图像分割与SVM的城市森林植被分类技术研究

Applying segmentation method and SVM technology on urban forest regetation classification research

为了实现城市森林植被种类的信息提取,文中采用面向对象结合支持向量机的分类方法,基于无人机影像数据对上海某校区内的城市森林进行了植被分类技术的研究。利用影像数据中各类地物的光谱、纹理等特征信息,将城市森林植被类别分为四类,并将分类结果与最大似然法分类结果进行对比分析。结果表明:该方法实现了高分辨率遥感影像的城市森林植被分类,不仅消除了分类过程中的"椒盐现象",而且有效提高了植被分类精度。最大似然法分类结果的分类精度为55.12%,面向对象结合支持向量机的分类精度达83.60%,提高了28.48%。实验结果满足精度要求,可为城市规划提供数据支持。

In order to classify urban vegetation types,the object-oriented combination of support vector machine( SVM) classification method was applied to one campus urban forest in Shanghai based on UAV image data. Using the spectral and texture features of the image,the urban forest vegetation in the study area was classified into five categories,which were compared to the results of the traditional maximum likelihood method. The results show this method realizes high resolution remote sensing image classification of urban forest vegetation. This method not only solves the 'salt-and-pepper noise'in the classification processing,but also improves the accuracy of vegetation classification. The classification accuracy of maximum likelihood is 55. 12%,while the object-oriented combined with support vector machine( SVM) is 83. 60%,increased by 28. 48 percentage points. The experimental results meet the precision requirement,and can provide data support for urban planning.