关于视觉随机街景三维仿真建模研究

Research on Visual Random Street View 3D Simulation Modeling

在视觉随机街景的三维仿真地图服务中,要求具备高质量和高精度的图像。由于环境混合光照产生波动时,不同视觉环境下混合颜色与拍摄角度行成相互干扰因素,使得不同视觉随机拍摄下的图像关键信息受到干扰,传统算法只能根据同一种类视觉下的单帧图像建立街景仿真模型,当视觉种类不同时无法避免上述所说干扰因素造成的影响,导致建模精度低的问题。提出采用三维视觉的随机街景仿真模型的建立方法,首先标定街景图像中建筑物的消失点,然后利用手动的方法得到建筑物各个顶点的几何信息,最后利用旋转平移法获得世界坐标系下的街景仿真模型。由于每一种建筑物都可以看作几种几何体的组合形式,改进算法具体较好的适应性。仿真结果表明,改进算法能够提高随机街景仿真模型的精度。

Three -dimension （3D） simulation of visual random street view of has important application value in the map service. A simulation modeling method based on 3D visual random street view is presented, the vanishing point of buildings in street view images is calibrated, then the geometric information of every vertex of buildings is obtained manually, and the street view simulation model in world coordinates by using the rotating translational method. Because each kind of building can be seen as a combination of several kinds of geometry forms, the improved algorithm has good adaptability. The simulation results show that the proposed algorithm can improve the accuracy of random street scene simulation model.