printer friendly version

The video analytics conundrum

Video analytics has been around for many years, however the effectiveness of the solution has been in question as false positives have been the biggest challenge to successful video analytics, as has seamless integration into the surveillance platform.

However before we go in to this discussion I believe it prevalent to go into video analytics and how it works. We also need to understand the difference between video motion detection and video analytics.

Video motion detection (VMD) is a way of defining activity in a scene by analysing the differences that occur in a series of images. This is usually done by pixel matching or frame referencing. Any change between frames is regarded as ‘detection’. The functionality can either be built into a network video product or made available with video management software.

An in-picture alarm feature will allow users to define specific areas of a screen where any visual changes should be detected. Video analytics analyses live or recorded video and generates data for pre-emptive action or data mining purposes. It is generally regarded as superior to VMD as it identifies items of interest, and eliminates the cause of many false alarms, which can provide more useful information.

A set of software algorithms are run to analyse the individual frames presented by the device to do the image analysis. Basically the software automates the monitoring and analysis of video images. A specific process is followed by the algorithms in order for the software to give feedback (acquire, process, identify, analyse, and report.) Specific algorithms are written for different purposes, for example, intrusion detection, vehicle monitoring, licence-plate recognition, abandoned object detection, people counting and loitering. It can even detect camera tampering and failure.

The purpose of video analytic was to enable two areas:

1. To allow less monitoring agents to monitor more cameras and be proactive in responding to the analytic alarms.

2. For the generation of metadata for analysis, an example being people counting.

The video analytic solutions on the market today are either supplied from manufacturers and an activation licence is required, or a third party. The third-party vendors have been around for longer and have a wider platform acceptance.

Although the capability of video analytics software offered by various manufacturers differ, the general functionality is in most cases based on rules, filters or algorithms which can be embedded in cameras, network video servers (NVRs) and other video hardware.

Immaterial of the environment, false positives have been reduced and with the maturing of the technology algorithms rules and alarms, we believe the point where video analytic is ready to go mainstream with widespread acceptance is here. With the escalating costs of staffing, the barrier to market, cost, is no longer an issue and weighted against the monthly cost of monitoring agents and or onsite guarding staff video analytics it is definitely an alternative option.

Kelly Mclintock

What to consider when installing video analytics

Do not believe the sales person: Try it before you commit to the technology.

Run a trial before investing heavily in video analytics. Purchase just a couple of VCA units and run them at a site where you have easy access to the configuration and maintenance interface. Every installation will present the VCA system with different challenges and you should verify the system meets your requirements in terms of detection rate and false alarm performance before embarking on bigger projects. If your installation is outdoors, make sure the system can handle environmental changes without generating false alarms. Check that the system does not go 'blind' in these situations and can still meet your requirements.

Which analytics product

There are many manufacturers of video analytics systems, each with different strengths and weaknesses. Some manufacturers offer systems targeted at specific situations, such as intrusion detection, traffic surveillance and so forth. Make sure you know what you want and if you have multiple specific scenarios it might pay to use different manufacturers for different applications. Look for manufacturers that have proven solutions in your vertical market

Edge or server based?

Many new cameras and IP encoders are hitting the market with the ability to run the analytics inside the camera, known as 'analytics at the edge'. In a large number of cases, edge devices offer benefits over server-based systems. Since edge devices run the analytics locally, they can be configured to send a video clip only when alarms are detected, lowering the bandwidth requirements, thus saving costs. On the other hand, server based analytics might be the right solution if you need to run video stored on your NVR through analysis in retrospective viewing. One thing worth considering: you can make an edge device behave like a server system, but not the other way around.

Camera placement

Despite what the sales material will tell you, VCA is still a maturing technology and it is not a magic solution that will work in all situations. The performance of VCA systems has improved over recent years, but camera placement can have a dramatic effect on how well your VCA functions. Avoid locating cameras where nuisance alarms are likely to occur, for example where there is a lot of moving foliage, waves, changing tides or where there are numerous light changes, etc.

If this is not possible, then problem areas should be masked at setup stage. If you plan to install analytics on a PTZ camera, make sure it can support automatic detection.

Share this article:

Further reading: