光学学报, 2012, 32 (7): 0711002, 网络出版: 2012-05-31
简易高斯灰度扩散模型的误差分析及适用性研究
Error Analysis and Applicability Study on Simplified Gaussian Gray Diffusion Model
图像处理 星图模拟 高斯分布 归一化 灰度重心法 image processing star map simulation Gaussian distribution normalization gray weighted centroid method
摘要
为了验证简易高斯灰度扩散模型的适用性,与传统高斯灰度扩散模型进行了对比分析。将两种高斯模型做归一化处理,设定检验像素,分析检验像元灰度的归一化值的相对误差;进行星图模拟,得到4个不同高斯半径(σ)下灰度赋值相对误差与像点映射位置偏离值的关系曲线,整体上误差随σ的增大而减小;对星图模拟得到的系列星像点采用灰度重心法提取质心,质心误差随σ增大而减小,传统模型模拟像点的质心提取精度比简易模型高约2个数量级。无噪声条件下σ=0.671时,简易模型模拟像点最大质心误差仅为0.033 pixel。仿真结果表明:单就像点外形仿真而言,当σ较小时,简易模型不再适用;但针对像点的质心定位及后续算法,简易模型带来质心误差量级可以忽略,运算量更小,适于应用。
Abstract
The applicability of the simplified Gaussian gray diffusion model needs to be further verified theoretically, whose error analysis method relies on the analytical comparison with the typical traditional one. The two models are normalized respectively, a test pixel is chosen, and the evaluating criterion for the simplified model is established by analyzing the relative error of the normalized gray value regarding the test pixel. Star image simulation is conducted, and four relative error curves to the mapping position deviation from the pixel central coordinate are plotted, each of which represents a different case of Gaussian radius σ, manifesting that error decreases as σ increases as a whole. Gray-weighted centroiding method is carried out to the two series of simulated star images generated respectively by the two models. Centroiding absolute error decreases as σ increases, absolute error resulted from simplified model is about 2 orders of magnitude higher than that of the traditional one. Under the conditions of no noise and σ=0.671, the maximum absolute error from the simplified model is only 0.033 pixel. Simulated results show that the simplified model is no longer applicable when Gaussian radius σ is rather small in terms of shape simulation of image point. Considering that latter algorithms of a star sensor, such as star matching and attitude determination, whose focus is just centroid, the simplified model still possesses applicability due to its advantage of small calculating amount, also because its error magnitude is small enough to be neglected anyway.
王海涌, 周文睿, 赵彦武. 简易高斯灰度扩散模型的误差分析及适用性研究[J]. 光学学报, 2012, 32(7): 0711002. Wang Haiyong, Zhou Wenrui, Zhao Yanwu. Error Analysis and Applicability Study on Simplified Gaussian Gray Diffusion Model[J]. Acta Optica Sinica, 2012, 32(7): 0711002.