Light RGB Sensor

[knine143]

Hey all

I need a light sensor that can detect RGB values or recognize a specific RGB value. This will be used for detecting an orange cone.

This sensor looks pretty good: http://www.sparkfun.com/commerce/pro...oducts_id=8618

The problem is that the evaluation board needed to hook up the sensor is out of stock...

Are there any other sensors out there that could do the job? What do you recommend? Also, we are connecting the sensor to the FRC robot controller, so it would be nice if it sent analog data.

Thanks

[Uberbots]

Why not just use a CMUcam? you already have those prepackaged with previous year's kits...

if not id order the sensor anyway. half the fun is figuring out how to wire it.

[whytheheckme]

Originally Posted by Uberbots

Why not just use a CMUcam? you already have those prepackaged with previous year's kits...

Indeed. Beautiful piece of equipment.

Mind telling us a bit more about the project? *says the curious programmer*

-Jacob

[SL8]

Originally Posted by Uberbots

Why not just use a CMUcam? you already have those prepackaged with previous year's kits...

if not id order the sensor anyway. half the fun is figuring out how to wire it.

I believe there is just a little programming complexity there.

[DonRotolo]

Just be aware that the sensor you linked to is small and not an imager. That is, it essentially senses one pixel. So, if you wanted to "find" an orange cone, you'd really have to so some fancy mechanical tricks so that it pointed right at the cone.

The CMUcam, on the other hand, is a camera. It will tell you the size and centroid of an object of a certain color within its field of vision. This can be very useful for locating objects.

Don

.

[knine143]

Originally Posted by whytheheckme

Indeed. Beautiful piece of equipment.

Mind telling us a bit more about the project? *says the curious programmer*

-Jacob

The project is called RoboMagellan. We are given GPS coordinates of an orange cone and have to autonomously navigate to it avoiding trees and other obstacles along the way including varying terrain. There are also bonus cones that can be touched to score higher.


I was thinking about using the CMUcam, but my team has never done this before. Would i be able to easily program it to know where a cone is? With the light sensor, I was thinking to just have the robot rotate in place when it knows its near the cone and once the sensor detects orange, it will know that the cone is in front of the sensor.

[DonRotolo]

The CMU cam can be taught to recognize a "color", anywhere in RGB-space, and can also be given 'tolerance' parameters to account for some lighting or color variations.

Instead of having your controller process image data (ouch!), the CMUcam does this for you, providing signals that say "I've found that color", "This is how 'big' the color is" (i.e., how much of that color is in the field of vision), and "this is the center of the 'blob' of color". Exactly what you need to 'see' a cone from 30 feet or more and go to it.

Programming it is also not difficult, but you are well advised to read all of the instructions explaining it. Also search here. The part most folks have the most trouble with is calibrating the camera - setting the color and tolerance, so that the object is found, but only the object and not every stray cat that runs past the robot. This takes a little tweaking by trial and error. The LabView calibration routine makes it a LOT easier to do well.

Your color sensor is meant to have light shine upon it. Unless the cone glows brightly, and you have a good lens system to focus the light from a small area (don't want it 'seeing' a 45 degree field of vision...) the RGB sensor you propose will not work.

Don