基于DOM航空影像单木树冠提取

准确提取单木树冠边界是获取森林数量参数的重要基础,是高分辨率遥感图像林业应用的技术难题。基于DOM航空影像数据源,采用面向对象的方法对研究区内的2个树种的林分进行了单木树冠边界提取研究。首先利用桉树和杉木的空间分布矢量数据对DOM航空影像进行掩膜处理,在掩膜区域内进行多层次多尺度图像分割得到初步树冠分割结果,并剔除非树冠信息;再以树冠信息种子对象为基础,使用区域增长算法对树冠信息种子对象增长得到单木树冠范围;最后使用形态学滤波的方法优化单木树冠边界,完成林区内桉树和杉木两类树种的单木树冠边界提取。结果表明,由于不同树种的树冠存在尺度和形态差异,进行单木树冠分割时需要设置不同的参数才能到达较好的分割效果。本研究中桉树和杉木的单木树冠提取总体精度分别为86.75%与89.21%,可满足林业部门获取森林单木树冠的精度需求。

Individual Canopy Boundary Extraction Based on DOM Aerial Images

Accurate extraction of individual canopy boundary is an important basis for obtaining forest quantity parameters and a technical difficulty in forestry application of high-resolution remote sensing images.Based on DOM aerial image as data source,this paper adopted object-oriented method to extract the individual canopy boundary of two tree species in the study area.Firstly,the spatial distribution vector data of Eucalyptus robusta and Cunninghamia lanceolata were used to mask DOM aerial images,and the multi-layer and multi-scale image segmentation were carried out in the mask region to obtain the preliminary tree canopy segmentation results,and the non-tree canopy information was removed.Then,the tree crown information seed object was taken as the base,the regional growth algorithm was used to grow the crown information seed object to obtain the range of individual tree crown.Finally,morphological filtering method was used to optimize the canopy boundary of individual trees.Extraction of individual canopy boundary of E.robusta and C.lanceolata was completed in the forest area.The results showed that,due to the differences in the size and morphology of the crowns of different tree species,different parameters needed to be set to achieve a better segmentation effect.In this study,the overall precisions of single tree canopy extraction of E.robusta and C.lanceolata were 86.75%and 89.21%,respectively,which could meet the requirements of forestry department for the accuracy of obtaining forest single tree canopy.