一种基于三维激光点云数据的单木树冠投影面积和树冠体积自动提取算法

An Automatic Extraction Algorithm for Individual Tree Crown Projection Area and Volume Based on 3D Point Cloud Data

树冠投影面积和树冠体积是研究单木生物量估测、三维绿量测算等的重要测量因子。针对树冠结构复杂、形态各异,树冠因子难以精确测量等问题,为实现单木树冠投影面积和树冠体积的自动提取和精准测量,以三维激光扫描获取的树木点云数据为数据源,运用平面散乱点集凸包算法———Graham扫描算法和不规则体切片分割累加算法,以VC++6.0和Matlab 7.0混合编程,实现树冠投影面积和树冠体积的自动提取。以实验区选取具有代表性的22株不同树种的样木为研究对象,并与传统人工测量方法进行了对比分析。试验结果表明,树冠投影面积人工测量(A4)结果与点云数据的自动计算(AV)结果相关性较好,r=0.964(p<0.01),A4高于AV的平均值为25.5%。近似规则几何体方法树冠体积测量(VC)结果与基于点云数据的自动计算(VVC)结果之间呈现较强的相关性,r=0.960(p<0.001),VC低于VVC的平均值为8.03%。该方法基于高精度单木点云数据,实现了树冠结构的高精度快速重构、单木树冠投影面积和树冠体积的无损自动提取,可以为单木树冠结构的研究提供参考,在精准林业领域具有推广应用价值。

Tree crown projection area and crown volume are the important parameters for the estimation of biomass ,tridimen-sional green biomass and other forestry science applications .Using conventional measurements of tree crown projection area and crown volume will produce a large area of errors in the view of practical situations referring to complicated tree crown structures or different morphological characteristics .However ,it is difficult to measure and validate their accuracy through conventional measurement methods .In view of practical problems which include complicated tree crown structure ,different morphological characteristics ,so as to implement the objective that tree crown projection and crown volume can be extracted by computer pro-gram automatically .This paper proposes an automatic untouched measurement based on terrestrial three-dimensional laser scan-ner named FARO Photon120 using plane scattered data point convex hull algorithm and slice segmentation and accumulation al-gorithm to calculate the tree crown projection area .It is exploited on VC＋6.0 and Matlab7.0 .The experiments are exploited on 22 common tree species of Beijing ,China .The results show that the correlation coefficient of the crown projection between AV calculated by new method and conventional method A4 reaches 0.964（p＆lt;0.01）;and the correlation coefficient of tree crown vol-ume between VVC derived from new method and VC by the formula of a regular body is 0.960（p＆lt;0.001） .The results also show that the average of VC is smaller than that of VVC at the rate of 8.03% ,and the average of A4 is larger than that of AV at the rate of 25.5% .Assumed Av and VVC as ture values ,the deviations of the new method could be attributed to irregularity of the crowns’ silhouettes .Different morphological characteristics of tree crown led to measurement error in forest simple plot survey . Based on the results ,the paper proposes that ：（1） the use of eight-point or sixteen-point projection with fixed angles to estimate crown projections ,and （2） different regular volume formula to simulate crown volume according to the tree crown shapes .Based on the high-resolution 3D LIDAR point cloud data of individual tree ,tree crown structure was reconstructed at a high rate of speed with high accuracy ,and crown projection and volume of individual tree were extracted by this automatical untouched meth-od ,which can provide a reference for tree crown structure studies and be worth to popularize in the field of precision forestry .