應用手持式光達系統於闊葉次生林之林木點雲萃取分析

森林資源管理者透過森林資源調查，來獲取林分之性態值、林木生長、森林發展狀態、林地空間資訊等在森林經營管理上之重要資訊。隨著科技進步，各種儀器設備發展精進且快速，有別於傳統森林資源調查，於大面積調查需投入大量人力、物力與時間。光達系統之出現與導入森林資源調查，使其調查有著更進一步的發展，又手持式光達在重量及體積相對地面架站式光達較為輕便方便攜帶，應用於森林資源調查中更具有優勢，根據本研究結果顯示，使用手持式光達一方面可減少負重，使野外樣區的可及性提高、於調查人員可步行之範圍內，皆能執行掃瞄任務，調查更加靈活；在有效掃瞄範圍內可以迅速完成掃瞄任務並完整地蒐集現場資料，對於日後的重新檢核或複查作業有著莫大幫助；透過即時定位與地圖構建技術(SLAM)，對於林木及障礙物相互間的遮蔽效應，能夠以改變行走路徑之方式避免遮蔽；最後所掃瞄之點雲資料，透過大量的自動化操作，將林木萃取以進行分析，能夠提高測量林木性態值之效率。於本校達仁林場，選擇不同坡度、林分密度、立木徑級之樣區，以規劃之路徑進行掃瞄確保立木能被完整掃瞄，再以內業進行2種點雲萃取模式(手動測量模式、半自動萃取模式)與實測值之比較。本研究對於立木數量之結果顯示，手動測量模式萃取漏授誤差為0.8%、半自動萃取模式漏授誤差為9.3%；2種模式萃取各樣區胸徑之平均數與實測值皆不具有顯著性差異，且2種點雲萃取模式之間R2為0.93，具高度相關性，顯示地形不影響2種模式對於胸徑之萃取；與現地測量相比，2種點雲萃取模式分別節省15.6%、28.9%之時間耗費。

關鍵字

闊葉次生林；森林資源調查；手持式光達；點雲資料分析；林木點雲萃取

並列摘要

Forest managers use forest resource inventory to obtain important information, such as forest stand status, tree growth, forest development status, and forest space information. The development of various instruments is more and more efficient. This is different from the large amount of manpower and time required for traditional investigations. The introduction of LiDAR system allows further development of the investigation. In addition, the hand-held LiDAR is lighter and easier to carry than terrestrial LiDAR. The hand-held LiDAR can reduce the load, increase the accessibility of the field plot, making the survey more flexible. And this is of great help to the re-measured or re-inspection in the future. Through simultaneous localization and mapping (SLAM), it is possible to avoid the shadowing by changing the walking path. Through a large of automated operations, the analysis can improve the efficiency. In Dajen experimental forest station of our university, this research select 7 plots with different slopes, stand densities, and stand DBH grades, with the planned path to ensure stands can be scanned completely. And compare the two of point cloud extraction modes (manual measurement mode and semi-automatic extraction mode) with the actual measured value. The results of this research on extracting number of stands shows that omission error in manual measurement mode is 0.8%, and that in semi-automatic extraction mode is 9.3%. There is no significant difference between the average DBH and the measured value of each sample area extracted by the two modes. The R2 between the two point cloud extraction modes is 0.93, which is highly correlated. It shows that the terrain does not affect the extraction of the DBH. Compared with on-site measurement, the two point cloud extraction modes save 15.6% and 28.9% of time respectively.