A Real Time Dynamic Approach for Management of Vehicle Generated Traffic
Hemant Kumar Chaudhary, Kartikeya Saraswat, Harshita Yadav, Hrithik Puri, Asha Rani Mishra, Sansar Singh Chauhan
Abstract
India, with a population of 1.3 billion and almost 300 million vehicles, is one of the biggest contributors to traffic jams, vehicle- specific pollution, and chronic lung diseases. To manage the footfall of these gigantic vehicles, continuous effort and research is taking place in the direction of both active and passive traffic management. This research paper aims towards showcasing a dynamic, fully autonomous deep learning model that uses real-time feeds from existing traffic junctions/ intersection cameras, process them and provide an intensity score based on the density of traffic in each adjoining lane.The proposed CNN model which is based on YOLO framework uses 10 seconds wait analysis time. The proposed system manages traffic, based on the intensity scores which assign traffic a Go time to each lane using an optimal traffic time in the range of 10-50 seconds.The model also scans for emergency vehicles in each lane, to provide a priority pass to such vehicles. Evaluation of model performance Mean Average Precision (mAP) is used.