For more efficient security, see how this surveillance system “learns” to i.d. people and filter out false alarms

by Brianna Crandall — September 27, 2017 — Hikvision, a supplier of video surveillance products and solutions worldwide, has launched its DeepinMind Network Video Recorder (NVR), said to be the world’s first embedded NVR benefiting from “Deep Learning” functionality. This intelligent NVR effectively “learns” to identify people captured by video surveillance cameras. The NVR will also learn to filter out false alarms that would have previously been triggered by non-threatening moving objects. This new 32-channel iDS-9632NXI-I8/16S NVR will be the first in Hikvision’s new range of DeepinMind products.

Hikvision DeepinMind NVR

The Hikvision DeepinMind Network Video Recorder uses “Deep Learning” to identify people captured by surveillance cameras and to filter out false alarms.

Hikvision is taking the Deep Learning concept and applying it to the security industry, yielding a series of completely new products. Imitating human beings’ synaptic learning and memory processes, the DeepinMind Series NVR incorporates advanced algorithms to achieve astonishingly accurate and consistent Video Content Analytics (VCA) performance, claims the company, as well as new levels of alarm-activated monitoring.

False alarms triggered by animals, leaves, shadows, changes in lighting, and other insignificant objects regularly plague security personnel, costing too many man-hours and monetary resources, points out Hikvision. But with its ability to detect human bodies, the iDS-9632NXI-I8/16S NVR is designed to effectively filter out such false alarms.

The DeepinMind NVR identifies and triggers human activity with what the company says is unprecedented accuracy — exceeding 90%. Its high-speed Graphics Processing Unit (GPU) performs accelerated computing while Deep Learning algorithms are said to improve accuracy over against current NVR models that still rely on conventional CPUs. Moreover, these features reportedly enable authorized users to search recorded footage and find targets far more quickly than with a traditional NVR.

Keen Yao, vice president (VP) at Hikvision International Business Centre, stated:

With traditional edge-based VCAs employed in camera heads, alarm activation accuracy relies on strict requirements for a scene’s background. To date, the accuracy of their intelligent recognition and analysis in comparable scenarios can be inconsistent. With the introduction of the DeepinMind embedded NVR, Hikvision is offering installers and end-users a whole new level of alarm activation accuracy, and with that, more efficient surveillance systems management.

The iDS-9632NXI-I8/16S DeepinMind embedded NVR can simply replace an existing DVR or NVR, introducing intelligent learning previously unavailable with traditional surveillance systems in just one step. Such an upgrade enhances conventional video surveillance performance for markedly more efficient security, says Hikvision.

The DeepinMind embedded NVR also features up to 32 channels of video input for Internet protocol (IP) cameras (up to 12 MP), high-definition (HD) video output and decoding, hard drive hot-swapping, RAID (0, 1, 5, 6, and 10) configurations for up to eight 8-TB hard disk drives (HDDs), and dual-NICs (network interface controllers).

Hikvision has also taken the Deep Learning technology and innovated a family of products to maximize its use, including the DeepinView IP camera range and DeepinMind video analytics server. Keep an eye on the Hikvision Web site to see launches of these products as well.