【作者】 张晓东；

【导师】 杨元喜；

【作者基本信息】 解放军信息工程大学 ， 大地测量学与测量工程， 2013， 博士

【摘要】 高精度动态定位定姿技术、数字传感器技术、近景摄影测量技术以及自动控制技术的飞速发展和集成融合，使得移动状态下的地面遥感快速测量成为可能，从而催生出一个新的应用领域——移动测量。移动测量技术为三维空间信息的快速高精度测量和更新开辟了一个新的途径。移动道路测量技术具有动态定位定姿测量速度快和地面近景摄影测量信息量大的特点，提高了野外空间数据获取的效率，获取的可量测立体影像也使得数据处理和应用更为灵活多样。将移动道路测量系统的可量测立体影像与GPS（Global PositionSystem）、IMU（Inertial Measurement Unit）数据有机结合，为移动测量系统的高精度、高可靠性定位定姿提供了一种新的途径。移动测量系统是一个典型的多传感器集成系统。一方面，遥感影像需要利用POS提供的高精度定位定姿数据实现与地理信息的融合，实现无控制、快速遥感立体定位；另一方面，来自无控制移动测量系统的DMI（Digital Measurable Image）本身是一个具有严密几何传递参数的序列影像，具有定位能力，作为一种新的数据源，可以与GPS、IMU数据进行集成融合定位，从而构建一个集GPS/IMU/DMI定位定姿技术于一体的更为复杂的组合导航系统，可望进一步提高移动测量系统定位定姿的精度和可靠性。本文以无控制移动测量系统中用于直接地理参考的定位定姿测量为研究对象，研究可靠的、高精度的多传感器集成定位定姿方法与技术，包括基于GPS、IMU的高精度定位定姿方法、GPS/IMU紧组合定位定姿技术、基于移动测量系统的序列可量测影像定位定姿方法。在此基础上研究基于GPS、IMU与可量测序列影像集成融合定位定姿模型与算法，探讨有可量测影像参与的GPS/IMU/DMI组合定位定姿方法，实现多传感器数据融合的组合POS，提高GPS/IMU定位定姿处理的精度和可靠性，为无控制移动测量技术的实际工程应用提供理论和技术支撑。本文围绕着基于陆基移动测量系统的定位定姿技术，主要开展了以下研究：1）将POS定位定姿传感器与遥感成像CCD相机集成，实现无控制移动测量。本章首先总结了多传感器集成与无控制移动道路测量的理论与方法，分析了无控制移动测量误差模型，在此基础上提出了无控制移动测量的关键技术。详细讨论了无控制移动测量误差模型与关键技术，分析了陆基无控制移动测量的数据特征，构建了基于陆基无控制移动测量多源数据融合定位定姿的基本模型，探讨了其技术可行性。2）对陆基移动测量系统定位定姿的方法进行了系统的总结和分析，包括GPS定位定姿方法和惯性测量定位定姿方法。惯性测量定位定姿主要包括SINS定位定姿、DR定位定姿和3G1M（3个陀螺和1个里程计）定位定姿。在此基础上，重点讨论了GPS/IMU定位定姿模型，对GPS/IMU松组合定位定姿的原理、算法特点进行了分析，以陆基移动测量系统为平台进行了GPS/IMU松组合定位定姿算法测试；介绍了GPS/IMU紧组合定位定姿算法模型，包括基于IMU的GPS整周模糊度固定算法和IMU辅助GPS周跳探测方法；以陆基移动测量系统获取的数据进行了算法实测，验证了GPS/IMU紧组合定位定姿算法的有效性和可靠性。最后，提出一种紧组合下的惯性辅助三星定位方法，以实测数据进行了算法验证，表明采用IMU辅助可以实现三星定位。3）对影像视觉定位定姿技术及其国内外研究现状进行了总结与分析，提出了基于无控制移动测量系统可量测序列影像的定位模型，对序列DMI定位定姿的误差进行了分析和讨论，指出序列影像间连接点特征提取和匹配是影响DMI定位定姿的关键因素；可量测序列影像定位本质上也是一种视觉定位方式，将视觉定位与POS融合集成，可以提高组合定位定姿精度与可靠性；此外，DMI本身是一个具有严密几何传递参数的序列影像，作为一种新的数据源，也应该与GPS/INS数据进行集成融合，从而构建一个集GPS/INS/DMI定位定姿技术于一体的组合导航系统；最后，为了提高基于DMI定位定姿的精度和可靠性，提出了基于SIFT的序列DMI影像自动匹配方法。4）在移动测量系统中，利用DMI、GPS和IMU进行定位定姿时各自所具有的不同优势和特点，将不同传感器数据进行融合通常采用卡尔曼滤波算法。但是卡尔曼滤波需要给定合适的状态模型、状态噪声矩阵和观测噪声矩阵，如果噪声矩阵不能较可靠地反映状态参数的不确定性，必然会影响滤波解，导致估值次优甚至滤波分散。为克服上述问题，针对DMI观测数据的特点，介绍了DMI的单因子自适应滤波、多因子自适应、分类自适应滤波和抗差自适应滤波；研究了自适应因子的构造方法，提出了不同的自适应因子模型，并对其在DMI定位定姿滤波中的应用情况进行了分析；提出了基于多传感器空间约束条件的自适应卡尔曼滤波，并将其模型拓展到GPS、IMU和DMI的融合定位定姿中；进一步提出了基于多传感器的抗差自适应GPS/IMU/DMI融合定位定姿和基于多传感器空间约束条件的融合定位定姿模型。5）介绍了陆基移动测量系统及实验系统，以实验系统获取的不同实验场地的大量DMI、GPS、IMU实测数据，分别开展了GPS/IMU松组合实验、GPS/IMU紧组合实验、DMI与GPS/IMU融合定位实验，对实验结果进行分析和讨论，结果表明，本文构建的方法是合理和有效的。最后，对可量测影像DMI与GPS/IMU融合集成定位定姿的应用模式进行了探讨。总之，采用多传感器集成理论方法、多传感器多源数据抗差融合处理方法将直接的定位定姿GPS、IMU数据和间接的可量测序列影像数据结合起来，一方面，可通过GPS、IMU紧组合融合提高直接定位定姿的精度；另一方面，利用来自陆基移动测量系统的可量测序列影像DMI定位定姿推算，提高定位定姿的可靠性和适用范围。更多还原

【Abstract】 The high-precision dynamic position and orientation technology, digital sensor technology,close-range photogrammetry technology, and autocontrol technology have had greatdevelopments and integration and fusion, which makes it possible for close rangephotogrammetry and terrestrial remote sensing, and as a result, a new subdiscipline namedvehicle borne mobile measurement come into being, which has opened up a new way for fasthigh-precision measurement and information updating in the three-dimensional space. Themobile road measurement technology charactered by high-speed in dynamic position and largeamount of information in terrestrial close range photogrammetry improves the efficiency offield space data acquisition. The stereo images obtained by mobile road measurementtechnology make it flexible in data processing and applications. With the organic combinationof the stereo image and the GPS/IMU data from the mobile road measurement system, a newapproach is provided for the high-precision and high reliability of the mobile measurementsystem.The mobile measurement system is a typical multi-sensor integrated systems. On one hand,the high-precision data provided by POS are used in the fusion of the remote sensing imagesand the georeferenced information, and then the uncontrolled and fast remote sensing arerealized. On the other hand, the DMI（Digital Measurable Image）with position capabilityobtained by uncontrolled system is a sequence of images with strict geometry pass parameters.When the DMI as a new data source is fused with the GPS/INS data, a more complex integratednavigation system with GPS/INS/DMI positioning and orientation technology is built, whichmake it possible to improve the position and orientation precision and reliability of the mobilemeasurement system in the future.In this thesis, the study object is the position and orientation measurement based ondirectly georeferenced of the uncontrolled mobile measurement system, and the reliable andprecise position and orientation technologies based on multi-sensor integration are researched,which include the high-precision methods based on GPS, INS and odometer, thetightly-coupled GPS/INS technologies, and the methods based on the sequence of measurementimages of mobile measurement system. On the basis of the above studies, the position andorientation models and algorithms based on the integration and fusion of GPS, INS, odometerand the sequence of measurable images are researched. And then the GPS/IMU/DMI integratedmethods for position and orientation based on the combination of the measurable images and the vector position recursive parameters are discussed. At last, the multi-sensor data fusionintegrated POS is designed and the process precision and reliability of GPS/IMU integratedposition and orientation system are developed, which provides theoretical and technicalsupports for uncontrolled mobile measurement technology.Surrounding the Position and orientation technologies based on land-based mobilemeasurement system, the following studies have been made in this thesis.1) Uncontrolled mobile measurement can be realized by integrating the position andorientation sensors of POS with the CCD cameras of remote sensing system. Firstly, thetheories and methods of multi-sensor integration and uncontrolled mobile road measurementare summarized and the uncontrolled mobile measurement model is analyzed, based on whichthe key technologies of uncontrolled mobile measurement are advanced. Secondly, the errormodel and the key technology of uncontrolled mobile measurement are detailed. Then, the datafrom land-based uncontrolled mobile measurement are analyzed. Finally, the basic model ofmulti-source data fusion for position and orientation is aerated, and its technical feasibility isdiscussed.2) The position and orientation methods of land-based mobile measurement system aresystemically summarized and analyzed, they consist of position and orientation methods usingGPS and the methods using inertial measurement. The position and orientation methods usinginertial measurement mainly include SINS, DR and3G1M. Next, on the basis of above studies,the GPS/IMU position and orientation model is discussed, the principles and algorithms ofGPS/IMU loosely-coupled integration are analyzed, and the GPS/IMU loosely-coupledintegration algorithms are tested in the land-based mobile measurement system. Then, thetightly-coupled GPS/IMU integration position and orientation model is presented, it consists ofthe fixed algorithm based on the IMU GPS ambiguity and the detection algorithm of IMUassisted GPS cycle slip. Immediately, the validity and reliability of the algorithms are verifiedby using the data obtained from land-based mobile measurement system. Finally, the tri-satellitepositioning method aided by INS in the case of tightly-coupled integration is put forward, and itis tested by the real data. The result shows that the method is applicative.3) Based on the analysis of image position and orientation technology and the presentresearch status at home and abroad, the following researches have been done. Firstly, thepositioning model based on the digital measurable image (DMI) of uncontrolled mobilemeasurement system is proposed, the error is analyzed, and the key factors in position andorientation are pointed out. Secondly, due to the fact that the measurable sequence imagespositioning is visual positioning essentially, the precision and reliability can be improved by theintegrated fusion of POS and visual positioning. Thirdly, the integrated navigation system with GPS/INS/DMI is built, and the precision and reliability are further improved by the way ofusing the integrated fusion of the DMI and GPS/INS data. At last, the sequence DMIautomatically matching method based on SIFT is proposed to improve the precision andreliability of DMI position and orientation.4)In the mobile measurement system, DMI,GPS and IMU have different advantages andcharacteristics in position and orientation. Whatever it is based on what kind of sensor, Kalmanfiltering is the most commonly used method to realize the data fusion of different sensors.Kalman filtering needs appropriate state model, state noise matrix and measurement noisematrix. If the given noise matrix is not appropriate to reflect the uncertainty of state parameters,the filtering solution will be affected, which leads to the suboptimal evaluation or filteringdispersion.In order to solve the above problems, in consideration of the characteristics of DMI data,the single-factor adaptive filtering and multi-factor adaptive filtering based on DMI areproposed. At first, the method of adaptive factor structure is studied, some adaptive factormodels are given, and the application in DMI position and orientation is analyzed. Then,adaptive Kalman filtering based on multi-sensor space constraints is put forward, and it isapplied in the GPS/IMU/DMI integrated position and orientation. Finally, the robust adaptiveGPS/IMU/DMI integrated position and orientation model and space constraints integratedposition and orientation model based on multi-sensor are presented.5)The land-based mobile measurement system and the experimental fields are introduced.Then, on the basis of a large amount of DMI, GPS and IMU data obtained from differentexperimental fields, the GPS/IMU loose combination experiments, the GPS/IMUtightly-coupled integration experiments and the DMI/GPS/IMU integrated position experimentsare conducted, and the experimental results are analyzed, which show that the proposed methodis effective and reasonable. At last, the mode of DMI/GPS/IMU integrated position andorientation is discussed.In conclution, with the combination of the firsthand data obtained from position andorientation GPS/IMU and the secondhand measurable images by using the theoretical methodsof mult-sensor integration and the Robust approach of multi-sensor multi-source data, tworesults can be obtained: one is that the position and orientation precision can be improved bythe way of tightly-coupled GPS/IMU integration, and the other one is that the reliability isenhanced and the utility range is broadened by using of DMI position and orientationdeductions from the measurable images obtained by land-based mobile measurement system.更多还原