基于国产高分遥感的人工种植红树林种间分类方法研究——以广西茅尾海红树林为例 |
Pixel-based and object-based comparative analysis of classification among mangrove species in Maoweihai |
|
DOI: |
中文关键词: 红树林 基于像素 面向对象 茅尾海 种间分类 |
英文关键词:mangrove pixel-based object-based Maoweihai Interspecific classification |
基金项目: |
|
摘要点击次数: 1136 |
全文下载次数: 707 |
中文摘要: |
红树林的种间结构组成对红树林生态系统的健康和发展至关重要,而红树林种间分类问题一直以来都是基于遥感手段的红树林监测中的难点。针对该问题,本文以人工种植为特点的广西茅尾海红树林遥感种间分类为例,基于面向对象的分类思想,提出了一种现场样本与分割对象相结合的红树林种间分类方法。利用GF-2 PMS1高分辨率卫星遥感影像数据,开展了广西茅尾海红树林湿地典型植被精细分类和空间分布研究,并将分类结果与基于像素和传统面向对象SVM分类方法进行了对比。结果显示:总体上,面向对象分类方法更适合用于茅尾海红树林湿地典型植被分类;对于局部混生明显的区域使用基于像素SVM分类方法效果会更好;传统面向对象分类方法中将整个影像分割对象单元作为训练样本 可能会在某种程度上造成负面影响,因此,使用本文提出的样本选择新方法进行面向对象分类精度最高,总体精度达到了93.13%,Kappa为0.89。 |
英文摘要: |
The structure of mangrove forests is of great importance to the health and development of mangrove ecosystems, and the classification of mangroves has long been a difficult part of the mangrove monitoring based on remote sensing. In view of the above problems, this paper takes artificial planting mangrove in Maoweihai of Guangxi as an example to conduct remote sensing inter-species classification. Based on the idea of object oriented classification, a mangrove interspecies classification method based on the combination of field samples and segmentation objects is proposed. Based on the high-resolution satellite remote sensing data of GF-2 PMS1, the fine classification and spatial distribution of typical vegetation in mangrove wetlands of Maoweihai in Guangxi were studied, and the classification results were compared with the classification methods pixel-based and traditional object-based SVM. The results show that the object-based classification method is more suitable for the classification of typical vegetation in mangrove wetland of Maoweihai. It is better to use the pixel-based SVM classification method for local mixed areas. Using the whole image object as training sample in the traditional object- based classification method may have a negative impact to some extent. Therefore, the new method of sample selection proposed in this paper has the highest accuracy in object-based classification, with an overall accuracy of 93.13% and a Kappa of 0.89. |
查看全文 查看/发表评论 下载PDF阅读器 |