全向视觉中数据噪声对参数估计的影响

针对全向视觉中特有的数据特性,对图像中不同区域的数据噪声提出了不同假设,并在此假设下分析了数据噪声对线性参数估计的影响,研究表明,尽管远离图像中心区域的数据误差较大,但这些数据对参数估计的影响是非常大的.此外,还提出了一种分步选择数据点的方法,首先选择较高精度的数据得到初始参数估计值,然后逐步选择更多的数据点以保证得到尽可能多的数据点,而且所选择的数据点尽可能覆盖图像的边缘区域.实验结果表明,文章提出的方法可以得到更精确和更稳定的参数估计.

Impact of Data Noise on Parameter Estimation in Omnidirectional Vision

Based on the characteristics of data noise in context of omnidirectional vision,we assume different noise models for different regions in an omnidirectional image,and the effect of data noise on linear parameter estimation is studied.The results indicated that those data far from the image center have very important impact on the parameter estimation,though the data are less accurate.Furthermore,we also give a data selection method which is performed gradually.The initial model is estimated by using high pr...