Archived

This topic is now archived and is closed to further replies.

Khaos

Image Algorithm

Recommended Posts

I saw a special on TechTV about this guy''s invention. It prevents his cat from bringing in dead animals. Here''s how it works: There''s a small box or house that the cat walks into near the main house door. There is a white background and a strong white spotlight that projects a circular spot of light on the white background. The cat walks through and right when he is in the spotlight it triggers a digital camera to take a shot of the cats cross section or silhouette. This image is then compared by a computer with a "normal" cat image that''s the default image of the cat without any animals in his mouth. The computer compares the images and if it detects the cat has something in his mouth (something in the image that shouldn''t be there, or isn''t at all like the default image) it won''t let the cat through. Otherwise it opens the door. So anyways, how would the algorithm work to do this? It is loosely comparing two images, note the loosely aspect. There is nothing exact about it. And secondly, how does the algorithm work that will take an image of any colored picture and break it down into black pixels and white pixels? There seems to be two of them. One that produces more of a solid black and solid white parts, and one that produces more of lots of little black and white dots. Any ideas? If you can just point me in the general direction on any of these topics I''d appreciate it. Thanks!

Share this post


Link to post
Share on other sites
I would guess he''s using some sort of neural network to compare the two images (check out www.generation5.org for some good info on NNs). A properly structured NN is capable of doing this type of loose comparison.

As for colored to black and white images, read through the following thread and I''m sure you''ll find what your looking for.

http://www.gamedev.net/community/forums/topic.asp?topic_id=127864

digital radiation

Share this post


Link to post
Share on other sites