【作者】 张莉；

【导师】 周清波；

【作者基本信息】 中国农业科学院 ， 环境工程， 2012， 硕士

【摘要】 水稻是世界也是中国主要的粮食作物之一。获取大范围的水稻种植空间分布、面积和产量信息对指导水稻生产、合理分配水资源、以及检测大气环境变化等均具有重要的意义。由于水稻生产具有覆盖面积大、季节性强、单位面积的经济效益低等特点,通过地面调查方法获取每年的农作物种植信息,不论在技术性还是经济性方面都比较困难,而利用遥感技术则是解决这个问题的可行且有效的方法。与常规的统计方法相比,应用遥感方法获取作物种植信息具有他独特的优势,由于遥感具有覆盖面大、短时间内可重复观测、以及成本相对较低等特点,并与地理信息系统GIS和全球定位系统GPS相结合,不仅可以提取农作物种植面积,并且可以实现空间布局的准确定位。本论文以湖南省为研究区域,将遥感技术RS和地理信息系统GIS技术相结合,利用EOS/MODIS数据空间覆盖面大和时间分辨率高的优势,选取湖南省范围的数据,实现对湖南省晚稻种植和生长信息的提取。研究目标是解决大范围快速监测晚稻种植情况,并探索结合中高空间分辨率遥感影像来解决低分辨率遥感影像分类的混合像元问题。研究的主要内容包括：基于SPOT4遥感影像的晚稻识别及网格化处理、基于MODIS遥感数据的晚稻面积识别和晚稻识别精度评价。具体如下首先,根据湖南省的晚稻物候期选取适当时相的SPOT4影像,对数据进行几何校正、非监督分类、人工目视解泽等处理,得到湖南省晚稻的分布。同时以此影像为基础建立500米×500米的网格,并与MODIS遥感数据的像元完全重合,分别计算各个网格内晚稻面积占网格面积的百分比,并根据数值的不同进行分级,将分级后的网格抽取一部分作为训练样本数据,其余作为验证数据。其次,构建MODIS遥感数据的植被指数序列数据集,其中植被指数包括EVI和1LSWI,并对植被指数进行去噪处理。然后对相应时相的MODIS波段影像、植被指数进行主成分分析并生成新的数据集,之后利用训练样本对此数据集进行监督分类,得出湖南省晚稻面积及其分布。最后,利用验证数据、统计数据等对湖南省晚稻的面积及空间分布进行精度验证,并得出结论：此方法对于在水稻生育期的早期进行大范围水稻监测精度总体上是可信的,同时对解决利用中低空间分辨率的遥感影像进行作物分类研究时不可避免的混合像元问题进行了有益的探索。更多还原

【Abstract】 Paddy rice is one of the main food crops in the world and also in China. Gaining the information of large-scale rice planting spatial distribution, area, and output is significant for guiding rice production, reasonable allocation of water resources, and detection of the atmospheric environment change, etc. Since the rice production features large coverage area, strong seasonality, and low economic benefit of unit area, it is technically and economically difficult to gain annual crop planting information by the ground investigation method. However, the remote sensing technology is a feasible and effective method in solving this problem. Compared with conventional statistical methods, the remote sensing method in gaining the crop planting information has its unique advantages. With the features of large coverage area, repeated observations within a short time, and relatively low cost, the remote sensing technology, combining geographic information system (GIS) and global positioning system, may extract crop planting area and realize the accurate positioning of spatial distribution.This paper takes Hunan Province as the research area. It combines the remote sensing technology and GIS technology, and selects the data within Hunan Province to realize the information extraction on late rice planting and growth in Hunan Province by the large coverage area and high time resolution of EOS/MODIS data. The research objective is to solve the problem of extensively and rapidly monitoring the late rice planting situation, and explore the solution of the mixed pixel of low resolution remote sensing image classification combining remote sensing images of middle and high spatial resolutions. The main contents in the research are:late rice identification and grid treatment based on SPOT4remote sensing image, and late rice area identification and late rice identification precision evaluation based on MODIS remote sensing data. The detail is shown as follows:First, select a SPOT4image in proper time phases to conduct geometry correction, unsupervised classification, human visual interpretation, and other treatments in order to get the late rice distribution of Hunan Province according to the late rice phenological period of Hunan Province. Meanwhile, establish a grid of500mx500m based on this image, which completely coincides with the pixel of MODIS remote sensing data to respectively calculate the percentage that late rice area occupies the whole grid area in every grid, classify them according to value differences, and then take some pixels after the classification as the training sample data and the rest as the validating data.Second, construct the vegetation index sequence data set of MODIS remote sensing data, and the vegetation index includes EVI and LSWI, and denoising treatment on the vegetation index. Then conduct principal component analysis on MOISD band images and the vegetation index of corresponding time phases and generate a new data set, and the conduct supervised classification on this data set with the training sample to get the late rice area and distribution of Hunan Province.Finally, conduct the precision validation on the late rice area and spatial distribution of Hunan Province by using the validation data and statistical data, etc. and come to the conclusion:as a whole, this method is reliable in conducting the large-scale rice monitoring precision in the rice growth stage; in the meantime, conduct beneficial exploration on solving the problem of inevitable mixed pixels in crop classification research by using the remote sensing image of low and middle spatial resolution.更多还原