基于纹理的影像分类的特征、波段与窗口选择问题研究

Research on Selection of Features,Bands and Windows in Classification Based on Texture Feature

结合纹理特征进行分类是提高影像分类精度的重要手段之一。纹理分析结果的好坏直接影响到最终的影像分类精度,而纹理分析结果又与很多因素有关。本文选取武汉某地区的局部多光谱遥感影像,分别在不同纹理特征组合、不同波段组合以及不同分析窗口下利用灰度共生矩阵进行纹理分析,然后采用最大似然法进行影像分类。通过精度分析实验选取出最优纹理特征组合、波段组合以及纹理分析窗口。

Using texture feature in classification is an important mean for improving the accuracy of remote sensing image classification.The results of texture analysis have direct influence on final classification accuracy,and texture analysis result is in connection with a lot of factors.This paper selected a part of multi-spectral image of Wuhan and analyzed the image texture based on gray-level co-occurrence matrix in different combination of texture features,different combination of bands and different window sizes.Then classed this image in the way of maximum likelihood.Finally,selected the best combination of texture features,optimal bands combination and appropriate window size from the results of precision analysis.