I represent my one of closed projects.
This projects is to study simple object recognition to use significant concept.
Here is the procedure to recognize object.
Take informations (covariance matrix including average and variance of pixel) of a background.
Take information of objects in front of the same background.
Get Mahalanobis distance of the covariance matrix of the background
Classify between background and object using SVD
Reduce noise using blob coloring
Get histogram of each classified informations
Recognize object to use bhattacharyya coefficient
You can watch register three different objects and detect it.