The development of advanced intelligent video analytics has been a positive step for those designing smart surveillance systems. The depth of choices available to many integrators and installers is immense, with a wide range of detection criteria covered. Some of the algorithms are specific, such as loitering detection, traffic flow monitoring, object left or removed, etc.. Others can be manipulated to fit a number of requirements. One which is often overlooked is line cross detection.
The growth of intelligent video analytics (IVA) in recent years has created a wide range of options for integrators and installers looking to deliver smart video surveillance solutions. As IVA becomes ever more flexible, so the potential for the technology grows. End users are looking at their system-based investments to identify return on investment, and the deployment of IVA can deliver this.
In terms of security, modern IVA algorithms offer high degrees of accuracy and reliability (when implemented correctly), and provide a degree of event filtering which allows the classification of events based upon risk factors or priorities. This enables surveillance systems to respond to alarms in differing ways, dependent upon site conditions or status.
Often, the core functionality of an analytics algorithm is created by the developer or manufacturer. Integrators and installers can then tweak the parameters to ensure the algorithm works as expected in any given location. These configurations can be the reason a specific algorithm is a success or failure. If the IVA engine cannot make sense of the data it is receiving, the performance will always be compromised. However, when correctly configured in terms of perspective, target sizes and trigger events, most IVA from credible manufacturers can deliver consistent and credible performance.
However, not all algorithms are the same. For integrators and installers, it is critical to first understand what is needed at the site in question. Once a clear operational requirement is established, it is then vital to ensure any considered IVA algorithms can meet the need.
For example, consider missing object IVA rules. Some algorithms will track objects in a detection zone which are then moved from that zone into another part of the scene, while others will look for objects which ‘disappear’. This latter definition might be that the object isn’t where it should be (even if it is in another part of the scene), while others will track when the object is no longer in the total scene (it has been concealed in a bag, for example). While all three can be classified as ‘object removed’ IVA, they all work in different ways.
There are a number of IVA algorithms which offer niche usage. These include (but are not limited to) loitering detection, wrong way traffic, overcrowding, queue management, heat mapping, etc.. There are also a number of algorithms which the integrator or installer can use in more creative ways. By adapting the flexibility inherent in the IVA, it is possible to create bespoke applications, simply via the configurations. One example of this is line cross detection.
It’s not uncommon for integrators and installers to consider line cross detection for perimeter applications, and it’s fair to say that’s the main reason the analytic rule was developed. In such deployments, it works well and can deliver fast and accurate results. Indeed, that the technology is used at many high risk sites and critical infrastructure applications speaks volumes for its suitability in such applications.
However, line cross detection can also be utilised for a wide range of other detection duties, including security-related tasks, process control and automated site or building management.
The versatility of line cross detection is only limited by the imagination of the integrator or installer. When the right algorithms are selected (in terms of configuration options), it is possible to build in a high level of event filtering, allowing multiple actions to be triggered dependent upon which detection scenario has been activated.
By selecting an appropriate line cross IVA algorithm, integrators and installers can create smart solutions which are efficient, credible, robust and cost-effective.
Line cross detection is predominantly used for perimeter protection applications. It is ideal for such use, and while the algorithms offer a good level of performance, correct implementation is important, especially where the perimeter does not include a physical demarcation.
In many IVA-based perimeter applications, it is considered best practice to create a sterile zone, ensuring that no one should enter the protected space. Therefore, any activity in the monitored area can be considered as an exception. This approach does allow sensitivity to be increased. In reality, creating a sterile zone can often increase costs due the need to install physical barriers such as fences. The approach is arguably less critical when working with line cross analytics instead of general motion detection rules.
The selection of an appropriate line cross analytic package for perimeter use hinges on a few considerations. First, often perimeters are not linear, or may not appear to be so when viewed from the location of a camera. Therefore, there must be an ability to create segmented lines.
Some IVA line cross offerings either limit lines to consisting of a single segment, or have a restriction on the number of segments which can be created. Even in a straight line, the need to move the line around obstructions or to deal with blind spots could result in a need for multiple segments. For example, on a single perimeter run, the need to allow for an inset gateway would require a minimum of five segments: three to create the box around the inset and two for the continuation of the perimeter line on either side.
Therefore, for perimeter applications to be successful, the IVA should allow the use of multiple segments. If it does not, or if the number of available segments is not sufficient for the site’s requirements, the IVA shouldn’t be considered for perimeter applications.
As a note, while some IVA line cross rules allow the inclusion of multiple lines, this isn’t always a workable alternative for segmented lines. In some cases the multiple lines each require a dedicated configuration and are treated as separate entities. With complex perimeters, the risk of errors in configurations increases, as each individual line requires its own configuration.
That said, the ability to create multiple lines can be benefit in many applications. For example, a second line could be used as a ‘double-knock’ option to prevent intrusion, especially if the physical perimeter isn’t substantial. Crossing the first alarm could generate a pre-alarm event, such as triggering a PA announcement, initiating a local audible or visual alarm, or sending a push notification to a local member of personnel. If this is ignored and the second line is crossed, a full alarm is triggered.
Alternatively, a second line could be used to filter alarms if, for example, someone crosses the perimeter to retrieve a ball or recapture a runaway dog, before returning – within a defined time window – to the public side of the perimeter.
It will be of little surprise that line cross IVA is a good tool when implementing perimeter protection, but integrators and installers should ensure lines can be segmented (and in sufficient numbers), and if additional services are required, that multiple lines with differing configurations are supported.
Line cross IVA can be used to monitor flow, whether that be traffic in and out of a site, building or specific area by people of vehicles. To best deliver workable benefits, IVA algorithms must be able to deploy multiple lines in a scene, with individual configurations supported for each line. The number of lines deployed will be based on the need.
With multiple lines, these can be used to differentiate actions, filter alerts or even clear alarms. For example, an initial line could be used to detect entry into a protected area. This can trigger a process such as real-time recording, sending a notification, alerting an operator, etc.. If there are multiple exits from the area, crossing one of many secondary lines will define the action to be taken. Using an accepted route might result in the event being logged but with no action taken. If the target attempts to enter a restricted area, an alarm could be raised or another security action initiated.
Equally, if the target leaves the area, a second line with a different directional filter might clear down the alert, freeing up the operator to tackle other tasks.
Where correctly implemented, line cross IVA can provide additional benefits for end users. Health and safety is a good example. If the IVA supports configuration where an alarm is generated if a percentage of the sensed object crosses the virtual line, this allows detection of people in dangerous situations such as those too close to a train track or leaning over a balcony in a retail centre.
Where IVA algorithms include accurate size parameters or can make use of object classification, this also allows vehicles entering a parking area to be ignored, but pedestrians generate an action.
Bosch: Intelligent Video Analytics
Intelligent Video Analytics allows the use of up to eight rules. Alongside line-crossing (including multiple lines in a logical row), the camera also supports object within detection field, loitering, condition change, person following route, camera tampering, object left, object removed, entering a defined area, exiting a defined area, crowd detection and people counting. It also includes standard VMD and metadata generation.
Because Intelligent Video Analytics automatically creates metadata, which can be recorded along with the video stream or be archived on its own, detection criteria can be added or tweaked, even after the event.
Configuration for Intelligent Video Analytics is carried out using the camera’s web page via a browser. The flexibility on offer from Intelligent Video Analytics is very good, and considering it’s a licence-free feature, it is certainly one of the better ‘on-board’ IVA engines you are likely to find in an edge device.
Axis: AXIS Fence Guard
AXIS Fence Guard is an advanced line crossing application which can be used with any Axis camera loaded with firmware version 7 or later. It supports up to 10 profiles, with one virtual fence of up to ten segments per profile.
AXIS Fence Guard creates virtual fences in a camera’s field of view, and triggers an alarm when it detects a moving object crossing the virtual line. It makes use of an intuitive user interface with real-time visual confirmation to deliver validation of correct configuration. The application integrates with the camera’s Action Rules management feature to automate actions when an alarm event occurs.
AXIS Fence Guard can support different profiles, which can have unique configurations. This allows specific rules and actions to be applied dependent upon a schedule, or to be manually switched.
It works well, and the filters allow the vast majority of nuisance activations to be filtered out. The application delivers triggers which can be used by the Action Rules function, and that gives wide range of options with regard to managing events.
Hanwha Techwin: Wisenet X
Hanwha Techwin’s Wisenet X camera range offers a number of video and audio analytics rules. The range also supports Hanwha’s open-platform approach, allowing the use of third-party applications to enhance functionality.
Integral analytics rules include virtual line crossing, zone entry and exit, directional detection, object appear and disappear, loitering detection, motion detection, tamper protection and face detection.
With regard to line crossing, up to eight lines can be supported but may only make use of two nodes and cannot be segmented. Each line can have different directional discriminations applied.
To ensure flexibility, each line can have different directional filtering applied, but the associated alarm actions for all lines in a specific scene must be the same. This makes sense if multiple lines are being used to construct a polygonal barrier or to follow a perimeter, but does rule out using multiple lines to create staged alarm events.
Given the relative simplicity of the integral line crossing IVA, it works as expected and in many mainstream systems will be enough for most needs.
Hikvision: Smart Feature Set
Hikvision’s Pro Series cameras include what the manufacturer refers to as a Smart Feature Set. This includes numerous intelligent video analytics algorithms: line crossing, intrusion detection and entry/exit from a defined region. It reduces potential nuisance activations by classifying objects as human or vehicles through the use of deep learning algorithms.
With line crossing, up to four virtual lines are supported, and each can support individual configurations. The use of four lines allows more complex scenarios to be created.
Directional discriminations can be applied, as can size criteria to help filter out nuisance alarms. Sensitivity of detection is adjustable, and this is based upon the percentage of the target object which must cross the line in order to generate an alarm event.
The ability to protect perimeters or set up staged alarms is limited, but for detecting line cross events in gateways, doorways or corridors it does what you’d expect from an integral licence-free analytic on a budget camera.