中文摘要：

本文以北京昌平为研究区域，针对农作物的分类特点，结合ASAR的VV极化、新型PALSAR的HH、HV极化以及TM的多光谱数据进行细化分类。首先，使用MIMICS模型对该地区主要农作物玉米和果林的后向散射特性进行了模拟分析，并跟SAR实际观测数据进行对比。在充分认识农作物后向散射的机制和数值大小关系的基础上，构建一种基于BP神经网络和止态模糊分布函数的模糊神经网络模型，结合双频多极化SAR数据和多光谱数据进行农作物类型的识别。研究结果表明：双频多极化SAR数据能够提供有利于作物类型识别的信息，并产生重要的可分离性，其结合多光谱数据进行作物类型识别是一种有效的途径，具有较大的优势。

英文摘要：

This paper, taking Changping District of Beijing as a study area, conducts a detailed analysis on crop characteristics combining ASAR image, new PALSAR dual-frequency images and TM multi-spectral image. Firstly, MIMICS is used to simulate the backscattering characteristics for spring maize and orchard based on extracted baekscattering coefficient from images, and the modeling result is compared with the radar measurement result. Then, a classifier based on BP neural network and normal fuzzy distribution function was established utilizing redundancies and complementary of multisource remote sensing data sufficiently. This classifier used training samples generated by Gaussian fuzzy distribution function, taking the advantage of the study ability of fuzzy neural network to adjust membership functions and fuzzy rules, completed fuzzy inference, consequently made the system have adaptive characteristics. Results indicated that：the dual-frequency and multi-polarization radar data can afford more crucial information for crop classification, this kind of classifier is able to be used in crop recognition after training, and the classification precision increases by about 10% averagely, it is an effective crop recognition method.