2023 Volume 79 Issue 20 Article ID: 23-20034
In the future, a mixed pedestrian-vehicle traffic flow simulation environment in a shared space where pedestrians and vehicles are mixed is necessary, but it is difficult to observe the simulation input in a limited environment such as a roadside station. However, the detection accuracy of pedestrians and bicycles is low, and there are issues to be solved in order to perform constant observation. Among them, there is a particular problem in the detection accuracy of pedestrians and bicycles, so in this research, network cameras were installed at 10 locations in Nagoya University campus and pedestrians and bicycles were observed for one month. Then, we constructed a people flow counting system using YOLO and Strong Sort, which detect in real time from the observed images. In order to improve the accuracy of the counting results, we were able to greatly improve the accuracy of pedestrian measurement by constructing an algorithm that can be applied to the whole rather than taking measures for each measurement area. Based on the analysis and results of the above, we will verify the possibility of applying it to continuous observation in the city.
