论文部分内容阅读
森林植被优势树种类型信息的提取是遥感影像分类中的难点.面向对象分类方法是用高空间分辨率遥感数据实现精确类型信息提取的新方法.本文以2013年Quickbird影像作为基础数据,选择福建省三明市将乐林场为研究区,采用面向对象多尺度分割方法提取耕地、灌草地、未成林造林地、马尾松、杉木和阔叶树等类型信息.分类特征融合植被的光谱、纹理和多种植被指数3类特征信息,建立类层次结构,对不同层次分别用隶属度函数和决策树分类规则,最终完成分类,并与只用纹理与光谱特征相结合的方法进行对比.结果表明:融合纹理、光谱、多种植被指数的面向对象的分类方法提取研究区优势树种类型信息的精度为91.3%,比只用纹理和光谱的方法精度提高了5.7%.
The extraction of dominant tree species information in forest vegetation is a difficult point in remote sensing image classification.The object-oriented classification method is a new method to extract accurate type information with high spatial resolution remote sensing data.Taking the Quickbird image in 2013 as the basic data, Sanming City, Le Lin field as the research area, using object-oriented multi-scale segmentation method to extract cultivated land, shrub and grassland, non-forestation afforestation, masson pine, fir and broadleaved tree and other types of information. Classification features Fusion vegetation spectrum, texture and multiple vegetation index 3 types of feature information, the establishment of the class hierarchy, the different levels of membership functions and decision tree classification rules, the final classification, and with only the combination of texture and spectral characteristics of the method for comparison.Results show that: fusion texture, spectrum , The object-oriented classification method with multiple vegetation indices extracted the dominance tree species type information in the study area with an accuracy of 91.3%, which was 5.7% higher than the method using only the texture and the spectrum.