基于光谱与纹理特征的高分二号竹林信息提取——以庐山自然保护区为例

Extraction Bamboo Forest from GF-2 Satellite Based on Spectral and Texture Features:Taking Lushan Nature Reserve as an Example

研究以庐山自然保护区为研究区域,进行基于高分二号影像光谱特征与纹理特征的竹林信息提取研究。为了提高竹林的识别精度,依据竹林的纹理特征与周围植被的差异性,本文开展引入纹理特征的分类研究,选取两种不同的实验方案(光谱特征、光谱特征+纹理特征),采取马氏距离、最大似然法,支持向量机等分类方法进行比较。结果表明:1)引入纹理特征信息的分类精度比纯光谱特征分类精度更高;2)基于光谱特征+纹理特征的支持向量机的分类方法精度最高,竹林分类的生产精度和用户精度均达到90%以上。

Taking Lushan mountain nature conservation as the research area,the bamboo forest information extraction based on the spectral features and texture features of GF-2 image was studied.In order to improve the identification precision of the bamboo forest,according to the texture characteristics of bamboo and the surrounding vegetation diversity,this paper study on the introduction of texture feature classification,selecting two different experimental program(spectrum,spectral characteristics+texture feature),adopting Markov distance,maximum likelihood method,support vector machine(SVM)classification methods such as comparison.The results show that:(1)the classification accuracy of texture feature information introduced is higher than that of pure spectral feature classification;(2)the classification method based on spectral feature+texture feature has the highest accuracy,and the production accuracy and user accuracy of bamboo forest classification both reach over 90%.