Object recognition on OpenCV

Hello,

We need a new software solution to perform object recognition using opencv computer vision library.

The software needs to accept a jpg image, compare it with a previously trained database of images and return matches from the list of trained images / objects. The system needs to handle hundreds or even thousands of trained images and perform matching on them in reasonable time. We need the ability to organize the trained objects into "collections" and allow matching against a preset collection only. For example, find matches of a given image among "cars" or "corporate logos" etc. and return list of labels.

The application needs to work from the command line. It needs to accept input parameters for file name, label to be trained etc. When performing matching, the solution should output match results to a text file or standard output. Feature points (descriptors) of trained images should be stored in an easy to access/manage database.

requirements:
- Linux or Windows platform
- command line interface to train new images and assign to new or existing collection
- command line interface to match images against all or subset from database
- open source database solution preferable

In your proposal, please include the suggested database solution and estimated project completion date. Previous experience with similar projects is desired.

Attached you will find sets of images for training and then testing recognition matches against the trained sets.

Thank you,
Gabriel