基于随机森林与J48决策树的面向对象方法的土地覆被分类

21世纪以来,面向对象的影像分析方法快速发展,在高分辨率影像中的应用日益增加,同时也体现了其优势。目前,基于机器学习算法的分类方法开始普遍运用,不再局限于基于规则集的面向对象影像分类方法,这些算法相对于基于规则集的分类算法,精度有所提高,文章基于随机森林与J48决策树机器学习算法,利用WorldView2影像,进行了中卫市西南部分地区的土地覆被分类,并得到了显示效果较好的土地覆被分类图。文章的结果表明,面向对象的机器学习算法精度较高,并能适用于各种复杂的研究区,其中,针对文章研究区影像,基于随机森林的分类方法精度更高。

Since the 21st century,object-oriented image analysis methods have developed rapidly,and their applications in high-resolution images have increased day by day,which also reflects their advantages.At present,classification methods based on machine learning algorithms have begun to be widely used,and are no longer limited to object-oriented image classification methods based on rule sets.Compared with the classification algorithms based on rule sets,these algorithms have improved accuracy.This article is based on random forest and J48 decision making The tree machine learning algorithm used WorldView2 images to classify land cover in parts of the southwestern part of Zhongwei City,and obtained a land cover classification map with a better display effect.The results of this paper show that the object-oriented machine learning algorithm has high accuracy and can be applied to various complex research areas.Among them,for the images of the study area in this paper,the classification method based on random forest has higher accuracy.