UKF滤波算法在数据同化中的应用

介绍了一种新的数据同化算法（UKF，Unscented Kalman Filter），该算法不需要计算伴随矩阵，就能够解决模式的非线性问题。以Lorenz系统为例，进行了数据同化的数值试验。结果表明：基于UKF的同化方案与背景场的初始值无关，它能有效地抑制状态变量误差的增长，同化结果精度高。     

宁波非气象雷达回波的人工智能识别及滤波

统计分析2003—2006年宁波雷达的非气象杂波, 影响杂波主要为地物杂波, 包括普通地物杂波 (NP杂波) 和异常地物杂波 (AP杂波) ; 并分析这些杂波在反射率因子、径向速度、谱宽产品上不连续的离散状分布等特征。基于以上特征及传统杂波识别的经验而设计出一种近似模拟人眼模糊识别的稳定安全的多维线性近似的杂波识别及其过滤算法, 在 190 个个例测试中大都效果较好, 尤其在 150 km 以内, 能够在确保降水数据完整、准确的基础上有效过滤非气象杂波。     

多普勒雷达降水产品优化

通过对雷达1h、3h降水产品进行应用检验，找出降水产品的局限性及其测量误差；利用反射率因子及自动站雨量资料，通过调整雷达混合扫描面弥补了降水产品因山脉等地物挡角大而缺测面大的缺陷；结合雨量站资料运用最优插值、卡尔曼滤波等方法对本地Z-I关系及降水产品进行调整优化；调整后的降水产品估测降水的误差明显减小，产品质量大为提高。     

利用MM5模式输出产品制作空气质量预报方法探讨

根据2004年中尺度数值预报模式MM5输出产品和临沂市环境监测中心逐日监测资料建立了各污染物浓度预报方程,在2005年夏季的试报中,该方程的试报效果较差,其预报准确率明显低于其历史拟合率。为了提高预报准确率,利用逐步回归筛选的因子及统计模型研究中的有关数据,探讨了使用卡尔曼滤波方法制作空气污染物浓度预报的问题。分析发现,利用卡尔曼滤波方法制作空气质量预报可以取得比较满意的效果。     

Integrated Navigation System and Experiment of a Low-Cost and Low-Accuracy SINS/GPS

When SINS (strap-down inertial navigation system) is combined with GPS, the observability of the course angle is weak. Although the course angle error is improved to some extent through Kalman filtering, the course angle still assumes a divergent trend. This trend is aggravated further when using low-cost and low-accuracy SINS. In order to restrain this trend, a method that uses AHRS to substitute for SINS course angle information is put forward aimed at the hardware component characteristic of the low-cost and low-accuracy SINS including AHRS (attitude and heading reference system) and IMU (inertial measurement unit). Real static and dynamic experiments show that the method can restrain the divergent trend of the navigation system angle effectively, and the positioning accuracy is high.     

Comparison of Two Fading Filters anc Adaptively Robust Filter

Two kinds of fading filters and their principles are introduced. An adaptive robust filter is given with corresponding principle. The basic abilities of the fading filters and adaptively robust filter in controlling the influences of the kinematic model errors are analyzed. A practical example is given. The results of the fading filter and adaptively robust filter are compared and analyzed.     

A method to improve the alignment performance for GPS-IMU System

Zero velocity and zero east component of rotation rate relative to local geographic frame have been traditionally applied as measurements for fine alignment of a GPS-IMU system. The performance of the fine alignment, however, will be affected by several types of inertial sensor errors, which could cause part of the Kalman filter states to be unobservable. To overcome this problem, a new method that utilizes the total outputs of the gyro triad and the accelerometer triad as part of the measurements has been proposed by the authors. The initial results have confirmed the effectiveness of the method. In this paper, the observability for both traditional and new alignment methods will be first reviewed. The results from computer simulations will then be presented to compare the traditional and the new alignment methods for the purpose of evaluating the performance of the proposed new method. Data acquired from real inertial sensors will also be applied to assess the traditional and new alignment methods by analyzing their innovation sequences from the Klaman filter. Based on a paper presented at the 18th International Technical Meeting of the Satellite Division of the Institute of Navigation, Long Beach, California, September 2005.     

基于自适应Kalman滤波的BP神经网络及其在高程拟合中的应用

针对传统BP神经网络容易发散、泛化能力差等问题，采用自适应Kalman滤波方法训练神经网络的连接权。与Kalman滤波训练连接权和传统的BP算法相比，该方法提高了BP神经网络计算精度，增强了泛化能力。实测数据的计算结果证明了该方法的可行性与有效性。     

支持向量机与卡尔曼滤波集合的西太平洋副热带高压数值预报误差修正

基于T106数值预报产品资料,提出了支持向量机和卡尔曼滤波相结合的方法来进行夏季西太平洋副热带高压数值预报的误差修正与预报优化。首先采用支持向量机方法建立了西太平洋副热带高压面积指数的误差修正模型。基于支持向量机预报优化模型尽管有比较好的拟合精度和预报效果,但与实际副热带高压指数尚有一定的差异。究其原因,除预报对象(副热带高压)本身比较复杂、模型优化因子不够充分以及数值预报误差自身的随机性以外,优化模型的输入、输出基本上是一个静态映射结构,因此前一时刻的预测误差难以得到有效的反馈、调整和修正。为考虑前一时刻预报误差的反馈信息,动态跟踪副高的变化趋势,随后引入卡尔曼滤波方法建立支持向量机-卡尔曼滤波模型,对支持向量机模型的输出结果作进一步的调整和优化。试验结果表明,该方法模型的预报优化效果优于T106数值预报产品以及单纯的神经网络修正模型和卡尔曼滤波修正模型的优化效果,能够较为客观、有效地修正西太平洋副热带高压指数的数值预报误差,改进和优化西太平洋副热带高压的数值预报效果。该方法为副热带高压等复杂天气系统和要素场预报提供了一种新的思路,表现出较好的应用前景。     

一种应用于城市区域的自适应形态学滤波方法

数学形态学滤波是从激光雷达数据中识别地面点、创建数字高程模型的一种重要方法,在应用中取得了较好的效果,但也具有一些明显不足。在分析现有方法优劣及城市地形特点的基础上,提出一种应用于城市环境的自适应形态学滤波算法。该方法采用分层识别策略：首先通过分割得到地面主体部分;之后利用正规化高度值寻找其余地面区域;最后估计地物覆盖区域地面点取值。采用三个不同分辨率、不同地形/地物特点的数据集进行了实验,结果表明该方法能有效识别地面和地物点,并解决了窗口尺寸限制、粗差误判等问题。