基于MRF场的侧扫声呐图像分割方法

为了利用侧扫声呐进行水下目标自动探测和识别,首先必须将声呐图像分为目标高亮区、海底混响区和目标阴影区.由于声呐图像有强背景噪声,传统的图像分割方法显得无能为力,故采用基于MRF场的图像分割方法来准确地分割.根据侧扫声呐目标的成像特点,建立了分割的约束条件;利用阴影与目标的灰度均值比很小这一特点进行初始分割,然后根据分割后目标与阴影的宽度差来剔除虚假目标,由初始分割的结果求得MRF模型初始参数,再采用迭代条件估计得到最终的模型参数和准确的分割结果.由于考虑了相邻像素间的依赖关系,具有抗噪性强、分割效果好的优点,从理论上说是合理的.实测数据分析也证明了这种算法的优越性.     

基于MRF与引导滤波的声纳图像去噪方法

声纳具有对大范围水下场景探测的能力,一直以来都是水下设备感知外界环境的重要手段,但由于声纳图像分辨率低,海洋环境噪声干扰较为复杂,所以在诸多方面的应用都受到了限制。文中提出了一种基于马尔可夫随机场和引导滤波的声纳图像去噪与增强方法,使用马尔可夫随机场对声纳图像进行预分割,然后采用中值滤波方法对原始图像进行简单滤波处理,最后将该图像作为引导图像对马尔可夫随机场(Markov Random Field,MRF)分割后的图像进行引导滤波实现了对声纳图像的去噪与增强。该方法有效地去除了背景和影子内的噪声,对目标区域内部噪声起到了很好的抑制作用,消除了MRF分割产生的伪轮廓效应,具有较好的边界保持和增强效果。     

Wavelet analysis of bathymetric sidescan sonar data for the classification of seafloor sediments in Hopvågen Bay - Norway

In this paper we examine the use of bathymetric sidescan sonar for automatic classification of seabed sediments. Bathymetric sidescan sonar, here implemented through a small receiver array, retains the advantage of sidescan in speed through illuminating large swaths, but also enables the data gathered to be located spatially. The spatial location allows the image intensity to be corrected for depth and insonification angle, thus improving the use of the sonar for identifying changes in seafloor sediment. In this paper we investigate automatic tools for seabed recognition, using wavelets to analyse the image of Hopvågen Bay in Norway. We use the back-propagation elimination algorithm to determine the most significant wavelet features for discrimination. We show that the features selected present good agreement with the grab sample results in the survey under study and can be used in a classifier to discriminate between different seabed sediments.     

Mapping benthic habitat in regions of gradational substrata: An automated approach utilising geophysical,geological, and biological relationships

The relationship between acoustic backscatter, sediment characteristics and benthic habitat is examined using high-resolution sidescan sonar data collected at the Loch Linnhe artificial reef site on the west coast of Scotland. The site is typical for the continental shelf of NW Europe, with a mix of seabed environments from muddy to coarse, stony substrata on a 10–100 m length scale. A sidescan sonar mosaic was produced and classified according to derived backscatter parameters (mean, median and standard deviation of the backscatter values) using an unsupervised classification procedure. The accuracy of the final classified map was assessed by comparison with a ground-truthing survey in which the biological habitat was derived from underwater video footage. The sidescan correctly predicted seabed surface characteristics of observed biological habitat with 78% accuracy. A second, and more challenging test of the acoustic data to correctly predict biological habitat was made by comparing it with data from 21 grab sampling stations. These stations were divided into three groups using multivariate statistical techniques based on their backscatter properties. Benthic assemblage structure was found to be significantly distinct between the high and low, and the medium and low backscatter stations. There was a low to moderate but significant correlation between the multivariate patterns of acoustic backscatter, benthic assemblage structure, and particle size distribution. The work shows that even in areas with subtle and gradational changes in substratum, the sidescan was able to predict biological community with an acceptable accuracy.     

Geometric corrections on sidescan sonar images based on bathymetry. Application with SeaMARC II and Sea Beam data

Acoustic backscatter images of the seafloor obtained with sidescan sonar systems are displayed most often using a flat bottom assumption. Whenever this assumption is not valid, pixels are mapped incorrectly in the image frame, yielding distorted representations of the seafloor. Here, such distortions are corrected by using an appropriate representation of the relief, as measured by the sonar that collected the acoustic backscatter information. In addition, all spatial filtering operations required in the pixel relocation process take the sonar geometry into account. Examples of the process are provided by data collected in the Northeastern Pacific over Fieberling Guyot with the SeaMARC II bathymetric sidescan sonar system and the Sea Beam multibeam echo-sounder. The nearly complete (90%) Sea Beam bathymetry coverage of the Guyot serves as a reference to quantify the distortions found in the backscatter images and to evaluate the accuracy of the corrections performed with SeaMARC II bathymetry. As a byproduct, the processed SeaMARC II bathymetry and the Sea Beam bathymetry adapted to the SeaMARC II sonar geometry exhibit a 35m mean-square difference over the entire area surveyed.On leave at the Naval Research Laboratory, Code 7420, Washington D.C. 20375-5350.     

Active learning for detection of mine-like objects in side-scan sonar imagery

A data-adaptive algorithm is presented for the selection of the basis functions and training data used in classifier design with application to sensing mine-like targets with a side-scan sonar. Automatic detection of mine-like targets using side-scan sonar imagery is complicated by the variability of the target, clutter, and background signatures. Specifically, the strong dependence of the data on environmental conditions vitiates the assumption that one may perform a priori algorithm training using separate side-scan sonar data collected previously. In this paper, a novel active-learning algorithm is developed based on kernel classifiers with the goal of enhancing detection/classification of mines without requiring an a priori training set. It is assumed that divers and/or unmanned underwater vehicles (UUVs) may be used to determine the binary labels (target/clutter) of a small number of signatures from a given side-scan collection. These sets of signatures and associated labels are then used to train a kernel-based algorithm with which the remaining side-scan signatures are classified. Information-theoretic concepts are used to adaptively construct the form of the kernel classifier and to determine which signatures and associated labels would be most informative in the context of algorithm training. Using measured side-looking sonar data, the authors demonstrate that the number of signatures for which labels are required (via diver/UUV) is often small relative to the total number of potential targets in a given image. This procedure designs the detection/classification algorithm on the observed data itself without requiring a priori training data and also allows adaptation as environmental conditions change.     

Sidescan Sonar Segmentation Using Texture Descriptors and Active Contours

This paper is concerned with the application of active contour methods to unsupervised binary segmentation of high-resolution sonar images. First, texture features are extracted from a sidescan image containing two distinct regions. A region-based active contour model of Chan et al. [J. Vis. Commun. Image represent, vol. 11, pp. 130-141,2000] is then applied to the vector-valued image extracted from the original data. Our implementation includes a new automatic feature selection step used to readjust the weights attached to each feature in the curve evolution equation that drives the segmentation. Results are shown on simulated and real data. The influence of the algorithm parameters and contour initialization are also analyzed.     

侧扫声纳系统在海底障碍物扫测中的应用

针对海道测量中侧扫声纳位置精度低的问题,利用GPS定位数据、单波束水深数据和侧扫数据,研究侧扫声纳扫海测量中海底障碍物信息精化问题。海上试验数据表明：文中提出的多源水声测量信息联合扫测海底障碍物的方法,可改善单一侧扫系统目标探测的精度,对海底目标物扫测具有较好的适用性。     

Postprocessing and corrections of bathymetry derived from sidescansonar systems: application with SeaMARC II

A procedure for postprocessing bathymetry data provided by a phase-measuring sidescan sonar system is presented. The data were collected with the SeaMARC II system, and are generally characterized by a high level of noise and uneven spatial sampling. Before any spatial filtering is applied, data are selected to remove most of the obvious artifacts and to retain instantaneous depth profiles whose slant ranges increase monotonically from a central location to the edges of the swath. An extrapolation scheme, patterned after a potential field, is proposed to fill gaps in the coverage or to extend the bathymetric swath to that of the corresponding sidescan image when regridding the data to a rectangular frame. To fill the near nadir gap typically found in these data, a specific interpolation methodology is developed that takes into account the slant range of the first bottom return as received by the sidescan sonar itself or by a shipboard echo-sounder. Spatial low-pass filtering is applied through convolutions with parabolic windows whose width is proportional to the footprint of the acoustic beam along track and roughly 1/8 of the swath width across track. Mismatches of contour lines between adjacent tracks are reduced through a statistical method design to correct systematic profile errors     

Concurrent mapping and localization using sidescan sonar

This paper describes and evaluates a concurrent mapping and localization (CML) algorithm suitable for localizing an autonomous underwater vehicle. The proposed CML algorithm uses a sidescan sonar to sense the environment. The returns from the sonar are used to detect landmarks in the vehicle's vicinity. These landmarks are used, in conjunction with a vehicle model, by the CML algorithm to concurrently build an absolute map of the environment and to localize the vehicle in absolute coordinates. As the vehicle moves forward, the areas covered by a forward-look sonar overlap, whereas little or no overlap occurs when using sidescan sonar. It has been demonstrated that numerous reobservations by a forward-look sonar of the landmarks can be used to perform CML. Multipass missions, such as sets of parallel and regularly spaced linear tracks, allow a few reobservations of each landmark with sidescan sonar. An evaluation of the CML algorithm using sidescan sonar is made on this type of trajectory. The estimated trajectory provided by the CML algorithm shows significant jerks in the positions and heading brought about by the corrections that occur when a landmark is reobserved. Thus, this trajectory is not useful to mosaic the sea bed. This paper proposes the implementation of an optimal smoother on the CML solution. A forward stochastic map is used in conjunction with a backward Rauch-Tung-Striebel filter to provide the smoothed trajectory. This paper presents simulation and real results and shows that the smoothed CML solution helps to produce a more accurate navigation solution and a smooth navigation trajectory. This paper also shows that the qualitative value of the mosaics produced using CML is far superior to those that do not use it.     

基于BP网络对模拟声呐信号分类

针对常规的主动声呐调查设备,在简单海洋分层模型的基础上,模拟了多波束类单频信号、侧扫类单频信号、Ch irp调频信号和混合信号4类声呐接收信号,并针对接收信号特征构造了3层BP网络模型,将隐藏层神经元数目设为可调节;利用时间域脉冲宽度和水深与频率域功率谱密度相结合的特征参量,成功地对模拟信号进行了分类。采用改进的BP网络模型,用训练成功的BP网络对102个检测信号进行了分类测试,结果表明,分类成功率较高,可达76%~84.6%,因而利用BP网络可以对不同类别设备的模拟声呐接收信号进行分类。     

Shadow Enhancement in Synthetic Aperture Sonar Using Fixed Focusing

A shadow cast by an object on the seafloor is important information for target recognition in synthetic aperture sonar (SAS) images. Synthetic aperture imaging causes a fundamental limitation to shadow clarity because the illuminator is moved during the data collection. This leads to a blend of echo and shadow, or geometrical fill-in in the shadow region. The fill-in is most dominant for widebeam synthetic aperture imaging systems. By treating the shadow as a moving target and compensating for the motion during the synthetic aperture imagery, we avoid the geometrical shadow fill-in. We show this to be equivalent to fixing the focus at the range of the shadow caster. This novel technique, referred to as fixed focus shadow enhancement (FFSE) can be used directly as an imaging method on hydrophone data or as a postprocessing technique on the complex SAS image. We demonstrate the FFSE technique on simulated data and on real data from a rail-based SAS, and on two different SAS systems operated on a HUGIN autonomous underwater vehicle.      

基于声呐图像的深海采矿车组合定位导航算法

实现高精度的定位导航是深海采矿车完成海底工作任务的基础条件。在采矿车行进过程中,声呐设备生成的图像信息能够反映海底场景的变化,从而体现采矿车本身的运动,由此建立了一种声呐图像里程计,并将其与轮式里程计和USBL测量数据相结合提出了一种深海采矿车组合定位导航算法。首先对多波束前视声呐图像进行预处理,然后使用Canny算法进行特征检测并对特征点云进行配准,再结合声呐成像原理构建了声呐图像里程计运动模型,最后通过轮式里程计运动模型推导预测方程、声呐图像里程计运动模型和USBL测量数据推导更新方程,利用EKF（extended Kalman filter）算法实现基于多传感器融合的定位与姿态估计。海试数据验证了该组合定位算法能实现轮式里程计、声呐里程计和超短基线在速度、位置、艏向角估计、定位速率的精度互补,具有一定的有效性和精确性,该算法为深海采矿车的定位与导航算法研发提供了参考。     

基于DeepLabv3架构的高分辨率遥感图像分类

针对目前使用机器学习解决高分辨率遥感图像分类主要存在下采样导致的细节信息丢失问题,提出了一种基于DeepLabv3架构的小波域DeepLabv3-MRF(Markov random field,MRF)算法。选择当前较为普遍的DeepLabv3架构分类算法,能够获得更为精确的分类结果;采用小波域DeepLabv3-MRF算法,还能够获得更为清晰的边缘细节信息。选取南方某地区高分辨率无人机遥感图像进行分类实验,通过小波变换的方向性、非冗余性以及MRF变换像素空间的交互性这三个方面,将分类结果与原始DeepLabv3架构分类结果对比分析。结果表明,所提出的分类方法精度明显高于原始DeepLabv3架构分类算法的精度,总体精度可提升3%左右,并且可以充分表达高分辨率遥感图像细节信息。     

TOBI image processing-the state of the art

TOBI (Towed Ocean Bottom Instrument) is a deep-tow sidescan sonar vehicle from which sidescan sonar data are now routinely collected and archived. This paper describes the algorithms developed for detailed processing of TOBI data. Sonar imagery has a characteristic set of processing challenges and these are addressed. TOBI provides a very large sonar dataset, and to limit the difficulties of handling and processing these data, the raw data are subjected to a data reduction technique prior to further processing. Slant-range correction is improved by editing vehicle altitude data using a median filter. Noise on TOBI imagery can appear in two main forms; speckle noise and line dropouts. Speckle noise is removed by a small median difference kernel and line dropouts are removed using a ratio of two box-car filters, each with appropriate thresholding techniques. Precise geocoding of the imagery requires an accurate estimate of vehicle location, and a method of calculation is presented. Two optional processing algorithms are also; presented; deblurring of imagery to improve along-track resolution at far range, and the suppression of a surface reflection return which may occur when TOBI is operated in relatively shallow water. Several of the techniques presented can be transcribed and modified to suit other datasets     

Correlation of sidescan backscatter with grain size distribution of surficial seabed sediments

The dependence of acoustic backscatter on sediment grain size distribution is examined using dual frequency (100 and 410 kHz) sidescan sonar and 22 sediment grab samples from the Loch Linnhe artificial reef site on the west coast of Scotland. The sidescan data were processed to remove an empirically estimated average grazing angle dependence on backscatter. The processed data were analysed by forming histograms of pixels extracted from a 20 m² box around each ground truth site. A positive correlation (r=0.73) between mean backscatter intensity and mean grain size was obtained, i.e., the coarsest samples had the brightest backscatter. A positive correlation (r=0.59) was also found between the standard deviations of the backscatter and grain size distributions, i.e., poorly sorted sediments gave the most variable backscatter. The performance of the sidescan data was compared to results from a co-incident single-beam echo-sounder RoxAnn survey. The RoxAnn roughness index E1 compared well with the sidescan, whilst the RoxAnn hardness index E2 did not. This may be due to a physical link between the acoustic measures. The comparison showed the sidescan to have delivered a significantly higher-resolution image of the seabed for a similar amount of ship-time. Imaging of the artificial reef modules themselves was found to be frequency dependent.     

大深度近海底精细地形地貌探测技术

船载低频多波束测深声纳、侧扫声纳可以对深海海底地形地貌进行快速、高效、大面积探测,但其测量精度有限,难以满足深海科学考察、资源勘探开发对高精度海底地形地貌的需求。随着各类大深度水下移动载体(如深海拖体、水下机器人、遥控潜器和载人潜水器)的涌现,特别是各类耐高压测绘声纳的商业化,使大深度近海底精细地形地貌探测成为可能。首先分析了多波束测深声纳、侧扫声纳和测深侧扫声纳等3种测绘声纳的基本原理,然后分别介绍了各类测绘声纳的国内外典型商业化产品,并通过典型实例分析了其在大深度近海底精细测绘中的应用情况。