基于点云的树木枝干形态参数提取技术  

Morphological Parameters Extraction of Tree Branches Based on Point Cloud

作　　者：高士增[1] 张怀清[1] 刘闽[1] 朱中华 

机构地区：[1]中国林业科学研究院资源信息研究所,北京100091 [2]湖南攸县林业局黄丰桥林业工作站

出　　处：《东北林业大学学报》2014年第4期109-114,共6页Journal of Northeast Forestry University

基　　金：国家自然科学基金项目(31170590);国家"863"计划课题(2012AA102002)资助

摘　　要：以三维激光扫描的不同分辨率树木枝干点云数据为数据源,分别使用迭代的凸包算法和三角网生长法提取其不同层次的等值线模型和Delaunay三角网格模型,并使用射线碰撞检测技术提取以上模型的形态参数。通过对比分析表明,当分层数相同时,使用不同分辨率的模型提取的参数精度相差不大,与实际测量数值相比误差在5%以内;而当分辨率一定时,层数越多的模型,其效果越好,提取的参数精度越高。因此,在三维激光扫描仪的使用过程中,可适当降低分辨率以减少扫描时间,而在建模的过程中,通过提高分层数量以此获得较好的模型表现效果。With branches point cloud of different resolution as the data source, we extracted contour models and Delaunay trian- gular mesh model with different levels using iterative convex hull algorithm and triangulation growth algorithm respectively, and extracted the morphological parameters of these model by ray collision detection technique. By comparison, the accuracy of parameters extracted using models with different resolution differs little when stratified with the same number, and the error is less than 5% compared with the values in the actual measurement. When the resolution is certain, the more layers the model has, the better the effect is and the higher the accuracy of the extracted parameters is. Therefore, it may be ap- propriate to reduce the resolution to reduce the scan time in the process of 3D laser scanner, and increase the number of hierarchical to get a better model performance results in the modeling process.

关 键 词：三维激光扫描技术 DELAUNAY三角网 参数提取 

分 类 号：S791.27[农业科学—林木遗传育种]