Design Team Members: Sri Artham, Tim Lee
Supervisor: Dr. M. Kamel
Background
With the exponential growth of computing capabilities computers are becoming more and more pervasive in solving everyday problems. These applications are especially beneficial when they improve safety and reduce the incidence of hazardous situations.
One such hazardous situation occurs when a driver attempts to find a street in an unfamiliar area. Other drivers first notice this problem in the form of erratic driving behaviour, such as sudden acceleration and braking, incorrect use of turn signals, deviation from the proper lane, and non-compliance with traffic controls. These driving behaviours yield potential harm to both pedestrians and other drivers sharing the roadway.
Project description
The objective of this design project is to develop a vision-based system that will recognise and read street signs as a first step towards solving the distracted driver problem. Due to the computation effort involved and technology’s current inability to produce a product which would respond in a timely manner, the main application considered will be that of a vision-impaired person navigating an unfamiliar area. Although the technological hurdles are nearly identical in both cases, a mobile system carried by a vision-impaired individual would have less stringent response-time requirements as pedestrians are travelling at a much slower pace than automobiles.
The proposed system, upon identifying a street sign, will communicate the street name to the user, giving them an increased degree of freedom and independence.
It is hoped that with the developments realised in this project a functional prototype that recognises street signs will be created. With some further work and more advanced technology, an automated street-sign recognition device for operation in a moving vehicle should be readily feasible.
Design methodology
A proof-of-concept system is to be developed with a digital camera used for image acquisition. Upon acquiring the image, the algorithms developed will operate on the image to identify and isolate any street signs in the scene. The street name will then be verbalised via a speech synthesis program.
Processing of the acquired image will occur through a multi-stage process involving segmentation (dividing the image into objects), feature extraction (measuring various aspects of each object) and classification (determining whether or not each object is a street sign based on the obtained measurements). The results of this process will then be processed by an OCR (Optical Character Recognition) routine to produce a readable string for the speech synthesis program.
A Windows-based user interface for the overall system is to be designed for ease of use, and the entire system is to be implemented on a laptop computer for portability. The Windows-based interface is to be implemented in C or C++ in order to improve performance and cohesiveness of the entire system.