Do I need to write a logic of pattern recognition (classification of objects) or OpenCV already have such a function?

edit

OpenCV is a wonderful thing my friend :)

Take a look at the Machine Learning section of the library, it contains implementations of many state-of-the-art machine learning and pattern recognition techniques, you don't need to reinvent the wheel. And the interface is pretty simple: MachineLearningMethod::train() and MachineLearningMethod::predict()

You can find implementations for:

• Statistical models
• Normal Bayes Classifier
• K-nearest neighbors
• Support Vector Machine
• Decision trees
• Random forest
• EM
• Neural Networks
• ...

It doesn't work only with face recognition, if you provide the training data you can learn anything.

Happy pattern recognition!

Here are a couple of tutorials to start:

1. Face and eyes detection.
2. Detection of a known planar object.
3. Template matching.
4. Detecting lines and circles.

So, you can see that there are a number of methods for particular tasks. And Martin is right, you should also study the opencv_ml module.