基于稀疏信号动态反演的远距离空间非合作目标形姿识别研究

The shape and the pose are the fundamental properties used to identify the type of space target and to diagnose its working status. The spatial structure and timing structure of distant non-cooperative space target signal in real environment are degenerated so greatly that its observability of the shape and the pose is deteriorated sharply. To overcome this difficulty, the proposal studies the law between the shape and the pose of space target and the intensity and the appearance of target signal, and then proposes the theory and method to estimate the shape and the pose of distant space target based on sparse signal dynamic inversion. The main contents of the proposal are as following: (1) Revealing the mechanism that the shape and the pose of distant space target depredates the spatial structure and the timing structure of target signal;(2)Developing the sparse Bayesian learning theory to evolve the target state and analyzes the coupling rule of shape and pose in target signal;(3)Establishing the method to recognize the shape and the pose of distant space target based on sparse signal Bayesian dynamic inversion to break through the bottlenecks that the performance of the shape estimation and the pose recognition are low under fluctuating noise; (4)Establishing the systematical experimental technology including theoretical calculation, simulation experiment and measured signal experiment to test and validate the mechanism of target signal degradation sparsely and the estimation performance of the shape and the pose of distant space target. The proposal can provide a strong technical support for target recognition and status identification.

空间目标的几何形状和运动姿态是目标类型识别与工作状态诊断的基本依据。实际环境中远距离空间非合作目标信号的空间结构和时序结构信息退化导致其几何形状和运动姿态可观测性急剧恶化。为克服该困难,本项目探索空间目标的几何形状和运动姿态与目标信号强度和形态时间曲线的演变规律，建立基于稀疏信号动态反演的远距离空间目标形姿识别理论与方法。研究内容涵括：(1)揭示远距离空间目标的几何形状和运动姿态稀疏退化目标信号机理；(2)发展稀疏贝叶斯学习理论演化目标信号状态变化规律，剖析几何形状和运动姿态对目标信号的耦合规则； (3)建立贝叶斯动态反演远距离空间目标形姿识别方法，突破起伏噪声下目标形姿估计准确度和分辨力低的瓶颈； (4)建立理论计算、模拟试验和实测信号相结合的试验技术，有步骤地测试和验证目标信号稀疏退化机理和空间目标形姿反演性能。本项目的成功实施能为空间目标的类型识别和运行状态辨识提供有力的支撑。

空间目标的几何形状和运动姿态是目标类型识别与工作状态诊断的基本依据，是空间环境感知的基本任务，对维护空间安全具有不可或缺的地位。尤其是自旋或翻滚运动，空间非合作目标信号的空间结构和时序结构信息退化，这导致其几何形状和运动姿态可观测性急剧恶化。针对空间目标形姿测量的难题，本项目主要研究有：（1）采用数值仿真和试验验证方式，分析了湍流效应、运动姿态和环境噪声弱化空间目标信号形姿的严重程度，构建了蕴含几何形状和运动状态的空间目标信号空时结构原子及其超完备字典。空时结构原子体现了空间目标信号状态的差异度，也是反演空间目标信号重要的正则项。（2）通过实测数据分析了空间目标信号稀疏系数在时空域上相关性和连续性，以空间目标信号稀疏系数的相关系数进行空时稀疏贝叶斯建模和演化目标信号状态变化规律。该方法提高了空间目标信号稀疏表征程度，继而通过更具代表性的非结构化原子动态刻画空间目标的几何形状和运动姿态。（3）针对弱化后的空间目标信号的几何形状和运动姿态求解非唯一性难题，发展了基于信号空时稀疏先验的空间目标图像盲反演方法，提高了目标信号细节信息；采用广义高斯函数构建超完备字典，高效表征与提取空间目标结构参数，提高目标时序结构稀疏表征准确度。同时，从反演后的图像序列中联合推演空间目标的几何形状和运动姿态，即在通过特征点重建三维几何结构的基础上估计目标旋转运动姿态。本项研究中空间目标信号空时稀疏性建模，以及稀疏贝叶斯提取和演化空间目标的形状与姿态等成果，有望提升翻滚姿态变化大的空间目标几何形状和运动姿态估计性能，提高空间目标感知能力。