A Real-Time Multiple Traffic Violation Detection System Using DeepSORT

As population growth accelerates, the number of vehicles on the roads increases, leading to rise in road accidents. A significant cause of these accidents is traffic rule violations, such as speeding and wrong-way driving. To address this issue, an advanced real-time system is needed to detect traffic rule violations using CCTV cameras. The proposed traffic signal violation detection system automates the monitoring process, performing tasks such as speed and wrong-way driving detection without human intervention. The system focuses on three primary tasks: vehicle detection, vehicle tracking, and violation detection. Vehicles are first located using the YOLOv3 deep learning method. The Kalman filter combined with DeepSORT is then used to track the detected vehicles, assigning each a unique ID. Algorithms for wrong-way driving and speed violations are subsequently applied to identify rule-breaking vehicles. This method ensures high efficiency and accuracy in detecting vehicles that violate traffic rules.