摘要
在基于像素的高光谱影像分类方法的基础上,结合面向对象图像分析理论与方法,提出面向对象的高光谱遥感影像分类方法,并具体分析探讨了面向对象高光谱遥感影像分类的关键技术,包括多尺度分割、最优波段选择、人机交互和知识库的建立等。试验表明,面向对象的分类方法应用于高光谱影像较传统分类方法有较高的精度,有很大的应用潜力。
The object-oriented image classification for hyperspectral remote sensing is proposed based on the conventional pixelbased processing approach and novel object-oriented image processing and analysis methods. Some key techniques inferred in the object-oriented image classification for hyperspectral remote sensing are analyzed, including multi-scale segmentation, optimal wavebands selection, man-machine interactive system and establishment of knowledge base. By experiment, it is showed that the object-oriented image classification for hyperspectral remote sensing can get higher accuracy than conventional algorithms and has more potential applications.
出处
《遥感信息》
CSCD
2007年第4期29-32,I0003,共5页
Remote Sensing Information
基金
国家自然科学基金(40401038)
地理空间信息工程国家测绘局重点实验室开放基金
中国矿业大学科学基金(D200403)资助项目
关键词
高光谱遥感
面向对象的图像分析
分类
图像分割
尺度
hyperspectral
object-oriented image analysis
classification
image segmentation
scale
