Traffic Management using Convolution Neural Network
Gaurav Dhingra, S Supreeth, Kohli Neha, R V Amruthashree, D Eshitha · Zenodo (CERN European Organization for Nuclear Research) · 2019
Traffic is one of the major problems in most of the metropolitan cities. Classifying the traffic conditions are important for determining traffic control strategies and management. Traffic congestions have negative impact on society, as a lot of time is wasted in it and controlling the congestions is necessary. By classification we can get to know which lane has traffic, from which we can further check the reasons for traffic and to take appropriate decisions to improve the performance. Video on traffic data is suitable source for traffic analysis. In this paper, video surveillance data is used for classification of road traffic using Convolution Neural Network. Convolution Neural Network requires minimal preprocessing when compared to other classification algorithms and is known for its accuracy. The video is classified based on rating of the traffic of its content. The Convolution Neural Network is first trained and then it is evaluated and updated using validation set. Once the model is completely trained it is tested with the testing set. This trained model is capable of processing the live streaming video and classifies each of the frames and gives the rating of the traffic for each lane, which can be helpful for traffic management.