-
摘要近年来，形态学剖面（MP）已成为图像处理和模式识别研究领域的一个热点。其重要算法和理论已经应用到遥感图像研究，特别是高光谱图像处理方面。近些年来，稀疏表示在遥感图像分析领域应用十分广泛，形态学剖面特征正成为一个新的方法来为遥感图像分类。本文提出了一种基于高光谱数据集的分类方法，即遥感图像的DMP特征提取算法。方法首先将原数据集进行PCA降文，然后用differential morphological profiles（DMPs）来组合并表示原数据集的光谱和空间的数据信息，把提取到的DMP特征通过使用SVM分类器进行训练和测试，就可以将未知的图片像素点确定地理类型。34039
- 上一篇：OSG人体运动捕捉系统的数据采集软件开发
- 下一篇：手势识别系统的设计与实现
根据所考虑的属性的类型以及形态属性转换，算法当中不同的参数特征可以被模拟。实验分析首先证明了形态剖面将获得一个场景描述，利用算法提取差异图像的几何结构特征，构造深入描述图像结构化信息的特征向量空间；然后，在多文特征空间中对图像进行变化与否的判别；接着，利用数学形态学方法进行特征提取；最后，训练样本并进行测试。
通过对不同数据集和不同地形属性的实验证明了这种方法可以被应用到各种不同的高光谱数据集中，作为判断地形地貌类型的一种方法。
关键词 差分形态学剖面 PCA降文 SVM分类器 遥感高光谱图像
毕业论文设计说明书外文摘要
Title DMP Feature Extraction Algorithm Based On Remote Sensing Image
Abstract
Morphological profile(MP) has become a hot spot research field of image processing and pattern recognition in the recent years. Its major algorithms and theories have been applied to the study of remote sensing images, particularly in terms of hyperspectral image processing. As we know, the sparse representation is often widely used when operating with remote sensing image analysis. However, morphological characteristic is another useful key point for remote sensing image classification. This paper proposes a method of classification based on hyperspectral data set, which uses differential morphological profiles(DMPs).
This method first uses PCA dimension reduction to reduce the dimensions of the original data set and then uses DMP algorithm to combine and extract necessary features of the original picture. Finally, it puts the extracting characteristics into a SVM classifier, which includes training and testing part. If the parameters of the experiments are different, the result will be different too. Thus it’s also important to choose proper parameters for extracting features. The analysis of the experiment proves that the profile form will receive a scene description by using algorithm to extract geometric structure characteristics. What is important is that it can construct feature vector space to describe the original image.
Through the experiments based on different data sets and various environments, we conclude that this method can be successfully applied to practice as a means to verify the type of the exact terrain.
Keywords: DMP PCA Algorithm SVM Classification Hyperspectral Image
目 次
1 引言 1
2 高光谱图像… 2
2.1 高光谱图像数据的组成 … 2
2.2 高光谱图像的特征 … 2
2.3 高光谱图像的处理 … 3
2.4 高光谱图像的应用 … 3
2.5 高光谱图像的校正处理 … 4
3 PCA降文理论… 5
4 形态学运算… 6
4.1 腐蚀 … 6
4.2 膨胀 … 6
4.3 开运算 7
4.4 闭运算 7
5 特征提取算法 8
5.1 形态学剖面（MP）… 8