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。

用于水下声呐目标检测的弱特征共焦通道调控方法

声呐图像视觉检测是复杂水域资源勘探和水下异物目标探测领域的重要技术之一。针对声呐图像中小目标存在的特征微弱和背景信息干扰问题,本文提出弱特征共焦通道调控水下声呐目标检测算法。为了提高模型对弱小目标的信息捕获和表征能力,设计弱小目标特征激活策略,并引入先验框尺度校准机制匹配底层语义特征检测分支,以提高小目标检测精度。应用全局信息聚合模块深入挖掘弱小目标的全局特征,避免冗余信息覆盖小目标微弱关键特征。为解决传统空间金字塔池化易忽视通道信息的问题,提出共焦通道调控池化模块,保留有效通道域小目标信息并克服复杂背景信息干扰。实验结果表明,本文所提模型在水下声呐数据集的9类弱小目标识别的平均检测精度达83.3%,相较基准提高了5.5%,其中铁桶、人体模型和立方体检测精度得到显著提高,分别提高24%、8.6%和7.3%,有效改善水下复杂环境中弱小目标漏检和误检问题。

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。

基于随机共振的水下目标微弱回波信号增强检测

针对低信噪比下主动声呐探测水下目标的微弱回波信号检测难题,本文提出了一种基于随机共振的微弱回波信号增强检测方法。该方法通过建立频移变尺度随机共振系统,与输入高频微弱回波信号、随机噪声相匹配,实现微弱回波信号的共振增强;进而利用乘积谱与谱峭度构建目标回波亮点增强检测指标,有效提升微弱回波亮点的能量集中程度,抑制杂波及虚假成分干扰。仿真结果表明:在低信噪比下,随机共振系统输出信噪比增益可达5 dB以上,微弱目标回波亮点检测结果更加直观便捷,杂波干扰得到大幅抑制。海试数据处理结果进一步验证了所提方法的有效性和可靠性。