Open Sesame: Unlocking Smart Building Management with AI
by Puneet Gupta (Senior Manager, System Solutions)
“Open Sesame!” The thieves of the ancient, Oriental story ‘Ali Baba and the Forty Thieves’ used this incantation to guard their treasure hidden in a mystical cave. If only we had a similar magical way to control and manage our modern premises ranging from shopping malls and concert halls to football stadia and office/apartment buildings! Well think again. If not magical, we do have something more powerful and hi-tech to enable smart building management.
We are talking about Machine Vision & Deep Learning that are rapidly changing the building management paradigm. By commissioning low cost smart cameras and AI devices, they provide analytics to automate building/facility access control. This evolution is what is driving the movement towards smart buildings and smart cities.
Security, crowd monitoring, queue management and accident prevention are important challenges facing today’s city planners and administrators. A primary focus in all these applications is managing the people who walk in through the gates/doors of a monitored premise.
Traditional access control techniques involved the use of trained staff, magnetic cards or punch-in passcodes. While these techniques are effective in certain ways, they heavily rely on human vigilance or privacy protection of access codes. In contrast, Machine vision and deep learning pave way for efficient Smart Building or Facility Management by automating the whole process and ensuring greater control.
Let us investigate this using three different access control scenarios:
A camera is typically mounted on top of the garage door to look for any vehicle driving towards the door – either facing the camera (front number plate visible) or reversing towards it (rear number plates visible). The camera system is trained to detect color, size, shape/contours of the vehicles. Besides, it can detect and read the license plates of the approaching vehicle (using ALPR techniques).
To allow access, the camera system then matches the license plates using machine learning inferences. Unless a match is found, the garage doors stay shut.
A facial recognition based override enables the home owner to open the garage doors for rental vehicles whose details are not entered in the authorized vehicle database of the camera system. Additional security features can be built in too. For instance, if an unauthorized vehicle continues to stay parked in front of the garage door, a loitering/intrusion alert can be triggered to the home owner’s smartphone.
In this case, the crowd management solution comprises multiple cameras mounted at strategic locations to allow unobstructed view of the viewers in the pavilions/stands as well as the entrances/exits to these spaces.
The entrances closer to the emptier pavilions are open and the ones near more densely populated pavilions can be closed, while people pour into the stadium for the game. During intervals, the entries and exits to snack counters can be regulated to evenly distribute the crowd.
Additional machine vision and analytics features can be built in for smooth crowd management and emergency response. This includes detection of unruly behavior, reporting acts of vandalism, and raising medical alerts when someone faints/collapses/falls.
There are several machine vision solutions deployed in a smart factory – driven by the Industry 4.0 revolution. For instance, the entrance to various locations is managed strictly based on permissions. To allow seamless and fraud-free roll call (attendance), many factories have moved to face recognition and/or biometrics based entry or exit to record the movements and payroll of personnel.Further, visitors are often assigned temporary privileges and access to select areas of the factory only when accompanied by an authorized personnel.
Video cameras installed in the alleyways, factory floor and production halls continuously examine visitors during their presence on the factory premises. Supplementary surveillance feeds are also used to promptly report and handle incidents of factory intrusion or perimeter violation.
The team at Ittiam is dedicated to finding innovative solutions for enabling these and more applications for smart cities and smart buildings. Check out adroitVista SDK, a suite of software solutions for machine vision, deep learning and visual analytics that leverages DSP/GPU acceleration to realize real-time visual analytics.