Friday, February 10, 2012

played with Tracker

Downloaded Tracker 4.62, the Video Analysis and Modeling Tool software by Douglas Brown. Fired up the Java app. Started reading the Getting Started topic in the help. And immediately felt overwhelmed. What was I getting myself into... But slowly, I figured some things out.

I loaded my video. Set the start and end frames. I was very pleased to see a step option; I adjusted this to output 50 or so data points from my trial. I calibrated the scale, although I wasn't sure what to use at the unit of measurement—I used the real time. I set the reference frame origin and angle. And then I tracked the stars. I was glad I only chose 50 points. It already felt tedious marking the 50 frames. I saw the matrix build up with rows showing t, x, and y values. I suddenly I realised that I was actually getting somewhere. I exported the results to a spreadsheet and stared at the numbers.

I saw that y was the delta from the origin. In other words, it was showing the separation in arc-seconds. If I took an average, it would compensate for the seeing, the wavering. But then I realised that y was constant in this model. And that was not really correct. And I'd have to take into account the declination along with the drift speed and/or time to get a real separation value. And that was a little off-putting. The angle though... That should be easy!

The t value was how far we are along the calibration stick. I had set it to the time from 0 to 83.47 seconds.

The x value was where the companion was in relation to the primary. As it was a touch to the left, it showed as a negative number in the first row. I saw that it was always slightly less than the corresponding t value. Which was correct for my video. If I subtracted the x value from t, I'd have the relative position.

I realised it was a simple matter then, using the net-x and y values, to determine the angle between the two stars. Uh. Yeah. But that would mean trig. Nooooo! Damn it. I just can't get away from trigonometric math. It keeps following me...

I had to dive into the web for some pointers on how to do this, it had been so long. Crikey. Would need the arc tangent function. Meh. In Excel, I tried

= ATAN( x / y ) * ( 180 * PI( ) )

and received a good result. I then tried

= DEGREES( ATAN( x / y ) )

which was a bit cleaner. Copied it into the matrix. Not bad.

The average angle was 10.1 degrees. But there was a lot of variance! I saw a minimum value of 4.9 and a max of 16.6. Wow. I really needed to take into consideration that both the stars were moving, both shifting with the seeing conditions. I needed to plot both stars.

I was pleased with the new outcome. The average returned was 12.1 with a min of 6.7 and a max of 17.6. The average was pretty close to my target. Still. It was a little frustrating to miss by about a degree.

Hold on... I got an idea. What if... I wondered what the frequency distribution might look like. So, back to the web for some tips. Found the Data Analysis plug-ins. Launched the Histogram. And then I made the graph.

Holy cow! Would you look at that. The peak was right on target! Now this was interesting... Eerie.

No comments: