Robust real-time traffic light detection and distance estimation using a single camera. Issue 8 (15th May 2015)

• Record Type: Journal Article
• Authors: Diaz-Cabrera, Moises; Cerri, Pietro; Medici, Paolo
• Is Part Of: Expert systems with applications. Volume 42:Issue 8(2015)
• Page Start: 3911
• Page End: 3923
• Publication Date: 2015-05-15
• Subjects: Traffic light detection -- Advanced driver assistance systems -- Intelligent transportation systems -- Image recognition -- Intelligent vehicles; Expert systems (Computer science) -- Periodicals; Systèmes experts (Informatique) -- Périodiques; Electronic journals; 006.33
• DOI: 10.1016/j.eswa.2014.12.037
• Languages: English
• ISSNs: 0957-4174

Abstract

Highlights: A method to detect traffic lights both during day and night is designed. A mixed method based on fuzzy logic and sequential rules is developed. The distance between vehicle and traffic light is calculated using Bayesian filters. Different results and tests are presented to validate the method. Abstract: This paper presents a robust technique to detect traffic lights during both day and night conditions and estimate their distance. The traffic light detection is based initially on color properties. To enhance the color on the video sequences, the acquisition is adapted according to the luminosity of the pixels on the top of the image. A fuzzy clustering provides a better division of the traffic light colors. The traffic light color properties have been estimated from registered sequences including both colors from LED spot lights and from traditional light bulbs. The filters rules based on the traffic light aspect ratios as well as the tracking stage are used to decide whether the spots on the frames are likely to be traffic lights. Then, the distance between traffic lights and the autonomous vehicle is estimated by applying Bayesian filters to the traffic lights represented on the frames. The tests are validated with more than an hour in real urban scenarios during day and night. The paper shows that the developed advanced driver assistance system is able to detect the traffic lights with 99.4% of accuracy in the range of 10–115 m. The utility of this system has been demonstrated during the Public ROad Urban Driverless car test in Italy in 2013.