利用位能場做三維物體之形狀比對與辨識

Abstract

本篇論文之目的在探討利用位能場模型來協助三維物體之形狀比對與辨識 。此模型假設三維形狀樣版周界帶有均勻的電荷。若將縮小的待辨識物體 模型置於形狀樣版之三維物體內，則彼此之間的推斥作用經由平移、旋轉 會自動將二者對正。如果待辨識物體逐次增加其尺寸，則正確的形狀樣版 將可使待辨識物體放大至最大體積。據此，形狀之比對與辨識得以完成。 相較於其他的形狀比對方法，上述方法具有下列的優點：(1)待測物與樣 版不需先旋轉對正，(2)兩者尺寸不需先做正規化，(3)待辨識物體周界可 以不連續。由於牛頓場模型不能保證兩個三維物體之周界不接觸，我們利 用位能場隨兩點距離做較迅速遞減之廣義場模型來產生物體間所需要之推 斥作用。 The goal of this thesis is to investigate the shape matching and recognition of 3D objects using artificial potential fields. The potential-based approach recognizes the shape of a 3D object by identifying the best match from a selected group of template objects. The proposed model assumes that boundary of every 3D template object is uniformly charged. An initially small input object, represented by its boundary samples, placed inside a template object will experience the repulsive force and torque arising from the potential field. A better match in the shape between the template object and the input object can be obtained if the input object translates and reorients itself to reduce the potential while growing in size. The input object with the largest final size corresponds to the best match and represents the shape of the giveninput object. The potential and the associated repulsive force and torque between the input object and the template object are analytically tractable. The proposed approach is intrinsically invariant under translation, rotation and size changes of the input object.