摘要近年来,高光谱遥感得到了广泛的应用,如地物分类、地质勘查、环境监测、精细化农业等。虽然高光谱数据为人们带来了更加丰富的地物信息,但由于信息量的大大增加,在影像分类方面给传统分类方法带来了挑战。因此,有必要在经典的针对多光谱影像的分类方法上,进一步探讨适用于高光谱遥感数据的分类方案。32364
本研究以我国黑河流域为试验区。采用欧空局PROBA-CHRIS高光谱多角度数据,对其进行包括去条带、表观反射率计算、几何校正、特征提取降文等预处理。通过应用ISODATA法、支持向量机、最大似然法分别对黑河流域上、中、下游的不同地表类型的CHRIS 0度数据进行分类。在对结果分析比较的基础上,得出流域不同区域下垫面条件下分类精度最高的分类方法。然后针对黑河流域的不同区域,对5个角度的影像,分别选择精度最高的分类方法进行分类。结果表明,针对本研究的地表景观特点,最大似然法对城区和农田为主的地表分类方法是最为适宜的,而支持向量机法对以山地和荒漠为主的地表分类最为适宜。分类结果的误差主要来源于预处理的误差、降文带来的一些信息缺失等。
毕业论文关键词:高光谱遥感,多角度遥感,PROBA-CHRIS,分类,黑河流域
Abstract Hyperspectral remote sensing has been widely used in various applications, such as land cover classification, geological exploration, environmental monitoring and agriculture utilizations. Although hyperspectral data can afford more detailed information of land features, there also have more challenges in data processing, especially in image classification.
This study selects the Heihe River Basin as study area, and PROBA-CHRIS observations are used. The hyperspectral and multi-angular data are firstly pre-processed, such as destriping, radiation calibration, atmospheric correction, geometric correction and feature extraction. Secondly, several methods are used to perform image classification, including Support Vector Machine (SVM), Maximum Likelihood Method (MLM) and Iterative Self-Organizing Data Analysis Technique (ISODATA). These approaches are assessed by using classification results obtained over different regions of the study area with nadir CHRIS observations. On this basis, the most robust method suitable to each different landscapes are used for all observations with 5 incidence angles. Results show that according to the landscape characteristics of this research, the MLM can give best result in urban areas and farmland surface. The SVM is superior for mountainous and desert surfaces. Error involved in this study may be caused by the error of data pre-processing and classification step.
Key words:Hyperspectral remote sensing, multi-angle remote sensing, PROBA-CHRIS, classification, Heihe river basin
目 录
摘要Ⅰ
AbstractⅡ
1 绪论1
1.1 研究背景1
1.1.1高光谱遥感概况1
1.1.2 PROBA-CHRIS卫星传感器2
1.2 高光谱遥感分类国内外研究现状 4
1.3 主要研究内容和技术路线5
2 研究区与数据7
2.1 研究区 7
2.2 遥感数据 7
3 PROBA-CHRIS数据预处理11
3.1 去条带 11
3.2 表观反射率计算13
3.3 几何校正16
4 PROBA-CHRIS高光谱多角度数据分类18
4.1 特征波段提取与降文 18
4.2 研究区地表分类方法与结果 19
4.2.1 ISODATA法19
4.2.2 最大似然法19
4.2.3 支持向量机法19
4.2.4 分类结果精度评定 20
5 结论和展望31
致谢 32
参考文献 33
1 绪论
1.1 研究背景
遥感即以不接触的方式从高空对地球进行探测,通过对传感器接收到的数据进行分析,以研究被探测物的性质、状态、变化过程等。遥感科学综合了数学方法,物理学原理以及相关地学规律,是一门交叉性的学科。遥感的出现,把人们的观测视线从地球表面推向了高空,对地观测从此步入了一个崭新的阶段。 基于PROBA-CHRIS高光谱数据的黑河流域地物分类研究:http://www.751com.cn/gongcheng/lunwen_28902.html