A Novel Approach for Detection of Traffic Light Timer with Voice based Assistance

An efficient method to detect traffic lights and recognize signal count-down timings is developed. Traffic scenes are often complex as it include lot of information, keeping constant attention on the traffic signs is not an easy task for drivers. Therefore some traffic data or signs can be missed for several causes such as the complexity of the road scene, the high number of visual information, or even the bad weather condition or visual fatigue. The countdown timings are segmented and extracted from the image based color segmentation. These signals also use specific shapes for each color, which aids visually impaired people in distinguishing signal aspects. Traffic light detection is done by color segmentation of the input image, based on the RGB color thresholding. The predominant glowing traffic light can be very well identified by this method regardless of the size and shape of the traffic lights. Although the shape of traffic lights is not considered, as a criterion for identification, only the round traffic lights are taken into account. Finally the detected digit in the timer is displayed on the screen and converted into voice.