In recent years, the number of traffic fatalities caused by illegal right turn vehicles in Korea has increased by approximately 2 times, from 0.34 to 0.64. However, there is currently no dedicated enforcement system in place for detecting illegal right turn vehicles, unlike the speed violation enforcement systems. To address this issue, this study aims to develop an algorithm for the detection of illegal right turn vehicles. You Only Look Once version 5 (YOLOv5) is used for the recognition of vehicles, pedestrian crossings, and pedestrians. AI Hub dataset and Common Objects in Context(COCO) dataset are utilized as the training dataset. Specifically, 36,808 samples from the AI Hub pedestrian crossing video dataset are used for training to detect the pedestrian crossing, while the COCO dataset is divided into a training set of 118,287 samples, a validation set of 5,000 samples and it is used for detecting pedestrians and vehicles. For testing the algorithm, news footage related to illegal right turn vehicles provided by the Korean news channel JTBC is used as the test dataset.
Detection for Illegal Right Turn Vehicles in the Korea Traffic System
09.08.2023
2191699 byte
Aufsatz (Konferenz)
Elektronische Ressource
Englisch
Behavior discrimination method for illegal driving into downstream through right turn
Europäisches Patentamt | 2024
|Traffic illegal behavior detection method for non-motor vehicles and drivers and system
Europäisches Patentamt | 2021
|