MCMS-YOLO: A Lightweight Small Object Detection Network for Unmanned Aerial Vehicles

Unmanned Aerial Vehicles (UAVs) face inherent challenges in autonomous object detection, such as complex backgrounds, small targets, and frequent occlusions. Moreover, the limited computational power and onboard memory of UAV processors further constrain the performance of traditional detection algorithms. To address these issues, this paper proposes a lightweight algorithm based on YOLOv11. The proposed model introduces a novel multi-scale feature fusion mechanism to alleviate feature degradation during feature propagation and enhance cross-scale interaction capabilities. In addition, a lightweight multidimensional collaborative self-attention mechanism is designed to improve the algorithm’s perception of small targets. Comparative experiments conducted on the VisDrone2019 dataset validate the effectiveness and superiority of the proposed method. Compared with the baseline YOLOv11, the model demonstrates stronger robustness and generalization capability. Furthermore, when compared with other state-of-the-art lightweight detectors, MCMS-YOLO achieves the best balance between detection accuracy and computational efficiency, providing a highly practical and effective solution for UAV-based object detection tasks.