Development of Smart Traffic Light Controller System with Deep Learning Capability in Image Processing

Abstract

The traffic congestion at the junction is becoming one of the major issues for many cities all around the world. One of the reasons causing this issue is due to the inefficient of the existing traffic light system at the traffic junction. This paper proposes a Smart Traffic Light Controller System (STLCS) with deep learning capability in image processing. The developed STLCS is comprised of Altera DE2 board, personal computer and Intel Neural Compute Stick 2 (NCS2). The personal computer is used as the vehicle detection system of the STLCS by performing various computer vision tasks and inference. The tasks include image acquisition, processing, and vehicle detection and counting. The smart feature of the system can detect the vehicles by using deep learning model and compute a flexible green time for each lane according to the density of traffic in each lane. The vehicle detection emphasizes the image processing by using the deep learning algorithm from the pre-trained model to increase the efficiency and computing time of the system. The efficiency of the vehicle detection system is about 94.73%.

Keywords

• deep learning
• image processing
• inference engine
• traffic congestion
• vehicle detection