You are on page 1of 5

e-ISSN (O): 2348-4470

Scientific Journal of Impact Factor (SJIF): 4.72


p-ISSN (P): 2348-6406

International Journal of Advance Engineering and Research


Development
Volume 4, Issue 2, February -2017

A Survey On Real Time People Counting Method With Head Localization And
Tracking System Consideration.
1
Mr. Ganesh S. Patel, 2Mr. Jaykumar S. Dhage
1
(Computer Science and Technology, Dr. BAM University/MITCollege, Aurangabad (MS)
2
(Dr.BAM University/MIT College, Aurangabad (MS)

Abstract the people counting method plays a vital role in marketing research like traffic management, tourists flow.
With the help of people counting system it is possible to collect the statistical information of the people present inside the
organizations environment, So that in the case of an emergency by referring the count of people, proper action can be
taken to perform rescue operation.
Initially people counting process was manually conducted by security guards but the limitation of this approach
is that guards may not have attention for 24*7 and required to pay monthly amount to them.
There are several techniques to implement people counting, It can be implemented using Vertical Kinect depth
sensor, blob descriptor, Gaussian mixture model GMM, K-Mean clustering approach, HOG feature and kalman filter. In
this paper we use different target taking after with direct nearest neighbor data relationship to Estimate people.

Keywords People counting, LSVM, Foreground and background segmentation.

I. INTRODUCTION

People counting is an important term in security providing organizations because the security system needs to take
care of people and originations property. Considering scope in security field lots of work is done in people counting
system. Today more research is published to resolve problem of people counting using video camera. People counting
help in to the security or people management. Several ways to implement people counting are as follows. In [5] to
implement people counting Matlab simulink programming tool is used and it tries to measure and compare its accuracy
with another system using laser beam sensor. Indoor scenes are more difficult than outdoor for counting people because
of slightly moving people and occlusion. For this propose [1] head-shoulder detection based crowd counting framework
for indoor scenes is used.

On the off chance that huge measure of individuals is goes ahead railroad station then they have to oversee high
stream rate. There are many problems like shadow elimination, dynamic background and also occlusion. Zhang et al [2]
use group tracking to solve the problem of multiple human segmentation, but dynamic background problem was not
solved. The surveillance application require for safety, site management, security etc. which include retail outlets, entry
points monitoring, ATM, parking. From last few years background subtraction algorithm [4] proposed using Gaussians
average, kalman filter. So for this purpose we can use GMM because of its robustness.

Detection counting approach has two unsolved problem the first one is different bounding boxes suppression if
people interact and second is some object occluded so they are difficult to find. Regression counting is difficult and time
consuming for large dataset to give labeling. In Zhang et al. [5] present an approach for counting people using Kinect
depth sensor. Kinect depth sensor uses structural light technique to generate image information. Real time people
counting [6] is based on blob feature and develop system for estimating "how many people are and where they are in
video sequences" in real time. The real time people counting reduce the human monitoring work. The depth imagery
helps to localize the head which are later track in video frames using superior method.

II. TECHNIQUES

A people counter is a gadget used to tally the quantity of people on foot strolling through an entryway or hall [1]. More
often than not, this framework is utilized at the passage of a assembling so that the aggregate number of guests can be
recorded.

@IJAERD-2017, All rights Reserved 218


International Journal of Advance Engineering and Research Development (IJAERD)
Volume 4, Issue 2, February -2017, e-ISSN: 2348 - 4470, print-ISSN: 2348-6406

CURRENT
FRAME TRACKING
BACKGROUND SEGMENTATION AND
ESTIMATION MODULE MODULE BLOB?
COUNTING
PREVIOUS MODULE
FRAME

STOP

Fig.1. Counting People Module

Fig. 1 is real time people counting video based on Matlab simulink programming tool. To implement this system each
frame is compared with previous frame using background estimation module and segmentation module. After
comparison if blobs are found then tracking and counting module is applied on frame, else counting process stops. This
method is used to implement people counting in indoor scenes. Previously [1] implemented methods not have good
performance because of occlusion and randomly changing indoor environment.

Fig. 2. Overall framework for people counting in indoor scenes.

Computer vision based techniques have been widely used in the area of people counting. Many problems
challenge the real application, such as occlusion, shadow elimination and dynamic background. [2] Many kinds of
approaches have been proposed to handle these problems. The approaches can be roughly divided into two categories,
one is based on background subtraction and the other is based on pedestrian detection and tracking.

Background subtraction based methods are famous because they are easy to implement and very fast. HOG features have
been extended a lot of improvement in the performance of people counting. HOG and LBP features are used to combine
improve detection accuracy but performance is less as compared to previous research.

The main problem in video detection system is presence of shadow. Because [4] moving shadow may treated as
object in video, so to have efficient people counting a robust real time people tracking and counting incorporating
shadow detection and removal is used.

@IJAERD-2017, All rights Reserved 219


219
International Journal of Advance Engineering and Research Development (IJAERD)
Volume 4, Issue 2, February -2017, e-ISSN: 2348 - 4470, print-ISSN: 2348-6406

Fig.3. System Model for the Proposed Counting method.

Fig.3. is proposed counting method of a robust real time people tracking and counting incorporating shadow
detection and removal. Captured or live video is taken as input, then moving object is detected using GMM then
HORPRASERT COLORMODEL is applied on detected object to remove shadow. Features are extracted and classify the
object either as a single or multiple objects. In the above work bi-directional people are counted. To describe the
background, GMM is used. In proposed method GMM is used to extract the foreground region and to classify the shadow
region from original a color model is used, then to remove noise morphological operation is applied. Some geometric
feature from detected object is extracted which will be used to tracking process.

This paper use input depth maps produced by kinect depth sensor in real world environment [3] and propose a simple
pre-processing to improve the quality of depth maps. For Head tracking base segmentation is used.

Fig. 4. is people counting system is based on a vertical kinect depth sensor. The kinect sensor uses structural light
technique to generate the calibrated depth information of the scene.

The outline of the proposed system is illustrated in fig.5. Frames are extracted from videos and taken as input.
Blobs of moving object are form using background subtraction [5] and shadow elimination. For each blob feature are
extracted and they are normalized according to its approximate size to its real size. Finally number of people counted in
each blob by using neural network.

@IJAERD-2017, All rights Reserved 220


220
International Journal of Advance Engineering and Research Development (IJAERD)
Volume 4, Issue 2, February -2017, e-ISSN: 2348 - 4470, print-ISSN: 2348-6406

Fig.5. Outline of the people counting system

The neural network is trained to established relation between the feature value and the number of pedestrians in a blob
refers to training data. In fig.5. Proposed system background subtraction is used for segmentation.

III. COMPARISON TABLE

Sr. no. Paper Description Research Gap


1. Real-time people In this propose a head-shoulder detection based The head-shoulder
counting for indoor crowd counting framework for indoor scenes. detection is the main
scenes People counting performed in indoor problem in this approach
environment because of strange clothes.

2. Real-time people It proposes a system for counting the number of Counting based on whole
counting using blob pedestrians in real-time. This system estimates body of pedestrians so
descriptor how many pedestrians are and where they are in many times it gives wrong
video sequences using blob descriptor. output.

3. Real-Time Depth Map The main idea of counting by detection is using Time consuming labeling
Based People Counting the multi-scale window slide over the image. To in regression if training
determine if there is a person within a window instance is large.
or not, a binary classifier is used.
4. A Detection And A head and shoulder shape hog feature is Performance is less
Tracking Based Method extracted to detect Pedestrians and particle compared to previous
For Real-time People filtering method with color based appearance research.
Counting model, is used to track pedestrians.

5. Counting people by It present a vision based method for counting Different installation
RGB or depth overhead the number of persons which cross a virtual requirements for large
cameras line. To determine the number of persons that dataset and At the doors of
cross the virtual line and providing the crossing buses, trains, metros
direction for each person performance is low.

6. A Robust Real Time To describe background scene Gaussian mixture It fails to classify the
People tracking and model is used. shadows from moving
Counting incorporating foreground objects.
shadow detection and
removal

V. CONCLUSION

The paper shows people counting method which increase performance of the crowd analysis. People counting helps to
monitoring the people in Public places, Metro Stations, ATM system etc. People counting reduce the burden on humans.

@IJAERD-2017, All rights Reserved 221


221
International Journal of Advance Engineering and Research Development (IJAERD)
Volume 4, Issue 2, February -2017, e-ISSN: 2348 - 4470, print-ISSN: 2348-6406

In this paper, the method uses depth information to accurately segment the foreground from the background. The Multi-
Target Tracker Monitors the in and out people. The Proposed system use head localization and people tracking
Techniques.

REFERENCES

1. Jun Luo, JinqiaoWang, HuazhongXu, HanqingLu Real-time people counting for indoor scenes. E-mail addresses:
junjing2218@gmail.com (J. Luo),jqwang@nlpr.ia.ac.cn (J. Wang), wutxhz@163.com (H. Xu),Accepted 30
October2015

2. Jingyu Liu Jiazheng Liu Mengyang Zhang. A Detection And Tracking Based Method For Real-time People Counting.
jyl_999@163.com978-1-4799-0333-7/13/$31.00 2013 IEEE

3. Frantisek Galck and Radoslav Gargalk .Real-Time Depth Map Based People Counting. J. Blanc-Talon et al. (Eds.):
ACIVS 2013, LNCS 8192, pp. 330341, 2013. Springer International Publishing Switzerland 2013.

4. J. L. Raheja, Sishir Kalita, Pallab Jyoti Dutta, Solanki Lovendra A Robust Real Time People tracking and Counting
incorporating shadow detection and removal International Journal of Computer Applications (0975 8887) Volume
46 No.4, May 2012

5. Satoshi Yoshinaga, Atsushi Shimada, Rinichiro Taniguchi Real-time people counting using blob descriptor Kyushu
University, 744, Motooka, Nishi-ku, Fukuoka, 819-0395 Japan Received November 13, 2009; revised December 4,
2009; accepted December 10, 2009

6. Supervisors: Jon Y. Hardeberg, Faouzi Alaya Cheikh, Pierre Gouton Real Time People Counting system using Video
Camera. Master of Computer Science, Image and Artificial Intelligence 2007

7. Jingyu Liu Jiazheng Liu Mengyang Zhang .Counting people by RGB or depth overhead cameras.jyl_999@163.com
978-1-4799-0333-7/13/$31.00 2013 IEEE

8. Nikolai Smolyanskiy , Christian Huitema, Lin Liang, Sean Eron Anderson Real-time 3D face tracking based on active
appearance model constrained by depth data0262-8856/ 2014 The Authors. Published by Elsevier 4 August 2014

9. Antoni B. Chan, Nuno Vasconcelos Counting People With Low-Level Features and Bayesian Regression IEEE
transactions on image processing, vol. 21, no. 4, april 2012

10. Ya-Li Hou, Grantham K. H. Pang, People Counting and Human Detection in a Challenging Situation ieee
transactions on systems, man, and cyberneticspart a: systems and humans, vol. 41, no. 1, january 2011

11. Thou-Ho (Chao-Ho) Chen, Che-Wei Hsu, An Automatic Bi-Directional Passing-People Counting Method Based on
Color Image Processing 0-7803-7882-2/03/$17.0002003 ieee

12. Ismail Haritaoglu, David Harwood,Larry S. Davis W4: Real-Time Surveillance of People and Their activities ieee
transactions on pattern analysis and machine intelligence, vol. 22, no. 8, august 2000

13. Jae Hoon Lee, Yong-Shik Kim , Bong Keun Kim and Kohtaro Ohba Hirohiko Kawata, Akihisa Ohya and Shinichi
Yuta. Security Door System Using Human Tracking Method with Laser Range Finders International Conference on
Mechatronics and Automation August 5 - 8, 2007, Harbin, China

@IJAERD-2017, All rights Reserved 222


222

You might also like