Society of Robots - Robot Forum

General Misc => Misc => Topic started by: dmcb2009 on November 08, 2013, 11:56:20 AM

Title: Uni Project
Post by: dmcb2009 on November 08, 2013, 11:56:20 AM
Im new to all this robotic stuff and to be honest im a bit lost in it all.
So for uni I have to do a project where a machine has to electronically differentiate between a tennis ball and an orange golf ball and then this machine has to harvest all the tennis balls.
The balls are being placed in a straight line at various intervals and this machine is to be mounted to the back of a tractor.

I'm fine with creating a mechanism for lifting the ball but the problem I have is that I have no idea how to differentiate between the two balls and what sensors to use ect.
Another requirement is that the golf balls must not be disturbed from there original position.
Any ideas or help is much appreciated, Thanks.
Title: Re: Uni Project
Post by: Roman505 on November 08, 2013, 03:17:33 PM
Hello dmcb, I am interested to know what are your thoughts about possible sensors? Given the parameters, would you go for touch, distance, reflectance or width recognition or do you have other thoughts so far?

I read that you can not disturb the golf balls but you know that the balls will be in a neat line and they have differing heights and are spherical in nature so, have you looked up specifications of close range distance or reflectance measuring such as with IR vs ultrasonic? Is colour a useful option or are orange and yellow (the tennis balls are yellow?) too close? If this is a uni project then presumably you have access to sensors for testing. Have you tested measuring a golf ball and a tennis ball from a height a short distance above a tennis ball? Do spec sheets suggest one option might work better? Web sites such as pololu group sensors on the site and offer ready comparisons.

Your description is unclear on whether you are supposed to follow the line automatically or whether the tractor is guided by a person. This will affect whether you need line following (about which much can be found here) and the extent to which you might need arm positioning. What more do you know about those please?

Please forgive my slightly didactic tone. Learning by researching and doing, doubtless why you are asking and searching here, is something I favour, hence I (at least) am denying you a pat answer despite liberally strewing clues and options about the place. :)

I will be interested to know on what you finally settle and I am sure you will find help here if you run into implementation problems further down the track, if not different clues from other people now.
Title: Re: Uni Project
Post by: dmcb2009 on November 08, 2013, 05:10:51 PM
At the minute I have been thinking of using either some sort of reflectance sensor or a colour sensor of some type, but as I am inexperienced with all this stuff I really don't know which to go for.

I have read a bit on the IR sensor but I don't know how close to the balls the sensors would have to be and also don't know if it would be possible to use the ir whilst the tractor is moving. Can the IR be used for some sort of texture recognition ?

The compact tractor will be driven by a person so isn't going to be as accurate as a machine with auto steer. If I had an idea of the ranges that these sensors operated in id be able to determine if the tractor driver could drive accurate enough to allow the sensors to work.
Thanks
Title: Re: Uni Project
Post by: Roman505 on November 08, 2013, 06:28:48 PM
At the minute I have been thinking of using either some sort of reflectance sensor or a colour sensor of some type, but as I am inexperienced with all this stuff I really don't know which to go for.
Frankly, neither do I know for those objects, hence my suggestion of testing. There is a combination of reflectance and distance differences between the two objects of interest. This may make them very readily distinguishable without too much fuss about precise distance, provided they are within range of your sensor and approximately lined up with it. Do you know what will be the background surface, by the way? Will it be constant? Measure that too, during your testing. You will need to distinguish it (partly for positioning).
Quote
I don't know how close to the balls the sensors would have to be and also don't know if it would be possible to use the ir whilst the tractor is moving.
Well, they are used on moving robots. :)
Quote
Can the IR be used for some sort of texture recognition ?
Yes, see reflectance above. In fact, one of the issues with IR is precisely the different reflectance from different surfaces or materials. For example, try pointing an IR camera out of a closed window. Pretty much all you will see is yourself. This might be to your advantage, a golf ball being glossy and a tennis ball not.
Quote
The compact tractor will be driven by a person so isn't going to be as accurate as a machine with auto steer. If I had an idea of the ranges that these sensors operated...
Widely varying. Check out pololu or similar sites. They tell you quickly what are the distance options for sensor types and for models within sensor ranges.

Consider also the possibility of two angled sensors rather than one.
Title: Re: Uni Project
Post by: jwatte on November 08, 2013, 08:30:59 PM
The recommendation on the Trossen Robotics forum for this same request is to use a webcam with the OpenCV library and a small computer of some sort (like Raspberry Pi or Odroid or whatever.)
I'd just like to cross reference here, as it seems to be the same problem.