Targeted Object Tracking In a Live Video

    3 Votes

Video tracking is the process of locating a single or multiple moving object in time using a camera. An algorithm analyses the video frames and outputs the location of moving targets within the video frame. The main difficulty in video tracking is to associate target locations in consecutive video frames, especially when the objects are moving fast relative to the frame rate. Here, video tracking systems usually employ a motion model which describes how the image of the target might change for different possible motions of the object to track. Examples of simple motion models are

  • To track planar objects, the motion model is a 2D transformation of an image of the object (e.g. the initial frame)
  • When the target is a rigid 3D object, the motion model defines its aspect depending on its 3D position and orientation
  • For video compression, key frames are divided into macro blocks. The motion model is a disruption of a key frame, where each macro block is translated by a motion vector given by the motion parameters
  • The image of deformable objects can be covered with a mesh, the motion of the object is defined by the position of the nodes of the mesh.
  • The role of the tracking algorithm is to analyze the video frames in order to estimate the motion parameters. These parameters characterize the location of the target.

Temporal difference (DT) and template correlation matching are the 2 methods used for tracking targets in real time applications.  In Temporal difference, video frames separated by a constant time and are compared to find the changed regions.  In template correlation matching, each video image is scanned for the region which best correlates to an image template. If we implement these methods one at a time, these methods have significant shortcomings. Temporal difference and template correlation matching are the 2 methods used for tracking targets in real time applications. In Temporal difference, video frames separated by a constant time and are compared to find the changed regions.  In template correlation matching, each video image is scanned for the region which best correlates to an image template. If we implement these methods one at a time, these methods have significant shortcomings.if there is significant camera motion, DT tracking is impossible. If we want to make this algorithm work, a well defined image stabilization algorithm should be developed. It will fail if the target becomes occluded or stops its motion. For Template correlation matching, target object should remain constant. If object's size, direction or lighting condition change, correlation matching won't an ideal solution.

These methods can complement each other. If the target is stationary, template matching is the best method to track targets. If the target is in motion, Temporal difference would be the ideal choice where template matching won't fit. So it will be good to combine two methods. Idea behind this is to use DT to detect moving targets and train the template matching algorithm. These targets are then tracked using template matching guided by the DT stage. This combination obviates the need for any predictive filtering in the tracking process as the tracking is guided by motion detection.

References

http://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.73.2875&rep=rep1&type=pdf

Popular Videos

communication

How to improve your Interview, Salary Negotiation, Communication & Presentation Skills.

Got a tip or Question?
Let us know

Related Articles

Travel Planner using Genetic Algorithm
Data Recovery and Undeletion using RecoverE2
PC CONTROLLED ROBOTIC CAR
Routino Router Algorithm
Data Leakage Detection
Scene Animation System Project
Data Structures and Algorithms Visualization Tool
Paint Program in C
Solving 0-1 Knapsack Problem using Genetic Algorithm
Software Watermarking Project
Android Gesture Recognition
Internet working between OSI and TCP/IP Network Managements with Security Features Requirements
Web Image Searching Engine Using SIFT Algorithm
Remote Wireless Sensor Networks for Water Quality Monitoring Requirements
Ranking Spatial Data by Quality Preferences
Scalable Learning Of Collective Behaviour
Computational Metaphor Extraction And Interpretation
Designing a domain independent Rules Engine For Business Intelligence
Graph Colouring Algorithm
Gesture Based Computing