Paper 14.378

P. Costa et al., "3D Reconstruction of Body Parts Using RGB-D Sensors: Challenges from a Biomedical Perspective", in Proc. of 5th Int. Conf. on 3D Body Scanning Technologies, Lugano, Switzerland, 2014, pp. 378-389, doi:10.15221/14.378.


3D Reconstruction of Body Parts Using RGB-D Sensors: Challenges from a Biomedical Perspective


Pedro Costa 1,2, Hooshiar Zolfagharnasab 1,2, Joao P. Monteiro 2, Jaime S. Cardoso 1,2, Helder P. Oliveira 2

1 Faculdade de Engenharia, Universidade do Porto, Porto, Portugal;
2 INESC TEC, Porto, Portugal


The patient 3D model reconstruction plays an important role in applications such as surgery planning or computer-aided prosthesis design systems. Common methods use either expensive devices or require expert personnel which are not available in every clinic. Thus to make patient-specific modelling more versatile, it is required to develop efficient methods together with feasible devices. Body parts such as head and torso present valid challenges with different degrees of complexity, especially because of the absence of relevant and abundant features.
Considering Microsoft Kinect, it is a low-cost and widely available sensor, which has been successfully applied in medical applications. Since single depth-map acquired by Kinect is often incomplete and noisy, different approaches have been proposed to perform the reconstruction by merging multiple depth-maps, by registering single view point clouds generated form each point cloud. As human body is a non-rigid model, most of previous reconstruction methods using Kinect fail to perform accurate reconstruction since they do not address non-rigid surfaces.
In this paper we present the challenges of using low-cost RGB-D sensors to reconstruct human body. Additionally, we analysed coarse registration stage to understand its impact on the quality of reconstruction on both rigid and non-rigid data. Also comparative research has been performed to study different coarse registration methods such as Spin Image (SI), Curvedness, and Principal Component Analysis (PCA). Studies showed that the quality of reconstruction is directly related to robustness of reconstruction method to the rotational and translation noise. Regarding analytical comparisons, results indicate the positive impression of applying coarse registration on both rigid and non-rigid data. Moreover, evaluations show PCA presents better results among other considered methods. Finally it is shown that down-sampled models present less error.


Full paper: 14.378.pdf
Proceedings: 3DBST 2014, 21-22 Oct. 2014, Lugano, Switzerland
Pages: 378-389
DOI: 10.15221/14.378

Copyright notice

© Hometrica Consulting - Dr. Nicola D'Apuzzo, Switzerland,
Reproduction of the proceedings or any parts thereof (excluding short quotations for the use in the preparation of reviews and technical and scientific papers) may be made only after obtaining the specific approval of the publisher. The papers appearing in the proceedings reflect the author's opinions. Their inclusion in these publications does not necessary constitute endorsement by the editor or by the publisher. Authors retain all rights to individual papers.

Proceedings of 3DBODY.TECH International Conferences on 3D Body Scanning & Processing Technologies, © Hometrica Consulting, Switzerland