You are on page 1of 5

2015 International Conference on Advances in Computer Engineering and Applications (ICACEA)

IMS Engineering College, Ghaziabad, India

Survey Paper on Various Techniques of Recognition


and Tracking
Namrata Jaiswal Vishan Kumar Gupta Asmita Mishra
M.Tech, Computer Science and Department of Computer Science M.Tech, Computer Science and
Engineering, IMS Engineering and Engineering, IMS Engineering Engineering, IMS Engineering
college Ghaziabad,India. College, Ghaziabad, India college Ghaziabad,India.
namrata01nov@gmail.com vishangupta@gmail.com asmitamishra22@gmail.com

Abstract- This paper describes different methods used for


identification and tracking of player and license plate. Although of the same kind(homogeneous) regions in order to separate
object (Player and license plate) recognition have projected many the number of uniform from the background. Digital
challenges because the algorithms did not give correct results. characters Size and pipe like attribute are used to filter the
But time to time, several techniques are discovered to identify the players. The algorithm mentioned by Viola and Jones is for
object. In this survey paper, we describe and compare many face identification. There are two stage approaches for number
techniques of identification through table and bar graph. recognition. The first stage is to grab and selection of image
corners by Harris detection method while second stage is to
Keywords- character recognizer, license plate recognition, plate
region extraction, segmentation, localization, vehicle number plate, extraction of maximally stable external regions. Both of these
vehicle plate extraction. steps segment the image to process by OCR software.
Real time player recognition and tracking is difficult
I. INTRODUCTION because of the below reasons-
The Player recognition is to identify the individual team 1. The player size sometimes is just about 15-30 pixels.
at real time during the match in order to know particular
player performance. Identifying the player, tracking the 2. There are very often occlusions between the players.
player, counting the total team has advantage in many fields. 3. The illumination conditions vary during the game and even
For example to know about the current activity of the player. for each frame, when stadium shadows are considered.
The evolution of team recognition at live time is moving
towards a future of endless possibilities. The basic idea behind 4. Player shapes and colors tend to be blurred and non
this application is image processing of individual. Image distinctive, due to low resolution, motion blurring or image
processing manipulated the HSV characteristics of the player compression artifacts.
jersey, the jersey number to different them. The image To solve the player recognition task despite these
processing technique Morphological operation is to isolate the complications, we combine simple but powerful perceptual
number mentioned on jersey from captured image and process cues in a probabilistic manner, in order to achieve robustness,
further. In this survey paper, Map detection system will be speed as well as the required precision
used. It will different the uniform from whole image and then
segment the image for further process to guess the number of All potential player regions unlikely to belong to either
a player. the field or the surroundings are segmented .Likelihood-maps
The input of the title project is image, data of the team are generated by template-matching using color, compactness
name and uniform type while the output is the image along and height constraints. Hypotheses given by the prediction
with player name and number. Automatic Player recognition module of the Tracker are fused together with template-maps.
would be of great importance in game video. Video image is Players are localized using their estimated center of gravity
actually classified into caption text which is scene text and their uncertainties are estimated from the resultant maps
available already in the image. Team player number is the in image coordinates.
scene text. Because of the different orientation and angles the Following techniques are applied to increase robustness and
text detection is often tough task to find exact result. accuracy:
Several methods are used to find the better output. As 1. Spatial variance filtering and probabilistic color-models
mentioned earlier that image is segmented by vector (mixture of Gaussians) are used to segment player regions.
quantization algorithm that surrounds pixels in to limited color 2. Predicted camera parameters are used to determine
thresholding values and context constraints such as expected
978-1-4673-6911-4/15/$31.00©2015 IEEE player size.

921

Authorized licensed use limited to: Malaviya National Institute of Technology Jaipur. Downloaded on May 18,2023 at 12:09:43 UTC from IEEE Xplore. Restrictions apply.
2015 International Conference on Advances in Computer Engineering and Applications (ICACEA)
IMS Engineering College, Ghaziabad, India

3. The Mahalanobis distance to probabilistic color-classes is player number localization and recognition .First uniform area
used in color-template matching. were extracted by region growing algorithm.HSV color space
We analyze segmentation which defines the process. is used for segmented the region and uniform and number
Second process is to define player extraction using internal region. The main method was used for players number
contours in processed image. And finally identify object and localization was internal contour. It removed the false number
players. of players on the region and aspect ratio and took enhance
image for smoothing and rotation normal.
II. LITERATURE SURVEY [5] Introduced modern methodology for development of
[1] Presented new algorithm for Vehicle recognition an Automatic Vehicle Identification System(AVIS).The
which was an effective technique of image processing device could be divided in three main parts: vehicle image
technique. It was used to recognize vehicles plate numbers. processing, license Plate feature extraction, Hidden Markov
Many application were used in many field which is based on Model(HMM).It gave identification rate 98%.Hidden Morkov
this technique. This technique was used in many region model was a simple foundation for making probabilistic
traffic, travel, marketing tool, border control etc .Many model of linear serial ‘labeling problems .It gave toolkit for
method were used to explain this technique .There were two building hard models just by gene finding, profile searches,
techniques used in this paper edge detection method and many sequence alignment and regulatory site recognition.
image processing tools.[12] Edge detection was a set of [6] Identification of players and tracking number of
mathematical terms which target was to recognize points in players in sports video took with a (single pan-tilt zoom)
digital image. The use of image processing tools were resize, means it is ip camera, user can control movement and position
scan the plate area .Image processing tools convert image in of lens had many application. It was very tough and
gray scale image and filter the image in order to small objects challenging task. In this paper handled difficulties task. The
and identify plate number which I got object. After that system made the ability to detect and track multiple numbers
Guassian filter used to control the obtain image. of players and used homography between video frames and
[2] Presented changeable software based platform. It was court, and recognize the players. The recognition system
used for recognition application of number plate .In this paper added three weak visual cues, exploit both temporal and
software plug-in used a Dll (Dynamic link library) interface. mutual exclusion parts in a conditional Random video clip and
Processing is broken in many explicit modules. The modular reduce the number of training data which was require to learn
structure provide maximum facilitates during the comparison the recognition system and supervised learning was used .The
between different algorithm in a particular steps .It allows result was tracking, homography estimation and identification.
code to be reused among application.Dll interface define all [7] Identifying and tracking players in sports videos took
items of data and function which was known to be exported in a single moving pan-tilt-zoom camera had many application
by a program in the system. Dllexport and dllimport was .It was a challenging problem due to fast camera motions,
declared for all items. unpredictable player movement and unreliable visual
[3] There were many application of .Automatic Vehicle constituent. It introduced a system to handled problem based
recognition .Automatic Vehicle Identification used in traffic on conditional random fields .System required large number
system(highway electronic tax collection, red light violation of image for training. Semi supervised learning was used to
enforcement ,border and custom checkpoints) etc. License trained and recognition.
plate recognition was more effective form of AVI system. [8] Finding player’s identity was challenging task in
There was easy and smart algorithm for Automatic vehicle sports video. In case of basket ball player identification was
recognition by plate recognition.[12] The given algorithm effective and precise solution. Uniform number could have
contains three main parts- Extraction of plate region, been contained a picture text and problems in localization and
segmentation of character and character recognition. There identification present in variations in orientation, size,
were many types of algorithm used in segmentation of plate illumination, motion etc. They applied modern method for
area smearing algorithm, filtering and morphological localization and identification and HSV color space used.
algorithm. For extraction of plate area we used different [9] They Have presented good and strong method for
algorithm edge detection algorithm and smearing algorithm. detecting and recognize Farsi license plate .It defined space
Statistical based template matching was finally used for plate between cars (how car is far away), rotation(angle between
alphabet recognition. The performance was based on real digital camera and car) and contrast (being dirty, reflected or
image. We noted these algorithms give best result for plate deformed) and considered more than one car. This method
recognition. extracted edges and determine the candidate areas by using
[4] Player recognition was very important task sports adaptive image target is adapt quality, size and resolution
video. In Basketball player’s numbers could be exploiting to enhancement and applying window movement and finally
perform identification. Uniform number could be contained as license plate were detected. The area constituents were
three step picture text and major problems in localization and working according to the license plate geometric structure,
identification were low quality of being determined continuity and parallelism. After searched license plate, solve
(resolution). Low brightness, color aliasing, unconstrained rotation angle and try to compensate it. There were 53 features
background etc. Two different techniques used in this paper

922

Authorized licensed use limited to: Malaviya National Institute of Technology Jaipur. Downloaded on May 18,2023 at 12:09:43 UTC from IEEE Xplore. Restrictions apply.
2015 International Conference on Advances in Computer Engineering and Applications (ICACEA)
IMS Engineering College, Ghaziabad, India

were presented and input was artificial neural network


classifier.
[10] Introduced license plate localization and character
segmentation and identification were the researched vehicle
license plate recognition (VLPR).That was advance technical
which was used in this paper. In license plate they fixed local
region part, O tsu binarization was solved to get the plate –
person, areas. Optimal adaptive Binarization Was used to find
the accurate result of license plate localization. After the plate
localization, character devision part and statistical attribute
were took out, for plate identification multilevel classification
RBF neural network was used.
[11] In this paper they defined automatic license plate [14] Many researchers had been developed many
number. This process was done by three steps: 1) extraction 2) different technique for License plate recognition in past and
character division part 3) recognition. Compared to other got different result. When they implemented this paper they
countries in India license plate was not followed strictly. kept in mind localization method and Indian vehicle license,
Characters on license plate were in many types In Indian license plate recognition specification. They defined of license
language. There occur many changed representation of license plate by using license plate. They made used morphological
plate after that extraction, character division and identification operation like variation, used, dilation and erosion and used
was difficult task. After that they presented extraction Ycbcr color model space for special license plate. They
segmentation and identification in English characters. Sobel defined process in two steps: first color segmentation process
filter and morphological operation was used for plate and checked skewed plates after applied test and got approx
extraction, connected component and vertical projection was result.
used for segmentation, Support vector Machine was used for [15] In this paper they used new method for recognizing
identification. player motions in video those was support vector machine and
[12] License Plate Recognition (LPR) has been deeply optical flow. Optical flow was used for traced area and
read in almost countries. Many different types of number improve robustness.SVM was used for recognize player
plates had been used, the utilize of an automated region of action.
number identification system are different for every TABLE I. COMPARISION TABLE
authorized university. In general terms of any License region
identification system was localized and identification, specific Author Year Approach Description
number plate region. License plate recognition image was Allam Mousa 2002 Software plug-ins They proposed a flexible
through dynamic software based platform
captured by digital smart camera, smart phone camera. The link libraries for license plate
Proposed framework for (LPR) system worked in three parts: identification and
1) Localization 2) Segmentation 3) Recognition. Localization applications described.
of the number plate was done by using morphological Serkan 2007 Template They Proposed application
Ozbay and matching. software designed for the
operation, horizontal and vertical edge processing. Character Ergun identification of car name
recognition was done by neural network classifier; Ercelebi plate.
Segmentation was done by connected component labeling. Matko 2008 Internal contour They proposed a new
Sari,Hrvoje and hsv color method for localization and
Dujmi,Vlada model identification of player
[13]Correctly discover human being in a visual n Papi and numbers in basket ball
surveillance system was very big task for different application Nikola Rozi video was proposed.
region which included person identity, gender, event detection Matko 2009 Algorithm region They proposed algorithm
,human motion characterization, fall detection for elderly Saric,Hrvoje growing based, that used temporal
Dujmic,Vlad hsv color model, redundancy and increased
people .they defined first step which was detection process to
an internal contour identification.
detect object after that background subtraction, optical flow Papic,Nikola
and spatio-temporal filtering techniques. Rozic
M.M.Shidore 2011 Sobelfilter, They introduced algorithm
,S.P.Narote morphological for vehicle license plate
operation, svm extraction, segmentation
and identification.

Baoming 2011 Otsu binarization They introduced otsu for


Shah and multilevel rbf clustering of region and rbf
neural network neural network for
identification.
Wei-Lwun 2011 Semi supervised They introduced the use of
Lu,JO-Anne learning semi supervised learning to
Ting,James decrease the amount of

923

Authorized licensed use limited to: Malaviya National Institute of Technology Jaipur. Downloaded on May 18,2023 at 12:09:43 UTC from IEEE Xplore. Restrictions apply.
2015 International Conference on Advances in Computer Engineering and Applications (ICACEA)
IMS Engineering College, Ghaziabad, India

J.Little,and labels needed during III. CONCLUSION


Kevin training.
P.Murph The conclusion of this paper, first specify various stage of
object detecting, tracking, classification, identification has
S.Adebayoda 2011 Hidden markov They applied hidden been studied. And Literature review of object in real time
ramola,E.Ad model markov model for took out
etiba,A.U. license plate image and
environment and non real time environment using many
Adoghe,J.A. identification. different approaches. The main aim of this paper identify an
Badejo,I.A object approx and using different methods and many
Samuel and approaches to get final result.
T.Fagorus.
Mehran 2011 Neural network They introduced algorithm REFERENCES
Rasooli, classifier for license plate detection
Sedigheh and artificial neural [1] A.Mousa, “Canny Edge-Detection Based Vehicle Plate Recognition”
Ghofrani and network for identification ,International Journal of signal processing,Image processing and pattern
Emad and got accuracy. recognition vol.5 NO.3,September,2012.
Fatemizadeh [2] D.G.Bailey,DIrecki,B.k.Lim and L.yang ,” Test Bed for number plate
Wei-Lwun 2013 Linear They proposed algorithm recognition application” IEEE International Workshop on Electronic
Lu,Jo-Anne programming (linear programming Design,Test application (DELTA.02),2002 .
Ting,James relaxation relaxation) for analyzing [3] S.Ozbay and E.Ercelebi,” Automatic vehicle identification by plate
J.Little,Kevin algorithm the best player recognition recognition”World Academy of Science,Engineering and
P.Murphy in a video clip. Technology,International Journal of Electrical,Robotics,Electronics and
M.M.Kodaba 2013 Morphological They used morphological Communications Engineering,vol:1No:9,2007.
gi,Mr.Vijaya opearation, , horizontal and vertical [4] M.Saric,H.Dujmic,Vladan Papic,Nikola Rozic “Player Number
mahantesh S. horizontal and operation for extraction Recognition in Soccer Video using Internal Contours and Temporal
Kanavi,” vertical edge ,character segmentation Redundancy “.WSEAS International Conference on AUTOMATION &
processing, neural was used for labeling INFORMATION,23 March 2009.
network component neural network
[5] S.Adebayodaramola,E.Adetiba,A.U. Adoghe,J.A.Badejo,I.A Samuel
for recognition license
and T.Fagorus.”Automatic Vehicle Identification System Using License
plate.
Plate” International Journal of Engineering Science and Technology
Saravana 2013 Ycbcr color model They proposed accurate (IJEST), Vol. 3 No. 2 Feb
Kumar.G, and matlab license plate which used
Ram serial operation to localize [6] Wei-Lwun Lu,Jo-Anne Ting,James J.Little,Kevin P.Murphy ,” Learning
Kumar.T, license plate. to Track and identify Players from Broadcast sports Videos” IEEE
Senthil TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE
INTELLIGENCE,July 2013.
Kumar.M,
Manoj [7] Wei-Lwun Lu,JO-Anne Ting,James J.Little,and Kevin P.Murphy,”
kumar.W Semi-supervised Learning for Identifying Players From Broadcast sports
Manoranjan 2013 spatio-temporal They proposed optical flow Videos with Play-by –play Information”2011.
Paul*, Shah filtering and spatio-temporal filter [8] M.Sari,H.Dujmi,V.Papi and N. Rozi ,” Player Number Localization and
M E Haque techniques and techniques and addressed Recognition in Soccer Video using HSV color space and Internal
and Subrata optical flow some issue to some extent Contours” world academy of Science, Engineering and
Chakraborty of fixed object and their technology,Vol:2,2008-07-23.
motion. [9] M.Rasooli, S.Ghofrani and E.Fatemizadeh,” Farsi License Plate
Detection based on Element Analysis and Characters
Gang Liu, 2014 SVM and optical They proposed method for Recognition”International Journal of Signal Processing,Image
Deming flow identifying player motions. Processing and Pattern recognition,Vol:4,No.1,March 2011.
Zhang and
Hui Li [10] B.Shah,” Vehicle License Plate Recognition Based on Text-line
Construction and Multilevel RBF Neural Network” JOURNAL OF
COMPUTERS,Vol:6,No.2,February 2011.
[11] M.M.Shidore,S.P.Narote ,” Number Plate Recognition For Indian
Vehicles “IJCSNS International Journal of Computer Science and
Network Security, VOL.11 No.2, Feb. 2011.
[12] D.Wazalwar, E. Oruklu, Jafar Saniie,” A Design Flow for Robust
License Plate Localization andRecognition in Complex Scenes” Journal
of Transportation Technologies, 2012.
[13] M.Paul, S. M E Haque and S. Chakraborty,” Human detection in
surveillance videos and its applications,Springer open journal.,2013.
[14] M.M.Kodabagi,Mr.Vijayamahantesh S. Kanavi,” License Plate
Recognition System For Indian Vehicles,” International Journal of
advance research in Engineering an technology(IJARET),2 March-
April2013.
[15] Saravana Kumar.G, Ram Kumar.T, Senthil Kumar.M, Manoj
kumar.W,”Accurate Number Plate Recognition” International Journal of
Innovations in Engineering and Technology (IJIET), Vol. 2 Issue 1
February 2013.
[16] G. Liu, D.Zhang and Hui Li,” Research on Action Recognition of Player
in Broadcast Sports Video”, International Journal of Multimedia and
Ubiquitous Engineering, Vol. 9, No. 10 (2014).

924

Authorized licensed use limited to: Malaviya National Institute of Technology Jaipur. Downloaded on May 18,2023 at 12:09:43 UTC from IEEE Xplore. Restrictions apply.
2015 International Conference on Advances in Computer Engineering and Applications (ICACEA)
IMS Engineering College, Ghaziabad, India

[17] Mayank Arya Chandra, Dr. Ravindra Purwar, Dr. Navin Rajpal,"A [18] Ekansh Agarwal, Shaili Gupta, Mayank Arya Chandra, "Data Hiding
Novel Approach of Digital Video Encrytion", International Journal of using Lazy Wavelet Transform Strategy" IJCA Proceedings on
Computer Applications (0975 – 8887) Volume 49– No.4, July 2012. International Conference on Advances in Computer
Engineering and Applications ICACEA(5):5-8, March 2014.

925

Authorized licensed use limited to: Malaviya National Institute of Technology Jaipur. Downloaded on May 18,2023 at 12:09:43 UTC from IEEE Xplore. Restrictions apply.

You might also like