RepDNet:A re-parameterization despeckling network for autonomous underwater side-scan sonar imaging with prior-knowledge customized convolution

Side-scan sonar(SSS)is now a prevalent instrument for large-scale seafloor topography measurements,deployable on an autonomous underwater vehicle(AUV)to execute fully automated underwater acoustic scanning imaging along a predetermined trajectory.However,SSS images often suffer from speckle noise caused by mutual interference between echoes,and limited AUV computational resources further hinder noise suppression.Existing approaches for SSS image processing and speckle noise reduction rely heavily on complex network structures and fail to combine the benefits of deep learning and domain knowledge.To address the problem,Rep DNet,a novel and effective despeckling convolutional neural network is proposed.Rep DNet introduces two re-parameterized blocks:the Pixel Smoothing Block(PSB)and Edge Enhancement Block(EEB),preserving edge information while attenuating speckle noise.During training,PSB and EEB manifest as double-layered multi-branch structures,integrating first-order and secondorder derivatives and smoothing functions.During inference,the branches are re-parameterized into a 3×3 convolution,enabling efficient inference without sacrificing accuracy.Rep DNet comprises three computational operations:3×3 convolution,element-wise summation and Rectified Linear Unit activation.Evaluations on benchmark datasets,a real SSS dataset and Data collected at Lake Mulan aestablish Rep DNet as a well-balanced network,meeting the AUV computational constraints in terms of performance and latency.

An Underwater Robot Inspection Anomaly Localization Feedback System Based on Sonar Technology

This article introduces an underwater robot inspection anomaly localization feedback system comprising a real-time water surface tracking,detection,and positioning system located on the water surface,while the underwater robot inspection anomaly feedback system is housed within the underwater robot.The system facilitates the issuance of corresponding mechanical responses based on the water surface’s real-time tracking,detection,and positioning,enabling recognition and feedback of anomaly information.Through sonar technology,the underwater robot inspection anomaly feedback system monitors the underwater robot in real-time,triggering responsive actions upon encountering anomalies.The real-time tracking,detection,and positioning system from the water surface identifies abnormal conditions of underwater robots based on changes in sonar images,subsequently notifying personnel for necessary intervention.

A Novel CCA-NMF Whitening Method for Practical Machine Learning Based Underwater Direction of Arrival Estimation

Underwater direction of arrival(DOA)estimation has always been a very challenging theoretical and practical problem.Due to the serious non-stationary,non-linear,and non-Gaussian characteristics,machine learning based DOA estimation methods trained on simulated Gaussian noised array data cannot be directly applied to actual underwater DOA estimation tasks.In order to deal with this problem,environmental data with no target echoes can be employed to analyze the non-Gaussian components.Then,the obtained information about non-Gaussian components can be used to whiten the array data.Based on these considerations,a novel practical sonar array whitening method was proposed.Specifically,based on a weak assumption that the non-Gaussian components in adjacent patches with and without target echoes are almost the same,canonical cor-relation analysis(CCA)and non-negative matrix factorization(NMF)techniques are employed for whitening the array data.With the whitened array data,machine learning based DOA estimation models trained on simulated Gaussian noised datasets can be used to perform underwater DOA estimation tasks.Experimental results illustrated that,using actual underwater datasets for testing with known machine learning based DOA estimation models,accurate and robust DOA estimation performance can be achieved by using the proposed whitening method in different underwater con-ditions.

YOLOv5-Based Seabed Sediment Recognition Method for Side-Scan Sonar Imagery

Seabed sediment recognition is vital for the exploitation of marine resources.Side-scan sonar(SSS)is an excellent tool for acquiring the imagery of seafloor topography.Combined with ocean surface sampling,it provides detailed and accurate images of marine substrate features.Most of the processing of SSS imagery works around limited sampling stations and requires manual interpretation to complete the classification of seabed sediment imagery.In complex sea areas,with manual interpretation,small targets are often lost due to a large amount of information.To date,studies related to the automatic recognition of seabed sediments are still few.This paper proposes a seabed sediment recognition method based on You Only Look Once version 5 and SSS imagery to perform real-time sedi-ment classification and localization for accuracy,particularly on small targets and faster speeds.We used methods such as changing the dataset size,epoch,and optimizer and adding multiscale training to overcome the challenges of having a small sample and a low accuracy.With these methods,we improved the results on mean average precision by 8.98%and F1 score by 11.12%compared with the original method.In addition,the detection speed was approximately 100 frames per second,which is faster than that of previous methods.This speed enabled us to achieve real-time seabed sediment recognition from SSS imagery.

Underwater Terrain-Aided Navigation Based on Multibeam Bathymetric Sonar Images

Underwater terrain-aided navigation is used to complement the traditional inertial navigation employed by autonomous underwater vehicles during lengthy missions. It can provide fixed estimations by matching real-time depth data with a digital terrain map, This study presents the concept of using image processing techniques in the underwater terrain matching process. A traditional gray-scale histogram of an image is enriched by incorporation with spatial information in pixels. Edge comer pixels are then defined and used to construct an edge comer histogram, which employs as a template to scan the digital terrain map and estimate the fixes of the vehicle by searching the correlation peak. Simulations are performed to investigate the robustness of the proposed method, particularly in relation to its sensitivity to background noise, the scale of real-time images, and the travel direction of the vehicle. At an image resolution of 1 m2/pixel, the accuracy of localization is more than 10 meters.

Sonar Image Processing System for an Autonomous Underwater Vehicle(AUV)

Sonar image processing system is an important intelligent system of Autonomous Un-derwater Vehicle.Based on TMS320C30 high speed DSP,it is used to realize sonar imagecompression and underwater object detections including obstacle recognition in real time.Inthis paper,the software and hardware designs of this system are introduced and the experi-mental results are given.

Application of A* Algorithm for Real-time Path Re-planning of an Unmanned Surface Vehicle Avoiding Underwater Obstacles

This paper describes path re-planning techniques and underwater obstacle avoidance for unmanned surface vehicle(USV) based on multi-beam forward looking sonar(FLS). Near-optimal paths in static and dynamic environments with underwater obstacles are computed using a numerical solution procedure based on an A* algorithm. The USV is modeled with a circular shape in 2 degrees of freedom(surge and yaw). In this paper, two-dimensional(2-D) underwater obstacle avoidance and the robust real-time path re-planning technique for actual USV using multi-beam FLS are developed. Our real-time path re-planning algorithm has been tested to regenerate the optimal path for several updated frames in the field of view of the sonar with a proper update frequency of the FLS. The performance of the proposed method was verified through simulations, and sea experiments. For simulations, the USV model can avoid both a single stationary obstacle, multiple stationary obstacles and moving obstacles with the near-optimal trajectory that are performed both in the vehicle and the world reference frame. For sea experiments, the proposed method for an underwater obstacle avoidance system is implemented with a USV test platform. The actual USV is automatically controlled and succeeded in its real-time avoidance against the stationary undersea obstacle in the field of view of the FLS together with the Global Positioning System(GPS) of the USV.

DcNet: Dilated Convolutional Neural Networks for Side-Scan Sonar Image Semantic Segmentation

In ocean explorations,side-scan sonar(SSS)plays a very important role and can quickly depict seabed topography.As-sembling the SSS to an autonomous underwater vehicle(AUV)and performing semantic segmentation of an SSS image in real time can realize online submarine geomorphology or target recognition,which is conducive to submarine detection.However,because of the complexity of the marine environment,various noises in the ocean pollute the sonar image,which also encounters the intensity inhomogeneity problem.In this paper,we propose a novel neural network architecture named dilated convolutional neural network(DcNet)that can run in real time while addressing the above-mentioned issues and providing accurate semantic segmentation.The proposed architecture presents an encoder-decoder network to gradually reduce the spatial dimension of the input image and recover the details of the target,respectively.The core of our network is a novel block connection named DCblock,which mainly uses dilated convolution and depthwise separable convolution between the encoder and decoder to attain more context while still retaining high accuracy.Furthermore,our proposed method performs a super-resolution reconstruction to enlarge the dataset with high-quality im-ages.We compared our network to other common semantic segmentation networks performed on an NVIDIA Jetson TX2 using our sonar image datasets.Experimental results show that while the inference speed of the proposed network significantly outperforms state-of-the-art architectures,the accuracy of our method is still comparable,which indicates its potential applications not only in AUVs equipped with SSS but also in marine exploration.

Multi-beam Sonar and Side-scan Sonar Image Co-registering and Fusing

Multi-beam Sonar and Side-scan Sonar compensate each other. In order to fully utilize all information, it is necessary to fuse two kinds of image and data. And the image co-registration is an important and complicated job before fusion. This paper suggests combining bathymetric data with intensity image, obtaining the characteristic points through the minimal angles of lines, and then deciding the corresponding image points by the maximal correlate coefficient in searching space. Finally, the second order polynomial is applied to the deformation model. After the images have been co-registered, Wavelet is used to fuse the images. It is shown that this algorithm can be used in the flat seafloor or the isotropic seabed. Verification is made in the paper with the observed data.

Improving Yolo5 for Real-Time Detection of Small Targets in Side Scan Sonar Images

Side scan sonar(SSS)is an important means to detect and locate seafloor targets.Autonomous underwater vehicles(AUVs)carrying SSS stay near the seafloor to obtain high-resolution images and provide the outline of the target for observers.The target feature information of an SSS image is similar to the background information,and a small target has less pixel information;therefore,accu-rately identifying and locating small targets in SSS images is challenging.We collect the SSS images of iron metal balls(with a diameter of 1m)and rocks to solve the problem of target misclassification.Thus,the dataset contains two types of targets,namely,‘ball’and‘rock’.With the aim to enable AUVs to accurately and automatically identify small underwater targets in SSS images,this study designs a multisize parallel convolution module embedded in state-of-the-art Yolo5.An attention mechanism transformer and a convolutional block attention module are also introduced to compare their contributions to small target detection accuracy.The performance of the proposed method is further evaluated by taking the lightweight networks Mobilenet3 and Shufflenet2 as the backbone network of Yolo5.This study focuses on the performance of convolutional neural networks for the detection of small targets in SSS images,while another comparison experiment is carried out using traditional HOG+SVM to highlight the neural network’s ability.This study aims to improve the detection accuracy while ensuring the model efficiency to meet the real-time working requirements of AUV target detection.

基于法线微分的3维声呐点云自适应简化方法

在不损失原始点云数据质量的前提下,大幅约简点云数据量是减少存储空间、降低后期计算强度的重要预处理步骤。针对这一需求,提出了针对水下3维声呐点云数据的自适应简化方法。首先,定义法线微分算子来识别点云中几何尺度的骤变,从而实现原始点云中边界部分点云和主体部分点云的分割。其次,对于点云的边界部分,应用移动最小二乘法来对边界点进行优化,降低噪点的影响,并保持其曲面的几何一致性;基于体素栅格结构,在边界上使用八叉树进行降采样,并在此基础上实施局部最远点采样,在实现均匀简化的同时保证已简化点云的边界部分具有各向同性,有效保留边界部分点云的几何特征信息。再次,对于点云的主体部分,为保持简化后点云整体的各向同性,使用体素中心采样法来减少数据量。然后,通过高斯滤波平滑点云表面,最后,整合简化后的边界点云和主体点云,得到简化结果。实验结果表明,提出的简化方法计算成本低、处理速度快,在与现行典型算法保持一致简化率的情况下,对水下点云数据的简化速度提高了约32%。另外,通过表面密度对比与几何失真分析,证明了提出方法对水下3维点云边界点及整体分布的优化作用。综上,此方法能提高水下作业目标探测效率,得到保留重要几何特征信息并具有各向同性的水下任务目标点云简化结果。

基于被动声呐音频信号的水中目标识别综述

基于被动声呐音频信号的水中目标识别是当前水下无人探测领域的重要技术难题,在军事和民用领域都应用广泛.本文从数据处理和识别方法两个层面系统阐述基于被动声呐信号进行水中目标识别的方法和流程.在数据处理方面,从基于被动声呐信号的水中目标识别基本流程、被动声呐音频信号分析的数理基础及其特征提取三个方面概述被动声呐信号处理的基本原理.在识别方法层面,全面分析基于机器学习算法的水中目标识别方法,并聚焦以深度学习算法为核心的水中目标识别研究.本文从有监督学习、无监督学习、自监督学习等多种学习范式对当前研究进展进行系统性的总结分析,并从算法的标签数据需求、鲁棒性、可扩展性与适应性等多个维度分析这些方法的优缺点.同时,还总结该领域中较为广泛使用的公开数据集,并分析公开数据集应具备的基本要素.最后,通过对水中目标识别过程的论述,总结目前基于被动声呐音频信号的水中目标自动识别算法存在的困难与挑战,并对该领域未来的发展方向进行展望.

基于ZYNQ的轻量化YOLOv5声呐图像目标检测算法及实现

针对声呐图像存在的模糊、样本量不足的现象,本文提出一种基于YOLOv5的声呐图像目标检测改进算法。利用几何滤波、垂直翻转等方法,对声呐图像数据集进行数据增强。添加融合注意力机制模块,使算法更好地关注声呐图像小目标的特征。同时,针对目前大多数目标检测算法运行在云端,无法做到实时性声呐图像检测的问题,本文利用替换轻量级网络和NCNN边端移植技术,同时在颈部网络中采用GSConv模块,将算法成功移植到ZYNQ平台,实现声呐图像的嵌入式端实时检测。实验表明,本文提出的算法在降低了56%参数量的同时,在map50和map50-95上分别提高2.2%和2.5%。改进后的算法性能提升明显,证明所提出的方法在轻量化声呐图像目标检测任务上具有一定的可行性与有效性。

基于卷积神经网络的声呐图像水下目标检测综述

声呐探测技术作为水下探测的主要手段之一,在海洋环境中具有广泛的应用,随着卷积神经网络(CNN)在计算机视觉领域表现出卓越的性能,该技术越来越受到研究人员的重视。文章综述了卷积神经网络在声呐图像水下目标检测中的应用与发展,重点讨论基于CNN的典型水下目标检测方法,包括基于候选区域与回归的检测算法在声呐图像上的改进与应用,并分析各类网络模型在处理声呐图像特有问题上的创新策略,如小样本检测、小目标检测及CNN与传统算法融合等。最后总结了当前基于CNN的水下目标检测面临的挑战,并预测该领域技术发展趋势。

基于非负矩阵分解的双曲调频信号目标回波增强

双曲调频信号是主动声呐中常用的多普勒不敏感信号。在实际使用中,由于信道变化和混响影响,目标回波较弱。针对水下航行器主动声呐探测性能下降问题,本文提出基于非负矩阵分解的双曲调频信号目标回波增强方法。该方法对目标回波和干扰信号的基矩阵分别定义,将发射信号的时频图作为目标回波的频率信息矩阵,对目标回波的时间信息矩阵做出约束,迭代处理完成目标回波的增强,获得较高增益的匹配回波检测结果。数值仿真和海上试验结果表明:本文方法与传统匹配滤波方法相比,其峰均比值提高了2 dB以上,提高了回波检测结果。

基于解卷积的高精度水下声信标测向方法

针对水下声信标搜寻定位声呐对搜寻定位效率的需求,文章引入一种基于空域解卷积的高精度水下声信标测向方法,对常规波束形成结果进行解卷积处理,提高了声信标搜寻定位声呐的处理增益和测向精度,提高了搜寻定位效率,并且使得为搜索发射频率为37.5 kHz的声信标设计的声呐,同时也具备了对发射频率为8.8 kHz的声信标的搜寻定位功能。通过仿真和实测数据验证了文中方法的有效性。

基于目标特征的数据压缩预处理方法

随着水下自动监测技术的不断发展,对于数据的实时传输效率和测量精度的要求也越来越高,而高质量无损数据压缩传输技术还存在严重不足。文章利用在广西西江航道上建立的侧扫声呐船舶吃水自动监测系统,提出了一种基于目标特征的数据压缩预处理方法。该方法根据散货船的船体结构和声图成像特点,在进行数据压缩前分别在时间和空间两个维度上实现对目标信号的识别和提取,完成对无关冗余数据的剔除。该预处理方法不仅可以大幅提高数据压缩的压缩比,提高传输效率,节省存储空间,还可以保证用于测量计算的目标关键特征信息不丢失,为进一步突破无损数据压缩的压缩比限制提供新思路。

基于语义分割的侧扫声纳管线目标检测方法

为提高侧扫声纳图像中管线目标检测的自动化程度及效率,提出了一种基于语义分割的水下管线目标检测方法。首先通过构建高效语义分割网络主干,提高网络计算速度并降低网络对计算机硬件性能的需求;其次给出了一种针对管线目标特点的加权交叉熵损失函数,解决了因类间数量不均衡导致的网络训练困难问题。以多种复杂条件下侧扫声纳实测数据进行了水下管线检测试验,结果表明,该方法在取得和经典网络相近精度的情况下,速度提升了2.7倍,可达52.6FPS,实现了水下管线的快速、准确检测。

基于占空比特征的高频水声信号粗分类方法

高频水声信号主要包含水声通信信号和主动声呐信号,频带范围通常在3~40 kHz之间,具有信噪比高、隐含信息丰富等特点。图像识别和机器学习类方法受无人平台能量供给限制,难以实现工程应用。针对测距类主动声呐信号与水声通信信号低复杂度分类问题,提出了一种基于占空比特征的高频水声信号粗分类方法,利用水声通信信号和主动声呐信号占空比差异实现两类高频水声信号粗分类。最后,通过仿真信号和实测数据验证了算法的有效性。

量测不确定性条件下的箱粒子滤波目标跟踪方法

在复杂水下环境中,主动声呐的量测值在距离和方位分辨率上存在较大的不确定性,即一个目标回波的能量可能覆盖声呐距离-方位能量谱的多个相邻位置网格。并且,当环境中混响较强时,上述量测不确定性将引起多个区域性的杂波干扰。为了减小状态空间估计的偏差,基于粒子滤波(PF)的跟踪方法需要大量粒子来近似后验概率密度,跟踪的实时性急剧降低。针对上述问题,该文提出一种基于区间量测的箱粒子滤波跟踪方法(IBPF),对主动声呐量测值进行区间表示,即用一个表示距离和方位区间的箱粒子代替点值量测,用区间表示这种量测不确定性,在提高状态估计稳定性的同时,极大程度地减少了后验概率密度估计所需的粒子数,从而进一步提高计算效率。实验结果表明,所提IBPF与PF相比,能以更高的计算效率获得更优的跟踪性能,对目标的跟踪时间缩短了18.06%,跟踪成功帧数增加了4.29%。