Video analytics on the edge

Kevin Waterhouse, Executive Vice President of Sales of VCA Technology, predicts that demand for edge based Video Analytics will continue to grow as a result of end-users increasingly wishing to take advantage of the superb quality of the images captured by high definition cameras.

It was not that long ago that government and businesses were resigned to paying telecom companies large annual sums for dedicated leased lines. Typically these were analogue fibres for transmitting video back to a central control room from remote cameras, or sometimes they were private digital circuits for linking control rooms and remote locations together. Needless to say the annual revenue cost of leasing these private lines has been extremely high, if not prohibitive.

Then along came broadband and the opportunity to view live images or retrieve previously recorded video over the Internet became a possibility for almost any business, and it delivered the added significant benefit that any authorised member of staff could view the video from any PC, and more recently, even from a  tablet or smartphone from anywhere in the world.

The broadband challenge

Broadband has opened up the opportunity for IP network based video surveillance solutions and this has included the option to take advantage of the amazing quality images which can be captured by HD cameras. We have all got used to seeing the quality and sharpness of the images which can be delivered into our homes via high definition TVs and it is not surprising therefore that users are increasingly looking to deploy the same technology to capture evidence grade images from their video surveillance systems. This presents both a network bandwidth and storage challenge as high definition cameras require large amounts of both. The arithmetic is simple, i.e. a 1080p stream at full-frame rate captured over 31 days by a single 2 megapixel camera will require about 1,000 GB of network bandwidth and storage, and so for a site with a large number of cameras it is unlikely to be realistic or affordable to monitor or record all the cameras in real time.

The solution for most video surveillance projects is to compromise either on frame rates or resolution or both. It is therefore not unusual for a system to be configured whereby high resolution images are recorded in real time and stored for just say two or three days, with a separate video stream of low resolution or lower frame rate images stored for longer periods.

The Video Analytics Option

Video Analytics can provide a cost effective method of reducing the amount of video that has to be transmitted when there is an alarm event. 

When first introduced, Video Analytics was only available as a centrally based server solution. As it has to work with raw uncompressed video, there was a need for large amounts of network bandwidth in order to be able to transmit images to a remote control room where it could be analysed. Recent improvements however in the design and processing power of Digital Signal Processor (DSP) chipsets has made it possible for Video Analytics software to operate highly effectively at the edge. By this we mean the software is incorporated into individual cameras, (or into encoders connected to existing analogue cameras), instead of having to be located on a central PC or server. This offers a major benefit to IP network based video surveillance systems in that HD cameras can be deployed without excessive network bandwidth utilisation.

It works on the basis that users can take advantage of easy to use on-screen tools to define what events are important to them.  Detection lines and zones can be configured to ensure that only specific activity is recorded. The accuracy of the software is enhanced by the fact that direction of movement can be defined so that alerts are only triggered when individuals or objects move in a particular direction across a zone, or when they move into a zone but not when they leave it. Filters provide the software with the ability to only detect adult people of a specified size and ignore (false alarms) from small animals and children. Operating effectively with both indoor and outdoor cameras Video Analytics can be used for a wide range of applications including intrusion detection, perimeter protection, vehicle monitoring, abandoned object detection and ‘dwell’ detection, i.e. alerts are generated if one of more people remain in an area for more than a defined period of time. It can also detect camera tampering and failure.

The key to success is to specify video analytics software which has a proven track record of working reliably in a wide range of environments. A high detection rate and a low false alarm rate is essential in order for the software to have credibility.

The ability of Video Analytics to detect specific events means that users can save on both storage and bandwidth requirements. When a pre-defined event takes place, all of the relevant video can stored on the camera’s SD card, whilst the event data and a small video clip can be instantly transmitted to a remote location where the operator can take decide if they wish to retrieve all of the video relating to the event and/or view live images from the site. The viability of recoding and storing video on the edge has been enhanced by the fact that 64Gbyte SD cards are now readily available and 128GB & 256GB SDs are on the horizon. Video Analytics help ensure that the maximum use is made of this storage capacity available storage space.

With the technology now available it would therefore seem to make sense to do as much compression, storage and analytics as possible at the edge before transmitting video, (which could include pre-alarm images), linked to specific alerts through to a central monitoring station, or perhaps direct to a patrolling security officer or a small business owner where alerts may require urgent action.  The same principles apply when Video Analytics is used for other applications such as the detection of fire/smoke, abandoned objects, tailgating or for object or people counting.

Edge based Video Analytics and Cloud storage offer a winning combination 

The decision on where to transmit video to and where to store it will very much depend on the size and type of surveillance system and what it is intended to be used for, as well as the costs linked to bandwidth usage, installation and the overall total cost of ownership, including the skill set of the people required to operate and maintain the systems.

If users decide on a remote ‘centralised’ video storage solution, there is now the added option of Cloud based services offered by server warehouses. The charges for utilising these services are more than likely to be justified by not having to incur the cost of purchasing, housing, managing and maintaining in-house servers, as well as providing back-up should there be a major failure.  As more service providers offering cloud based services enter the market, the resulting competition will inevitably push charges in a downward direction. It may also not be too long before we see those ‘cloud’ based service providers expanding their portfolio of services to include an alternative to what is provided by traditional remote video monitoring and alarm handling centres. Video Analytics will certainly make it an efficient way to do so. However, without the utilisation of edge based Video Analytics, a Cloud based service (or VSaaS as it is often referred to), because of bandwidth restrictions, is unlikely to be suitable for video surveillance systems which have more than four or five cameras. On the other hand, with proper utilisation of Video Analytics, the number of on-site cameras becomes almost irrelevant.

The business case for Video Analytics

Video Analytics has established its credibility as a reliable and robust method of detecting predefined behaviour or events and is increasingly being deployed as an integral component of video surveillance systems. Retail chains in Denmark, France, Finland, Hungary and Spain, an Italian car factory, a prison in China, a University in Estonia, a London sports stadium and hospitals in Japan, are just a few examples of the environments where Video Analytics is being deployed.

The business advantage which cameras equipped with Video Analytics can provide, is compelling. In addition to offering security personnel a very powerful tool to detect and combat theft and other criminal activity, Video Analytics is capable of multi-tasking and so, depending on the field of view, a camera equipped with the technology should be able to simultaneously generate reports with supporting video to provide merchandising, human resources, security and other operational departments, with the information they need to achieve their objectives. Every retail store will, for example, have its own set of challenges, but common to all is likely to be a desire to fully understand why a particular store is performing better than others and if specific marketing activities or promotions have been effective. Video Analytics is able to make a significant and profitable contribution to finding out the answer to these crucial questions.

The good news is that more and more video management software (VMS) developers and network camera manufacturers are recognising that Video Analytics is the next big opportunity for installers and system integrators to win new business, and most importantly deliver added value to users who are looking to maximise their return on investment in a video surveillance system.