摘要
分析总结了立体视觉对应点匹配研究现状,针对单相机双抛物反射镜面共轴结构的特定折反射立体全向成像装置,将原始捕获的立体全向图像对投影成柱面全景图像对,并结合动态规划和灰度段匹配思想给出了一种改进的对应点匹配算法用于柱面全景图像的稠密深度估计。基于虚拟装置和虚拟场景的仿真实验说明了该装置结构设计和对应点匹配方法的有效性。
After available corresponding points match algorithms in computer stereo vision were analyzed and summarized, aiming at the coaxial structure design of a certain catadioptric omni-directional imaging system with one camera and two parabolic reflectors, the stereo cylindrical panorama pairs were generated from the originally captured omni-directional image pairs, and then an improved dynamic programming and grayness segment based corresponding points match algorithm was proposed to estimate the dense depth information of the cylindrical panorama pairs. The simulated experiment using virtual equipment and virtual scenes shows the validity of the system structure design and the corresponding points match algorithm.
出处
《系统仿真学报》
CAS
CSCD
北大核心
2009年第22期7185-7189,共5页
Journal of System Simulation
基金
国家自然科学基金(60872150
60705013)
关键词
立体视觉
折反射全向成像
柱面全景
动态规划
对应点匹配
stereo vision
catadioptric omni-directional imaging
cylindrical panorama
dynamic programming
corresponding points match