發明
美國
16/194,847
US 10,685,474
METHOD FOR REPAIRING INCOMPLETE 3D DEPTH IMAGE USING 2D IMAGE INFORMATION
國立中央大學
2020/06/16
現今的彩色3D影像掃描技術通常結合兩種模組,一個是2D的彩色相機模組,一個是3D空間資訊擷取模組,常見的3D空間資訊擷取模組會使用結構光技術或是time of flight (TOF)技術,然而這兩種技術皆會受到環境當中雜訊影響(通常是紅外線雜訊),而造成點雲破損。我們提出一個能從不完整的點雲,透過其二維影像資訊,提供物件位置、面資訊與邊界,還原出物件與點雲的演算法。其執行步驟依序為對2D影像進行切割,以判斷可進行重建之特定區塊,並判斷此特定區塊的邊界資訊,接著將待判斷特定區塊內殘存的3D點雲資訊進行曲面或平面延伸,並將2D的邊界投影到3D空間中的對應錐面,最後由殘存的3D點雲延伸結果與對應錐面的交集可獲得此特定區塊之3D邊界與3D曲面資訊。 We propose an algorithm that can restore objects and point clouds from an incomplete point cloud, through its two-dimensional image information, to provide object location, surface information and boundaries. The execution step is to cut the 2D image in order to determine the specific block that can be reconstructed and determine the boundary information of the specific block. Then, the 3D point cloud information to be judged in the specific block is subjected to the surface or plane. And the 2D boundary is projected into the corresponding cone in the 3D space. Finally, the 3D boundary and 3D surface information of the specific block can be obtained by the intersection of the residual 3D point cloud extension and the corresponding cone.
智權技轉組
03-4227151轉27076
版權所有 © 國家科學及技術委員會 National Science and Technology Council All Rights Reserved.
建議使用IE 11或以上版本瀏覽器,最佳瀏覽解析度為1024x768以上|政府網站資料開放宣告
主辦單位:國家科學及技術委員會 執行單位:台灣經濟研究院 網站維護:台灣經濟研究院