Check out the on-demand sessions from the Low-Code/No-Code Summit to learn how to successfully innovate and achieve efficiency by upskilling and scaling citizen developers. Watch now.
Surveillance cameras already dot or blanket major cities, but it’s not necessarily easy to continuously track a person or object moving through multiple locations and cameras. Now researchers from the Indian Institute of Science are applying advanced artificial intelligence algorithms to this complex challenge, creating a software platform called Anveshak — Hindi for “investigator” — that simultaneously manages the specifics of tracking and the larger issue of working within a city’s limited computing resources.
Anveshak has the big-picture ability to know the locations and overlap points of 1,000 camera feeds, as well as possible paths an object (such as a stolen car) or person could take through those feeds, critical details in limiting what would otherwise be an unfathomably large quantity of video data coming from multiple cameras. The system creates a “spotlight” on the tracked subject, dynamically adjusting the size of that spotlight based on known gaps in camera coverage; for example, four cameras might be monitored for the subject’s arrival in situations of ambiguity, decreasing to only two cameras where their coverage is better and the subject’s route more obvious. Where computing power is limited, Anveshak can automatically cut video quality to reduce bandwidth rather than stalling or stopping tracking activity.
The research is significant for technical decision makers for two reasons: Anveshak has obvious public safety applications, ranging from crime fighting to the automated routing of ambulances, but it also has future enterprise potential in optimizing the performance of factory, mass retail, and other large-scale computer vision systems for enterprises. Beyond the “spotlight’s” thoughtful algorithm for focusing computing resources on only relevant data, Anveshak supports multiple wide-area networks that both cities and enterprises are adopting, including cloud, edge, and fog computers. Unlike existing bespoke multi-camera analysis platforms, which are built solely for specific hardware and software, Anveshak allows users to employ custom tracking strategies, computer vision tools, and reusable algorithms.
While the researchers are already focusing on Anveshak’s privacy implications for surveillance, including limitations that could block the system from being used to track people, or allow it to track adults but not children, there’s clear value in automating the monitoring of anything’s or anyone’s location across multiple cameras. For now, Anveshak is limited to tracking one object at a time, but the researchers are working on ways for the system to track multiple objects at once — something that will no doubt increase processing demands considerably.
Intelligent Security Summit
Learn the critical role of AI & ML in cybersecurity and industry specific case studies on December 8. Register for your free pass today.
Early access to the research paper, titled “A Scalable Platform for Distributed Object Tracking across a Many-camera Network,” is available now in the January 2021 IEEE Transactions on Parallel and Distributed Systems. Research on Anveshak has been ongoing for several years, and the technology was part of an award-winning entry in the 2019 IEEE TCSC SCALE Challenge.
VentureBeat's mission is to be a digital town square for technical decision-makers to gain knowledge about transformative enterprise technology and transact. Discover our Briefings.