基于模糊熵的單木骨架重建方法

摘" 要： 在對點云數(shù)據(jù)進(jìn)行采集時，由于激光雷達(dá)設(shè)備和拍攝高度等因素導(dǎo)致點云質(zhì)量存在差異，將影響后續(xù)重建結(jié)果。針對輸入點云質(zhì)量不高或拓?fù)浣Y(jié)構(gòu)不完整的問題，文中采用一種基于骨架的方法提取樹木的原始骨架，通過最短路徑算法構(gòu)建一個最小生成樹（MST），通過模糊熵迭代方法去除冗余部分簡化原始骨架。利用基于優(yōu)化的擬合圓柱體序列進(jìn)行重建，以近似模擬樹枝的幾何形狀。實驗結(jié)果表明，在點云數(shù)量非常有限的情況下，該方法的單木重建精度也可以達(dá)到85%左右。而在點云數(shù)量較多的情況下，重建精度可提高至90%左右，且能夠處理不同形狀和結(jié)構(gòu)的樹木，對樹木主干以及具有高密度點云的矮樹具有良好的重建效果。

關(guān)鍵詞： 激光雷達(dá)； 樹木骨架； MST； 模糊熵迭代； 擬合圓柱體序列； 重建精度

中圖分類號： TN958.98?34" " " " " " " " " " " " "文獻(xiàn)標(biāo)識碼： A" " " " " " " " " " " "文章編號： 1004?373X（2025）07?0139?07

Fuzzy entropy based reconstruction method for single tree skeleton

ZHAO Yonghui1， YAN Peiyu1， GU Juntao2， LI Zhen1， PENG Xunhui1， LI Chao1， LIU Shuyu1

（1. College of Computer and Control Engineering， Northeast Forestry University， Harbin 150040， China;

2. Heilongjiang Cyberspace Research Centre， Harbin 150090， China）

Abstract： During the acquisition of point cloud data， the point cloud quality varies due to factors such as LiDAR equipment and shooting height， which will affect the results of subsequent reconstruction. In view of the poor quality or incomplete topology of the input point clouds， this paper adopts a skeleton?based method to extract the original skeleton of the tree， constructs a minimum spanning tree （MST） with the shortest path algorithm， and removes the redundant parts by the fuzzy entropy iterative method， so as to simplify the original skeleton. Reconstruction is carried out with an optimization?based sequence of fitted cylinders to approximate the geometry of simulated tree branches. The experimental results show that the accuracy of single tree reconstruction can reach about 85% even in the case of a very limited number of point clouds. In the case of a larger number of point clouds， however， its reconstruction accuracy can be increased to about 90%， and it can deal with trees of different shapes and structures， and the reconstruction results are good for tree trunks， as well as dwarf trees with high?density point clouds.

Keywords： LiDAR; tree skeleton; MST; fuzzy entropy iteration; fitted cylinder sequence; reconstruction accuracy

0" 引" 言

激光雷達(dá)能快速、準(zhǔn)確、廣泛地獲取三維空間信息，從不同角度生成不同分辨率的點云，取代多種測量儀器來獲取單木的胸徑、樹高和樹冠等結(jié)構(gòu)參數(shù)，生成不同質(zhì)量的點云，點云對一些重建方法影響較大，導(dǎo)致拓?fù)浣Y(jié)構(gòu)不完整[1?2]。樹木拓?fù)浣Y(jié)構(gòu)的重建方法大致分為兩類：基于分割和基于骨架。基于分割的方法先將樹點云分成小的子集，再將它們連接起來形成樹木結(jié)構(gòu)，但這種方法對輸入數(shù)據(jù)質(zhì)量高度依賴；而基于骨架的方法是通過原始輸入點云直接提取骨架。文獻(xiàn)[3]提出基于動態(tài)圓柱擬合的方法對單木骨架進(jìn)行提取。……