Re: Camera Issues -- So close yet so far.
 

I made such a comment and I realize how difficult it is. We spent a lot of time getting our camera interface working reliably. As was mentioned above, our robot scanned the 8 tetra positions to determine what to do. The scanning was really nothing more than pointing the camera at a desired location and looking at a tiny window to determine if the tetra was there. To make it even harder, we had to scan with our lift tilted which complicated the positioning of the camera. Once the object was determined, the camera was manually pointed at it and then was told to scan. Using this metod, the camera was already locked on to the tetra and had no reason to be looking elsewhere.

I made my original comment about focusing as an argument for why a team needs to consider all of the elements that may possibly affect their robot's performance.

I don't know the exact situation of 66's mishap, but is it possible that they were the blue team and that their camera picked up the blue triangle by the autoloader? The camera had a really hard time distinguishing between blue and green....just a thought...

I 100% agree with you on this. The lighting between the practice field and the stadium was completely different. We had to download new numbers depending on where we were...but as much as that was a pain, it was something that we had to deal with.

A solution that we found for this was to set up halogen work lights around our practice area. You may want to try this and see if it helps.

Re: So close yet so far.
 

I agree, and I'd like to put a whitepaper together along with some of the code we used. However, I can't promise anything just yet. As you might imagine, our interest in working on anything FIRST related diminishes rapidly after Championships as we catch up on the rest of our lives. I was already starting to work on something the other night before you even mentioned this though so maybe it'll get done.. ;)
I don't think anyone was (intentionally) suggesting that any of this was easy to do. The fact that not a single team was able to cap the center goal in autonomous is proof-positive that this is really hard stuff.

Re: Camera Issues -- So close yet so far.
 

On 177 we had managed to code our camera well enough that it was redundant enough to be able to change calibration on the fly during the pre-auton time, as well as using a non-auto-servo-mode tracking method that was able to locate the tetra about 75% of the time (from my estimates).

The only problem was, the rest of our autonomous was being a bit sketchy, so we never did anything except drive towards the vision tetra a little bit...

I would love to see FIRST use the camera again next year. Just hopefully for a slightly easier task...

Re: Camera Issues -- So close yet so far.
 

You have to be very careful when making general statements like this. Please keep in mind that all teams using the camera had different mounting systems and different software. We were able to get on the Archimedes field to take some calibrations and discovered that the numbers we got from admin were close, but not exact. However, when we cast shadows on the green and yellow, the numbers were WAY off. Through a little trial and error, we discovered that if the program was calibrated to green and yellow WITH shadows, it still recognized the colors without them but the opposite was not true. In other words, if you calibrated the program without shadows on the colors, it lost them when a shadow fell on the tetra or goal marker.

The fact that you could not get readings off the "distractions" only proves that on that field, at that time, you were unable to get readings off the distractions. I KNOW that our camera locked onto a volunteer's shirt for the following reasons: 1) If the camera does not recognize the color it's looking for, the drive-train shuts down (safety feature) 2) In the match in question, the robot had to turn approximately 30 degrees to line up with the volunteer with the tan shirt and then 3) moved directly toward him. If the camera had not recognized the tan shirt as yellow (this was after we already picked up the tetra) the robot would not have moved toward him.

I could get on my soapbox and spell out everything that FIRST could have done differently, but I would rather just congratulate all the teams that tried to work with the less than optimal parameters at Nationals. In retrospect, the real world is seldom how we would like it to be, and this was an excellent example of how you have to adapt in order to function.

Re: Camera Issues -- So close yet so far.
 

one thing though. the calibration values varied from field to field, and from what ive heard there have been different results in each arena also. adding to that, archimedes, specifically, didnt have the right values--i say this because we realized that though they were right theoretically, they didnt account for the shadows casted on the field. had there been better and more omnidirectional or overhead lighting, they probably would have worked.
heres the deal: the shirts are only picked up under ONE circumstance--that you account for shadows on the field and calibrate the yellow you're supposed to track with that in mind. interestingly enough, to pick up the shirts required the exact situation that was necessary to calibrate effective numbers on our field.

thats what we found, and took precautions not only for our ability but for their safety.