CMUcam Results!

[Terry Sherman]

We tested out to 55 ft and tracked but we noticed our tracked blob size was 1 pixel so we're not confident that the racking is exceptional. We also noted the further away we went, the camera tilted less and less (as expected) but in an attempt to quantify it we noticed that ~45-55 feet we got the same angle feedback from the camera, so yes we could track and track consistently, so we're accurate, but the precision is not very great. We're doing some analysis to figure out where we feel we can safely determine trajectories etc...
Have fun!
-Terry

[Matt Adams]

Originally Posted by Terry Sherman

in an attempt to quantify it we noticed that ~45-55 feet we got the same angle feedback from the camera, so yes we could track and track consistently, so we're accurate, but the precision is not very great.

I don't want to dillute the thread here... but does your team (does any team?) really think they're going to launch from 45 feet away? I'm not trying to pick a fight, but I think that your definition of precision might be a little optimistic.

At those distances, the angle is 25 and 20 degrees for 45 and 55 feet away respectively. That's +/-5 degrees at that distance. If this was happening at 20 and 25 feet, I would be disappointed.

Feel free to answer honestly: Have I set my personal 'realistic & reliable' shooting distances of 20 feet (absolute max) too low?

Matt

[Brandon Holley]

Honestly, to be safe shooting from 20 feet is a very nice distance. I'm sure most teams are going to be going for some sort of range around there or closer to the target. BUT, what happens if you have an awesome defender on you, and you cannot get any closer with timing running out, maybe the only way to make a shot is to fire from 40 ft???? Even if you are not as accurate as you are when you are closer (which is a given), you can still have a CHANCE to make a shot.......If you unload all of your balls at the target, maybe you can hit a few, and turn the match around in your favor. I think having the long ball could end up winning a lot of matches!

[Terry Sherman]

One of the inputs to our design team working on the shooter is how realiable the targeting system can be. There is no need for them to try and get distance of the 40+ range if we can't target there. Realistically we're not likely to shoot anywhere near that but, and if we could it would require a great deal of investigation and effort over say a 20 foot shooter. We wanted to determine if they could grant us that type of system, could we really use it. From our testing it doesn't seem likely we could use it so our shooting distance will likely be shorter...

-Terry

[seanwitte]

Originally Posted by Terry Sherman

We tested out to 55 ft and tracked but we noticed our tracked blob size was 1 pixel so we're not confident that the racking is exceptional. We also noted the further away we went, the camera tilted less and less (as expected) but in an attempt to quantify it we noticed that ~45-55 feet we got the same angle feedback from the camera, so yes we could track and track consistently, so we're accurate, but the precision is not very great. We're doing some analysis to figure out where we feel we can safely determine trajectories etc...
Have fun!
-Terry

With a blob size of 1 pixel was there any decrease in the tracking reliability? If you move the camera (or the target) would it reliably stay locked on? I only ask because I'm interested to find out how small a blob the camera will track. Its great that you all have been able to make it work out of the box and is really due to Kevin Watson's hard work.

[devicenull]

Originally Posted by seanwitte

With a blob size of 1 pixel was there any decrease in the tracking reliability? If you move the camera (or the target) would it reliably stay locked on? I only ask because I'm interested to find out how small a blob the camera will track. Its great that you all have been able to make it work out of the box and is really due to Kevin Watson's hard work.

Depends on your noise filter settings, I think. If you totally disable the noise filter, then it may be locked on then, but you will run into other problems.


We may not be shooting from 50ft away, but the camera tracking that far away means we can run full speed at our goal from the other side of the field, and be ready to shoot before we are actually in range. Saves time

[Terry Sherman]

Yes, Kevin put in a great deal of hard work and we're thankful for it. We took each measurement 6 times to verify repeatability. The first 3 were from a static state and the last the was us moving the camera to that position and having it track along the way, each time the same result was determined.
We are using Noise Filter 1.

-Terry

[Jared Russell]

Originally Posted by seanwitte

With a blob size of 1 pixel was there any decrease in the tracking reliability? If you move the camera (or the target) would it reliably stay locked on? I only ask because I'm interested to find out how small a blob the camera will track. Its great that you all have been able to make it work out of the box and is really due to Kevin Watson's hard work.

Looking to get a new lens eh?

Oh, and to the guy who thinks precision is overrated - our shooter made the goal CONSISTENTLY from ~50 ft the other night.

Granted, our exit velocity might be a bit high...

[Jon236]

I'm glad everyone is getting such good results. The one piece of data lacking from this thread is ambient light level. Our school recently upgraded to super-efficient (and super-bright) lights. With them on, we're tracking at 15 feet; with lights off, 30+. How bright will the arenas be?

Jon Mittelman
Team236

[Jared Russell]

Fairly subdued lighting - we are talking about university gymnasiums for the most part. If you get it working in a room with normal light, it will work fine in the arena, probably better.

I'm surprised that you can only get 30ft with the lights off, though.

[Damian Manda]

Our previous testing had resulted in a distance of about 20ft, because we had put a threshold of 10 pixels to track due to earlier interference. We have now tuned the parameters and decided to try again.

We got a maximum distance of about 92ft today, in a fairly well lit hallway with the standard system. It would see 1 pixel, or 2-3 if placed on the floor (enlargement due to reflection). At this distance, the camera had a hard time tracking, but would re-lock on if it lost the light. In later testing with the light properly mounted on a goal, we saw about 6 pixels from a distance of 27 feet (half of the field).