【作者】 黄凯；

【导师】 徐文；

【作者基本信息】 浙江大学 ， 信息与通信工程， 2017， 硕士

【摘要】 随着海洋经济与技术的发展,海底地形探测技术受到了越来越多的关注,并已逐渐成为海洋资源开发、环境调查等工作的基础。较之于传统声纳的单点测量,多波束测深系统利用广角度发射、多通道接收与阵处理技术,可以单次形成高密度的条幅式数据,测深效率大幅提高,在水下资源调查、海洋划界等领域具有重要的应用价值。论文针对多波束测深精度进一步改善的需求,引入地形跟踪的思想并发展了相关算法。首先介绍了多波束测深系统的发展历史和特点以及声线跟踪技术和多波束测深系统的基本原理。由于传统的直接利用声线跟踪来获取海底地形的方法很大程度上会受到测量噪声声速畸变以及干扰回波的影响,本文提出了一种将环境起伏特性与回波到达时间、回波到达角以及声速剖面结合起来的状态-空间模型,并给出了序贯滤波的解法。首先重点阐述了无迹卡尔曼滤波算法的相关理论和算法执行流程,然后通过仿真数据对同一信噪比下序贯滤波算法与传统算法的跟踪性能进行了比较,其次研究跟踪性能与信噪比的关系,比较条件均方误差与后验Cramer-Rao下限的关系,最后结合2016年7月的海试数据,进一步验证了序贯滤波算法的优势。针对序贯滤波算法计算量大、效率低的局限,本文提出一种减少未知参数个数的降维方法,该方法使用经验正交函数表征海底深度特征,显著降低了算法的复杂度。在此基础上,基于经验正交函数的正交性进一步扩展了状态-空间模型,使得表征更为复杂的地形演变成为可能,并运用仿真数据对改进后的算法进行了验证。研究了多波束测深系统的常见误差来源及编辑准则,针对文中涉及的几种算法,引入了基于支持向量机的机器学习方法,训练出二分类模型用于判别算法的优劣并剔除异常值,进行数据融合,实现了测量可靠性与水底跟踪精度的提高。论文还介绍了中科院声学所和浙江大学等机构联合开发的全海深多波束测深系统,阐述了其中基于现场可编程门阵列的信号预处理模块的硬件设计与数字信号处理软件算法,并将运用此系统进行信号处理所得的结果与MATLAB仿真所得数据进行对比,验证了所设计的信号预处理算法的正确性。更多还原

【Abstract】 With the development of marine economy and technology,measurement of seafloor bathymetry has become the foundation of the numerous human ocean activities such as environmental investi-gation,ocean engineering design and resource exploitation.Compared to the single point measure-ment with traditional echo sounder,the multi-beam echo sounder(MBES)generates high density strip depth measurements by exploiting wide-swath directional transmitting,multi-channel receiv-ing and array processing technique.which improves the efficiency of the seafloor bathymetry and is of great application value in fields such as underwater investigation of resources and maritime delimitation.To further improve MBES measurement accuracy,this thesis introduces a new concept of bathymetry tracking and develops relevant signal processing techniques.We start from reviewing the main features and the fundamental theory of the MBES including ray tracing.The tradition-al way of ray tracing for measuring of seafloor bathymetry is always affected by measurement noises and sound speed changes.Considering the environmental variation with time and space,a sequential filtering method is proposed and demonstrated,which constructs a state-space model u-tilizing the relationship among the time-of-arrivals(TOAs),direction-of-arrivals(DOAs)and sound speed profile(SSP)over different pings.The theory and operation procedure of the Kalman filter-ing especially the Unscented Kalman Filter(UKF)is then presented.The tracking performance of the sequential filtering algorithm and the traditional algorithm under the same Signal Noise Ra-tio(SNR)is then compared through the simulation data,and the relationship between the tracking performance and the SNR is then evaluated.The effectiveness of the sequential filtering algorith-m is shown by the comparison of conditional mean square error and posterior Cramer-Rao lower bound.To overcome the problem of large computation and low efficiency with the sequential filtering,a method of unknown parameter dimensionality reduction is proposed,which introduces empiri-cal orthogonal function(EOF)representation for seafloor bathymetry,thus remarkably reduces the complexity.Based on the orthogonality of the EOF,the state-space model is further extended to characterize more complicated seafloor evolutions.The developed approach is validated by the simulation data.Study the common error sources and editing rules of MBES,and a binary classification model based on support vector machine(SVM)is then applied to the referred algorithms to enhance the reliability of the measurement and the precision of the tracking.In this development,the obtained binary classification model is utilized to evaluate the performance of the algorithms and eliminate the exceptional points.At last,a deep sea MBES developed by the institute of acoustics of the Chinese academy of Sciences jointly with some other organizations including Zhejiang University is introduced.The hardware design of the field programmable gate array(FPGA)signal preprocessing module and the relevant digital signal processing algorithm are described,and its implementation is verified via comparison with the MATLAB based simulation.更多还原