Dynamic time distribution system monitoring on traffic light using image processing and convolutional neural network method

Abstract

The number of vehicles that continues to increase every time can cause congestion and increase the potential for accidents. One type of violation that often results in accidents is motorists who break through traffic lights when the red light is on. One factor that encourage motorists to not obey traffic light signs is due to the distribution of green time which tends to be fast on relatively dense tracks. This final project aims to get dynamic green light on traffic light according to the volume of the vehicle for each lane using convolutional neural network method and can detect traffic light violations in the form of breaking the stop line markings on zebra cross. The sensor used is camera. There are three cameras used in each line. The number of cars detected will be sent to microcontroller via laptop. The data on the number of cars is used as input for the convolutional neural network to produce output data that serves as reference for obtaining the length of time for the green light. If the car stops above or exceeds the stop line on zebra cross when the red light is on, the car will be detected as violation.