航空图像中基于DSM的多尺度树冠分割

提出了一种基于数字表面高程模型（DSM）和多尺度权聚类图像分割的算法。首先通过DSM选择需要分割的候选树冠顶点，利用航空摄影测量的方法将树冠顶点投影到航空相片作为种子点；然后对树冠进行分割，其方法是在一个全局分割测度的约束下，利用代数多网格的数值计算方法，逐步进行权聚类，形成多尺度的分割图像，通过不同的选权参数，获得有意义的分割区域，从多尺度的分割图像中，根据种子点所在位置和分割图像大小确定分割的树冠。经过实际图像验证，此算法能对不同角度投影的树冠进行有效分割。

Multiscale Segmentation of Tree Crown from Aerial Images Based on the DSM

A segmentation method by multiscale weight aggregation is presented for segmentation of the tree crowns based on the digital surface model(DSM).The tree top from the DSM for the segmentation of the tree crown is firstly selected.Then the tree top is projected to the aerial image as the seed of the segmentation.Based on a global measure,an algebraic multigrid technique is used to construct the pixels to a multiscale segmented image by weight aggregation.Through the position of the seed and the image size,the segmented tree crown is obtained.The experimental results demonstrate that the algorithm can accurately segment the tree crown which is projected to different viewing angle aerial images.