基于光谱匹配技术的青藏高原典型植被识别与提取

发布时间：2018-04-28 08:01

本文选题：平滑滤波 + 垂直条纹去除　；参考：《中国地质大学(北京)》2017年硕士论文

【摘要】：光谱匹配技术是成像光谱地物识别的关键技术之一,如何针对不同的应用选择不同的光谱匹配技术以及能否发展更完善的技术一直是人们研究的热点。受限于青藏高原的特殊地理环境特点,目前国内对于青藏高原植被分类的研究较少,而植被类型及其变化对于青藏高原生态环境演变及其全球变化响应研究又具有重要的意义。本文采用HJ-1A高光谱影像和ASD实测高光谱数据进行了青藏高原可可西里典型区高原植被的识别与提取研究。首先对实测高光谱数据利用Savitzky-Golay滤波器进行光谱数据的平滑与去噪处理;并采用全局去条纹法解决了HJ-1A高光谱数据存在的垂直条纹问题;然后基于预处理后的实测波谱建立了高原植被光谱数据库;最后采用了交叉相关光谱匹配、光谱角匹配、光谱特征拟合、小波变换等多种光谱匹配方法对小蒿草、针茅、青藏苔草等高原植被进行了匹配识别试验和提取效果对比研究。结果表明:(1)光谱特征参数和光谱吸收指数能较好的提取出绿色健康植被特有的光谱特征;交叉相关光谱匹配在进行相近地物光谱的区分时需要有峰值和偏度系数的最优权值;小波变换光谱匹配受限于光谱匹配数据的误差阈值,对于青藏高原的复杂环境和卫星遥感数据的低空间分辨率,小波变换光谱匹配方法的能力发挥有所局限;光谱角匹配和波谱特征拟合都是常见的光谱匹配方法,光谱角匹配更多地考虑了光谱的形状特征,而光谱特征拟合着重于光谱的本质,更能有效地提取出具有明显光谱吸收特征的地物光谱。(2)本文通过上述多种光谱匹配方法的试验对比,发现都能较好的识别出研究区小蒿草光谱;小波变换光谱匹配和光谱特征拟合都能较好的识别出研究区针茅光谱,交叉相关光谱匹配在针茅光谱与青藏苔草光谱之间无法区分;最后,采用光谱角填图方法提取了相应研究区小蒿草和针茅的植被覆盖区域,并用1:100万中国植被图进行了对比验证。
[Abstract]:Spectral matching technology is one of the key technologies in the recognition of imaging spectral features. How to select different spectral matching techniques for different applications and whether to develop more perfect techniques have been the focus of research. Limited by the special geographical environment characteristics of the Qinghai-Xizang Plateau, there are few researches on vegetation classification in China at present, and the vegetation types and their changes are of great significance to the study of the ecological environment evolution and the global change response of the Qinghai-Xizang Plateau. In this paper, HJ-1A hyperspectral image and ASD hyperspectral data are used to identify and extract vegetation in the typical area of Hoh Xili, Qinghai-Xizang Plateau. Firstly, the hyperspectral data are smoothed and de-noised by Savitzky-Golay filter, and the problem of vertical stripe in HJ-1A hyperspectral data is solved by global striping method. Then the vegetation spectral database of the plateau was established based on the measured spectrum after pretreatment. Finally, the cross correlation spectral matching, spectral angle matching, spectral feature fitting, wavelet transform and other spectral matching methods were applied to Artemisia annua and Stipa chinensis. The matching identification test and extraction effect of the vegetation in Qinghai-Xizang Plateau were studied. The results show that the spectral characteristic parameters and spectral absorption index can extract the special spectral characteristics of green and healthy vegetation, and the cross-correlation spectral matching needs the optimum weight of peak value and deviation coefficient when distinguishing the spectrum of similar ground objects. Wavelet transform spectral matching is limited by the error threshold of spectral matching data, which is limited to the complex environment of Qinghai-Xizang Plateau and the low spatial resolution of satellite remote sensing data, so the ability of wavelet transform spectral matching method is limited. Both spectral angle matching and spectral feature fitting are common spectral matching methods. Spectral angle matching takes more account of spectral shape features, while spectral feature fitting focuses on the essence of spectrum. In this paper, it is found that the spectrum of Artemisia annua in the study area can be better identified by the experimental comparison of various spectral matching methods mentioned above. Wavelet transform spectrum matching and spectral feature fitting can identify the spectrum of Stipa przewalskii, and cross-correlation spectrum matching can not be distinguished between the spectrum of Festuca and Qinghai-Xizang Carex. The vegetation cover areas of Artemisia annua and Stipa arundinacea were extracted by spectral angle mapping and compared with 1: 1 million Chinese vegetation map.
【学位授予单位】：中国地质大学(北京)
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：Q948

【相似文献】