Error Analysis of Ground-based LiDAR Apply on Stem Mapping and Application

傳統森林調查是非常費工耗時的工作，但是森林調查除了可以為森林經營者提供更新林分資訊，使得管理者更有效管理林分。在森林調查項目中，除了林分資訊外，單木之基本資訊也是非常重要，例如胸高直徑(DBH)，樹高以及立木位置。 當中立木位置可以協助管理者通過資料，對應現場木位的相應資訊。這對於長期的森林經營和學術研究貢獻很大。 本研究通過對照傳統以經緯儀和3D雷射光達掃描儀分別製作立木位置圖，比較兩者差距。結果差距在 0.1 公尺以下的立木的佔 47.5%，差距 0.2-0.5 公尺的立木佔50.8%，差距 0.5 公尺以上的立木佔 1.7%。 除此之外，本研究以牛頓迭代法（Newton-Raphson method）進行自動化定義圓中心。於整個林分各樣木在不同的掃描站數量下，3D雷射光達掃本身之誤差測試。結果在只有一個測站的情況下，成功被掃描之樣木當中，最大的誤差為8.5公分，29.79%之誤差低於1公分，65.96%之誤差低於5公分。在只有三個測站的情況下，成功被掃描之樣木當中，最大的誤差為9.8公分，41.18%之誤差低於1公分，82.35%之誤差低於5公分。 再進一步發現，同一棵樹同時被兩個測站掃描可以增加其位置判別的準確度。單被一個測站掃描，如點雲不足的，誤差可到10公分，但如果是兩個測站或是上，其誤差可以解到2公分以下。同樣方法對樹圓周被掃描百分比進行分析，結果呈現最少要有90度(25%)的圓被掃瞄出來，其所定義出來的立木位置誤差可到1公分以下，而如果被掃瞄到只有45度(12.5%)，其所定義出來的立木位置在4公分以下。 遮蔽效應和掃描站架設有重大的關係，如遇到林分密度高的林分，需要在較密之區域增加額外之掃描站，以確保每棵立木都被最少兩個測站掃描以及其收集之圓控制上25％以上，以有利於提升繪制立木位置圖之精準度。而為了確保每棵立木都能成功地被掃描，除了考量單木的精準度外，樣木數量的精準度也是非常重要。實驗結果顯示，三個掃描站便可收集樣區中百分之九十以上的樣木，但在實際應用上，需要百分之百的樣木數量情況下，五個掃描站是強調建議被使用的。Traditional forestry inventory is a time consuming and costly task. However, it can provide up to date information of the stand to the forest manager in order who can have a better management strategy. There is not only information of stand will be collected during the inventory, as well as the individual trees information. For instance, Breast height diameter (DBH), tree height and location are also essential information. Stem mapping can link the field and the data together; it contributes lots not limited on long-term forest management and research. This study compares the stem maps drawn by Total station theodolite (TST) and terrestrial LiDAR. Result the difference are below 0.1m, 0.2m-0.5m and above 0.5m are 47.5%, 50.8% and 1.7% respectively This study use the Newton-Raphson method to determine the center of a circle of each stem automatically. Mapping the whole stand by various number of scanning stations. Under one station scenario, the maximum error is 8.5cm and 29.79% of scanned stems under 1 cm error while 65.96% below 5cm. In three stations scenario, the maximum error is 9.8cm and 41.18% of scanned stems under 1 cm error while 82.35% below 5cm. Furthermore, the frequency being scanned affecting the accuracy, which scanned by two stations with higher accuracy than that scanned by one station. Also, if there are not enough effective point clouds, the error could be as high as 10cm. experiment repeated by the same method to find out the effect of circumference percentage scanned. Result showed that at least 25% (90∘) of circumference being scanned provide accuracy less than 1 cm while 12.5% (45∘) of circumference being scanned only give 4cm accuracy. The Occlusion issue and allocation of scanning station with high correlation, additional scanning station needed in the dense stand to ensure each stem scanned by at least two stations and collect 25% or above of circumference to maintain higher accuracy of stem map. To ensure all stems can be scanned, both number of stems and accuracy of stem location are important. The result of this study showed 3 scanned stations collected above 90% of stems. However, during application in forest inventory, 100% of stem is needed, so 5 stations setting is strongly recommend