Collaboratively modeling buildings from large-scale street-side and remote sensing images

This paper proposed an approach to rapidly generate 3D models for street-side buildings from street side images, remote sensing images, and GPS information. Traditional modeling methods usually require a single user to fulfill all processes of 3D modeling, and lack efficient management on massive data. Thus, they are unsuitable for large-scale modeling. In order to overcome those drawbacks, we employ the pipeline idea. The modeling processes are divided into several steps, which are further assembled to form a pipeline elaborately. We allocate computing resources into each step according to their requirements; on the other side, human resources are deployed into the pipeline to achieve team-work modeling. To demonstrate the effectiveness of our approach, we developed a distributed prototype system for rapid large scale urban 3D modeling. This system was tested to build an urban scene with more than 500, 000 street-side images. The results show great improvement on modeling efficiency compared with traditional methods, and it reveals a great prospect in applications.