基于专家知识的决策树分类

基于ENVI运用专家知识的决策树分类、监督分类、非监督分类手段来实现对于Landsat TM影像和快鸟影像数据的图像分类处理,简单介绍了决策树的建立过程,以及如何进行预处理、分类后处理、精度评价。通过3种分类方式的比较,基于专家知识的决策树分类法具有分类判别规则十分灵活、分类决策树看起来很直观、分类条件清晰、分类效果好、运算效率高等特点。实验中,其分类方式的缺陷表现为分类过多、过于复杂,可能会产生错误的速度加快;决策树判别规则复杂,树形分枝多导致用户难以识别、理解、应用。

Classification Based on Decision Tree Expert Knowledge

This paper is based on ENVI using expert knowledge of decision tree classification,and supervised classification and unsupervised classification means to achieve for image classification of Landsat TM images and quick bird image data processing,and briefly introduces the decision tree building process,and how to carry out the preprocessing and post classification processing,accuracy assessment.Through the comparison of three kinds of classification,decision tree classification based on expert knowledge with classification rules of discrimination is very flexible and classification decision tree looks very intuitive,clear classification conditions,to the effect of strong,operational efficiency higher characteristic.Experiment,the classification of defects for classification too much too complex situation,may be produced speed error;decision tree classification rules are complex,tree branches lead to users to identify,understand and apply.