一种高分辨率遥感图像单木树冠信息提取方法

A Method for Individual Tree-crown Extraction from High Spatial Resolution Imagery

单木树冠信息是森林管理和相关科学研究的基础数据。为解决面向对象方法在树冠相互连接时无法有效提取单木树冠信息的问题,提出了一种基于面向对象技术和水文分析的高分辨率遥感图像单木树冠信息提取方法。方法首先对图像进行预处理,然后利用面向对象方法从融合图像获取树冠的分布范围,利用水文分析技术从全色波段图像获取单木树冠的潜在分布范围,最后将提取的数据进行叠加相交处理,完成单木位置信息提取和单木树冠描绘。实验结果表明,所提出的方法能有效提取单木位置和描绘单木树冠;采用修改的遥感分类精度评价指标进行的精度评价结果为,单木探测总体精度为87.63%,单木树冠描绘总体精度为84.89%。

The information of individual tree-crown is a basic data for forest management and relevant scientific research.In the past few years,some techniques,such as the object-oriented method,were used to extract the crown information of individual trees.However,this problem remains unsolved due to the complexity caused by the connected crowns.In this paper,a new method,which integrates the object-oriented technique and hydrological analysis functions,was used to deal with this problem.This method is a procedure that requires four steps：firstly,image preprocessing was done;secondly,the spatial distribution of the crowns was extracted from a fusion image using the object-oriented method;thirdly,the potential distribution area for individual tree crown was obtained from a panchromatic image based on hydrological analysis functions;finally,the locations of individual trees were detected and individual crowns were delineated by overlaying the two layers mentioned above.The results showed that the proposed method is effective for detection of individual trees and delineation of the individual tree crowns.A promising agreement between the detected results and manual delineation results was achieved in counting the number of trees and individual tree-crown effective delineation using modified classification accuracy assessment.The overall accuracy for detection of individual trees is 87.63%,and the one for delineation of individual crowns is 84.89%.