Faced with the challenges of numerous insulators in power transmission lines, low inspection efficiency of UAVs, and instances of missed detections, this paper proposes a lightweight insulator detection algorithm based on an improved YOLOv5 to enhance inspection efficiency. The algorithm leverages the lightweight Ghost network to replace the original backbone network and uses the SimSPPF (Simplified Spatial Pyramid Pooling - Fast) module instead of the traditional SPP (Spatial Pyramid Pooling). This adaptation improves the model's flexibility to different input sizes, effectively reducing model parameters and computation, and lowering resource consumption. Furthermore, the algorithm introduces the combination of DSConv (Dynamic Snake Convolution) and the EMA (Efficient Multi-scale Attention Module) to enhance the capture capability of key insulator features, thereby improving target detection accuracy. Experimental results demonstrate that the proposed lightweight model achieves a mean average precision () of 99.67%. Compared to the traditional YOLOv5, the model reduces parameters by 37.5% while increasing the mean average precision by 2.97%. Additionally, compared to other existing algorithms, the proposed algorithm achieves a detection speed of 86.1 FPS, significantly enhancing insulator detection accuracy while maintaining a lightweight structure, making it suitable for resource-constrained application scenarios.