Πλοήγηση ανά Συγγραφέας "Mbatha, Nelson-Stefanos"

Τώρα δείχνει 1 - 1 of 1
  • Μικρογραφία εικόνας
    Τεκμήριο
    Model-based 3D hand tracking using the lighting information as a visual cue.
    (Τ.Ε.Ι. Κρήτης, Σχολή Τεχνολογικών Εφαρμογών (Σ.Τ.Εφ), ΔΠΜΣ Προηγμένα Συστήματα Παραγωγής, Αυτοματισμού και Ρομποτικής, 2017-09-13) Mbatha, Nelson-Stefanos; Μπάτα, Νέλσον-Στέφανος
    The goal of this thesis is to exploit lighting as a visual cue for the development of a 3D hand tracking system. The proposed method follows the hypothesize and testing paradigm. Under this paradigm, the hand pose estimation boils down to comparing the RGB hand observation with several rendered hand hypotheses. An objective function calculates the discrepancy between the real hand pose and a rendered hand pose. In this thesis, we propose to use lighting information and a simple RGB sensor to build the objective function. This approach compares favorably with previous state of the art methods that require depth sensors and RGB-D observations to build the objective function. Under the assumption of a known single light source we are able to photo-realistically render the hand model. To do so we use the principles of the Phong reflection model usually used in computer graphics to develop a lighting model. Also, an object detection technique able to extract the foreground information of the RGB images is developed. Several objective functions with different properties are created and tested. The experiments showed which lighting objective functions are the most efficient and that under certain lighting conditions the use of the lighting information can be used to build a robust 3D hand tracking system which overcomes depth ambiguities in complex cases were the real hand pose either rotates or the fingers are bending.