面向对象的高分辨率遥感信息提取技术分析-地理学(城市与区域规划)专业毕业论文.docxVIP

华中科技大学硕士学位论文 华 中 科 技 大 学 硕 士 学 位 论 文 II II Abstract High spatial resolution remote sensing technology in recent years made significant progress. IKONOST and QuikBird make obtaining high-resolution remote sensing images easily. High resolution remote sensing image shows more information of surface features. How to make use of abundant information of high resolution remote sensing image is a much-needed problem. A new method based on new ideas of object-oriented remote sensing image analysis is used to solve the problem of low accuracy faced by traditional methods. This method utilized multi-scale and multi-features to optimize information extraction result. PGF filter is a capable method for high-resolution remote sensing image pre-processing, which remove noise effectively with boundary information preserved. Watershed algorithm and region-grow algorithm with shape and spectrum combined as homogeneity measurement can divide raster images into similar objects effective. Classifier, which based on regional characteristics of object-oriented information, is trained with the pre-entered samples features. Information of remote sensing image classification is obtained efficiently by this way. This method has achieved the precision of 90 percent in applications on high-resolution remote sensing images. It overcomes the lack of surface texture and shape information in traditional extraction algorithm. It‘s proved to be an effective object oriented remote sensing image extraction method based on multi-scale and multi-feature. Keyword: Information Extraction; PGF; Multi-Scale Segmentation; Multi-Feature Fusion PAGE IV PAGE IV 目 录 摘 要 I Abstract II 1 绪论 1.1 选题背景 (1) 1.2 遥感信息提取技术研究现状 (2) 1.3 研究内容与结构安排 (5) 2 面向对象遥感信息提取系统设计 2.1 多尺度遥感图像分割算法设计 (8) 2.2 面向对象的遥感图像分类算法设计 (10) 2.3 本章小结 (13) 3 多特征的遥感图像分割算法 3.1 预处理算法设计 (14) 3.2 分割算法设计 (16) 3.3 本章小结 (25) 4 面向对象的分类算法 4.1 特征选择与提取 (26) 4.2 分类算法研究 (29) 4.3 本章小结