乔木树种遥感监测波段窗口研究

发布时间：2018-04-19 20:27

  本文选题：遥感 + 监测　； 参考：《中南林业科技大学》2017年硕士论文

【摘要】：随着高光谱技术的不断发展,高光谱遥感在树种识别方面的应用越来越广泛。传统遥感只能识别植被、水体、裸地等差异较大的地类,无法满足乔木树种的精细识别要求。高光谱数据具有光谱分辨率高、波段数目多、数据量庞大的显著特点,为乔木树种的精细分类与识别提供了可能。乔木树种的分类一直是林业遥感监测研究的一个技术难点,为寻找便于识别乔木树种高光谱数据的最佳监测窗口。本文以黄丰桥国有林场为主要研究区,以杉木、马褂木、马尾松和樟树为研究对象,以JM(Jeffreys-Matusita)距离和变换离散度为最佳监测波段的筛选算法,然后使用马氏距离判别法对计算所得的最佳波段窗口进行验证,从而确定最佳波段区间的范围。此外,利用MATLAB软件GUI界面构建框架,结合开发组件与高光谱数据处理算法,开发相应的数据处理软件,实现对高光谱数据的一键式处理,快速、高效的得出最佳监测波段窗口。本文历时2年多,对杉木、马褂木、马尾松和樟树进行定点定期观测,共采集数据2000多条,利用以上观测数据,进行高光谱数据的预处理、相关性分析和相对辐射校正,然后使用JM距离算法和变换离散度算法计算树种之间在所有光谱区间(340nm~2500nm)上的区分距离值,以大于等于1.9的波段区间为最佳监测窗口,再用马氏距离判别法对得到的最佳波段窗口进行验证和精度评价,最终确定乔木树种遥感监测的最佳波段窗口。为快速、高效处理高光谱数据提供一定的参考,为乔木树种遥感监测提供最佳的波段窗口。主要研究结果有:(1)对于实测乔木树种地面高光谱数据最佳监测波段的选择,研究得到高光谱数据滤波、相关性分析、相对辐射校正等比较成熟有效的高光谱数据预处理方法。(2)对于实测乔木树种地面高光谱数据最佳监测波段的选择,研究得出JM距离算法作为乔木树种遥感监测波段筛选算法比较理想。(3)研究得出乔木树种遥感监测最佳波段窗口为:杉木与马褂木为1572～1591nm;杉木与马尾松为993～1013nm;杉木与樟树为2139～2159nm;马褂木与马尾松为729～742nm,1601～1606nm;马褂木与樟树为1710～1729nm;马尾松与樟树为1596～1625nm。(4)研发了一套一键式、快速管理与分析高光谱数据的数据处理系统。通过可视化界面,实现研究人员对数据的快速、高效处理,以及数据对人的及时响应,形成人与数据之间的直接价值连接,将数据存储管理与数据处理分析集成到一起。使得系统可以完成高光谱数据的管理与存储,又能完成对高光谱数据的深入处理和分析,形成一个数据管理与分析处理的一体化平台。
[Abstract]:With the development of hyperspectral technology, hyperspectral remote sensing is more and more widely used in tree species identification. Traditional remote sensing can only identify vegetation, water, bare land and other land species, which can not meet the requirements of fine identification of tree species. The hyperspectral data have the characteristics of high spectral resolution, large number of bands and large amount of data, which provides the possibility for the fine classification and recognition of tree species. The classification of tree species has always been a technical difficulty in forestry remote sensing monitoring. In order to find the best monitoring window for identifying hyperspectral data of tree species. In this paper, Huangfeng Bridge National Forest Farm is taken as the main research area, Chinese fir, mandarin, Masson pine and camphor tree as the research object, and the distance and transform dispersion of JMJM Jeffreys-Matusita as the best screening algorithm for monitoring band. Then the Mahalanobis distance discriminant method is used to verify the optimal band window and to determine the range of the best band range. In addition, using the MATLAB software GUI interface to build a framework, combined with the development of components and hyperspectral data processing algorithm, the development of the corresponding data processing software, to achieve the hyperspectral data one-key processing, fast and efficient to obtain the best monitoring band window. In this paper, the Chinese fir, mandarin, Pinus massoniana and camphor tree were observed at fixed points for more than 2 years. More than 2000 data were collected, which were used for preprocessing, correlation analysis and relative radiation correction of hyperspectral data. Then the JM distance algorithm and the transform dispersion algorithm are used to calculate the distance value of the tree species on all spectral intervals (340 nm-1). The best monitoring window is the band range greater than 1.9. Finally, the best band window of tree species is determined by using Markov distance discriminant method to verify and evaluate the precision of the best band window. It provides a certain reference for fast and efficient processing of hyperspectral data and provides the best band window for tree species remote sensing monitoring. The main results of this study are: (1) selection of the best monitoring band for the ground hyperspectral data of tree species measured. The filtering of hyperspectral data and the analysis of correlation are obtained. Relative radiation correction is a mature and effective preprocessing method for hyperspectral data. The selection of the best monitoring band for the ground hyperspectral data of tree species measured. It is concluded that JM distance algorithm is an ideal selection algorithm for remote sensing monitoring band of Arbor species.) the optimum band window of remote sensing monitoring for Arbor tree species is 1572 ~ 1591 nm for Chinese fir and Pinus mandarinensis; 9931 ~ 1013 nm for Chinese fir and Masson pine; and 931 nm for Chinese fir and camphor. A one-button model was developed for mandarin and Pinus massoniana at 2139nm; Masson pine and Pinus massoniana for 1601nm; mandarin and camphor for 1710101729nm; Pinus massoniana and camphor for 1596Nm. A data processing system for fast management and analysis of hyperspectral data. Through the visual interface, the researchers can deal with the data quickly and efficiently, and the data can respond to the people in time, form the direct value connection between the data and the data, and integrate the data storage management and the data processing analysis together. The system can complete the management and storage of hyperspectral data, and also complete the in-depth processing and analysis of hyperspectral data, and form an integrated platform for data management and analysis.
【学位授予单位】：中南林业科技大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：S771.8
，

本文编号：1774596