基于快速稀疏贝叶斯学习算法的雷达数据融合技术研究

发布时间：2018-04-22 08:52

本文选题：多雷达数据融合成像 + 信号稀疏表示　；参考：《南京理工大学》2014年硕士论文

【摘要】：由于单雷达成像系统的分辨率受到信号带宽和相干积累角的约束,近年来,多雷达数据融合技术作为一种新兴的雷达成像技术,在军事上得到了重视并拥有着广阔的应用前景。多雷达数据融合技术是一种综合不同视角、不同频带雷达回波数据,利用信号处理的方法获得高精度目标模型参数的技术,它突破了单雷达分辨率的约束,在成像过程中可获得更高分辨率的清晰图像。本文主要讨论了高频区雷达的数据融合问题。利用几何绕射理论模型,雷达数据融合问题可以转化为信号稀疏表示问题,信号稀疏表示方法作为一种有效的数据分析方法,将其应用于雷达成像处理中,可准确估计出目标散射中心参数,大幅提高最终成像质量,便于后续的分析和处理。本文主要包括以下四部分内容：第一部分主要介绍了雷达数据融合技术的理论基础,包括目标电磁散射模型的建立,信号稀疏表示的相关理论。第二部分详细说明了同视角多频带雷达数据融合技术。首先给出了一维雷达回波的信号稀疏表示模型,然后针对多子带观测情况进行了分析,选择使用稀疏贝叶斯学习方法求解信号稀疏表示问题,并分别详细介绍了期望最大化方法、求导方法和快速边缘似然函数最大化方法三种求解超参数的方法。第三部分主要分析了一种基于信号稀疏表示的相干配准方法。在第一部分信号稀疏表示相关理论的基础上,基于幅相补偿参数的稀疏特性,对引起两部雷达之间不相干的固定相移和线性相移进行估计,在提高了估计精度的同时使算法更具鲁棒性。第四部分重点介绍了多视角多频带雷达数据融合技术。在给出目标散射场二维模型的基础上,构建了多视角多频带雷达数据的信号稀疏表示模型,最终用稀疏贝叶斯学习方法对该信号稀疏表示问题进行求解,并用仿真算例验证了该方法的有效性。
[Abstract]:Because the resolution of single radar imaging system is constrained by signal bandwidth and coherent accumulation angle, in recent years, as a new radar imaging technology, multi-radar data fusion technology has been paid attention to in military and has a broad application prospect. Multi-radar data fusion technology is a kind of technology that synthesizes radar echo data of different visual angle and different frequency band and obtains high precision target model parameters by signal processing method. It breaks through the constraint of single radar resolution. A clear image with higher resolution can be obtained during the imaging process. This paper mainly discusses the data fusion of high frequency radar. Using the geometric diffraction theory model, the radar data fusion problem can be transformed into the signal sparse representation problem. As an effective data analysis method, the signal sparse representation method is applied to radar imaging processing. The scattering center parameters of the target can be estimated accurately, and the final imaging quality can be greatly improved, which is convenient for subsequent analysis and processing. This paper mainly includes the following four parts: The first part mainly introduces the theory foundation of radar data fusion technology, including the establishment of target electromagnetic scattering model and the theory of signal sparse representation. In the second part, the data fusion technology of multi-band radar with same view angle is described in detail. Firstly, the signal sparse representation model of one-dimensional radar echo is given, then the multi-subband observation is analyzed, and the sparse Bayesian learning method is chosen to solve the signal sparse representation problem. Three methods to solve the hyperparameter are introduced in detail, including the expected maximization method, the derivative method and the fast edge likelihood function maximization method. In the third part, a coherent registration method based on sparse signal representation is analyzed. Based on the correlation theory of signal sparse representation and the sparse characteristic of amplitude and phase compensation parameters, the stationary phase shift and linear phase shift which cause incoherence between two radars are estimated. The estimation accuracy is improved and the algorithm is more robust. The fourth part focuses on the multi-view multi-band radar data fusion technology. Based on the two-dimensional model of target scattering field, the signal sparse representation model of multi-view and multi-band radar data is constructed. Finally, the sparse Bayesian learning method is used to solve the signal sparse representation problem. The effectiveness of the method is verified by a simulation example.
【学位授予单位】：南京理工大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TN957.52

【参考文献】