一种改进二维投影模型运动参数估计算法  

An improved algorithm for motion parameters estimation of two-dimensional projective model

作　　者：赵歆波[1] 邹晓春[2] 马钟[1] 

机构地区：[1]西北工业大学计算机学院,西安710072 [2]西北工业大学电子信息学院,西安710072

出　　处：《中国体视学与图像分析》2010年第4期372-376,共5页Chinese Journal of Stereology and Image Analysis

基　　金：国家自然基金项目(60872144)

摘　　要：二维投影模型运动参数估计是图像序列运动分析的关键技术,针对Levenberg-Marquardt算法在解决二维投影模型运动参数估计问题时计算量大的缺点,提出了一种改进的二维投影模型运动参数估计算法。该方法通过对非线性最小平方误差函数的线性化,重新定义了目标函数,采用极小化目标函数的方法求解参数矢量,得到只与已知特征点对应关系有关的系数矩阵后,利用线性算法求解图像序列相邻两帧间二维投影模型的运动参数。实验结果表明,该算法在保证精确度的条件下,提高了计算效率。The motion parameters estimation of two-dimensional projective model is a key technique in image sequence analysis.Aiming at the shortcoming of the algorithm for Levenberg-Marquardt algorithm,i.e.,huge computational cost,an improved algorithm for motion parameters estimation of two-dimensional projective model is proposed in this paper.Firstly,the algorithm redefines the objective function by linearizing the non-linear least-square error function.Then,minimization algorithm is used for the function to get the vector of the parameter.As a result,the coefficient matrix is only related to the feature-correspondence that we have known.Finally,the motion parameters of two-dimensional projective model are linearly solved.Experimental results show that the proposed algorithm ensures the precision and improve the computational efficiency.

关 键 词：运动参数估计 二维投影模型 最小平方误差 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]