Next steps
22 May, 2007 - Post a comment - Permalink
We had a meeting today to discuss the next step for the depth vision project I will be doing this summer (RGBZ camera). It was... fascinating. There were a lot of ideas thrown around as to what we could do with the potentials that this lovely camera possesses. There were talks about animating a virtual person that mimics the body pose of a person standing in front of the camera, pinpointing the location of the hands and moving the hands of the virtual character accordingly.
However... I want to start with just getting some kind of output from the camera and call that pretty good progress. I think that baby steps are necessary for this kind of project and to make sure that you are not overwhelmed by the whole thing. This is going to be a very complex project and I'm being very careful in taking the smallest steps I can. Then you can speed up the steps and before you know it you have a working product to show.
The first thing I want to do is, as I said before, get some kind of output from the camera and then assess exactly how accurate it is.
Lets imagine this. We've got two cameras. One is the depth camera and the other is just a regular RGB camera. We have the cameras pointed at a person and lets say that we wanted to only show the pixels, from the RGB camera, that the person occupies. If there is noise around the person, a glow as it were, then the numbers that the depth camera gives us are going to be skewed and we'll end up with a big blob that contains the person but also some of the background that was behind the noise that was present in the image that the depth camera saw.
Lost you? No? Good.
Let's just hope that this camera is a precise as we hope it is.
However... I want to start with just getting some kind of output from the camera and call that pretty good progress. I think that baby steps are necessary for this kind of project and to make sure that you are not overwhelmed by the whole thing. This is going to be a very complex project and I'm being very careful in taking the smallest steps I can. Then you can speed up the steps and before you know it you have a working product to show.
The first thing I want to do is, as I said before, get some kind of output from the camera and then assess exactly how accurate it is.
Lets imagine this. We've got two cameras. One is the depth camera and the other is just a regular RGB camera. We have the cameras pointed at a person and lets say that we wanted to only show the pixels, from the RGB camera, that the person occupies. If there is noise around the person, a glow as it were, then the numbers that the depth camera gives us are going to be skewed and we'll end up with a big blob that contains the person but also some of the background that was behind the noise that was present in the image that the depth camera saw.
Lost you? No? Good.
Let's just hope that this camera is a precise as we hope it is.
Post your comment
Feel free to share your opinion.
Comments
No comments posted yet...