遥感图像分割算法研究与实现电子书 摘 要 图像分割是由图像处理过渡到图像分析的关键步骤。一方面,它是目标表达的基础,对特征测量有重要的影响。另一方面,因为图像分割及其基于分割的目标表达、特征提取和参数测量等将原始图像转化为更抽象更紧凑的形式,使得更高层的图像分析和理解成为可能。遥感图像通常表现出:灰度级多、信息量大、边界模糊、目标结构复杂等等。由于遥感图像的这些特点,使得对遥感图像的分割没有完全可靠的模 型进行指导,因而在一定程度上阻碍了分割技术在遥感领域的应用。本文针对遥感图像表现出的特点,研究并改进了传统的图像分割技术,使之适应遥感图像的特点并应用于遥感图像分割中。主要工作包括如下几个方面。首先,针对当前主流的图像分割算法进行了分析、分类、归纳和总结,指出了各类方法的优缺点,为人们在不同的应用场合及不同的图像数据条件下选择不同的分割算法提供了一定的依据。其次,分析了Otsu 算法进行阈值分割的原理和有效性,针对Otsu 多阈值分割算法计算效率低的缺点,提出了用Nelder-Mead 单纯形法对Otsu 的多阈值分割进行优化,并完成仿真实验。接着,分析了SUSAN 算法进行边缘检测的原理和有效性,在此基础上对SUSAN 算法进行了边缘检测精度的改进,并用于地物边界提取中,使地物边界提取更清晰,准确,层次感分明。最后,针对区域生长分割算法的生长点确定问题,与分水岭算法结合起来,既弥补了分水岭算法的过分割问题,又解决了区域生长法生长点确定问题。本文综合两种算法并结合异质性最小区域合并准则对遥感图像分割做了理论分析,并进行了仿真实验。 关键词 遥感图像;分割算法;边缘检测;阈值;区域生长;面向对象 目 录 摘 要··················································································································· I Abstract················································································································II 第1 章 绪论········································································································· 1 1.1 研究背景··································································································· 1 1.2 国内外研究现状······················································································· 2 1.3 本文研究的内容······················································································· 5 1.4 研究意义··································································································· 6 1.5 本文组织结构··························································································· 7 第2 章 图像分割方法概述················································································· 8 2.1 图像分割定义··························································································· 8 2.1.1 图像分割的地位················································································ 8 2.1.2 图像分割定义···················································································· 8 2.2 经典图像分割方法··················································································· 9 2.2.1 边缘检测分割法················································································ 9 2.2.2 阈值分割法······················································································ 12 |
查看评论
已有0位网友发表了看法