Automated Traffic Light Detection for Day and Night Condition Using Geometric Moment Features

Abstract

This paper proposes a novel technique  to detect traffic lights, based on Geometric moment features (GMF). First, a colour image is captured from the camera mounted on the vehicle. From the captured colour image candidate region of traffic lights are extracted by using morphological operations. The input image is converted to YcbCr colour. After candidate region selection, the required shape of candidate is selected using parameters such as length, area, area of boundary box, etc. Geometric moments based values between candidate region and adjacent region are used to improve the detection performance.  Through this test, we show that the proposed method using GMF reaches up to 95 % of detection rate with computation time of average 82 s. The proposed method using GMF has got High detection rate and less computational time in comparison with other existing algorithms.