Detecting Motorcycle Crime Gangs in CCTV Video Footage Using YOLOv9 and CNN

Street crime, such as criminal motorcycle gangs, has become a severe problem that concerns Indonesian society, especially for those who live in big cities. Criminal Motorcycle gang members often disturb the surrounding community. One of the activities of motorcycle gangs that often cause disturbance is the activity of street convoys using motorbikes. Various solutions to reduce motorcycle gang crime have been conducted by the authorities. One of them is patrolling. However, this effort is less efficient due to limited time, workforce, and coverage of the area that can be monitored. Another preventive solution is to install CCTV. This solution also requires human resources to monitor CCTV footage. This certainly increases the possibility of human error. Some studies have been conducted to automate CCTV surveillance by detecting anomalies or crimes. In this research, motorcycle gang detection consists of three stages. The process begins with detecting and tracking motorcycles in the video using YOLOv9, which achieves an AP50 of 93.2%, alongside ByteTrack. The second step involves mapping each motorcycle's center coordinates to represent each motorcyclist's motion patterns. Finally, these patterns are examined and classified using CNN to detect motorcycle gangs. This method achieves 93.4 % accuracy in detecting motorcycle gangs' presence in a video.