The frequent illegal use of civilian low-altitude small unmanned aerial vehicles (UAV) poses a serious threat to the harmonious development of the country and society. In response to the problems of missed detection, false detection, and low detection accuracy in small target detection of UAV based on visible light, a YOLO-SCConv ATT (YOLO-SCAT) algorithm model was proposed to reconstruct the ELAN (efficient layer aggregation networks) structure and reduce redundant spatial and channel features. At the same time, the attention mechanism CBAM (convolutional block attention module) was introduced to enhance feature extraction in spatial and channel dimensions during training, thereby improving the average detection accuracy of the model. The experimental results show that the accuracy, recall, F₁ score, mAP@0.5 and mAP@0.5:0.95 can reach 94.4%, 94.4%, 94.4%, 94.7% and 52.9%, respectively, which proves that the YOLO-SCAT model improves the detection and recognition ability of small targets in complex visible light scenes, and can better meet the practical needs of anti-drone systems.