融合光學影像與雷射掃瞄技術產製高細緻三維點雲

Abstract

於現代三維空間資訊測繪技術中，利用光達掃瞄技術所取得之高精度高密度的點雲資料受到各界領域廣泛應用。然而，受限於其資料獲取邏輯與機制，光達掃瞄所獲得的點雲資料時常有解析度不均、資料缺漏等問題。為此，本研究發展以影像與光達點雲為基礎之高細緻三維彩色點雲產製技術，利用光達資料內高精度的幾何資訊以及外部光學影像之輔助資訊以提升點雲成果品質。首先於影像的外方位參數求解上，應用影像匹配技術發展半自動之求解程序，提高資料整合作業效率。在完成影像對位程序後，利用攝影測量共線條件以及物空間中的直線與平面特徵，於幾何上重構點雲缺漏並密化原有點雲資料；於點雲顏色品質之改善上，萃取影像所包含豐富的光譜訊息賦予至已完成密化與重建後之點雲，產製高細緻的三維彩色點雲資料以提供後續應用。實驗成果顯示，本研究提出之方法能充分地融合點雲及影像資料，所取得彩色點雲成果於色彩或幾何完整性上皆有顯著提升，能夠有效降低現行自動化三維資料獲取技術實務應用之侷限，提升原始資料之實用效能與價值。

The Light Detection and Ranging (LiDAR) technique is a popular 3D surveying technique nowadays, which is capable of swiftly and precisely acquiring point cloud data with high density, for various engineering monitoring tasks. However, due to the limitation on the mechanism of the used technique, the point cloud data acquired by the LiDAR system frequently reaches unfavorable results, which have the flaws, such as uneven resolution and missing areas. In order to solve this problem, a combined approach for delivering high quality 3D colored point cloud is developed in this study, making the quality improvement of point cloud can be accomplished by integrating accurate geometric data from the LiDAR point cloud with the auxiliary information from images. The proposed approach, firstly determines the camera exterior orientation parameters by applying a semi-automatic process based on image matching technique. Secondly, for improving the geometric integrity of the raw data, the collinearity condition, as well as the geometry information from edge and planar features, will be used for point cloud reconstruction and densification. Finally, the abundant spectral information from multi-view images will be extracted and assigned to the refined point cloud data for generating high-quality result, which can be serve as a reliable and stable data source for further applications. The experiment result indicated that the proposed approach can adequately utilize the information from image data to increase the geometrical detail of the LiDAR point cloud. In conclusion, the quality improvement can be achieved through the use of data from different sources, which can successively reduces the limitation of current 3D surveying techniques, and also provides much more reliable and stable 3D information for applications.