郝振帮, 林丽丽, 余坤勇, 等. 基于无人机三维信息的杉木新造林林分参数遥感估测研究[J]. 西南林业大学学报(自然科学), 2023, 43(1): 108–116 . DOI: 10.11929/j.swfu.202111044
引用本文: 郝振帮, 林丽丽, 余坤勇, 等. 基于无人机三维信息的杉木新造林林分参数遥感估测研究[J]. 西南林业大学学报(自然科学), 2023, 43(1): 108–116 . DOI: 10.11929/j.swfu.202111044
Hao Zhenbang, Lin Lili, Yu Kunyong, Liu Jian, Zhao Gejin, Li Minghui, Song Xianfen, Yang Liuqing. Remote Sensing Estimation of Stand Parameters in a New Chinese Fir Plantation From UAV Three-Dimensional Information[J]. Journal of Southwest Forestry University, 2023, 43(1): 108-116. DOI: 10.11929/j.swfu.202111044
Citation: Hao Zhenbang, Lin Lili, Yu Kunyong, Liu Jian, Zhao Gejin, Li Minghui, Song Xianfen, Yang Liuqing. Remote Sensing Estimation of Stand Parameters in a New Chinese Fir Plantation From UAV Three-Dimensional Information[J]. Journal of Southwest Forestry University, 2023, 43(1): 108-116. DOI: 10.11929/j.swfu.202111044

基于无人机三维信息的杉木新造林林分参数遥感估测研究

Remote Sensing Estimation of Stand Parameters in a New Chinese Fir Plantation From UAV Three-Dimensional Information

  • 摘要: 以福建顺昌埔上国有林场的杉木新造林为研究对象,采用大疆Phantom 4 Multispectral无人机分2次获取研究区的无人机影像,并以无人机影像为数据源,从研究区的数字表面模型(DSM)中提取冠层高度模型(CHM)。根据局部最大值算法和分水岭算法,从CHM中获取研究区杉木的树高和冠幅数据;同时在研究区设立15个标准地,采用测量杆测定各标准地内所有杉木的树高和南北冠幅;以随机选取、且在影像中具有精确位置的265棵杉木为单木水平的实测数据,以及各标准地内杉木的平均树高和平均南北冠幅为林分水平的实测数据,分别从单木和林分角度对杉木树高和冠幅的遥感估测精度进行评价。结果表明:2次飞行作业之间树高的估测精度分别为90.86%和91.34%,南北冠幅的估测精度分别为83.55%和83.95%;在单木水平上,遥感估测的树高精度为R2=0.89、RMSE=22.37 cm、EA=91.00%;南北冠幅精度为R2=0.70、RMSE=27.33 cm、EA=82.22%;在林分水平上,树高的估测精度为R2=0.95、RMSE=12.27 cm、EA=94.61%;南北冠幅的估测精度为R2=0.82、RMSE=11.24 cm、EA=92.20%。遥感估测的树高均值比野外测量的树高均值小0.07 m,南北冠幅均值比野外测量的均值小0.04 m。基于无人机三维信息实现了研究区杉木树高和冠幅的精确估测,且在飞行参数一致的情况下,不同飞行区域和飞行批次之间的估测精度相近。研究可以为杉木新造林快速、稳定的监测和经营管理策略的科学制定提供基础数据。

     

    Abstract: A new plantation of Chinese fir(Cunninghamia lanceolata) in Pushang national forest farm, Shunchang county, Nanping city was used as the study object. The flight was conducted twice for UAV imagery collection using a Phantom 4 Multispectral. The Digital Surface Model(DSM) from UAV imagery was used to derive Canopy Height Model(CHM) in the study site. Then according to the algorithms of local maxima and watershed segmentation, the tree height and tree crown were detected from the CHM. At the same time, 15 standard sites were established in the study site. A measuring rod was used to measure all Chinese fir heights and north-south crown sizes. Taking the randomly selected 265 Chinese firs in the field with corresponding accurate positions to trees on the UAV imagery as the field data for tree-level, and the average tree height and average north-south crown size in each standard site as the field data for stand-level. Then the accuracy of remote sensed tree height and tree crown detection of Chinese fir was evaluated from the tree-level and stand-level, respectively. The accuracy of estimated tree height between the two flight operations were 90.86% and 91.34%, respectively, and the accuracy of estimated north-south crown were 83.55% and 83.95%, respectively. For the tree-level, tree height assessment yielded R2=0.89, RMSE=22.37 cm, and EA=91.00%; north-south crown assessment yielded R2=0.70, RMSE=27.33 cm, and EA=82.22%.At the stand-level, R2=0.95, RMSE=12.27 cm, and EA=94.61% for tree height assessment; R2=0.82, RMSE=11.24 cm, and EA=92.20% for north-south crown assessment. The average remote sensed tree height was 0.07 m lower than the average field measurement tree height, and the average north-south crown size was 0.04 m lower than the average field measurement tree height. The results highlight that UAV 3D scene can accurately estimate the height and tree crown of Chinese fir in the study site, and the estimation accuracy is similar between different areas and flight batches in the case of the same flight parameters. It can provide a reference for rapid, effective monitoring of newly planted Chinese fir, and scientific forest management.

     

/

返回文章
返回