高级检索
陈旺, 张茂军, 崇洋, 熊志辉. 单-双镜面组合结构的全向深度获取系统[J]. 计算机辅助设计与图形学学报, 2010, 22(6): 1004-1009.
引用本文: 陈旺, 张茂军, 崇洋, 熊志辉. 单-双镜面组合结构的全向深度获取系统[J]. 计算机辅助设计与图形学学报, 2010, 22(6): 1004-1009.
Chen Wang, Zhang Maojun, Chong Yang, Xiong Zhihui. Omni-Directional Depth Estimation System with Single-Double Mirrors Architecture[J]. Journal of Computer-Aided Design & Computer Graphics, 2010, 22(6): 1004-1009.
Citation: Chen Wang, Zhang Maojun, Chong Yang, Xiong Zhihui. Omni-Directional Depth Estimation System with Single-Double Mirrors Architecture[J]. Journal of Computer-Aided Design & Computer Graphics, 2010, 22(6): 1004-1009.

单-双镜面组合结构的全向深度获取系统

Omni-Directional Depth Estimation System with Single-Double Mirrors Architecture

  • 摘要: 为避免成像系统对于单视点约束的要求,提高场景深度获取的准确性和灵活性,设计了一个由单透视相机、双抛物反射镜面同轴放置所构成的非单视点全向深度获取系统.首先分析了实现单视点约束存在的困难及基于非单视点成像系统获取深度的可行性;然后提出了在双镜面全向成像和单镜面全向成像2种不同方式下的深度计算方法;最后通过误差分析研究了镜面面型、基线距离及空间点位置对深度恢复结果的影响,并由此优化系统配置参数,使深度估计的总体误差保持在2%以内.该系统不仅可基于双反射镜面全向立体成像恢复水平方向360°的场景深度,也可仅利用单反射镜面全向成像获取空间线特征的三维信息.实验结果表明,文中系统具有较高的准确度,可用于不同精细程度需求下的摄影测量、三维重建、机器人导航等场合.

     

    Abstract: In order to avoid the enforcement of the single-viewpoint constraint and to improve the reconstruction accuracy,a new omni-directional depth estimation system with single-double paraboloid mirrors is proposed.At first,the difficulty of manufacturing a real single-viewpoint system in practice is illustrated,and the feasibility of our proposed approach is proved.The detailed system design is then presented,and formulations of depth estimation from images reflected by single and double conicoid mirrors are investigated.Finally the system parameters are optimized to make sure the error rate of depth estimation usually lies within 2%.This system can not only obtain depth of 360° field of view via omni-directional stereo technique by the double-mirror configuration,it also can locate space lines based on single omni-directional image captured with the single-mirror structure.Our proposed system could be used for various applications such as photogrammetry,3D reconstruction,and robot navigation.

     

/

返回文章
返回