This thesis aims to investigate how AI and computer vision can be used for long-term traffic monitoring and detect abnormal behaviour on the road. The setup used a Raspberry Pi and a stillcamera to gather images of traffic at a roundabout nearby Jönköping University. The images were run through the YOLO26 model to detect and count vehicles over time. As a result, different traffic patterns were identified by splitting the data into hourly, daily, and weekly counts. The collected data illustrates clear differences between weekdays and weekends, along-side peak traffic hours during the afternoon. The findings from this thesis proves how AI-based trafficmonitoring help low-cost solutions to analyse traffic flow and identify unusual traffic events. These outcomes can be useful for future studies on smart traffic systems, abnormal detection behaviour and better infrastructure. To sum up, combining computer vision, deep learning, and long-term traffic patterns leads to more effective and automated traffic monitoring systems.