MRP-YOLO: An Enhanced Lightweight Network for Small Object Detection in Complex UAV Imagery

Object detection in Unmanned Aerial Vehicle (UAV) imagery plays a critical role in military reconnaissance and security monitoring. However, it is constrained by tiny object sizes, complex backgrounds, and limited computational resources on edge devices. To address these issues, this paper proposes MRP-YOLO, an enhanced lightweight detector based on YOLOv11s. The model integrates a C3k2-MSCB module to strengthen multi-scale feature extraction and an RCSOSA module to optimize contextual information fusion for dense targets. Additionally, a lightweight Partial Convolution-based Detection (PCD) head is constructed to significantly reduce computational complexity. Experimental results on the VisDrone2019 dataset demonstrate that MRP-YOLO achieves 43.7% mAP@0.5 (surpassing the baseline by 5.1%), while achieving 204.5 FPS with a model size of only 8.5 MB. These metrics confirm that MRP-YOLO provides strong technical support for intelligent surveillance with a superior trade-off between accuracy and efficiency.