基于主动轮廓模型的遥感图像海岸线检测方法

发布时间：2018-06-28 11:07

本文选题：遥感图像 + 主动轮廓模型　；参考：《大连海事大学》2014年硕士论文

【摘要】：海岸线检测在自动导航、地理勘探、海域资源管理、海域环境保护和海域发展规划中具有重要地位,遥感图像是海岸线检测的重要手段之一。传统人工解译方法费时费力,因而逐渐被计算机解译方法代替。在计算机解译方法中,传统的图像处理方法存在较多的图像后处理,图像过分割问题。而主动轮廓模型的方法依靠其连续光滑的曲线自动逼近海岸线,近年来一直受到学术界的广泛关注。目前在海岸线检测中,主动轮廓模型方法主要采用基于边缘的方法,存在初始轮廓位置敏感、易受噪声影响等问题。为解决上述问题,本文针对三种具有不同特征的含海遥感图像,分别给出了三种不同的改进主动轮廓模型,主要研究内容如下： 1)给出了一种基于区域局部均值递减的改进主动轮廓模型。针对背景区域(海面)均匀,目标区域(陆地)不均匀,并含有伪点和近似背景的弱区域的遥感图像,构建一种基于区域均值递减的改进主动轮廓模型,并对改进的能量泛函进行了公式的推导,给出了算法的描述,最后在遥感图像上对这种改进模型进行了实验对比和分析,并且在其他非遥感图像上进行了拓展。实验表明,改进模型具有良好的效果,较快的运行时间。 2)给出了一种基于区域局部均值递增的改进主动轮廓模型。针对背景区域(海面)含强度噪声,目标区域(陆地)不均匀,含有伪点的遥感图像,构建一种基于区域均值递增的改进的主动轮廓模型,并对改进的能量泛函进行公式的推导和局部均值收敛性的必要性证明,同时给出算法的描述,最后在遥感图像上进行了实验对比和分析,并且在其他非遥感图像上进行了拓展。实验表明,改进模型具有良好的效果,较快的运行时间。 3)给出了一种基于边缘的区域梯度改进主动轮廓模型。针对背景区域(海面)较为均匀,目标区域(陆地)不均匀,且像素分布较为复杂的遥感图像,推导出改进的边缘主动轮廓模型,给出了算法的描述,在遥感图像上进行了实验对比和分析。实验表明,改进模型具有良好的效果。
[Abstract]:Coastline detection plays an important role in automatic navigation, geographic exploration, sea area resource management, sea area environmental protection and sea area development planning. Remote sensing image is one of the important means of coastline detection. The traditional artificial interpretation method is time-consuming and laborious, so it is gradually replaced by the computer interpretation method. There are many image post-processing, image over segmentation problem. And the active contour model method relies on its continuous and smooth curve to automatically approach the coastline. In recent years, the active contour model has been widely concerned by the academic circle. At present, in the coastline detection, the active contour model method mainly uses the edge based method and has the initial contour. In order to solve the above problems, this paper presents three different active contour models for the three different features of the sea - bearing remote sensing images.
1) an improved active contour model based on regional local mean diminishing is presented. A modified active contour model based on regional mean decline is constructed, and an improved active contour model is constructed based on the region (sea surface) uniform, the target region (land) inhomogeneous and the weak region of the pseudo point and the approximate background. The description of the algorithm is given. Finally, the experimental comparison and analysis of the improved model on remote sensing images and the expansion of the other non remote sensing images have been carried out. The experiment shows that the improved model has a good effect and a faster running time.
2) an improved active contour model based on regional local mean increment is given. In view of the intensity noise of the background region (sea surface), the target region (land) is not uniform, the remote sensing image containing the pseudo point, an improved active contour model based on the increasing mean value of the region is constructed, and the formula for the improved energy functional is derived and local. The necessity of mean convergence is proved, and the description of the algorithm is given at the same time. Finally, experimental comparison and analysis are carried out on remote sensing images, and the expansion of other non remote sensing images is carried out. The experiment shows that the improved model has good effect and faster running time.
3) a regional gradient improved active contour model based on edge is given. The improved edge active contour model is derived for remote sensing images with more uniform background region (sea surface), uneven target area (land) and complex pixel distribution, and the description of the algorithm is given. Experimental comparison and analysis are carried out on remote sensing images. The experiment shows that the improved model has good effect.
【学位授予单位】：大连海事大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TP751

【参考文献】