基于See5.0算法的决策树构建与分类研究

A Study of Construction and Classification of Decision Tree Classifiers Based on SEE5.0 Algorithm

文中以地形较为复杂的济南市长清区为研究区域,综合TM影像的光谱特征、纹理特征与区域内的地形特征、植被特征信息作为样本集的属性值,结合研究区内6种主要地物类型训练数据集,基于See5.0的决策树分类法进行分类实验,并在ENVI中提取土地的利用现状信息。实验结果表明:综合利用不同的特征数据会有效的提高分类精度;See5.0和ENVI相结合可以高效、高精度地、客观地实现土地覆盖分类,是基于知识的遥感影像分类的有效的方法。

Combining TM spectrum, texture, terrain and NDVI as attributes of sample set to do class test using the decision tree algorithm based on See5.0,conjuncting with six major land cover types over Changqing District in Ji'nan, Shandong province ,and extracting the present situation information of land utilization in ENVI. The test show that the precision of classification, integrating different attributes data, has been improve obviously ,and the combination of See5.0 and ENVI is an effective method of remote sensing imagery classification based on knowledge to classify land cover types with high efficiency, high accuracy and reliability.