Two months ago we launched a new type of activity alert. Our goal was to cut down on the false alerts that are common when using motion-based video analytics in outdoor environments. It’s a significant industry problem: When too many false alerts are sent, users tend to shut out all alerts, which defeats the purpose. Our idea was to apply our cloud-based computer vision (a type of artificial intelligence that recognizes objects or activity in an image or video) to verify the presence of a vehicle, person or both in the video analytics-generated alert before sending it. As part of this initiative, we also overhauled the alerts user experience (UX), allowing customers to define and edit their own alerts, and improving the user interface significantly.
In going through the data in detail we found remarkable results. The computer vision (CV) filtering has cut down on false alerts by 70%, and by also improving the way we track vehicles and people across multiple cameras at a site, we have reduced the total volume of alerts by 90%. Our customers are noticing the difference too:
“I want this on all my sites (…) much better experience” – Security Lead, major oil and gas producer
“This is great. Please put all our sites on the new alerts right away” – Control Centre Manager for midstream oil and gas company
This is just one example of how we can integrate, train and deploy computer vision algorithms to solve real industry challenges. Look for new CV-based industry solutions from us in 2018, and get more information about our approach to computer vision here.