Innovations in crowd management: An integration of visual closure, anthropometry, and computer vision

Abstract

This paper reports the application of a bio-inspired computational artificial intelligent (A.I.) real-time crowd monitoring and management system that integrates ergonomics, anthropometric database, computer vision and decision analytics. The system matches and fits anthropometrically customized 3D human models into a 3D space that is dynamically constructed from videos captured by one or more surveillance cameras. This approach is consistent with the human visual closure effect when we estimate the number of people in moving crowds. Dynamic human movement data are optimally extracted from the video data and used to construct and train a crowd movement profile detector. Learning algorithms have been developed to detect deviations from the normal profile. Results of validations show that there remains a huge gap in the performance between a bio-inspired computational A.I. model and a normal human-being in the surveillance tasks in terms of reliability, but this is a notable first step of a reliable crowd management system not emphasizing on facial feature extraction.