共查询到19条相似文献，搜索用时 156 毫秒
1.
2.
3.
侧扫声呐图像特征自动提取的难点在于特征地貌边缘检测较困难,依据图像灰度突变检测得到的边缘比较粗糙、不连续,而且有断口和小洞。本文在对图像进行预处理和阈值化的基础上,采用数学形态学方法对图像进行处理,即用具有一定形态的结构元素去量度和提取图像中的对应形状,得到连续化、粗化、圆滑的特征区域边缘填充目标内部阴影且消除背景噪声。基于数学形态学的侧扫声呐图像特征自动提取的主要步骤为:首先对侧扫声呐图像进行预处理,然后进行灰度阈值化,接着采用数学形态学方法进行处理,最后对处理后的图像进行边缘检测,提取出特征地貌边缘。实验表明,采用数学形态学方法进行处理后,错断、离散的海底目标物变得连续,背景噪声大大减少,自动提取结果准确可靠。 相似文献
4.
淦智权 《数字海洋与水下攻防》2022,5(4):381-386
声呐图像分割是图像分割技术发展中的组成部分,是水下目标识别与检测的重要一环。传统方法中基于有监督分割方法的算法往往代价较大,表现出试验周期长、实时性较差、运行速率较慢等不足。并且由于声呐图像的成像质量差、分辨率不高、边缘条件不清晰、人工标注工作量大等客观因素,不易建立用于有监督模型训练的大规模数据集,使得传统分割方法越来越不适应当前实际应用的多方面要求。将基于无监督学习卷积神经网络引入到声呐图像分割任务中,分割模型通过对单帧声呐图像进行训练和测试,最后经过推理得到将阴影区和目标高亮区分割后的声呐图像,得到分割出来的水下目标。通过对实验的分割结果进行各项指标分析,证明此方法有着更好的运行效率和分割精度,并且实时性较高,综合性能优于传统方法。 相似文献
5.
针对水下小目标探测与识别难的问题，开展基于侧扫声呐的声呐图像滤波、图像分割及目标提取方法研究。常规滤波方法难以有效清除图像中存在的噪声，从而造成图像质量下降。采用非局部均值滤波算法与 GPU 加速的方法，在获得声呐图像较好处理效果的同时，满足水下小目标检测实时性的要求；同时， 采用膨胀算法与 Canny 边缘检测算法相结合的方式，实现了水下真假目标的有效区分。 相似文献
6.
侧扫声呐回波信号是形成侧扫声呐图像的基础,是侧扫声呐系统对水下目标的最直接观测量, 将一维小波变换与非线性增强方法相结合,提出了一种基于小波变换的侧扫声呐回波信号非线性增强算法, 用以改善侧扫声呐图像对比度低、噪声强度大的问题。首先利用改进的 Bayes 阈值对侧扫声呐 ping 信号进行一维小波分解,提取信号特征信息;然后利用 2 种不同的非线性函数对高、低频小波系数进行处理;最后利用小波反变换重构信号,形成增强后的侧扫声呐图像。实测数据验证结果表明：利用该算法对侧扫声呐 ping 信号进行处理,实现了侧扫声呐图像对比度的增强和对噪声的抑制,可以获取较好的图像视觉效果。 相似文献
7.
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。 相似文献
8.
基于声呐的水下目标检测是具有重要意义的研究课题。由于声呐图像质量差、对比度低、边缘模糊等,基于特征提取的识别方法在精度和速度上无法满足现有要求。鉴于此,将基于卷积神经网络的 YOLOv3 引入声呐图像的目标识别任务中,首先通过动态亮度分配和中值滤波对图像进行预处理建立数据集,然后对 YOLOv3 模型进行训练和测试,最后根据静态目标的位置不变去除虚假目标,降低误检率。实验结果表明： 较之现有的特征提取识别算法,本算法具有更高的效率和更低的误检率。 相似文献
9.
提出噪声图象边缘信息提取的一种实用方法。该方法以锐化处理结果为基础,二次滤波完成图象预处理。研究表明,单纯对噪声图象进行锐化处理,只可增加视觉感观效果,不会使边缘提取效果增强;必须对锐化结果再次进行滤波处理后,才可获得满意的边缘信息。 相似文献
10.
程文博 《数字海洋与水下攻防》2023,6(3):353-358
由于受到复杂水声信道影响，侧扫声呐生成图像一般存在分辨率低、细节模糊等问题，这给基于侧扫声呐图像的小目标检测与识别过程带来了一定难度。常见的图像超分辨率处理算法大致可以分为插值算法、稀疏表示的重建算法等。传统插值算法存在图像边缘模糊、细节不清楚的问题，因此提出了一种改进型超分辨率生成对抗网络(SRGAN)结构。该结构可在保留图像中的小目标细节的基础上提升细节特征显示度，并且可使目标边缘更加清晰，可为后续的小目标检测和识别在一定程度上提供技术支持。 相似文献
11.
12.
13.
利用可视化编程语言ＶＢ开发出一套成像光谱仪资料处理系统，用于海面叶绿素浓度信息提取，具有操作方便的图形用户界面，大大简化了各种振兴中华，去了众多参数值的反复输入。图像处理模块可加入许多常用的图像处理功能。并可调入一些商业化软件，寿命完成许多图像处理功能。 相似文献
14.
色彩处理是卫星影像处理应用过程中的重要环节,大区域的资源一号02C影像处理的难点在于真彩色模拟和色彩一致性处理。通过分析02C影像特点及大区域色彩处理的主要影响因素,提出基于最小二乘拟合的02C卫星影像真彩色合成方法和面向多时相大区域02C卫星影像的色彩一致性处理技术。在保证色彩真实的前提下,兼顾了大范围影像色彩的局部差异和合理过渡,从而使图面效果达到最佳。通过甘南地区的生产实验,表明提出的色彩处理技术的可靠性得到充分验证。 相似文献
15.
The digital processing requirements of several algorithms for extracting the spectrum of a detected synthetic aperture radar (SAR) image from the raw SAR data are described and compared. The most efficient algorithms for image spectrum extraction from raw SAR data appear to be those containing an intermediate image formation step. It is shown that a recently developed compact formulation of the image spectrum in terms of the raw data is computationally inefficient when evaluated directly, in comparison with the classical method where matched-filter image formation is an intermediate result. It is also shown that a proposed indirect procedure for digitally implementing the same compact formulation is somewhat more efficient than the classical matched-filtering approach. However, this indirect procedure includes the image formation process as part of the total algorithm. Indeed, the computational savings afforded by the indirect implementation are identical to those obtained in SAR image formation processing when the matched-filtering algorithm is replaced by the well-known "dechirp-Fourier transform" technique. Furthermore, corrections to account for slant-to-ground range conversion, spherical earth, etc., are often best implemented in the image domain, making intermediate image formation a valuable processing feature. 相似文献
16.
《Ocean Engineering》2004,31(8-9):1037-1062
In this paper, we present an algorithm for range image processing to reduce effects of underwater environments on the quality of range finding. This algorithm of image processing is based on the principle of the least-squares error method, which fits discrete image data to continuous piecewise curves. To simply the fitting of image data, the interval of each row and column is subdivided into several subintervals. Then a straight line is used to fit the image data within the subinterval. To merge two adjacent lines together, a weighting technique with a linear weighting factor is imposed. After the image is processed, it will provide a better imaging quality than the original one if values of the design parameters are properly assigned. Thus, a series of design of experiment process runs are carried out to study effects of the design parameters on quality of range finding. To make the proposed algorithm robust against noises, both environmental illumination and turbidity are forced into the experiments by utilizing an outer array. From the results of the range finding experiments, it was found that the proposed algorithm of image processing in truth has potential of increasing quality of range finding. The results also show that the proposed algorithm can achieve high quality of range finding only having the processing of row elements. Also, the quality of range finding by using the proposed algorithm of image processing is superior to that of using a bandpass optical filter. 相似文献
17.
确定区分星点与背景的阈值是星空图像处理的重要环节。介绍了根据背景灰度统计特征确定阈值的算法,根据星空图像灰度直方图的走势,提出一种基于灰度直方图的阈值算法;通过对实际拍摄大视场星图的处理,基于统计的阈值容易将亮背景中的亮噪声误认为星点提取,而基于直方图的阈值能够滤除这些噪点;针对星图中星点亮度不一的情况,提出在全局阈值的基础上应用Bernsen算法计算局部阈值提取暗星点,实验证明该算法能较充分地提取暗星图像。 相似文献
18.
The natural motion of a ship at sea produces undesirable effects on sonar displays. To overcome this problem some form of stabilization is desirable. A practical sonar display system, which uses image processing techniques to stabilize the displayed sonar image against rotational motion is described. The stabilization method provides the capability of image enhancement as well as improved image interpretation due to an increased area of view and a geometrically correct conversion to polar display. Details of the three processing stages required in the system are provided. These processes are ship motion analysis, image integration, and polar display conversion. Stabilized results are given for simulated ship motion 相似文献
19.
Chan-Young Park Seung-Ho Park Choon-Woo Kim Jung-Keuk Kang Ki-Hyune Kim 《Marine Georesources & Geotechnology》1999,17(4):371-386
In order to develop the mineral resources contained in manganese nodules of the deep sea, the Korea Ocean Research & Development Institute (KORDI) has explored the area allocated by the United Nations in the Clarion-Clipperton Fracture Zone in the northeastern Pacific. During research cruises, the seabed surface was photographed every 30 s by the KORDI Deep Tow Imaging System (DTIS). Features such as the coverage and size distribution of manganese nodules on the photographs serve as the essential information to determine the potential mining areas. This article presents (semi)automatic procedures to extract the useful features from the photographs of the seabed surface using digital image processing techniques. The 35-mm films are first digitized by the film scanner. The depth information written on the film is then recognized to compensate for distortions due to nonuniform illumination. The nodule areas on the digitized image are recognized and separated from the background based on the characteristics of the nodules. The nodule coverage and distribution of nodule diameters are then calculated from the processed image. The proposed technique has been applied to sample photographs of the seabed surface. Experimental results indicate that the technique could be utilized as an efficient tool to process the massive collection of photographs of the seabed surface. 相似文献