Date post: | 24-May-2015 |
Category: |
Technology |
Upload: | mahfuzul-haque |
View: | 40 times |
Download: | 0 times |
A Hybrid Object Detection Technique from Dynamic Background Using Gaussian Mixture Models Mahfuzul Haque, Manzur Murshed, and Manoranjan Paul
Gippsland School of Information Technology, Monash University, Victoria 3842, Australia Email: {Mahfuzul.Haque, Manzur.Murshed, Manoranjan.Paul}@infotech.monash.edu.au
The images shown in the header has been taken from http://www.informationliberation.com [1] Mahfuzul Haque, Manzur Murshed, and Manoranjan Paul, On Stable Dynamic Background Generation Technique using Gaussian Mixture Models for Robust Object
Detection, IEEE International Conference On Advanced Video and Signal Based Surveillance (AVSS), New Mexico, USA, 2008.
[2] Mahfuzul Haque, Manzur Murshed, and Manoranjan Paul, A Hybrid Object Detection Technique from Dynamic Background Using Gaussian Mixture Models, IEEE
International Workshop on Multimedia Signal Processing (MMSP), Cairns, Australia, 2008.
[3] Mahfuzul Haque, Manzur Murshed, and Manoranjan Paul, Improved Gaussian Mixtures for Robust Object Detection by Adaptive Multi-Background Generation,
International Conference on Pattern Recognition (ICPR), Tampa, Florida, USA, 2008.
The Complete System 8
Background Modelling
Frame 1
Frame 2
Frame t
..
Ki titittit XwXP
1 ,,, ),,()(
)()(2
1
2/12/,,
1
||)2(
1)(
ttT
tt XX
nttt eX
Gaussian Mixture Model (GMM)
for each pixel
Input scenes
A pixel model is constructed and updated for each pixel which
maintains a mixture of Gaussian distributions for modelling multi-modal
distribution caused by moving foregrounds and repetitive background
motions [1-3].
1 Abstract
Adaptive background modelling is essential for robust
object detection in real-world scenarios while basic
background subtraction offers improved stability with
varying operating environments [1]. This research
presents a new technique by exploiting the strengths of
both approaches, which not only reduces noise, shadow,
and trailing effects, but also maintains superior stability
across variable operating environments.
P(x)
Existing Models
Intensity
New Model
2
Qualitative Evaluation 7
Improved Model Quality by New Model
Induction Scheme 3
Model Quality Visualisation 4
Hybrid Detection Algorithm 5
Red pixels correspond shadows and reflections detected with
probabilistic subtraction, green pixels represent the low contrast regions
detected using basic background subtraction, while yellow pixels are
detected by both approaches. The hybrid algorithm utilises
neighbourhood statistics of both decisions to minimise the red regions
while maximising the green regions.
Quantitative Evaluation 6
Single Model
Double Models
Many Models
0 127 255 First
Frame
Test
Frame
Ideal
Result
Lee’s
Tech. Prop.
Tech.
Lee’s
Tech.
Prop.
Tech.
Comparisons of best error rates achieved by three techniques and
standard deviations of error rates across three learning rates.
Input Frames
Visualisation
Test Sequence S&G Lee Proposed S&G Lee Proposed
1. PETS2000 1.78 1.89 1.56 1.66 1.31 0.01
2. PETS2006-B1 5.50 8.22 3.89 3.20 1.18 0.52
3. PETS2006-B2 2.48 3.58 2.38 2.12 0.25 0.25
4. PETS2006-B3 2.46 3.42 2.36 2.66 1.08 0.31
5. PETS2006-B4 5.67 9.09 5.58 3.45 1.10 0.90
6. Bootstrap 11.83 13.26 11.80 1.89 2.14 1.26
7. Camouflage 13.01 25.46 10.08 10.99 9.64 2.32
8. Fground Aper. 16.31 52.39 15.74 21.65 7.39 0.03
9. LightSwitch 13.77 28.05 56.44 40.14 32.86 14.81
10. MovedObject 0.83 0.51 0.00 1.78 3.29 3.78
11. TimeOfDay 4.69 4.07 5.69 2.72 7.04 0.59
12. WavingTrees 13.68 18.38 12.91 1.79 0.45 0.09
13. Football 17.19 18.14 16.91 4.99 10.84 2.42
14. Walk 0.27 0.46 0.21 2.60 0.63 0.07
Error Rate (%) Stdev.