|
|
|
![]() |
|
|||||||
|
||||||||
![]() |
|
|
Thread Tools | Rate Thread | Display Modes |
|
|
|
#1
|
|||
|
|||
|
Re: Vision Failing
You may want to try using the HSL color space rather than RGB. We've had relatively good luck with it.
Also loosening the particle filter a little bit (on the low end) may help. I would also suggest doing some null checking for each of the different image pointers you get while performing your threshold, filtering, etc. especially if your getting vision task is crashing/throwing errors. In addition to the null checking, I would recommend checking to ensure the height and width of the original camera image are non-zero, as this does occasionally happen for some reason, and will cause the vision to stop working. Lastly, you say you look for a width of 33px to determine when your 12' away? Have you considered using height instead? By using width, your dependent on your horizontal viewing angle; if you use height instead, your distance equation should work from almost any angle (as long as your camera isn't moving up and down). |
|
#2
|
||||
|
||||
|
Re: Vision Failing
I'll 2nd DJ's advice -- great advice DJScribbles! All of these items are necessary. In a realtime system, the world is never kind. Odd things do happen and null pointers and failures do occur occasionally. As programmers, we like to program for perfect systems as it makes our jobs easier, but alas this isn't reality. By carefully taking each step, you'll likely find that things work much more reliably.
bob |
![]() |
| Thread Tools | |
| Display Modes | Rate This Thread |
|
|