CCTV Test: Loitering Detection (Part 1)

The growth of video analytics has seen a diverse number of options made available to installers and integrators. Analytics can be embedded on cameras and encoders, added to open platform-based devices as an uploadable app, run on a centralised server or accessed via the cloud. The potential for deploying video analytics has increased as the rules on offer also become more diverse. One area that is challenging due to the nature of the action being detected is the identification of loitering.

READ PART 2 OF THE LOITERING DETECTION TEST WITH SOLUTIONS FROM XTRALIS, IPS INTELLIGENT VIDEO ANALYTICS, INTUVISION AND AGENT VI

The rise of video analytics can only be seen as a positive for many security applications. The technology – which delivers benefits to security and business intelligence systems – creates a proactive element which adds an ever-increasing number of diverse deployments for video surveillance.

Typically the effectiveness of any analytics rule will be dependent upon the complexity of the scenario it is trying to detect. Video analytics work best when the ‘violation’ is simple to define.

For example, applications such as line-crossing simply require the analytics engine to detect if an object passes over a defined line. Obviously events can be filtered with discriminations: object size and direction of travel can be added to reduce nuisance activations. However, the definition of the rule remains fairly straightforward.

The same can be said for object disappear rules. The analytics engine is made aware of an object within the video scene to be protected. If that object disappears, then a violation has occurred.

These two examples underline how some analytics rules can be very clearly defined. The clearer the definition of a violation, the more effective an analytics rule will be. This is because the ‘state’ that needs to be detected can be very easily defined. In effect, the analytics engine ‘understands’ what it is looking for.

There are many analytics rules which can be defined in very simple terms and it is therefore of no surprise that these are the more established options in the market. After all, if you had to code an analytics engine you would start with the simplest tasks too! Unfortunately life isn’t always that straightforward and in order to offer credible protection, analytics must also address some less well-defined situations.

When considering the performance of video analytics it is important to remember the purpose of the technology in any application. Analytics cannot and will not apply reasoning or understanding to a chain of events. Anyone who believes the video analytics will consistently make the correct decision will eventually be disappointed with the technology.

The role of analytics is to identify certain behaviours, thus making the job of the system operator – who can apply reason and understanding to a chain of events – simpler.

Detecting loitering is of great value for many applications. Often a precursor to a more significant event, making operators aware of someone loitering or flagging such activity in a recording can assist in the prevention and/or investigation of an incident.

The challenge for video analytics is how to effectively define loitering. Is someone who is waiting for a friend or family member loitering? How do you define people in a doorway sheltering from rain? What about someone just taking a few minutes from their busy day to watch the sun go down?

All of these activities will have similarities with a criminal watching an ATM and waiting for a vulnerable victim. Whilst an operator viewing live or recorded footage will be able to assess the situation, video analytics will not. Therefore, it is not reasonable to expect video analytics to differentiate between loitering with intent and innocuous loitering.

Deciding when and where to implement loitering-based video analytics is just as important to their success as is the selection of the right analytics engine. Deploying the technology in areas where people may be waiting, sheltering or queueing could lead to a high level of nuisance activations caused by everyday activity.

Equally it should be remembered that loitering per se is not an offence: loitering with intent, as defined under the Vagrancy Act 1824 was abolished when the Criminal Attempts Act was introduced in 1981. This changes the offence from loitering to attempting to commit a criminal act. If a site has public areas then care needs to be taken to ensure any system reflects the user’s customer service guidelines.

Because loitering is in itself an activity without clearly definable attributes, it is important to consider the discriminations that can be applied by a video analytics engine in relation to the needs of any given site. It is also important to carry out a thorough risk assessment and to ensure the end-user has clearly defined operational requirements. This will help to identify which said of analytics are best suited for their needs. It is vital to the success of any system that the customer’s expectations are realistic, more so than when a simpler-to-define analytics rule is deployed.

ANALYTICS OPTIMISED?
As video analytics becomes smarter, it is important to ensure that the system hardware does not become the weak link in the system. For those who want to deliver effective and robust analytics, read about the Secure Logiq HA Series of Analytics Optimised Servers.

Bosch: Dinion IP starlight 7000 HD

The Dinion IP starlight 7000 HD from Bosch Security is an HD1080p camera which features licence-free intelligent video analytics as a standard function. The analytics includes loitering detection along with a host of other roles: object detection, line crossing, enter/exit, route following, stationary or removed object, counting and occupancy, crowd density, condition change and flow monitoring. Rules allow discriminations based upon duration, size, speed, direction, colour and object classifications. The camera also offers tamper protection, audio detection and face recognition.

The Dinion IP starlight 7000 makes use of a 1/2.8 inch CMOS sensor to deliver H.264 and M-JPEG streams at frame rates of up to 50fps. The day/night camera offers enhanced lowlight performance, as you would guess from the ‘starlight’ designation. Sensitivity is quoted as 0.0069 lux for a 30IRE image. The camera also features high dynamic range of up to 120dB.

Multiple streaming is supported with configurable frame rates and bit rates. Regions of interest are also supported. Other video-based features include dynamic noise reduction, privacy masking, two-way audio streaming, edge recording via a microSD card and defogging. A range of tracking options is also included.

Installation of the Dinion camera is very straightforward. Power is PoE; 12V DC supplies can also be supported. The camera makes use of the IP Helper utility which is used to set the initial network configurations. This needs to be downloaded from the Bosch website; whilst you are doing this is also worth downloading the MPEG ActiveX plug-in as this will also be required if you wish to configure the camera via a browser.

The IP Helper immediately found our test camera and allowed the network configurations to be set. The process is very simple and effective. During initial login you will be prompted to set passwords for each of the three user levels. The next task is to install the plug-in which was mentioned earlier. Once that is done you can start the camera configurations.

Whilst the test is focused upon the loitering-based video analytics, it is worth noting that the camera’s performance is as you would expect from a high-end device. Image quality is very good with sharp and well detailed images, and colour fidelity is high. The set-up menus are intuitive and easy to follow and allow installers or integrators high degree of flexibility to optimise the image. Bosch upgraded its user interface a short time ago, so if you haven’t configured one of the devices for a while you might be surprised at how much better the process is nowadays.

The video analytics configuration menus are found in the Alarm section. Most of the settings are then made via the VCA screen. This is relatively straightforward. Along with the configurations there are also two images: one is a live preview and the other is a reference frame.

The menu itself is split into two sections. The first is for the VCA configuration and the second covers more generic elements. These are for tamper detection with a setting for sensitivity and trigger delay, global scene change with a setting for level, and brightness. The latter will detect if the viewed scene is too bright or too dark and each element can have an independent threshold set.

The first task is to set a general VCA configuration. There can be two configurable profiles for VCA which allow the installer or integrator to create bespoke options. There are also three pre-defined profiles: silent VCA, scheduled and event triggered.

Silent VCA allows the creation of metadata to assist in searching, but does not create alarms or events. This profile cannot be customised. The scheduled option allows a weekly schedule to be created for either or both of the configurable profiles, and the event triggered option allows profiles to be switched or disabled based upon inputs or virtual alarms.

In order to create an analytic rule to detect loitering, one of the profiles needs to be selected and configured. As a default, Profile #1 is set as ‘detect any object’ and is set the whole screen area. Whilst this is unlikely to be retained it does give a quick reference to how the Bosch IVA functions.

Each profile can contain multiple tasks, which allows additional filtering in some applications. The Settings screen has three tabs: tasks, metadata generation and metadata inspection. Calibration of the camera which is essential for optimum performance, is carried out via the meta data generation tab. This allows the mounting height, angle, roll and focal length to be entered, thus allowing the analytics engine to understand the viewed scene.

Once Loitering has been set as a task, the next choices to whether the rule is applied to the whole scene or to a specific field. The creation of new fields is simple and quick, and nodes are simply dragged to identify the boundaries. Each field can have an independent debounce time (the period an object must be inside the field before an alarm is generated) and intersection trigger. The latter defines how much of the target object must be inside the field before it is judged to be active.

Once this is done the trigger can be defined. This includes specifying how far an object needs to move (the radius is setting metres) and how long an object needs to remain in the field in seconds. Obviously with a loitering rule the object motion setting will probably be minimised. There is then an option to filter by object class. The options are for an upright person, a bike, a car or a truck. Object classifications are only used with 3D tracking. Filtering can also occur due to object conditions. These include minimum and maximum object size, direction, etc..

With the camera configured, the loitering section element works well. We found in our tests all rule violations were detected. Of course, there is no way of rationalising what constitutes innocuous loitering and intent to commit an offence. However, there is enough flexibility in the IVA configurations for a bespoke configuration to maximise accuracy for a given application.

The ability to specify a range of movement for the loitering radius does deliver flexibility with regards to eliminating innocuous patterns of behaviour that may occur in certain environments. Activity outside of the loitering radius is therefore ignored, and this is an important element when configuring the rule.

Hanwha Techwin: XND-6080RP

The XND-6080RP from Hanwha Techwin is a model from the company’s new Wisenet X range. The camera is a two megapixel model delivering HD1080p streams. The camera supports licence-free intelligent video analytics. These include loitering, tamper and defocusing detection, motion detection, line crossing, directional motion, enter/exit, appear/disappear, face detection, audio detection and auto-tracking. It also supports the use of alarm inputs and outputs.
Alarm actions include file upload, email notification, recording to edge storage, output triggering or displaying a digital PTZ preset.

The camera is a static dome model and utilises a 1/2.8 inch CMOS sensor to deliver H.265, H.264 and M-JPEG streams at up to 50 frames per second (M-JPEG streams have a maximum of 25 frames per second).

Sensitivity is quoted as 0.015 lux. The optical element is a motorised varifocal lens with a focal length of 2.8-12mm. There is a simple focus mechanism to assist with setup.

The day/night camera supports WiseStream II technology. This is Hanwha Techwin’s dynamic encoding engine which is designed to minimise bandwidth consumption. It does this by analysing image data to reduce the image quality where there is no important detail. The camera also includes the Hanwha Techwin smart codec, multiple streaming, two-way audio, the latest incarnation of the SSNR noise reduction technology, privacy masking, defogging and a host of other features. Power input is PoE; 12V DC inputs are also supported.

The camera is supplied with a paper quick start guide and a miniature CD which contains the full manual, a network setup guide and the IP Installer utility. The latter is used to set network configurations for the device. It runs from the CD and does not require installation. Once the configurations have been changed you will require the default password to complete the process.

On the first log into the device you are required to set a new password; Hanwha Techwin has implemented a secure password policy. If passwords are eight or nine characters, then a combination of upper case and lower case letters, numbers and special characters are required. If the password is between 10 and 15 characters, then only two are required. Experience has taught us to opt for a longer password and avoid the use of special characters as some ONVIF implementations don’t like the latter.

If you are setting the camera using a browser, Wisenet supports the use of Internet Explorer, Chrome or Firefox. If opting for Internet Explorer, then version 11 is required.

On initial login you may be somewhat surprised at the quality of the image and its low frame rate. Also, there will be no video showing in the various configuration windows. To resolve this go to the plug-in drop-down menu on the main screen and ensure that the viewer has loaded. This will resolve the various issues.

Whilst the test focuses very much upon loitering detection, it is worth noting that performance from the camera is very good. Images are sharp and detailed, colours are accurate and motion is smooth. Deploying H.265 compression with the WiseStream dynamic encoding function set at medium also reduces the bit rate requirement. Although our test camera was a two megapixel model, we would be confident that the five megapixel version will not be excessively bandwidth-heavy when correctly setup.

The menus are uncluttered and straightforward, giving the camera an intuitive feel. The various video and audio settings are all as expected, and during the setup process we didn’t encounter anything to give concern.

There is a dedicated menu section for Analytics, and this has subsections for motion detection, tampering detection, defocus detection, fog detection, face detection, IVA, audio detection and sound classification. It is the IVA menu which handles the general analytics. In order to configure these that tick-box to enable the functionality needs to be active. There are then four tabs which cover the various options.

The tabs for virtual line, virtual area, exclude area and common. The selection of the analytic type is dependent upon the setup screen being used. The virtual line screen handles line crossing, whilst the virtual area screen manages loitering, object appear/disappear, enter/exit and intrusion. To configure loitering detection the first step is to create the detection zone. This is done by clicking within the image to create nodes. These can be dragged to resize the zone.

When this is complete the interface will assign the detection area a number. Up to eight areas can be supported. Then you simply tick the box for loitering analytics, and the only loitering-specific discrimination that can be applied is a minimum duration time which is measured in seconds. The range is from 1 to 60 seconds.

Once this is done event actions can be selected. The choice is for FTP transfer, email, record or trigger an alarm output. The final task is to set whether the loitering analytic is always active or is scheduled.

With the loitering alarm configured, it is possible to tweak some general settings in the Common tab. These include overall sensitivity along with a minimum and maximum target size. These are set by simply resizing two colour-coded boxes. Alarm actions can be further tweaked in the Event menu.

It is fair to say that the loitering detection in the Wisenet X is effective but basic, and if you are looking for a more comprehensive analytics tool, then the best approach may be to use and Open Platform app, which this camera also supports. That said, so long as your expectations are realistic the loitering detection does work and for many mainstream applications will be sufficient.

During the test period all loitering incidents that fell within the configured parameters for the system were detected. By optimising the size of the detection zone, along with object size and loitering duration, it was possible to achieve a configuration which gave a good degree of confidence in the analytics.

FLIR: CB-5222-11

The CB-5222-11 from FLIR is an IP-based bullet camera delivering HD1080p streams. The camera utilises the analytics engine from Ioimage; the IVA provider was acquired by DVTel in 2010, which was in turn acquired by FLIR in 2015. The camera features integral licence-free analytics. Claimed by the manufacturer to be ‘military grade’, the detection types include loitering, intrusion, zone entry, virtual tripwire, fence tress pass, unattended baggage, object removal, stopped vehicle and tamper protection. The analytics engine also features automatic calibration.

Alarm actions can include notifications and video transmission over email and FTP, as well as triggering relay outputs.
The day/night camera utilises a 1/2.8 inch CMOS sensor and supports streams at 25 frames per second. Compression is H.264 or M-JPEG, and streams can use a bit rate of up to 8Mbps. Sensitivity is quoted as 0.2 lux at 15 frames per second. The camera is equipped with a motorised varifocal lens, with a focal length of 3-10.5mm.

Other features include 2D and 3D noise reduction, true and digital WDR up to 96dB, bidirectional audio, digital slow shutter and autofocus. Power input is PoE; 12V DC inputs are also supported. The camera is rated to IP66 for moisture and dust ingress, and IK10 for vandal resistance.

The camera is supplied with connector blocks for traditional power and alarm inputs and outputs, plus mounting screws and a hex head key. There is no documentation in either paper or digital format and no utilities. We did find a user and installation guide on the company’s website, and this does indicate that the product should include – or at least used to include – a quick start guide and a CD. The manual alludes to a utility (DNA IP – this stands for discovery network assistant) which we found on the company’s website after a search.

The physical installation of the camera is relatively straightforward. The unit includes a short fly lead with the various connections. These include an RJ45 socket for LAN and PoE, a modular plug for alarm inputs and outputs, audio in and out, a BNC output for installation and a 12V DC input.

The DNA IP utility took a couple of scans before it found our test camera. However, as the scans took around three seconds each this was no hardship. With the device found it is possible to change the IP address if required. The utility can also be used to upgrade the firmware or to change a number of admin settings.

With the network configurations set you can log into the camera. It takes a fair while for the ActiveX element to load. Indeed, the browser appeared to have crashed for a couple of minutes before the dialogue came up to load the element. The camera does not have a secure password policy, meaning it can be left on default. This is a negative point.

Whilst the test focuses very much upon loitering detection, the video performance of the CB-5222-11 is good. Detail is high and whilst it lacks the crispness of some cameras, there is nothing that would be considered a negative in terms of image quality. Running at the highest bit rate we did see an occasional dropped frame but nothing that adversely affected the credibility of the unit. Colour fidelity was good and motion was relatively smooth. The camera did display a little more latency than the other units on test.

The menus are fairly minimal which does in turn make them simple to follow. They are intuitive and any installer or integrator competent with video surveillance will not have any issues.

When the Analytics menu is expanded it reveals seven pages: depth, rules, responses, scheduled actions, on-screen display, firmware and backup and restore. Depth sets the cameras calibration and process can be carried out automatically or manually. Setting this manually involves recording a video clip during which somebody walks around various locations of interest in the camera view. Once this is completed markers are drawn to indicate the typical height of people, and guidelines are used to represent distances between objects.

The automatic calibration is a simpler affair. All you need to enter is the height of the camera (which must be a minimum of 4 metres), and you must also ensure that the ‘horizon’ is less than 30 per cent of the total field-of-view. Then it’s a single button click to initiate the process. During the automated calibration you need to have somebody walking along the vertical access of the field-of-view to ensure the results are correct.

It is worth considering the type of application when looking at this camera. Whilst many general IVA applications will include external areas including street scenes and large campuses, loitering detection will typically be deployed in smaller locations. These could include lobbies, doorways, ATM machines in vestibules, etc.. For some such applications a requirement of 4 metres as a minimum might not be suitable.

The Rules menu screen allows the creation of the various analytics. Once a detection zone is specified, then the ‘human loitering’ option is selected. This then gives two Attributes tabs: basic and advanced. These two tabs are where discriminations are added for the rule. The basic tab gives to options. These are for time in the zone, and an option to enable detection of small crawling or slow intruders. The latter is a standard option on the various rules and is not really relevant for loitering detection.

Switching to the advanced tab gives one further discrimination: size.

Once the rule has been created, alarm responses can be configured. There are three stages to this process. The first is to set the triggering event (in this case the loitering role). The second is to configure the action. Options include to activate a relay, clear alarms, arm camera, disarm camera, enable rule or disable rule. The final stage is to set a schedule for the response. There is also an option to set scheduled actions; these do not require a triggering event.

The on screen display menu can be used to select which information is shown. The alignment of messages can be configured, as can the displayed captions and colours.

The CB-5222-11 works well and does deliver a good degree of loitering detection. However it is true to say that the analytics engine is better suited to the protection of large open areas or busy and more complex scenes. For some basic scenarios the requirement for a minimum four meter mounting height with a horizon of less than 30 per cent might rule this camera out. In larger spaces the automatic calibration may appeal to some.

Panasonic: WV-SPW532L

The WV-SPW532L from Panasonic is an HD1080p integrated camera designed for external use. In its standard form it offers basic VMD, but can support i-VMD via extension software which does require an additional licence fee. This supports loitering detection, intrusion-based detection, directional detection and object-based analytics. Up to 8 detection areas can be included in a detection profile. The camera also supports face detection.

When using analytics there is an option to set image ‘depth’, ensuring that perspective in the viewed scene is accurate. Aside from this, there are no other filters specific to the loitering analytic.

The WV-SPW532L makes use of a 1/3 inch MOS image sensor to deliver HD1080p or 3MP (2048 × 1536 pixels) streams at frame rates of up to 25fps. The camera includes day night functionality and also offers enhanced lowlight performance. Sensitivity is quoted as 0.07 lux. The camera also features wide dynamic range.

Multiple streaming is supported, and the camera’s codec can support H.264 and M-JPEG. Other video-based features include privacy masking, two-way audio streaming, edge recording via an SD card, fog compensation, lens distortion correction and VIQS. The latter allows installer or integrator to designate areas (up to 8 are supported) which retain higher quality. Compression in other areas is increased to reduce bandwidth load.

Installation of the WV-SPW532L is straightforward. Power is PoE only, so there is a single connection to be made once the camera is physically mounted. The camera is supplied with a brief installation guide and a CD containing full manuals and a utility for installation. Unfortunately our CD was defective; the various manuals and the utility can be found on Panasonic’s website for download.

With the camera mounted and connected, the utility found the camera immediately and allow the network configurations to be adjusted. On initial login you are prompted to load the viewing plug-in. The result of this is a partial screen load followed by a fairly lengthy wait! Just as you suspect that the installation has gone wrong, a pop-up appears for verification that the plug-in should be loaded. Whilst the utility works well, loading the browser plug-in for configuration is very slow.

In that element of the process tries your patience, then you will experience a similar frustration when you first set up the loitering detection. The setup menus appear in a separate window which requires an additional plug-in. As with the initial viewer, the loading process is slow and there is no indication that anything is happening. If you retry loading the configurations screen you receive a warning that another user is currently programming the VMD. If you receive this message, you will need to wait a few minutes until it times out.

Whilst the test focuses on the loitering-based video analytics, video performance is as you would expect. Image quality is good with well detailed images, and colour fidelity is high.

The video analytics configuration menus are found in the Alarm section of the set-up menus. The first task is to select whether you will be using VMD or i-VMD. With the appropriate selection made, the configuration needs to be confirmed using the ‘set’ button. This will then change the tabs in the menu to suit the selection. To deploy loitering analytics you will need the i-VMD option.

The first step is to set up the detection area. There are a number of options as to how this is drawn. Detection areas can either be rectangular or polygonal; there is also an option to draw lines, but this is not used for loitering detection. However, each detection profile can have up to 8 detection areas, each of which can have a separate detection mode. This allows a virtual line to be used for crossing detection prior to somebody entering a loitering detection zone, for example.

Once the detection area has been created, loitering is selected from a drop-down menu. Once this is completed the perspective is set using the depth control. This simply requires the creation of two marker shapes designating foreground and background. When these are set the menu imposes a grid on the image so that correct orientation can be verified.

The final stage is to configure the camera action required when alarm event takes place. There are numerous options for notifications using either the Panasonic alarm protocol or HTTP notifications.

Given that i-VMD is an additional licensed software element, it is arguably somewhat light in terms of discriminations and the ability to link events. The detection functionalities are basic, and other devices in this test offer a greater degree of flexibility without the requirement of an additional licence fee.

Filtering out innocuous alarm events is made more difficult due to a lack of filtering based upon loitering dwell time.

Admittedly there are not many discriminations that can be applied to loitering detection, aside from size, duration of dwell and loiter radius. The WV-SPW532L only provides one of these, and as such somewhat limits the ability of the installer or integrator to maximise the camera’s efficiency.

Tyco Security Products: Illustra IPS02D21CWIT

The Illustra IPS02D21CWIT from Tyco Security Products is an HD1080p static dome camera. In its standard form it offers basic VMD, but the latest firmware upgrade adds a suite of video analytics to the camera. These include dwell and linger detection, along with object detection, abandoned/removed item, directional detection, enter/exit, crowd detection and queue monitoring. This is licence free for use of up to three analytics rules. To deploy more rules, an additional licence fee needs to be paid.

There is a subtle difference between dwell and linger: a target is classed as lingering if it remains in a region of interest for a specified time. It can move around within the area, so long as it remains within the defined region. Alternatively, the detection of dwell is based upon a predominantly static object been detected within the region of interest for a specified time. As persons loitering with intent are unlikely to remain totally static, the linger analytics rule is of greater interest.

The Illustra IPS02D21CWIT is an HD1080p static dome camera which delivers streams at up to 30fps. The day/night camera makes use of a 1/3.2 inch CMOS sensor to stream H.264 and M-JPEG format video. Sensitivity is quoted as 0.03 lux. The unit supports integral infrared illumination. Our test unit was fitted with an auto-focus 3.9mm lens.

Other video-based features include wide dynamic range and privacy zones. Edge recording is also supported, as is blur detection to prevent tampering. The camera supports PoE as well as traditional low-power inputs.

The camera is supplied with a printed quick start guide and a CD containing full manuals and installation utility. Our test unit had older firmware on it and so an upgrade was required to access the video intelligence functionality. This was a straightforward process and completed quickly.

The supplied utility works well and detect cameras on the network. We did note that if the camera is on a different subnet, the utility tends the default IPv6. This can be changed in the settings menu if required.

With the new firmware installed, initial login generates a screen which enforces a default password change. If configuring the camera via a browser, performance will be limited unless Apple QuickTime is loaded. To be honest, QuickTime is not the best video codec and we preferred to set up the camera with limited browser video display, performing the final tweaks once it was connected to a VMS.

General performance from the Illustra camera is good. Detail is crisp and clean, colour fidelity is high and motion is smooth. When compared with other streams the video does seem marginally muted, but this can be adjusted by tweaking the exposure settings. In all fairness is the kind of thing you’ll only pick up if you’re being pedantic!

Configuring analytics is a relatively simple task. The setup menus reside in the ‘events and actions’ section, under the analytics heading. This presents the installer or integrator with four tabbed menus: ROI, motion detection, video intelligence and blur detection. The ROI screen also includes settings for face detection.

With the video intelligence tab selected, a tick box needs to be used to activate the functionality. Once this is done rule can be created. This process is relatively simple and requires a name to be set for the rule, an action to be specified and the rule type to be selected from a drop-down menu. Dependent upon the rules selected, the remaining menus will change to present options for relevant discriminations.

When setting up a rule to detect lingering, other options include percentage of overlap and lingering time. The overlap function can be used to prevent false activations if lingering is detected close to the region of interest. The lingering time, in seconds, denotes how long a target needs to be in the region of interest before an alarm is generated. There is also a colour filter, although how much you use this may be in loitering detection is up for debate.

When creating the detection zone there are options to draw rectangles, polygons or to create freehand boundaries. Once all the elements have been completed the rule is simply saved.

Up to five event actions can be created, although only one can be applied to each rule. Options with regard to actions include triggering an output, recording video to an SD card, notification by email, FTP transfer, CIFS transfer or the triggering of an audio alert. Single or multiple actions can be included, giving a higher degree of flexibility with regard to how responses are configured when events occur.

The lingering detection analytics work well, and gave consistent performance. Where spurious activations were caused by innocuous activity, this was relatively easy to filter out by tweaking the configurations. Whilst the dwell analytics rule can be a touch hit-and-miss if used to detect loitering-type behaviour, the lingering function works well, even when a target is moving around in the detection zone.

BENCHMARK VERDICT

Bosch: Dinion IP starlight 7000 HD

The Dinion IP starlight 7000 HD makes use of the integral Intelligent Video Analytics functions which are included in all Bosch cameras and encoders. As has been proven in other tests, these work well if set up correctly. Loitering will always be a more operator-based algorithm, but given that the Bosch analytics worked well, and the flexibility with regard to alarm actions was also a plus point.

The general image quality was also good, and other functionality makes this a camera worthy of consideration for a wide range of applications. As such, it has to be recommended.

Hanwha Techwin: XND-6080RP

The XND-6080RP forms part of the new Wisenet X range, and on this showing that series will win a number of friends in the installation and integration communicities. While the camera is only delivering HD1080p streams, the image clarity is very sharp. The inclusion of video analytics is a benefit and the loitering rules work well. However, they are basic and as such any further refinement will need to be achieved with system design.

Set-up is quick and simple, and the general performance is what you’d expect for a mainstream camera. As a result, it is recommended.

FLIR: CB-5222-11

The CB-5222-11 was supplied without any documentation or utilities, and whilst this is increasingly common it was the only unit in the first part of the test which lacked both. The menus are basic and the implementation of IVA is simple. However, it terms of alarm actions it did fall behind the other units.

The unit did deliver loitering detection, but for some scenarios the camera might not be the right fit. It works better in larger and more complex applications. Because of this it is recommended, but with the proviso that the site in question is right for the implementation.

Panasonic: WV-SPW532L

The WV-SPW532L from Panasonic only offers additional intelligent video analytics as an extra carrying a licence fee. As such, you might expect that a paid-for IVA product would include a higher degree of discriminations and functionality, that actually the available options are a bit thin on the ground. This detracts from its appeal where more specific details make the difference between useful and superfluous activations, such as is the case with loitering.

Benchmark has tested Panasonic’s ‘externsion’ software before, and certainly some of the rules have a greater depth of configuration. However, for loitering detection, the software is noit recommended, based solely upon the other options available.

Tyco Security Products: Illustra IPS02D21CWIT

The Illustra IPS02D21CWIT from Tyco Security Products features linger and dwell detection: if you’re looking to detect typical loitering behaviour, then the option for liner works well. This ensures that a target trying to appear as if going about their everyday activities but actually remaining in the protected area is quickly detected.

The ability to enter multiple rules into one scene is something of a bonus, but these would need to be linked via a VMS. The ability to link these in the camera would be on our wishlist. The Illustra device is recommended for loitering detection.

READ PART 2 OF THE LOITERING DETECTION TEST WITH SOLUTIONS FROM XTRALIS, IPS INTELLIGENT VIDEO ANALYTICS, INTUVISION AND AGENT VI

BENCHMARK
Benchmark is the industry's only publication for installers and integrators which is dedicated to technological innovation and the design and implementation of smarter solutions. With an unrivalled level of experience in technology-based systems, Benchmark delivers independent and credible editorial content.

Related Posts