Concerns regarding the green paneling/green shirts

[Al Skierkiewicz]

Originally Posted by scottm87

The vision system has a few different modes. One mode (that most directly applies to the challenge) returns the location and bounding boxes of blobs of color. In the code, it most likely takes these bounding boxes, sorts them by the similarity to the target color, and plots a path depending on the blob's X position or something. It would not be difficult to A) decrease the range of colors or B) apply another level of filtering to determine the height of the blob off the playing field to filter out humans or robots (tall ones at least).

All I am waiting for is someone to pick up a serial CompactFlash interface that are always in Circuit Cellar, and write the frame captures and show a post-match movie .

I think Scott and a few others are on to something here. The sensor is a camera capable of color matching and shape matching. It is not just a color sensor. I think we are blowing this out of proportion. Thunder chickens, don't change your shirt color and other teams don't worry about your robot color and MCs don't go for black shirts until we get some real data about the software and the camera. There are some pretty cool and knowledgeable people at IFI and FIRST, don't you think they would have thought of the problems.

[dlavery]

A few items that you might want to consider before you go heading down too many blind alleys:

- There are three possible starting locations for your robot. Pick any of the three. You know what the field should look like from that vantage point.

- There are eight possible starting locations for the vision tetras. Those locations relative to the robot starting locations are known and fixed.

- The number of robot location/vision tetra position permutations is finite.

- The CMUcamII allows you to trim the field of view to a pre-defined area of the video frame, so extraneous items that are not of interest (ie. off-field items) can be ignored.

- For a given field of view, if two or more objects of the target color are in view and the CMUcamII is requested to locate the target, it will return the location (in X-Y coordinates within the frame) of the largest color blob of the desired color.

- For a set of given obejcts of the same size, closer objects will appear larger (duh!)

So you really have only one question to ask: can you use the camera to look from a given point toward another given point, and determine if the second point is occupied by a certain known object or not?

-dave

[Petey]

Thanks, Dave. I'm fairly sure I understood some fraction of that.

I'm not sure if it completely solves the problem, but it provides some direction, at least. Although - For a given field of view, if two or more objects of the target color are in view and the CMUcamII is requested to locate the target, it will return the location (in X-Y coordinates within the frame) of the largest color blob of the desired color. scares me.

I'll forward this on over to the controls people.

I'm still waiting on the "official answer" from the FRC Q&A forum, for anyone who is breathlessly awaiting it.

[elknise]

During our Kickoff meeting, we brought this up. The CMUcamI, which we used in a different competition last year to track a orange cone, could pick it up at around 30-45 feet depending on lighting, so, this being CMUcamII, it would imply that it might have a longer range. One solution to just finding the colors on the field would be to mount it low to the ground and to angle it forwards, to limit the distractions it can recieve from its sorroundings.

Edit: During the competition, one of the other robots did chase someone from the crowd who was wearing a brightly colored red shirt when it was going for a orange cone.

[Salik Syed]

hmmmm.... maybe they don't want it so easy.....
i'm sure you could program in SOFTWARE to ensure that it is a vision tetra...
i.e after recognizing the green color you measure the length of the green... like move the camera left and right and see how long it takes to "lose" the green ....
i'm sure you could also adjsut the camera to dif light settings based on the beginning of the match... maybe make your robot drive forward to a known positon of color and register that value to calibrate the camera... however if there is ALOT of variable settings like diamond plate reflecting more light near the starting point etc.. some difficulties may come up!

[KevinRo]

Take a good look at the camera documentation. You will find several things that will help you out.

1) Each object is tracked by a bounding box. This is key. Remember it. Its important. Don't forget it! A bounding box is the smallest possible box that will enclose the color object. Consider the difference in view between a person in a shirt and a vision tetra. What can the bounding box tell you?

2) The field of view of the camera isn't as wide as you would think.

3) The percieved size of a distant object is related to the sine function. A good hint: Percieved Height is sort of = Height * (1 / radius). Draw a picture, think about what sine is, and see if you can figure that one out.

Kevin

[Gary C]

i wonder how the sensor would work if a competition was held at chatsworth high school all the lights are ORANGE!!!! hahahahahahah

[Robotics Dude]

Originally Posted by Mark McLeod

Now that'll be a sight to see.
Wait until your first regional and five other robots converge on MOE, hoist it up and cap a goal with you.

That would be something to see. But Id be a little scared when i saw a robot picking up another 120 lb robot without breaking.

[Al Skierkiewicz]

Guys,
Dave has spoken, read and believe, then get to work.
RoboticsDude, we've done that, many times. No damage, dents build character.

[Tytus Gerrish]

Teams Like Us Moe And Element All Have Bright Lime Green robots. im shure if you want your robots sensors to work you should probaly take some time to discriminate dark green from lime green

[Swan217]

Update from the FIRST Q&A. Please take note that these aren't official until they are addressed in an official manual update (as Q&A's are occasionally contridicted or flip flopped).

ID: 981 Section: 5.3.4 Status: Answered Date Answered: 1/11/2005
Q: Regarding <R26>: Will there be any precautions taken--or any precautions necessary--to ensure that the cameras do not become attracted to robots or humans that are colored like field elements (eg clad in red, blue, green, yellow, silver?)
A: No.
ID: 982 Section: 5.3.4 Status: Answered Date Answered: 1/11/2005
Q: Clarifying Q. 981: Even if the intent is not to distract the robot? Since the cameras can be calibrated to sense many different colors, must some colors (& if so, which) be avoided by teams in decorating their robots and themselves?
A: No.

[JBotAlan]

Which mode do you think the cam was in when they had the demo during the kickoff? I saw it followed the guy's green shirt, but his shirt's color is different from vision tetra green. If they had it calibrated for tetra green and it followed his shirt, can it really discriminate between his shirt and the tetra? I still have yet to even power up the cam, but that should happen today.

[dlavery]

Originally Posted by Dan Swando

Update from the FIRST Q&A. Please take note that these aren't official until they are addressed in an official manual update (as Q&A's are occasionally contridicted or flip flopped).

...

Good, because I'd hate not to have our Robot Skirt again this year just because the color of the tetras is ThunderChicken Green.

Just a thought here. Yes it is LEGAL to have your robot any color that you may want. But, and this is absolutely not an "Official FIRST Position" or anything, just an independent thought (I am allowed to have one per week... ), you have to ask yourself if it is SMART to have your robot certain colors. If you pick a color for your robot that you know is very close to the color of certain objects that will be sought by the vision system, what will be the result?

You know that the vision systems of the robots on the field will be calibrated to look for green. If your robot is a similar hue, then you know there is a significant probability that it will be detected by the vision system of the other robots - INCLUDING YOUR ALLIANCE PARTNERS. After about the second time that an alliance partner robot is prevented from grabbing the vision tetra and scoring bonus points because your robot was a confusing color and distracted their vision system, your team is likely to become a pariah. If I were an alliance captain picking teams for the elimination rounds, I don't care how freaking good your robot may be, I would never ever pick you as a partner if I saw you repeatedly screw up the autonomous period performance of your own alliance. The risk to the alliance would just be too great. There are lots of other implications as well, but I think you get the point.

So, once again, it comes down to a simple question: coloring your robot green (or even yellow, the color of the targets in the goals) may be legal, but is it smart?

Originally Posted by JBotAlan

Which mode do you think the cam was in when they had the demo during the kickoff? I saw it followed the guy's green shirt, but his shirt's color is different from vision tetra green. If they had it calibrated for tetra green and it followed his shirt, can it really discriminate between his shirt and the tetra? I still have yet to even power up the cam, but that should happen today.

For the kickoff demo, the camera was calibrated for the color of my shirt. It was able to properly discriminate between the shirt and the green vision tetras.

-dave

[Kyle]

Once everyone is able to use the sensors and see how well they are able to detect different shades or a color is when we will all know if it is truly smart to have a green robot and what shade if green at all. Thank you Dave for clearing up most of this mess about the sensors and how smart it is to use green or yellow color bots. If MOE green does have a negative effect with the sensors then MOE can be mostly black with green on it but we most likely will always keep MOE green on our shirts, robots and everything else that we can get our hands on. Teams that also use green as there color should be able to use another color that is identified with their team. ex. MOE is green and BLACK so using black is still ok because it is still one of our colors. Another question about the sensors, Last year our robot had "kilts" that were colored black and green plad pattern would that also draw the sensors to us or would the non flat color keep us safe?


Lets wait a week or so and play with the sensors and see what we learn then the question will be answered.

[MikeDubreuil]

Originally Posted by Kyle

Once everyone is able to use the sensors and see how well they are able to detect different shades or a color is when we will all know if it is truly smart to have a green robot and what shade if green at all. Thank you Dave for clearing up most of this mess about the sensors and how smart it is to use green or yellow color bots. If MOE green does have a negative effect with the sensors then MOE can be mostly black with green on it but we most likely will always keep MOE green on our shirts, robots and everything else that we can get our hands on. Teams that also use green as there color should be able to use another color that is identified with their team. ex. MOE is green and BLACK so using black is still ok because it is still one of our colors. Another question about the sensors, Last year our robot had "kilts" that were colored black and green plad pattern would that also draw the sensors to us or would the non flat color keep us safe?


Lets wait a week or so and play with the sensors and see what we learn then the question will be answered.

I think what Dave is saying is that if you decide to use green on your robot you are playing with fire. Honestly, most people know MOE's traditional team colors but nobody is going to care and will become irate with you if you mess up their autonomous mode. Quite frankly, if you keep a green color on your robot knowing it might jeopardize someone's autonomous mode you are being completely selfish.