抄録
However, the conventional pedestrian models using these systems are intended to predict the behaviour of the pedestrian as a single mass. In addition, the pedestrian speeds in the system have been assumed almost constant during the crossing. Therefore, we have analysed the pedestrian image of the live camera that is installed in Shibuya scramble intersection in this study, especially to analyse walking speed change. Furthermore, in order to examine the change of leg joints on the condition, we have conducted experiments in the 10-meter-long road, which is divided into four conditions. We have constructed the prediction model by the Kalman filter using the participant behaviour. As a result, the prediction model is substantially equal to the observed value; the error was 0.74% and 0.62%, 0.66%, 0.75%. It seems to be possible to estimate the change of state of the pedestrian from this approach.
