The study on the estimation of vehicles speed using a dashboard camera

Mobile and handheld electronic products such as smartphones and dashboard cameras are frequently used to record occurrences of traffic accidents nowadays. Most existing studies require at least one entire lane marking as a reference to calculate a certain amount of driving distances and are not adapted to a targeted vehicle with a rapidly changing speed in a short time. Therefore, existing approaches cannot provide the desired accuracy of speed estimation in real-life scenarios. In this study, we obtain dynamic time and space data from recorded footage using dashboard cameras and then apply photogrammetry and cross-ratio methods to estimate the vehicle speed. In addition, the proposed method is applied to other cars' speed estimation and ego-speed estimation even though both cars are moving. The experimental results show that given the frame rate on the recorded footage, our proposed method only needs one object in each frame to estimate the vehicle speed even at a rapidly changing speed. Our proposed method shows that the difference between the estimated speed and the reference speed by the Global Positioning System (GPS) is smaller than 1 km/h when only one car is on the move, and smaller than 3 km/h when both cars are on the move. Nevertheless, the difference between the estimated speed and the reference speed is between 1.46 km/h and 5.28 km/h when one car moves by AUPD method. To our knowledge, there is no existing method that estimated the front vehicle speed when both cars are on the move using dashboard cameras.