An ever-increasing proliferation of video cameras are being used to capture intrusion, prevent crimes, collect information, monitor processes and detect hazards. As the cost of these systems has decreased, the number of cameras is exponentially increasing.
According to the 2004 National Retail Security Survey (NRSS), 82,2% of the respondents use CCTV with 24,3% planning increased use.
The 2002 US Census retail classified establishments are just over four million. If 82,2% of the retail market uses CCTV surveillance, there are an estimated three million retail establishments using existing CCTV cameras. In addition to retail, the number of companies that currently fall under new legislation requiring new 'physical security' measures is astounding especially when you consider that most of these companies have multiple sites requiring large arrays of cameras.
The US government is instituting tighter controls and conformance legislations that require a higher level of physical security and penalties for those who fail to conform.
If the retail industry is any indication of the growing popularity of CCTV, then an 82,2% share of this sample would be well over two million of those establishments required by government to implement 'physical security' measures.
CCTV installations range in size from a small retail shop to downtown public areas of major cities, such as London, New York, Chicago or Washington DC.
The Security Equipment Marketplace Demand from 2001 to 2011 will more than double reaching 96 billion dollars (Freedonia Group).
In the latest Pro AV Magazine dealer survey, security sales and integrators responded that video surveillance components and systems account for 20% of their total revenues.
By 2011 with an estimate at 96 billion, that 20% translates into a US$19,2 billion in yearly market sales of video surveillance components and systems.
What is on the video?
Adding a camera is not expensive, staffing someone to watch it can be an ongoing sizable expense because watching video takes time and effort. The question is raised "who can successfully watch this ever-expanding flood of video?"
Considerable studies have focused on how effective human observation is. These studies have raised serious security issues in regards to boredom, distractions, multitasking, as well as what is known as 'change blindness' (ie looking but not seeing or losing connection with reality). The results of studies show that very often, human observation is considerably less than expected. Thus, increasing the number of cameras being watched has a converse effect on how likely a security officer will be to spot a threat on any given video.
In addition, if the video is not being watched effectively, then failure to intercede on security events is the likely result. This is a missed opportunity as the critical time for interceding has passed and whatever damage, if any, is done. After the fact, someone will need to search and review the uneventful footage in order to find the security events that went unnoticed.
This leads to the question of how to use technology to augment human observation and overcome limitations. By using computers to take on many of the mundane and time intensive tasks, security can become more effective. In the security industry, computerised vision has been a goal of a number of technologies.
The promise of a smart computer that can effectively watch videos and monitor their images has huge ramifications for trimming down the variable cost of security surveillance, as well as augmenting vigilance and providing cost efficient unmanned surveillance with remote capabilities.
In addition, the cost of storing and retrieving video surveillance events from 24/7 video, begs solutions on how to reduce the need to record hours of eventless unchanging video footage.
The primary technology for video monitoring is with intelligent video systems (IVS).
High-tech monitoring with intelligent video systems
VMD, the predecessor of intelligent video, is not new. Surprisingly, it has been around for sometime; that is the automated analysis of video to detect change in the scenes.
Although the justification for using intelligent video systems are strong, the known limitations and the shortfalls of the traditional intelligent video and VMD, create some trepidation for security managers who face leaner budgets and must meet greater security demands.
Devices range from rudimentary motion detection, such as those found on cellphone cameras, low-calibre detection such as IP cameras that offer non-complex preset size detection, single scenario specialised intelligent video detection, and state-of-the-art intelligent video for multi-task complex scenarios detection.
In general, there are devices that offer realtime counter-active alarm with event generation systems and then there are simple digital video recorders (DVRs) upgraded to offer some intelligent video technologies and forensic searching for after-the-fact investigation. Although DVRs have made limited progress towards the intelligent video market, their architecture requirements, specialty-niche mass analogue-recording, and alarm interface requirements (if available) are fast becoming antiquated. In this article, we will concentrate primarily on the realtime intelligent video with alarm interfacing.
The following is a list of some of the benefits of IVS intrusion and threat detection has over other security devices that detect movements:
* Provide instant alarm verification, displaying a video picture to security personnel (Allows for remote analysis of the scene for gauging response and scene safety).
* Wide-area detection, not restricted to straight line or short distance detection, such as infrared beams used in PIR or motion detectors.
* Provides some form of definable detection area where the user can define different areas of the scene.
* Can often be retrofitted to existing cameras.
* Provide a level of security-personnel accountability for surveillance.
* Can be used to record evidence (latest technology offers pre-buffer of video so recording begins before the incident.)
* Threats that move around a site can be followed autonomously by cameras in order to prevent risk to responders created when the intruder's location is no longer known. (PTZ tracking).
* Offer some level of differentiation and qualification of threats (size, speed, direction, distance etc.)
Despite these benefits, the shortfalls of intelligent video technologies made its use too complicated and annoying to use. The justification for its use was often torpedoed by high false-alarm rates, less than acceptable probability-of-detection (POD), unwillingness of personnel to learn the science and costly ongoing configurations that produced less than satisfactory results in real world indoor and outdoor applications.
Intelligent video system technologies and concepts
IVS technologies range in their complexity. Products on the market may offer low-end detection using a single concept or may combine several concepts together to provide greater detection capabilities.
Very old systems actually attached a detection device to the glass of a monitor screen and watched the rendering on the screen. Later technologies analysed digitised pixels using software.
All modern IVS use the pixels of video images sampled from video at frame intervals (pixels are nothing more that plotted colour points or greyscale points that are used to paint the images on a computer screen.)
On many systems, these digitised images are converted into a standard for multimedia known as MPEG-4 using a variety of picture-sampling and compression techniques, such as DCT (discrete cosine transform) compression. Lossy compression algorithms, such as this, have a tradeoff as they lower image quality in order to achieve faster transmission and reduced processing requirements.
Many IVS technologies use the post MPEG-4 images for detection, this allows faster processing because of the reduced image information, however its reduced quality lowers detection capabilities, distorts the image and may lose detection in hard-to-see scenarios, such as shadows or at dusk. On the up side, the MPEG-4 format reduces cabling requirements as it can be transmitted over an IP network (eg, wireless networks) instead of more expensive analogue wiring.
Better technologies use the images before they are compressed into MPEG-4. This means higher quality images and better sampling for detection. It also means a greater memory usage provided the IVS technology does not cut down on the sampling rate or percentage of pixels used. The use of analogue non-MPEG-4 compressed images often translates into greater cabling requirements unless the processing is done local to the cameras prior to MPEG-4 compression.
IVS technologies that choose to ignore frames by sampling less often, decrease the memory requirements, however they require a greater time (sampling) needed to detect movements, as the frames are further apart in time. This is detrimental for detection in high-speed scenarios where the amount of time an object is in the field of view is extremely brief.
Only a few systems, which are considered top-end, sample pre-compression images using all pixels and at a high frame rate. The following are some concepts that may be used in commercially available IVS products.
* Localised pixel change detection: the concept is very simple, on the surface. The digital images are displayed in pixels. These pixels are converted into greyscale and then regions of the pixels are checked for pixel changes that exceed a given greyscale threshold. On most systems, these thresholds are usually static presets; more advanced systems allow adjustments and only the most advanced technologies provide some form of automatic threshold adjustments.
* Global pixel distribution: primitive in concept and limited in utility, a global count (entire picture) of greyscale is used to calculate distribution ratios and to watch for significant changes in distribution within greyscale totals. If a sizable shift is detected in the totals, an alarm is triggered, regardless of the cause, whether the image change is local or global shifts in lighting.
* Motion patterns: detects if the changes in the image pixel's greyscale provide a detectable pattern of greyscale-flow from image to image.
* Object dimensioning: more advanced than localised image change, seeks to identify and classify the shape and area of greyscale change in pixels from image to image. Using dimension measurements, systems can eliminate false alarms if they can properly classify the object and how to handle it.
* Behaviour: innovative intelligent-video detection that learns through analysis of greyscale changes image to image. Couples numerous concepts to detect anomalies and deviance in objects of a scene while providing leeway for background activity, supporting directional restrictions, and overcoming nuisances. Uses different object identification and movement behaviours to construct complex logic and condition-based sequences for qualifying detection of behaviour scenarios.
There are various types of intelligent video in the IVS arena. The spectrum goes from simple devices that register any movement whatsoever, such as those built into low-end digital cameras, to advanced systems that differentiate objects and behaviours while compensating for false alarms and spurious activity.
Only a few offer flexibility for complex real-world detection scenarios. Most are limited to specialised detection with one-size-fits-all configurations. For example, some may offer one type of detection at a time. An example would be offering unattended baggage detection but not detection of illegally parked cars. In addition, the detectable scenarios may not handle the numerous possibilities that a security guard would look for, eg, baggage thrown from a moving vehicle as opposed to baggage carried into a scene, or dropped from a balcony above.
Single detection scenario approaches or one-at-a-time detection devices tend to leave loopholes for real-world security.
Complex challenges for IVS
Whether indoors or outdoors, there are a large number of factors that can cause changes in image pixels or trigger false alarms. The ability to sort out the threats and ignore non-threatening items is something human vision takes for granted and one of the greatest challenges for consistent and accurate visual motion detection.
Commercially available IVS technologies fail on one or more obstacles. Additionally, like a dog chasing its tail, making correction for one obstacle elevates another or opens security holes that cause detection of real threats and intruders to fail.
The corrections and fine-tuning for virtually all of these systems require that the operators (security guards) learn and know the advanced concepts as well as understand difficult setup and configuration procedures that are ongoing daily activities. This makes the technologies difficult to deploy and maintain, requiring a level of technological capabilities from security guards and increase the training costs of new employees.
Worse, some leading systems on the market today have become so complex that their stable operation requires only an expert from the company can make changes in a system where a single typo can bring the system to its knees.
Overconfidence in a system that is not properly adjusted can cause a situation where security personnel may miss intruders. Conversely, if the system is over sensitive the security personnel may spend too much time dealing with false alarms or may become too casual in dismissing alarms.
Because real-world objects are three-dimensional, matching an object to a classification presents challenges, as the size and shape of an object facing head on is not the same as a side or angled view. Because of this, some IVS have strict requirements on how a camera can be mounted for perspective and are reliant on subjects to behave and dress in an expected manner (eg no head shrouds, no crawling, no piggyback, no camouflage etc.).
Need for proven effective technologies
Within the video security surveillance market, CCTV and security surveillance has been steadily increasing. Since 1997, their sales surpassed burglar and fire alarm systems.
With the video-security-surveillance market so ripe, new products are racing to market to try to capture a piece of the pie.
A Pinkerton study of Top Security Challenges for fortune 1000 companies, found that respondents ranked second the 'Need to justify/show effectiveness of programme'.
This is in part due to an ever-increasing number of untried untested products. Security administrators are finding themselves in an awkward position of vouching for the worthiness of security expenditures and this responsibility requires them to weed out products that do not have track records of success.
Chicago city now touts the largest video surveillance system of its kind in the world with a fibre-optic grid almost 1600 km long with cameras and biochemical sensors. Not stopping there, at the beginning of 2006 the mayor of Chicago proposed that all businesses open more than 12 hours a day, install indoor and outdoor security cameras. Any businesses that agreed to pay an undisclosed fee could have cameras monitored by the 911 centre. (Chicago Sun Times, 31 January, 2006)
Kissimmee City Commission approved the installation of surveillance cameras. It is hoped the cameras will reduce crime and alleviate the negative high-crime perception in certain areas; authorities will install surveillance cameras in North Kissimmee. From several blocks away, via monitors in police cars, officers will be able to spot and respond to any crimes viewed in the video feeds. (Osceola News-Gazette, 1 June, 2006)
© Technews Publishing (Pty) Ltd | All Rights Reserved