高分系列遥感影像高大地物阴影检测方法研究

发布时间：2018-04-11 15:04

本文选题：高分系列卫星 + 阴影检测　；参考：《兰州交通大学》2017年硕士论文

【摘要】：在高分影像中,普遍存在大量地物阴影。阴影的存在降低影像质量造成失真,并且使部分信息难以获取,严重影响地物信息的识别与统计。因此,高效、准确地检测和消除高分辨遥感影像中的地物阴影已成为业内研究、应用热点和难点。高分系列影像(GF)是指国产最高分辨率卫星遥感影像,通过国内外文献分析,还未发现针对GF-1及GF-2等我国高分影像的阴影检测算法,因此,论文提出3种针对GF-1影像的阴影检测方法和1种针对GF-2影像的阴影检测方法。主要研究成果如下:(1)提出了针对GF-1的基于多特征的影像阴影检测方法首先,对GF-1影像多光谱数据、全色数据进行正射校正和信息融合,以达到光谱与分辨率信息最大化利用,突出阴影信息,增大阴影与其他地物的差异。其次,建立基于主成分变换的阴影检测方法PC1/NIR和图像特征计算Rg_nir,得到初始阴影信息和水体信息;最后,用初始阴影信息与水体信息做逻辑非运算,从而达到剔除水体的目的,并利用形态学闭运算使得阴影区域更加连续,饱满。(2)提出了针对GF-1的基于HSV与PCA集成的阴影检测方法首先,利用主成分变换方法分割阴影与非阴影区域,分离后的阴影区域含有水体、深色地物信息;其次,对HSV色彩空间的V分量利用阈值法分割阴影和非阴影区域,分离后结果含有暗色植被,但不含有水体跟深色地物信息。最后,对两次计算结果进行逻辑与运算,从而剔除混合阴影区域中水体、深色地物以及暗色植被等信息,获得高精度阴影区信息。(3)提出了针对GF-1的集成特征分量的多尺度分割方法首先,对GF-1影像多光谱数据、全色数据进行正射校正和信息融合;其次,集成特征分量(主成分第一分量PC1、亮度分量V、绿光波段G、归一化植被指数NDVI)以增强阴影信息;最后,对集成后的影像进行多尺度分割,并利用特征分量构建规则集,最终实现阴影信息提取。(4)提出了针对GF-2的基于主成分分析的阴影检测方法首先,对GF-2影像多光谱数据进行典型地物光谱信息统计、分析,并利用主成分变换与波段计算方法分割阴影与非阴影信息;其次,利用归一化计算得到植被细信息;最后,对两次计算结果进行逻辑非运算,从而得到准确阴影信息。大量实验分析表明,论文提出的阴影检测方法即能高效、准确的提取出GF-1、GF-2影像阴影,又能削弱水体、深色地物、暗色植被的干扰,有效解决了国产高分系列影像的阴影检测的技术难题,为后续的阴影补偿和应用提供了有效地数据支持。
[Abstract]:In high score images, there are a large number of shadows of ground objects.The existence of shadow reduces the distortion of image quality, and makes part of information difficult to obtain, which seriously affects the recognition and statistics of ground object information.Therefore, detection and elimination of ground shadow in high resolution remote sensing image efficiently and accurately has become a hot and difficult point in the field of application.High score series image (GFF) refers to the highest resolution satellite remote sensing image made in China. Through literature analysis at home and abroad, no shadow detection algorithm has been found for high score images such as GF-1 and GF-2 in China, so,This paper presents three shadow detection methods for GF-1 images and one shadow detection method for GF-2 images.The main research results are as follows: (1) A multi-feature based shadow detection method based on GF-1 is proposed. Firstly, forward correction and information fusion for multi-spectral data and panchromatic data of GF-1 images are carried out to maximize the utilization of spectral and resolution information.Highlight shadow information and increase the difference between shadow and other features.The shadow detection method based on the integration of HSV and PCA for GF-1 is proposed. Firstly, the principal component transform (PCA) is used to segment the shadow and non-shadow regions.Secondly, the V component of HSV color space is divided into shadow and non-shadow regions by threshold method. The result of separation contains dark vegetation, but does not contain information of water and dark features.Finally, the logic and operation of the results of the two calculations are carried out to eliminate the information of water body, dark features and dark vegetation in the mixed shadow area.This paper presents a multi-scale segmentation method for the integrated feature components of GF-1. Firstly, the multi-spectral data and panchromatic data of GF-1 images are corrected and fused.The feature component (principal component first component PC1, luminance component V, green light wave segment G, normalized vegetation index NDVI) is integrated to enhance shadow information. Finally, the integrated image is segmented at multiple scales, and the rule set is constructed by using the feature component.Finally, shadow information extraction is realized. (4) A shadow detection method based on principal component analysis (PCA) for GF-2 is proposed. Firstly, the multispectral data of GF-2 image are analyzed and analyzed.The method of principal component transform and band calculation is used to segment the shadow and non-shadow information. Secondly, the fine information of vegetation is obtained by normalized calculation. Finally, the results of the two calculations are logically non-calculated, and the accurate shadow information is obtained.A large number of experiments show that the shadow detection method proposed in this paper can extract the shadow of GF-1 / GF-2 image efficiently and accurately, and weaken the interference of water, dark features and dark vegetation.It can effectively solve the technical problem of shadow detection of domestic high score series images and provide effective data support for shadow compensation and application in the future.
【学位授予单位】：兰州交通大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：P237

【参考文献】