Open AccessAdaptive Traffic Light Controller Based on Congestion Detection Using Computer Vision
Author(s) -
Rakhmad Gusta Putra,
Wahyu Pribadi,
Ivan Yuwono,
Dirvi Eko Juliando Sudirman,
Bambang Winarno
Publication year - 2021
Publication title -
journal of physics. conference series
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.21
H-Index - 85
eISSN - 1742-6596
pISSN - 1742-6588
DOI - 10.1088/1742-6596/1845/1/012047
Subject(s) - truck , traffic congestion , computer science , traffic flow (computer networking) , real time computing , floating car data , traffic congestion reconstruction with kerner's three phase theory , traffic optimization , automotive engineering , traffic bottleneck , controller (irrigation) , simulation , intelligent transportation system , process (computing) , transport engineering , engineering , computer network , agronomy , biology , operating system
The transportation sector plays an important role in realizing a smart city. The increase in the number of vehicles is currently not supported by an increase in road capacity. Traffic jams or congestion will occur in many places. Congestion will increase the accident rate, bad effect on economic growth, and increase gas emissions. Effective traffic management is necessary to reduce congestion levels and its side effects. A traffic light is one of traffic management methods. Traffic lights control the flow of traffic at road intersections, zebra crossings, and other traffic flow points. Conventional traffic lights work on a pre-programmed time sequence. This system is effective if the vehicle density is relatively constant. The density of vehicles from various directions fluctuates with time. To increase the effectiveness of using traffic light, an adaptive system is needed. In this study, a simple adaptive traffic light mechanism was developed based on congestion on the road using computer vision. Vehicle congestion is detected using the YOLOv3 object detection which detects the type of vehicle. The detection system used by YOLOv3 with pretrained weight COCO has a true positive value for motorbikes of 60%, cars (light vehicles) 93%, and trucks/buses (heavy vehicles) 100%. The processing speed of the Jetson Nano mini-PC with the OpenCV library on the GPU is 2 times faster than the process with the CPU.