From BRS Labs, AISight is an adaptive learning video analytics solution for large-scale deployments that provides awareness and real-time responsiveness to security watch centers. AISight learns what is normal activity for any given scene or environment, and then accurately recognizes and alerts whenever abnormal activities or behaviors occur. In-house and field security personnel receive alerts in real time via consoles and mobile devices so they can respond quickly and proactively. The system adaptively learns changes to the scene and environment over time and continually updates itself to improve its performance without human input or configuration. This greatly improves accuracy, increases scalability and reduces the number of false alarms.
The idea behind a system that identifies abnormal behavior through an adaptive learning function is that it addresses several business problems currently challenging the video surveillance marketplace. As an example: law enforcement often leverages security cameras to solve a crime after it occurs, even though these cameras rarely detect security breaches in progress nor can they identify new threats before they happen. One option is to invest in more security personnel to monitor the output of hundreds of video surveillance cameras—an incredibly difficult task to perform in real time. Added costs also make this route impractical for most companies and government agencies experiencing tightening budgets – and do not necessarily offer commensurate security improvement.
As a second route, companies have invested in rules-based video analytics software, which tracks pre-programmed behaviors recognized as aggressive, abnormal, or suspicious. Many of these products require human programmers to regularly update the software program to detect newly defined threats, increasing implementation and maintenance costs while still allowing for vulnerabilities that updates fail to take into account. While these systems work for certain applications, other complex environments require a scalable system less dependent on human programming and monitoring.
A system that can automatically observe and refine its model of a scene to identify abnormal behavior will in turn be able to detect, track, and classify behaviors more efficiently over time. As a result, this kind of system minimizes labor and software upgrade costs and improves the effectiveness of operators and security personnel by allowing them to focus on events that have the highest probability of being threats.
