基于数据融合的多特征遥感图像分类

以多光谱图像为研究对象，综合利用遥感图像的光谱、纹理和数学变换特征，提出了一种基于数据融合的多特征遥感地物分类方法。该方法针对不同的特征分别构造了神经网络分类器和K-均值聚类器，并对前者利用Adaboost算法进行提升，然后再将各特征的分类结果利用证据理论合成公式融合得到最终结果。实验结果表明，该方法的分类效果要优于单特征的分类结果。

Multi-Feature Remote Sensing Image Classification Based on Data Fusion

Remote sensing image classification is a key application of pattern recognition in the remote sensing field. Multi-spectral images are studies under the situation in which exact training data are absent. A method for multi-feature remote sensing image classification based on data fusion is proposed. Classifiers by using ANN and K-Means technologies are constructed according to different features with ANN boosted by the Adaboost algorithm. The final map is an integration results with different features. Experiments show that the fusion classification is superior to the result of any classifier with a single feature.