CMUcam1

Half-Lucas

Hello all, guess what I got for Christmas? A CMUcam for my Boe-Bot!

So this leads me to a new mini-project. Having the Boe-bot navigate the house using only the Camera. This means avoiding objects as well as drops such as stairs and edges. I have an idea of how the camera works, after reading the different commands I can issue to the camera, but I still have no clue of how I can incorporate depth perception...any help?

Franklin

Depth perception will be hard to do with only one camera. I would start by having fun with the camera and using other sensors to watch for the dangerous stuff then after a bit you may need to build some device that either moves the camera side to side or uses mirrors to get two points of view the camera can compare. Most of all have fun.

▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
- Stephen

Bob and Herbert

You might also be interested in computer vision software at: roborealm.com

Half-Lucas

Franklin said...
Depth perception will be hard to do with only one camera. I would start by having fun with the camera and using other sensors to watch for the dangerous stuff then after a bit you may need to build some device that either moves the camera side to side or uses mirrors to get two points of view the camera can compare. Most of all have fun.

Is it possible to use IR sensors with the CMUcam? I thought the camera takes up all of the pins...
I'm using the Board of Education USB rev B, so there is a special pin set (9x2) for the camera. My breadboard is free, but aren't all the pins used for the camera?

sdy

I have a CMUCam on my bot, but not a Parallax one. I use one line serialout cmds. Depth Perception can only crudely be done by checking the target size. Part of the T packet return is the top-left & bottom-right x,y coords of the corners of a bounding box around the target. You get a feel for how close you are by the size of the target.
Get a real good mabnual on the CMUCam here:
http://www.ifirobotics.com/docs/legacy/cmucam2-documents-and-software1.zip