability of transmission lines.