• 主题
高级搜索  >
历史搜索 清空历史
首页> 外文会议>Asian Conference on Remote Sensing >A HIERARCHICAL MULTISCALE OBJECTED-ORIENTED CLASSIFICATION METHOD FOR GF-2 WALNUT FOREST EXTRACTION
【24h】

A HIERARCHICAL MULTISCALE OBJECTED-ORIENTED CLASSIFICATION METHOD FOR GF-2 WALNUT FOREST EXTRACTION

机译：GF-2核桃林提取的分层多尺度面向对象的分类方法

获取原文
页面导航
  • 摘要
  • 著录项
  • 相似文献
  • 相关主题

摘要

Walnut forest is one of the most important economic forest in south area of Shannxi Province. It is a combination of ecological forestry and people's livelihood forestry. The remote sensing monitoring of walnut forest(WF) acreage is very important for the develop of forestry industry. However, walnut trees' spectral is very similar with other green vegetation, and their canopy is very similar with apple trees. WF in the study area is block distribution not connected distribution, which resulted in less pure walnut forest pixels in Landsat like image. In order to get a high accuracy classification map for WF. high resolution remote sensing image(HRRSI) is priority. But complexity objects and large data volume become another problem in HRRSI. so the extraction of WF is a relatively difficult job. In this paper, a GF-2 HRRSI was used in this work, which cover the south part of Linwei District. The contribution of this work is that a hierarchical multiscale objected-oriented decision tree classification scheme was proposed for GF-2 WF classification. Firstly, fractal net evolution approach-FNEA in eCognition was used for GF-2 multiscale segmentation, and four scale different objects were produced; secondly, rule based decision tree was applied for coarse level classification, for example, water body, green vegetation, rural area, road, and bare soil; lastly, green vegetation was fine classification by CART classifier. WF is one of the fine classification class type. The overall classification accuracy is 87.98%, and the WF omission error is 11.32%. The method adapted different decision tree model for different classification level, it got high accuracy result effectively. It is expected that other type economic forest mapping may also use this method.
机译：核桃森林是陕西省南部最重要的经济森林之一。它是生态林业和人民生计林业的结合。核桃林（WF）面积的遥感监测对林业行业的发展非常重要。然而，核桃树的光谱与其他绿色植被非常相似，它们的树冠与苹果树非常相似。研究区域中的WF是块分布未连接分布，导致Landsat中的纯核桃森林像素较少。为了获得WF的高精度分类图。高分辨率遥感图像（HRRSI）是优先级。但复杂性对象和大数据量成为HRRSI的另一个问题。因此，WF的提取是相对困难的工作。在本文中，在这项工作中使用了GF-2 HRRSI，涵盖了临威地区的南部。这项工作的贡献是提出了一种分层多尺度对象的决策树分类方案，用于GF-2 WF分类。首先，evognition中的分形净进化方法-FNEA用于GF-2多尺度分割，并产生四种不同的物体;其次，施加规则的决策树，用于粗级分类，例如水体，绿色植被，农村，道路和裸土;最后，绿色植被是购物车分类器的精细分类。 WF是精细分类类类型之一。整体分类准确性为87.98％，而WF遗漏误差为11.32％。该方法适用于不同分类级别的不同决策树模型，有效地获得了高精度。预计其他经济森林映射也可能使用此方法。

著录项

相似文献

  • 外文文献
  • 中文文献
  • 专利
获取原文

客服邮箱：kefu@zhangqiaokeyan.com

京公网安备：11010802029741号 ICP备案号：京ICP备15016152号-6 六维联合信息科技 (北京) 有限公司©版权所有

  • 客服微信

  • 服务号