Object Recognition with OpenCV

OData support
Dr. Szirmay-Kalos László
Department of Control Engineering and Information Technology

Identifying objects in digital images is not always easy, even for human beings.

Developing a computer algorithm for it is a great challenge. Hovewer, in the past

decades many work has been done in this field, which could help object recognition

happen in certain situations.

Objects can be identified if we have a reference picture of them, and we use the

features in this reference to find similarities. In other cases we may only have statistical

data about the objects' features, although this does not mean that the their recognition in

other pictures is not possible.

Robustness is a key to every artificial intelligence algorithm, and it has a huge

significance in the field of image processing, because two pictures showing almost the

same scene can differ in many ways.

One feature based recognition technique which is robust enough is Scale

Invariant Feature Transform (SIFT). In this work I will give details about how it works.

The system I have developed is also based on this method.

I will also provide details about my work in the field of statistical image

processing. I have developed a method for localizing price tags in pictures taken in

arbitrary shops.

I implemented a software which demonstrates the use of the aformentioned

methods, with the help of the OpenCV library. Comparing this program to other similar

software have shown that my results are similar, and in some fields they are even better.


Please sign in to download the files of this thesis.