遥感影像融合与分类在城市边缘带扩展监测中应用研究

探讨了TM30m分辨率波段与SPOT10m分辨率全色波段通过融合来提高城市扩展动态监测精度的方法和应用潜力。首先采用IHS变换完成TM的多光谱波段与SPOT全色波段融合，增强变化信息在光谱和几何特征上的表征，然后采用最大似然分类方法对融合图像进行分类。实验结果表明光谱与纹理特征组合在用户精度上比单纯光谱、纹理特征分类分别提高21．87％和10．22％；在生产者精度上比各自分别提高8．4％和17．88％；Kappa系数分别提高0．10和0．21。通过高几何分辨率图像与多光谱波段融合方法可以，增强变化信息，纹理特征参与变化信息提取可以提高变化类型的分类精度。

Study of Remotely Sensed Image Fusion and Classification Applied to Urban Fringe Expansion

This paper discusses the image fusion method of TM multi-spectral bands and SPOT panchromatic band to improve urban expansion monitoring accuracy and the potency applied tochange detection. Firstly IHS transform is applied to finish TM multi-spectral bands and SPOT panchromatic band fusion, then MLC method is used to classify the fusion imagery; Finally theability of differing spectrum, texture and their combination applied to urban expansion extractionis compared. The result indicated that combination improves 21.87% and 10. 22% in user accuracy; a 14. 26% and 14. 12% in producer accuracy compared to spectrum only and texture only. For the whole classification, their combination improves 8. 47% and 17. 88% in overall accuracy; 0. 10 and 0. 21 in Kappa Coefficient. It is concluded that the fusion of high spatial imagery and multi-spectral bands can enhance change information, and the fusion of texture character can improve the classification results.