【作者】 刘海波；

【导师】 谭吉春；

【作者基本信息】 国防科学技术大学 ， 物理学， 2012， 博士

【摘要】 近年深空自主探测和微小卫星技术发展迅速，采用成本相对较低的光学姿态敏感器实现角速度测量，具有很好的应用前景。针对深空自主导航和微小卫星姿态测量的需求，本文研究无陀螺条件下利用星敏感测量角速度的新方法，使得航天器减小对陀螺的依赖。针对星敏感器角速度测量中的关键技术，从角速度估计方法、星点提取和内部参数的校准三个方面开展研究，主要工作包括：⑴基于星敏感器测量的角速度估计方法研究提出了利用星敏感器测量角速度的矢量差分法和恒星跟踪滤波法。矢量差分法采用同一恒星在星敏感器连续两帧图像中的测量矢量的一阶向后差分，构建角速度测量方程。恒星跟踪滤波法利用Singer模型的目标跟踪滤波，获得恒星投影位置和速度信息，构建角速度测量方程，实现角速度估计。与矢量差分法相比，恒星跟踪滤波法无需差分和上一时刻星敏感器的测量信息。针对矢量差分法和恒星跟踪滤波法，设计了自适应卡尔曼滤波器，估计角速度。与现有的基于卡尔曼滤波估计角速度方法不同，该滤波器以当前时刻星图中导航星数目的增加作为时刻变化，更新状态量，获得角速度的最优估计值；不依赖于航天器动力学模型与转动惯量信息，以及星体参数向量或姿态的任何信息。对低轨卫星轨道仿真测试，以及对地球自转角速度测量的结果表明，在星敏感器采样频率10Hz条件下，该滤波器估计角速度的精度可以达到10-6rad/s量级。⑵星点提取方法研究提出了依赖姿态信息和不依赖姿态信息的两种卡尔曼滤波星点提取方法。具体步骤为：根据卡尔曼滤波的预测方程，预测星点位置；然后，在以预测位置为中心的m×m像元星图窗口，提取星体目标，采用质心法获得星点质心位置的测量值；最后，利用设计的卡尔曼滤波器，获得星点位置信息的最优估计。卡尔曼滤波的星点提取方法，可以在减小星点提取位置的随机噪声的同时，提高提取的速度、减小伪星出现的概率。依赖于姿态信息的方法，根据星敏感器的姿态矩阵和姿态角速度，构建预测方程，预测星点位置。不依赖姿态信息的方法，采用Singer模型构建卡尔曼滤波的预测方程，具有不依赖于姿态矩阵和角速度信息的特点，能同时获得星点位置和速度，是恒星跟踪滤波法测量角速度的基础。⑶内部参数的实验室校准提出了一种星敏感器实验室校准的装置和方法。首先，利用激光自准直的方法测量光学中心位置；然后，采用单星模拟器结合二维可调平面反射镜，利用激光反射的方法监测平面反射镜转角，采集校准数据，估计星敏感器安装角、焦距和高阶畸变系数。该校准方法简单、无需高成本的精密转台；在采集星点数据时无需转动星敏感器，可以消除传统实验室校准方法中星敏感器安装误差对精度的限制。误差分析和实验结果表明，校准后的恒星测角误差小于4.0×105rad，通过增加激光反射距离的简单方法，校准的精度还有进一步提高的空间。此外，还提出了一种新的量化的星敏感器调焦方法，该方法借助于上述装置，通过高斯曲面拟合法，测量不同视场、不同离焦位置的成像光斑的高斯半径，进而选择星敏感器“最佳”离焦位置。⑷内部参数的在轨自主校准研究基于姿态估计和基于角距的星敏感器内部参数在轨自主校准方法，并进行了地面观星测试实验。基于姿态估计的校准方法，在进行姿态估计的同时，估计光学中心、焦距以及高阶畸变系数。基于角距的校准方法，利用角距不变原理构建测量方程，设计结合最小二乘迭代法的卡尔曼滤波器，估计内部参数。当导航星数目较少时，基于姿态估计的校准方法内部参数和姿态存在耦合。与已有的扩展卡尔曼滤波校准方法相比，基于角距的校准方法估计结果不受初值的影响，且引入了高阶畸变模型，具有更高的校准精度。对观星数据处理结果显示，校准后的恒星角距误差为105rad量级上述研究方法和结论可为无陀螺角条件下角速度测量提供参考。更多还原

【Abstract】 With the development of autonomous navigation for deep space and smallspacecraft, there is often a desire to do away with gyros and use other means todetermine the angular rate. This dissertation focuses on proposing the novel approachfor gyroless angular rate measurement using a star tracker; and its key techniques, e.g.,angular rate estimation, star spot location and camera parameters calibration, have beendeveloped. The main contents and contribution of this dissertation are as follows:1. Gyrosless angular rate determination from measurement of star trackerTwo approaches (i.e., the difference method and star tracking filter method,respectively) have been put forward to determine the angular rate, while the attitude andstar reference vectors are not required to be known. In the difference method, theangular rate is determined by the backward difference of the vector measurements,which are obtained directly from the star tracker. In the star tracking filter method, theangular rate is determined according to the star spot location and its velocity, which areachieved using the tracking filter of the Singer model. Therefore, the needs fordifference and previous vector measurements of star tracker are eliminated.Adaptive Kalman filters are designed for the difference method and star trackingfilter method respectively. The Kalman filter states are updated on different guide starsin a same star image frame. Thus, the best estimates of angular rate can be achievedwithout the information of dynamic model. The simulations based on the orbit data ofnearly Earth-pointing satellite and experimental test with night sky observation areperformed. Both results indicate that the accuracies of the estimated angular rate are inthe order of10-6rad/s magnitude, when the refresh rate is10Hz.2. Star spot location estimationTwo approaches for star spot location estimation with the Kalman filter (i.e.,attitude-dependent and attitude-independent, respectively) have been proposed, whichconsist of three steps. The approximate locations of the star spots in successive framesare predicted by the Kalman prediction equation firstly; then the measurement locationsare acquired by defining a series of small m×m pixels windows around each predictivelocation. Finally, the star spot location is updated using the designed Kalman filter. Theremarkable advantage of the Kalman filter over other methods for star spot location isthat it combines the predictive location to give more accurate estimates than that byusing the achieved centroids alone. Furthermore, as the threshold scan is only performedin small windows, it takes rather less time and can avoid false star spots effectively.In the attitude-dependent approach, the prediction equation of the Kalman filter isconstructed according to the attitude matrix and angular rate. However, in theattitude-independent approach, the prediction equation is constructed according to the famous Singer model, which can estimate star spot location and velocity directly. It laysthe foundation of the star tracking filter method for angular rate determinion.3. Laboratory calibration for star tracker cameraA simple and available calibration approach for star tracker camera has been putforward. The calibration procedure consists of two steps:(1) the principal point isestimated using autocollimation adjustment;(2) the focal length and distortions areestimated via least-squares iteration, taking into account the extrinsic parameters.Compared with traditional method, in which the calibration data are acquired viarotation of the star tracker installed on a two-axes rotating stage, the star data ofdifferent field angles are acquired via rotation of the adjustable plane mirror in theproposed approach. It could decrease the setting error during the rotation of the startracker and has the advantages of simplicity, low cost, and high accuracy. Thetheoretical analysis and experimental results both indicate that the uncertainties of themeasured star direction vectors are less than4.0×105rad after calibration, which can befurther improved.Besides, an approach for defocus adjustment has been proposed for star tracker.This approach can measure the intensity distribution and spot size of pointsource imagebased on the calibration setup and Gaussian surface fitting method.4. On-orbit autonomous calibration for star tracker cameraThe calibration methods, i.e. attitude-dependent method and inter-star-anglemethod, are introduced to estimate parameters of on-orbit star tracker camera. And thenight sky observation is adopted to test their performance. In the attitude-dependentcalibration method, the camera internal parameters are combined with attitude matrices(or external parameters) determination. The error of attitude estimate will influence theerror of camera parameter estimate and vice-versa. Thus, the calibration accuracy isaffected by the number of stars identified as the control points. In the inter-star-anglecalibration method, a modified version of the least-squares iteration algorithmcombining Kalman filter is put forward on the assumption that the angle between twovectors is invariant. Compared with the traditional EKF (Extended Kalman Filter)method, the estimates are converged even for the poor initial guesses with nonlinearcamera distortions. The experimental results indicate that the deviation of the measuredinter-star angle is in the order of10-5rad magnitude after calibrated.The research method and conclusion would have significance in improving theperformance of gyroless angular rate measurement using star tracker.更多还原