基于C5.0决策树算法的西北干旱区土地覆盖分类研究——以甘肃省武威市为例

西北干旱区面积广阔,由于土地利用类型多样,成因复杂,对环境变化敏感、变化过程快、幅度大、景观差异明显等特点,在影像上表现出的“同物异谱”现象明显 |利用常规目视解译、监督非监督分类、人工参与的决策树分类等方法在效率或精度等方面各有其缺陷。采用机器学习C5.0决策树算法,综合利用地物波谱、NDVI、TC、纹理等信息,根据样本数据自动挖掘分类规则并对整个研究区进行地物分类。机器学习的决策树可以挖掘出更多的分类规则,C5.0算法对采样数据的分布没有要求,可以处理离散和连续数据,生成的规则易于理解,分类精度高,可以满足西北干旱区大面积的土地利用/覆被变化制图的需要。

The Study of the Northwest Arid Zone Land-Cover Classification Based on C5.0 Decision Tree Algorithm at Wuwei City,Gansu Province

In the broadly northwest arid regions,frequently,same object has different spectral characters because of the special characteristics of land cover change such as complex causes of formation,sensitivity to environment change,rapid and violent change and obvious differences in landscape. The conventional methods of classification including visual interpretation,supervised classification,unsupervised classification,and artificial decision tree classification have disadvantages in the efficiency or the accuracy. In this paper,machine learning algorithm based on C5. 0 decision tree was used to classify the entire study area automatically according to the sample data mining classification rules. Spectral features,NDVI,TC,texture and other informations were involved in the algorithm. More classification rules could be mined by machine learning decision tree. C5. 0 algorithm handling with both continuous and discrete data is independent of the distribution of sampling sites,The classification rules mined by this algorithm were interpretable. Other superiority of this algorithm included the fast speed of training and higher accuracy than many other classifiers. Thus,it is able to be used in the mapping of land use/cover change in a large scale in northwest arid regions.