Color Tracking Motion Detection - PowerPoint PPT Presentation

1 / 11
About This Presentation
Title:

Color Tracking Motion Detection

Description:

(1) When the object is moving, the centroids of some color bins ... NO: continue to bi-party the small bins, till each tiny bin belongs to only one object. ... – PowerPoint PPT presentation

Number of Views:32
Avg rating:3.0/5.0
Slides: 12
Provided by: mathT3
Category:

less

Transcript and Presenter's Notes

Title: Color Tracking Motion Detection


1
Color Tracking Motion Detection
  • CIS 750 Final Project
  • Advisor Professor Latecki
  • Presentation X. Wen

2
Agenda
  • Introduction
  • Algorithm
  • Experiments
  • Conclusion

3
Introduction
  • We make the bins histogram first, and then use it
    to track the object.
  • Based on the fact
  • (1) When the object is moving, the centroids
    of some color bins(belongs to the object) is
    moving the same way with the object.
  • (2) The background bins' centroids is also
    moving but in a relatively slowly way and moving
    in the opposite direction of the objects.

4
The classical method Algorithm
  • Take the differences of the frames(usually to a
    background frame),
  • Get the object's position in each frame,
  • Obtain the object's moving path.

5
Color tracing method Algorithm
  • Group bins for each frame
  • Find the moving path of the bins, (differences
    of the centrods of the bins)
  • Identify the moving object. (connected component
    method.)

6
Assumptions
  • Objects in the frame have different colors.
  • The colors of the pixels belonging to objects'
    body don't change with time.
  • The object do not make large rotations during
    moving. The movement of each
  • pixel of object parallels to that of the
    object.

7
Algorithm (Step 1)
  • (1) Make each bin belongs to at most one object.
  • Split a bin to equal two parts, test whether
    the small bin parts' centroid move with the same
    trajectory as the original bin.
  • YES the bin is considered belongs to only
    one object.
  • NO continue to bi-party the small bins,
    till each tiny bin belongs to only one object.

8
Algorithm (Step 2,3)
  • (2) Group bins
  • A Drop the bins with small trajectory
    length.
  • B Combine bins with with same slope and
    same length of the trajectory.
  • (3) The trajectory of the bins left show us the
    corresponding objects' moving paths.

9
Experiments
  • Extremely good when the objects have different
    color with the background.
  • Very bad when the background is moving.
  • Not very good when the objects have same with the
    background

10
Conclusion
  • The method is first suggested by Professor
    Latecki,
  • It works good when the objects have different
    colors with the background.
  • It is still under development.

11
Thank you!
Write a Comment
User Comments (0)
About PowerShow.com