高级检索
    马静怡, 黄华国, 黄侃, 邢路. 基于16线阵TLS数据的单木识别及林分断面积估测研究[J]. 北京林业大学学报, 2018, 40(8): 23-32. DOI: 10.13332/j.1000-1522.20180016
    引用本文: 马静怡, 黄华国, 黄侃, 邢路. 基于16线阵TLS数据的单木识别及林分断面积估测研究[J]. 北京林业大学学报, 2018, 40(8): 23-32. DOI: 10.13332/j.1000-1522.20180016
    Ma Jingyi, Huang Huaguo, Huang Kan, Xing Lu. Individual tree detection and stand basal area estimation based on 16 linear array TLS data[J]. Journal of Beijing Forestry University, 2018, 40(8): 23-32. DOI: 10.13332/j.1000-1522.20180016
    Citation: Ma Jingyi, Huang Huaguo, Huang Kan, Xing Lu. Individual tree detection and stand basal area estimation based on 16 linear array TLS data[J]. Journal of Beijing Forestry University, 2018, 40(8): 23-32. DOI: 10.13332/j.1000-1522.20180016

    基于16线阵TLS数据的单木识别及林分断面积估测研究

    Individual tree detection and stand basal area estimation based on 16 linear array TLS data

    • 摘要:
      目的地基激光雷达(TLS)可以对林冠下层进行快速、非破坏性的三维测量,与传统森林参数调查相比,节省了大量人力、物力和时间,在林业调查中有广泛应用。目前的研究集中在基于全方位TLS数据的参数提取,全方位扫描获取的点云数据量庞大,所需扫描时间较长,而针对快速扫描的多线阵点云数据的研究较少,相关算法有待提出,多线阵激光雷达数据的应用能力也有待验证。
      方法以北京市东升八家郊野公园和奥林匹克森林公园内的人工林为研究对象,基于多个单站扫描采集的16线阵TLS点云数据,提出了一种新的树干识别算法。该算法利用点云到达目标单木及周围其他物体距离的差异,检测出树干表面点云,并结合随机采样一致性(RANSAC)算法拟合圆,提取单木胸径;在此基础上引用角规抽样技术,进行林分平均胸高断面积的估测。
      结果对于多个单站扫描数据,单木检测率均在80%以上,株数密度最小的样地单木检测率可高达95%;对于单站数据,单木平均检测率随着扫描半径的增加而下降,在10 m左右范围内可以达到较高的检测率。以样地中被正确检测到的单木胸径估测值与实测值建立回归方程,单木胸径估测的决定系数R2在0.72~0.82之间;计算各样地单木胸径实测值与估测值的平方平均数,林分平均胸径估测精度均在90%以上,最高可达到99%,表明在样地水平上有较好的胸径估测效果。由TLS提取的胸径值结合角规抽样原理计算林分平均断面积估测值,与实测值相比,林分平均断面积估测精度可以达到90%左右。
      结论本文提出的算法能够基于单帧16线TLS数据提取单木参数,实现林分平均胸径及单位断面积的快速高效估算,为林业调查提供了一种新方法。

       

      Abstract:
      ObjectiveTerrestrial laser scanning (TLS) can be used for fast and non-destructive 3D measurement of forest canopy with less manpower, material resources and time consumption than traditional methods, and it has been widely used in forestry inventory. Most current studies use 360 degrees of scanning to conduct parameter extraction, which requires relatively long time to scan forest trees and process the big data. Fewer study has been found on the faster multi-linear array scanning; and corresponding algorithms on processing the sparse point cloud data have not yet developed. As a first trial, the application capability of multi-linear array TLS on forest parameter extraction needs to be verified.
      MethodA new stem detection algorithm was proposed based on the 16 linear array TLS data from a single-station scanning. The key to the algorithm is using the distance difference to the scanning station to differ tree stems from other objects around them. The random sample consensus (RANSAC) algorithm for circle was used to fit the stem point cloud at breast height and extract DBH. After detection of all stems, the angle gauge method was introduced to estimate the basal area of the forest. A case study was performed in two places: the Dongsheng Country Park and the Olympic Forest Park in Beijing.
      ResultThe detection rate of individual trees was above 80% if averaging multiple single-scanning TLS data; the detection rate in the sparse plot was about 95%; the detection rate decreased with the increasing distance from the scanning station, where 10 m was the maximum to achieve a relatively high detection rate. The determination coefficient of estimated DBH of individual trees calculated by the regression equation ranged in 0.72-0.82. The accuracy of stand average DBH was higher than 90%, which was good at the plot level. The forest plot basal area was estimated based on the DBH extracted from TLS using the angle gauge method. Compared with the field measurement, the relative accuracy of basal area was about 90%.
      Conclusion This method can accurately acquire single tree DBH and stand average DBH as well as basal area based on the 16 linear array TLS data with high efficiency, which provides a new choice for forestry survey.

       

    /

    返回文章
    返回