生物启发的运动人群瓶颈效应感知视觉神经网络

打开文本图片集

关键词：人群异常活动分析；人群瓶颈效应；LGMD；蝗虫视觉神经系统；神经尖峰响应；智能视频监控中图分类号：TP183 文献标志码：A 文章编号：1001-3695（2025）08-004-2274-09doi：10.19734/j.issn.1001-3695.2025.01.0015

Bio-inspired visual neural network for perceiving moving crowd bottleneck effect

Lu Xianpeia，b，Hu Bina，b.ct（.StateKeybooub，lofueSe&lellceehsUniversity，Guiyang ，China）

Abstract：Thecrowdbotteneck effct inlarge-scaleeventsserves asan important precursortodisasters suchascrowdingand stampedes.However，therearefewstudiesoncomputationalmodelsforcrowdbotleneckeffctdetection.Duetotherandomnessof this efectandthecomplexityof therowdstate，thetraditionalcomputationalmodelsarenotidealfordetection.Toaddressthis issue，this paper proposed a crowdbotteneck efect perceiving visual neuralnetwork（CBEPVNN），drawing onhe neural structure characteristicsofthe locust visual systemand therisk perceiving mechanismoflocustLGMD.The model simulatedthe visualinformationprocesingcharacteristicsoflocustsandmammals，integratedvisualmovementinformationfromumanactivities，andemployed the danger perception mechanism of LGMD neurons to establish a peak threshold mechanism. Thismechanism tuned the neural network output toperceive thebotleneck efect ofmoving crowd invisual scenes.Experimentalresultsdemonstrate that CBEPVNNeffctively perceives the movingcrowdbotteneck effct invideosequences andproduces a strong preference response.This work elucidates the mechanism of crowd dynamic visual information procesing inspiredbybiologicalopticneural mechanisms，providing newideasand methodsforabnormal crowdactivitydetectionandbehavior analysis in intelligent video surveillance.

Key words：analysis of abnormal crowdactivity；crowd botteneck effct;LGMD（lobula giant movement detector）；visual nervous system of locust；neural peak response；intelligent video surveillance

0 引言

Solmaz等人[1]将现实世界中常见的特定人群行为分为瓶颈、源头、车道、拱门和阻塞五类。（剩余23366字）