Open AccessVehicle Classification and Counting System Using YOLO Object Detection Technology
Author(s) -
Jian-Da Wu,
BoYuan Chen,
WenJye Shyr,
Fan-Yu Shih
Publication year - 2021
Publication title -
traitement du signal/ts. traitement du signal
Language(s) - English
Resource type - Journals
eISSN - 1958-5608
pISSN - 0765-0019
DOI - 10.18280/ts.380419
Subject(s) - intelligent transportation system , computer science , convolutional neural network , artificial intelligence , computer vision , object detection , image processing , real time computing , object (grammar) , line (geometry) , engineering , pattern recognition (psychology) , image (mathematics) , mathematics , transport engineering , geometry
The intelligent transportation system is one of the most important constructions of urban modernization. Traffic flow monitoring technology is the most essential information in the intelligent transportation system. With the advancements in instrumentation, computer image processing and communication technology, computerized traffic monitoring technologies have become feasible. This study captures traffic information using surveillance cameras installed at higher locations. The YOLO object detection technology is used to identify vehicle types. The system principle uses image processing and deep convolutional neural networks for object detection training. Vehicle type identification and counting are carried out in this study for straight-line bidirectional roads, and T-shaped and cross-type intersections. A counting line is defined in the vehicle path direction using the object tracking method. The center coordinate of the object moves through the counting line. The number of motorcycles, small vehicles, and large vehicles were counted in different road sections. The actual number of vehicles on the road was compared with the number of vehicles measured by the system. Three separate counting periods were used to define the results using the confusion matrix.