This paper presents a new algorithm for 3D face tracking intended for clinical infant pain monitoring. The algorithm uses a cylinder head model and 3D head pose recovery by alignment of dynamically extracted templates based on dense-HOG features. Drift reduction is obtained from re-registration in combination with multi-pose state estimation by a square-root unscented Kalman filter. Results on videos of moving infants in hospital show good tracking for poses up to 50 degrees from upright-frontal, with mean eyelocation error relative to inter-ocular distance below 9%.
|Status||Gepubliceerd - 2016|
|Evenement||The Netherlands Conference on Computer Vision (NCCV 2016) - Lunteren, Nederland|
Duur: 12 dec 2016 → 13 dec 2016
|Congres||The Netherlands Conference on Computer Vision (NCCV 2016)|
|Verkorte titel||NCCV 2016|
|Periode||12/12/16 → 13/12/16|