High Resolution Bottom Detection For Multi-beam Echo Sounder:Algorithm Study And System Implementation

Information on bottom topography is the foundation of numerous human ocean activities including marine environmental investigation, resource exploitation and navigation security. The goal of a bathymetry measurement system is to measure the water depth and further map the seabed topography. As such a high-tech instrument, multi-beam echo sounders (MBES) can generate high-density strip depth measurements by exploiting wide-swath directional transmitting and multi-channel receiving. During the past 50 years, the MBES technology has been improved significantly; it continues to improve toward even wider coverage, higher resolution and better precision.As an underwater acoustic system, performance of the MBES is ultimately determined by the physical mechanism of the measurement process. The backscattering signal of the seabed is continuously distributed in both space and time. Compared to the case of point objects studied in conventional sonar applications, detection and estimation of signal parameters in MBES are more complex, and the associated performance analysis is even more challenging.The bottom detection process in MBES is a hypothesis testing problem with unknown parameters, i.e., signal detection is jointly done with parameter estimation. On one hand, data samples, for which the hypothesis testing is true, are needed in parameter estimation; on the other hand, detection is supported by effective estimation results. Here, the target is the seabed, and the problem facing includes target direction-of-arrival (DOA) and time-of-arrival (TOA) estimation, backscattering signal detection, and model-based change detection.This thesis is to improve the depth accuracy of the outer beams and the reliability of the measurement results, using the shallow-water MBES system funded by the national 863 program as the research platform. Through an interactive linkage among detection, estimation and model, considering main factors impacting the measurement lines (the position and depth of measuring points, and the seabed model), the thesis research focuses on three key aspects of MBES signal processing:beamforming. bottom detection, and model-based bottom change detection.Firstly, a seabed backscattering signal model is developed. For a shallow-water environment, the signal in the near distance propagates in the form of a spherical wave; in this case, use of the plane wave model will result in some deviation of the measured position. As such, a combination of the near-field focused beamforming and plane wave beamforming is presented to make the beams pointed more precisely and improve the localization accuracy. Considering the large amount of calculation required for focused beamforming, the thesis proposes an approximate algorithm. The difference between the approximate and the ideal beam patterns is discussed; and simulations validate the effectiveness of the approximate algorithm.The depth of measuring point is determined by DOA and TOA estimation. The estimation of DOA can be obtained from the beam angles; however, broadening of the footprints of the outer inclined beams reduces the resolution significantly and affects the accuracy of depth estimation. In fact, the topographic relief within the footprints of the inclined beams makes the echo present point-like features. Hence the high resolution algorithms can be used to resolve simultaneously echo signals and obtain a sub-beamwidth resolution. In this thesis, estimation of signal parameters via rotational invariance techniques (ESPRIT) with multiple-angle subarray beamforming is developed to reduce the requirement for the signal-to-noise ratio (SNR), the number of snapshots, and computational efforts. Experimental data processing results show that, compared to the traditional bottom detection algorithms such as weighted mean time (WMT) and bearing deviation indicator (BDI), ESPRIT with multiple-angle subarray beamforming can provide better fine-structure resolution.On the other hand, because of the system coverage/resolution requirement and the space constraint of installation, an MBES usually employs a receiving array of special shape, which leads to that the high resolution algorithms based on a uniform linear array (ULA) can not be applied directly. Meanwhile the short-term stationary of the echo signals makes it difficult to estimate the covariance matrix required, further increasing the complexity in applying a high-resolution algorithm. In this thesis, a high resolution bottom detection algorithm is developed, which can be used on a planar array of any shape, by combining the ESPRIT with multiple-angle subarray beamforming with virtual array transformation and multiple subarray forward-backward spatial smoothing. Computer simulations and experimental data processing results verify the effectiveness of the developed algorithm.The DOA and TOA estimation can be converted to the horizontal distance and depth of a measuring point. Due to the impact of operational errors and external interferences, there are outliers in the results of bottom detection, while the uneven distribution of measuring points is not conducive to obtaining a continuous bottom profile. In the thesis, the LS-SVM (least squares support vector machine) approach is studied to model the bottom profile. Based on this model, the outliers are removed to further improve the fitting precision. Experimental data processing results show that the model can not only fill the measurement gap, but also achieve the fusion of the results with a variety of algorithms, and improve the reliability of the results.During the thesis research, two kinds of bottom detection algorithms—WMT and BDI have been designed on a digital signal processor for the MBES. Numerous tank tests, lake and sea trials have demonstrated that the signal processor and its algorithms can work effectively and reliably.