A Journal of Software Engineering and Applications, 2013, 6, 1-5
doi:10.4236/j s ea.2 013.65B001 Published Online May 2013 (http://www.scirp.org/journal/jsea)
Copyright © 2013 SciRes. JSEA
1
An Adaptive P art i cle F i lt er Based Method for Real Time
Face Tracki ng
Wei-Ming Chen, Yi-Lung Lin, Ya-Hsiung Hsieh
Institute of Computer Science and Information Engineering, National Ilan University, Chinese Taipei.
Email: wmchen@niu.edu.tw, r9943013@niu.edu.tw, r0143002@niu.edu.tw
Received 2013
ABSTRACT
The video surveillance systems of recent years, usually major focus on the Human-Face of observation and detection.
Human-Face is the most characteristic and prominent feature of a human, therefore, detection and tracking of Hu-
man-Face has beco me an important indicator of the study. This paper discusses video surveillance of public places and
maj ors in automated face detection and face tracking. The main detection method is the use of Haar-Like Feature-based
and through the Cascade classifier of the Adaboost face detection. In the tracking mechanism is based on particle filter
and we modified SURF (Speeded up Robust Features) particle filter tracking, and thus enhance the detection and tra ck-
ing accuracy.
Keywords: Particle Filte r; Adaboost; Haar-Like Feature; SURF
1. Introduction
As the information technology made progress and the
hardware has been improved in recent years, the calcula-
tion speed is much higher and the application is more
diverse. In the image-processing field, it is easier to get
images with better quality thru current hardware device
and we can easily do analysis and process via images.
Plus that there are a verity of multi-media products, such
as digital camera, digital video, 3D digital TV, digital
monitoring system, HMI, image-processing software,
etc…, which makes image-processing technology more
popular and important. The application of human-face
detection/recognition and human-face tracking in mul-
ti-media field is getting more diverse. For example, the
auto-focus of human face inside digital camera can let
people take pictures with better quality. The function of
human-face re c ogni t io n in mo ni to r in g s ys te m o r ent rance
security indicates the importance of human face tracking
and recognition in image-processing technology.
The technology of human face recognition is to judge
whether there exists human-faces or not inside a picture
or video. If there exists a human face, it will locate the
size and position of human face by the algorithm of hu-
man-face location. Meanwhile, it will predict the p ossible
position of next human-face by tracking technology and
associated in formation.
2. Skin Color Model Face Detection
The technology of image tracking has wide applicatio ns
in man y fie lds, suc h as monit oring s ystem, r obot naviga-
tion, missile tracking etc… You can find image tracking
in these fields. The most common application upon im-
age tracking is Kalman Filter. Recent years, there exists a
tracking technology which can perform more efficiently
& accurately and also can deal with more complicated
environment & nonlinear Gaussian Model. This tech-
nology is called Particle Filter, which is a latest tracking
method.
The direction of this thesis is the technology of hu-
man-face detection & tracking. The technology of hu-
man-face detection is to judge whether there exists any
human-face in a picture or video sequence and to locate
the size & position of human-face. The general human-
face detection technology is to do detection & tracking
based on the model of skin colour, but this method will
be affected by illuminations easily. This thesis uses the
method of Adaboost [1] to do human face detection. In
human-face tracking, we improve Particle Filter to do
human-face tracking, in order to achieve a human-face
tracking system with better efficiency and accuracy.
3. Particle Filter
We usually use Kalman Filter to track targets under com-
plicated environment, but the target must follow Gaus-
sian distribution by using Kalman Filter. Ho wever, there
are many reasons to cause the existence of nonlinear and
non Ga uss ia n d ist ri b utio n i n r ea lit y, suc h a s j u mpi ng c ha ng e
of tracking tar get. This will lead to a bad trac king result.
A Adaptive Particle Filter Based Method for Real Time Face Tracking
Copyright © 2013 SciRes. JSEA
2
Particle Filter can solve the algorithm of nonlinear and
non-Gaussian problems. The distribution of post proba-
bility is a group of discr e te samples with weighing factor.
During the process of target tracking, it will do the sam-
pling for each frame of imaging sequence. Then, do the
prediction of particle, then weigh and output, so as to
predict the position of tracki ng target.
There are three steps of Particle Filter, i.e. sampling,
deliver prediction, and measurement.
3.1. Sampl ing
In the sampling process, the particle groups with high
weighing factor could be our tracking target. The higher
the weighing factor is, the more similar to our tracking
target the particle is. Then, we will delete the particles
with smaller particles.
3.2. Prediction
After finishing the sample process, the particle groups
with smaller weighing factor are deleted. In order to
make sure the uniformity of each particle number, the
particle groups with higher weighing factor will be kept.
We will also make extensions on kept particles and
transfer in different directions & speeds to make predic-
tions, in order to make sure the uniformity of possibility
on each particle.
3.3. Measurement
The last step is to finish the prediction. We use dynamic
models to predict the positions of tracking target. The
new weighing factor of particle group will be obtained by
applying a measuring model with high accur acy.
4. System Architecture
Here are the system structures of human-face detection &
tracking methods applie d in this thesis. T he f irst step is to
get the image sequences of video frames. The second is
to use the received image sequences to do some pre-process,
in benefit to the po st calculation of human-face detection
and tracking. The third step is to do human-face detec-
tion by detector, then find the related information of hu-
man-face position. (Figure 1)
4.1. The Model-Measurement Improvement of
Particle Filter
Particle filter usually uses skin color model to do hu-
man-face tracking. If human-face is shaded or the illu-
mination changes severely, the increased tracking devia-
tion will cause the tracking failure. So, we propose an
improvement o f particle filter tr acking model, in ord er to
locate the position of human face and calculate the best
window of human face in position, size and angle.
Process of model-measurement
First, use skin color model to estimate human-face posi-
tion, then judge whether there exists any human face in
video frame. Meanwhile, we will collect the samples of
particle filter. If the collected sample of human-face is
lower than 30%, it means to use skin-color as human-
face tracking fails. Then, it will activate human-face de-
tection mechanism to locate human-face. If the collected
sample of human-face is higher than 50%, it means to
use skin-color as human-face tracking works well. If the
collected sample of human-face is between 30% and
50%, SURF mechanism will be activated to update the
position of human-face. (Figure 2)
Figure 1 . Syst e m archi tec t ur e.
Figure 2. Process of model-meas u rement.
A Adaptive Particle Filter Based Method for Real Time Face Tracking
Copyright © 2013 SciRes. JSEA
3
In the step of Speeded up Robust Feature modified
particle model, we will duplicate the human-face de-
tected by Adaboost, and then receive the information of
width & height of Face ROI. In the step of human-face
trakcing, we can keep using the information of Face ROI
continuously until the human-face vanishes, if the track-
ing target of human-face does not change. In order to
speed up the real-time calculation speed of SURF, we
use the received width and height information of Face
ROI to locate the center of human-face target, expand
twice the size as Face ROI searching area, in order to
lower the calculation of SURF and rise the detection rate.
4.2. System Flo wch art
Firs t, u se v i s io n device to ge t t he i ma ge s eq ue nce. In o r d er
to increase the detection speed in human-face tracking,
we use edge-detection rate to get rid of so me Feature Set
not belongs to Adaboost. By using human-face detection
to locate human-face, we will copy the ROI, and then go
to tracking step. While particle filter makes sampling, it
will remove some confirmed non-tracking targets. Then,
re-predict the possible area of human-face. In the mea-
suring model, we improve original Skin Color Model b y
adding SURF to do careful human-face prediction upon
measuring model. (Figure 3)
Figure 3. System flowchart .
A Adaptive Particle Filter Based Method for Real Time Face Tracking
Copyright © 2013 SciRes. JSEA
4
5. Experimental Result
B y usi n g i ma ge seque nc e o f e x pe ri mental te st, t hi s chapter
compares Skin Color Model, here called Original Method,
and our improved par tic le filter in efficienc y. We use two
image sequences to compare the results of human-face
detection & tracking.
5.1. Translational Motion
Image sequence of translational motion : under a com-
plicated environment to monitor human-face to go from
left to right then out of monitor area, we use this se-
quence to test Skin Color Model and improved particle
filter added with SURF, then to compare the strength of
these two methods in tracking algorithm. This video has
226 frames in about 7 seconds, moving from left to right
until out-of monitor area.
By the experimental statistics of translation motion
image sequence to compare Skin Color Model and im-
proved particle filter, there exists a big vibratio n in Orig-
inal human-face detection. It means that it does not work
well i n human -face tracking. By using our improved par-
ticle filter to tr a ck human-face, the tracking result is good
and it can cope with human-face planar motion & detec-
tion of human-face out-of monitor area. (Table 1)
Using Skin Color Model to do average human-face
detection, the average detection rate is 59.97%; using our
improved particle filter method to track human face, the
average detection rate of human-face is 82.31%. (Figure
4 and Table 2)
6. Discussed and Future Work
This thesis provides a detection & tracking system of
human-face. It uses Adaboost to do human-face tracking
upon video frames, and then use SURF to extract &
match features, to help particle filter do tracking and in-
crease the accuracy of matching results in particle filter.
Traditional tracking system can only track object in con-
stant movement. While tracking object is shaded or sud-
denly stops or change moving direction, it will increase
the difficulty of tracking. By combining particle filter
and SURF, it will impro ve t he o b j ec t tr a cki ng u nd er non -
constant motion, non-linear, and non-Gaussian environ-
ment .
By using our proposed method, it can do the human-
face tracking more accurately and efficiently than tradi-
tional human-face tracking. By improving particle filter.
In the future, we will expand the applications of our
human-face detection & tracking system. We will com-
bine the matching of facial features to do fast tracking of
human-face position in image sequence, then calculate
the distance between features, inclined angles, and other
Table 1. Detect ion rate of human-face.
Frame Original Our
1 71.58% 97.60%
6 33.26% 90.49%
9 84.78% 96.05%
12 65.41% 87.00%
15 2.67% 90.51%
20 83.25% 9 0.46%
24 73.66% 88.75%
28 57.16% 9 3.07%
32 43.47% 91.74%
38 37.00% 8 4.09%
44 76.34% 74.47%
137 35.95% 77.05%
141 68.47% 76.29%
144 51.04% 70.54%
154 69.92% 88.87%
157 37.60% 79.75%
160 64.80% 77.90%
163 35.06% 76.57%
167 81.82% 79.78%
170 72.15% 78.70%
174 80.94% 77.17%
180 97.46% 74.77%
187 56.56% 76.50%
190 60.06% 83.42%
197 57.12% 74.78%
205 50.53% 82.14%
208 74.85% 75.21%
215 70.11% 79.65%
220 50.39% 80.02%
226 55.66% 75.98%
Figure 4. Translational motion detection rate of human-
face.
A Adaptive Particle Filter Based Method for Real Time Face Tracking
Copyright © 2013 SciRes. JSEA
5
Table 2. Trans lati onal motion Ave rage A c curacy Rate.
Average Accuracy Rate
Original 59.97%
Our 82.31%
related information. By extracting these information of
human-face, we will increase more interactive function
in later extended situation, in order to reach the interac-
tive system of HMI.
REFERENCES
[1] R. L. Hsu, M. Abode-Mottaleb and A. K. Jain, Face
Detection in Color Images,IEEE Transactions on Pat-
tern Analysis and Machine Intellgence, Vol. 24, No. 5,
2002.
[2] Viola and P. Jones, “Rapid Object Detection Using a
Boosted Cascade o f Simple Features ,” Proceedings of the
2001 IEEE Computer Society Conference, Vol. 1, 2001,
pp. 1511- 1518.
[3] Viola and P. Jones, Robust Real-Time Face Detection,”
International Journal of Computer Vision archive, Vol.
57, No. 2, 20 04 .
[4] Rainer Lienhart and Jochen Maydt, An Extended Set of
Haar-like Features for Rapid Object Detection,” Image
Processing , Vol. 1, 2002, pp. 900-903.
doi: 10.1109/ICIP.2002.1038171
[5] Rainer Lienhart, Alexander Kuranov and Vadim Pisa-
revsky, Empirical Analysis of Detection Cascades of
Boosted Classifiers for Rapid Object Detection,” MRL
Techni c al R e por t, 2002.
[6] Giusto, D. D. Massidda and F. Perra, “A Fast Algorithm
for Video Segmentation and Object Tracking,” Digital
Signal Processing, Vol. 2, 2002, pp. 697-700.
[7] R. E. Kalman, “A New Approach to Linear Filtering and
Prediction Problems,” Transactions of the ASME Journal
of Basic Engineering, Vol . 82, 1960, pp. 35-45.
doi: 10.1115/1.3662552
[8] G. Welch an d G. Bishop, “An Introduction to the Kalman
Filter,” Department of Computer Science University of
North Carolina at Chapel Hill, 1997, pp. 27599-3175.
[9] R. Lienhart, A. Kuranov and V. Pisarevsky, “Empirical
Analysis o f Det ectio n Cascad es of Bo ost ed C lassi fier s fo r
Rapid Object Detecti on,” MRL Tec hni c al R e por t, 2002 .
[10] N. J. Gordon , D. J. Salmond and A. F. M. Smith, “Novel
Approach to Nonlinear/non-Gaussian Bayesian State Es-
timation,” IEEE Proceedings F Radar and Signal
Processing, Vol. 12, No. 2, 1993, pp. 107-113.
[11] D. G. Lowe, Object Recognition from Local
Scale-invariant Features,International Conference of
Compu t er Vision, Vol. 60, 1999, pp. 1150-1157.
[12] H. Bay, T. Tuytelaars, and L. Van Gool, “SURF: Speeded
up Robust Features,” Proceedings of European Confe-
rence on Computer Vision, 2006, pp. 404-417.
[13] L. Juan and O. Gwun, A Comparison of SIFT,
PCA-SIFT and SURF,” International Journal of Image
Processing , Vol. 3, 2009, pp. 143-152.