Human Computer Interface V2 (virtual touchscreen)

Using my laptops web-cam and the opencv framework I have created a virtual world allowing interaction with both virtual objects and controllers.

The demo allows up to 3 fingers to be detected according to color (red, green, blue).

Written in c using opencv, I created a virtual 2D controller out of each finger and allowed it to interact (with a small phisycal mass with other virtual objects.

The demo also demonstrates the 2 button modes I have created, On-Off button and state buttons.

Next I would-like to loose the color detection and allow a real finger detector, I have achieved this only when there is no hand/face interaction. So more work needs to be done.

Using 2 web-cams I have been trying to find the z factor for each finger allowing real 3d interaction. Again more work needs to be done.

Enjoy

Category:

Science & Technology

Tags:

• opencv
• virtual touchscreen
• virtual touch screen
• virtual mouse
• bubbles
• bubble game
• computer vision
• web cam mouse
• educational
• technology
• human
• computer
• interaction
• mouse
• game
• screen
• touch
• air
• virtual
• touchscreen
• color
• segmentation

[Shaunakde]

Hi,

I am trying to use color based object detection too (Using Aforge libraries on c#). I am applying the HSL filter, but I notice that HSL calculations take way more time (and make my video stutter)

Do you notice this problem in OpenCV? Or do you avoid the HSL colorspace altogether?

[amizorach]

@Shaunakde Hi

I have not used the HSL algorithm as I needed better results

I used a combination of spectrum changes.

I have however experimented with the HSL in open CV and found it to work fast enough

[Shaunakde]

@amizorach

Hi, Thanks for the reply. I will give OpenCV's c# wrappers a try.

I did not quite understand what you meant by "combination of spectrum changes". I am very interested in finding out more about your technique. Could you please direct me to relevant documentation/papers? Or perhaps outline the method (if its Ok with you)

[amizorach]

I will try to elaborate when I have some more time

But the main idea was to simply find the point of interest (according to colour) and then try to approximate where it might be next (movement). this allows me to scale down the entire frame into an expected block and run a more detailed search with a smaller variant on the smaller rectangle.

[amizorach]

@Shaunakde I also used a feature I was developing to segment the picture and I was hoping to be able to parse the moving objects so I can follow them even if they are a bit covered (a hand moving in-front of the another one) although I was able to get a pretty good response when using the colour markers I struggled when it came to regular hands and differentiating them from the face.

[Nikola98Jakovljevic]

My name is Nikola Niki Jakovljevic (Gta San Andreas Ivanjica)

[Nikola98Jakovljevic]

Or send me link on my Facebook profile on wall

[Nikola98Jakovljevic]

Please link

:D

Please, Please, Please

My Friend How To Make This Program?

[Shaunakde]

@amizorach Thanks for the reply. I have somewhat of a clear idea now and am beginning to understand the algorithm. Thank you! :)

[Shaunakde]

@amizorach Thanks for the reply. I have somewhat of a clear idea now and am beginning to understand the algorithm. Thank you! :)