Identification of Traffic Flow Using Multi-Convolutional Neural Networks

Ching-Ta Lu ； Yu Huang ； Jia-An Lin ； Ling-Ling Wang

《International Journal of Network Security》 24卷4期 (2022/07) Pp. 747-754

https://doi.org/10.6633/IJNS.202207_24(4).18

摘要

Many countries suffer from traffic jams on a freeway or expressways. Moreover, governments significantly consider treating traffic jam-related problems in most modern cities. Therefore, developing a valid and reliable automatic analysis method for detecting expressway traffic flow for the control of each interchange gateway is essential. This study uses two convolutional neural networks (CNNs) to recognize the expressway's traffic flow. Initially, a road-region CNN is employed to recognize the lane regions of a captured image. At the same time, a traffic-flow CNN is used to identify the traffic flow of the image processing by the road-region CNN. The identified results contain three categories: block, more cars, and smooth. The experimental results reveal that the recognized precision rate of the traffic flow can reach 92.5%. Accordingly, the recognition results can control the number of vehicles entering and expressway in the interchange gateways, preventing traffic jams.

關鍵字

Convolutional Neural Network ； Deep Learning ； Traffic Flow Detection ； Traffic Jams