摄像测量像机链中多头像机固连关系标定方法研究

发布时间：2018-04-19 16:40

本文选题：路基沉降监测 + 多头像机标定　；参考：《国防科学技术大学》2015年硕士论文

【摘要】：在水利水电、交通、高层建筑、采矿等领域大型工程建设中,这些基础设施的路基会存在沉降变化。严重的路基沉降不仅影响工程的施工、运营安全,更可能造成不可挽回的损失。路基沉降的实时自动监测需求大量存在,而位移传递像机链摄像测量方法是解决该类问题的有效途径之一。高精度标定出像机链中这些固连关系的多头像机间的安装参数是准确监测沉降变化的必要任务。本文系统调研了多像机系统外参数标定方法的历史发展与研究现状,着重对已有方法进行了分类和比较,从多方面综合评价了各类方法的性能,总结和归纳出多头像机标定待研究的关键问题。讨论了路基沉降摄像测量监测系统中的相关误差来源,重点研究了双头像机光轴不重合引起的模型误差,并通过仿真实验分析了两个像机光轴不重合对测量精度的影响。仿真结果表明虽然双头像机光轴不重合引起的合作标志像点变化是个小量,但是由于传递级数的增加会放大测量误差,而且路基沉降测量精度一般都要求在亚毫米量级,即使微小的误差也会影响沉降测量精度。因而,标定出双头像机间的固连关系,利用它们之间的相对关系消除这一误差是很有必要的。为了解决路基沉降像机链中多头像机间没有公共视场的难题,本文研究了基于机器人手眼定标和利用平面镜反射辅助视场的多头像机标定方法。由于受到双头像机固连的约束,在标定过程中像机组能够运动的范围较小,这极大地限制了手眼定标方法的精度。与传统的手眼定标方法相比,本文应用的两种共轭旋转平均的手眼定标方法对噪声不敏感,在配置条件不太理想的情况下精度更高。为了使多头像机的标定更加简单易行,本文在平面镜辅助视场位姿估计基础上研究了基于平面镜的多头像机标定方法。利用克拉美罗下界知识推导了平面镜辅助视场下位姿估计的理论误差下界,并分析了像点误差、视图幅数以及平面镜参数对位姿估计精度的影响。设计了两种基于平面镜的多头像机标定方案,最后通过仿真实验和实物实验验证了两种标定方案的可行性与准确性。在路基沉降摄像测量中,多头像机固连关系的标定是一个极其重要的基础性环节。解决多头像机的标定问题对于能够有效减少沉降位移传递中的多级误差累积、完善沉降监测系统,具有重要的研究意义。
[Abstract]:The subgrade of these infrastructure will change in the construction of large projects in the fields of water conservancy and hydropower, transportation, high-rise building, mining and so on.Serious subgrade settlement not only affects the construction and operation safety of the project, but also may cause irreparable losses.There is a large demand for real-time automatic monitoring of subgrade settlement, and one of the effective ways to solve this kind of problem is the camera measurement method of displacement transfer camera chain.It is necessary to accurately calibrate the installation parameters between the multiple cameras in the camera chain.In this paper, the historical development and research status of the calibration methods for external parameters of multi-camera systems are systematically investigated, and the existing methods are classified and compared, and the performance of various methods is comprehensively evaluated from many aspects.The key problems in the calibration of multi-head camera are summarized and summarized.This paper discusses the sources of relative errors in the monitoring system of subgrade settlement measurement by camera, focuses on the model error caused by the non-coincidence of the optical axis of the double-headed camera, and analyzes the influence of the non-coincidence of the optical axes of the two cameras on the measurement accuracy through the simulation experiment.The simulation results show that although the variation of cooperative symbol image points caused by the misalignment of optical axis of dual-headed camera is small, the measurement error will be magnified by the increase of transmission series, and the accuracy of submillimeter settlement measurement is generally required to be in the order of sub-millimeter.Even small errors can affect the accuracy of settlement measurement.Therefore, it is necessary to calibrate the fixed connection relationship between two head cameras and eliminate this error by using the relative relation between them.In order to solve the problem that there is no common field of view between multi-head cameras in roadbed settlement camera chain, this paper studies the calibration method of multi-head camera based on robot hand-eye calibration and using plane mirror reflection to assist the field of view.Due to the constraint of the binocular camera, the range of the image unit can move is small in the calibration process, which greatly limits the precision of the hand-eye calibration method.Compared with the traditional hand-eye calibration methods, the two conjugate rotation average hand-eye calibration methods used in this paper are not sensitive to noise, and the accuracy is higher when the collocation conditions are not ideal.In order to make the calibration of multi-head camera more simple and practical, this paper studies the calibration method of multi-head camera based on plane mirror based on plane mirror aided field of view orientation estimation.The lower bound of the theoretical error of the attitude estimation of view field assisted by plane mirror is derived by using the lower bound knowledge of Clemero. The effects of image point error, view amplitude and plane mirror parameters on the precision of pose estimation are analyzed.Two calibration schemes of multi-head camera based on plane mirror are designed. Finally, the feasibility and accuracy of the two calibration schemes are verified by simulation experiments and physical experiments.In the imaging measurement of subgrade settlement, the calibration of fixation relation of multi-head camera is an extremely important basic link.To solve the calibration problem of multi-head camera is of great significance to reduce the accumulation of multi-stage errors in the transfer of settlement displacement and to improve the settlement monitoring system.
【学位授予单位】：国防科学技术大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：U418.5;P23

【参考文献】