基于水稻特征波段的决策树分类研究

针对种植结构复杂、地形复杂的水稻种植面积遥感提取精度不高现象,结合多时相遥感影像中反映水稻物候规律的特征波段,以南京江宁丘陵山区为例,选择典型水稻物候期时相的TM数据,基于多特征波段构建决策树分类提取水稻种植面积。结果表明:纹理、植被指数、湿度因子、坡度因子等多特征参与决策树分类能够提高总体精度;在具有两期物候数据时提取精度和效率较好,而加入了地形特征的水稻抽穗期数据比水稻灌浆期数据获取效果略好。因此,利用合理的作物物候期数据和该遥感影像的特征波段可有效提高分类精度,为地块破碎区作物种植面积提取提供有效手段。

Decision Tree Classification Based on Multi-temporal Characteristic Bands of Rice

In order to study the paddy distribution in hilly mountainous areas,Jiangning District is taken as an example to discuss the method of combining features of multi-temporal LANDSAT TM image to improve the accuracy of extracted information using CART-based decision tree model.By the rules of CART algorithm classification,vegetation index,textural,slope characteristics as test variables gets better effect in extracting paddy area.The experiment proves that the CART-based decision tree result using multi-temporal image can get higher accuracy compared with the same method using single image.Especially,the CART-based decision tree model using feature bands of rice heading TM image can also use to get the paddy distribution for further study.