Has anyone here been able to get their Vision tracking code to target the reflective tape on the pyramid goal? We have implemented the 2013 Rectangular Target example into our code, and it works without a hitch for detecting the two and three point targets, but when we aim at the pyramid, we don’t pick up any targeting information.
I suspect that this issue has something to do with the fact that the other rectangles that it can detect are hollow on the inside before the convex hull is applied. I’m trying to figure out how to work around this, but I’m stumped. Can anyone offer some advice on where to start with this?
I’ve attached the pertinent code (please excuse the disorganization).
If you process an image of the pyramid goal, I suspect you’ll see score numbers printed next to the large particles. The example code measure a number of factors and scores them according to the 2 and 3 point goals, so the scores will likely be very low. You may want to add or make a specialized version that uses the aspect ratio, and rectangularity measurements and scores according to pyramid. Hollowness probably doesn’t make sense, and I think you can remove the convex hull operation as well.
If you put up an image, I can help a bit more with the measurements and scoring.
Thanks for replying, Greg. The camera does process the targets as part of the binary image, but it doesn’t display any scores. I’ll take a look at modifying the aspect ratio .vi to include that of the pyramid target. I’ll get some pictures for you in a few hours.
EDIT: I’ve attached the unprocessed image, and the processed binary image.
You likely need to lower the Area Size limit % then, and/or increase the Particle Count limit. Note that depending on how good of a binary image you are getting this may have have an impact on processing speed by requiring particle reports and line averages on substantially more particles.
Thanks for the reply, Ruffles. We already had the area size limit at 0%, and I tried increasing the particle count limit from 10 to well into the thousands with no results.
Greg, could you offer some insight on how to modify the example’s Aspect Ratio Score.vi to let it compare the binary image to the pyramid targets’ aspect ratios as well as the middle and high goals?
The photo is dark, but it doesn’t look like much of it is showing to the camera, does it? And the threshold has barely anything at all in it. A bit of horizontal stripe and no vertical stripe.
Once you have a good binary image, you can decide which elements you want to key in on. Perhaps you want to look for the upside down T, or perhaps it is too specific to field location and you want to look for just the horizontal or vertical strip.
To track the pyramid, I track the Alliance Wall. We shoot at the pyramid from the corner, and that is a constant solution in respect to the Alliance Wall. So, our driver drives until the solutions we tested match the ones we found to work. And very excitingly, the labview programmer dwelled into PIDs, so he can make those micro adjustments automatically.
This was my work around to tracking the pyramid. Of course, this only works from one spot, but it is effective. The reason I do this is because of the FOV of my camera. I cannot see the reflective tape at the base of the pyramid, and when I am far enough away to see it, the shot is near impossible and we arent protected like we are when touching the pyramid.
