A WEIGHTED FACTORIZATION APPROACH FOR ARTICULATED MOTION MODELLING
João K. Fayad, Alessio Del Bue, Pedro M. Q. Aguiar · 2009
Keywords: Human Body modelling, Biomechanics, Structure from Motion, Factorization. Abstract. Human body models and their analysis are nowadays used in a wide range of applications, spanning from medicine to security and surveillance. In this work we focus on the automatic creation of biomechanical human models for clinical and sports applications. State of the art motion capture systems are nowadays able to measure the 3D coordinates of reflective markers placed above the skin with sufficient accuracy. Clinical gait analysis is currently facing problems like the repeatability of the measurements, the creation of subjectspecific models, and the compensation for soft-tissue movement. Recent developments on structure from motion approaches have allowed the efficient recovery of articulated bodies from a set of 2D images. In a similar mathematical framework, we present a set of computational tools able to create subject specific 3D models form motion capture data. Our aim is to obtain fully data-driven algorithms for the localization of the joints and their properties. Specifically, our goal is to obtain a 3D model of the body parts, which is able to deal with the soft-tissue artifacts that are very common in this type of inference. The proposed approach has the distinguishing feature of being able to deal with marker occlusions that may occur during the extraction of the measurements from the motion capture system. Finally, we use synthetic data to assess the performance of the algorithms and compare the results with ground truth data. Qualitative analysis of real data sequences are also presented. 1 J. Fayad, A. Del Bue, P. Aguiar 1