A computer vision approach for the load time history estimation of lively individuals and crowds. (15th April 2018)

• Record Type: Journal Article
• Authors: Celik, Ozan; Dong, Chuan-Zhi; Catbas, F. Necati
• Is Part Of: Computers & structures. Volume 200(2018)
• Page Start: 32
• Page End: 52
• Publication Date: 2018-04-15
• Subjects: Human structure interaction -- Monitoring -- Vibration serviceability -- Jumping -- Crowd loading -- Load modeling -- Stadium -- Grandstand -- Computer vision; Structural engineering -- Data processing -- Periodicals; Electronic data processing -- Structures, Theory of -- Periodicals; 624.171
• DOI: 10.1016/j.compstruc.2018.02.001
• Languages: English
• ISSNs: 0045-7949

Abstract

Highlights: Vision based algorithms are introduced to estimate the forces due to lively crowds. A unique instrumented grandstand simulator is utilized for experiments. Computer vision based methods are validated to be an alternative with high accuracy. The methods are verified via a real-life field measurement. Achievements, limitations and future directions are discussed. Abstract: A computer vision approach for measuring the load time history due to individuals and crowds jumping and bobbing is investigated. The method comprises of tracking the displacement trajectories of individuals and crowds using optical flow based algorithms followed by generating force time histories. Laboratory experiments, in which individuals and groups perform jumping at regular beats and songs on a force platform and on a grandstand simulator, are conducted. The estimated trajectories are compared directly with conventional sensors as well as indirectly with responses acquired from finite element models. The method is further validated via a field demonstration. Limitations of the method and future work for improvement are discussed. The proposed methods along with their applications on a real structure, and findings from a laboratory grandstand simulator that can accommodate experiments for groups of different sizes and structural configurations show great promise for computer vision based load modeling. In this sense, the study is taking an important step in support of creating a database for crowd loading that is needed as it is pointed out in the literature.