Back to the future: continuous recording within grasp (once again)

Timo Sachse

Outside of the security sector, the perception that many people have of video surveillance is shaped by what they’ve seen in movies and TV shows; cameras always recording, with the footage saved onto video tape and stored until whichever detective or private investigator wants to review an incident.

There’s some truth in this perception, of course, but it’s very outdated. It’s largely based on the world of analog surveillance cameras, where the limitations of the technology demanded that they were either on or off, and when they were on, they were always recording. This had some obvious advantages, of course, with no incident seem within the camera’s field of view being missed and available for review by security personnel and law enforcement. At least, that is, for as long as the footage was retained. And this created a problem in physical storage.

With video tapes being relatively large and only having capacity for a few hours of video, the space needed to keep tapes quickly became a problem. As a result, footage would only be kept for a certain length of time – maybe 30 days – before the tapes were used again and new footage recorded over old. This was problematic for a number of reasons: the most obvious being that previous footage would be lost forever, but also that the quality of recorded video would deteriorate over time as tapes became worn.

The arrival of digital video surveillance, and data

continuous recording
Left: The first sketch of AXIS 200; Right: 2021 Bullet Camera sketch.

Fast-forward (excuse the pun) and the arrival of digital network video surveillance seemed to solve the physical storage issues. Hard disks were able to store huge amounts of digital video in a relatively small physical space – amounts that would require vast warehouses to store the equivalent saved on video tape.

Problem solved? Not quite. Network video cameras evolved rapidly, particularly in terms of the quality and resolution of the images captured. New cameras brought higher resolution, increased frame rates and bitrates and, in short, created more data. Much more data.

With surveillance systems containing tens, hundreds and even thousands of connected cameras, the storage demands of digital video surveillance – along with the bandwidth needed to send this information between cameras and data centers – quickly became an issue.

Sub-optimal surveillance as a result

Organizations quickly started looking for solutions to the issue of data. An obvious approach – one which digital surveillance cameras readily enabled – was to only capture video when ‘something happened’.

This was most often achieved through motion detection algorithms, with cameras being activated when a certain threshold of movement in the camera’s field of view was met. However, this is an imperfect solution. Depending on the sensitivity thresholds set on the motion detection sensors, cameras could be activated too often, leading to a high number of false alarms. If the motion thresholds were increased as a result, meaningful activity might be missed.

Other solutions involved reducing the amount of data produced by the surveillance camera itself. Video compression, decreasing frame rate and lowering video resolution are all ways to achieve this, but also impact the quality of the video images captured.

Whichever way the data burden was addressed, continuous recording fell out of fashion in many environments. Organizations were willing to accept the compromise of incomplete or lower-quality video to reduce the storage and bandwidth burden. But this no longer needs to be the case.

The new opportunity and need for continuous recording

Most people would agree that when it comes to video surveillance, continuous recording is a good thing. It’s obvious to say that not missing any event in a camera’s field of view is of great value. New innovations in surveillance camera technology now make this a possibility, without increasing the data burden (and in some environments, actually decreasing it in comparison to previous stop/start approaches).

The simplest approach to video compression is to reduce bitrate – essentially the amount of video transferred in a certain period of time, typically expressed as bits (or megabits) per second. This is often achieved through a combination of one or more of three things: reducing video resolution, reducing frame rate, or increasing video compression, all of which risk causing a loss in quality.

Video compression remains an important aspect of reducing data, but it’s critical that none of the detail of video is lost in the process. There are a number of approaches that can combine to achieve the objective of a reduction in data, while maintaining the best possible quality, image resolution and forensic details.

One approach is to apply different levels of video compression to specific areas of interest in a camera’s field of view. In a wide view of a hospital corridor, for instance, the view of walls will be compressed to a higher degree than that of the corridor itself. The overall impact can be a significant reduction in video data produced.

Another method is to use algorithms within cameras to only send data relating to changes in a scene, with images from areas which remain static sent less often. This can be useful in recording quiet areas out of hours – an office lobby at night changes very little, but if someone does enter the full detail is recorded.

Technology can also be employed through which a camera captures and analyses video at full frame rate, but before transmitting the data unnecessary video frames are omitted from the stream. A static scene will be encoded with radically reduced frame rate, sometimes as low as one frame per second. When changes occur in the scene, the frame rate is automatically increased to capture every important detail.

Finally, an emerging technology is average bitrate control, which – if implemented correctly – allows the camera to automatically adjust its bitrate in relation to the available storage and video retention time. Specifically designed for continuous recording, is gives optimal control over the full retention time without simply applying bitrate limitations across all video capture.

These combined technologies significantly reduce storage and bandwidth demands – often by more than 50% – without losing the detail of any events taking place. They’re fundamental to enabling continuous recording.

A new world of surveillance insight

The benefits of continuous recording aren’t limited to ensuring that no event or incident is missed. Increasingly sophisticated analytics – both edge analytics in the camera and based on the server – will ensure that more value is derived from the video captured.

Video from existing cameras will benefit from server-based analytics, while newer, deep learning-enhanced cameras which supply extremely useful meta data alongside video itself, will come together to enhance security and operational efficiency.

A new layer of information will emerge between real-time alarms and alerts, and the forensic search of video in relation to known incidents. Analytics applied to continuously recorded video – recognizing patterns and anomalies – will start to highlight valuable ‘unknowns’. The insight from these will result in optimization of security operations and be applied to other aspects of organizational efficiency.

It is ironic that in the move from analog surveillance cameras to digital – and the enhancements in quality that this brought – many organizations felt the need to compromise on quality in order to manage the data implications. But increasingly this isn’t a compromise which needs to be made. The future takes us back to the days of continuous recording, without the drawbacks.

Future of Edge Analytics