Evaluation of traffic light detection algorithms for automated video analysis

Abstract

Vehicle and Traffic Safety is a growingly important research topic among the automotive industry and academia. Being able to analyse driving behaviour and collecting data is key for gaining understanding about potential risks affecting traffic safety. Traffic lights are important in terms of traffic safety, therefore it is of importance to have a solution to detect them without having to spend time to find their occurrence manually in a video analysis. The goal of this this paper was to evaluate a traffic light detection algorithm for automated video analysis. The study was conducted as a case study with a quantitative research method, and present an evaluation of the implemented algorithm. The implemented algorithm is benchmarked and evaluated on a dataset exceeding one million frames coming from videos of naturalistic driving in different conditions. The result of this study covers an evaluation of the algorithm based on the benchmark. This study concluded that using Haar feature-based cascade classifiers for traffic light detection is a suitable method if some trade-offs can be made. This paper also presents recommendations for developers facing similar problems in terms of automated detection of objects connected to the real world. The process of designing and creating a solution for an automated video analysis is emphasized in a top-down approach, giving an insight for developers facing similar challenges.