Webcam-based object recognition

Using the color information from a webcam feed to maneuver geometry inside of Rhino. Once again relying heavily on Andy Payne's new vision components in Firefly. The 3D projection techniques (especially the depth) are only approximations right now, tweaked to look right using Galapagos and a little patience. Still working on the math to make it properly accurate....if AR Media can do it, I can too....

The rotation didn't really show up much in this video. Usually works a little better. Also, the red box's form is not derived from the green square. The script does roughly translate the square into Rhino geometry, but I'm only using it as a marker for the webcam to recognize.

8 members like this

Get Embed Code

Tags:

Comment

You need to be a member of Grasshopper to add comments!

Comment by Minsoo Thigpen on January 14, 2015 at 1:36pm: Hey Scott, I saw that this was a while ago but I am trying to project a geometric object in rhino based on live-feed of a graphic indicator (or just a specific color) from the webcam. I am using the latest version of firefly add-on but the most I can figure out is picking one RGB value and desaturating the rest, then setting a point to go to the average of that one color in the whole video frame.

However, even if setting up the video, I put a white screen behind me and only hold either a red, green, or blue object -- it's still rudimentary at best since red, green, and blue tend to be everywhere and this method is not very specific. How did you achieve isolating only the green square?