【作者】 邓旺华；

【导师】 范少辉；

【作者基本信息】 中国林业科学研究院 ， 森林培育， 2009， 博士

【摘要】 针对我国传统的竹资源监测存在时间间隔长、人力投入大、调查难度高和统计精度低等许多不足,以福建省顺昌县为研究区域,开展不同季相竹林及其周边地物反射光谱特征,以及竹林面积遥感信息提取技术和动态监测技术的相关研究,主要结果如下:(1)利用ASD光谱分析仪于2007年11月、2008年4月、7月、10月进行了4次野外光谱测定,研究对象包括:毛竹、杉木、马尾松和木荷,以及水体、旱田和水泥路面等其它地物。4个不同时期毛竹、杉木、马尾松和木荷的光谱特征研究表明:毛竹与其它植被类型具有可分性,较为理想的区分波段为500～600nm的可见光波段和670～800nm的近红外波段,而尤以556nm和718nm处区分效果最优。在遥感信息源时相选择方面,以春季的毛竹换叶期前后的时相为佳,其次为秋季。但由于南方的雨季影响,难以在该季节得到理想的影像数据。因此,秋季应为数据获取的重点时期。(2)利用ASD光谱分析仪于2007年11月对闽北地区常见的毛竹林、杉木林和马尾松林纯林,以及竹阔混交、竹杉混交、杉阔混交等混交类型进行光谱特征测定,并将各类型光谱数据按TM的1~4波段进行光谱反射率模拟和相似性矩阵分析。结果表明:毛竹与杉木的可区分度较小,而与其它类型树种之间具有较好的区分性。(3)以2001年10月的ETM数据为信息源,在采用ISODATA法、最大似然法和子像元分类法等3种常规分类方法的基础上,提出“基于光谱片层—面向纹理类”的竹林专题信息提取方法。实验结果表明,“基于光谱片层—面向纹理类”的竹林专题信息提取方法的分类精度可达84.8%,总Kappa系数达0.8235,能有效地解决由于地形因素引起的阴影问题,是提高山区竹林信息的分类精度的一种有效方法。(4)选取顺昌县1988~2007年之间的4个时相的TM/ETM影像,采用分类后比较法进行竹林面积动态监测研究。结果表明,顺昌县域的竹资源分布范围广,主要分布于边远的四周山区,而中部和西北部则分布较少。从竹林面积增长情况来看,1988 ~1992年的年平均增长面积为0.0645万公顷,增长率为2.2%;1992 ~2001年的年平均增长面积为0.0262万公顷,增长率为0.8%;2001 ~2007年的年平均增长面积为0.1629万公顷,增长率为4.8%。2001年以来,是顺昌县竹林产业发展的快速增长时期。通过野外实测不同季相条件下竹林及其周边地物的光谱特征,确定竹林与其他植被种类及生态环境因子的光谱特征差异与区分度的大小,可为竹林资源监测中卫星影像的选择和信息提取提供科学的光谱学依据,也为遥感技术在竹林资源长势、变化和健康等监测提供理论基础。通过对比分析不同的竹林面积提取精度的图像处理方法,为遥感技术在我国竹资源清查、面积变化、灾害预报及周边生态环境等动态监测中的应用提供理论基础和技术支持,对我国开展竹资源调查和监测研究具有重要的借鉴意义。更多还原

【Abstract】 According to the deficiency of long time intervals, heavey human input, difficult to survey and low statistical accuracy of the traditional bamboo resources monitoring. Taking Shunchang county as study area, carried out the reflectance spectra features characteristic of bamboo and its surrounding ground objects with different phases, research on bamboo forest area extraction with remote sensing classification techniques and dynamic monitoring technology. The research results were as follow:(1) The spectral characteristic of Phyllostachys pubescens, cunninghamia lanceolata, Pinus massoniana, Schima superba,water body, farmland and cement pavement were measured respectively in November 2007, April 2008, July 2008 and October 2008. According to the spectral characteristic of Phyllostachys pubescens, Cunninghamia lanceolata, Pinus massoniana, Schima superba in four different phases, it was shown that bamboo can be compartmentation with other vegetation types. The distinction between 500 ~ 600nm in the visible band and 670 ~ 800nm in the near-infrared band is significantly, optimal distinction result appeared on 556nm and 718nm. In the aspect of selecting phase of the remote sensing information, the best one was pre and post spring-leaf stage ,followed by the fall. However, due to the impact of the the rainy season in south area, it was difficult to get the ideal data, so the fall was the key period to get remote images data.(2) The spectral characteristic of Phyllostachys pubescens, cunninghamia lanceolata, Pinus massoniana pure forest and bamboo mixed wide, bamboo mixed cunninghamia lanceolata, cunninghamia lanceolata mixed Schima superba were measured respectively in November 2007. Then makes the human simulation of the It also makes all types of spectral data according to TM1~4 bands resampled and similarity matrix analysis. The results showed that: bamboo and cunninghamia lanceolata can distinguish little degrees, and with other types was more.(3) The research relating to remote sensing classification extraction of bamboo forest areas was carried out with the ETM data which was acquried in October 2001.Based on three kinds of conventional methods which are ISODATA method, maximum likelihood method and sub-pixel classification categories, A new thematic information extraction method which called‘Base on the spectral sheet and texture typeswas proposed’in this paper. The experimental results show that, the classification accuracy of this method can be 84.8%, and the total Kappa coefficient can be 0.8235, which can effectively solve the shadow problem caused by terrain factor. It was an effective way to improve the classification accuracy of the information about bamboo in mountain area.(4) Four phases of TM/ETM images between 1988 and 2007 in Shunchang country were selected to making research about dynamic monitoring of bamboo forest areas through post-classification comparison method.The results showed that there was wide range of bamboo resources in Shunchang county, which mainly located in remote mountainous areas and the central and north-west areas little. On the point of rising situation of the Bamboo area, the average annual growth between 1988 and 1992 was 0.0645 million hectares with the growth rate 2.2%.The average annual growth between 1992 and 2001 was 0.0262 million hectares with the growth rate 0.8%. The average annual growth between 2001 and 2007 was 0.1629 million hectares with the growth rate 4.8%.The third period was the rapid growing period of industrial development about bamboo in Shunchang County.Through measured the spectral characteristics of bamboo forest and surrounding surface features in different seasons, differences among bamboo, other vegetation and eco-environmental factors were determined to distinguish, and the spectroscopy theory of satellite images choice and information extraction were provide, and a scientific basis of remote sensing technology in the bamboo forest resources, growth, change and health monitoring was provided. By comparing the different extraction methods of bamboo forest area precision, provided technical support for bamboo resource inventory in China, and covering change, disaster forecasting and the surrounding ecological environment of the application of dynamic monitoring with remote sensing, and it was important to carry out the bamboo resource investigation and monitoring in China .更多还原