Authors:
V. D. Shepelev
1
;
A. I. Vorobyev
2
;
E. V. Shepeleva
1
;
I. D. Alferova
1
;
N. Golenyaev
1
;
G. Yakupova
3
and
V. G. Mavrin
3
Affiliations:
1
South Ural State University, 76 Lenin Prospekt, Chelyabinsk, Russia
;
2
Moscow Automobile and Road Construction State Technical University, 64, Leningradsky Prospect, Moscow, Russia
;
3
Kazan Federal University, 18 Kremlyovskaya str, Kazan, Russia
Keyword(s):
Monitoring, Neural Networks, Statistical Analysis, Traffic Capacity, Vehicle Speed, YOLOv3.
Abstract:
Most of the previous works dealing with road traffic organization have been focused on optimizing the setup of traffic signals, assuming that the traffic flow speed is fixed or adheres to a given distribution. In our study, we focused on real-time determining the vehicle speed and assessing its influence on the vehicle delay time. Vehicle detection and speed determination are based on real-time processing of video streams by a convolutional neural network (YOLOv3). The developed system can identify and classify traffic flows into eleven types, as well as track the motion path and speed of vehicles throughout the entire functional area of a signal-controlled intersection. While analysing the data, we identified two important factors corresponding to the presence of a queue of vehicles waiting for the green traffic light: 1. We identified the nature and statistically significant measure of reducing the free vehicle movement speed, depending on the size of the queue; 2. We determined th
e acceptable queue size, which does not affect the dynamics of crossing the intersection by group vehicles moving from the previous intersection. The obtained data allows us to optimize the operation of the adaptive traffic light control of intersections and to optimize the synchronization of road network signals based on speed indications.
(More)