PDF version of this entire document

Summary

I have presented an outline and developed a user-friendly application/framework for crudely identifying and tracking objects on the road. Considering time limitations, it is safe to say that the full potential of this project was not fulfilled. The code is made free/open source software, however, encouraging others to carry on from the point reached. Utilisation of the framework depends on one's ability to comprehend the underlying concepts, primarily those associated with pattern recognition and real-time tracking. Nothing particularly novel has yet been attempted.

This document detailed and showed compartments of the framework, then demonstrated how it adapts to various form factors, depending for the most part on screen resolution. This was tested in emulation mode for the most part. It then delved deeper into the methods explored and the resultant performance. Data for training classifiers was explained, then some tests of the classifiers were shown. Many demos were produced in video form and uploaded to the World Wide Web. In them, evidence of gradual improvement over time - depending on the size of the training set and the evolving implementation - exists. Further details about the projects, including videos, source code, binary executables, and further explanations, can be found online at http://schestowitz.com/Research/Cars/Car_Navigation/http://schestowitz.com/Research/Cars/Car_Navigation/.

Acknowledgements: the project was funded by the http://erc.europa.eu/European Research Council.