We use a chroma key solution now in our cycling aerodynamics tool, but we want to get rid of the green screen in our product. My idea is to use depth filtering using a fixed depth camera. I bought the R200 and SR300 to do some experiments.
The main purpose of our application is capturing the frontal area of a cyclist on a static bike. The subject can be placed in such a way that there is at least one metre of empty space between it and its surroundings. I would like to use one or a few static cameras and filter the data based on depth, basically having a chroma key without hard color and lighting requirements.
Would the current RealSense cameras be a good match for this setup? Is there anything related to depth filtering built into the SDK, or would I get to work using the RAW depth map (probably filtered for stable results)? If not possible yet, will the next generation (R400) be better for this? Or should I start looking at stereo cameras if I want stable results?
Looking forward to get to work on this!
Here's the filter options from the SR-300's data sheet.
The full data sheet PDF can be viewed here.