Measuring In-Building Spatial-Temporal Human Distribution through Monocular Image Data Considering Deep Learning–Based Image Depth Estimation

Abstract

This research estimated the spatial-temporal distribution of humans in buildings through image sensing. Inputs were the in-building network, image sequences recording the movement of human, and camera parameters. Object detection and tracking models were utilized to discover humans in the images. Image depth estimation, clustering, and the camera model were integrated for the association of human and the in-building space in the image coordinates with the real world coordinates. The temporal human count for each in-building space was acquired. To validate the approach, two real cases in a school building, at a corridor and a hallway, were tested, and a synthesized case was carried out to exclude error from the detection and tracking steps. The proposed approach achieved results comparable to those of manual counting.