Research on Insulator Self-Explosion Defect Detection Method Based on Improved YOLOv8 Algorithm

Aiming to improve the detection accuracy of insulators, a new algorithm based on YOLOv8 is proposed. This aims to improve the detection performance of insulators and their spontaneous detonation defect in high-voltage transmission lines to ensure that insulator damage can be identified and located during daily inspections to prevent safety hazards and power interruption. Specific improvements have been made to the model, which are outlined below. In the first instance, a new small target detection head has been added to the original structure of the YOLOv8 model. The goal of this modification is to improve the network’s capability to detect small-sized targets. In the second case, Deformable ConvNets v2 (DCNv2) replaces the standard convolution in the C2f module. The purpose of this replacement is to improve the model’s ability to capture and perceive target features; Introducing Inverted Residual Mobile Block (iRMB) in the backbone network enables the model to better capture local features and complex spatial relationships, further focusing on target feature information; The Weighted Inverse Orthogonal Unconstrained (WIOU) function is employed to optimize the training process and improve detection accuracy. The results demonstrate that the enhanced algorithm outperforms the original algorithm, achieving a higher mAP on the transmission line insulator dataset. The improved algorithm is effective for detecting small defects and is ready for practical use.