Object Detection and Contour Extraction of Transmission Towers Based on Computer Vision

Transmission towers play a pivotal role in the power system, responsible for transmitting electricity from power plants directly to end users. Ensuring their safety and reliability is of paramount importance. Given the vast distribution and intricate structures of transmission towers, conventional manual inspections face multiple challenges in meeting the real-world requirements. Utilizing machine vision technology, this study aims to achieve target detection and contour extraction of transmission towers. Images of transmission towers were captured across different scenarios, culminating in the construction of a dataset. Leveraging the YOLOv8s algorithm, we developed a model for power tower detection. Experimental results reveal that the model achieves a map0.5 accuracy of 0.98 and a map0.5:0.95 accuracy of 0.821. Recognized frames from the transmission towers were extracted and subsequently processed using the Canny edge detection algorithm. The experimental outcomes demonstrate the efficacy of our proposed algorithm in power tower detection and contour extraction.