Question/suggestion: target tracking and calibration
Phil Hourglass
2 Posts
Hello,
I have been using Tracker occasionally for research (mechanics/impact testing) and I really like it to look at video and do some quick tracking of objects.
I have questions/suggestions on two topics:
1) Calibration For motion in a plane, a planar calibration can be used to improve the tracking in case the camera is not perfectly perpendicular to the plane of motion. Typically, a pattern of round targets (e.g. four circles at the corners of a rectangle of known dimensions) or a checker pattern can be used for that.
Is there a way to do that in Tracker (I can only find sticks affecting both directions)? If not, that could be nice to implement to increase the accuracy. (Note: Allowing multiple planes within the software could also help).
2) Tracking: I would like to detect and track automatically a round target (i.e. a round object with high contrast put on a moving or fixed body). These are commonly used in testing (e.g. for impact, along with checker targets).
The circle option seems to require defining manually points for the fitting, and then tracking the zone with the autotracker could be used to compute the displacement of the region (seems like the circle is static).
However this would likely be less reproducible (and perhaps accurate) than finding the most likely center of an object automatically by fitting (e.g. after clicking inside). This could also help calibrating the image in a reproducible way (between two targets at a know distance, or more cf. 1).
Thanks! Philippe PS: I can provide links to illustrations if needed
Re: Question/suggestion: target tracking and calibration -
Douglas Brown
450 Posts
Hello Philippe, thanks for your questions/suggestions.
I think the Perspective Filter does exactly what you want. Search for "perspective" in Tracker Help to learn how to use it.
I did at one time consider automatically fitting a circle to or tracing the edge of a region of interest but eventually abandoned it for the template-matching algorithm because it is much more flexible. If your targets are high contrast and your lighting is uniform, then (after correcting for perspective distortion) the template-matching should work well. If your circular target is not too big then you can resize the template so the entire target is inside it.
Re calibrating videos in a reproducible way, if you mean you want to calibrate multiple video that share the exact same camera setup then you can do this by copying the Coordinate System from one tab (Edit|Copy Object|Coordinate System) and pasting into another (Edit|Paste Coordinate System). You can do the same thing by importing the Coordinate System--search for "import coordinate system".
I hope this helps. Doug
You wrote:
For motion in a plane, a planar calibration can be used to improve the tracking in case the camera is not perfectly perpendicular to the plane of motion.
I would like to detect and track automatically a round target...by finding the most likely center...automatically by fitting (e.g. after clicking inside). This could also help calibrating the image in a reproducible way...
Re: Question/suggestion: target tracking and calibration -
Phil Hourglass
2 Posts
Hello Doug,
yes it helps. Thanks.
I had not seen the perspective in the image filter or the copy filter capability.
I was hoping for something in the calibration side not to degrade the image through resizing and just perform the change of coordinates at the end. Resizing seems to create some artefacts around some targets (don't know how the resising is done). I have attached an example (grid for perspective on top: taken on a calibration shot, actual image after removing the calibration grid, without perspective filter on the left, with on the right).
Regarding reproducible, I meant different operators or the same operator on a different day Something with an automatic center detection would remove the operator dependency (and give the same result for the same video and software parameters).
Re: Question/suggestion: target tracking and calibration -
Douglas Brown
450 Posts
Hi Philippe, I'm very surprised to see that discontinuity in the filtered image you attached. It would be very useful if I could attempt to reproduce your result on my machine. Would it be possible for you to send me a single frame from the calibration video (at full resolution)? If it's easier you can send directly to me at dobrown@cabrillo.edu. Thanks! Doug