Development of The Method for Estimating Traffic Volume of Pedestrians in An Underground Mall by Use of Watch Cameras
Naoya Takagi, Atsushi Takizawa · eCAADe proceedings · 2016
This paper describes a method for estimating pedestrian traffic volume by using video cameras. In the Umeda underground mall in Osaka City, we estimated the traffic volume without tracking technology and while protecting pedestrian's privacy. We developed an original algorithm that roughly estimates the traffic volume of pedestrians from sequential images of video cameras. We focused on a line on each image cut out from video and made a new image which shows the spatiotemporal distribution of pedestrians. We defined this image as 'time historical image of pedestrian spots (THIPS)'. In a THIPS, a pedestrian is regarded as a cluster of connected pixels with the same label. We captured the spatiotemporal distribution of pedestrians by using these images. We found that this algorithm requires a THIPS to estimate the number of pedestrians who passed the spot for a few minutes and plural THIPSs to estimate their traveling directions. Finally, we concluded that this algorithm is an efficient means of estimating pedestrian traffic volume.