Sensor Fusion Example #2
Boolean Logic vs Sensor Fusion - Pedestrian During Wind
Our second example has a similar windy condition, striking the North fence of the installation in both zones A and B. However, in this case, there isn’t an intruder but a pedestrian (green person) who’s walking by the site.
Since there is both fence motion and people detected within zone A, the Boolean response to this situation is the same as in example #1 – in other words, four alarms are reported for each subzone between 0 and 100 meters (zone A) while no alarms are reported for zone B. However, because the person is just walking by and not attempting to scale the fence, it means that there are four nuisance alarms triggered in this case.
The sensor fusion system can distinguish between the walking pedestrian and intruder because it understands nuances present within the camera video feed and the fence motion sensor data stream due to the machine learning used to train it. For example, the person is facing to the side rather than straight on and the person is slightly away from the fence. Similarly, the fence sensor indicates a non-human activity moving the fence. These behaviors would have been one of thousands of different benign and malevolent test inputs used to train the system before it was deployed. Thus, even though there is alarm-triggering activity happening at the sensor level, the system matches the signal patterns with non-intrusive behavior to correctly determine there is no threat, and no nuisance alarms are created.