Please use this identifier to cite or link to this item:
http://tdr.lib.ntu.edu.tw/jspui/handle/123456789/92025
Title: | 以多維影像融合進行表面邊緣重建與關鍵尺寸量測 Multi-dimensional image fusion for surface edge reconstruction and critical dimension measurement |
Authors: | 陳俞安 Yu-An Chen |
Advisor: | 陳亮嘉 Liang-Chia Chen |
Keyword: | 光學量測,三維光學檢測,三維邊緣重建,關鍵尺寸,影像融合, optical metrology,3D measurement,surface edge reconstruction,critical dimension,image fusion, |
Publication Year : | 2021 |
Degree: | 碩士 |
Abstract: | 三維光學量測技術由於具備量測物體深度與不易受量測視角變化影響的性質,在自動化檢測領域中有許多二維量測無法與其相比的優勢。然而三維光學量測在碰上銳利邊緣或深度的急遽變化時,因為難以捕捉該處的反射光,經常導致重建出的三維點雲邊緣經常出現量測雜訊及資料破碎的情況,進而產生量測誤差。
本文提出一種結合三維與二維方式新的量測演算法,目標為結合三維可量測深度以及二維易於判斷邊緣位置的優勢,以重建出可量測關鍵尺寸的三維形貌資訊。首先使用結構光投影獲得原始三維點雲,藉由前處理去除雜訊與其他不必要的資料,得到目標尺寸所在之表面點雲。接著透過邊緣偵測法從二維影像獲得二維邊緣資訊,將其於相機模型中的投影與表面點雲擬合之B-spline曲面相交,以得到物體邊緣在三維空間中的關鍵位置,進一步獲得自動化光學檢測所需的關鍵尺寸。 為了測試此方法重建邊緣的可行性與可靠度,進行實際量測數種不同性質的樣本及其尺寸,並與三次元座標接觸式量測儀及既有之三維點雲邊緣偵測法所得之結果進行比較。從實驗結果可知,提出之演算法能夠以優於0.07 %的精度測量寬度、距離等關鍵尺寸,在量測難度更高的曲面特徵也有0.12 %的誤差百分比。證明本研究可重建出精確三維邊緣及擷取可靠之關鍵尺寸。 With its excellent ability to measure the depth of object profile and insensitive to viewing angles, 3D optical metrology has unique advantages in the field of automated optical inspection over 2D measurement. However, when 3D optical metrology measures onto sharp edges or sudden changes of a surface, it becomes rather difficult to capture the reflected light, leading to undesired measured errors, thus obtaining unacceptable surface and inaccurate critical dimensions. In this work, the proposed method integrates 3D and 2D measurement into a fused imaging method, in order to obtain depth detection by 3D imaging and critical dimension by 2D imaging simultaneously. First, the original 3D point cloud is acquired by structured light projection, and the surface point cloud of the target is obtained by pre-processing to remove image noises. Following this, the 2D surface edges or contours are extracted from the two-dimensional image using the proposed edge detection. By projecting the 2D edge along the camera view direction which is defined by the calibrated camera model, The geometric intersection between the projecting edge vector and the extrapolated B-spline surface generated from the surface point cloud represents the 3D surface edge or contour underlying measurement. To verify the feasibility and reliability of the proposed method, several samples with different geometric properties and critical dimensions were measured and compared with the results obtained by a calibrated coordinate measuring machine, as well as other existing methods. Verified by the experimental results, the proposed algorithm is able to measure the critical dimensions for width and distance with measured errors less than 0.07 % for regular analytic surfaces, and 0.12 % for freeform surfaces. It is demonstrated that this proposed method has capability for the reconstruction of accurate 3D surface edges and the extraction of object critical dimensions. |
URI: | http://tdr.lib.ntu.edu.tw/jspui/handle/123456789/92025 |
DOI: | 10.6342/NTU202200241 |
Fulltext Rights: | 同意授權(限校園內公開) |
Appears in Collections: | 機械工程學系 |
Files in This Item:
File | Size | Format | |
---|---|---|---|
ntu-110-2.pdf Restricted Access | 35.9 MB | Adobe PDF | View/Open |
Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.