Rapidly Street-side Buildings Modeling System From RMS, GPS And VIDEO

As the development of urbanization, digital city and city plan play a more and more role in the development of city. Urban modeling is the base of 3D digital city system. Large-scale urban modeling often requires a lot of resources including manpower, material and funds. So it's a major challenge of modeling that how to build models with geometry and texture high efficient, large-scale and low cost.The first problem that urban modeling faces is the acquisition of large-scale street-side images(VIDEO). The acquisition velocity of exist acquisition system is enough for the demand, but the quality of image need further improvement, especially the color consistency of image. In this paper we improve the color consistency of image.In this paper we present a new collaborative modeling approach which is able to produce 3D building models from remote sensing images(RMS),GPS and street-side images(VIDEO) by computer vision techniques. The approach adopts the idea of pipelining manufacture. It divides the modeling process into several steps, which consists of a pipeline. Some steps can be done automatically, the others require human intervention. Furthermore, the computing resources are allocated to these steps according to their requirements. That is, some steps are performed on single workstation, while the others are done by a cluster of workstations. Base on this idea, we develop a prototype distributed system for modeling large-scale urban scene by team. In this system, the modeling workload is balanced along the pipeline, and the waiting time between steps is greatly reduced. Moreover, the human intervention that the system needs is simple and the system's learning curve is short for ordinary users.