基于决策树模型的海岸带分类方法研究

海岸带地物分布复杂,地物混淆常造成海岸带提取困难。该文以江苏省粉砂淤泥质海岸为研究对象,运用图像光谱特征、纹理特征并引入地学知识,构建研究区遥感图像分类决策树模型,并利用ETM 图像进行海岸地物分类研究。结果表明:采用的决策树模型可以较好地结合纹理信息和地学知识,解决遥感图像中复杂地物分类过程中的混淆现象,分类精度达89.26%,比最大似然法分类精度提高了15.19%。

Land Type Classification of Coastal Zone Based on Decision Tree Model

Coastal zone is characterized by the mixing of various land types.Classification confusion is a serious problem when processing the coastal image.In the present study,the ETM image is used to classify the coast zone land types for the mud-silt coast zone of Jiangsu Province.Decision tree model was developed based on spectral characteristics,texture characteristics and geo-knowledge for the study area.It is shown that,by building a classification decision tree with well combination of texture features and geo-knowledge for the study area,the problem of classification confusion for some land types can be well solved.Compared with conventional maximum likelihood method,the decision-tree method increases the classification accuracy by 15.19%,achieving an overall accuracy of 89.26%.