一种基于ELM的土地覆盖遥感影像分类方法研究

An Approach Based on ELM for Land Cover Remote Sensing Image Classification

数据分类是将遥感数据转化为专题图信息的一个关键步骤.找到一个好的分类方法,提高数据处理的准确性是高度挑战性问题.提出一种极限学习机和支持向量机相融合的遥感图像分类模式,选ELM为基础分类器,以SVM来修正改善分类效率.仿真实验结果表明,该算法不仅具有较高的分类精度,而且消除了一些训练样本标签对分类的负面影响.结合PSM图像,与SVM、ANN(Artificial Neural Network)方法进行对比分析,表明该方法的鲁棒性.

Data classification is a key step of the remote sensing data into thematic map information. Find a good classification method and improve the accuracy of data processing are highly challenging problems. In this pa- per, we propose a fusion classification model based on Extreme Learning Machine （ELM） and Support Vector Ma- chine （SVM） for remote sensing image classification. Choose ELM based classifier and correction to improve the classification efficiency by SVM. Simulation experiment results show that the algorithm not only has higher classifi- cation accuracy, and eliminate some of the training sample tag on the negative impact for classification. Combining PSM images, comparative analysis with SVM and Artificial Neural Network （ANN）, show the robustness of the meth- od.