Thanks to advances in vision sensors, compression, communication, storage, processing and data analysis algorithms, retailers are able to turn video surveillance from a defensive, security-focused cost overhead to a proactive, customer-oriented revenue-generating asset.
How many times have we watched “Caught on Video” store crimes and bemoaned the low-res, grainy video that made perpetrator identification almost impossible? This assumes the act was caught and stored to begin with, and that it also wasn’t erased to make room for the next recording cycle.
Those artifacts of a bygone era are fading fast as the Internet of Things pulls video surveillance into the 21st. Leveraging the work of companies such as Nexcom and Kontron, retailers will be able to not only prevent crimes before they even happen, but also to use surveillance to enhance the shopping experience – and their own bottom lines.
The innovation that makes this all possible starts with high-definition CMOS image sensors from the likes of OmniVision, Canon and Samsung, all of which have been competing vigorously in the mobile handset category. The fruits of their labor have transferred to in-store video cameras where size and power consumption are not as much of a concern, but high definition can be critical. Canon just introduced a 250-megapixel CMOS sensor last year. Combined with good optics (lenses) to get a wide-area view, and smooth motor drives for accurate and response zoom, the sensor and camera are where the video and imaging journey begins for retail.
Figure 1. Enabled by high-resolution, low-cost CMOS sensors and combined with wide-area optics and good zoom drives, high-quality video capture is only the starting point for the IoT-enabled retail store.
There are many stages upstream, but three of the most exciting for retail in the context of IoT are storage, analysis and the communication of outcomes.
In storage, the massive amounts of digital data coming from high-resolution cameras would have swamped yesteryear’s storage systems. Now, the rapidly falling costs of both semiconductor and hard-drive memory means low-cost networked video recorders (NVRs) can store from gigabytes to terabytes, and can scale easily if necessary. However, advances in compression/decompression schemes, such as H.265, mean that not as much memory is needed to store a given frame, and not as much bandwidth is required to transmit the information. That eases the pressure on networks and storage, even as they become more capable.
That said, coordinating and making use of the video feeds from multiple cameras requires a lot of processing horsepower and a dedicated system to host it. That’s where Nexcom comes in. It introduced the NViS 1410 1U-size NVR last summer, breaking ground with an Intel Pentium N700-based system that can store 4K video, uses H.265 (HEVC) and a 1 x 3.5” hard-disk drive bay for up to 6 Tbytes of storage.
A key aspect of any such device is the number of inputs, outputs and what speeds they can support so they can be matched to the store layout plan, which determines the number of cameras required. The NViS 1410 has one HDMI 1.4 and one DVI-I video output can support up to 16 network cameras. It also has mSATA and eSATA support.
Figure 2. The NViS 1410 from Nexcom supports up to 16 networked cameras and features an HDMI 1.4, DVI-1, and regular VGA port. However, its use of H.265 HEVC is what gives it the added value for flexible, strategic deployment of fish-eye or PTZ cameras.
For larger deployments, companies like Aaeon offer the likes of the NVR-B75, a behemoth with up to five hot-swappable SATA bays and based on the Intel Core i3/i5/i7 processor and B75 chipset.
But the size of the NVR isn’t what’s important here: it can scale as needed, though opting for a solid, scalable foundation is generally good practice.
What’s important here, and where Nexcom has really shone a light, is what’s possible once the data is acquired. Allow me to get personal here, just for a minute.
My wife is a physical therapist and she can identify a friend two blocks away by their gait: everyone is different. She doesn’t even have to see their face or hear them talk. She’s also armed with massive memory capacity and instant recall. She beats me to the punch on Jeopardy almost every time.
Now, take that expert knowledge and capability to a retail store, and a system can identify a would-be thief or assailant from gait analysis and recognition alone. Face masks don’t matter anymore. The algorithms are being put in place to do just that as we speak, and much more. From zooming in to review contested transactions to monitoring customer traffic and patterns, down to eye movements that may indicate pleasure or disgust, a high-resolution surveillance system, when combined with the right analysis, can affect how much revenue a retail store can generate and increase its immunity to crime.
The next step is to communicate the information and alert both store owners and emergency services to unexpected events. That’s where systems such as the NViS 1410 come into play too. They can send alerts to mobile phones and remote displays, anywhere, while also being connected to the store’s security services.
Figure 3. NVR 1160 is the older brother to the NViS 1410, but the principle is the same: a store can be video connected, data stored and analyzed, and stakeholders notified whenever an incident occurs.