Single Object Tracking System by Using Labview
Uploaded by
Editor IJRITCCSingle Object Tracking System by Using Labview
Uploaded by
Editor IJRITCCInternational Journal on Recent and Innovation Trends in Computing and Communication ISSN: 2321-8169
Volume: 4 Issue: 5 52 - 56
_______________________________________________________________________________________________
Single Object Tracking System By Using Labview
Divya M Dr. Ravi kumar A V
Department of Electronics & Communication Engineering Department of Electronics & Communication Engineering
S. J. B Institute of Technology S. J. B Institute of Technology
Bengaluru, India Bengaluru, India
[Link]@[Link] ravi_sjb@[Link]
Abstract Moving object detection is an interesting area in the field of video tracking. When the object is moving then tracking that is
challenging task in vision area. This paper is about detecting, tracking of moving objects from the source. Many image processing methods to
detect and track the moving object from the source was proposed, but those methods require more time to implement and is complex. This paper
gives information about the moving object detection based on the tracking mean shift algorithm. LABVIEW vision module introduces a recent
algorithm for object detection and tracking, in this algorithm current location is searched based on the histogram of the object in the previous
image frame and result of mean shift is used to find the peak of probability density function near the object old position. Initial location of the
target object need to be determined for this to work correctly. Mean shift is an effective algorithm that has so many application especially in the
field of computer vision like clustering, object contour detection etc.
Keywords LabVIEW, object tracking, mean shift
__________________________________________________*****_________________________________________________
I. INTRODUCTION II. MEAN SHIFT ALGORITHM
Object detection and tracking is important research in the felid Mean shift algorithm for moving object tracking was initially
of computer vision . Tracking is an important approach in proposed byfukunaga et al, comaniciu [10] in the estimation of
various fields like, in military uses such as anti-aircraft, probability density function et al,[8].if we have a distribution
artificial intelligence, monitoring and security. Some other points . Then according to the mean shift algorithm modes or
application of object detection are vehicle navigation, peaks in density function is determined. This method is called
biometrics, robotics, video games , and industrial automation non-parametric method [9].this method of tracking tracks the
applications. object for long time and more robust compare to other tracking
algorithm.
In this paper an object can be defined as anything that is
To find the new location of the object that we are going to
stimulating for inspection for example .people walking on the
track, we need to find a vector which can suggest the direction
road , movement of cars on the road, face in a motion etc.
of the moving object. This vector is called mean shift vector
If we take an object as a vehicle then the vehicle detection and [10].First we need to draw the ROI around the object and get
tracking is widely used in the vehicle information system, in the data points, approximate location of the mean of this data.
intelligent transport and traffic systems. Most of the research Then estimate the exact location of the mean of the data by
works are on this field from last decade by using image determining the mean shift vector from the initial mean.
procesisng technolgy and computational capacity. Now a days
vehicle detection and tracking is key factor for the road traffic In the first frame, tracking object is selected and
controller. Some of the appliction of vehicle detection and object model has probability distribution of color
tracking system are traffic signal controller, automatic vehicle histogram. If 0 is the center of an object, then the
accident detection and traffic density estimation.[4]. position of pixels are i=1N, where N is the
number of pixels in the image. statical histogram
Video object tracking can be stated as detection of an
distribution model of target area given by
object when it is moved around the [Link] research work
is very intresting in the field of both civilian and military =C =1 2 .
applicattion. At the current frame the statistical histogram
distribution given by
A well-configured object tracking method track the object 0
apart from blur, noise, or particle occusion of object . any 0 = =1 || 1
gradual changes in the tracked object including geomatric
transformation such as shifting, rotation by NI vision is Computing the measurement between the object and
tolerent for object tracking. candidate template by Bhattacharyya coefficient.
1
Object tracking in LABIEW NI vision can be used in gray P ( (0), )= =0 (0 ) (2)
scale(U8, U16) as well as color(RGB32).
52
IJRITCC | May 2016, Available @ [Link]
_______________________________________________________________________________________
International Journal on Recent and Innovation Trends in Computing and Communication ISSN: 2321-8169
Volume: 4 Issue: 5 52 - 56
_______________________________________________________________________________________________
Weight of the window of pixels in tracking window [Link] AVI2 open VI
1
= [( ) ]
(0)
=0
This VI used to opens the existing file. It contain
New object position search by mean shift value given AVI file path, which is complete pathname, including drive,
by directory, file name, for the AVI file. Extension of the file
1
=0 (
2) name must be .avi.
= ( 1 2)
(3)
=0
Then computing the Bhattacharya coefficient given [Link] AVI2 get info VI
by
P 1 , = 1
=0 (1 )
This VI acquire the information about the AVI file.
Comparing coefficients and update the candidate Such as width of the image, height of the image , frame
window. number per second, number of frames contained in the AVI
If ||1 0 || < iteration stops, and going to step 2. file etc.
III. OBJECT TRACKING IN VIDEO SEQUENCE [Link] create
Object tracking in a video sequence can be defined as process
of locating the position of a moving object over the time using
a camera. Application of video tracking is security and this VI creates the temporary memory location for an
surveillance, traffic controller, medical imaging [Link] aim image. using the IMAQ create in conjunction with IMAQ
of the object tracking is tracking the target object in a video dispose VI we can create or dispose the images in LabVIEW.
frames. First step in this work we have to read the AVI video
file. AVI file can be read ion LabVIEW. The basic parameter [Link] read frame
to initialize, create an AVI file can be set by IMAQ AVI2
create function. It includes path name, including drive, file
name, directory etc. file extension must be AVI. AVI frame
rate should be equal to rate at which images are acquired. This VI used read the image from the AI file
IMAQ AVI get info is used to obtain the information about the specified by AVI refnum. Image is reference to the image that
AVI file with AVI reference. To read the frame from the video we want to contain the AI image after we call this VI. Frame
IMAQ AVI read frame is used, with the frame number is number identifies the which frame of AVI file to read. Image
specified. It will read an image from the AVI file which is out reference to the output image or destination image
specified by AVI refnum. Which is reference to the AVI file
output put from IMAQ AVI2 create or open VI. Technique of
tracking an object in a video sequence of frame is directly IV. PROPOSED METHODOLAGY
depending on the object representation being used. Figure 1.
Shows how the AI file can be read using LabVIEW. Once the avi video file read we have select the algorithm for
object tracking. Algorithm used in this paper is traditional
mean shift algorithm. It will track the user defined object by
iteratively updating the location of the object.
Next step is to wait for the user to draw the region of interest
(ROI), if the user fails to draw the ROI loop repeated until user
draw. IMAQ add tracking object is used to add the object
information from the ROI stored in the object tracing section.
Once get the information about the object next step is to check
frame by frame.
Figure 1 block diagram of acquiring AVI file
53
IJRITCC | May 2016, Available @ [Link]
_______________________________________________________________________________________
International Journal on Recent and Innovation Trends in Computing and Communication ISSN: 2321-8169
Volume: 4 Issue: 5 52 - 56
_______________________________________________________________________________________________
[Link] object tracking options VI
This VI specifies the option to which
algorithm we are going to track the object. There are two type
of option mean shit option and shape adaptive option.
IMAQ mean shift option specifies the algorithm for mean shift
tracking. IMAQ shape adaptive option specifies the algorithm
for shape adaptive object tracking.
[Link] add tracking object
Figure 2. Code for draws the ROI on an image
Adds the object instance from the list of instance
stored n the object tracking session. Object tracking section in
Once the user draws the ROI it will exit from the loop. IMAQ is reference to object tracking session. Image is the image
add tracing object adds an object from the list of instance contains an object to track.
stored in the object tracking session .IMAQ track object VI
gives the information about the tracked object. And tracks the [Link] track object VI
user defined object from frame to frame. Once the object I
found draw the bounding box. Then the object will be tracked
using IMAQ track object VI, by using the IMAQ overlay
multiple lines 2 VI we can see the tracking of an object. We This VI use to track the user defined objects from
can change color of the bounding box. This process will be image to image. Image is reference to the image that contains
repeated number of frames times. This method effectively the object to be track. Objects in this VI refer to the object
checks the every frame to track the moving object. Figure found in the image.
illustrate how the object being tracked using LabVIEW.
V. EXPERIMENTAL REULTS & PERFORMANCE ANALYSIS
The project implementation has three front panel components.
The first component is file path, which gives the location of
the source file. The second component display indicator which
is used to display the output .third component is Stop button
which is used to terminate the program.
Experiment 1: The result of moving vehicle tracking in a
video sequence
Figure 3. Code for tracking object
[Link] create object tracking session VI
Create the new object session we use this VI. Object
tracking session in this VI reference to the new object tracking
session.
Figure 4. Before tracking the object
54
IJRITCC | May 2016, Available @ [Link]
_______________________________________________________________________________________
International Journal on Recent and Innovation Trends in Computing and Communication ISSN: 2321-8169
Volume: 4 Issue: 5 52 - 56
_______________________________________________________________________________________________
Figure [Link] tracking the object. Experiment 2: test results of tracking human in a video
(a) 45th frame sequence
Figure 6: before tracking the object.
(b) 61st frame Figure 7: After tracking the object.
(d) 31st frame
104th frame
(e) 39th frame
(f) 42nd frame
Experiment 8: tracking of moving ball.
55
IJRITCC | May 2016, Available @ [Link]
_______________________________________________________________________________________
International Journal on Recent and Innovation Trends in Computing and Communication ISSN: 2321-8169
Volume: 4 Issue: 5 52 - 56
_______________________________________________________________________________________________
VI. CONCLUSION
LabVIEW method is cost effective. It is also low on
maintenance so labVIEW program is easy to use. The drag and
drop feature reduces the time to implement program.
Many algorithm of moving objects detection achieves by
image processing, but in this algorithm which is easy to be
affected by light change of surround. So this paper provides
algorithm of moving object detection and based on tracking
mean shift algorithm. Mean shift is an application,
independent tool suitable for real data analysis. The
experiment is performed on the moving vehicle, moving
human, moving ball. The results at figure 4-5,6-7,8-9 shows
Figure 9: after tracking the object the movable object tracking successful. Experiment on actual
AVI (Audio video interleaved) files shows that the algorithm
is effective with high accuracy.
REFERENCES
[1] Ravi kumar A.V, Dr. nataraj .R , Dr. rekhka K, [Link]
real time video edge detection in labVIEWIJCSIT, Val.
3(2),2012,3808-3811
[2] Ravi kumar A. real time face selection and tracking based
camera motion control systemIJECE, val 7, no 2(2014), pp.89-
97
[3] Ravi kumar A.V , sunil tiwari N, multiple moving object
detection and tracking using Harr features with smart video
(g)15th frame surveillence systemIJERT, vol 3-issue 6(june 2014),2278-011
[4] Morris, B.T; Cuong tran; scora, G; Trivedi, M.M;Barth,
M.J.,Real-time video-based traffic measurement, Intelligent
Transportation system, IEEE Transactions on, vol.13, no.4,
pp.1667,.1678,Dec.2012J.
[5] Wei zhan, xiaolong ji, b algorithm research on moving
vehicles detection
[6] Khatoonabadi, S.H bajic, I.V, video object tracking in the
compressed domain using spatiotemporal markov random
fields,IEEE International on image processing 2013
[7] Amer, A, voting-based simultaneous tracing of multipe video
objects,IEEE transaction on circuits and system for video
technoogy.2005
[8] Mahesh kumar chouhan, rahul mishra, [Link]
nitnawwremovabe object trcking by using mean shift method
(h)27th frame with adjusted background histogram volume 2, issue 7, july
2012.
[9] Xiaojing zhang,object tracing approach based on mean shift
algorithm journal of multimedia, vol.8, No 3, june 2013
[10] [Link], .ramesh and p. meer, real time tracking of non-
rigid object using mean-shift, IEEE [Link]. vis. Pattern
recog.,pp. 673-678,2000.
[11] Jaideep jayakar [Link] babu and K.R ramakrishnaan,
roubst object tracking with backround weighted local kernels,
comput. Vis. Image understanding, ol.112,pp.296-309,2008.
[12] Fukunaga, keinosuke, andlarry hostetler. the estimation of the
gradient of a dencity funtion , with application in pattern
recognition. Information theory, IEEE transaction on
21.1(!975):32-40.
(i)88th frame
56
IJRITCC | May 2016, Available @ [Link]
_______________________________________________________________________________________
Common questions
Powered by AIChallenges and limitations in object tracking using the mean shift algorithm primarily arise from environmental factors such as lighting variations, object occlusion, and background noise. Mean shift can be impacted by sudden changes in lighting conditions, which may alter the histogram representation of the object . Occlusions or overlapping objects pose a significant challenge as the algorithm relies on continuous visibility for accurate tracking . Additionally, background clutter or similarity in color distribution between the object and background can lead to tracking inaccuracies . Despite these challenges, the mean shift's capability to adapt to gradual changes in object appearance moderately addresses some dynamic environmental variations.
The mean shift algorithm is considered advantageous for moving object tracking in LabVIEW due to its non-parametric nature, which allows it to determine modes or peaks in the density function efficiently . This robustness makes it suitable for long-term tracking and handling gradual changes in the object's appearance, including geometric transformations like shifting and rotation . Furthermore, the algorithm's integration with LabVIEW’s vision module simplifies the implementation process, leveraging LabVIEW’s capabilities for effective object contour detection and tracking without being affected much by noise or occlusion .
In LABVIEW, object tracking in a video sequence is facilitated by various VI (Virtual Instrument) functions that handle video file reading and image processing. The IMAQ AVI2 Create Function sets up the AVI file with essential parameters like file path and frame rate . The IMAQ AVI Read Frame function reads specific frames from the video . For tracking, the IMAQ Add Tracking Object function incorporates the object information by defining a region of interest, while the IMAQ Track Object VI tracks the object across frames, utilizing the mean shift algorithm . These functions enable efficient processing and manipulation of video data for object tracking purposes.
Applications of object detection and tracking span diverse fields including security and surveillance, traffic control, medical imaging, and automatic vehicle response systems . Advancements in algorithms like mean shift enhance these applications by providing more accurate and efficient tracking capabilities. For instance, in traffic management, improved tracking algorithms enable better vehicle navigation and congestion analysis . In surveillance, they help maintain continuous identification of individuals or objects despite environmental changes or occlusions, thus improving security outcomes . These advancements facilitate real-time monitoring and reduce errors associated with manual oversight or simpler algorithms.
The mean shift algorithm adapts to geometric transformations during object tracking by continuously updating the target’s probability distribution based on real-time frame analysis. It evaluates the object's histogram model against potential candidate regions, adjusting to transformations like scaling, rotation, and translation . Each iteration involves recalculating the mean shift vector, which directly influences the object's updated position, allowing it to align with transformed object orientations . This adaptability is achieved through analyzing changes in feature space distribution, ensuring the algorithm remains robust against such transformations throughout the tracking process.
Object tracking remains effective in the presence of blur and noise by employing robust algorithms like the mean shift, which focus on maintaining the object's histogram model rather than relying on precise image details . These algorithms work on higher-level features, such as color distributions, which are less affected by noise or blur. Additionally, the algorithm's iterative process, using techniques like the Bhattacharyya coefficient, evaluates and refines object position based on cumulative data across multiple frames, mitigating short-term noise effects . This ability to average out variations over time ensures consistent tracking even in imperfect visual conditions.
LABVIEW's drag and drop feature simplifies the implementation of object tracking algorithms by providing a graphical programming interface rather than relying on code-based methodologies . This reduces the complexity of programming as users can visually assemble functions and components by dragging and dropping them onto the interface. It enhances accessibility, particularly for those without extensive programming experience, and accelerates the development process, allowing straightforward configuration of the tracking system without needing to write extensive code .
The implementation of the mean shift algorithm for object tracking in LabVIEW involves several key steps. Initially, the ROI (Region of Interest) around the object is defined, allowing the extraction of data points and the estimation of the object's initial mean position . The mean shift vector is calculated to suggest the moving object's direction, updating its position iteratively . In each iteration, Bhattacharyya coefficients compare the histograms of the tracked object and the candidate, refining the tracking process . The iterative process stops when the difference between consecutive object positions is below a threshold, indicating convergence to the object's new location . These steps ensure robust tracking by continuously adapting to the object's movement in video frames.
Compared to traditional tracking algorithms like Kalman filters or particle filters, the mean shift algorithm offers robustness particularly in scenarios with non-linear motion and appearance changes . It does not assume any specific parametric form of the target's dynamics, making it more flexible and capable of handling abrupt changes in the tracking scene . However, mean shift can be computationally intensive due to its iterative computation of the density gradients and similarity measures, yet it remains relatively efficient in converging to the target's position by leveraging the simplicity of histogram comparisons . In contrast, parametric algorithms like Kalman filters may perform faster but struggle with non-linear dynamics and sudden changes.
In the mean shift algorithm for object tracking, the Bhattacharyya coefficient is used to measure the similarity between the target model's histogram and the candidate region's histogram in the current frame . It is computed using the formula P(p_h(y_0), q_h), where p_h(y_0) is the histogram of the candidate region and q_h is the histogram of the target model . This coefficient serves as a metric to evaluate how well the candidate region matches the target object, guiding the algorithm to adjust and refine the object's tracked position .