基于模糊熵和ABC算法的SAR图像分割的研究

发布时间：2019-04-17 15:30

【摘要】：随着合成孔径雷达(Synthetic Aperture Radar, SAR)技术的快速发展,SAR图像在卫星遥感、军事侦查、海洋监测以及农林业监测中发挥着十分重要的作用。SAR图像分割就是将图像中所需要的目标部分提取出来,以便于对图像进行更精确地研究。由于其独特的成像特点,SAR图像会带有相干斑噪声,所以它和普通光学图像有着很大的区别,从而在分割图像上也有着不同。本文对近年来国内外的多种图像分割方法进行了综述,其中阈值分割在图像分割中应用最为广泛,一些智能的优化算法用于图像分割当中对图像的分割有着很好的优化效果。针对SAR图像分割存在的问题,本文对基于最大模糊熵和改进蜂群的阈值分割算法的SAR图像分割进行研究,主要的工作和研究成果如下：首先,针对一维模糊熵算法只考虑到像素的灰度特性而没有考虑到邻域的空间灰度特性的弊端,引进了基于二维直方图的模糊熵的算法(简称二维模糊熵),得到灰度-邻域平均灰度的二维直方图,并将二维直方图模糊划分成目标区域和背景区域,分别求出模糊熵,计算最大阈值,用求得的阈值来对图像进行分割。算法的提出为后续工作提供了基础。然后,针对二维模糊熵算法抗噪声能力较弱的问题,本文将二维模糊熵算法进行了改进,将二维直方图改进成灰度-灰度梯度的二维直方图,使二维模糊熵算法具有更强的抗噪性。最后,由于二维模糊熵算法速度较慢,本文引进了人工蜂群算法,将二维模糊熵作为人工蜂群算法的适应度函数,通过求出最优阈值来进行图像分割。为了避免人工蜂群算法出现早熟收敛的现象,本文又将人工蜂群算法进行改进,提出了基于模糊熵和人工蜂群的算法,该算法分别改进了蜜源搜索的方式和转移概率的计算方式,大大加快了算法的速度。在图像分割开始时,先对图像进行灰度形态学预处理,然后在用该算法对图像进行分割,提高抗噪性。最后通过仿真实验,验证了提出算法的有效性和抗噪性,并证明了提出的算法能够提高算法的速率。
[Abstract]:With the rapid development of synthetic Aperture Radar (Synthetic Aperture Radar, SAR) technology, SAR images in satellite remote sensing, military investigation, Marine monitoring and agroforestry monitoring play a very important role. SAR image segmentation is to extract the part of the target needed in the image to facilitate the more accurate study of the image. Because of its unique imaging characteristics, SAR images will have speckle noise, so it has a great difference from ordinary optical images, so it is also different in segmented images. In this paper, a variety of image segmentation methods at home and abroad in recent years are reviewed, among which threshold segmentation is the most widely used in image segmentation. Some intelligent optimization algorithms for image segmentation have a good optimization effect on image segmentation. Aiming at the problems existing in SAR image segmentation, this paper studies the SAR image segmentation based on maximum fuzzy entropy and improved beehive threshold segmentation algorithm. The main work and achievements are as follows: firstly, In view of the disadvantages of one-dimensional fuzzy entropy algorithm, which only considers the gray-scale characteristics of pixels but not the spatial gray-scale characteristics of the neighborhood, a two-dimensional histogram-based fuzzy entropy algorithm (called two-dimensional fuzzy entropy) is introduced. The two-dimensional histogram of gray-neighborhood average gray level is obtained, and the two-dimensional histogram is divided into target region and background region. The fuzzy entropy is calculated, the maximum threshold is calculated, and the obtained threshold is used to segment the image. The proposed algorithm provides a basis for the follow-up work. Then, in view of the weak anti-noise ability of the two-dimensional fuzzy entropy algorithm, this paper improves the two-dimensional fuzzy entropy algorithm, and improves the two-dimensional histogram into the two-dimensional histogram of grayscale-grayscale gradient. The two-dimensional fuzzy entropy algorithm is more robust to noise. Finally, because of the slow speed of the two-dimensional fuzzy entropy algorithm, the artificial bee colony algorithm is introduced in this paper. The two-dimensional fuzzy entropy is used as the fitness function of the artificial bee colony algorithm, and the optimal threshold is obtained to segment the image. In order to avoid premature convergence of artificial bee colony algorithm, this paper improves artificial bee colony algorithm and proposes an algorithm based on fuzzy entropy and artificial bee colony. The algorithm improves the way of honey source search and the calculation of transfer probability, and greatly accelerates the speed of the algorithm. At the beginning of image segmentation, the image is preprocessed by gray-scale morphology, then the image is segmented by this algorithm to improve the anti-noise property. Finally, the simulation results show that the proposed algorithm is effective and anti-noise, and that the proposed algorithm can improve the speed of the algorithm.
【学位授予单位】：东北大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TN957.52

【相似文献】