利用对象和支持向量机的遥感信息提取方法探讨

土地利用/覆被专题信息的快速、高效、准确提取是遥感图像处理研究的重要方向。传统的遥感分类方法常依靠像元的光谱值,未充分利用影像的空间信息。本文将面向对象影像分割和支持向量机方法相结合,复合光谱和纹理信息,建立了Object-SVM分类模型,并与面向对象的模糊函数和基于像元的SVM方法相比较,探寻区域尺度土地利用/覆被信息提取方法。结果显示,Object-SVM模型有效地提高了遥感图像的分类精度和分类效率,对于区域尺度影像的快速、准确、客观的信息提取具有实际意义。

Remote sensing information extraction based on object-oriented and support vector machines

It is the important for remote sensing image processing to quickly,efficiently and accurately extract the land use/cover information.Traditional remote sensing classification makes use of the single-source spectral information to extract the land use/cover information without spatial information.In this paper,Object-oriented and Support Vector Machines are incorporated to extract the land use/cover information with spectral and texture information.Object-SVM model is set up for the extraction of land use/cover information in area coverage,comparing with Object-oriented fuzzy logic and SVM based on pixels.The research shows that Object-SVM can achieve better accuracies and efficiencies for actual applications.