3D Facial Model Reconstruction By Object-Space Matching
近景攝影量測 ； 顏面三維模型 ； 物空間匹配 ； close-range photogrammetry ； three-dimensional facial models ； object-space matching
|Volume or Term/Year and Month of Publication||
|Academic Degree Category||
現今電腦硬體設備的快速發展，使得軟體處理速度以及效能都大有提升，在電腦上展示三維虛擬環境讓使用者能更直觀的接收視覺資訊，如果三維模型足夠精密，更可在軟體上測量，省時省力。三維立體模型的建構已應用在如醫療、假牙製作、整形手術及大型手術模擬等領域上，然而市面上關於醫療方面的測量儀器價格仍高。應用在醫療用途上之立體模型，精度應至少需達到公釐等級以下，在量測進行中還需顧及患者的感受，快速完成，並盡量不接觸到患者的皮膚表面。基於上述理由，建立可量測及預視的立體模型，提供醫師及患者作為術前參考是很有幫助且必要的。本研究以攝影測量技術，搭配一般市售平價之非量測型數位相機，以非接觸量測方式重建顏面模型，並檢驗精度評估應用於人體之可行性。 本研究主要分為四部分：(1)經設計實驗證明在結合適當演算法下，採用立體像對測量方式測量已知點位的位置及高程等物理量，本實驗架構之量測水準確實可達到公釐等級。(2)應用物空間匹配法(Object-Space Matching)，重新匹配產生顏面模型，與ATOS逆向工程系統製作之模型比對檢驗誤差，並以上述方法建立真人的顏面模型。(3)模型透過虛擬實境模組語言(Virtual Reality Modeling Language)展示。(4)實驗後半階段選擇10個顏面特徵標的(Landmark)進行測量，比較四種測量方法並統計每種測量方法之標準差，選擇最佳測量方法，實際測量並統計30位平均年齡22歲之成年男女性臉部線條長度與角度提供參考。
Rapid development of computer hardware has lead to enhanced efficiency and performance. Nowadays, three-dimensional virtual environments displayed on computers provide viewers more direct access to visual information. Three-dimensional models possessing high accuracy may enable users to conduct measurement procedures through software programs, saving them time and effort in comparison with conventional measuring processes. These types of three-dimensional models have been utilized for medical purposes, such as dentures (artificial teeth), plastic surgery and pre-operation simulation. However, such surveying instruments are costly. In addition, literature has suggested that stereo models used for medical purposes should possess at least 1 mm accuracy. A rapid surveying process which minimizes the contact with the skin is also a necessary feature to reduce discomfort of the applicant. Based on these requirements, establishing a stereo model that can be subsequently used for measurement and previewing through software is greatly helpful and necessary for pre-operation simulation. In this study, photogrammetric techniques were applied along with non-metric digital cameras to reconstruct the facial image of an alabaster sculpture via non-contact measuring methods. The procedure was then inspected for its applicability on human faces. This study consists of three parts. First, algorithms were applied with a stereo-pair measurement method to measure the locations and heights of existing dots. Experimental results confirmed that the measurement method achieved 1 mm accuracy. Secondly, the Object-Space Matching theory was applied for matching and construction of facial models. These models were then compared with models obtained by Reverse Engineering ATOS, and inspected for the errors. The above approach was applied to construct and display actual facial models through Virtual Reality Modeling Language (VRML). Finally, ten facial landmarks were selected and measured via four different methods. The standard deviations of each method were calculated in order to determine the best measurement method. This method was then applied to measure the facial length and angles of 30 adults (with an average age of 22); statistical results are provided as reference.
工程學 > 土木與建築工程
工程學 > 市政與環境工程