基于三维虚拟视觉的导视系统设计研究

Research on Sign System Design Based on 3D Virtual Vision

对曲面物体的角度和坐标的准确描述一直是导视系统设计的难点。传统的办法采用多角度叠加逼近的方式完成复杂曲面的设计,但是误差较大。提出采用三维虚拟视觉的导视系统设计方法。该方法先采集多组不同角度的原始图像,依据图像中各个坐标系间的相互关系,获取不同角度的成像几何模型(照相机模型),并对各祯图像中场景的深度信息进行估计,采用面片逼近曲面的方法获取精确度较高的图像间的像素匹配,并将各帧图像的深度信息融合为一个整体的三维初始模型,对建立的初始模型进行全局的迭代优化,消除图像中存在的误差点,输出最优的三维重建模型。仿真结果表明,采用三维虚拟视觉的导视系统设计方法精确度高,能够设计出高质量完整的三维图像导视系统。

Accurate description for the angle and coordinate of the curved surface object has always been a diffi- culty of sign system design. In traditional method, the way of multi - angle superposition approach is used to com- plete the design of the complex curved surface, but the error is too big. A sign system design method was proposed in the paper based on 3D virtual vision. In above mentioned method, multigroup original images with different angles were collected. On the basis of the mutual relationship between each coordinate system in the image, the imaging ge- ometric model （ camera model） of different angles was obtained, and the depth information of scene of each frame im- age was estimated. The method of surface patch approaching the curved surface was used to obtain the pixel matching between images with higher accuracy, and the depth information of each frame images was fused to an integral 3D ini- tial modal. The established initial model was optimized globally by iterating, to eliminate the error point of the image and output the optimal 3D reconstruction model. Simulation result proves that the precision of the method is high, it can design complete 3D image sign syste with high quality.