User:TheLarry/Notebook/Larrys Notebook/2009/06/09

From OpenWetWare
Jump to navigationJump to search
Owwnotebook icon.png Project name Report.pngMain project page
Resultset previous.pngPrevious entry      Next entryResultset next.png

Automating Tracking Software

Turning the 16 bit image into a binary image allows me to use a .vi that can find particles. I can use this .vi to find all the initial particles in the image. So now the user doesn't need to click on anything to get it started. I have to threshold to create this binary image and as of right now this is set to constants. At some point I am going to have to switch this to a program decribed numbers so it can be used for many different movies.

I must be doing something stupid because i go from 16 to binary to 16 to binary. I can definitely make this better.

In in the particle analysis, one of the information you can get out is orientation. This might be able to help when two particles overlap. Because best place to guess is in the particles orientation, but i am having my doubts because once they overlap this .vi will think it is one particle. I gotta think about this problem more later.

Remove Particle .vi is in here. This'll be useful when I add in the ability to take out MICROtubules if they are too small or whatever.

Here is the graph of one of the tubules I tracked. For a long time it doesn't look too bad. I assumed a second between frames so the average speed is something like 10 pixels/second. Everything is cool until frame like 30, 31, 32, 33 where there is a wild deviation. The three different graphs are the tracking of the front edge, back edge, and the average between the two. I don't know why there is such a huge error in the middle, so I'll have to spend time looking at this.

Those frames are when the particles starts to turn a bit. This might be the problem. The tubule then starts to go straight again. I might be able to fix this with a larger order polynomial describing it. Or maybe the polynomial is missing the edges entirely. I will have to work this out. It might be possible that switching over to binary and using the first pixel information may be better in general. Yeah that looks like it. It seems to miss the point when it turns.