Abstract:
In the modern planet, the number of vehicles is increasing worldwide continuously at a remarkable rate and in parallel, road accidents are also linearly increasing globally. Specifically, in Ethiopia, traffic accidents are becoming more destructive and horrifying day-to-day. To minimize the rate of accidents, different countries are exercising different policies and rules which could possibly decrease it and make the flow of traffic safer and smoother.
Commonly placing traffic signs across the road is one way of minimizing the accidents, because traffic signs provide an important information to the drivers about the road and other things in that specific place. But drivers are sometimes insensitive to those signs due to carelessness, lack of attention and other different reasons. As a consequence, the probability of having an accident goes high with signs left unnoticed and unapplied by the drivers.
Hence to resolve this threat of road accidents in this era of technological advancement, different countries are continuously working towards developing different Driver Assistance Systems like Traffic Sign Detection and Recognition system (TSDR) system. TSDR system is a system which is used to alert/warn and disburden the driver by detecting traffic signs on the road using the camera mounted on the top of the car. But since traffic signs are different in different countries developing TSDR system is a country-specific task. So in this study, VoiceBased TSDR system is developed for three warning and one priority Ethiopian traffic signs using image processing technique. After a specific traffic sign is recognized, Amharic voice notification is generated by the system as a warning. The experiment is done on 230 image datasets using two different color spaces namely RGB & HSI color spaces. Segmentation on HSI color space is done using adopted fixed thresholding values for detecting red color from the image and the RGB color space images are segmented with both Otsu and Global thresholding techniques. For all those three combinations, Histogram of Oriented Gradient (HOG) is used as feature extraction with Support Vector Machines (SVM) classification technique. And according to F measure of each tested experiment, the performance of TSDR system on HSI color space is shown better than the others with the F-measure value of 76.27%.