复杂工况背景下星图预处理相关技术研究

发布时间：2018-03-25 19:36

本文选题：星敏感器　切入点：工况背景　出处：《中国科学院研究生院(光电技术研究所)》2016年博士论文

【摘要】：航天器姿态的高精度测量是实现航天器姿态高精度控制的重要前提。而星敏感器在众多姿态敏感器中具有最高的定姿精度,具备广阔的应用前景。对星敏感器在复杂工况背景下星图预处理相关技术的研究具有重要的意义。本文的研究内容主要分为五个方面:星点的高精度质心定位方法研究、基于积分时间调整策略的动态星点成像方法的研究、高动态条件下自主星跟踪方法的研究、复杂工况背景下星点提取算法的研究及其实现。借助傅里叶分析,从理论角度分析质心法及其改进形式的系统误差,证明了系统误差的像元周期性。分析像元尺寸、质心提取窗口、星点质心位置及加窗的形式等因素对系统误差的影响。不考虑噪声影响时,选择尺寸更大的质心提取窗口可以有效地减小系统误差。考虑噪声因素对质心定位精度的影响,理论推导了高斯白噪声模型下和泊松噪声模型下质心定位误差的理论下界-克拉美劳下界。通过数值仿真分析得出:低信噪比条件下Gamma运算有于提高质心定位精度,减阈值操作能有效抑制背景光的干扰;而在高信噪比条件下限制质心定位精度进一步提高的误差来源是系统误差。从星敏感器的光学系统模型出发,推导了绕星敏感器坐标系三轴角速度影响下星点质心在探测器平面上的运动轨迹。结合静态星点的成像模型,建立高动态条件下星点的成像模型。动态条件下,绕星相机坐标系X轴方向的角速度驱使星点质心沿Y轴方向滑动;绕星相机坐标系Y轴方向的角速度驱使星点质心沿X轴方向滑动;而绕星相机Z轴的角速度驱使星点质心沿轴某段圆弧滑动。星点的最大灰度值与三轴角速度和星相机积分时间相关。当结合三轴角速度大小调整星敏感器曝光时间使得质心在探测器平面上的滑动距离大约为一个星像斑宽度时,星点的灰度值的峰值能量达到最大值,同时能有效地减小动态拖尾现象。建立星点质心在星敏感器探测器平面上的运动模型。为满足高动态条件下的星跟踪需求,在差分星跟踪方法的基础上提出融入星点运动的加速度及急动度等信息完善运动模型,在高动态条件下准确地预测跟踪窗口的位置。在不依赖于先验姿态信息的前提下,基于扩展卡尔曼滤波提出对应的位置预测器。通过仿真分析,实现了10/的大角速度下的星跟踪。通过观星试验表明,本文方法至少能满足4.0/的大角速度条件下的星跟踪需求。分析月光视场、高动态成像等工况背景下星图的特征,提出基于星像素灰度分布特征的星点提取方法。与全局阈值法和局部自适应阈值法分割星像素相比,本文方法仅需一次星图扫描即可完成星像素提取。提出采用双边阈值分割星像素以克服强光辐射的图像区域背景的复杂变化,采用区域生长的策略提取超过星像素提取窗口的星像素,并依据种子点的位置正确地实现星像素灰度值和坐标值的乘、乘累加及求和等操作。充分利用FPGA可并行处理的特点,完成复杂工况背景下星点提取算法的硬件实现。在星像素输出的同时提取星像素、合并目标、统计目标面积和能量强度。同时,执行像素灰度值乘累加等操作。一帧星图输出完毕即可执行质心计算,满足星图预处理实时性的需求。
[Abstract]:High precision measurement of spacecraft attitude is an important prerequisite for achieving high precision spacecraft attitude control. The star sensor has the highest accuracy among the attitude sensor, has broad application prospects. It is important to study the star sensor in complex conditions under the background of star image preprocessing technologies. The main contents of the thesis five aspects: the research of high precision star centroid localization method, study the dynamic star imaging method of integral time adjustment strategy based on the study of star tracking method under high dynamic condition, complicated working condition under the background of the research and implementation of star extraction algorithm. By means of Fourier analysis, from the theoretical point of view of the centroid method and system error improved form, prove the pixel periodic system error. Analysis of pixel size, centroid extraction window, centroid position and window form The influence factors of the system error. Without considering the noise effect, select a size larger centroid extraction window can effectively reduce the system error. Considering the noise influence on the precision of centroid location, theoretical derivation of the lower bound theory error of centroid localization Gauss white noise model and Poisson noise model under the Cramer Rao lower bounds. By numerical simulation analysis shows that under the conditions of low SNR Gamma operations are to improve the centroid accuracy, reduce the interference threshold operation can effectively restrain the background light; while in high SNR conditions limit the centroid accuracy and further improve the system error. The error source is starting from the optical system of star sensor model, derivation around the star sensor coordinate system three axis angular velocity under the influence of centroid trajectory in the detector plane movement. Combined with the static model of star imaging, established under the conditions of high dynamic satellite The imaging model. The dynamic condition, the angular velocity around the star camera coordinate system X axis driven by centroid along the Y axis sliding; angular velocity around the star camera coordinate system Y axis driven by centroid along the X axis sliding around; angular velocity star camera Z axis drive along the axis of a centroid arc sliding. The maximum gray value associated with the star three axis angular velocity and star camera integration time. When combined with the three axis angular velocity to adjust the size of the star sensor exposure time makes the centroid sliding distance in the detector plane is about a star like spot width, peak energy star gray value reached the maximum value at the same time can effectively reduce the dynamic trailing phenomenon. Centroid motion model is built in the detector plane of star sensor. In order to meet the dynamic conditions of star tracking requirements, put into operation in star based on star tracking method The acceleration and jerk information improve the motion model of dynamic, accurate prediction of the tracking window position in high dynamic conditions. In the premise does not rely on a priori attitude information, based on the extended Calman filter is proposed. The corresponding position predictor through simulation and analysis, to achieve the 10/ angular velocity under the star tracking. By viewing test this paper shows that the method can meet the angular velocity of at least 4.0/ condition. The demand analysis of moon star tracking field, characteristics of star high dynamic imaging conditions such as background, extraction method of star pixel gray distribution features of the star is proposed. Compared with the global threshold and local adaptive threshold segmentation method of star pixel, this method only needs a map scan can be completed. The complex changes of star pixel extraction using bilateral threshold segmentation to overcome image pixel star light radiation background, by Region The growth strategy to extract more than star star pixel pixel extraction window, and on the basis of the seed point in the right position to achieve star pixel value and the coordinate values of multiplication, multiplication and accumulation and summation operation. Taking advantage of FPGA parallel processing, complete the complex working conditions under the background of star extraction algorithm for hardware implementation. Star pixels. To extract the target in the star pixel output at the same time, the target area and energy intensity. At the same time, implementation of locomotive operation. The cumulative pixel frame chart output can be executed after the centroid calculation, to meet the needs of real-time star image preprocessing.

【学位授予单位】：中国科学院研究生院(光电技术研究所)
【学位级别】：博士
【学位授予年份】：2016
【分类号】：TP391.41

【参考文献】