基于序列图像特征配准的摄像机旋转补偿算法  

Video camera rotation compensation algorithm based on feature matching of sequence image frames

作　　者：王会峰[1] 刘上乾[1] 汪大宝[1] 牛建军[1] 

机构地区：[1]西安电子科技大学技术物理学院,陕西西安710071

出　　处：《光学精密工程》2008年第7期1330-1334,共5页Optics and Precision Engineering

基　　金：国家自然科学基金资助项目(No.60377034)

摘　　要：提出了一种基于序列图像特征点配准和基于最小二乘估计的摄像机绕轴旋转运动精确估计算法。该算法利用序列图像帧间的强相关性,通过分析相邻两帧图像的运动,由Harris算子进行特征点检测。基于这些特征点用模板匹配方法对帧间图像进行配准,然后根据多个特征点的运动矢量用最小二乘估计获得摄像机的运动参数,并用获得的运动参数实现对摄像机绕轴旋转运动的精确补偿。实验结果表明,该方法在帧间旋转角度<10°时,摄像机绕轴旋转运动角度估计误差<5%,但是当旋转角度>14°时,相对估计误差>20%。从系统实际应用(帧间旋转<5°)来看,该方法克服了成像测量中摄像机旋转对图像处理和测量精度的影响,弥补了因像机旋转引起的测量误差大的缺陷,提高了测量精度,可满足使用要求。In order to eliminate the measurement error caused by video camera rotation during its moving along the axis, a high precision video camera moving estimation algorithm based on the feature point matching of sequence image frames and the least-square estimatimation is presented. In the algorithm, the strong correlation of sequence image frames is used to analyze the moving parameters of two continuous image frames. Then, the feature points of sequence image frames are detected by Harris operator and matched by template, and the moving parameters of the video camera are obtained with the least-square algorithm. Finally, the obtained moving parameters are used to compensate the video camera rotation precisely. The experimental results show that the rotation angle error of the video camera is less than 5% when the rotation angle of inter-frame is less than 10°,but more than 20% when it is more than 14°. Because the rotation angle of inter-frame is less than 5° in practical application,the algorithm eliminates the influence of video camera rotation on measurement precision, compensates its measurement error,and improves the measurement precision, which meets the requirement of practical application.

关 键 词：图像测量 角点检测 图像配准 运动估计 最小二乘法 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]