Video Analytics – An Introduction to Its Types, Advantages And Disadvantages

shutterstock_163084580Ever wondered where those surveillance footages that are captured at train stations, airports, shops, petrol pumps and every other establishments that you see each day go? Also, whatever happens to those hours and hours of video surveillance footages later on? Do they ever get to be seen by someone or something? You betcha! Surveillance footages are an important tool in the hands of criminal investigators. In the event of a crime or a suspicion that a crime is about to be committed, law enforcement agencies will pursue every clue, including surveillance footages taken at various establishments, to establish the identity of the perpetrators. In the event there is a threat, whether or not that threat is genuine, analysis of video surveillance footages are conducted to establish, eliminate or discount the threat.

Human Analysis of Surveillance Footages

Before the days of computerized video footage analysis, humans had to do the bulk of the work. Hours and hours of video data had to be seen, any related information had to be assessed and then passed on to the next level for the purpose of establishment of a crime. The process was tedious and cumbersome. Not to mention, the inherent problems that came from the whole process being subject to human errors or oversights. In a research done to assess the concentration power of a human to go through video surveillance footages, it was discovered that, on an average it is impossible for someone to concentrate for more than 20 minutes on the screen. Additionally, live data feeds that came from multiple sources / cameras took additional effort to be monitored by anybody sitting in the front desk or a dedicated security area. Thus, surveillance teams took turns to do the job round the clock. The simple notion was, two pairs of eyes are better than one. Still, there was a need to eliminate human intervention for the processing of surveillance data. Machines were faster and they did a much more efficient job of assessing live data as well as scrutinizing historical data recorded over a period of time. The natural transition was to make the whole process computerized in order to make it faster and more efficient. Computerized video analytics was thus born.

Effectiveness of Computerized Video Analytics

The process of analyzing of surveillance footages, the gathering of intelligence from them and the establishment of link between a crime and the perpetrators is known by different other names. You may alternately hear ‘video analytics’ or ‘video content analysis’ or ‘computerized video analytics’ in the course of any research. They essentially all point at the same thing.

Humans are inefficient when it comes to doing repetitive and mundane jobs. Computers excel in such conditions. The benefits of having a scalable and workable surveillance system is beyond contestation. Innumerable times they have been found to be effective as a counter-terrorism and counter-criminal tool, silently detecting movements, recording events and producing evidence that makes it easier for the criminal investigation and justice systems to work effectively. It is the processing bit that was human-dependent and that slowed everything down and made things less efficient.

Video analytics have a much wider ramification though, and beyond the obvious applications in the areas of surveillance. Its other important uses are in the healthcare, transport, retail as well as home & commercial security sectors.

Hardware Requirements of a Video Analytics System

Unlike surveillance systems, which are dependent on the type of technology that is being used to record the actual videos, analog or I.P. based, video analytics (VA) systems, to some extent, are not dependent on the hardware. However, a mention is needed to make a difference between I.P. cameras and analog cameras as their quality does go a long way towards the effectiveness of the surveillance system and resultantly the quality of the video analytics process. Another thing is some types of cameras are unsuitable for particular VA systems.

In the case analog systems are used, the quality of the video feed tends to be of a lower standard than where I.P. based cameras are used. Analog cameras are at the best only of DVD quality. Although, it may sound state of the art, the fact is they are not. Try blowing up a DVD quality image to 200% or even 400% and see for yourself what happens. The image breaks up resulting in pixelated displays.

I.P. based cameras are much more powerful and they offer a lot of advantages to surveillance system engineers to set up an effective system for monitoring any premises. The biggest advantage is their megapixel resolution. Megapixel resolution allows images to be enhanced, blown for facial or number-plate recognition and in general has a much better quality compared to the images coming out of analog computers. I.P. based cameras are capable of being integrated to any existing Ethernet system allowing their faster deployment and more robust applicability. They can, alternatively, be used stand alone because they come with built-in memory card slot. You can plug in a SDHC memory card and record the video directly on the camera itself. The only downside of I.P. cameras seems to be the price and they can be prohibitive.

Architecture of a Typical Video Analytics System – Server Based Implementation

Server based implementation of video analytics systems is usually preferred in a situation where there is a hybrid system in operation. It is also suitable for an existing analog system that is trying to upgrade or use I.P. camera for some of its more critical locations. Basically it is a server based VA implementation where the software to analyze the video feed in installed on the server. The video feeds from various cameras across the network are channeled to a central server which receives the feed, processes them, analyses and then flags relevant sections as and when required.

Advantages and Disadvantages of Server Based VA Implementation

There are some inherent advantages and disadvantages to the Server Based VA Implementation. The first thing is this type of implementation will support older analog cameras, thereby increasing their effectiveness for surveillance requirements. Additionally, server based video analytics allows the business to implement different analytics software for different segments of the business. Let’s take an example. Say, a large I.T. company requires video analytics for its HR, security and Fire-safety departments. While the HR team may only be concerned with the presence / absence of the employee along with the billed hours, the security team would be concerned with whether an employee is doing anything illegal while inside the premises or there is any third party intrusion. A fire safety team, on the other hand, will be concerned with the safety of the premises and would be monitoring indications of fire on the premises. All of these can be done using separate video analytics software if the need be. This (using separate VA software) is impossible to achieve using an Edge VA systems. Additionally, using a server based solution allows for faster search of archived data compared to Edge based systems.

On the flip side, however, each of the cameras have to be wired to the server which receives the feed and analyses them. This means that at any given point of time there is a huge network overload depending on the number of cameras that are being used to record surveillance videos. Compression of the video for streaming over a network further degrades the quality. While feed coming from I.P. cameras may not be affected that much, analog camera feeds are affected quite a bit. Needless to say, the very reason for which these feeds are channelized to an analytics server is defeated, as performance of video analytics is hampered due to the lower quality of the compressed videos.

Server based implementation are usually not preferred as a scalable solution. Establishments which require many cameras such as hospitals, defense, large departmental retail stores would find it problematic to maintain more than a dozen or so cameras and then have the video feed analyzed automatically using the video analytics software. This is because servers have a finite capacity to handle, no matter the technology and or quality that is used.

Embedded Video Analytics Systems or Edge Based Systems – Advantages

The fundamental different between server based VA and edge based systems is that in the latter the analysis bit is done in-camera rather than on the server. An important consideration for edge based VA systems is thus that it uses I.P. cameras, as analog cameras are unsuitable for such tasks. It also means that only cameras which have the necessary computing resources will be able to handle the requirements of an implementation. The cameras shoot the video, and then based on the on-board software analyses them and only sends a gist of the surveillance feed to the central server. It usually sends details such as the speed of any moving objects in the surveillance zone, their vector of movement and thus follows a checklist of signs of warning as prescribed on the central server. Resultantly, there is no massive demand on the central server to do a continuous analysis of the video feed. The central video monitors may or may not be monitored on a round the clock basis as the data is being continuously processed on-location by the smart cameras.

In server based VA implementations the video feed must be first compressed to be transmitted over a network. At the server end the compressed video is then uncompressed. A considerable amount of data is lost as a result. In edge based implementations there is no compression and subsequent decompression, resulting in no loss of video quality. A edge based video analytics system can  be programmed at the camera level to only send in video feeds for archiving purposes when a motion is detected. This eliminates the unnecessary clogging of network and over flooding the storage of the archiving system.

Disadvantages of Edge Based Implementation

As discussed previously not all cameras can be used for an edge based implementation. This means systems where there are a mix of I.P. cameras and analog cameras this system will have only limited use. For only analog camera based systems this implementation is unsuitable. At the current state of technology, high end I.P. cameras costs very high and for any degree of computing abilities, one has to dole out sufficient cash. As such, if the establishment really has a very high need for security edge based system costs are prohibitively high, but well worth the investment. Want to know more about security surveillance cameras?