遥感影像自动变化检测技术中的样本选择研究

摘要：随着遥感技术的发展，高空间、时间、光谱分辨率的遥感影像数据急剧膨胀，如何自动的从数据中快速准确的获取目标变化信息成为当前最具有挑战性的问题之一。对于变化检测技术，利用样本选择可以快速有效的分析运用这些大量的数据。本课题针对遥感影像的变化检测技术将研究四种不同的样本选择方法，这四种样本选择方法分别为：（1）手工选取样本；（2）最大最小值法选取样本；（3）多阈值法选取样本；（4）基于超像素分割选取样本。

本文运用四种样本选择方法分别对四组遥感图像选取训练样本，并进行变化检测的试验，之后统计并比较这四种样本选择方法的综合评价指标值。实验表明手工选取样本的分类结果最优，基于超像素分割选取样本的时间复杂度最低。

关键词遥感影像变化检测样本选择最大最小值法多阈值法超像素分割

毕业设计说明书外文摘要

Title Research on Sample Selection for Automatic Change Detection of Remote Sensing Images

Abstract：With the rapid development of remote sensing technologies, high spatial, temporal, and spectral resolution remote sensing images are increasing explosively. How to detect change rapidly, accurately and automatically from remote sensing data becomes one of the most challenges in remote sensing fields. This paper analyzes four sample selection methods using several group of remote sensing images. The four sample selection methods are: (1) select the samples manually; (2) select the samples by maximum and minimum values; (3) select the samples by multi-thresholds; (4) select the samples based on the super-pixel segmentation.

After four sets of remote sensing images were tested by the four sample selection methods respectively, qualitative and quantitative evaluations of the four methods are given in the experimental section. The experiment indicates that the classification result of selecting the samples manually is optimal while the time complexity of selecting the sample based on the super-pixel segmentation is the lowest.

Keywords remote sensing image change detection sample selection multi- thresholds super pixel segmentation

1 绪论 1

1.1 研究背景及意义 .. 1

1.2 相关技术的现状 .. 1

1.3 总体技术方案及其社会影响 . 3

1.4 技术方案的经济因素分析 3

1.5 论文结构安排 . 3

2 样本选择方法 5

2.1 手工选取样本 . 5

2.2 最大最小值法选取样本 .. 6

2.3 多阈值法选取样本 7

2.4 基于超像素分割选取样本 9

2.5 本章小结 . 11

3 支持向量机 12

3.1 线性判别函数 12

3.2 最优分类超平面 . 13

3.3 核函数变换与非线性支持向量机 .. 15

3.4 libsvm 16

3.5 本章小结 . 17

4 基于遥感影像变化检测的样本选择的实验研究 18

4.1 实验结果评价指标 .. 18

4.2 实验结果 . 18

4.3 实验分析 . 25

4.4 本章小结 . 26 结论遥感影像自动变化检测技术中的样本选择研究:http://www.chuibin.com/jisuanji/lunwen_205622.html