Road extraction method of vehicle trajectory data based on geo-referenced videos

With the increasing use of driving recorder equipment, people need new methods for the analysis of vehicle trajectories. The extraction of road information from vehicle trajectory data is one of the focuses in the field of geographic information. In this paper, we proposed two vehicle trajectory extraction methods, a fast vehicle trajectory extraction method based on GPS points and a vehicle trajectory extraction method based on the field of view. First, we gave a problem definition for the video trajectory display method. Then we expounded on the field of view of the Geo-referenced video [1] and its related information. The first method connects the location points, and each segment of the trajectory line indicates the current driving direction of the vehicle. The other method introduces the concept of perspective on this basis. It not only shows the direction of the trajectory line but also extracts the perspective of keyframes to accurately describe the trajectory of the vehicle. Next, we used a time and distance-based spatiotemporal clustering algorithm to extract points and demonstrate them through experimental results. We visualized the extracted vehicle trajectories and displayed them on a map. Finally, we compared the efficiency and accuracy of the traditional vehicle trajectory extraction method and the two methods proposed in this paper. The results showed that the vehicle trajectory extraction methods proposed in this paper are superior to the traditional vehicle trajectory display method in accuracy.