Real-Time 3D Road Scene Based on Virtual-Real Fusion Method

Abstract

Road monitoring plays a vital role in emergency disposal, traffic accident liability determination, traffic trend analyzes, and other fields. This paper designs a real-time virtual-real fusion framework for multiview surveillance of large-scale scene, in which 2D panorama, satellite texture, and 3D model are merged. The system is implemented with the following contributions. First, a novel automatic mosaic algorithm for multiview overlapped images is proposed. Feature points between adjacent cameras are detected via comotion statistics map. In addition, transformation model is obtained using random sample consensus algorithm. Second, multiview images are morphed to a same overhead viewpoint. Finally, 2D panoramic road image and 3D scene model are combined based on high-precision ground control points. The proposed framework has been successfully applied to a large road intersection. With the virtual-real fusion approach, observers can freely monitor and ramble in the large-scale 3-D road scene.