eye controlled wheelchair

[dheek]

thanks a lot klemen !!! that was really helpful. 

thank u loads..

[dheek]

hello klemen,

I am using Labview 2012. The attched file doesnt run on Labview 2012. what to do? 

[Klemen]

Here you go. You can also use the conversion board:

http://forums.ni.com/t5/Version-Conversion/bd-p/VersionConversion

Best regards,

K

[dheek]

hey klemen,

i am having problems with opencv.

Cant i perform a specific action after pattern matching by using case structures and all? 
is there any other method, other than HAAR transform in labview?

any method that would divide the eye into quadrants and determine whether its on the positive or negative side of the axis, by keeping the origin at the center?

[Klemen]

Hello,

there are many methods.

Search for example:

Real Time Eye Tracking and Blink Detection with USB Cameras (template correlation)

or

Real Time Eye Tracking and Blink Detection Using Low Resolution Web Cam (optical flow)

or

etc...

Best regards,

K

[dheek]

hello,

i am getting results on this VI. but its not exact.

the lower part of the program is for the bounding box. the upper part values is for postion determination.
I have been roughly using some values(in comparison during "greater than" and "less than"). they denote the pixel values right? I had used boolean led's to determine if there the eye is getting determined in the right direction or not. But its not giving accurate results.

someties the 

there is no bounding box coming when i tried pattern matching later. Could you tell me what might be the problem?

How do we specify the ROI in an image thats being continuously captured through webcam?

[Klemen]

Hello,

like I said before, I think pattern matching is not an appropriate way to go for eye detection. Maybe if you select the initial template for the detection and learn new template for each iteration (each acquired camera frame) or each n-th iteration (this can only be determined by testing), the pattern matching would work. This way your template gets updated, increasing the roboustness of the application. BUT, I have not tried this!

"How do we specify the ROI in an image thats being continuously captured through webcam?"

You can specify ROI manually (drawing the ROI directly on image) or programatically. Both are done through the ROI "property node". Right click on image (block diagram) -> create -> property node -> ROI. Check an example here:

http://www.ni.com/example/30249/en/

Again, you can take a look at the (upgraded) example of iris detection here:

https://decibel.ni.com/content/blogs/kl3m3n/2014/02/24/real-time-iris-detection-in-labview-using-ope...

It uses combination of OpenCV (face and eye detection) and Labview (iris detection). Using this method you could determine the location of the iris in the eye ROI and thus determining the direction of the gaze.

Best regards,

K