基于近景摄影测量技术的关键林分参数的获取研究  

Study of Obtain of Key Parameters of Forest Stand Based on Close Range Photogrammetry

作　　者：张志华[1] 云挺[1] 薛联凤[1] 李卫正[2] 张天安[1,3] 孙圆[4] 

机构地区：[1]南京林业大学信息科学与技术学院,南京210037 [2]南京林业大学现代分析测试中心,南京210037 [3]南京林业大学生物与环境学院,南京210037 [4]南京林业大学林学院,南京210037

出　　处：《科学技术与工程》2017年第12期85-92,共8页Science Technology and Engineering

基　　金：国家自然科学基金(31300472);农业部儋州热带作物科学观测实验站开放课题基金(RRI-KLOF201501);国家“973”计划(2012CB416904);中国博士后面上基金(2016M601823);省级大学生实践创新计划立项项目(201610298013Z)资助

摘　　要：为了研究利用摄影测量技术来获取和反演关键林分参数的准确性,从而为推算其他林分因子提供相关思路,以实现快捷高效的森林资源调查。选取了一片亚热带人工混交林作为实验样地,并利用样地中6个优势树种的实测数据分别构建了多个胸径-树高、冠幅回归模型,并得到了对应的最优模型;在此基础上,利用从点云数据中获取的树高和冠幅,反演出单株树木的胸径;最后通过分析实测胸径与基于点云计算得到的胸径之间的相关性来判断算法的准确度。结果表明:(1)6个树种的实测树高与基于点云获取的树高相关性较高,R^2均在0.7以上;(2)多数树种的实测冠幅与获取的冠幅相关性偏低,其中仅喜树的冠幅R^2高达0.88,其余树种的冠幅R^2最高只有0.61;(3)最优回归模型拟合胸径与实测胸径的回归分析中,发现高大落叶乔木比如马褂木、喜树的R^2均在0.7以上,而落叶小乔木比如樱花树以及柏类树木比如圆柏、龙柏等R^2最高仅0.25,雪松的R^2则为0.54。In order to study the accuracy of obtaining and retrieving key parameters of forest stand by using the photogrammetry ,and relevant means was provided for calculating other stand factors, then achieved a fast and efficient investigation of forest resources. A subtropical mixed plantation was selected as experimental plot, and several regression models using measured tree height H and crown diameter C were constructed to predict diameter at breast height D of six dominant tree species of the plot, then the corresponding optimal regression model of each tree species was yielded. Based on this,the H and C obtained from point cloud to retrieve the Z） of a single tree were used,then the accuracy of approach was predicted from the relationship between manually measured D and the calculated value using our method. It is obtaimed： （T） here, a high correlation between the measured H and the H acquired from point cloud data （PCD） was obtained, the R2 values are greater than 0. 7 , ？ the correlation between the measured C and the C acquired from PCD for most species is at a low level, only camptotheca acuminatas R^2 value is up to 0. 88, the highest R^2 value of the other species is 0. 61. ？ in the regression analysis of the D fitted from the optimal regression models and the manually measured D , was found the R^2 values of tall deciduous trees （ e. g. Liriodendron chinense and Camptotheca acuminate） are above 0. 7, while the highest R^2 value of small deciduous （ e. g. Sakura） and Cupressaceae trees （e. g. Juniper and Sabina chinensis） is only 0. 25, and the R2 value of Cedrus deodara is 0. 54.

关 键 词：近景摄影测量 林分参数反演 树冠提取 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]