2018 International Conference On Advances in Communication and Computing Technology (ICACCT)

Amrutvahini College of Engineering, Sangamner, Ahmednagar, India. Feb 8-9, 2018

Moving Object Tracking using Kalman Filter

Mr. Pramod R. Gunjal Miss. Bhagyashri R. Gunjal Mr. Haribhau A. Shinde
Asst. Professor, Department of Electronics Department of Electronics Engineering, Asst. Professor, Department of Electronics
and Telecommunication Engineering, Amrutvahini College of Engineeirng, and Telecommunication Engineering,
Amrutvahini College of Engineeirng, Sangamner, SPPU, Pune, MH, India Amrutvahini College of Engineeirng,
Sangamner, SPPU, Pune, MH, India [email protected] Sangamner, SPPU, Pune, MH, India
[email protected] [email protected]
Mr. Swapnil M. Vanam Mr. Sachin S. Aher
Asst. Professor, Department of Electronics and Asst. Professor, Department of Electronics and
Telecommunication Engineering, Telecommunication Engineering,
Amrutvahini College of Engineeirng, Amrutvahini College of Engineeirng,
Sangamner, SPPU, Pune, MH, India Sangamner, SPPU, Pune, MH, India
[email protected] [email protected]

Abstract— In this paper we described a method for moving object of ground targets are used for airborne sensors. It has some
detection and tracking using Kalman filter. Basically, estimation advantages such as low visibility and target density is high.
process is very important in the surveillance system. This process Therefore, it has to improve the quality and continuity of
is for finding out the location of the target. The decomposition is tracking, as well as also used for the previous information
also helpful for the estimation process, in this process first step is
should be utilized as much as possible, which is highly
the tracking the video, and then the video is converted into
frames in the initialization period and every frame is made up of desirable. The research on the identification of moving targets
a piece of picture. In further step, the targets in each frame are are provides us with the information and the important for
identified by means of color recognition; next position is the implications for us to understand the treatment of video
moving target and to identify the center coordinates and next images deeply.
another last step the coordinate of the previous and current
frames is inputted and find out the location of the moving target
which is present frame. And this frame is estimated by filter. The II .LITERATURE REVIEW
tracking is very important for different object. The objects are
tracked with the help of Kalman filter. This filter is used for the
Surveillance systems are very most important for detection
pixel wise subtraction of current frame. As well as also used to be and tracking of the object. The first step is to detect the any
find out the error between actual position of the ball and object and second step is tracking object. In detection of
estimated position value with the help of this filter. moving object can be done by using the various methods.
Hong-Son was proposed the system which is depend on
Keywords— Object Detection, kalman filter, tracking object. pedestrian detection method [1]. They pointed out the
processing rate has improved detection performance.
Also, this method has highly potential as compared to other
I .INTRODUCTION method. In [2] Tested on designed Omni -directional robot
base and also present a robust real-time framework. This
Motion analysis is commonly based on a pixel per pixel
frame work used for moving object detection, on the base of
estimation of the instantaneous displacement of the underlying
estimation and prediction of trajectory. Sunil Kumar Vengalil
physical points. In wide area applications of moving target are
and Neelam Sinha proposed the system which is Based
used in many fields, like that used in computer vision with the
Texture Detection Algorithm [3]. They also focused on the
typically such as the video monitoring, image encoding, robot
types of the Texture. Sam Schauland, Joerg Velten, and Anton
technology, etc., there for it is applied in research for detection
Kummert were proposed digital filter [4]. They pointed out the
and tracking. The Moving object tracking are the tracking
separation of dimensional signals. Also get the good
based on Kalman algorithm. The Kalman algorithm means to
performance for motion-based object detection
predict and track moving targets, where to recognize the
In This paper we have focused on the types of methods. The
moving target is the first priority. The Ground surveillance
method is consisting of the first is temporal differencing. The
aims to determining the situation a large military ground with
temporal differencing has depended upon the basic principle
numerous targets as well as also determining the situation of
of pixel wise subtraction of current. It has disadvantages of it
weather and conditions for the daylight. The art sensor
is required to sensitive to dynamic change. Second detection
technology is used for the tracking. This technology is called
method is optical flow it has also required the large amount of
as airborne ground moving target indicator.
calculation. Next object tracking method is simple template
In this subject is providing high-quality tracking of military
matching is depending upon matching region of interest in
equipment, also providing the vehicles such as cars? The
video for algorithm. It has required equivalent model for each
tracking of target is used to support surveillance. The tracking
region of interest for each image.

978-1-5386-0926-2/18/$31.00 ©2018 IEEE 544

The tracking of the object has consisted of partical filter. The measurements. It has consisted of two types of the equations:
partical filter is depending upon the recursive filter algorithm. predict equations and correct equations. The predictions are
But it has computational timer to moderate to high. And it has responsible for project covariance system. It’s also called as
complex background also required the big calculations. in time. The important task of the correct update equations is
So Kalman filter are overcome the partical filter problems feedback. This feedback has considered the main important
and also used to Kalman filtering algorithm to track the object. task is to convert the new correct update equation in to the
judge the value. It has used the to get the new judge value in
the processes. Then last the find out the noisy value.
III .METHEDOLOGY
The Kalman filter is a set of mathematical equations which
has provides estimate the state of a process, in a way that
minimizes the mean of the squared error. The filter is very
powerful in several aspects: it supports estimations of past,
present, and even future states, and it can do so even when the
precise nature of the modeled system is unknown. It is very
important for first step is find out the detection.
Reorganization and tracking of the object sequence. Images
or video are firstly detected and track is very important and
risky tasks, to find out the object. Such as this is used in
surveillance system, military and automatic robotic system. It
has roll is to finding out the object which is present in the
frame of images or video. Next tracking methods are
consisting of the track the object for every frame. There for
the camera are used for the Object tracking process of locating
an object. If increasing availability for the high-powered
computers, then they also are increasing availability of high
quality .so by default increasing need for tracking algorithms.
Basically, tracking algorithms has consisted of three key steps.
First step is analysis of video, second step is detection of
moving objects; third step is tracking of objects which is
present in every frame. Therefore, the tracking is most
important tasks of the system which is nothing but the
recognition of motion based.

A. Kalman Filter
It is recursive estimator. The judge the value for the current
state, which is used for the current measurement. The type of
the filter is consisting of two variables, which is state time and
error covariance.
XK\K : estimate the state at time K
PK\K : error covariance matrix .

B. Kalman Filtering Algorithm

The kalman filter is estimation. The estimation processes have
Fig 1: Kalman filter prediction estimation cycle
consisted of important previous step & current measurement
step. Fig. 1 Shows the estimation cycle of kalman filter.
Basically, Kalman Filtering Algorithm is consisting of the two C. Filtering Problem
methods. Fist methods are time update and second method are
Kalman filtering is found out the problems of filtering and
measurement.
next to give the simple solution. Fig. 2 is shows the typical
The feedback controls are used for the estimation process. The
application of the Kalman filter. A physical system has
process has consisted of the state at some time for the time
consisted of the mobile robot, chemical process, satellite. The
updates and then last step is to get the feedback, which is in
system has given the two inputs, first input is control and
the form of corrected update. The time updates are nothing but
another is system error sources. The system state is the output
the prediction and measurement update nothing but the
which is nothing but the desired. The system state and
correction. These measurements are also called as the noisy
measurements error sources have two inputs are given to the

545
 measuring devices. The measuring devices have obtained the
observed measurements. Finally, Kalman filter are obtained
the optimal estimate of the system state.

Fig 4: Experimental simulation snapshot

Above Fig. 4 shows the experimental results. The X axis shows
the horizontal & y axis shows vertical position. The ball moving
from top to bottom which is tracked by using kalman filter. By
Fig 2: Typical application of the Kalman Filter using background subtraction methods to find out the contrast
between ball and floor.
So, this is used of the filter. This filter is useful in the When ball is detected then kalman filter shows it’s predicts
navigation, bus stop station, railway station, military are etc. current state.
The general filtering problem has used for following When ball is missing then kalman filter solely relies on the
Formulate, consider previous state and predicts the ball current location. So find out
the actual position of the ball.
x (k + 1) = f(x(k), u(k), w(k))

y (k) = h(x(k), v(k)) V. CONCLUSION

We have proposed an algorithm for segmentation and
detecting an object by using Kalman filters. It is observed that
IV.RESULTS less the error good are the results and we also observed that
when we achieve less error value, very fine object is detected.
The Kalman filter is used for prediction and filtering
Algorithm. The additional conditions are included on system;
this system is nothing but the dynamics system. The Kalman
filter is also called as the steady-state filter. So, to find out the
error between actual position of the ball and estimated position
value with the help of this filter.
When ball is missing then kalman filter solely relies on the
previous state and predicts the ball current location. So, find
out the actual position of the ball.
This proposed system is used for the navigation for security
system.

VI. REFERENCES

Fig 3: Experimental simulation snapshot [1] Hong-Son Vu, Jia-Xian Guo, Kuan-Hung Chen, Shu-Jui Hsieh, and
De-Sheng Chen.“A Real-Time Moving Objects Detection and
Classification Approach for Static Cameras”,978-1-5090-2073-
7/16/$31.00 ©2016 IEEE
Above Fig. 3 shows the input ball moving from top to bottom
[2] Hao Sun, Zehui Meng, Xiaotong Shen, and Marcelo H. Ang Jr. “
on the floor. In simulation snapshot the red colour are indicates Detection and State Estimation of Moving Objects on a Moving Base for
the pixels, which are detected by using foreground detector Indoor Navigation” , 978 1-5090-3329- 4/16/$31.00 ©2016 IEEE
methods. [3] Tao Zhou1, Harish Bhaskar, “Online Learning Of Multi-Feature
Weights For Robust Object Tracking”, 978-1-4799-8339-1/15/$31.00
©2015 IEEE.

546
[4] Sam Schauland, JoergVelten, Anton Kummert “Motion-Based Object [12] K S Chidanand Kumar “Phase-Edge Based Approach For Pedestrian
Detection Using 3d Wave Digital Filters”, 978-1-4244-2358- Segmentation Using Nir Camera And Tracking For Driver Ass”, 978-1-
3/08/$20.00 © 2008 IEEE 4673-6101-9/13/$31.00 ©2013 IEEE
[5] Zhiguo Song, JifengSun,Wanyi “Object Tracking Via Sparse [13] Chung-Ching Lin, et. al, “Dynamic Multi-Vehicle Detection And
Representation Of Dct Features ” , 2015IEEE Tracking From A Moving Platform”, 978-0-7695-4990-3/13 $26.00
[6] Mohamad Mahdi Saemi, John See , “Lost And Found: Identifying ©2013 IEEE, DOI: 10.1109/ Cvprw.2013.117.
Objects In Long-Term Surveillance Videos”, 978-1-4799-8996- [14] Lin Yang*, “Prediction Based Collaborative Trackers (Pct): A Robust
6/15/$31.00 _C 2015 IEEE And Accurate Approach Toward 3d Medical Object Tracking”, IEEE
[7] Sunil Kumar Vengalil “Modified Oriented Gaussian Derivative Filter Transactions On Medical Imaging, Vol. 30, No. 11, November 2011
Based Texture Detection Algorithm And Parameter Estimation”, 978-1- [15] M. Cossalter, M. Tagliasacchi, G. Valenzise” Privacy-Enabled Object
4799-9985-9/15/$31.00 ©2015 IEEE. Tracking in Video Sequences Using Compressive Sensing’’, 978-0-
[8] Jega Anish Dev “Human Computer Interaction Advancement By Usage 7695-3718-4/09 $25.00 © 2009 IEEE, DOI: 10.1109/Avss.2009.
Of Smart Phones For Motion Tracking And Remote Operation” , 978-1- [16] C. Hue et.al, “tracking multiple objects with particle filtering”, IEEE
4799-7646-1/14 $31.00 © 2014 IEEE transactions on aerospace and electronic systems vol. 38, no. 3 july 2002
[9] Yang Lu, Tianfu WuAnd Song-Chun Zhu , “Online Object Tracking, [17] Boris Babenko “Robust Object Tracking with Online Multiple Instance
Learning And Parsing With And-Or Graphs”, 1063-6919/14 $31.00 © Learning”, Senior Member, Ieee And Serge Belongie, Member, IEEE.
2014 IEEE Do 10.1109/Cvpr.2014.443 [18] Yang Hua KarteekAlahari Cordelia Schmid, “Online Object Tracking
[10] El-Sayed Abdoul-Moaty, et.al, “A Particle Filter For Mutistatic Radar With Proposal Selection”, 1550-5499/15 $31.00 © 2015 IEEE Doi
Tracking”, 978-1-4799-1622-1/14/$31.00 ©2014 IEEE. 10.1109/Iccv.2015
[11] Abhishek Kumar Chauhan “Moving Objec Tracking Using Gaussian
Mixture Model and Optical Flow” © 2013, Ijarcsse.

547