摘要
基于随机采样最小冗余子集的新概念，依据单视图特征点集的模型数据和图象数据，开发了一项目标三维视觉信息鲁棒精确复原新方法在强噪声高出格率的恶劣条件下，该方法仍可高精度地复原目标的三维信息实验表明，对于由１００个特征点组成的数据集而言，当出格率高达９０％，内点信噪比低达２８ｄＢ时。
Based on the new idea of randomly sampling the redundant minimal subset,a new method is developed of robustly and accurately restoring the 3-D vision information of an object from the model data and image data of its key point set of a single perspective view.The technique is able to normally work with high accuracy under very hard condition of heavy noise and high outlier rate.The experiments demonstrate that for an image date set consisting of 100 key points,when the outlier rate is as high as 0.9 and the SNR of its inliers is as low as 28dB,the algorithm is still able to restore the 3-D coordinates of the key points with relative error of 1%.
出处
《哈尔滨工程大学学报》
EI
CAS
CSCD
1998年第3期51-57,共7页
Journal of Harbin Engineering University
关键词
计算机视觉
三维信息复原
鲁棒估计
单视图
computer vision
restoration of 3-D information
robust estimation
redundant minimal subset
random sampling
parameter refinement
