【作者】 卞和方；

【导师】 张书毕；

【作者基本信息】 中国矿业大学 ， 大地测量学与测量工程， 2013， 博士

【摘要】 精密单点单位技术（PPP）不需要坐标已知的固定参考站，不受基线长度的制约，单台接收机即可在全球范围内进行静态或动态定位，并且能直接得到全球同精度的ITRF框架内的精密三维坐标。本文以大面积矿区高精度变形监测为背景，结合PPP关键技术展开相关研究，结合实例分析了基于区域增强PPP在矿区变形监测中的应用。本文主要研究成果和贡献如下：1）提出质心拉格朗日插值法对PPP钟差和轨道数据进行预处理，该方法只需在拉格朗日插值的形式上稍作改变，简单方便。实验结果表明插值阶数为13时，稳定在毫米级。无论是最大误差还是RMS，Z方向插值效果均优于X、Y方向。2）利用区域CORS站点联合周围IGS跟踪站，研究了基于IGU预报轨道实时估算卫星钟差的方法，并探讨了非差模糊度固定的精密单点定位方法。该方法对于提高PPP精度和可靠性具有一定的参考价值。3）针对PPP的随机模型和函数模型，分别提出了先验信息定权法和EMD修正非差观测值法。该方法能够准确地确定观测值的权重和有效地减弱非建模误差的影响。4）提出PSO优化BP神经网络，改变BP算法依赖梯度信息指导网络的方法来调整权重，利用PSO算法全局性搜索的特点，寻找最为合适的网络连接权值和阈值，提高BP神经网络的泛化能力。BP神经网络具有很强的函数逼近和模式识别能力，通过PSO优化BP神经网络能够提高修正动力学模型和进行异常检测的能力。5）分析PPP历元间差分模型的实时解算精度，当考虑对流延迟的影响，并顾及对流层水平梯度引起的信号延迟。结果表明可以显著消除系统误差，一定程度上提高定位精度。6）分析了精密单点定位用于大面积矿区变形监测的精度，实例计算结果表明经过4小时的连续观测即可获得毫米级的定位精度。该技术因无需建立变形监测基准点，在大区域变形监测应用中有广泛的前景。为了确定监测点的瞬时历元的精确坐标，提出了基于速率信息和实测结果的抗差卡尔曼滤波模型。更多还原

【Abstract】 GNSS precise point positioning (PPP) can determine the three dimensionalcoordinates in the framework of worldwide, and does not require a fixed referencestation which coordinates are known. Therefore, PPP technology can position throghstatic or dynamic model without constrains of baseline length in ITRF. This paperstudy the regional augmentation based GNSS precise point positioning and itsapplication in mining deformation monitoring, major research achievements andcontributions in this paper as follows:1) Barycentric Lagrange interpolation method was proposed for PPP datapreprocessing, and the method is convenient and simple simply, which can beexpressed from the Lagrange interpolation by slightly change. The experimentalresults show that the interpolation order for13, achieved stable millimeter. Either themaximum error or the RMS, and z direction interpolation effects are superior to the x,y direction.2) Joint the regional CORS sites and the nearby International GNSS Service(IGS) stations, the satellite clock error can be estimated real time based on IGUforecast orbit, and the zero difference ambiguity fixed method of precise pointpositioning was discussed. The method has a certain reference value to improve theaccuracy and reliability of PPP.3) The prior information weighting method and the Empirical ModeDecomposition (EMD) correction zero difference value were proposed based on therandom model and function model. The above method can accurately determine theweight of observations and effectively weaken the impact of non modeling error ofPPP.4) The Particle Swarm Optimization (PSO) optimization the neural network wasstudied by changing the gradient information of BP neural network to guide networkadjustment the weights. PSO can improve the generalization ability of BP neuralnetwork based on the PSO algorithm to find the most suitable network connectionweights and thresholds through global searching. Therefore, BP neural network cancorrect dynamic model and anomaly detection by training a large number of learningsamples.5) The real time positioning accuracy of PPP Epoch differenced was analyzed, and when consider delayed effects of convection, and taking into account the level ofthe troposphere gradient caused the delay signal, the results show that systematicerrors can be significantly eliminated, to some extent, improve positioning accuracy.6) The precision of Mining Deformation Monitoring for large area based onPrecise Point Positioning precision was analyzed. The calculation shows that afterfour hours of continuous observation, millimeter positioning accuracy can be obtained.The technology without having to establish a reference point of deformationmonitoring, has broad prospects deformation monitoring applications in a large area.In order to determine the precise instantaneous coordinates of monitoring points, theKalman Filtering model was proposed combined the rate information and themeasured results.更多还原