AI Video Analytic System for Escalator Monitoring in Market (REF:C-0005)
E&M InnoPortal Trial Project Ref. No.:


Overview

The AI Video Analytic System for Escalator Monitoring is a comprehensive solution that utilises cameras, an edge device, an AI computer and associated remote monitoring and network equipment. Its purpose is to detect and monitor unwanted events and behaviors on escalators. These include passenger falls, blockages at entrances/exits and reversed movement, misuse of escalators with bulky objects and maintenance misconduct, such as over-wetting during cleaning. The system also logs maintenance activities and monitors escalator status including On/Off condition and speed to optimise operations and maintenance processes.

Problem Addressed

A number of wet markets in Hong Kong have multiple storeys with packed food stalls. Escalators serve as an essential transportation for buyers traveling between levels. To enhance escalator safety, it is necessary to create an innovative system to monitor escalator status to optimise traveler experience and prevent accidents. The AI Video Analytic System for Escalator Monitoring is tailor-made for escalator safety monitoring. Notifications are sent to the management to report issues and snapshots / videos of each camera are displayed nearly real time, so that immediate actions can be taken to prevent accidents.

Innovation

The system mainly comprises network cameras, an AI computer, an email bridge, network storage and uninterruptible power supply. Multiple cameras installed at the escalators are supported by an artificial intelligence video analytic computing system. If unwanted events are detected, SMS/email alerts will be triggered and videos of the event will be uploaded to EMSD's AWS Cloud. For example, an alarm will be triggered when bulky objects approach the escalator entrance. A monthly report will also be generated to summarise the event occurrences.

Key Benefits
  • Utilises a combination of more than one AI models for high accuracy, achieving a false positive rate of less than 1% and a true positive rate of 85%
  • Highly customisable AI models can potentially detect other safety events
  • Run on cost-effective hardware with high processing throughput and horizontal scalability
  • Readily integrates with existing CCTV system and alarm systems
  • Provide face redaction functionality to protect individual privacy by masking human faces in videos
  • Adjustable threshold for alarms and real-time display of detection results
  • Conducts quantitative performance assessment to verify system performance
Patent and Award
  • Intellectual Property Department, Short-term Patent: HK30048625, HK30043745 and HK30045770
  • The 48th International Exhibition of Inventions of Geneva (Bronze Medal)
Project Reference

Tai Shing Street Market