Lightweight insulator defect detection algorithm based on improved YOLOv8

In transmission lines, insulators are one of the important components. Due to long-term exposure of insulators to harsh environments, insulators on transmission lines are prone to faults such as flashover and damage, which can lead to a decrease in their insulation performance and seriously affect the safe and stable operation of the power grid. Drone inspection has become the mainstream way of inspecting power transmission lines, and the detection of insulator defects is an important part of drone inspection. Therefore, a lightweight insulator defect detection algorithm based on improved YOLOv8 is proposed. Firstly, use lightweight Ghost convolution instead of regular convolution; Then, use a small object detection head to enhance the model's ability to detect small targets and improve accuracy; Next, introduce GAM (Global Attention Mechanism) to improve the performance of deep neural networks. Finally, use the loss function SIoU to optimize the algorithm and improve model performance. The experimental results show that the average accuracy of insulator detection has increased by 1.2 percentage points, and the model size has decreased by 29.3%. The improved algorithm model not only improves detection accuracy but also becomes lighter, which can achieve rapid detection of insulator defects.