Insulator Damage Identification Method Based on Lightweight Improved YOLOv8

Aiming at the problems of complex overlapping image background and large calculation amount of algorithm model in the detection of damaged insulator of power line, a new method of damaged insulator identification based on lightweight and improved YOLOv8 is proposed. An efficient multi-scale attention mechanism is introduced into the YOLOv8 backbone network, a lightweight convolutional module is used to replace the standard convolutional module, and a WIoU loss function is used to improve the original loss function. The experimental results show that compared with the original YOLOv8 algorithm, the average precision of insulator breakage detection based on the lightweight improvement is increased by 4.35%, and the number of model parameters is decreased by 23.53%, indicating that the algorithm has improved in terms of algorithm accuracy and lightweight improvement, which provides a possibility for the deployment of the algorithm on edge inspection equipment of power lines.