Loading...
Please wait, while we are loading the content...
Similar Documents
D Description of the Human Body Shape Using Karhunen-Loève Expansion
| Content Provider | Semantic Scholar |
|---|---|
| Author | Azouz, Zouhour Ben Rioux, Marc Lepage, Richard |
| Copyright Year | 2002 |
| Abstract | The Karhunen-Loève expansion is used for a compact 3D description of the human body shape. In this paper we show that a compact 3D description can be achieved using a small set of eigenvectors. Experimental results using a database of three-dimensional body scans are presented. Applications are also discussed. 1.Introduction The measurement of the human body (Anthropometry) is an essential part of the engineering design of cars, aircraft, workspaces, and clothing, to name a few. Traditionally such measurements involve the linear distances between anatomic landmarks and the circumferential values at predefined locations. Practically, such measurements are limited to a set of about 100 values. There are, however, problems inherent with the use of traditional anthropometry as outlined in Robinette et. al [1]. First, the set of collected values is so small that the only valid attempt to reconstruct the original shape is limited to very simple geometries. Moreover, in most cases the set of unconnected (unregistered) values does not provide an accurate reconstruction of the original subject. Reconstruction ambiguities are due to the lack of a coordinate reference system. Furthermore, there can be large differences in measurement values between observers who collect the data even if they use the same measurement protocol. Full body 3D digitizing, a recently developed technology as outlined in Robinette et. al [1], allows one to increase the number of measurements from a hundred to a million, in ten’s of seconds and thus provides a better description of the human body surface. Furthermore each value in the data set is related to a common coordinate system, allowing an accurate and detailed 3D reconstruction. Int. J. of Information Technology Vol 8, No. 2 CAESAR project (Civilian American and European Surface Anthropometry Resource) as outlined in Robinette et al [2], is the first three-dimensional surface anthropometry survey performed in both U.S. and Europe. During this project, body measurements have been taken for people between the ages of 18 and 65 in three countries: U.S, Netherlands, and Italy. Besides traditional measurement, full body 3-D scans have been recorded in three postures as shown in figure 1. Two scanners were used to record full body 3-D scans: a Cyberware WB4 scanner[3] in the United States and Italy, and a Vitronic scanner [4] in the Netherlands. Figure 1. The three CAESAR postures The CAESAR project is a collaborative effort with partners from several countries such as the Air Force Research Laboratory (AFRL) in Ohio, the Society of Automotive Engineers (SAE), the Netherlands Organization of the Applied Scientific Research (TNO) and the National Research Council of Canada (NRC). After collecting a large number a body scans, the challenge is to analyze the human body shape variability using the provided 3-D database in order to improve the sizing strategy in the design of many products. To achieve this goal, a compact description of the human body shape is required. Even though a 3D scan provides a good description of the body surface, it represents a large amount of data (thousands of polygons), which is not convenient to study the variability of the human body shape. The description should also provide a faithful reconstruction of the original shape; otherwise it will not be more valuable than traditional measurements. This paper proposes a technique to achieve this goal. Among the methods proposed in the literature of 3-D objects modeling, three-dimensional implicit functions such as algebraic functions, superquadricss and hyperquadrics provide a compact description. Algebraic functions used by Sullivan et. al [5] and Keren et. al [6] require high order terms and several constraints to represent significant shapes. Superquadrics introduced by Pentland [7] and modified by Barr [8] and Solina et. al [9] are a family of parametric shapes which are extensions of ellipsoids. Hyperquadrics proposed by Hanson [10], are volumetric shape representation. Even though they have more representation power than superquadrics, hyperquadrics have a difficulty in representing objects with concavities. Ohushi et.al [11] developed the extended hyperquadrics to overcome this problem. However it has been demonstrated that the presented implicit functions are not adequate to represent complex objects such as the human body. |
| File Format | PDF HTM / HTML |
| Alternate Webpage(s) | http://www.icis.ntu.edu.sg/scs-ijit/82/82-3.pdf |
| Language | English |
| Access Restriction | Open |
| Content Type | Text |
| Resource Type | Article |
