Density Based Traffic Management Using Deep Learning and Computer Vision

https://doi.org/10.46610/JOSP.2023.v09i01.002

Abstract

Our project is focused on developing a traffic control system that will dynamically adjust signal timing in response to traffic congestion. Traffic congestion is an acute problem in numerous world cities, and it's time to replace our current time-dependent traffic management system with one that has decision-making capabilities. Current traffic management systems are fixed, which can be hugely inefficient when one lane is occupied more than the others. To optimize this problem, we've created a framework for controlling traffic efficiently that takes into account the density of traffic at any given moment. Sometimes higher traffic density at one side of the junction demands longer green light times – rather than the allotted time. So that all lanes can move along smoothly. We propose here a mechanism in which the green light of a stop sign or traffic signal doubles as the glowing indicator for when it's safe for drivers to cross based on the number of vehicles captured by cameras stationed around the area. Once the number of vehicles is calculated, this information is sent to a microprocessor (Raspberry Pi), and the red-light timer for that particular lane is dynamically adjusted accordingly.