無人機光達測量於都市林樹高測量之應用 ―大安森林公園為例

Abstract

林木材積計算中樹高是重要參數，過去調查人員使用測高桿與雷射測距儀量測樹高，調查時間冗長且判釋樹高標準不一，本研究使用無人機光達測量大安森林公園樹木樹高，探討光達與其他測量樹高方法差異。 光達點雲利用布料模擬法分類地面點雲後使用KNN-IDW建立數值高程模型(DEM)，使用Pit-free方法建立數值地表模型(DSM)，DSM減掉DEM來建立樹冠高程模型(CHM)。使用樹木座標為圓心，分別在CHM網格資料建立半徑1到5公尺搜索範圍，比較不同搜尋半徑之均方根誤差後，選擇搜索半徑2公尺來搜索公園樹高。 四分位距挑選2024年樹高生長量異常林木達1280棵，現場實測林木樹高後，比較光達與實測樹高，結果平均樹高差距2.97公尺。七個樹種在顯著水準α = 0.05下，實測與光達樹高符合常態分佈及成對樣本t檢定有顯著差異。 公園平均樹高前三名樹種為肯氏南洋杉、黑板樹與白榕，平均樹高介於13到15公尺，利用實測樹高修正光達樹高資料後，公園樹高分布更符合實際情況。

Tree height is an essential variable in calculation of tree volume. In the past, surveyors used telescoping measuring poles and laser hypsometer to measure tree total height, but this method was time-consuming and had inconsistent standards. With advancements in LiDAR technology, this study explores using UAV LiDAR to measure tree height in Daan Forest Park and evaluate the differences between LiDAR and conventional tree height measurement methods. Ground points were classified using the Cloth Simulation Filter (CSF). A digital elevation model (DEM) was generated using the KNN-IDW method. A digital surface model (DSM) was generated using the Pit-free algorithm. The canopy height model (CHM) was derived by subtracting the DEM from the DSM. Using the tree coordinates as the center, search radius ranging from 1 to 5 m were established on the CHM to identify the highest canopy point. Based on comparisons of the root mean square error (RMSE), a search radius of 2 m was selected as the optimal radius for estimating tree heights in the park. Using the interquartile range (IQR) method, 1,280 trees with abnormal height growth were selected from the 2024 inventory. Comparisons between LiDAR and tree height showed a mean difference of 2.97 m. For the seven species, the differences between LiDAR and tree height were normal distribution according to the Shapiro–Wilk test, and paired-sample t-test indicated significant differences at the α = 0.05 significance level. The top three species in the height were Araucaria cunninghamii, Alstonia scholaris and Ficus benjamina with mean height ranging from 13 to 15 m. After correcting the LiDAR using park inventory, tree height distribution that more accurately reflected the actual conditions of the park.