...
首页> 外文期刊>Remote Sensing of Environment: An Interdisciplinary Journal >On the use of binary partition trees for the tree crown segmentation of tropical rainforest hyperspectral images
【24h】

On the use of binary partition trees for the tree crown segmentation of tropical rainforest hyperspectral images

机译:关于二叉分割树在热带雨林高光谱图像树冠分割中的应用

获取原文
获取原文并翻译 | 示例
           

摘要

The segmentation of remotely sensed images acquired over tropical forests is of great interest for numerous ecological applications, such as forest inventories or conservation and management of ecosystems, for which species classification techniques and estimation of the number of individuals are highly valuable inputs. In this paper, we propose a method for hyperspectral image segmentation, based on the binary partition tree (BPT) algorithm, and we apply it to two sites located in Hawaiian and Panamean tropical rainforests. Different strategies combining spatial and spectral dimensionality reduction are compared prior to the construction of the BPT. Various superpixel generation methods including watershed transformation and mean shift clustering are applied to decrease spatial dimensionality and provide an initial segmentation map. Principal component analysis is performed to reduce the spectral dimensionality and different combinations of principal components are compared. A non-parametric region model based on histograms, combined with the diffusion distance to merge regions, is used to build the BPT. An adapted pruning strategy based on the size discontinuity of the merging regions is proposed and compared with an already existing pruning strategy. Finally, a set of criteria to assess the quality of the tree segmentation is introduced. The proposed method correctly segmented up to 68% of the tree crowns and produced reasonable patterns of the segmented landscapes. (C) 2015 Elsevier Inc. All rights reserved.
机译:对于许多生态学应用,例如森林清单或生态系统的保护和管理,在热带森林上采集的遥感图像的分割引起了极大的兴趣,对于这些应用,物种分类技术和个体数量的估计是非常有价值的输入。在本文中,我们提出了一种基于二分分割树(BPT)算法的高光谱图像分割方法,并将其应用于位于夏威夷和帕纳米亚热带雨林的两个地点。在构建BPT之前,比较了将空间和频谱降维相结合的不同策略。包括分水岭变换和均值漂移聚类在内的各种超像素生成方法被应用于减小空间维数并提供初始分割图。执行主成分分析以降低光谱维数,并比较主成分的不同组合。基于直方图的非参数区域模型,结合扩散距离以合并区域,可用于构建BPT。提出了一种基于合并区域大小不连续的修剪策略,并将其与现有修剪策略进行了比较。最后,介绍了一组评估树分割质量的标准。所提出的方法正确地分割了多达68%的树冠并产生了合理的分割景观格局。 (C)2015 Elsevier Inc.保留所有权利。

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