Follow the Red Box. Watch for yourself how AISight sees and responds to unusual behaviors in the following situations. Unlike other products in the market, no rules or trip wires were set up in these locations; AISight learned autonomously and sent alerts in real time.
Alert Examples
Traffic 1
With AISight's ability to detect and distinguish between normal and abnormal traffic flow, it signals when a truck crosses the center line- passing through the gate the "wrong" way.
Beach Perimeter Section 18
This alert catches a human, seen through a chain-link fence, in a secure area of a beach. Notice the moving trees and bushes; they do not cause a problem for AISight. No other analytic is able to handle the complex elements in this scene. For example, 1) The geometric nature of a chain-link fence make it nearly impossible to recognize objects on the other side. 2) The movement of the trees would cause hundreds of false alarms. 3) There is no distinct path in which to draw a trip line without generating erroneous alarms. 4) The distance from the camera to the human is too great for other systems, the human is not big enough to be detected, much less identified as such.
Rear Access Building 6B
This alert was generated when two individuals tried to place luggage in a populated area.
Check Point 14
At this security checkpoint, its normal for the driver to exit the vehicle occasionally but AISight alerts when the passenger tries to leave the car!
Public Access PTZ Position 4
AISight alerts when a human is seen walking in a roadway where only vehicles travel.
Transit Alert
At this transit customer location it is important to catch people that enter the rail area or leave suspicious objects around the tracks, however, it's the unforeseen or unexpected behaviors that most often indicate potential dangers. No one would have programed this camera to catch a person driving his car behind a train to save time on his morning commute. AISight labels moving objects with virtual "fingerprints"; detailing dozens of attributes like aspect, color, shininess, mass, and size. While trains of many sizes and utility vehicles have used this track in the past, AISight has never seen a personal automobile following a train so this alert was instantly generated.
Unusual Time of Day
Ten days after a deadly terrorist attack on this facility, AISight was installed. In this camera's view the activity has been very extensive as the facility went through new construction and upgraded security structures. AISight learned and adjusted for the ongoing activities of construction workers without issuing erroneous alerts. But when this person carrying a gun entered a secure gate at the wrong time of day, an alert was instantly issued. The fact that he was carrying a rifle did not factor into the alert, guards carrying guns is not unusual in this area. The alert was issued because this guard was in the area at the wrong time of day. AISight tracks behavior by time of day and day of week to make sure activities and objects considered normal in one time space does not preclude an alert when that same activity occurs at a time that it shouldn't.
Manufacturing Alert
In most manufacturing and processing environments it is impossible to foresee every possible danger that could arise, whether it's from criminals, terrorist, or normal day-to-day operational failures. In this AISight alert, a large tank containing toxic chemicals has just begun to vibrate and lose its secure mounting. This sort of early warning mechanism, only available with behavioral recognition, could be the difference between a maintenance call and an emergency 9-1-1 call. AISight has been credited with forewarning of other environmental dangers such as fire, smoke, steam, chemical leaks and spills long before conventional sensors were able to detect them.
Secure Facility
With today's critical security environment the activities outside of many buildings should be heavily scrutinized. In this alert, a truck is noticed backing up to the entrance door of a secure facility where the usual behavior is to drive forward and around the turning court.