【作者】 蔡丽玲；

【导师】 石军南；

【作者基本信息】 中南林业科技大学 ， 森林经理学， 2008， 硕士

【摘要】 土地利用实地调查费时、费力、耗资、效率低下。当今,随着遥感与地理信息系统技术的兴起与日益成熟,充分利用这些高科技成果,对土地资源进行调查已成为可能并为人们所共识。然而,遥感信息不确定的特性严重影响了遥感的功能、效率和灵活性,制约着遥感信息的产品化和实用化的进一步发展。本文以红水河流域册亨县弼佑乡的QuickBird影像数据为例,分析其在土地利用信息提取中出现的不确定性及原因,并结合其它信息,以遥感、地理信息系统为主要技术手段对影像中的不确定性信息进行处理研究,具有一定的针对性和相对有效性,不仅改善了遥感影像数据,也提高了由影像数据而得来的GIS土地利用调查数据质量,从而减少了影像识别错误,在以后的实际生产应用中具有实用价值。研究的主要结论如下：(1)运用遥感图像处理软件,辅以地形图、GPS实测点、DEM等相关的地学信息对研究区的QuickBird影像进行几何精校正、正射校正,校正后能够提高所获取的数据质量,减少影像的不确定性。(2)采用遥感数据定量分析的方法,计算出各波段的标准差、信息熵、空间相关系数、光谱相关系数,通过对比发现PCA分析法为最佳融合方法,融合图像的标准差、信息熵和相关系数均大于其它几种算法,融合后的图像在视觉上、保持空间信息细节上均具有较为明显的改善且目视区分度较好。为了突出植被信息,采用公式(float (b2)+float (b4))/2进行波段计算,减少和抑制了对被感知对象或环境解释中可能存在的多义性、不完全性和不确定性。(3)运用遥感图像处理软件和MATLAB相结合的办法对被薄云覆盖的遥感影像进行云噪声处理研究。基于R、G、B通道和基于HIS变换进行去云处理。基于HIS变换去云获得较好的效果,使云层底部地物得到清晰显示,边界清楚,内部纹理丰富,便于进一步精确识别与解译,但也使图像丢失了一部分光谱信息,图像中无云地区的背景细节被削弱了；基于R、G、B通道去云,提高了影像信息分辨率,图像背景信息保持得很好,图像信息更加丰富。(4)为解决部分纹理特征带来的不确定性问题,采用地学信息进行辅助处理。用特征点、线加入到不规则三角网中生成高精度的数字高程模型。用无洼地数字高程模型生成河流网络信息,采用数字高程模型提取坡度图。将河流网络信息、坡度图与遥感影像进行叠加分析,判别河流走向,区分河流支流与水流冲刷地,常年旱地与陡坡地。结果表明利用河流网络判定河流走向,区分河流支流与水流冲刷地,准确率为100%。利用坡度图区分常年旱地与陡坡地,陡坡地的准确率为92.1%,常年旱地的准确率为94.2%。(5)通过对影像的光谱特征与混合像元的空间分布特征分析发现影像上地形破碎处、多种地物类型交界处不确定性情况比较多。为降低“同物异谱”、“同谱异物”、混合像元带来的不确定性问题,利用ARCGIS中的Model builder工具进行可视化分析,透过视觉立体效果,探讨空间信息所反映的规律知识,将多个空间操作对像集中在一个模型中分层提取各个地类,充分利用几何形状和结构信息进行分析,降低了由“同物异谱”、“同谱异物”和混合像元引起的不确定性。更多还原

【Abstract】 Land-use survey in field consume time,energy and money efficiency is low. Today, with the development of remote sensing and geographic information system technology, making full use of these high-tech achievements to survey land resources has become possible for the people and by consensus.The uncertainty of remote sensing information will affect the remote sensing seriously, such as the function, the efficiency and the flexibility, which restricts the remote sensing information products and the further development.Taking the QuickBird image data of the Hongshui River valley Bute-you township of Ceheng County for an example, analyzing the uncertainty and the reasons in land-use information extraction, at the same time, in combination with other information, making remote sensing, geographical information systems as the main technical tools in the course of processing uncertainty information of images, which is more targeted and relative effectiveness, not only improve the quality of remote sensing image, enhance the GIS data quality obtained from the image data, but also reduce Image recognition errors.The main conclusions are as follows:(1) Making use of remote sensing image processing software, combined with topographic maps, GPS measurement points, DEM, and other relevant information to carry out geometric precision correction and ortho-corrected of the QuickBird images, which improved the data quality, reduced uncertainty of the image on the study area.(2) Using quantitative analysis method to analyse the remote sensing data calculating the standard deviation,entropy, space-related factors and spectral correlation coefficient of the band, found that the PCA analysis integration was the best way by contrast.The standard deviation, entropy and the correlation coefficient were higher than several other algorithms, after fusing, image in the visual, maintaining the spatial details information were better.In order to highlight vegetation, using the formula (float (b2)+float (b4))/2 to calculate ban information,then,the incomplete,uncertainties and the existence of more justice in the different environment were reduced.(3) Using remote sensing image processing software and MATLAB to process remote sensing images-noiseing coverd by little cloud. Using two methods to remove cloud, based on the R, G, B channels and HIS transformation, HIS-based transformation obtained better results, which made the bottom features clearly.The borders were more clear, the internal texture was more rich, facilitating the further accurate identification and interpretation.At the same time, making part of the image spectrum lost, cloud-free image in the background details were weakened. Based on the R, G, B-access to remove cloud, which improved the information-resolution of images, images background information was more abundant.(4)To Solute the uncertainty brought by the texture characteristics, using the geo-information for auxiliary processing. Features point, line were added to the TIN to generate high-precision digital elevation model. The river network information were generated by depression-free digital elevation model and the slope map were extracted by digital elevation model. The river network information, slope map and remote sensing images were superimposed analysis to determine the distinction between water erosion and the river,the steep slopes and the dryland. Useing the network river to determine the river and the water erosion, the accuracy rate was 100%. Useing slope map to distinguish between perennial dry and steep slopes, the accuracy of steep slopes was 92.1 percent, the accuracy of perennial dry was 94.2 percent.(5)Through the analysis of the spectral characteristics and mixed-pixel, found that the spatial distribution of the uncertainty are more on the broken terrain or a variety types of features at the junction. To reduce the uncertainty problem brought by "with the spectrum of differences", " Spectrum foreign body " or the mixed-pixel, using Model builder in ARCGIS for visual analysis, through three-dimensional visual, discussing the laws of knowledge reflected by spatial information. A number of space operations are concentrated in a hierarchical model to extract every category, make full use of geometry and structure information for analysis, reducing the uncertainty bring by "differences of the same spectrum", "foreign body in the same spectrum" and mixed-pixel.更多还原