• Citer

Crowd forecasting based on WiFi sensors and LSTM neural networks

par Singh, Utkarsh ;Determe, Jean-François ;Horlin, François ;De Doncker, Philippe
Référence IEEE transactions on instrumentation and measurement, 69, 9, page (6121-6131)
Publication Publié, 2020-09-01

Article révisé par les pairs

• Accès en ligne
• Détails
• Contenu
• Statistiques

Résumé :

To ensure effective management and security in largescale public events, it is imperative for the event organizers to beaware of potentially critical crowd densities. This paper, therefore,presents a solution to the above problem in terms of WiFi basedcrowd counting and LSTM neural network based forecasting.Monitoring of an actual event organized in Brussels has beendescribed, wherein crowd counts are obtained using WiFi sensorsin a privacy-preserved manner. The time-stamped crowd countsare used to develop univariate time-series, which are in-turnutilized for forecasting. Five different LSTM models are utilizedfor crowd time-series forecasting and analyzed for theirsuitability. A random walk model is used as reference forperformance assessment. Among different LSTM models,Convolutional LSTM delivered the best performance. Overallresults and analysis show that the developed system is suitable forcrowd monitoring.