I'm trying to measure a persons height using the Microsoft Kinect and LabVIEW. Ideally the Kinect will be in a fixed position, about 2m from the ground about 500mm away from where the person is going to be standing.
I've tried this with just a normal Webcam, but the background won't be clean (could be other people standing nearby), and the webcam couldn't differentiate between the person we were trying to see and the other people a bit further back - which is why we thought about the Kinect.
I've looked at the various examples available for the kinect but I haven't yet found one that is able to calculate this. Does anyone have any ideas?
If the height of the camera is fixed, all I need to do is detect the object (person) who in front of the Kinect, and then where the top of their head is in relation to the known height of the camera and the number pixels from the bottom of the image? ...in theory!
Welcome to the NI Discussion Forums, thanks for posting.
As far as I am aware, the Kinect Sensor has a hard time calculating the distance of objects that are very close up but the examples that I have seen display the depth information as a coloured pixel map, where red means the closest distance, from this would you be able to mask the object of the certain hue, and base your calculations off of this?
Quite possibly, I will give it a go - although I don't have a Kinect available to me right now.
I've read in some documentation (although I can't remember where), that the Kinect has a minimum depth of 500mm. I'd like to get as close to this as possible, but could go a little further away if it helped disguinish the objects.
Would I be able to clearly detect the top of that pixel map though? Or would it fade out? We need to be measuring the height to within 10mm ideally.
I've been reading some articles on distance sensing with the XBOX Kinect, a particular website caught my attention: http://mathnathan.com/2011/02/depthvsdistance/
It has some images which show a definite distinction between objects, at different depths with a clear edge (albeit a little distorted). Have a look and see what you think, personally I'd say that the distance calculation via camera is rather tricky, and trying to calculate height based on this is adding another level of complexity.
You may however be interested in a LabVIEW Project from Leeds University: "Kinesthesia" (https://decibel.ni.com/content/docs/DOC-20973) which comes with a toolkit to perform similar tasks.