The long-term operation of transmission lines under the influence of natural environments leads to various defects. In order to enhance the precision and efficiency of defect detection, this study proposes an improved strategy based on YOLOv8 and Variability Convolution. Due to the significant disparity in the quantity of various defect samples, resulting in a long-tail distribution problem, we suggest employing the SAM model to augment the data to enhance data balance. Within the Backbone, the introduction of DCNv2 dynamically adjusts the shape of convolution kernels to adapt to the features contained in various samples, thus enhancing generalization capability. Moreover, the integration of the Multi-CA attention mechanism guides the network to focus on the fusion information of each channel. Wise-IoU is utilized to guide model learning, enabling adaptation to anchor boxes of different qualities. Comparative analysis with other algorithms demonstrates an increase in detection accuracy of transmission line defects achieved by the proposed improved algorithm.