多光谱影像的边缘信息提取方法研究

How to use all of the spectral informations from remote sensing is an important research field, particularly for the extraction of edges as the methods proposed in all the textbooks and papers are only designed for single band of image and they can't work for multispectral or hyperspectral images. This project is performed to resolve the problem of the lack of the theories and methods for the edge extraction from multispectral or hyperspectral images. The analysis in spectral vector or image fusions are proposed for using the approaches which were applied only for single band image but now for multispectral images. The method of inter-correlation analysis with spectral vectors has been suggested for edge detections, as well as that based on the density analysis for spectral space, including the determinations of statistic unit for density analysis, and the regions of low density. The analysis approaches for choosing principal bands, for instance, the method based on spectral responsible curves of main types of ground features, and the comparison of differencial operator in 8 directions and inter-correlation analysis of neibour field with the data of those bands are proposed and tested. The results of researches have been introduced in the international and national acdamic conferences. Three papers related have been EI achieves...

本项目主要研究多光谱影像边缘信息提取的数学模型和方法，将针对单波段影像边缘提取算子、Marr-Hildreth理论和非线性检测方法等拓展到多波段影像，提出光谱空间密度的统计治龇椒ê驮诖嘶∩系谋咴敌畔⒓觳夥椒岸嘈诺婪椒ǖ龋猿涫岛头岣挥跋窭斫獾睦砺酆头椒ǎ谘跎辖徊娇寡芯亢驮谑导手性擞么蛳禄