本研究的目的是要使用既有的3D人體計測資料庫來發展3D人體動畫模型，並以5種身高和3種BMI分別建立男女各15個不同體型的動畫模型。這個人體計測資料庫稱為3D Human Body Bank，裡面記錄的人體計測資料是使用3D人體掃描儀取得的真實人體表面。我們會先為一個中等體型（M-Medium）男性和女性植入骨架，做為標的原型，而整個骨架建立的程序包括設計骨架、指定座標系統以及建立骨架系統與表面資料間的連動關係。在完成標的原型的建構之後，會更進一步地使用真實人體的表面資料來驗證標的原型的正確性和精確性並加以調整骨架模型，待驗證完成後再以類似的步驟套用至其他14個體型中。在這個骨架系統中包含了29個節點和7個終端點，並以多連桿的方式定義了35個細桿以及11個剛體，使用的座標系統為球座標系統，有（L, ψ, θ, α）4種自由度，整個資料結構起始於骨盆關節，並規劃了5個路徑，例如軀幹、2個上肢和2個下肢。並且為表面點指定相關節點的函數以建立表面資料與骨段間的連動關係。動畫人模建立之後，將其變換姿勢與其於8種不同姿式的3D人體掃描資料作疊合比較。比較結果顯示，所有人模中之最大疊合誤差範圍為25.2公厘至37.7公厘，平均疊合誤差範圍為4.32公厘至10.82公厘，於產品與工作場所設計應用而言，為相當精確而可接受的。之後，此系列的動態人模可用以模擬多種的工作姿勢，作為工作場所設計應用。

The objective of this study is to develop 3D animation human figures using an existing 3D anthropometrical database. A total of 15 figures of different body sizes, 5 heights and 3 BMIs, will be developed each for males and females. The existing 3D anthropometrical database, called 3D Human Body Bank, was measured using 3D body scanners. For both genders, we began with the establishment of the median body size one, called prototype model. The whole developing process consists of the design of skeletal system, the assignment of the coordinate systems and the establishment of the relationship between the surface and the skeletal system. After, the prototype model was further calibrated and validated on accuracy and precision using true scan data. Then the similar procedure was applied on the rest of 14 figures. The skeletal system consists of 29 rotation centers and 7 end nodes, organized in multi-linkage form of 35 stick-links, and 11 rigid bodies. The coordinate system is spherical which consists of 4 degrees of freedom, namely (L, ψ, θ, α). The data structure of the skeleton is originated in pelvis and organized in 5 paths, such as torso, 2 upper extremities, and 2 lower extremities. The relationship between the surface and the skeletal system is mainly established by assigned the surfaces points as the function of related nodes. Upon the construction of these digital models, they are manipulated to match and superimpose on those 8 different scanned postures. The results show that the maximum mismatch error for each digital model ranges between 25.2 mm and 37.7 mm, the averaged mismatch error ranges between 4.32 mm and 10.82 mm. With these maximum and averaged mismatch errors, they are quite promising for product and workplace design application. Later, this set of digital human models has been used to simulate many types of working postures for workplace design.