應用特徵圖像匹配於數位影像相關法之三維全場形貌及變形量測

Abstract

數位影像相關法 (Digital Image Correlation, DIC) 為基於影像的一種非接觸式全場量測技術，其具備跨尺度跨領域的優勢。數位影像相關法之核心概念為參考影像中定義一樣板子集合，再至目標影像中搜尋其位置，以此量測待測物之座標、位移、速度、加速度、應變等資訊，搭配迭代演算法，可得到高精度的解。然而其對於影像之變形會有一定的限制，例如旋轉、縮放等。 本論文應用了基於特徵之圖像匹配，以提高圖像變形時之量測精度，並應用平行計算於全場多點量測，以增進全場量測時之計算效能。在實驗部分，本論文進行了懸臂薄板之變形及立體形貌量測，搭配雷射位移計及 Comsol 驗證其精度，以及使用一水桶進行模擬儲油槽等大型結構物之結構檢測，並以敲擊的方式模擬其力學損傷之行為，以全場變形及立體形貌進行結果的呈現。本論文也進行了風扇葉片之三維旋轉立體形貌量測，以模擬風機葉片的旋轉過程。本論文最後進行了橡膠材料的拉伸試驗，並以數位影像相關法進行精度的驗證。

Digital Image Correlation (DIC) is a non-contact, full-field measurement technique based on images, which possesses the advantages of cross scales and domains. The core concept of DIC involves defining a subset in the reference image and searching for its location in the target image. This allows for the measurement of coordinates, displacement, speed, acceleration, strain, and other information of the object being measured. Using with iterative algorithms, high-precision solutions can be obtained. However, it has certain limitations in handling image deformation, such as rotation and scaling. This thesis applies feature-based image matching to improve the measurement accuracy during image deformation and employs parallel computing for full-field multipoint measurement to enhance computational efficiency during full-field measurement. In the experimental part, this thesis conducts deformation and 3D surface measurement of a cantilever thin plate, verifies its accuracy with a laser displacement sensor and Comsol, and uses a bucket to simulate the structural detection of large structures like oil storage tanks. It simulates its mechanical damage behavior by knocking and presents the results through full-field deformation and 3D surface. This thesis also conducts rotational surface measurement of fan blades to simulate the rotation process of fan blades. Finally, this thesis conducts a tensile test on rubber materials and verifies the accuracy using DIC.