I'm planning to implement user interface through color/fiducial markers in fingers in Android..
Can anyone help regarding this?
At its least level black marker on white bg is sufficient which passes its axes values to the corresponding function in Android interface class.
Pls help me about the user interface class by which i can do this...
If the shape of maker in the input frame is quite apparent you can use the below solution.
In the first you have to prepare the template image by the following preprocessing steps:
1-Use the sobel derivatives then opening the result image.
2-In this state you can compute the region of blob(ROI) then expand this region.
3-Crop the ROI then resize it to specific size .
4-Apply the threshold on image then compute center of mass .
5-Apply the canny algorithm over the image.
6-By cvLogPolar & center of mass point transfer the image to polar space.
In the enrolment step for all of the markers that you have to recognize them applay the above preprocessing step & save each template in database.
In the test step Apply the preprocssing on input frame then use cvMatchTemplate for match of test sample by all of template images in database. Each template that have the greatest peek is the answer of problem.
This is so helpful !! Thank u so much :) Pls clear this doubt for me... My ultimate aim is to control/provide interface to android OS by replacing touch-activities by the position of the marker from the camera feed... For Instance to zoom i have to move the markers(attached to fingers) away... Lets assume that i've found the co-ordinates (x&y)center of mass of the marker after these processing steps.. Ater this how do i pass these values so that i can use it as an interface.???..
Asked: Dec 30 '12
Seen: 1,160 times
Last updated: Dec 31 '12
