Adaptive Traffic Light Signal Control Using Fuzzy Logic Based on Real-Time Vehicle Detection from Video Surveillance

Zulfa Fahrunnisa, Rahmadwati Rahmadwati, Raden Arief Setyawan

Abstract

Intersections often become the focal points of congestion due to poor traffic signal management, reduced productivity, increased travel duration, gas emissions, and fuel consumption. Existing traffic light systems maintained constant signal duration regardless of traffic situations, resulting in green signals for lanes with no vehicle queues that increased waiting times in other lanes. Therefore, a real-time traffic signal optimization system using Fuzzy Logic control, utilizing vehicle queue and flow rate real-time data from video surveillance, is needed. This research used recorded video from surveillance cameras in Banten Province, Indonesia, during daylight conditions. Vehicle queues and flow rate data were used as parameters to determine traffic light signals. The YOLO algorithm obtained these parameter values, then served them as inputs for the Fuzzy Logic system to determine signal duration. The accuracy of the traffic situation estimation system fluctuated within a range of 40% to 100%. Simulation results showed an improvement of approximately 18% by evaluating the total number of vehicles that exited the queue and reduced vehicle waiting time by about 21% compared to the existing system on intersection efficiency. Consequently, the proposed system can reduce pollution and fuel consumption, contributing to urban sustainability and public well-being enhancement. Despite the improvements over the previous systems, the accuracy of the vehicle detection system may vary with traffic density based on the extent of occlusions present, which is an area that needs further refinement. This research’s contributions include utilizing real-time video footage from surveillance cameras above traffic lights to obtain real traffic conditions and identify potential errors such as occlusion of overlapping vehicle due to very congested roads. Another contribution is the adjustment of the Fuzzy membership function based on the vehicle detection system’s ability to ensure precise determination of green signal duration, even when the input data contains errors.

Keywords

Fuzzy Logic; Image Processing; Intelligent Transport System; Traffic Management; Computer Vision; YOLO; Video Surveillance

Full Text:

PDF

DOI: http://dx.doi.org/10.26555/jiteki.v10i2.28712