基于LOG算子的侧扫声呐图像水下小目标检测算法研究

针对侧扫声呐图像噪声干扰严重、分辨率低、目标轮廓模糊等特点，提出了一种基于LOG算子的侧扫声呐图像水下小目标检测算法。首先，根据侧扫声呐图像中水下小目标成像特点，对声呐图像进行滤波及聚类分割，大幅降低图像中噪声；然后，采用斑点检测思想，提取侧扫声呐图像中疑似目标区域；最后，基于自动阈值分割算法对声呐图像进行分割，获取目标区域二值图像，使用二阶矩估计目标尺度，剔除虚假目标，最终实现水下小目标准确检测。实验结果表明：该方法计算速度快、检测成功率高，对侧扫声呐图像中的水下小目标具有良好的检测效果。     

基于非锐化掩模引导滤波的水下图像细节增强算法研究

针对水下图像对比度偏低，细节模糊的问题，本文提出基于非锐化掩模引导滤波的细节增强方法。首先由原始图像做引导图进行滤波得到细节层图像，并对细节层使用噪声检测的中值滤波去除斑点噪声；然后对原始图像进行基于均值滤波的非锐化掩模，得到锐化图像，并将锐化图像作为引导图对原始图像进行引导滤波，获取基础层图像；最后将滤波后的细节层进行增益后与引导滤波获取的基础层进行叠加，达到增强水下图像细节的目的。并通过信息熵、局部对比度和平均梯度3种客观评价指标对图像处理结果进行了对比分析，主观和客观测试结果表明，本文采用的算法能够有效提高图像对比度以及增强细节信息，有利于提高水下图像资料解释的准确性。     

基于深度学习的水下光学图像超分辨率重建综述

图像超分辨率重建旨在从低分辨率图像中生成包含高频细节的高分辨率图像。随着近年来人工智能的快速发展，基于深度学习的超分辨率重建算法取得了突破性进展。然而，水下光学图像通常会产生严重的颜色失真、细节缺失、对比度下降与模糊等多种退化问题，重建难度远高于常规的自然光学图像。目前尚未有文献对基于深度学习的水下光学图像超分辨率重建进行系统性综述。首先，对自然图像退化方式和数据集进行分类总结，结合国内外最新研究现状将基于深度学习的单幅图像超分辨率重建算法分为针对一般退化、已知(非盲)多种退化、未知(盲)多种退化 3 个方面进行详细总结，为水下应用场景提供参考。然后， 介绍了水下光学图像退化方式，归纳了常见的公开数据集，总结并分析了水下光学图像超分辨率重建的最新进展。最后，对该领域未来可能的发展趋势进行了展望。     

水下激光线扫描探测系统的设计及试验

针对水下目标图像探测一直是困扰海洋界的1个难题，进行了水下探测的线扫描系统的设计和相关试验。从实验结果来看，线扫描系统可以很好地描述水下物体的三维外形细节，对于水下目标探测、识别十分有效。     

水下图像目标检测研究综述

在海洋开发的环境下,水下物体探测技术得到广泛应用,随着水下机器人与计算机技术的发展, 该技术越来越受到研究人员的重视。根据当前的水下图像目标检测研究进展,简要介绍水下图像目标检测流程(即图像采集、图像的预处理、以及图像检测的方法),对总结发展现状、发现技术的不足及挖掘未来的研究方向有重要意义。针对基于光学图像的水下目标识别问题,论述了图像采集、图像的预处理、以及图像检测等方面的主要进展,阐述了基于深度学习实现水下图像目标识别的技术发展现状。通过对水下目标处理过程的讨论和分析,指出水下图像目标识别领域中需要解决的问题,并预测该领域技术发展趋势。     

蓬莱附近 海区水下地形的星载合成孔径雷达遥感

本文依据山东蓬莱附近海区１９９２年４月１９日的ＥＲＳ－１ＳＡＲ图像，分析了该海区的水下地形并探讨了水下地形雷达成像机理。分析结果显示，ＳＡＲ图像较好地揭示了该海区四人洲、潮待洲、新井洲等水下地形。研究指出，流经水下地形的潮流对海洋表面波的水动力调制是水下地形ＳＡＲ成像的主要机理。     

基于小波的水下图像后向散射噪声去除

介绍水下图像成像的特点,分析影响水下成像的主要因素:吸收和散射。在此基础上,采用小波的方法对水下图像进行去噪。首先对低频系数采用了LUM滤波器进行滤波,然后对高频系数采用了NeighShrink方法和对其改进后的ENS方法进行去噪。通过实验证实,该方法去除水下图像噪声具有较好的效果。     

水下图像基于 GAN 去模糊的增强技术

由于水下环境具有不稳定性,水下图像可能会出现偏色、对比度低以及运动模糊等退化现象。针对这些问题,本文提出了适用于水下图像的增强算法,其实现需要依次经过颜色恢复和去模糊这2个阶段。在第一阶段中,本文增强算法先利用高斯滤波和均值漂移对图像进行锐化;然后,通过对比图像各颜色通道的均值得到补偿值对图像颜色进行校正;最后通过线性拉伸来调整图像的对比度。在第二阶段中,采用带有残差思想的生成对抗网络(GenerativeAdversarialNetworks,GAN),利用9个连续的残差网络能够很好地提取图像中的特征,可起到消除模糊和增强图像特征的作用。利用本文算法处理水下图像时,发现本文方法不仅能去除图像模糊,而且能消除图像的色偏现象且不携带红色伪影。同时,通过对比水下图像质量度量(UnderwaterImageQualityMeasures,UIQM)和水下彩色图像质量评估(UnderwaterColorImageQualityEvaluation,UCIQE)这2项指标发现,本文算法有较好的图像处理效果。     

基于联合生成–去除的水下图像增强方法

现有的基于深度学习的水下图像增强方法在仿真的水下图像上取得了良好的效果。但是,由于简化的仿真图像与复杂的真实图像之间存在较大差距,此类方法在处理真实水下图像时性能明显下降。为了解决真实水下图像增强问题,提出了一种联合生成–去除水下图像增强方法。该方法采用分解思路,将水下图像分解为干净的背景层和退化层,通过循环一致性损失和对抗性损失来更好地保留背景,进而实现真实图像和仿真图像之间的转换,既校正了图像颜色,又提升了图像对比度,实现良好的增强效果。实验结果表明, 本方法在真实水下图像数据集上处理的结果,在色彩、纹理细节和清晰程度方面均优于现有的对比方法。     

浅海水下地形的星载合成孔径雷达遥感

依据山东蓬莱附近海区１９９２年４月１９日的欧洲遥感卫星一号（ＥＲＳ－１）合成孔径雷达（ＳＡＲ）图像，分析了该海区的水下地形并探讨了水下地形雷达成像机理。分析结果显示，ＳＣＲ图像较好地揭示了该海区四人洲、潮待洲、新井洲等水下地形。研究指出，流经水下地形的潮流对海洋表面波的水动力调制是水下地形ＳＡＲ成像的主要机理。本研究表明星载ＳＡＲ作为浅海水下地形测量的新型手段具有广阔的应用前景。     

水声通信中基于小波变换的图像编码研究

提出了一种高误比特率传输条件下的图像编码方法，它适用于水声信道的图像传输。针对水下图像的特点，选取合适的小波基和变换参数对图像进行离散小波变换；依据小波系数的能量分布特性，对不同的子带采用不同的量化和定长编码，编码率为0．8比特／像素。水声通信试验表明，在传输误比特率达到10^-2时，仍能得到可接受的图像质量。     

Underwater Acoustic Image Transmission System Based on DSP

The underwater acoustic image transmission system based on the high-speed DSP device TMS320C549 has been studied.We use Goertzel algorithm for source decoding and MFSK for modulation.Turbo code is used for channel coding and decoding.The purpose is to implement underwater video image data transmission.     

A Method of Coding and Decoding in Underwater Image Transmission

A new method of coding and decoding in the system of underwater image transmission is introduced, including the rapid digital frequency synthesizer in multiple frequency shift keying,image data generator, image grayscale decoder with intelligent fuzzy algorithm, image restoration and display on microcomputer.     

Lossless compression of hydroacoustic image data

Despite rapid progress in improving mass-storage density and digital communication system performance, compression of hydroacoustic image data is still significant in many engineering and research areas since it can overcome data storage and transmission bandwidth limitations. In this paper, we present a novel and effective approach for lossless compression of hydroacoustic image data which consists of two stages. The first stage reduces the information redundancy. We propose several new techniques to remove redundancy between data samples, data blocks, and data frames. The second stage uses a newly developed cascade coding scheme. This simple scheme can achieve an efficiency of 97%. A decomposition algorithm is presented for finding the optimal cascade coding parameters. The algorithm decomposes a multivariable optimization problem into a series of one-variable optimizations. Our two-stage algorithm offers a compression ratio of 2-3 and provides an exact recovery of the original data. Because of its simplicity, the algorithm can be incorporated into a variety of echo sounder systems. The compression algorithms can also be implemented using low-level assembly language to meet the requirements of real-time applications     

A simulation involving MAPS video bandwidth reduction, acoustic channel errors, and error checking

The reliable acoustic transmission of images that have been processed by a video bandwidth reduction technique such as Micro-Adaptive Picture Sequencing (MAPS) necessitates a robust coding scheme. This is due to the fact that errors effect compressed data more seriously than uncompressed data. For this reason image quality is a function of three system variables: image compression, channel noise, and error checking. A real-time simulation has been developed to determine the relationship between the three system variables. The simulator uses two 68000 microcomputers connected by a serial link; one does image compression and noise modeling while the other performs error checking and image reconstruction. The noise model uses 17 characteristics and operating parameters of the acoustic channel to corrupt the image data with single bit random errors. At the receiver the error checking can correct up to 90 percent of the most serious single bit errors.     

水声通信中的鲁棒图像编码研究

由于受各种因素的影响，水下声信道是一种传输差错率较高的信道。标准化的图像编码系统(例如JPEC；H．263，MPEG等)使用了相似的压缩技术，它们往往存在严重的错误扩散，甚至单个错误比特就可能破坏整幅图像，所以一般不适合作为水下声信道图像传输的编码方案。文章针对常用的图像编码的缺点，利用定长编码技术，提出了一种高鲁棒性的图像压缩方案。实验表明在压缩率1．25比特／象素时，压缩后的图像仍然保持了较好的质量，并且能够较好地抵抗信道误码，提高了水下声信道图像传输的质量。     

基于地磁与红外双模探测的海洋浮标预警系统设计

设计了一种基于地磁检测与红外感应相结合、可对浮标周围异常目标进行探测和预警的控制系统。系统采用芯片级的微型磁感线圈,以及高集成度、低功耗的数据采集与总线技术,通过探测船体磁性对地磁场的扰动,监测船舶对浮标的靠近;采用芯片级的热释电红外传感器,通过探测人体红外辐射,监测浮标在正常工作期间未知人员的入侵。系统的预警采用声光报警和图像远程传输相结合的方式,现场采集的图像数据经过压缩编码后通过无线数传电台发送至远程岸基监测站,实现海洋浮标的远程预警与现场取证功能。     

XCTD剖面仪数据传输电路的设计与实现

针对XCTD(投弃式温盐深)剖面仪数据传输速率低、可靠性差的缺点,提出基于曼彻斯特编码的基带传输技术。在系统结构设计的基础上,着重讨论了基于FPGA的曼彻斯特编码器、位定时提取电路和自动增益控制电路的设计。通过测试表明,编码后的系统不仅可以使用基带信号直接传输,而且能够满足系统的通信速率要求。     

Acoustic transmission of images using a microprocessor-based video bandwidth reduction technique

The problems associated with the acoustic transmission of real-time television pictures are considered in this paper. A solution to the high data rate video source is found in the application of a spatial data redundancy reduction algorithm, Micro Adaptive Picture Sequencing (MAPS). This algorithm has been implemented on two commercially available microprocessors (M68000), which operate on a digitized image to produce a coded image which is transmitted and reconstructed. Using the spatial data reduction algorithm it is possible to transmit 1 to 2 frames per second of a96 times 96 pixel times 4bits gray level image over a 9600 baud link. Although these images do not possess the full dynamic range and resolution of home quality video, this work suggests that the acoustic channel could be utilized for control of an unmanned untethered submersible. Also, the transmission of high resolution information is made practical when some data reductions are performed prior to acoustic transmission.     

Impact Analysis of Coastal Engineering Projects on Mangrove Wetland Area Change with Remote Sensing

In the past decades, two large scale coastal engineering projects have been carried out in the Deep Bay surrounded by Shenzhen City and Hong Kong Special Administrative Region. One project is Shenzhen River channel regulation and the other is the sea reclamation along the seashore on the Shenzhen side. The two projects are very close to the two national nature reserves, specifically Futian in Shenzhen and Mai Po in Hong Kong, which are important wetland ecosystems worldwide. This paper aims to identify and monitor the mangrove wetland changes with time series of Landsat Thematic Mapper images pre and post to the two engineering projects being launched. Coupled analysis of the image interpretation results and tidal data acquired at the same time in the context of the two works reveals that the mangrove wetland area has increased from year 1989 to 1994, and has changed little from year 1994 to 2002. Binary coding is applied to reveal the distribution image of mangrove at each phase, and the coding image shows that the construction of the two coastal engineering projects has caused frequent changes in mangrove spatial distribution. The study also shows that the change is not significant regarding to the precision of the method and the natural evolution of mangrove wetland, and the projects do not cause apparently influences upon the two national mangrove conservation zones at least for the research time period.