Mouth Shape Detection using Raspberry Pi

asked 2017-02-01 22:04:45 -0500

Hi Folks. I am working on a personal project and I need some advice or pointers on what is the best way to approach this. The concept is akin to James Cameron's Performance Capture where a small camera is mounted in front of the wearer's face. In my situation the camera can only see the mouth as the other regions are covered. Just under the nose will be a reference point that all other co-ordinates can be based from. I am using the RPi and it's camera for portaility as this needs to be self contained and portable. What I want to do is to capture the lip shape (being a semi symmetrical polygon ) and then map various points around the mouth so I can gain an X/Y co-ord from it. This will then be used elsewhere to drive servo's

Ok - what is the best way to do this. Will Haar's pick up the mouth if the rest of the face is not visible. Should I put a series of dots around the lips at points that I am interested in and then work out how far it is from my reference point. Should I use colored points where blue is the reference points, red is the extremes of the mouth, and green are points around the lips where I want to read additional points.

Any suggestions would be greatly appreciated as I am fairly new to OpenCV. Many thanks Dave

edit retag flag offensive close merge delete