printer friendly version

Getting started with video analytics

Video motion detection (VMD) was first implemented more than four decades ago as an analogue system, but failed to deliver due to high false alarm rates and product and installation costs. The evolution of electronics, PCs and the Internet has led to a new form of digital VMD that has proven unsuccessful as a result of poor algorithm designs and the heavy processing required to accommodate outdoor scenarios. As a result of advances in miniaturisation technology and increased processing power, a new form of VMD was born: video analytics, a more cost effective, powerful and efficient solution which is also simple to set up and operate.

Architectures

Four different video analytics architectures can be identified in today’s market: software-based; OEM hardware-based; hybrid; and edge devices.

A decisive factor in choosing an effective video analytics system is its capability to maintain a low rate of false alarms while allowing a high rate of probability of detection (PoD). A system that generates frequent false alarms by reacting to extraneous stimuli is ineffective and counteracts the main purpose of video analytics, which is to deliver only relevant information. On the other hand, the system’s PoD needs to be high enough to ensure that relevant security events are not missed.

The following parameters allow users to achieve a fine balance between PoD and false alarms:

* Defining the detection zone is necessary in order to reflect the area of interest, since the camera has a large field of view (FoV) and only specific areas are relevant for detection purposes. There are three types of detection zones that can be defined by most systems: active (alarm zone); inactive (pre-alarm zone); and passive.

The active detection zone (or alarm zone) refers to the areas where an alert is required when a certain event occurs. The inactive detection zone refers to areas within the camera’s FoV but where activities are ignored. The passive zone (pre alarm zone), is defined as a zone where the system is aware of activities but will deliver an alarm only when an activity passes from that zone to an active zone – for instance, it could be a fence separating a public and private area.

* Detection rules define the types of detection that are of interest, such as intruders, tripwires, unattended baggage, loitering and stopped vehicles. Detection rules are important in order to reduce false alarms. For example, one may wish to identify persons entering a location from an unauthorised access point such as an exit, while allowing free passage to people exiting that location. In this case, the detection rule would define the sense of movement that is allowed.

* Depth setup enables a computer to translate 2D images into 3D viewing. The human brain has the innate capability of understanding the size of an object in relation to its distance. Depth setup gives a computer similar capability, enabling it to distinguish between an aeroplane in the background and a fly on the lens.

Other key elements that affect the reliability of video analytics in detecting security events and ignoring irrelevant information include lighting conditions (minimum required light intensity within the detection zone at night); defining the expected speed and size of an object; and selecting the appropriate camera lens for the desired coverage area.

A key challenge that system integrators face when dealing with video analytics is the customer’s desire for a system that can be rapidly installed, minimising surveillance downtime due to installation; and simple to operate, minimising the learning curve.

The basics of reliability

In tandem with the increased interest in video analytics over the past five years, many new companies have emerged in the market. With varying levels of performance and no unified standard, the user would be well advised to find out how well a system performs and seek installation references.

The military represents the perfect test bed for security products. With diverse locations and challenges such as rain, snow, wind, darkness, animals, cloud shadows, slow-moving or camouflaged intruders and far distance detection, the army is an ideal organisation to test and approve a new technology and provide feedback for improvement.

Many technologies available to the public today originated in military labs and testing grounds. A video analytics system that has passed the harshest tests in the most difficult terrains is obviously suitable for high-risk sites. The criteria for success in a military test include a low false alarm rate; high PoD; ability to perform well over long periods of time (sometimes measured in years); rapid deployment; and ease of integration. These factors, along with a bulletproof design, are all necessary for off-the-shelf purchasing and installing.

Probably most important, are references from actual system installations around the world. The more sites the system has been installed at, the more diversified is the manufacturer’s experience in dealing with myriad detection scenarios and false alarms reduction.

Share this article:

Further reading: