基于数学形态学的高空间分辨率遥感影像几何特征提取  

Geometrical Features Extracting from High Spatial Resolution Images with Mathematical Morphology Methods

作　　者：刘生[1] 王潇宇[2] 邱新法[2] 

机构地区：[1]南京信息工程大学遥感学院,南京210044 [2]南京信息工程大学计算机与软件学院遥感学院,南京210044

出　　处：《地球信息科学》2008年第2期251-256,共6页Geo-information Science

基　　金：中国气象局新技术推广重点项目(CMATG2006Z10);浙江省科学技术厅社会发展项目(2006C33054);南京信息工程大学科研基金项目(20070024;20070070)资助

摘　　要：本文结合高空间分辨率遥感影像的特点,利用数学形态学方法,选取具有不同尺度和包含全部方向的结构元素,设计了全方位结构元素多级加权滤波去噪算法和多尺度全方位形态学边缘检测算法,用于高空间分辨率遥感影像的处理;并通过图像边界追踪生成栅格数据对象和矢量数据对象,据此建立了高空间分辨率遥感影像的几何特征提取模型。结果表明:全方位结构元素多级加权滤波去噪算法很好地抑制了图像中的噪声,并保留了图像细节;多尺度全方位形态学边缘检测算法很好地解决了噪声抑制和精细边缘提取的矛盾,检测出的图像边缘比基本的边缘检测算子清晰,而且抗噪性能强。提取的几何特征信息可以结合遥感图像的光谱、纹理、统计等特征用于遥感图像地物识别与分类,也可以在GIS、摄影测量、计算机视觉等领域和气象、农林、地理、海洋、水利、国土资源和环保等行业使用。Based on mathematical morphology methods, two main algorithms were developed, such as multi-weighted combination filter algorithm with omnidirection structure elements and edge detection algorithm with muhi-scale omnidirectional structure elements. By tracing the image edge, both raster and vector data objects were generated, and geometrical features extracting model was built. Results suggest that： the purposed algorithms can successfully process remote sensing images with high-spatial-resolution. The former suppresses noise and preserves details effectively ; the latter resolves the contradictions between noise suppression and extraction of fine edge excellently, and anti-noise performance is strong. The extracted geometrical features information can be used in object recognition and classification of remote sensing image with its spectrums, texture, and statistics features. The extracted information can be used in the fields of GIS, photographic surveying, computer vision and the industries of meteorology, agriculture and forest, ocean, hydropower, land resources, and environmental protection.

关 键 词：高空间分辨率遥感影像 几何特征提取 数学形态学 

分 类 号：TP751[自动化与计算机技术—检测技术与自动化装置]