Host: The Japan Society of Mechanical Engineers
Name : Dynamics and Design Conference 2018
Date : August 28, 2018 - August 31, 2018
Recently, traffic congestions caused by increase of vehicle possession and complicatedness of traffic system have induced several serious social problems. In order to solve these problems, a lot of attempts have been carried out in many areas including new type of traffic signal system employing fuzzy control or neural network system. In addition to that system, a visualized miniature traffic simulation system based on real road system has been developed to examine the performance of the new traffic signal system and its effectiveness has been proved in several problems.
However, the simulation system is insufficient with the aspect of the model's validity because behavior characteristics of real traffic systems are not sufficiently reflected in the system. In this study, a traffic flow measurement system has been developed to extract traffic flow data by analyzing images from the fixed point cameras set up over two intersections. The measurement system has been developed by optical flow and R-CNN, and its performance was evaluated based on the recognition rate of the number of cars passing the intersection and the recognition rate of matching for same vehicle and the accuracy of the means speed are also estimated by the difference of passage time at two intersections. The examination result showed that the new system has been improved in matching for same vehicles compared to the previous study.
