Smart Lift Passenger Flow Analysis (REF:C-0004)
E&M InnoPortal Trial Project Ref. No.:

Overview

To optimise lift usage, a system for monitoring lift passenger traffic is developed, which can suggest passengers to proceed to the lift lobby with less traffic by collecting and analysing traffic data such as the location, running direction, load status and lift alarm of each lift. Passenger numbers are obtained with video analytics on edge computers optimised for artificial intelligence which enables real-time human recognition. The data are shown on a web-based dashboard, which displays not only data from the density estimation system but also other information retrieved from external systems.

Problem Addressed

In the past, the passenger flow of lifts can only be counted manually and the statistics are only one-off. Also, lift operation status is only accessible by supervisory panel but not building management staff or lift passengers. To address these, a people counting system has been developed to provide real-time passenger flow data of each lift and lift lobby. Lift users now can access the web-based dashboard on their mobile phones or computers to check the occupancy levels of lifts and lift lobbies. They can locate the lifts on each floor and decide which lift to take. Meanwhile, management staff can retrieve historical passenger flow data from different dates, during various time periods and on different floors for further analysis.
Innovation

Each people counting device consists of a mini-camera, an edge AI device and a 4G/5G modem. The edge AI device runs an AI model to detect real-time passenger numbers. The counts are submitted to the EMSD cloud for display and statistical analysis. A web-based dashboard shows the density of the running status of lifts and waiting people in lift lobbies. A centralised system is also designed to collect, store, display and manage data collected from the density estimation equipment to show a consolidated view of information across all lifts and all lift lobbies. The cloud server also exports the data from the database in text files on an hourly basis for further processing to identify the lifts with most usage.
Key Benefits
  • Efficiently monitors real-time lift lobby and lift usages, with data refreshed around every 10 seconds. The information is available on a cloud platform with graphical interface for easy access. Such data are essential for optimising people flow management inside buildings
  • The computer will optimise its own AI model based on the videos collected to enhance human recognition accuracy. The accuracy reaches around 90% in lift lobbies and 80% in lifts
  • The system requires no video storage or transfer and no cabling works is involved
Patent and Award

International Exhibition of Inventions of Geneva 2022 (Silver Medal)
Project Reference

Electrical and Mechanical Services Department headquarters