Building a traffic monitoring system for intelligent transportation systems (ITS) in the developing smart cities has drawn in a mass of consideration in the latest past. Since the majority of cities in the world are observing the increasing number of vehicles on the road, they are tending to accept an intelligent transportation system for resolving tedious issues like traffic density, count of traffic lines and their length, the standard speed of the traffic, and increase in the number of vehicles during weekends or for a particular time span in a day. Smart transportation systems are provided with traffic pictures and recordings by installed cameras on roads or signals. Also, different types of sensor-actuator pairs help check and deal with traffic issues. This paper proposed a method that uses customized convolution neural network (CCNN) on traffic images to classify images as per the traffic density and thereby provide driving assistance. The proposed system in the paper can monitor traffic using videos captured by installed cameras and then classify the current traffic situation into categories of high, medium, and low categories. The aim is to use this as a model to provide traffic density information from various places to expert systems and take other important decisions regarding traffic control. NVIDIA graphics processing unit (GPU) is used to parallelize the training process and implement complex deep neural networks to obtain better accuracy. Performance evaluation of the proposed system is done on a real-time dataset containing recorded footage of traffic from Pune city (India) and recorded footage by highway CCTVs at Seattle, WA, obtained from the Washington State Department of Transport. Experiment results classify traffic density into high, medium, and low based on existing traffic and predict it correctly up to 99.6%. Obtained testing accuracy is much better than the results given by existing algorithms.
Traffic Density Classification for Multiclass Vehicles Using Customized Convolutional Neural Network for Smart City
Lect. Notes in Networks, Syst.
2022-08-19
16 pages
Article/Chapter (Book)
Electronic Resource
English
Smart vehicles "see" city traffic
British Library Online Contents | 2001
|Traffic Congestion Pattern Classification Using Multiclass Active Shape Models
Transportation Research Record | 2017
|