Quantitative Inversion for Wind Injury Assessment of Rubber Trees by Using Mobile Laser Scanning
林木参数反演是森林资源管理与培育经营的关键环节。 迅速发展的激光探测与测量技术突破了传统测量方法, 可以快速的获取林木的空间三维信息, 在林业普查中具备高效率、 高精度的优势。 结合计算机图形学与图像学方法, 以中国最大的橡胶生产基地海南省儋州市长期受台风侵害下的两个不同品种橡胶林段(林段1 PR107, 林段2 CATAS7-20-59)为研究对象, 设计了面向离散激光点云的单株林木参数提取方法, 自动获取橡胶林木风害后的生物量指标。 首先, 通过人背负移动激光雷达获取林段点云数据并使用瑞利商求取台风造成的主枝干倾角, 以找寻每株橡胶树的树冠中心点。 其次, 对点云进行垂直投影, 并采用分水岭与Meanshift算法实现株株分离。 最后, 基于以上操作自动获取与实测值相近的林木参数, 例如冠幅、 冠径、 冠积、 叶面积密度、 叶面积分布以及主枝与分枝之间夹角等。 计算表明, 林段1与林段2东西冠幅分别为3.95和3.73 m, 与实测相差1.74%～6.27%; 林段1与林段2南北向冠幅分别为6.47和6.51 m, 与实测相差2.54%～4.02%; 林段1与林段2平均胸径分别为5.20和4.73 cm, 与实测相差0.64%～2.44%; 林段1与林段2平均冠积分别为168.01和141.80 m3, 与实测相差0.67%～0.85%; 林段1与林段2主枝干倾斜角分别为18.80°和13.11°, 与实测相差5.53%～7.09%; 林段1与林段2二级分支与主枝干的夹角分别为40.21°~69.23°和10.63°~32.14°, 它们相差62.63%; 林段1在不同天顶角下的叶面积指数均大于林段2。 通过对一定样本(150棵/每类林段)的分析结果与真实比对表明, 该方法对林木参数反演结果精度较高, 有效地评估了不同品种橡胶树在台风下的损伤度(如主枝干歪斜率、 叶面积密度及衰减分布)。 参数反演结果与实测值仅有8%的偏差, 此偏差主要由橡胶林分叶片稠密, 导致林分中叶面与枝干扫描数据获取缺失, 以及外界环境干扰如风力扰动、 点云拼接误差、 激光束发散率、 激光扫描范围等原因造成。 同时, 由于林段1(PR107)的橡胶树的主枝与分枝夹角、 冠积以及叶面积指数均大于林段2(CATAS7-20-59)的橡胶树, 导致林段1的橡胶树比林段的2橡胶树在台风侵害下更脆弱。 因此, 该研究可用于研究风力侵害对于不同森林地块的影响, 以及量化风害造成的生态系统紊乱的影响。 同时, 该方法解决了人背负激光扫描数据进行单株提取与林木参数反演的问题, 为激光测绘在林业中的应用提供了新思路。
The light detection and ranging (LiDAR) technique, which has the advantages of high efficiency and high accuracy in forest survey and is superior to the traditionalinformation acquisitionmethods, can be used to quickly obtain high-resolution mapping of morphological structures of forest. In this paper, two rubber forest plots (forest plot 1, clonePR107; forest plot 2, clone CATAS7-20-59) are taken as the study subjects, which are under the long-term impact of wind-induced disturbance severity and located in Danzhou city, the largest rubber production base of Hainan Island, China. First, point cloud of the forest plots through man-loaded mobile LiDAR was collectedand Ruili entropy method was designed to process the scanned data for calculating the slope angle of tree trunk (typhoon-induced) in order to find the canopy centre of each tree. Second, after the vertical projection of scanned forest points, Watershed and Mean shift algorithm were adopted to realize individual tree canopy delineation. Finally, many tree parameters, such as crown breadth, Diameter at Breast Height (DBH), crown volume, leaf area density, leaf distribution and included angle between trunk and main branches, were deduced automatically for analyzing the impact of typhoon disturbance on the two forest plots. Overall parameters obtained from our methods were compared with manual field measurements. The calculated average crown diameter in west-east direction of rubber trees in forest plot 1 and plot 2 using our method were 3.95 and 3.73 m, respectively, with false rate of 1.74% for forest plot 1 and 6.27% for plot 2. The calculated average crown diameter in north-south direction of rubber trees in forest plot 1 and plot 2 using our method were 6.47 and 6.51 m, respectively, with false rate of 4.02% for forest plot 1 and 2.54% for plot 2. The calculated average diameter at breast height (DBH) for forest plot 1 and plot 2 using our method were 5.20 and 4.73 cm, respectively, with false rate of 2.44% for forest plot 1 and 0.64% for plot 2. The calculated average crown volume for forest plot 1 and plot 2 using our method were 168.01 and 141.80 m3, respectively, with false rate of 0.67% for forest plot 1 and 0.85% for plot 2. The calculated average inclination angle of rubber trunk for forest plot 1 and plot 2 using our method were 18.80° and 13.11°, respectively, with false rate of 5.53% for forest plot 1 and 7.09% for plot 2. The calculated average included angle between trunk and branch for forest plot 1 ranged from 45.21° to 69.23°, and the calculated average included angle between trunk and branch for forest plot 2 ranged from 10.63° to 32.14°. Thedifference in the included angles of two forest plots was nearly 62.63%. Meanwhile, the leaf area index (LAI) of forest plot 1 derived fromhemispherical photos of various zenith angles was generally higher than forest plot 2. Compared with the in-situ measurements, the forest parameters from the subsample (scanned data of 150 trees per forest plot) were accurately retrieved using our method with a deviation of less than 8%. A variety of disturbance, such as the perspective occlusion caused by closed forest canopies, the error produced by multi-scan registration, vegetative elements moved by wind during the scanning process, beam divergence and scanning range constraint of the scanner, hampers the accurate scanned data acquisition and generates computer errors in the algorithm. Meanwhile, the included angle between trunks and branches, canopy volume and leaf area index of rubber tree clone PR107 (in forest plot 1) were overall higher than the parameters of rubber tree clone CATAS7-20-59 (in forest plot 2), resulting in the existence of higher vulnerability of clone PR107 than clone CATAS7-20-59 when wind damage propagation occurred in the forest plots. Thus, our research can be used to study the effects of wind disturbance on different forest plots and to quantify ecological instability of the forest in response to wind excitation. Our method makes a contribution to solving the problem of tree canopy delineation and forest parameter retrieval using man-loaded laser scanning technique, showing promise for further exploration of utilizing mobile terrestrial LiDAR as an effective tool for the applications in forest survey.
基金项目：国家重点研发计划(2017YFD0600900), 国家自然科学基金项目(31770591, 41701510), 中国博士后面上基金项目(2016M601823), 江苏高校优势学科建设工程项目资助
张艳侠：南京林业大学信息科学技术学院, 江苏 南京 210037
王佳敏：南京林业大学信息科学技术学院, 江苏 南京 210037
胡春华：南京林业大学信息科学技术学院, 江苏 南京 210037
陈邦乾：中国热带农业科学院橡胶研究所/农业部儋州热带作物野外观测实验站, 海南 儋州 571737
薛联凤：南京林业大学信息科学技术学院, 江苏 南京 210037
陈凡迪：南京林业大学信息科学技术学院, 江苏 南京 210037
备注：云 挺, 1980年生, 南京林业大学信息科学技术学院副教授
【2】Ardila J P, Bijker W, Tolpekin V A, et al. International Journal of Applied Earth Observation and Geoinformation, 2012, 15(4): 57.
【3】Karlson M, Ostwald M, Reese H, et al. International Journal of Applied Earth Observation and Geoinformation, 2016, 50: 80.
【4】Halperin J, Lemay V, Coops N, et al. Remote Sensing of Environment, 2016, 179: 170.
【5】Cartus O, Kellndorfer J, Rombach M, et al. Remote Sensing, 2012, 44(11): 3320.
【6】Castillo J A A, Apan A A, Maraseni T N, et al. ISPRS Journal of Photogrammetry and Remote Sensing, 2017, 134: 70.
【7】Ploton P, Barbier N, Couteron P, et al. Remote Sensing of Environment, 2018, 200: 140.
【8】Cao L, Coops N C, Innes J L, et al. Remote Sensing of Environment, 2016, 178: 158.
【9】Yun T, An F, Li W, et al. Remote Sensing, 2016, 8(11): 942.
【10】Pierzchaa M, Giguère P, Astrup R. Computers and Electronics in Agriculture, 2018, 145: 217.
【11】Jing L, Hu B, Noland T, et al. ISPRS Journal of Photogrammetry and Remote Sensing, 2012, 70(3): 88.
【12】Hu B, Li J, Jing L, et al. International Journal of Applied Earth Observation and Geoinformation, 2014, 26(26): 145.
【13】Cheein F A A, Guivant J. Computers and Electronics in Agriculture, 2014, 102(1): 19.
YUN Ting,ZHANG Yan-xia,WANG Jia-min,HU Chun-hua,CHEN Bang-qian,XUE Lian-feng,1CHEN Fan-di. Quantitative Inversion for Wind Injury Assessment of Rubber Trees by Using Mobile Laser Scanning[J]. Spectroscopy and Spectral Analysis, 2018, 38(11): 3452-3463
云 挺,张艳侠,王佳敏,胡春华,陈邦乾,薛联凤,陈凡迪. 基于移动激光扫描的橡胶林风害相关参数精准反演[J]. 光谱学与光谱分析, 2018, 38(11): 3452-3463