A Three-Stage Framework to Estimate Pedestrian Path by Using Signaling Data and Surveillance Video

Abstract

The estimation of pedestrian path is of great value in the study of crowd dynamics, and various data with location information can be used as the basis for path estimation. In this study, we jointly use the signal data obtained from the pedestrian interaction with the base station and the surveillance video in the study area, and estimate the pedestrian path in the network finally. This paper proposes a three-stage framework for pedestrian path estimation, in the first stage, the video is used to extract the pedestrian's trajectory in the monitoring field of view, and the road segment that the pedestrian certain to pass is determined; in the second stage, the location information contained in signaling data is used to predict the road segments that pedestrian may pass through. In the final stage, the road segments determined by surveillance videos and the road segments inferred from signaling data are integrated, then we use HMM model to determine the combination of road segments with the highest probability, so as to obtain the complete travel path of the pedestrian. To evaluate the framework proposed in this paper, we conducted simulation experiments based on CARLA. The experimental results show that the path of pedestrians in the road network can be estimated effectively through the cooperative application of signaling data and surveillance video. Compared with other methods relying on only one data source, the three-stage framework proposed has higher accuracy in path estimation.