机载LiDAR高差和高分影像的城市树冠提取比较

发布时间：2018-05-15 12:43

本文选题：LiDAR数据 + 高空间分辨率航空影像　；参考：《遥感技术与应用》2017年05期

【摘要】：在机载LiDAR(Light Detection and Ranging)数据和高空间分辨率航空影像的支持下,以城市为实验区,实现了单木树冠提取。首先通过LiDAR数据获取高差模型,将其作为包含林木的感兴趣区,再通过掩膜方式提取高分影像上的相同区域,然后采用标记分水岭分割算法分别对两幅感兴趣区影像进行树冠提取,最后以人工勾绘树冠结果为参考评价分割精度,比较了两种数据源提取树冠的优缺点。结果显示,利用LiDAR数据获取的高差模型中包含85.28%的林木信息,林木区域提取的效果显著;基于高分影像得到的分割结果较好,F值为57.14%,基于高度差值模型影像的分割结果较差,F值为42.47%。表明分水岭算法方便可行,且高分影像提供的二维信息更适用于树冠提取。
[Abstract]:With the support of airborne LiDAR(Light Detection and Ranging) data and high spatial resolution aerial images, single tree crown extraction was realized in the experimental area of city. First, the height difference model is obtained from the LiDAR data, which is regarded as the region of interest including trees, and then the same region on the high score image is extracted by masking. Then, two images of region of interest are extracted by using marked watershed segmentation algorithm. Finally, the segmentation accuracy is evaluated by using the results of artificially marked tree crown as a reference, and the advantages and disadvantages of the two data sources for tree crown extraction are compared. The results showed that the height difference model obtained by LiDAR data contained 85.28% of the forest information, and the effect of forest region extraction was remarkable. The better segmentation result based on high score image is 57.14, and the lower F value is 42.47 based on height difference model. The results show that the watershed algorithm is convenient and feasible, and the 2D information provided by high score image is more suitable for tree crown extraction.
【作者单位】：南京信息工程大学地理与遥感学院;
【基金】：国家自然科学基金项目“异速增长和资源限制模型结合多源遥感数据估算森林地上生物量研究”(41471312)
【分类号】：TN957.52

【相似文献】