【作者】 梁莹；

【导师】 谈振藩；

【作者基本信息】 哈尔滨工程大学 ， 导航、制导与控制， 2010， 博士

【摘要】 随着地磁测量技术及各相关学科的快速发展,地磁导航的综合优势日益突出。地磁匹配导航定位误差不随时间积累且具有长期稳定性,可弥补惯性导航系统(Inertial Navigation System, INS)误差随时间累积的不足,INS的短期高精度又可为地磁匹配提供位置参考,从而提高匹配效率和匹配精度。INS/地磁匹配组合定位能满足“长期、自主、实时、高精度、全天候”的导航需求,从而成为一种新型的组合导航技术。现有的大范围、导航用高精度磁测数据的数据源不足,针对此问题,本文在国际地磁学与高空大气物理学协会(International Association of Geomagnetism and Aeronomy, IAGA)提供的磁测数据Earth Magnetic Anomaly Grid 2 (EMAG2)基础上,提出一种插值算法对数据进行内插,以提高地磁图分辨率。空间数据内插中广泛使用的克里金法是无偏最优的,但它的“平滑效应”或“低通滤波”作用,难以获得两点间的局部细节变化和非均质特征。因此,本文将分形几何学思想引入到磁异常数据二维内插中,采用多重分形测度与克里金法相结合,提出一种多重分形测度克里金插值法。这种插值法不但能更真实地表征磁异常在几何域上的趋势变化,同时还能保持和增强磁异常的局部结构信息。地磁异常场是一种位场,存在大面积特征相似的部分,因此匹配时正确选择匹配区域很重要。根据磁异常具有非线性和自相似性的特点,通过定义特征指标,对某区域中18个子区域磁异常的估计精度、离散程度、单位表面起伏程度及不规则程度等进行了统计分析,在此基础上对区域适配性能做综合评价。在适配区的选择上,现有文献大多是通过多次试验,根据各区域的匹配概率和匹配误差人为划分选择标准,目前还没有通用的适配区选择方法。本文提出一种基于主成分分析的适配区选择法,可对区域适配性进行综合定量描述,使适配区的选择变得比较简单、直观、有效。以磁异常为特征量匹配时,磁异常值在个别局部范围内的数值差异小而导致配准率降低,针对这一问题,提出基于多重分维谱提取匹配特征量。多重分维谱通过描述磁异常的局部特征以及磁异常在其形成过程中不同层次的特征,能准确分辨磁异常相似区中数据间的差异。对于地磁匹配算法的研究,公开发表的文献中多是将地形匹配、图像处理的方法延伸或推广应用于地磁匹配导航中,并没有形成为地磁导航量身定做的匹配算法。由于地磁匹配导航可获样本少、非线性、维数高、要求全局最优等特点,本文提出一种基于支持向量机的地磁匹配算法,用支持向量机进行模式分类来实现匹配定位。考虑到磁异常表面的不规则性及自相似性,文中提出以分形几何中的自相似参数作为高斯核宽度参数。的初值,用启发式搜索共同选择。和惩罚因子C。给出了匹配算法实现的详细步骤。仿真结果表明,算法适用有效。对INS/地磁匹配组合定位的技术实现展开了研究。设计了组合定位滤波器。以INS的经、纬度误差等参量为状态量,以INS输出位置与地磁匹配位置之差为观测量,用滤波器估计了INS定位误差。对三个特征不同子区域的组合定位进行了仿真,结果显示地磁匹配可有效抑制INS长时间定位精度降低问题。更多还原

【Abstract】 With the fast development of the geomagnetic measurement technology and the related studies, the comprehensive advantages of geomagnetic navigation have become increasingly prominent. The positioning errors of geomagnetic matching navigation do not accumulate over time and have long-term stability, which can compensate the error accumulated over time in inertial navigation system (INS). INS can provide high-precision reference of short-term for geomagnetic matching positioning, which can improve the efficiency and accuracy of matching. INS/geomagnetic matching integrated positioning is becoming a new navigation technology, which can satisfy "long-term, self-contained, high-precision, all-weather" navigation requirements.The large-scale high precision data for navigation are inadequate. Based on the geomagnetic measurement data, Earth Magnetic Anomaly Grid 2 (EMAG2), provided by International Association of Geomagnetism and Aeronomy (IAGA), a new interpolation method is proposed for improving resolution of the geomagnetic map. Kriging method widely used in spatial data interpolation is unbiased optimal. It connects the two adjoining points with a straight line or a higher-order smooth curve that leads to "smooth effect" or "low-pass filter", so it is difficult to get the details of changes and non-average features between two points. Therefore, the fractal geometry theory is introduced to the geomagnetic data interpolation. Here the Multifractal Measure Kriging (MFMK) interpolation method is proposed. MFMK combines Kriging method with multifractal measure. It can not only show the trends of geomagnetic anomaly in geometric domain, but also maintain and enhance the local information of data structure.Geomagnetic anomaly field is a potential field, and there are large areas with similar characteristics. Therefore, selecting the suitable-matching region(SMR) is very important. Geomagnetic anomaly data are nonlinear and self-similarity. By defining feature indexes, the estimation accuracy, scattering, undulations within unit surface, irregular degree of geomagnetic anomaly regions are analyzed for selecting the SMR, and the performance of regional matching is evaluated on this basis. There has not been a universal method for SMR selection now. It mostly depends on the matching probability and matching error of each region to division standard by man-made in the existing references. In this paper, a SMR selection method based on principal component analysis(PCA) was proposed. It can evaluate the suitability of regions quantitatively and makes the selection problem more simple, intuitive, and effective.When geomagnetic anomaly is chosen as the characteristic variable for matching, the changes of geomagnetic anomaly values are small in some local areas, and that will lead to low precision and accuracy. To solve this problem, multifractal dimension spectrum is introduced for the extracting characteristic variable for matching. By describing the local features of the geomagnetic anomaly and the different features of the forming process of geomagnetic anomaly, multifractal dimension spectrum can distinguish the differences between the data from the similar geomagnetic anomaly areas correctly. In most of the published research reports about the geomagnetic matching algorithms, the terrain matching or image processing methods were generally applied for geomagnetic matching navigation. There are no special geomagnetic navigation matching algorithm. In terms of the characteristics of less samples, nonlinearity, high-dimension, high global optimality of geomagnetic anomaly, a geomagnetic matching algorithm based on support vector machine is proposed for pattern classification of latitude and longitude in this paper.. As self-similarity fractal parameter H can reflect the degree of the self-similarity and irregularity of geomagnetic anomaly surface, H was used as the initial value of Gaussian kernel width s. To improve search efficiency and real-time of algorithm, s and penalty factor C were selected together by heuristic search. The steps of the matching algorithm is given. Simulation results show that the algorithm performs effectively.The integrated technology of INS and geomagnetic matching is also studied. The integrated positioning filter is designed. Error of longitude and latitude and other parameters are set as state variables, and the difference of INS output position and geomagnetic location are set as observables so the INS position error can be estimated by filter. The results of integrated positioning simulation on regions of three different characteristics show that INS/geomagnetic matching integrated positioning can inhibit the reducing of INS positioning accuracy for a long time.更多还原