VIU-YOLO: Vehicle-Infrastructure-UAV Cooperative Perception with the Multi-Head Self-Attention Detection Model

Multi-perspective cooperative perception is a promising approach to improving the security and accuracy of unmanned systems in intelligent transportation scenarios. However, existing methods predominantly focus on full-view object cooperative within a closed-loop system, struggling with the simultaneous association of objects from multiple perspectives in complex, dynamic environments. To address this, we proposed a Vehicle-Infrastructure-UAV feature fusion method for multi-perspective cooperative perception, named VIU-YOLO, which can effectively complement multiple viewpoint features for accurate and robust object detection. We validated our method on a publicly available air-ground cooperative perception dataset and the results demonstrate both high detection accuracy and low inference latency.