本文采用在模糊监督分类中增加迭代过程的方法──模糊监督分类-迭代法, 在分解混合像元的基础上, 利用AVHRR数据求算水稻种植面积。根据稻田与旱地存在温度差异的特点, 在分析AVHRR数据统计特征的基础上, 增加了第3和4两通道参加模糊监督分类, 从而增加了分类像元值矢量维数, 增强了对水稻的鉴别能力。所采用的迭代法是收敛且有效的, 经本方法输出的水稻种植面积百分含量图与实际水稻分布十分吻合, 测算出的面积值与TM和实地调查结果相比, 有较高的精度。

Fuzzy supervised classification-iteration method was used in evaluating the rice planting area on a mixed pixel signal unmaxing basis. Channels 3 and 4 of AVHRR were used in this study according to the fact that the temperature of rice fields different from that of rainfalled-crop fields and based on the analysis of statistical characteristic. The additional channels in middle infrared and thermal infrared allowed rice to be identified easier than using the visible and near-infrared channels alone, and added the dimensional number of the pixel value vector in the characteristic. The iterative process is effective and convergen. The pixel proportion images of rice planting area produced in this study have a good consistence with the ground truth Compared with rice area abtained form TM image and field survey, this method has a high accuracy.