About this blog

So, this is a blog which will cover my work of my final year project for the bachelor degree in Informatics. The title of my thesis is "Implementation of image analysis algorithms for augmented reality smartphone applications" and it will be written with the support of the Ionian University and Information Technologies Institute in Greece.

The subject of this thesis is related with the fields of computer vision, pattern recognition and machine learning. The main goal is to build an augmented reality android application, which with the usage of visual recognition can classify objects into categories. Many visual recognition applications have been presented in the literature and this work will investigate the possibility to use them in mobile applications. The recognition process has high computational cost, and applications until now mostly use client-server architectures to perform these tasks on fast desktop PC's. With the processing power of current smartphones constantly increasing, the recognition tasks can be performed solely by the phone and avoid data transfer latency and bandwidth.

The OpenCV library will be used for both Windows and Android platforms, which contains more than 500 computer vision related functions. For this project some of the most helpful functions would be the k-means clustering, feature detection and extraction, fast approximate nearest neighbor searching (flann), MATLAB-like matrix operations and some basic image processing functions. There are many alternatives to OpenCV, probably the most popular of them would be the BoofCV library, but OpenCV seems easier to be ported on Android.

OpenCV website

Other software that were installed were the Eclipse IDE for Android development, Visual Studio 2010 for use with the desktop version of OpenCV, MATLAB and of course the android SDK.