基于三维激光点云数据的树冠体积估算研究  

Estimation of Tree Crown Volume Based on 3D Laser Point Clouds Data

作　　者：刘芳[1] 冯仲科[1] 杨立岩[1] 徐伟恒[2] 黄晓东[1] 冯海英[1] 

机构地区：[1]北京林业大学精准林业北京市重点实验室,北京100083 [2]西南林业大学计算机与信息学院,昆明650224

出　　处：《农业机械学报》2016年第3期328-334,共7页Transactions of the Chinese Society for Agricultural Machinery

基　　金：科技北京百名领军人才培养工程项目(Z131105000513003);北京林业大学青年教师科学研究中长期项目(2015ZCQ-LX-01)

摘　　要：树冠体积是预估树木生物量的重要参数之一。为了实现对树木冠体体积无损高精度量测,随机抽取了6个树种、共计30棵树木的三维激光点云数据作为数据源,对树冠体积的求算方法进行研究。首先,对三维激光点云数据进行匹配、拼接、去噪及压缩等处理,提取冠体点云数据;其次,提取每一棵样木树冠的边缘特征点;最后,应用不规则三角网TIN的原理算法计算冠体体积。本文所提取的边缘特征点能够最大限度地维持树冠冠体的整体不变形,还能够继续去除部分冗余数据,缩短了不规则三角网TIN的构建时间,提高了计算效率;此外,树种包含有针叶树和阔叶树,在冠形上既有针叶树所特有的冠体体态特征,又有阔叶树的冠体体态特征,其研究结果具有一定的代表性。本文采用的方法与已有文献计算结果对比表明:均方根误差为0.832,平均绝对误差为0.49,平均相对误差为1.75%,可看出二者之间差异较小;同时在30个样木中随机抽取5个样木的人工测量结果与本研究相比较,取得的精度相对较好。采用本研究所得结果精度较高,能够满足生产需求。The tree crown volume is one of the important parameters to estimate the biomass. In order to make an accurate measurement of the tree crown volume with non-destruction,this paper took 3D laser point cloud data,which were used as a data source,to calculate volumes of sample trees. The 3D laser point cloud data were randomly selected by six species,totally 30 trees. First of all,this paper extracted the point cloud data of tree crown volume from all points after matching, mosaic, denoising and compression etc. Secondly, it extracted the edge feature points from the tree crown through the programming algorithm. Finally,the crown volume was calculated by using the principle of irregular triangle net（ TIN）. In this paper,the edge feature points,extracted from the programming algorithm,can maintain the whole body of the crown. The algorithm can further remove the redundant data,shorten the construction time of TIN and improve the calculation efficiency. In addition,tree species also have certain representativeness. Because they included conifer and broadleaf trees,so the crowns not only have the crown body posture characteristic of conifers,but also have crown body posture characteristic of broadleaf trees. The results were as follows： the RMSE was 0. 832,the average absolute error was 0. 49,and the average relative error was 1. 75%. Comparing with the artificial measurement results by selecting five sample trees randomly,the precision was relatively good. It can be seen that there are few gaps between the two results,of which the accuracy can meet the requirements of production.

关 键 词：树冠体积 三维激光点云 边缘特征点 不规则三角网 

分 类 号：S758.1[农业科学—森林经理学] TP391.4[农业科学—林学]