Hi there, I am not sure whether i can ask this question here as this forum is more towards game AIs. Is there any links or url or article regarding of doing a image processing for hand palm? I will use a normal scanner. Well, simple thing is how do i differentiate a left or right hand palm? How do i recognize those lines in my fingers and thumb when comparing to other hand palms? Any idea? Thanks.

I think image processing would be the most relevant discipline here. You could perform edge-detection (look up sobel and laplacian operators) to find the lines, and some sort of erosion or skeletonisation to get these into a more easily-processed form. (some notes on this here, although it uses some other software. You can look up the key terms they use though.) From that point you might just look at the relative angles to determine whether it's a left or right hand. I don't know what else exactly you want to determine, but I guess that analysis of individual lines may be a little more difficult.

[ MSVC Fixes | STL Docs | SDL | Game AI | Sockets | C++ Faq Lite | Boost
Asking Questions | Organising code files | My stuff | Tiny XML | STLPort]

[edited by - Kylotan on February 13, 2004 7:25:39 AM]

This is related to AI as it deals with cognition.

Your biggest problem might be your scanner as it might not provide a good enough source image. Ideally, you want the handprint itself to be isolated from any background ''noise'', but this could prove difficult if any external light seeps in around the edges of the scanner lid. I''d suggest doing some experiments to work out the best way to get a good clean image.

As for the ''handedness'' detection, there are some techniques you can use for this. First of all, you could tweak the contrast and brightness of the image and perform an edge-detection process. You don''t need a full 24-bit colour image to work with, so downsample the image to 1-bit/pixel. That''ll give you a pure black and white image which you can then look at for patterns. The edge-detection processing should throw the lines on the palm into sharp relief, making their processing easier.

Most people''s palms have a line that runs in a curve from the wrist up, round the base of the thumb, and over towards the index finger. Find that and you can use this to work out which hand you''re looking at. Other lines and common features can be examined in a similar way.

HTH,

--
Sean Timarco Baggaley

Are there any sample apps on this?

I''m not clear on what precisely you wish to do. Do you want to: 1. Differentiate between left and right hands? 2. Identify/verify individuals based on their palm print? 3. Identify/verify individuals based on their hands, possibly using something other than the palm print, such as fingerprint or hand geometry?

-Predictor
http://will.dwinnell.com

Is hand geometry unique? I mean everyone had a different hand geometry?

i would go for 1st and 2nd.

quote: Original post by wenching
Is hand geometry unique? I mean everyone had a different hand geometry?

i would go for 1st and 2nd.

I don''t believe hand geometry is unique- at least, I don''t believe it has been feasible to identify individual by it- verify, yes, identify, no.

Differentiating between left and right hands should be the easiest of the tasks above. I would think that it would be very easy to use even a simple linear or quadratic discriminant to do that.

-Predictor
http://will.dwinnell.com

I''m not sure about hand scanning, but there is a ton of information about scanning fingerprints on Google.

Look in to a Sobel edge detection. It''s probably the most useful for what you want to do.

Best regards,
Will

You might also want to investigate contour tracking ... your problem is a simplified instance of the contour tracking problem, where the contours are not moving in time. That should make it a hell of a lot easier!

Timkin

[edited by - Timkin on February 15, 2004 9:13:49 PM]

why distinguish left and right hands, if one person puts in their right, then their left, they are going to be different lines and size aswell, just have the person scan either hand and put it under their name, and let the scanner find which hand