【摘要】 本文采用随机森林分类方法提取MODIS影像中的水体,根据水体和非水体在不同波段的反射率特征差异计算水体指数,选择一年内水体指数总和大于零的点构造分类特征,以全球30 m地表覆盖数据作为真值进行训练和验证.依据在随机森林中分类特征的重要性选出了10个分类特征,并通过一定量的实验统计选出有较好分类结果的随机森林模型参数.采用混淆矩阵及相关精度指标、Kappa系数等进行精度评价,获得较好的水体分类结果.更多还原

【Abstract】 In this paper,the random forest classification method was used to extract the water body from MODIS images. Firstly,the water index was calculated according to the difference of reflectance characteristics of water and non-water in different bands and the characteristics were constructed with the pixel values greater than zero. Then,the 30 meters spatial resolution land cover products were selected as training dataset and validation dataset. In the processing, ten classification characteristics were selected according to the importance of classification characteristics in random forest algorithm, and random forest model parameters with better classification results were selected by a certain amount of experimental statistics. Finally, the accuracy was evaluated by using the confusion matrix,correlation precision and kappa coefficient. The results showed that the method above achieved good classification results.更多还原