以布料模擬濾波(CSF)進行光達點雲分類之研究-以洛杉磯郡為例

Abstract

隨著光達技術進步，結合全球定位系統、慣性系統的運用。掃描後的點雲和產製數值地形模型的成果精度逐漸提升。產製數值地形模型前需要先將點雲進行地面點分類，目前業界進行點雲分類使用半自動化方式。原始資料先經由人工去除明顯高低雜點，再透過軟體設定參數自動化分類地面點與非地面點。分類成果會因地形、地貌、坡度差異，執行演算法時導致成果品質有異。本研究根據不同坡度組成的範圍區進行點雲分類參數調整，選洛杉磯郡不同坡度比例組成的實驗區進行參數設定分類測試，比較專業軟體之實驗成果和美國地質調查局網站的已知分類值差異性。透過參數設定尋找該不同坡度適合的數值，提升實際分類效率和精度品質。探索參數設定對於不同坡度的影響，演算法之細項參數對於地面點的萃取的意義及因果關係。

With the progress of lidar technology, global position system and inertial measurement unit, the accuracy of point cloud and the digital elevation model has been gradually improved. Nowadays, survey teams usually use a semi-automatic method for point cloud classification. First of all, analysts remove the high and low noise points from the original data manually. Furthermore, setting the parameters in the professional software is the next step for automatically classifying the ground points and non-ground points. Due to the difference in terrain, relief and slopes, even employing the same algorithms may cause the different result of classification. In the study, the parameters are adjusted to classify the point cloud data which areas are composed of different slope types in Los Angeles County. Compare the results of the experimental ground points in Cloud Compare and the known classification original classification values on the USGS’s website. The appropriate parameters for the different slopes can improve the classification efficiency and accuracy quality. Explore the impact of parameters settings to ground point extraction and check the significance of different kind of algorithms.