My personal blog: Currently focusing on AI methods applied to Kinect sensor data.
Welcome back, are you working on anything new?
Hi EddHeadI'm working on something new, yes. But it's not related to Kinect. But I hope to have this finished soon and to get back to Kinect.Regards,Stefan
Hi Steffan, I wonder if the method that you used to find the fingertips, it was you who created or already existed in the literature, because I'm doing a school project and would like to reference.
Hi JohnI guess there are quite a few papers that describe fingertip detection (like this http://www.iaeng.org/publication/IMECS2009/IMECS2009_pp489-493.pdf )But I must admit that I had not read any of them when I implemented my version. So I don't know if there is one that does the exact same thing.My method consists of some already existing and described algorithms though (k-means clustering, graham scan...) Regards,Stefan
Hello, stefan. I was referring to part of the algorithm that checks whether the points are aligned in http://blog.candescent.ch/2011/03/finger-tip-detection-part-2.html.
Hi JohnWhat I can say is that I've had that idea myself and that I don't know another project that does it like this. Most use a curvature algorithm:http://en.wikipedia.org/wiki/CurvatureRegardsStefan
Hello Stefan, I have a question, why didn't you use a curvature algorithm?
Hi JohnGood question.. as I said, I didn't read many papers, so I didn't know that I could use a curvature algorithm.The only thing I came up with was this simpler solution.Guess I'll have to try if the curvature algorithm would improve the detection.Here's someone who did this already:http://www.42spikes.com/post/Kinect-Hand-Tracking-Resources.aspxBut there seems to be no source available (yet?).Regards,Stefan
Hi Stefan,I just implement the k-curvature algorithm and it works very well, the detection accuracy is very similar to your solution.
Hi JohnSounds good! Guess I'll give it a try too when I find time!Regards,Stefan