As smart surveillance technologies come to the fore in the security market, manufacturers are adding functionality that delivers everyday benefits for end users. One area often addressed is making system operators more effective, and auto-tracking PTZ cameras falls under this requirement. There are two approaches to achieve this, but which is best?
It has long been recognised that one of the weaknesses of video surveillance is the human element. Even the most fastidious individuals have a hard time concentrating for prolonged periods of time, especially if the conditions under which they are working are not constantly engaging. Research offers up a number of figures, the most common being 20 minutes, and whilst it is possible to refocus on an on-going basis for many hours, low levels of mental stimuli do have a derogatory effect.
One area of concern for video surveillance operators is that incidents can be missed, purely due to a lack of concentration. An important event might only last a few seconds, and unless action is taken immediately, it may be too late. What makes the situation more challenging for system operators is that they often spend hours on end watching video of nothing happening. This makes refocusing a challenge.
In an attempt to address this, smart technology has increasingly been deployed in an attempt to reduce the time operators are looking at video where incidents or events are not happening. Using intelligent video analytics, technology can filter out scenes of little or no interest, enabling suspicious incidents or exceptions to be presented to operators.
Because every notification requires assessment and some form of action – events are either escalated or cleared from the system – it becomes easier for operators to refocus.
It has been proven that analytics can make operators more efficient, whilst also ensuring that incidents are not missed. This increases the return on investment for many surveillance projects.
The benefits of autotracking
Autotracking is not new. Video surveillance manufacturers have offered some type of autotracking using PTZs for many years. In truth, many of the early examples did leave much to be desired.
Typical issues included PTZ cameras following any activity regardless of what it was, often swapping the tracked target if two people walked passed each other. Some would lose the target in shadows or if they passed bushes or trees blowing in the wind. If a target stopped moving, the camera might move back to its home position.
Autotracking was something that many PTZ cameras included, but few integrators or installers deployed it in applications.
The reason that the manufacturers’ R&D departments kept working on the technology was because a good and credible implementation would tackle the significant challenge of ensuring operators dealt with actual incidents and events, rather than mundane monitoring.
The increase in processing power in recent years allowed video analytics to become significantly smarter, and the same is true of autotracking. Today’s systems can use other analytics to ensure that a specific target is tracked regardless of other motion in the scene, can ignore periods of inactivity and can even recognise behaviours that are exceptions to the norm. Handover to other cameras is also vastly improved.
However, while autotracking may have become more robust and credible, there is also a second option that many see as being a further improvement.
The typical approach to autotracking makes use of analytics inside the camera itself. This allows the camera to make decisions about what constitutes a target, or the operator can select targets of interest, allowing the camera to do the work when following them. This reduces the possibility of overshooting when manually tracking, but this shouldn’t be an issue for a competent operator.
If analytics are deployed to identify potential targets, the autotracking functionality then makes more sense. The camera can effectively detect a violation or exception and track those involved, sending a notification to the operator that an event needs investigation.
The credibility of autotracking varies from manufacturer to manufacturer, and essentially is dependent upon the way in which metadata is gathered and used. Some manufacturers who have focused on high quality IVA provide very good autotracking, whilst for others the cameras need to be operating in a sterile zone to ensure consistency.
A different approach, one that is gaining momentum, is geospatial tracking. This uses a combination of detectors and PTZ cameras. Because it not totally reliant on visual information being used by the internal IVA, in some applications it offers a high security alternative.
The alternative option is to use laser or radar detectors to detect and track motion around a site. Because these devices output alarms using X and Y coordinates, they effectively pinpoint the position of intruders. By using a site plan it becomes possible for the exact location of an intruder to be sensed, in real time, in relation to a site map.
The coordinates generated by the detector can be logged, either through dedicated mapping software, a plug-in or application on the PTZ camera, or via a compatible VMS, to produce accurate tracking information which is used to control the PTZ cameras.
If a group of people are detected and they then split up, the software can take a central point and zoom the camera accordingly to ensure an operator can monitor all of the intruders. Additionally, when coverage of one camera ends and another begins, or where building obscure views, the software allows the system to ‘hand over’ to other devices based on their actual positions.
This approach enables integrators to design smart and innovative solutions that rely on accurate detection, thus meeting the needs of high risk applications where pixel-based tracking might not be consistent enough.
Because of the accuracy of the detectors, the operator only needs to view footage, make an assessment of the situation and instigate an appropriate response.
Autotracking is a valuable tool for some applications. While pixel-based autotracking, using IVA and metadata gathered from the video images, does deliver a range of benefits, the option presented by tracking driven through accurate detection ensures that the technology can be deployed in higher risk applications.