Abstract: In 2023, the proportion of electricity consumption in the country's final energy consumption will reach 28%. Cables are used as a key component in the transmission of electrical energy. However, in high-altitude environments, the surface layer of the cable is susceptible to environmental erosion, and it is particularly important to carry out efficient detection of the cable in time. At present, the mainstream detection uses unmanned aerial vehicles (UAVs) to quickly obtain high-altitude images and transmit them to the network model for detection. The YOLO algorithm has been widely used in UAV missions because of its efficient detection ability. However, the surface defects of high-altitude cables are small and the image quality of bad weather shooting is low, which leads to the reduction of the accuracy and efficiency of UAV inspection. Therefore, a cable fault detection model based on improved YOLOv9, YOLOv9-SED, was proposed. Firstly, the dehazing network UnfogNet was added to the original YOLOv9 model to effectively enhance the image clarity of the model in the complex and harsh environment at high altitude. At the same time, the SEAM attention mechanism and the Shape-IoU loss function are introduced to improve the model's feature extraction ability for small targets. Finally, the DualConv convolution layer is used to replace the original Conv convolutional layer, which can enhance the performance of the model and reduce the complexity of the model. Compared with the original YOLOv9 model, the accuracy is increased by 1.7%, the average accuracy is increased by 3.5%, the model weight is reduced by 13MB, and the GFLOPS is reduced by 16 units, which provides a new scheme for the detection of cable defects in harsh high-altitude environments.