中文摘要：

利用水稻生育前期和后期两个不同时期TM影像分别进行穗帽变换生成亮度、绿度和湿度变量，并将它们合成为多时相影像，充分利用这3个具有物理意义的变量，特别是湿度变量进行水稻种植区影像分类和以水为背景的水稻面积提取，并使用亚米级GPS地面调查的数据进行分类验证。基于穗帽变换影像的分类方法有效提高了水稻面积提取精度，水稻分类的生产者精度和用户精度分别为84．30％和85．18％，这比原始合成影像数据的分类结果提高了约3个百分点；另外，总精度也由原始合成影像的74．12％提高到经过穗帽变换的78．04％。

英文摘要：

Rice cultivated area extraction is the premise of rice growth status monitoring and production estimation by remote sensing. One of the problems is the misclassification between rice and other green vegetations. In this study, two the-matic mapper （TM） images from early and late developmental stages were converted into brightness, greenness, and wetness variables using the tasseled cap transformation. And then these variables, which are associated with the important physical parameters of the land surface, were composed as multitemporal image and were fully applied to extract rice-cultivated area. The classified map was compared with the data from GPS in less than 1 meter resolution for classification validation. The results indicated that the method based on the tasseled cap transformation can extract the rice-cultivated area with relative high precision. The producers＇ accuracy and users＇ accuracy were 84.30% and 85.18% for rice, respectively, which were increased by about 3 percent point compared with the classified original images. The overall accuracy was also increased from 74.12% for the classified original image to 78.04% for the classified tasseled cap transformed image.