面向水下被动定位的稳健1-bit压缩感知关键算法研究

With the noise of underwater vehicles being reduced quickly, it is increasingly difficult for sonar equipment to detect and locate underwater targets under real ocean conditions. Therefore, new detection system and signal processing method are urgently needed to cope with this challenge. 1-bit compressive sensing has a remarkable application prospective, because of the potential to improve the shortcomings of the current passive sonar system of low positioning accuracy, close detection distance, and poor environment and noise robustness. The technique is also consistent with the trend of development of next generation underwater-locating systems: distributed, networking and stereoscopic. To this end, this project will focus on the applications of 1-bit compressive sensing in underwater passive localization research. First, put forward robust reconstruction algorithm for 1-bit compressive sensing with the oceanic noise to improve the reconstruction accuracy of the sparse coefficient. Second, propose the method of using the 1-bit compressive sensing algorithm for the DOA estimation and matching field localization under complex-value and multi-snapshot conditions. The array configuration conditions for target location are given and the positioning performance of the algorithm is analyzed. Third, carry out the studies on the adaptive meshing method and the off-grid method to improve both the localization accuracy and the computing efficiency. At last, carry out the verification experiment of water tank to optimize the sound localization method according to the experimental results. The project can provide technical support to improve the locating ability for underwater sonar in real ocean environment.

随着水下航行器降噪和消声技术的迅速发展，真实海洋条件下声纳设备对水下目标的检测和定位难度日益增加，因此迫切需要新的探测体制以及信号处理方法来应对这一挑战。1-bit压缩感知技术能够改进当前被动声纳系统定位精度低，探测距离近，对噪声和环境变化稳健性差的不足，符合下一代水下定位系统大规模、分布式、网络化和立体化的发展趋势，具有良好的应用前景。本项目将围绕1-bit压缩感知在水下被动定位中的应用这一具体问题开展研究，研究1-bit压缩感知在真实海洋噪声条件下的稳健重构算法，提高稀疏系数的重构精度；给出复数值和多快拍条件下利用1-bit压缩感知算法进行波达角估计和匹配场定位的方法，给出实现目标定位的阵列配置条件并分析算法的定位性能；开展自适应网格划分和网格失配校正算法研究，进一步提高水下目标定位的精度与计算效率；开展水池实验，结合实验数据进行算法验证和优化，为提高声纳的定位能力提供技术支撑。

本项目主要研究了1bit压缩感知在水下目标被动定位中的基础理论与应用方法。目标声源在水下空间中形成的稀疏特性是一种非常重要的先验信息，合理而充分地应用该信息能够突破传统阵列信号处理技术在应用方式和探测精度上的局限性；而1bit采样方法能够极大地降低系统的复杂性和成本，因此在分布式探测系统中具有很大的应用前景。基于以上背景，本项目探索了1bit压缩感知在复杂水声环境下的应用条件及方法，围绕与1bit压缩感知及水下被动定位相关的关键技术，如信号的压缩重构理论、优化模型与算法、展开式深度学习方法等基础理论开展了深入分析和研究。项目提出了复杂噪声条件下的稳健1bit压缩感知重构算法，研究了基于1bit压缩感知重构技术的水下被动定位方法与实现条件，研究了结合深度学习算法的1bit压缩感知重构算法与定位方法等，这些研究成果将对声纳装备的设计与应用具有理论参考与实际指导作用。