根据Kohonen网的生物学基础 ,基本结构和学习算法 ,提出了解决遥感影像分类的途径。依据实验区土地利用类别的光谱特征 ,采用主成分分析对遥感影像进行预处理 ,结合地理辅助数据的量化输入训练出Kohonen自组织图后对融合有地理辅助数据的影像进行土地利用分类 ,并与BP网和最大似然法分类结果进行分析比较。结果表明 ,地理辅助数据的参与对提高Kohonen网影像分类精度具有意义

According to the biological basis, basic structure and learning algorithms of Kohonen network, an image classification method is introduced.Pre-processing the image with principal component analysis method based on spectral characteristics of the land use types in the experimental area, and training Kohonen self-organization mapping with geographical ancillary data, land use classifications of Kohonen network are made by integrating image with geographic ancillary data.The classification results are analyzed and compared with the results obtained by Back-Propagation neural network and Maximum Likelihood.The result shows that the classification with geographic ancillary data can improve the image classification accuracy of Kohonen network.