摘要
森林类型识别技术是遥感分类中的重点和难点,采用面向对象的遥感影像分类方法是实现森林类型分类的新方法。资源3号遥感影像可为森林类型提取提供新方向。以资源3号遥感影像作为基础研究数据,采用面向对象的分类方法,选择分形网络演化法进行多尺度分层分割,并结合典型地物的光谱特征、纹理特征、几何特征以及植被指数,构建了适用于森林类型提取的决策树模型,并与分割尺度不同的支持向量机分类方法进行比较分析。结果表明:多层分割的决策树分类方法分类精度高于单层分割的支持向量机分类方法,分类精度分别提高了6.1%和12.5%。说明建立多层分割的决策树分类方法适用于森林类型的分类研究。
A new method of identification technology for forest types, an important and difficult part of remote sensing classification, uses object-oriented remote sensing image classification.It provides a new direction for forest type to extract which is based on ZY-3 remote sensing data. This study applied ZY-3 remote sensing data to the object-oriented classification method, chose hierarchical segmentation of a fractal network as an evolution method, and combined typical ground objects including spectrum features, texture features, geometrical characteristics, and vegetation indexes, to build a decision tree model which is applicable to forest types. Then, the different segmentation scale compared from the support vector machine(SVM) classification method. Results showed that classification accuracy of the decision tree classification method with multi-level segmentation(which increased 6.1% and 12.5%) was higher than the support vector machine(SVM) classification method with the different single segmentations. Thus, it would be suitable to build a decision tree classification with multi-level segmentation to the classification of forest type.
出处
《浙江农林大学学报》
CAS
CSCD
北大核心
2016年第5期816-825,共10页
Journal of Zhejiang A&F University
基金
国家高技术研究发展计划("863"计划)资助项目(2012AA102001)
关键词
森林测计学
面向对象
分型网络演化算法
信息提取
决策树模型
forest mensuration
object-oriented
fractal net evolution approach
information extractioni
decision tree model