基于林地激光点云的树木胸径自动提取方法

Tree Diameter at Breast Height Automatic Estimation Based on Forest Terrestrial Laser Scanning

摘要:

提出了一种基于随机RANSAC模型的树木胸径自动提取算法。首先，采用布料模拟滤波（CSF）算法对林地点云数据进行滤波，获取树木、地面数据与数字地面模型（DEM）并提取树木胸径处点云，然后进行欧式距离聚类，最后基于随机random sample consensus (RANSAC）模型拟合树木模型，实现自动化的树木胸径提取。使用上海市青浦区某区域两林区样地的地面激光点云数据对该算法进行验证，与实际人工测量树木胸径的平均偏差分别为0.79cm和0.52cm。实验对比结果表明，该算法在精度与时间性能上均优于基于Hough变换的算法与基于最小二乘的算法。

Abstract:

This paper presents a tree diameter at breast height （DBH）estimation algorithm using terrestrial laser scanning （TLS） based on the randomized random sample consensus （RANSAC） model. First， the forest cloud data were filtered by the cloth simulation filtering （CSF） algorithm. Digital elevation model （DEM） and the point cloud at the DBH of the tree were extracted. Then， Euclidean distance clustering was performed. Finally， the tree model was fitted on the basis of the randomized RANSAC model. The algorithm was verified on laser point cloud data from two sample plots in Qingpu District， Shanghai. The average biases from the actual diameter of trees were 0.79 cm and 0.52 cm， respectively. Experiment results show that the algorithm is better than those based on Hough transform or the least-squares in terms of the accuracy and running time.