Open DominikHorn opened 7 years ago
One image is enough for the edge detection algorithm to work, but I think it would make sense to average the result over a time of about a second or two. So a capture of about that timespan should be enough. This is the video I mentioned: https://www.youtube.com/watch?v=uihBwtPIBxM
I've conducted some testing and it turns out that the Kinect v2 will only capture color frame data at 1920x1080@30Hz. Uncompressed, this amounts to roughly 240MB of data per second. As there is no easy way to persuade it to capture less data: Should I spend time implementing video compression or is it sufficient to use one or more captured images to implement our fancy algorithm? @sschaeffner @hoelzel
These are what low light pictures taken with kinect v2 of edges on the other side of my room look like:
Given the overall modest precision of the kinect, using image recognition should not be our main implementation issue.