Facial feature point tracking algorithm based on eyes tracking and active appearance model step fitting

The facial feature point tracking algorithm by using Active Appearance Model (AAM) would cause tracking failure when face pose greatly deflected or initial position of facial AAM model deviated drastically from the target face. In order to solve the problem that the initial position of facial AAM mo...

Full description

Saved in:
Bibliographic Details
Main Authors: Rui, Zhong, Huai-yu, Wu, Ruo-hong, Wu
Format: Conference Proceeding
Language:English
Subjects:
Active appearance model
Algorithm design and analysis
Algorithms
Deflection
deflection model
Eyes
eyes tracking
Face
Facial
Fitting
Globes
Kalman filters
Mathematical models
Shape
step fitting
strong tracking Kalman filter
Support vector machines
Target tracking
Tracking
Tracking (position)
Online Access:Request full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:The facial feature point tracking algorithm by using Active Appearance Model (AAM) would cause tracking failure when face pose greatly deflected or initial position of facial AAM model deviated drastically from the target face. In order to solve the problem that the initial position of facial AAM model deviated drastically from the target face, this paper proposes an eyes-tracking algorithm by applying Adaboost to detect the face and predicts the position of eyes in next frame by using strong tracking Kalman filter algorithm. In the framework of multi-view AAM, this paper solves the problem of great face pose deflection by applying Support Vector Machine (SVM) to estimate the deflection angle of the present face and select the appropriate Multi-view AAM as the globe facial shape model. The globe facial shape model moved to the position of the eyes and the parameters of deflection model of face pose updated in real time. After completes the above steps, the algorithm would fit the eyes feature points firstly, then the residual facial components feature points would be fitted. The experimental results show that the proposed algorithm has good performance in the accuracy and efficiency of facial feature point tracking.
ISSN:2161-2927
1934-1768
DOI:10.1109/ChiCC.2015.7260211