摘要
尝试提出一种面向对象的图像分类方法,该方法的关键技术是影像分割,图像处理最小单元是一个个通过分割得到的大小不一的多边形"同质"图斑对象,分类过程综合考虑了光谱、形状、纹理以及拓扑等图像特征。以海岸带区域的QuickBird数据为研究对象,进行影像融合、影像分割以及采用最近邻法分类等图像处理步骤,最后进行精度评价。结果表明:对于海岸带这类地物混杂度大、噪声严重的高分辨率遥感影像,采用面向对象的图像分类方法,能有效避免"椒盐现象",抗噪能力较强,分类结果便于矢量化形成专题图,分类精度较高。
This paper tries to provide a object--oriented image classification method whose key technology is Image Segmentation. The small units of Image Segmentation are those homogeneous image object polygons classifying by image features such as spectrum, geometrics, texture and topology. We took QuickBird data of Coastal zone as study objects,chose methods such as image fusion, image segmentation, the nearest neighbor classification and did accuracy assessment finally. The result shows that to the coastal zone images of dynamic land types and serious noises, using Object-oriented Image Classification Method can avoid the "pepper and salt", increase noises resistance, change classification into vector thematically map and improve the classification accuracy.
出处
《三峡环境与生态》
2009年第3期27-30,共4页
Environment and Ecology in the Three Gorges
关键词
高分辨率遥感影像
面向对象
影像分割
海岸带
high spatial resolution remote sensed image
object-oriented
image segmentation
coastal zone