The Future of Enterprise Surveillance

Future of Surveillance and Access Control

The Future of Enterprise Surveillance

With the advancements of artificial intelligence, edge devices and 5G networks, traditional surveillance technology is being disrupted with cloud-managed AI powered video security solutions. This technology has become so powerful and commercially viable that native feature sets such as human analytics, face recognition and retail analytics have been built directly into the camera using AI-enabled chipsets, enabling rapid deployment and unlimited scalability across enterprise sites.

As AI processing becomes the new standard for surveillance solutions, opportunity to incorporate features such as face recognition for enhanced investigation, monitoring and public security to incorporate advanced human analytics built on deep learning algorithms and associated data becomes realistic. From use within public health environments, public safety, private safety – enterprises and government organizations alike can exploit such technology whilst pivoting from the traditional CCTV model.

Cloud Managed Surveillance Systems

When we discuss cloud-managed surveillance and associated edge technology for cloud recording, the question of how do you store CCTV footage in the cloud in predominant. Considering the future of surveillance is cloud based, this can be confused with streaming and storage of associated video data within the cloud. Naturally questions are asked in regards to bandwidth utilization on the network. The discussion of edge technology is now relevant as a hybrid-cloud solution such as Verkada smart surveillance utilize edge-storage and edge-compute architectures to achieve local recording storage whilst utilizing the cloud for metadata processing (features such as face recognition, analytics, command and control). Edge-compute cameras then have the ability to perform local AI calculations and storage without the need for processing power from a locally attached server or NVR – whilst maintaining low bandwidth consumption on a per device basis. Chipsets such as ambarella empower modern devices to perform such compute requirements. Traditional CCTV models would utilize a camera to network video recorder (NVR) based model, whereas modern systems will deploy without the need for NVRs or servers as the workload is performed at the edge (on the camera).

Advantages of considering security video cameras with edge capabilities opens opportunity for deployment within areas difficult for installation, such as regional environments with limited internet access. Will 5G ease the bandwidth problem or change the way cloud surveillance is deployed? Hybrid-cloud based systems can already be deployed on a low-latency based networks (such as LTE), the introduction of 5G networks will enable a greater opportunity for scalability and device management from a single connection without supporting hardware or security issues usually attached to NVR based CCTV systems. With increased bandwidth and cellular availability, edge-based devices running rich video analytics (including people and vehicle analytics) to perform video surveillance operations will become the most commercially sensible option, especially considering edge-security can be deployed across 5G connections via solar powered towers.

AI Powered CCTV

No servers or NVRs, this is the immediate advantage of deploying cloud-managed edge based surveillance services across the enterprise. From maintenance and cost reduction, to simplifying ICT operations to support cloud migration planning, the CCTV market and associated service providers are now having to pivot from traditional based deployment architectures to support enterprise strategy within the video surveillance system space that has been otherwise deployed and managed by non-ICT experts or technical leaders.

The disconnect between responsible stakeholders within enterprises is all to apparent, how can an ICT department take ownership of a system that was specified and installed by a group that has no experience in cyber-security, network engineering and associated ICT related workloads?

With localized AI processing built-in to each camera, IP cameras are now able to exploit machine learning capabilities to support facial recognition, number plate recognition and people analytics to move the system beyond traditional NVR based security systems. With AI, empowering features to support compliance, ICT and retail analytics from the cloud improves overall security management and moves the solution into a proactive state – meaning organizations can take advantage of the video recording as it happens, not after it happens. A traditional surveillance cameras will not contain processing power, but rather rely on a localized NVR or server clusters with heavy storage requirements to support digital camera fleets. With AI and edge-storage, the problems of storage calculations, scalability and associated hardware management have been eliminated – reducing operational risk to failed systems and extraction of specific events.

As with all new CCTV deployments, planning should be considered as to the exact requirements and features required – generally known as a functional specification. For organization’s deploying large cameras fleets over large spaces (eg an airport or stadium) a number of factors need to be considered such as storage retention policies, coverage requirements, indoor and outdoor requirements. With traditional systems, the more cameras added to the deployment the higher server storage and compute requirements are needed. With AI powered mass surveillance systems, problems involving storage and compute capacity planning can be eliminated and moved the edge. Bandwidth considerations are reviewed however considering cameras can idle at rest with low bandwidth consumptions as little as 20 Kbps, deploying 200+ cameras across a floorplan is easily achievable.

  • Review business requirements and evaluate deployment requirements across multiple sites.
  • Calculate the overhead reductions from both a hardware and maintenance perspective on elimination of traditional CCTV technology.
  • Review how AI can transform business operations and streamline compliance, security and actionable events.

This is where AI powered surveillance such as the Verkada smart surveillance system can make a true positive impact to surveillance operations, opening opportunity to enforce social distancing, facial recognition, number plate recognition, people counting and proactive alerting without the need for heavy client installs or thick-clients, but to move the management or administration component of such systems into the hands of the department stakeholders and frontline workers using ease of use applications on desktop (browser), mobile and tablet devices.

Human Analytics

With the use of AI being able to identify humans vs. non-human objects, data sets can be created to provide analytical reporting capabilities and proactive alerts that would otherwise be difficult or impossible with traditional systems (in most cases require a AI compute server to work alongside the NVR). Cameras can now identify human behavior such as recognition of an individual passing frame, humans being seen within an applicable zone or more complex behavior analysis such as motion heatmapping, foot traffic patterns, people counting, loitering, unattended objects and unusual behavior. Face recognition can be easily deployed within commercial environments for VIP identification or person of interest alerting (such as thief or known suspect) directly to the floor managers – enabling rapid decision making as the event occurs. AI powered video analytics are generally deployed either at the NVR or server level in traditional systems, but with hybrid cloud based systems we see the capabilities built directly into the IP camera itself.

Compliance operations see a greater benefit of utilizing AI-powered human analytics as investigation practices are often streamlined making it more efficient to filter through hours of dead footage looking for actionable events. Filtering of individuals from appearance and even face enables mass searching across a fleet of cameras and sites – this is the metadata storage of the command and control platform that frees up local resources or the need for on-premise server environments. Video analytics are immediately available directly from the cloud platform or mobile devices, making deployment to users and external groups streamlined and secure, without the need for firewall alterations or complex configurations. Artificial intelligence built directly into AI powered CCTV systems represents a shift within the CCTV market and video surveillance options.

From a retail analytics perspective, human behavior can be monitored at a granular level to enable smart product placement, marketing initiatives and strategic planning for brand placement. Staff awareness and customer service improvements are achievable by measuring density and activity across a floorplan using tools such as heatmapping, motion plotting and customer flows. The future of surveillance enables enterprises who are using such cloud manage surveillance capabilities across multiple stores consolidate related data within a single platform. The CCTV market offers a lot of different options around deploying retail analytic solutions, some more complex then others. No servers or NVR environments enable retail analytics to be scaled across the enterprise whilst consolidating operations within the cloud for fast analysis across unlimited sites and cameras.

Face Recognition

Can face recognition be fooled? How reliable is face recognition? Common questions often asked when discussing the technical aspects of such systems. The short answer is yes – face recognition can be fooled but in most cases is very reliable, it comes down to the design and methods used to capture and recognize faces including the indexing of such data. Take for example the question of do sunglasses stop facial recognition? In most scenarios sunglasses will not necessarily stop the identification of a person, but depending on the sensitivity of the alerting system some false positives could be produced. To accomplish face recognition systems a number of methods can be deployed, however the most common being deep learning algorithms in which the face profile is indexed and calculates accurate measurements across different key points on the face. Alternative methods include image comparison which is achieved by identifying a human, taking snapshots of individuals faces, indexing the image and then performing image matching across the database. Naturally, deep learning face recognition is more accurate but often more complex to deploy and involves additional considerations such as database storage. processing power and security.

For law enforcement, the ability to identify and find persons of interest can be achieved at scale without complex software or client installs. From the cloud, modern cloud managed surveillance platforms that include artificial intelligence will have the ability to express search across fleets of cameras and sites to find relevant events and POIs to then share out under strict conditions – such as no download of video with loss of rights and access restrictions based on schedule.

No Servers or NVRs

Is an NVR a server? Do you need NVR for IP cameras? Network video recorders, also known as NVRs are in the most part servers. Mostly unconfigurable in terms of the core OS and software flexibility when compared to a standard enterprise server. In traditional systems you always need an NVR or server should you wish to record and manage cameras. Generally the NVR either directly connects to the camera to provide power or they can be IP based and treat each IP camera as channel. This is very traditional setup and in more recent times NVRs are now being considered high-risk exploitable hardware in terms of cyber-security. Other problems with NVRs include scoping of video retention, on-going maintenance, uptime monitoring and lack of scalability across multiple sites. In a traditional sense, the more cameras and storage retention required, an increase in cost is going to be relative in initial deployments. For a system running 200+ cameras requiring 60-90 day retention, we could expect to see a bank of servers installed with limited warranty and unreliable hardware configurations (that are often prone to failure after 3-5 years). As new software is released, cameras and servers alike require manually installed upgrades and new firmware’s which attract maintenance and software configuration costs, that in some circumstances can result in system failures or on-going integration issues if not planned correctly.

Cloud managed surveillance built on edge technologies have essentially built the software directly within the camera and removed the need for NVRs and servers. This in return results in reduced overheads, maintenance requirements, ICT service desk reductions and long-term cost reduction especially when considered over a 10-year period. An attractive commercial advantage of eliminating the NVR and server model is the hardware itself can include extended warranties as the firmware built inside the camera is easily replaceable with a new unit should a failure occur – without impacting the overall system (isolated hardware and condensed risk environment).

Future of Surveillance

Which countries are adopting AI surveillance technology? Verkada smart surveillance systems are deploying their camera technology world-wide thanks to strategic partners such as AWS which provide cloud data services globally, allowing a single integrated platform that can store and distribute the solution across the internet without complex system setups or integration requirements. Large enterprise organizations such as Heinemann Australia upgraded their security systems to the Verkada smart surveillance platform to introduce AI-powered CCTV for compliance, ICT and retail analytics which provide an immediate cost saving of 30% compared to traditional vendors over a 10-year period (with Verkada offering a industry leading 10-year warranty on all hardware). A solution with over 200+ cloud-managed cameras with a range of analytical capabilities that is going to enhance their retail operations and service inline with the companies strategic plan, including people analytics for enhanced customer experience and product placement.

In conclusion, traditional CCTV deployment models have been implemented on a set and forget basis, resulting in system failures, high maintenance costs and ongoing loss of management. High-risk system failure is recognized when the priority events are required for retrieval only to be unavailable. Hybrid cloud-managed systems represent innovation on the surveillance front, resulting in a new perspective in managing physical spaces through modern infrastructure whilst empowering ICT teams to take back control of physical security from the cloud.


Next Steps

Learn more about the future of surveillance with a live demo and walkthrough.