How can I identify items in the real world?
Theoretically, how can I tak开发者_如何转开发e a picture of a room and identify items in the room? Let's say we have a table and a book on the table. Is there any way to identify them?
Lets say you have a book on this table sat next to a pile of papers which happen to be of the same dimensions? Or perhaps a larger encyclopedia? Define a book to a computer.
This is an incredibly difficult problem and certainly not something simple to answer. Where I am based there is extensive research going on into identifying things on a typical road - which is incredibly difficult.
Personally I would have a look into the Kinect SDK that Microsoft intend to release in the summer, I have no doubt people will try to use this to help develop some more frameworks, but remember that part of the problem is going to be differentiating between different items of the same dimensions. Non trivial.
Edits: Making sure it looks like I know how to type :P
Pattern matching with many different samples of the objects from all possible angles. The more images you have to analyse regions and shape the better your results will be.
one could use SIFT-Features if the items are known, have a look here: http://www.cs.ubc.ca/~lowe/keypoints/
See this project (especially the YouTube video) on how depth information (and therefore object separation) can be achieved with a video source.
It would be incredibly hard with a single picture due to there being little depth information. Binocular vision would add some depth information (i.e. a pair of images taken from two different angles but of the same scene). The above project uses the changes in the scene over time to calculate depth.
精彩评论