Research On Pedestrian Flow Reynolds Number In Crowd Motion Video Scenes

Research on crowd motion state plays an essential role for crowd motion analysis.The description of movement status is very important for analyzing crowd movement.Reynolds number is a feature in Hydrodynamics which used to judge whether it is laminar flow or turbulence flow.The paper aims to study the disordered motion state of pedestrian flows in crowd motion video scenes,thus the pedestrian flow is regarded as the flow of fluid particles in rectangular pipes.Then Reynolds number is introduced into crowd motion,thus Pedestrian Flow Reynolds Number(PFRN)is proposed,which is a novel feature descriptor derived from Hydrodynamics,to describe crowd motion state.The following will be studied from the global and local perspectives.From the global perspective,the pedestrian flow is regarded as a whole,this paper aims to investigate the chaotic characteristics of pedestrian flow as a dynamical system.First,the concept of Pedestrian Flow Reynolds Number is proposed.Later,the calculation method of Pedestrian Flow Reynolds Number in video is put forward to characterize the motion state of the pedestrian flow.Then experiments are carried out on two public crowd data sets to fully verify the effectiveness of the Pedestrian Flow Reynolds Number.Finally,nonlinear time series analysis tools,including time delay embedding and largest Lyapunov exponent are applied to verify the chaos of pedestrian flow's motion.Experimental result that all the largest Lyapunov exponent are positive could indeed demonstrate the complexity and chaos of crowd motion.Meanwhile,it turns out that Pedestrian Flow Reynolds Number put forward in the paper can effectively characterize the motion state of pedestrian flow.Pedestrian Flow Reynolds Number is applied to the open scene with simple motion and uniform crowd distribution,while the crowd distribution in the real scene is uneven.Local motion salient region crowd segmentation algorithm(LMSA)based on optical flow field is proposed to extract crowd concentrated areas in scenes with uneven crowd distribution,which are regarded as the research objects of pedestrian motion state analysis.Based on the Local Pedestrian Flow Reynolds Number,four crowd motion characteristics of Local Pedestrian Flow Reynolds Number,density,velocity and direction entropy are analyzed respectively to study the trend of the motion characteristics of local areas with time.Finally,the correlation analysis among the motion patterns of local areas is carried out.In summary,this paper analyzes the pedestrian flow motion state in video scenes from the global and local perspectives.Inspired by crowd turbulence,the feature descriptor of Pedestrian Flow Reynolds Number is proposed from the global perspective,which paves a new way for research on crowd turbulence.Further observation found that the crowd distribution is uneven and crowd gathering often occurs,so this paper extracts the crowd concentrated areas,studies the change trend of the Local Pedestrian Flow Reynolds Number with time,and intuitively shows the motion state and correlation among local regions.It could potentially be applied to crowd motion understanding.