Video Prediction for Automated Control of Traffic Signals Through Predictive Neural Network

Abstract

In today’s world of automated things, safety plays a very important role especially when it is concerned with the lives of people. Deep-learning algorithms are being used to help solve various tasks. But, using deep learning for automating traffic signals by using predictive neural network is a new challenge. In this project, the proposed model uses two predictive neural network algorithms. It is trained for next-frame prediction which learns to predict the future frames in a video sequence. This prediction can be used for improvising driving through automated control of traffic signals. A traffic simulator is made using MATLAB software. A four-way junction exists where the number of days can be provided depending on the dataset needed. An instance image of each hour shows the number of cars which appear in each direction of the road every hour. The learning algorithm uses this dataset to predict the future number of vehicles and hence, control the traffic. Using this model, the aim is to solve the traffic controlling system and make it very efficient by properly controlling and saving time through automation.