The videos taken from a dashboard camera mounted in a car could be used to determine if the corresponding driver is law-abiding or not, as a result, an insurance company can determine to increase the insurance rate if the driver is found to be aggressive or lower the rate if otherwise. In this paper we perform vehicle speed estimation and traffic light classification as the foundation for judging a driver’s behavior. Optical flow is used, together with histogram equalization to enhance image quality, and the frame’s GPS information is also incorporated to estimate the speed of a vehicle. Furthermore, traffic light classification, i.e., distinguishing whether the light is red, yellow or green, is achieved using YOLOv5 and YOLOv6.