Security Dealer & Integrator

NOV 2018

Find news and information for the executive corporate security director, CSO, facility manager and assets protection manager on issues of policy, products, incidents, risk management, threat assessments and preparedness.

Issue link:

Contents of this Issue


Page 45 of 70

November 2018 / Security Dealer & Integrator 43 algorithms embedded in the device can be limited due to processor and storage constraints. Standalone video analytics can be more expensive than embedded video analytics, involving dedicated hardware, wiring and the cost of the software itself; however, they tend to result in higher performing algorithms, additional software features and flexibility integrating with other sensors and/or existing security systems. 2. Rules-Based vs. AI: Whether to choose rules-based or deep learning video analytics comes down to the complexity of the scene and when your customer wants to devote manpower to the installation. Rules-based video analytics require the integrator to program the system for exactly what to detect and where – the more detailed you can be, the better. On the positive, these algorithms work the minute they are turned on; the downside is they detect as the rule as written, so if they are creating false alarms due to a difficult scenario (severe lighting, shadows, etc.), they will continue to do so until the rule is modified or the scenario is changed. Deep learning video analytics – commonly referred to also as AI or artificial intelligence – consist of a neural network that is “trained” to understand what the customer is and is not interested in. The software will come with a training set of data, but there will be a break-in period where a human in the loop must let the software know when it has done well, and when it is alarming on something that is not of interest. Out of the box, deep learning will not likely pass a stringent acceptance test; however, given the time and proper feedback, the system will lean to be more accurate and become adept at avoiding false alarms related to difficult scenarios. 3. Cloud vs. On Site Server: Selling cloud-based vs. on-site video analytics can be a difficult choice for the integrator. On-site video analytics is the traditional sale of hardware, software and installation directly at the customer location. Although there is opportunity for future hardware and software upgrades, this is likely a one-and-done opportunity. Cloud-based analytics involve little to no hardware at the customer’s site beyond the cameras themselves, with most of the heavy-lifting hardware and software being located remotely “in the cloud.” That means there is not a significant amount of installation labor or product to sell at a mark-up. Most hardware or software upgrades happen in the cloud, unbeknownst to the end-user. As such, the initial business case of these installs may not look very attractive from a resale aspect; however, often the integrator can receive a small commission (RMR) from the monthly cloud-based service. Even small amounts can really add up over time, and in most cases, this is a purely passive income stream. Don’t Forget Your Current Install Base An often-overlooked segment for video analytics is the retrofit market. Today’s video analytics easily integrate with existing NVRs and cameras; thus, the ability to add increased automation and detection may be as simple as adding an edge device or video analytics server. That means that every customer in your database is now potentially a new lead to update their system with video analytics. In addition to the sale itself, regaining the business of an existing customer is much less expensive than capturing the attention of a new customer. Approaching them with the value of adding video analytics to their system is a low-cost sales action, but despite the higher return on investment associated with retrofit opportunities, upselling video analytics to an existing install base is often an overlooked opportunity for integrators. Planning The sales and installation of video analytics always includes assessment of camera placement, camera views, camera mission and lighting – yes, always. A key aspect of a business case is whether this cost is accounted for. An integrator can choose to absorb the cost or include it as part of the installation charge to the customer. Deep Learning video analytics use neural networks that must be "trained" to understand what does and does not constitute an alarm. Photo: PureTech Systems

Articles in this issue

Links on this page

Archives of this issue

view archives of Security Dealer & Integrator - NOV 2018