Difference between revisions of "Tracker"

From MagnetoWiki
Jump to navigation Jump to search
(Fix perm redirect.)
Line 64: Line 64:
 
===Alternative Implementation===
 
===Alternative Implementation===
  
Use grayscale instead of B&W. Position, etc would be calculated based on weighted pixels. This would be able to pick up sub-pixel changes and provide smoother, more accurate data.
+
Use grayscale instead of B&W. Position, etc would be calculated based on weighted pixels. This would be able to pick up sub-pixel changes and provide smoother and more accurate data.
  
 
[[Category:Software]]
 
[[Category:Software]]

Revision as of 16:35, 29 December 2011

Tracker Program Description and Tutorial Video

Tracker is a program to track the path of animal movements in experiments.

Click on the screenshot below to see a brief tutorial video on how to use Tracker.

TrackerScreenShot.jpg

Source Code and Auto Build

Tracker Project Code

Tracker Auto Build

Proposed Functionality

Input:

  1. Quicktime movies of experiment

Outputs:

  1. Numeric output
    1. Instantaneous
      1. time
      2. position
      3. orientation
    2. Derived
      1. velocity
      2. distance
      3. direction of motion
      4. measure of wall taxis
  1. Generated Movies
    1. Extracted data graphic.
    2. Overlay of extracted data graphic and original video.


Proposed Implementation

Time determined from Quicktime movie time stamps.

Position is average position of blob pixels.

Orientation is average vector from average position to blob pixels. Position and orientation can be used to define oval around body.

Direction of motion, velocity and distance are calculated from Time, Position and Orientation in the usual way.


TrackerAngles.jpg

Proposed Technology

Cocoa application.

Use QuickTime filter to convert video to 2-color B&W via brightness cutoff.

Filter (either QuickTime or after frame extraction) to clip non-experiment portion of video.

Frame-by-frame code to detect blob and calculate position, etc.

Alternative Implementation

Use grayscale instead of B&W. Position, etc would be calculated based on weighted pixels. This would be able to pick up sub-pixel changes and provide smoother and more accurate data.