<p>Traditional manual inspection methods are fraught with limitations, including delayed information transmission, low detection efficiency, high costs, and potential safety risks. Although UAV (Unmanned Aerial Vehicle) technology and deep learning algorithms have made significant progress in the field of power transmission line inspection, most existing deep learning models are designed to detect only a single type of fault, making them ill-equipped to handle the complexities of multiple concurrent faults. To address this limitation, we propose the SPDFYOLO-SPD (Sobel Pooling Dense Fusion YOLO with Space-to-Depth Convolution) algorithm, which is built upon the YOLOv8 framework. This algorithm incorporates an innovative SPDF Block, a module designed to integrate features from both visible light and infrared images, thereby leveraging the complementary strengths of the two modalities to significantly enhance multi-fault detection capabilities. Additionally, the introduction of the SPD-Conv module further improves the model’s ability to detect small objects against complex backgrounds, enabling it to perform exceptionally well in multi-fault detection tasks. Experimental results demonstrate that, compared to baseline algorithms, SPDFYOLO-SPD achieves a substantial improvement in detection performance with only a minimal increase in additional parameters. The mAP50 metric is significantly enhanced by 7.5 to 12.1 percentage points. Moreover, the proposed algorithm outperforms the state-of-the-art real-time object detection framework YOLOv12 in terms of both accuracy and model size, with an mAP50 metric that exceeds by 9.8 to 13.5 percentage points, demonstrating its remarkable performance. The successful development of the SPDFYOLO-SPD algorithm provides an efficient and accurate solution for power transmission line fault detection, laying a solid foundation for the intelligent operation and maintenance of the power industry.</p>