Video Content Analysis: how far along are we now?
Alert operators.
An extra pair of eyes.
Fewer monitors in your control room.
It is possible if you have expanded your camera system with Video Content Analysis (VCA).
VCA analyses your camera images (in real-time), recognises certain situations and, if necessary, generates alarms.
In the event of an alarm, the operator is offered specific images showing the deviation.
If you combine VCA with other forms of detection, even more reliable detection is possible, so that hardly any false alarms occur. Video Content Analysis is no longer new.
The first serious systems emerged about 10 to 15 years ago.
In the meantime, the technology has improved considerably, especially in terms of reliability.
Developments in camera technology have also played an important role in this.
VCA detects moving objects and then determines the size of an object and its length/width ratio.
Based on this information, it can be determined to which ‘object class’ (e.g. human, animal, vehicle) the object belongs.
VCA also determines the direction in which and the speed at which the object moves.
The color of the object can also be determined.
If certain criteria are met, the system generates an alarm.
For example, a virtual line or a surface can be ‘made’ directly on the inside of the fence in the outdoor area.
Cameras with VCA can then detect all objects (or only the objects in a certain object class) that cross that line or come within the plane.
But it is also possible to determine that a machine stops because there is no more movement, or that a car stops in a place where it is not allowed.
VCA is also used for crowd management: counting the number of people in a certain area, on the basis of which certain decisions can be made.
This is already happening – albeit in test setups – at the Nijmegen Four Days Marches and during the Olympic Games.
VCA can also be used to detect manipulation, such as rotating or covering a camera.
After all, any deviation from the set standard is detected.
Most VCA systems can detect the following:
- Abandoned Object, Removed Object
- Defined Area Entered
- Climbing over
- Hanging out, stopping for too long
- Crossing barriers/lines, tailgating (trying to enter with someone)
- Counting objects (people, cars) for marketing purposes or personnel planning
- Sabotage (twisting or partially covering a camera)
- Fire and smoke
VCA improves control room efficiency
In general, people can only concentrate well for a short period of time.
This also applies to operators in a control room.
It has been shown that after a short period of time, they overlook things when many camera images have to be observed at the same time.
Even a monkey walking through the image is no longer noticed.
With VCA, an alarm is generated if certain criteria are met.
In many cases, this is an alarm notification including a camera image on the screen of the operators in the control room.
It goes without saying that a control room can look out many cameras very efficiently in this way.
After all, if there is a suspicious situation, they are actively warned.
In this way, hundreds of cameras can be optimally monitored with only a limited number of operators.
And because the addition of VCA detects more accurately, less unnecessary video is recorded and less storage is required.
This can result in significant cost savings.
Search for images based on Metadata
Many VCA systems generate metadata.
These describe the content of the analyzed image and can be stored together with the video images.
Metadata stores, for example, the color, size, direction, speed, and location of a moving object.
Later, this metadata can be used to investigate an incident, for example.
Metadata allows video to be searched for a particular event or information in a matter of minutes, days.
This is in contrast to standard video systems where no metadata is stored.
Here, it is usually only possible to search by movement.
If you search for specific events, you will spend hours or even days looking at images on these systems.
In addition, there is a chance that you will overlook a lot.
Restrictions
No matter how interesting (and affordable) the solution is, VCA also has limitations.
If you set the VCA too accurately, there will be too many false alarms and in the opposite case, you can miss situations.
However, it takes a relatively long time to set up the system properly and adapt it to local conditions.
After the initial installation, regular changes should be made to optimize the system.
The VCA systems that Mactwin implements are therefore connected to the Mactwin Operation Center as standard, so that the setup/adjustment can be done remotely from the MOC as much as possible and the costs for this are limited.
Increasing detection reliability
By combining Video Content Analysis with other forms of detection (e.g. VCA combined with radar detection), a very reliable system is created.
Thermal cameras, which show heat differences, can be used to detect objects under all conditions.
Even in pitch darkness, fog or with very bad weather.
In the images from thermal cameras, people in bushes can be clearly seen, as well as someone dressed in black against a dark background.
A combination of VCA with data from other systems, such as that of an ATM, is also very effective.
In that case, for example, it can be detected that someone is standing still at the machine for a long time without a debit card transaction being initiated within a certain time.
In some cases, this may indicate a suspicious situation that calls for an alarm.
VCA is not motion detection
Sometimes video content analysis and motion detection are confused with each other.
However, they are two different techniques.
Whereas motion detection can only detect that some of the pixels in the image are changing, VCA is much more intelligent and can actually observe what is happening in the image for a number of situations.
In conclusion, we can say that VCA offers many possibilities.
If you are considering using VCA, you must answer the following questions in advance in order to make the right choice:
- What are the objectives?
- Is it an indoor or outdoor situation?
- What are the possible interfering factors?
- What are the lighting conditions like?
- What is the perspective?
- What is the distance from the camera to objects to be observed?