<p>Small target detection in aerial imagery is critical for applications such as traffic flow monitoring and pedestrian behavior analysis. However, the challenges of feature loss in small targets and complex environmental conditions often hinder performance in such scenarios. As a solution, multimodal fusion methods, which leverage complementary information from different modalities, have demonstrated significant advantages in small target detection. Event cameras, as a novel bio-inspired motion vision sensor, effectively compensate for the limitations of traditional RGB images in capturing target contour information and adapting to complex environments. This paper proposes a novel multimodal fusion network, ARCAF-YOLO, to integrate event data with RGB data deeply. To effectively extract complementary information, this paper proposes an early fusion module (ARCAF), achieving deep feature integration of RGB and event data before the backbone network. In this module, an efficient RFCBAMConv module is introduced, which enhances feature extraction through channel and spatial attention mechanisms. Subsequently, the CSM module adaptively selects dominant modality features. Experimental results demonstrate that ARCAF-YOLO significantly improves small target detection accuracy while maintaining computational efficiency. On the public VisDrone dataset, ARCAF-YOLO achieves a mAP0.5 of 29.2%, surpassing YOLOv11s by 4%. It exhibits particularly strong performance in detecting small targets such as pedestrians and bicycles in nighttime scenes and complex environments, validating the success of event modality fusion.</p>