基于纹理特征的Spot绿地信息提取研究

城市绿地信息是城市园林(生态、城市)规划的重要信息,地物的几何结构和纹理信息更加明显的高分辨影像为获取更详细的绿地信息提供了可能。本文采用Spot卫星影像,以福建省泉州市鲤城区为例,借助Matlab和ERDAS图像处理软件,在详细分析研究区绿地不同类别纹理差异的基础上,利用统计和小波分析的方法提取绿地纹理特征,并将其引入基于光谱分类的最大似然法进行绿地信息的提取分类,相比直接用面向象元的监督分类的分类结果,其分类总精度由原来的76.5%提高到了80.2%,Kappa系数由原来的0.7023,提高到了0.7484,而且绿地类别被更详细地分成了一般绿地,农田,自然林地。不同绿地类别的分类精度明显提高,表明纹理特征在遥感绿地信息提取方面的有效性和本研究方法的可行性。本文是根据研究区特点,引入纹理特征提高绿地信息提取分类精度方法的探讨,更广泛的应用和精度的提高有待于进一步的完善研究。

Deriving Green Land from Spot Image Base on Texture Analysis

The green land information is important and absolutely necessarily in city(ecology) planning.High resolution image,which can display obviously the geometric structure and texture of green area,has provide the possibility for deriving green land information in detail.In this paper,taking the Lichen distract of Quanzhou city,Fujian province as a case,we present a methodology of derive green land texture character based on spatial variogram and wavelet analysis,and the classification result obtained from pixel-by-pixel classifiers simultaneously taking into account both radiometric and texture information.Comparing with the result from supervised classification,the total classification accuracy for the improved method has rised to 79.2 % from 77.5 %.Kappa change from 0.7023 to 0.7484.Furthermore,the green land have been classified into farmland,natural forest and grass detailed since there are two types green land when using supervised classification only.The calculation of the variogram and wavelet analysis,image process was realized by Matlab and ERDAS.The increase of classification accuracy of different kinds green land indicate the texture analysis is useful and the methods mentioned is feasible for deriving of green land information.The further improvement of classification accuracy and widely application depend on the more researches because this study mainly focused on the specific feature of green land in the study area.