阵列测向是阵列信号处理的重要分支，具有波束控制灵活、信号增益高及抗干扰能力强的优点，广泛应用于卫星导航、雷达预警和无线电监测等多个领域。阵列测向是利用多天线在时域、频域、空域对信号的幅相信息进行联合分析处理的新测向体制，在多信号、弱信号测向场景中展现了突出优势。然而，随着电磁辐射源空间密度逐渐增大，阵列测向系统所处的信号环境日益复杂，各天线接收到的环境噪声具有强相关性，且存在与目标信号相干的多径干扰，这会导致信号与噪声子空间能量泄露，使得主流子空间分解类算法的测向性能急剧恶化甚至失效。为摆脱对子空间的依赖，学者们提出稀疏重构类算法，利用信号空域稀疏性实现相干信号测向，具备抗多径干扰能力，为阵列测向带来新思路。但是，现有稀疏重构类算法对环境噪声敏感，在低信噪比情况下，存在性能退化严重、计算复杂度高，无法满足高精度、实时性需求的问题。因此，本文选取典型的均匀线阵和圆阵为研究对象，开展了环境噪声及多径干扰下的阵列测向技术研究，主要工作如下：1. 针对线阵测向算法在环境噪声及多径干扰下，信源数估计不准、测向精度差，甚至失效的问题，提出了波束空间差分矩阵加权重构测向算法。充分利用线阵对称结构及其阵列流形的范德蒙特性，构造差分矩阵并在波束空间进行重构实现信号增强，抑制环境噪声及多径干扰影响。仿真表明，相比典型的子空间分解类算法，信噪比为?10dB 时，所提算法的测向精度提升约3倍，角度分辨概率提升约35%。2. 针对圆阵测向算法受环境噪声及多径干扰的影响测向虚警高、精度低的问题，提出了高阶累积切片空间稀疏表示一维测向算法。通过高阶累积切片抑制环境噪声并利用其空域稀疏特征重构避免多径干扰的影响，实现连续域高精度测向。仿真表明，相比典型稀疏重构类算法，信噪比为?10dB时，所提算法的测向精度提升约4倍，角度分辨概率提升约80%。3. 针对圆阵二维测向中阵列测向精度与计算复杂度之间矛盾突出的问题，提出自适应动态感知矩阵二维测向算法。对固定高维矩阵进行动态降维处理，自适应缩小特征重构的空域范围，多分辨迭代逼近真实信号入射方向。仿真表明，相比典型二维测向算法，信噪比0dB 时，所提算法的方位角精度提升近1.5倍，俯仰角精度提升近2倍，计算时间实现数量级减小。

Involved in array signal processing, array direction finding has the advantages of flexible beam control, high signal gain, and strong anti-interference ability. It has been widely used in multiple fields such as satellite navigation, radar warning, and radio monitoring. Array direction finding is a new system that uses multiple antennas to perform joint analysis and processing of signals in the time-frequency and spatial domain, demonstrating outstanding advantages in multi-signal and weak signal direction finding scenarios. However, as the spatial density of electromagnetic radiation sources gradually increases, the signal environment in which the array direction finding system is located is becoming increasingly complex. The environmental noise received by each antenna is correlated and the interference is coherent with the target signal, which cause energy leakage in the subspaces where signal and noise are respectively located and lead to a sharp deterioration or even failure of the performance of mainstream subspace decomposition algorithms. To break away from the dependence on the subspace, scholars have proposed sparse reconstruction algorithms that use the space sparsity of signal to achieve direction finding of coherent signals, with anti-multipath interference ability, bringing new ideas to array direction finding. However, existing sparse reconstruction algorithms are sensitive to environmental noise. In low signal-to-noise ratio conditions, there are significant performance degradation and high computational complexity, making it difficult to meet real-time requirements. Therefore, this thesis selects typical uniform linear arrays and circular arrays as research objects, and carries out high-precision array direction finding technology research in environmental noise and multipath interference, mainly as follows:First, to address the problem that direction finding algorithms for the line array have poor accuracy or even fail in the presence of environmental noise and multipath interference, a direction finding algorithm based on weighted reconstruction of differential matrix in beam-space is proposed. The symmetric structure of the linear array and its Vandermonde array manifold are fully utilized to construct a differential matrix for signal enhancement and to suppress the impact of environmental noise and multipath interference through reconstruction in the beam space. Compared with typical subspace decomposition algorithms, the accuracy of the proposed method is improved by about three times and the angular resolution is increased by about 35% when the signal-to-noise ratio is ?10dB.Second, to address the problem of high false alarm rate and low accuracy of circular array direction finding algorithms under the influence of environmental noise and multipath interference, a direction finding algorithm based on space sparsity representation of high-order cumulative slicing is proposed. By suppressing environmental noise through the high-order cumulative slicing and reconstructing the space sparsity of incident signals to avoid the impact of multipath interference, then high-precision direction finding is achieved. Compared with typical sparse reconstruction algorithms, the accuracy of the proposed method is improved by about four times and the angular resolution is increased by about 80% when the signal-to-noise ratio is ?10dB.Third, to address the prominent contradiction between the precision and the computation complexity of 2D direction finding methods for the circular array, a direction finding algorithm based on an adaptive dynamic sensing matrix is proposed. The algorithm achieves dynamic dimension reduction processing of a fixed high-dimensional matrix, adaptively narrows the spatial range of feature reconstruction, and uses multi-resolution iteration to approximate the real signal direction of incidence. Compared with typical 2D direction finding algorithms, the proposed algorithm improves the azimuth precision by nearly 1.5 times, the elevation precision by nearly 2 times, and reduces the computation time by orders of magnitude when the signal-to-noise ratio is 0 dB.