基于特征组合的黄河源园区土地覆被分类

Classification of land cover in the park of the source region of the Yellow River based on the feature combination

为了解黄河源园区地物分布格局,本文以黄河源园区为研究对象,利用Landsat 8 OLI影像获取其光谱、植被、水体和纹理特征,采用Relief F算法对所有特征进行重要性排序,选出前8个特征作为优选特征;基于随机森林(RF)方法研究不同特征组合对分类结果的影响,为评估RF方法的效果,对优选特征组合,采用决策树、K近邻、感知机和支持向量机方法与RF方法进行对比;最后综合所有特征完成黄河源园区土地覆被分类。结果表明:(1)综合利用光谱、植被、水体和纹理特征,可以有效提高分类精度;(2)利用Relief F算法,可快速遴选出有效特征,大幅缩短模型运行时间;(3)相对于其他4种分类方法,随机森林的分类精度最高;(4)黄河源园区地物类型以中低覆盖度草地为主。

In order to understand the distribution pattern of surface features in the park of the source region of the Yellow River,this paper takes the park of the source of the Yellow River as the research object.Firstly,the spectrum,vegetation,water body and texture features are obtained by using Landsat 8 OLI image.Relief F algorithm is used to rank the importance of all features so as to select the first 8 features as optimal features;secondly,the influence of different feature combinations on the classification results is studied based on the RF method.In order to evaluate the effects of RF method,the optimal features are combined and compared with RF method by decision tree,K-nearest neighbor,perceptron and support vector machine.Finally,the classification of land cover in the park of the source region of the Yellow River is completed based on all the characteristics.The results show that:(1)The accuracy of the classification can be effectively improved by comprehensive utilization of spectrum,vegetation,water body and texture features;(2)Using Relief F algorithm can quickly select effective features so as to shorten the running time of the model;(3)RF method has the highest accuracy compared with the other four classification;(4)The surface features in the park of the source region of the Yellow River are mainly grassland with medium and low coverage.