View Full Version : Propeller Powered Rock Band Robot - Video analysis application question?
I am looking for input for a project I am starting, and have been reading up on the rock band robot that is prop powered thinking maybe I can learn from their work with the video analysis.· However, I may be looking for more then the prop can provide here...
What I am trying to do...is detect a trailer of yellow corn from the top and get the x/y value of the sides.· Basically, I want to have a camera looking straight down at the top of the trailer and detect where the trailer is at so that I can move a 'probe' over that part of the trailer and put it into the corn for testing.
I am open to all ideas here that anyone has.· I am thinking video might be the best right now, but maybe ir or another type of sensor would be better?· I would be open to using a PC and software also, if someone knows of a software out there that does this kind of thing?
Just a thought,
If looking down from above, could not an ultrasonic ranging sensor (Ping type sensor as they call it hereabouts) on an x/y servo mount detect the edges of the trailer?
Logically, the trailer is Much closer than the ground it stands on.
This kind of setup is used by many prop powered robots to find out about their environment.
Of course it may take a few seconds using stock servos to scan the 'field of vision' to find the edges, but assuming time is not an issue, the maths to find the center of the trailer is trivial and well within the bounds of propeller capability.
This of course assumes that all the trailers you see contain corn since ultrasonics have no way to tell 'what' they are looking at, corn, yellow or not http://forums.parallax.com/images/smilies/wink.gif
Search Propeller Forums (http://www.google.com/cse/home?cx=005038179529894410695:-i_2f0zjlbm&hl=en) easily thanks to StefanL38 (http://forums.parallax.com/member.php?u=47171)
Absolutely, the Prop has the power. On the imaging side of the issue you have a couple of choices. A while back Hanno developed a six bit?, nearly real time imaging system built around a $20 camera and a fast analog to digital converter. He then got occupied with other stuff. The hitch is that the a/d chip is smd only... Hanno tried to hand the product off, but somehow it is lost in translation. I talked to Nick from gadget gangster about getting something out that could be used by mere mortals and he seemed interested ... but he is also horribly polite.
I have been working on using a TSL1401, which is a line scanner... it can also produce data in real time, but I am currently stuck around 50K pixels per second. This is probably fast enough. You would need two, because the chip is limited to 128 in line, 12 bit pixels, and the effective resolution is determined by the distance of the sensor to the object being sensed.... so, you would want one sensor watching the truck... picking your coordinates for the other sensor, which would guide your robot to an ear of corn and not the space between. You can get effective color resolution... which greatly aids in object detection by simply putting an led on your corn sensor. This doesn't sound right, but it does work.
I didn't know that there was a Proprock band robot... hmmm does it have an agent, does it accept bookings for beer fests?:)
Great... I think...
rjo_:· I found the propellor rock band under one of the contests.· Here is the link...
I know this isn't what I need to do, but it was giving me a start into researching how to input video.· I would love any more info you could point me towards in reguards to Hanno's project.· I will keep searching for anything I can find.
As for the ping on a servo,·I did not think about that idea but it probably will not work in this application because of speed and it will be out in the elements.
I have attached two pictures as more of a drawing what I have in mind.· Currently there is a probe there that an operator has to sit and control for every truck.· It basically just has a joystick type control to run the unit.· I am either looking to control the existing unit that is hydrolic, or build my own system that is driven with large stepper motors.· In the pictures you can see that not only do i need to detect the sides of the truck, but need to watch out for any cross peice of metal going accross the top.
Hopefully the pictures will help show a little more what I have in mind.· Perhaps you have a completely new idea of how to do this and maybe I am over-complicating it with video.
PS....Try sending an email to Adam Stienecker and ask about booking the rock band for a beer party.· If you can get a tour schedualed, I will provide the first keg when it stops in South Dakota ;)
I am looking at the TSL1401 and thinking I will order at least 2 units to 'play' with on the bench. I am thinking that in the final project I would need about four? Perhaps so they can be about 10' above the truck and around the four corners area? I am guessing I should be able to detect the yellow corn to help find where the sides of the box would be. By doing this, I should also be able to detect where the metel cross peices of the trailer are at and stay away from them?
Thoughts everyone? Other then I may be in over my head...lol, but I feel pretty good having everyone on the forum to help me learn this type of sensor and how to impliment it....
05-19-2010, 02:42 AM
If the corn is the only yellow thing in range of your robotic arm you might use Phil's ColorPal to detect by scanning the arm over an X/Y grid of interest and searching for perimeter color changes to find the center of the box. If the color you want to detect changes, seems it would be a small code modification.
May the road rise to meet you; may the sun shine on your back.
May you create something useful, even if it's just a hack.