A Multi-Module Enhanced YOLOV8 for Lightweight and High-Precision Insulator Defect Detection in Edge Computing Environments

With the increasing application of edge computing devices, such as unmanned aerial vehicles (UAVs), in transmission line inspection, the demand for lightweight and high-precision target detection models has become increasingly urgent. However, due to the constraints of computing resources, most of the current target detection models can hardly be directly deployed on the hardware of UAV, and lightweight models often suffer from insufficient accuracy to meet practical inspection requirements. In this article, we put forward an enhanced lightweight defect - detecting model for YOLOv8 insulators. The GhostConv module is incorporated into the YOLOv8 network by the model. This helps in compressing the model's size, minimizing redundant parameters, and streamlining the network. Additionally, the RepViTBlock module is introduced to capture spatial information in images more effectively, enhancing the comprehensiveness and accuracy of the detection. In addition, the Generalized Attention Module (GAM) is installed within the main network structure. The intention is to heighten the network's capacity for extracting features of foreign items and cut down on non - essential details. The experimental findings indicate that the enhanced YOLOv8 algorithm exhibits excellent performance on the insulator fault detection dataset, attaining a recognition precision of <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\mathbf{9 2. 3 \%}$</tex>, which surpasses the performance of existing mainstream lightweight methods. The architecture reduces memory usage to <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\mathbf{4. 8 ~ M i B}$</tex> and computational complexity to <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$9.2 \times 10^{9}$</tex> Floating Point Operations (FLOPs), respectively, achieving both model lightweighting and high accuracy, thus meeting the stringent deployment requirements for transmission line insulator inspection.