| 摘要: | 車載雷射掃描系統可以直接獲取目標物三維點雲資料，因此可用來獲取詳盡的路廊資訊，路廊資訊則可應用於噪音模擬、道路安全、道路及相關設施維護、定位服務、汽車和行人導航，甚至發展未來駕駛協助系統。由於目前國內較缺乏以車載雷射掃描點雲自動萃取路廊地物的相關研究，因此本計劃預計全程執行期間為兩年，第一年（100年度）重點在於分析車載掃描點雲資料的特性並發展由點雲自動萃取並模朔路面地物，如道路面、燈柱、電線桿、及交通號誌等地物的演算法。第二年(101年度)重點則在發展由點雲自動萃取並模朔路旁地物，包含路樹、建物牆面、柱狀物、門、以及窗戶的演算法。兩年的研究期間亦將探討萃取及模朔其他路廊地物(如安全島、廣告招牌)的可行性。由於點雲資料量相當大且相關路廊地物眾多且複雜，因此演算法中將加入相關物空間知識並以每一掃描線斷面上的點雲為處理單元，處理過程中亦將引入強鈍估值法進行粗差偵錯協助每一掃描斷面基本特徵萃取與初步地物分類，並協助連續掃描斷面上分類點雲之群聚、叢集，進而發展萃取與模 Vehicle-based laser scanning system can be employed to directly huge 3D point clouds for the extraction of detailed road corridor information. The detailed road corridor information can be utilized for noise modeling, road safety, the maintenance of reverent road facilities, location-based services, navigation for cars and pedestrians, even for the development of future driver assistance system. Nowadays, the relevant studies on the automatic extraction and modeling for the terrain object in road corridors are lack in Taiwan, therefore the whole study period is two years for detailed study. The study in the first year focuses on the analysis of point clouds from the vehicle-based laser scanning system and the development of the automatic extraction and modeling for the terrain object on the road surfaces. The terrain objects on the road surface include road surface, pole-like objects (e.g. light lamps and traffic signs). The algorithms for automatic extraction and modeling of roadside terrain objects, including the trees, the building facades, building pillars, the doors and the windows, will be developed in the second year. The feasibility of automatic extraction and modeling for other objects, e.g. traffic islands and commercial boards in road corridors will be developed in this two-year study. Due to the huge data of point clouds and the complicated terrain objects in the road corridors, the algorithms will employ the relevant object knowledge in the basic processing unit, the scan line profile, for point calcification, grouping and clustering. Meanwhile, the concept of robust estimation for outlier detection will used for the basic feature extraction and classification and for the grouping and clustering point clouds belonging to the same object from the adjacent scan line profiles. It is expected to develop the practical algorithms for the automatic extraction and modeling the terrain objects in the road corridors. |
