Traffic Light Signal Detection: A Study

Abstract

This project involves comparative analysis of detection of traffic light signals with the use of image processing..
This paper focuses on a product of usage of monocular video camera for the detection of traffic lights under different
conditions i.e. adverse weather and low illumination, haze, fog etc. This paper provides a perception algorithm that works at
real time with high apprehension rate while using closeout camera for traffic light signal detection. By the use of this
algorithm credibility and immunity for UGV and drivers will step-up. The working of traffic light detection system is divided
into in three parts i.e. a monocular or color camera, then acquisition of image by that camera and a PC is required to
process the captured images. The RGB color space based algorithm helps to detect and extract the red, green and yellow
objects in the images and then illuminating the disturbing colors present in the locale. The final objective is to verify the
features and object identity and based on these the different types of traffic signals are examined. . The ideology is
qualitatively computed in various conditions, including driving in the heavy rain, at night and in city roads with dense traffic,
as well as their synergy. Quantitative computation on a publicly available manually commented database, scoring high
apprehension rate is also implemented.