基于几何活动轮廓模型的SAR图像海岸线检测

发布时间：2018-09-12 06:20

【摘要】：合成孔径雷达(Synthetic Aperture Radar,SAR)图像海岸线检测在海岸线管理、地图自动导航、船舰目标识别等方面发挥着重要的作用。几何活动轮廓(Geometric Active Contour,GAC)模型是在活动轮廓(Active Contour Model,ACM,又称为Snake模型)模型的基础上发展起来的,Snake模型是提取图像边界领域的重大突破性的发展,而且有非常实用的研究价值。近几年,随着Snake模型的广泛深入研究,GAC模型的思想受到了世界上广泛的关注,涉及的领域也越来越广。GAC模型在提取SAR图像边界的领域上也显示出强大的实用性。但是由于SAR图像具有边界模糊、对比度小、灰度等级多并且易受噪声干扰等问题,GAC模型的方法处理SAR图像仍然会遇到一些弱边界问题、迭代次数和迭代时间易受图像初始轮廓影响以及图像预处理对提取SAR图像的海岸线造成影响的问题。针对此问题,本文以SAR图像海岸线检测为应用背景,对其中涉及的弱边界问题、图像初始轮廓影响海岸线检测的迭代次数和迭代时间的问题及图像预处理对海岸线检测的影响进行了系统研究。经过研究SAR图像海岸线检测弱边界的特点,提出利用结合区域信息的改进符号压力函数为GAC模型的边界停止条件并对海岸线进行精确提取,这样能很好的弥补SAR图像中海岸线弱边界的不足,使得提取出的海岸线更加准确。在提取SAR图像海岸线存在弱边界问题进行研究的基础上,本文也对SAR图像初始轮廓的选取、图像预处理对提取出的海岸线的影响进行了研究。本文将研究SAR图像不同大小的初始轮廓对海岸线检测的迭代次数及迭代时间的影响,这是基于GAC模型不敏感于图像的初始轮廓。图像的初始轮廓选取越大,GAC模型的迭代次数越少,迭代次数越少则迭代时间越短。在图像的预处理中,因为SAR图像中的斑点噪声是乘性的,所以一般的图像增强方法、去除噪声方法已不适用于SAR图像,本文用灰度变换的方法对SAR图像进行增强处理,增加图像的对比度,用Lee滤波对SAR图像进行滤波处理。实验结果表明,这种图像预处理方法处理后的SAR图像,可以取得很好的检测效果。实验检测数据表明,文中方法不仅能有效的检测出SAR图像中的海岸线,而且与其他相关海岸线检测方法相比,其迭代次数减少了、迭代时间缩短了而且检测准确度得到了进一步提升,显示出该方法的有效性。
[Abstract]:Coastline detection in synthetic Aperture Radar (Synthetic Aperture Radar,SAR) images plays an important role in coastline management, map automatic navigation, ship target recognition and so on. Geometric active contour (Geometric Active Contour,GAC) model is developed on the basis of active contour (Active Contour Model,ACM, (also called Snake model) model. It is an important breakthrough in the field of extracting image boundary, and has a very practical research value. In recent years, with the extensive and in-depth study of the Snake model, the idea of the GAC model has been paid more and more attention in the world, and the domain involved in the field is also more and more extensive. The GAC model has also shown great practicability in the field of extracting the SAR image boundary. However, there are still some weak boundary problems in SAR image processing by using GAC model because of the problems of blurry boundary, low contrast, many grayscale levels and easy to be disturbed by noise. The number of iterations and the time of iteration are easily affected by the initial contour of the image and the influence of image preprocessing on extracting the coastline of SAR image. In order to solve this problem, this paper takes coastline detection of SAR image as the application background, and discusses the weak boundary problem involved in it. The effects of image initial contour on the number of iterations and iterative time of shoreline detection and the effect of image preprocessing on coastline detection are systematically studied. After studying the characteristic of detecting weak boundary of coastline in SAR image, the improved symbolic pressure function combined with regional information is proposed as the boundary stopping condition of GAC model and the coastline is extracted accurately. This method can make up for the weak boundary of coastline in SAR image and make the extracted coastline more accurate. On the basis of the research on the weak boundary of the shoreline extracted from SAR image, the selection of initial contour of SAR image and the influence of image preprocessing on the extracted coastline are also studied in this paper. In this paper, we will study the influence of the initial contour of SAR image on the number of iterations and the iteration time of coastline detection, which is based on the GAC model which is not sensitive to the initial contour of the image. The larger the initial contour selection of the image, the less the number of iterations and the shorter the iteration time of the GAC model. In image preprocessing, because speckle noise in SAR image is multiplicative, the general image enhancement method and noise removal method are no longer suitable for SAR image. In this paper, the method of gray level transformation is used to enhance SAR image. The contrast of the image is increased, and the SAR image is filtered by Lee filter. The experimental results show that the SAR images processed by this method can achieve good detection results. Experimental data show that the proposed method can not only effectively detect the coastline in SAR images, but also reduce the number of iterations compared with other relevant shoreline detection methods. The iteration time is shortened and the detection accuracy is further improved, which shows the effectiveness of the method.
【学位授予单位】：江苏科技大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TN957.52

【参考文献】