基于粗糙集规则提取的面向对象树种分类方法

摘要

树种分类是林业资源监测中的核心问题,而面向对象的树种分类是目前研究的重点,在面向对象分类方法中,难点在于规则集的建立.本文针对面向对象树种分类问题,以福建将乐国有林场为研究区域,ALOS影像为遥感数据源,探讨面向对象分类规则集的建立并进行树种分类.首先,选择最优的分割尺度,并对影像进行多尺度分割;然后,基于粗糙集进行对象特征的属性约简,从对象的34个特征中约简出最能代表树种分类的13种参数,包括波段比值、亮度值、平均灰度值等;最后建立分类规则集.应用该规则集,对实验区影像进行分类,分类精度达80.4509％,效果较好.实验证明本文所提出的方法可高效地利用遥感图像的信息,提高分类精度,为林业资源调查和植被监测提供了有效的辅助手段.%Species classification is a core issue in forestry resource monitoring,and the difficulty of object-oriented classification is the establishment of the rule set.In this paper,the establishment of the rule set of object-oriented classification and species classification based on ALOS image was studied.In the first place,by selecting the optimal segmentation scale,the image was performed multi-scale segmentation.Afterwards,it reduced the characteristics of the object attributes based on rough set,and extracted 13 kinds of parameters which was the most representative of the species classification from 34 characteristics of objects,including band ratio,brightness value,average gray value and so on.Finally,it established a classification rule set.The experimental image was carried out through this classification rule set,and the classification accuracy achieved 80.4509 ％,which made a better classification effect.The experimental results proved that object-oriented rules extraction method of trees can make full effective use of the information from the remote sensing image and improve the classification accuracy,which can provide an effective means for the investigation and monitoring of forest resources.