论文部分内容阅读
选取中国鄱阳湖南矶山湿地为研究范围,通过分析候鸟在越冬期的栖息特征,构建适宜于提取候鸟栖息地的湿地分类系统。以Landsat8卫星OLI遥感影像为数据源,采用面向对象分类方法,通过多尺度分割、特征提取和决策树建立等关键步骤,实现湿地信息的快速提取;通过与传统像元法的分类结果对比,系统分析了面向对象方法在基于中低分辨率遥感影像的湿地信息提取中的有效性。研究表明:在面向对象湿地信息提取中,构建不同等级的分割尺度,在多个尺度上提取同一地物,可以更好识别复杂的湿地景观类型;相比仅依据像元光谱特征进行分类的传统像元方法,面向对象方法综合利用光谱、空间、形状和纹理等多种特征,因此可以获得更高的精度(总体分类精度达到87.64%,Kappa系数为0.855 2);基于中低分辨率遥感影像,采用面向对象分类方法,能够获得较高精度的湿地景观分布,是一种成本较低且行之有效的技术手段。
By selecting the Nanjishan wetland of Poyang Lake in China as the research area, by analyzing the habitat characteristics of migratory birds during wintering period, a wetland classification system suitable for the extraction of migratory bird habitats was constructed. Taking Landsat8 satellite OLI remote sensing image as data source, the object-oriented classification method is adopted to realize the rapid extraction of wetland information through multi-scale segmentation, feature extraction and decision tree establishment. Compared with the classification results of traditional pixel method, The effectiveness of object-oriented method in wetland information extraction based on medium and low resolution remote sensing images is analyzed. The research shows that in the process of object-oriented wetland information extraction, constructing different levels of segmentation scales and extracting the same features at multiple scales can better identify complex wetland landscape types. Compared with the traditional classification based on spectral features of pixels The pixel-based method and the object-oriented method can comprehensively utilize various features such as spectrum, space, shape and texture, thus obtaining higher accuracy (overall classification accuracy of 87.64% and Kappa coefficient of 0.855 2). Based on the medium and low resolution remote sensing images , Adopting the object-oriented classification method to obtain the wetland landscape distribution with higher accuracy is a low-cost and effective technical means.