printer friendly version

Edge or server analytics

Many end users rely on integration companies to provide the best solution to protect their assets and property, however, the person making the decision needs to consider the following design criteria related to technologies being supplied in solving their security solution.

Over the years there has been much hype around analytics as well as the best option to be used, where integration companies have sold the product with a realistic performance outcome, they have been successful, however, many manufacturers and integrators have over-promised on performance of the product, which has led to many end users losing faith in their performance and reliability.

Today we see a marked improvement in this arena with many of the products offering an array of functions. However, as before, the integrator selling the product must understand the product and sell it correctly to meet performance criteria

When deploying these analytic applications, one should consider the following aspects:

1. Where is the processing of alarms taking place?

2. What are the phase costs associated with deployment of technology?

3. Updates related to improving performance.

1. Where is the processing of alarm taking place?

There are currently two options available:

Edge based analytics: An integrator can supply an encoder, camera, or device in the field upon which the analytics is loaded to allow selection of the applicable analytics required, this processing of alarm situations will be done on the edge and will not draw any additional processing power from the main server related to running or processing the analytic alarm.

Many of the edge-based devices now provide larger more powerful chipsets which provide a more comprehensive array of analytics with better performance. The analytics selected can be configured to inform operators on a wide range of real-time video or audio events requiring attention. Many of the devices already include video analytics such as loitering, line crossing, face recognition, entry and exit and direction detection. IP Audio devices can recognise and create an alarm for glass breaks, gunshots and screams. In addition, people counting, queue management and heat maps bring new opportunities for business owners.

There is a definite move towards AI (artificial intelligence) called deep learning. Deep learning takes machine learning to a new level based on neural network theory that mimics the complexity of the human brain. This AI analytics is relatively new to the market and is continuously improving.

These edge-based devices normally are cheaper over the long run to add to a system in a phased approach as you pay for analytics as you add to system; many of the Event Video Management System (EVMS) will allocate a device licence as you add channels. Obviously, this saves a huge amount of costs in terms of your server requirements for processing of alarms.

Server-based analytics: Server-based analytics is better suited for certain tasks related to running advanced analytic functions as well as incorporating AI functions. A good example of this will be automated number plate recognition (ANPR) as this can be done in the field on cameras, however, it offers better and more advanced features when run on a server, for example, checking colour, make, model and number plate of a vehicle, as well as comparing to databases and interfacing with products such as SNIPR (or the national database for stolen or known criminal syndicate vehicles). This will then allow data to be cross-referenced and flagged.

Likewise, facial recognition requires sophisticated database analytics for the same reasons. Server-based analytics is generally more expensive than edge devices with in-built analytics, as they have a limit to the number of channels they can handle in accordance with server specification and performance. This is normally a much larger cost and in many cases, you need to buy a server that may in essence only handle a few channels to start with and which allows later expansion, bearing in mind that many of the devices on the market already have analytics onboard for use at the edge.

One needs to plan the system correctly from the start in terms of what is the objective: to detect an object or a specific person, which can be done on the edge, or apply machine learning and deep learning algorithms that continuously learn and require the computational horsepower and storage that only a server can provide.

Server-based analytics with advanced features have their limits. An example, if a server is going to analyse video, it must start by decoding it, which consumes a fixed amount of CPU/GPU resources. Once the video is decoded, then it can be analysed. As the video channel count grows, it’s possible to saturate even a powerful server very quickly.

Edge devices do not have this problem because analytics are typically run right after the image is captured and before it is encoded into an IP video stream. The results of the analytics are then passed as data to the VMS, NVR or messaging system as metadata, which is quite small compared to a video stream.

You can imagine from the reasoning above that for powerful analytics processes, such as number plate or facial recognition, it will become more cost-effective to combining edge analytics with server-side processing and database management.

2. What are the phase costs associated with deployment of technology?

When looking at a server-based analytics system which will be able to decode and analyse 20 channels of video, the same server, processing metadata provided by the edge-based camera or device could analyse over 250 channels. The cost savings of such an implementation is substantial

3. Updates related to improving performance

In both cases it is important to understand the manufacturer you deal with, how often they update their products and add new features as well as the cost thereof. It is normally very easy to update, you receive a software patch that you upload on the network and then propagate to all devices, but bear in mind you may need to reconfigure devices if parameters have changed.

Analytics and AI on the edge

Analytics and artificial intelligence (AI) are proving to be a very valuable tool for customers to improve their security landscape as well as the operational efficiency of their business.

The additional business intelligence (BI) provided especially by edge analytic devices are allowing customers to adapt their processes and procedures to the next level and have huge advantages in terms of profitability and customer experience benefits.

Edge analytics in terms of facial recognition/collection or licence plate recognition allows companies to gather their intelligence and build up databases that will give them the ability to differentiate their product and services from mainstream competitors. Edge analytics is not only efficient and responsive but also has some key benefits in terms of data and communication costs associated with these offerings.

Companies that use edge analytics gain much deeper insights into customers’ demographics and behaviour. These features have proven to become more valuable as they are being integrated into normal business processes and procedures.

For more information, contact Secutel Technologies, +27 10 015 1401, sales@secutel.co.za, www.secutel.co.za

Share this article:

Further reading: