基于QuickBird影像的树冠提取研究

Research on Tree Crown Information Extraction Based on QuickBird Image

在单树木测量中最重要的组成部分是树冠特征。树冠测量可以用来预测对培育措施的反馈,也可以通过林分生长与收获模型来估测树木生长量。本文基于QuickBird高分辨率遥感影像数据,利用eCognition对图像进行多尺度分割,采用面向对象的影像分析技术有效地提取了影像中的树冠大小信息。结果表明:研究区内的树冠生产精度为86.67%,用户精度为92.86%, Hellden精度为89.66%,总体分类精度为88.89%, Kappa系数为0.817。采用面向对象的分类方法所提取到的树冠信息其精度是符合要求的。

The crown feature is the most important component in individual tree measurement.Tree crown measurement can be used to predict the feedback of cultivation practice and to estimate tree growth through stand growth and harvest models.Based on QuickBird high-resolution remote sensing image data,this paper uses eCognition for multi-scale image segmentation and uses object-oriented image analysis technology to effectively extract the tree crown size information in the image.The results show that the crown production accuracy,user accuracy,Hellden accuracy,overall classification accuracy and Kappa coefficient in the study area are 86.67%,92.86%,89.66%,88.89%and 0.817,respectively.The accuracy of tree crown information extracted by object-oriented classification method is in line with the requirements.