基于改进YOLOX的侧扫声纳图像沉船检测方法研究

针对传统侧扫声纳图像沉船目标识别精度低的问题,引入深度学习,提出了一种改进的You Only Look Once X(YOLOX)目标检测方法。首先对收集的侧扫声纳沉船图像进行预处理,根据实测过程拖鱼的姿态、仪器设备不同等造成的成像差异进行数据增强与扩充,并构建数据集;其次以YOLOX为基础网络,根据侧扫声纳图像缺少丰富特征信息的问题,对网络进行改进,在网络的Spatial pyramid pooling(SPP)结构引入Softpool池化替换原来的池化,提取更多的细节特征信息;最后对改进前后的网络模型进行精度评估,验证改进网络的可行性。实验结果表明,改进后的网络在平均精确率均值(mAP)等精度评定指标中相较于原网络都有着显著提升,识别效果更好。该研究可为侧扫声纳探测中实时目标物检测提供研究基础。     

基于改进YOLOv5网络的侧扫声纳图像目标检测方法

为构建适合不同测量平台搭载以及满足不同任务需求的高效侧扫声纳图像目标检测网络,将高性能轻量级目标识别网络PP-LCNet引入YOLOv5目标检测框架,并通过调整网络卷积层的通道数,构建了8种不同大小的高性能水下目标检测网络模型。以水下沉船目标为例进行了目标检测试验,结果表明,当网络主干改进为PP-LCNet-1.5x和PP-LCNet-2.0x时,检测精度(mAP)可分别达0.845和0.849,较原始YOLOv5s分别提升了0.024和0.028,同时参数量更少,能够满足普通硬件上的实时目标检测需求。     

侧扫声纳图像地理编码技术研究

在深入研究侧扫声纳图像地理编码理论的基础上，针对Q—MIPS格式的原始侧扫声纳图像进行了地理编码处理实验，并分析了重采样图像出现裂缝的原因，进而提出了一种较好的解决方法，最后给出了部分实验结果。     

侧扫声纳图像变形现象与实例分析

依据侧扫声纳成像原理，研究了由于声速变化、声波束倾斜和海底坡度改变等因素所致的声图像变形现象。结合飞机残骸、港池地形等典型声图像，分析了变形原因与判释方法。     

侧扫声纳数据库管理设计

海底地貌声纳图像作为新的遥感影像数据源，其数据管理问题日益突出。针对声纳图像的特点，分析研究了声纳图像的管理方法，并实现声纳图像数据库管理的功能。     

侧扫声纳数据的格式转换及应用

通过读取Qmips和XTF两种声纳数据格式，将其转换成通用的图像格式，用于图像的显示、分析和处理。介绍了声纳数据格式转换为Tiff及GeoTiff的方法，以及与数据格式相关的地理信息内容的获取，转换后的图像因其格式的头信息不同而显示结果不同。     

侧扫声纳图像地理编码方法研究

从侧扫声纳各类数据的特点出发,通过构建声纳图像的地理编码模型,提出侧扫声纳图像地理编码方法,将声纳回波数据与定位数据一一对应。实验数据结果表明:该方法是合理可行的,不仅较好地处理了定位数据,消除了拖鱼轨迹线上的折点和扇形裂缝,而且可实现海底回波点的地理定位。     

小波函数对侧扫声纳图像滤波效果的影响分析

侧扫声纳技术应用日益广泛,已成为海洋测量的重要工具,而去除噪声处理是对侧扫声纳图像进行正确判读的前提。利用小波函数滤波处理的方法,分别采用Haar、Daubechies、Coiflets、Symlets、Discrete Meyer、Biorthogonal、Reverse Biorthogonal等小波函数与中值滤波函数对侧扫声纳图像进行处理,并以平滑指数和边缘保持指数为评价指标,对滤波效果进行定量比较。试验表明,小波函数可以有效地平滑声纳图像,并能保持其较好的边缘效果。     

基于虚警函数的侧扫声纳水下目标实时检测方法

研究了侧扫声纳系统进行水下目标探测过程中目标信号的检测问题。通过分析海底回波信号的统计模型及其参数的估计,讨论了目标信号对统计模型拟合的影响规律,提出了侧扫声纳回波信号虚警函数和虚警率的概念,及其对Ping信号中目标信号的检测方法。算例结果表明,回波信号的三种分布模型中K分布拟合程度最优,在相同虚警率的条件下,基于K分布的虚警函数目标检测率最高。该法可为侧扫声纳回波信号中目标的实时报警提供技术支撑。     

高分辨率测深侧扫声纳

文中介绍了由中国科学院声学研究所和美国亚迪技术开发（上海）有限公司联合设计和制造的高分辨率测深侧扫声纳,它能够同时获得高分辨率的海底地形和地貌。该声纳由电子分机和分别安装在载体左右两侧的两条声纳阵组成,最大工作水深6000m。声纳阵由一条发射线阵和10条间距为λ／2的接收线阵组成,λ为声波波长,其中8条线阵接收声信号,两边的两条为哑元。声纳的多子阵海底自动检测-子空间拟合信号处理方法能克服水声信道多途和复杂海底的影响,正确检测到海底的直达回波。2003年11月和2004年7月,声纳在中国浙江千岛湖进行两次长时间的湖试,获得了高分辨率湖底等深线图和地貌图,正确检测出湖底边长为0．5m立方体目标的高度。     

Deblurring of GLORIA side-scan sonar images

In side-scan sonars such as GLORIA, along-track resolution is usually much worse than across-track resolution. This paper shows how along-track resolution may be improved by the application of an image restoration (deblurring) technique known as the Jansson-van Cittert method. Employing a model of the image formation process, this involves iterated convolution of the estimated deblurred image radiances with the theoretical alongtrack point spread function. The method and its implementation for GLORIA images are described. Smoothing of high frequency noise prior to restoration has been found to lead to an improved end-product. The restored images exhibit sharper edges and a greater clarity much appreciated by the interpreter. This visual impression is borne out by quantitative measurement. The technique is shown to be a useful adjunct to the battery of digital preprocessing techniques which can be applied to the sonar image prior to the information extraction stage.     

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.     

Geometric distortions in side-scan sonar images: a procedure fortheir estimation and correction

A procedure is introduced for the estimation and correction of geometric distortions frequently observed in side-scan sonar images as a result of motion instabilities of the sonar towfish. This procedure estimates geometric distortions from the image itself, without requiring navigational or altitude measurements. Estimates of the local degree of geometric distortion are obtained by cross-correlating segments of adjacent lines of the image. A mathematical model for the distortions is derived from the geometry of the problem and is applied to these estimates to reconstruct the sampling pattern on the seabed, under the assumption of a planar bottom. The estimated sampling pattern is then used for resampling the image to correct the geometric distortions. The model parameters may also be used for calculating approximate estimates of the attitude parameters of the towfish. A simulation is employed to evaluate the effectiveness of this technique and examples of its application to high-resolution side-scan sonar images are provided     

Fault scarp identification in side-scan sonar and bathymetry images from the Mid-Atlantic Ridge using wavelet-based digital filters

Digital filters designed using wavelet theory are applied to high resolution deep-towed side-scan sonar data from the median valley walls, crestal mountains, and flanks of the Mid-Atlantic Ridge at 29°10 N. With proper tuning, the digital filters are able to identify the location, orientation, length, and width of highly reflective linear features in sonar images. These features are presumed to represent the acoustic backscatter from axis-facing normal faults. The fault locations obtained from the digital filters are well correlated with visual geologic interpretation of the images. The side-scan sonar images are also compared with swath bathymetry from the same area. The digitally filtered bathymetry images contain nine of the eleven faults identified by eye in the detailed geologic interpretation of the side-scan data. Faults with widths (measured perpendicular to their strike) of less than about 150 m are missed in the bathymetry analysis due to the coarser resolution of these data. This digital image processing technique demonstrates the potential of wavelet-based analysis to reduce subjectivity and labor involved in mapping and analyzing topographic features in side-scan sonar and bathymetric image data.     

基于K-means聚类与数学形态学的侧扫声呐图像目标轮廓自动提取方法

针对侧扫声呐图像斑点噪声强、背景海底散射干扰严重,海底目标轮廓自动提取困难的问题,提出了一种基于K-means聚类与数学形态学相结合的海底目标轮廓自动提取算法。为克服噪声干扰,该算法首先利用中值滤波去除侧扫声呐图像中的强斑点噪声;然后采用K-means聚类算法对侧扫声呐灰度图像进行分割,并二值化,除去大部分海底背景噪声,初步提取出目标;接着利用数学形态学运算去除提取结果中的孤立噪点,并填充目标内部孔洞,得到连续化、圆滑的目标边缘;最后对处理后的侧扫声呐图像进行边缘检测,提取出目标轮廓。实验结果表明:该算法思想简单易行,具有很强的克服背景噪声的能力,自动提取的目标轮廓连续性较好,结果准确可靠。目前,在侧扫声呐图像目标轮廓提取过程中,主要采用人工方式,自动性较差,效率较低。本文算法可以实现目标轮廓的自动提取,提高效率,具有较强的实用价值。     

Interference fringes on GLORIA side-scan sonar images from the Bering Sea and their implications

GLORIA side-scan sonographs from the Bering Sea Basin show a complex pattern of interference fringes sub-parallel to the ship's track. Surveys along the same trackline made in 1986 and 1987 show nearly identical patterns. It is concluded from this that the interference patterns are caused by features in the shallow subsurface rather than in the water column. The fringes are interpreted as a thin-layer interference effect that occurs when some of the sound reaching the seafloor passes through it and is reflected off a subsurface layer. The backscattered sound interferes (constructively or desctructively) with the reflected sound. Constructive/destructive interference occurs when the difference in the length of the two soundpaths is a whole/half multiple of GLORIA's 25 cm wavelength. Thus as range from the ship increases, sound moves in and out of phase causing bands of greater and lesser intensity on the GLORIA sonograph. Fluctuations (or wiggles) of the fringes on the GLORIA sonographs relate to changes in layer thickness. In principle, a simple three dimensional image of the subsurface layer may be obtained using GLORIA and bathymetric data from adjacent (parallel) ship's tracks. These patterns have also been identified in images from two other systems; SeaMARC II (12 kHz) long-range sonar, and TOBI (30 kHz) deep-towed sonar. In these, and other cases world-wide, the fringes do not appear with the same persistence as those seen in the Bering Sea.     

基于分形纹理特征的侧扫声呐图像沉船识别方法研究

为提高侧扫声呐图像中沉船等目标信息的识别精度和识别效率,根据盒维数、毯维数与多重分形谱的侧扫声呐图像纹理特征提取算法,构建了基于分形纹理特征的Adaboost级联分类器沉船目标识别流程。结合实测侧扫声呐图像数据进行水下沉船识别实验,并与灰度共生矩阵和Tamura纹理特征的识别结果进行对比。研究表明,基于分形纹理特征的识别方法综合考虑了图像全局与局部纹理特征,且不依赖人工选取阈值参数与特征向量,可有效提高目标识别精度和识别效率。     

一种联合单波束测深的侧扫声呐海底线提取方法

论述了侧扫声呐的成像原理,针对目前常用侧扫声呐系统数据采集的相关特性,结合实际工程应用,提出了一种联合单波束测深的侧扫声呐海底线提取新方法,以便在数据后处理时进行快速准确地斜距改正。利用实测数据验证了该方法的有效性,并与成熟商用软件的提取结果和阈值法提取结果进行对比分析,结果表明该方法具有一定的可行性和优势。     

Active sonar detection in shallow water using the Page test

The use of active sonar in shallow water results in received echoes that may be considerably spread in time compared to the resolution of the transmitted waveform. The duration and structure of the spreading and the time of occurrence of the received echo are unknown without accurate knowledge of the environment and a priori information on the location and reflection properties of the target. A sequential detector based on the Page test is proposed for the detection of time-spread active sonar echoes. The detector also provides estimates of the starting and stopping times of the received echo. This signal segmentation is crucial to allow further processing such as more accurate range and bearing localization, depth localization, or classification. The detector is designed to exploit the time spreading of the received echo and is tuned as a function of range to the expected signal-to-noise ratio (SNR) as determined by the transmitted signal power, transmission loss, approximate target strength, and the estimated noise background level. The theoretical false alarm and detection performance of the proposed detector, the standard Page test, and the conventional thresholded matched filter detector are compared as a function of range, echo duration, SNR, and the mismatch between the actual and assumed SNR. The proposed detector and the standard Page test are seen to perform better than the conventional thresholded matched filter detector as soon as the received echo is minimally spread in time. The use of the proposed detector and the standard Page test in active sonar is illustrated with reverberation data containing target-like echoes from geological features, where it was seen that the proposed detector was able to suppress reverberation generated false alarms that were detected by the standard Page test