节点文献
ENVISAT ASAR森林分类研究
Study on Forest Classification by ENVISAT ASAR Image
【作者】 杨永恬;
【作者基本信息】 北京林业大学 , 森林经理, 2004, 硕士
【摘要】 在国家森林资源调查和监测中,既有宏观决策控制的需求,也有微观即局部小范围内调查监测的需要,尤其对后者,必须满足较高的森林识别精度。遥感技术的迅速发展为解决这一难题提供了有效的工具和手段。然而,可见光遥感的局限使数据的获取受制于天气或光照条件,数据源难以保证。雷达遥感具有全天时、全天候的工作能力,在林业应用上有很大的潜力。本文拟对多时相雷达数据应用于森林识别做一初步研究。 由于ERS-1和ERS-2 SAR影像仅采用单波段、单极化和固定入射角成像,无法利用多波段、多极化、多角度的雷达信息,对地类和森林的分类效果并非理想,所以我们选取双极化的ENVISAT ASAR为数据源来进行森林分类。 遥感的分类算法很多,但没有一种算法是普遍适用和最优的,这是由遥感影像本身的复杂性决定的。人们只能针对某一具体的遥感数据,不断探索新的分类算法,以求达到更好的效果。如何利用雷达影像的特征来提取更多的信息,提高森林的识别精度是本文研究的目的。通过研究得出以下结论: (1) 应用最大似然法直接对经过初步预处理的ASAR影像分类,要达到较高的森林识别精度比较困难,而通过一幅ASAR不同极化影像的加减运算和多时相ASAR影像来合成多波段影像,能提取更多的信息,从而使分类精度得到提高。 (2) 斜决策树分类方法有诸多优势,如相对简单、明确。另外,与已假定数据源呈一固定概率分布,然后在此基础上进行参数估计的常规分类方法相比,决策树属于严格“非参”,可提高森林识别精度, (3) 将H-α极化分解和分类技术应用于ASAR双极化数据(APS)的森林非森林识别,结果表明这种方法可以从一定程度上提高森林非森林的识别精度。
【Abstract】 There are not only macroscopical decision-making but also microcosmic investigation and monitoring needed in the course of national forestry investigation and monitoring. Remote sensing technology has become a powerful tool to solve the problem. But the visible light remote sensing technology is limited to receive the data because of the weather or light condition. With its all-weather, all-time capabilities, radar remote sensing has great potential for forestry management. In this paper we try to study new suitable method for forestry identification using multi-temporal ASAR data. From this study, we have got the following conclusions:(1) The multi-temporal ASAR image is practicable to forestry classification an it will be better with it’s dual polarization and multi-look angle characters(2) The classification accuracy is low if we use the maximum likelihood method to directly classify the pretreatment ASAR image. So it is necessary to use the operation of single or multitemporal ASAR image to acquire more information. So the accuracy will be improved.(3) The oblique classification tress have several advantages for remote sensing applications by virtue of their relatively simple, explicit and intuitive classification structure. In addition, decision tree algorithms are strictly nonparametric and, therefore, without assumptions regarding the distribution of input data the methods are flexible to general classifications among input features and class labels.(4) H- a polarimetric decompositon and classification technique have been applied to ENVISAT ASAR APS data for forest and non-forest classification, some encourage results have been achieved.
【Key words】 ENVISAT ASAR; polarization; multi-temporal Radar data; forest identification.;
- 【网络出版投稿人】 北京林业大学 【网络出版年期】2004年 04期
- 【分类号】S757.2
- 【被引频次】3
- 【下载频次】579