Real-time 4-way Intersection Smart Traffic Control System

Since traffic congestion is becoming a regular part of commuters' life, there is a pressing need for better traffic management. Most current traffic control systems are not sensitive to the current state of the roads being controlled, instead they are fixed, timed traffic signals that do not respond to unpredicted congestion. Solutions have been proposed to solve this problem including creating a large database for each traffic stop and determining the optimal traffic signals for the best vehicle flow based on the statistics collected, which does not react to data outliers. Other solutions suggest installing weight sensors under roads to detect the number of vehicles waiting then setting the duration of the next green light accordingly. This paper proposes an image analysis work flow to analyze the number of waiting vehicles as well as moving vehicles in each arm of a 4-way intersection. Then the collected data is utilized to control the state of the entire intersection to ensure the best traffic flow for all waiting and moving vehicles. Results from this approach yielded an absolute mean error of 0.559 detected representative vehicles with standard deviation of 0.93 on the first dataset and mean absolute error of 0.554 with 1.20 standard deviation for the second dataset. This level of accuracy conformed with the finite state machine control logic of the intersection, moving from one state to the other according to the analyzed images in real-time without causing starvation to any of the intersection arms.