Mike Reading, Product Engineering Team Lead at i-PRO Americas explores how AI scene change detection changes the game for video analytics.

Most professional video surveillance cameras have some type of motion sensing analytic that sends an alert when a line is crossed,or motion occurs in a predefined area.

As we know, pixel-based motion analytics are very prone to false positives because they can be triggered by something as mundane as a cloud passing by, causing a shadow to move across the camera’s field of view.

Great strides have been made with the introduction of AI-based object detection which recognizes people and vehicles and ignores shadows, cats or anything the wind moves. AI has taken video analytics to the next level by all but removing false alarms and providing a wealth of useful data about the movement and descriptive data attributed to known objects.

So what’s next in the world of AI-based analytics? Welcome to the exciting new world of AI scene change detection.

What’s different about AI scene change detection?

Unlike analytics that simply track when an object moves across a line or into a region, AI scene change detection considers a scene in its entirety.

To do this, it learns what the normal scene looks like and if something in the scene changes and remains changed after a preset time limit, the analytic will then sense an anomaly and send out a notification. For example, you might have a room with an outer door that is normally closed. While it’s not a problem if the door is opened occasionally, if the door is left open continuously for over 10 minutes, that might be cause for a notification to be sent to security.

AI scene change detection is particularly good at object left behind or illegal parking scenarios. It can also be used to monitor stock levels on store shelves or in warehouses. Having a security camera do double duty as a stock-out monitor are perfect examples of the enhanced utility this application can provide for both security and operations use cases.

As objects in motion in the scene, such as humans and vehicles, are excluded, false alarms are rare. This new level of logic and granularity is something that previous generations of analytics simply could not achieve.

Learning a scene

i-PRO’s edge-based AI Scene Change Detection app represents a first in the industry. It can optimize the scene detection algorithm by learning from captured images while also using AI-based object detection to automatically exclude people that may be entering and exiting the scene.

To configure the Scene Change analytic on location, an integrator or end-user simply captures a series of representative images at intervals using i-PRO’s configuration tool. This enables the application to learn about normal differences that might occur in the scene such as lighting variations over time. Once the scene is memorized, the user can then select specific regions within the scene to monitor. For each region, the user can set the amount of change necessary to trigger an alert and a time threshold in which to ignore brief changes to the scene.

As i-PRO’s AI Scene Change Detection is a free application for all i-PRO AI cameras, the new application will be deployable to more than 100 i-PRO AI camera models upon release. This demonstrates the benefit of having an open AI platform which allows new AI analytics to be added as they are developed.

The AI Scene Change Detection app can also be operated simultaneously with other apps such as AI-based motion detection of objects, allowing security professionals to conduct multiple detections with one camera. A high-end server is not necessary as all processing occurs on the edge. When the analytic detects an anomaly, the event trigger can be received by the i-PRO Active Guard plugin installed on popular VMSs such as Genetec, Milestone and Video Insight.

The possibilities for this new technology are endless and represent the next evolution of edge-based analytics for video security. Most importantly, it’s here today.

Michael Reading is the Product Engineering Team Lead at i-PRO Americas Inc. and has been working in the IT industry for over ten years.  He started his career at Video Insight and has held a variety of roles from Technical Support Specialist and Sales Engineering with extensive knowledge in product development and pre- and post-sales support.