The National Fire Protection Association (NFPA) compiles statistics on all fires reported in the United States, and provides a statistical overview in the research section of the NFPA website (www.nfpa.org). In the 2008 report “Fire Loss in the United States 2007” written by Michael J Karter, Jr, there were 11,500 reported structure fires resulting in $779mm in loss for the combined category of industry, utility, and defence during 2007. In addition, storage occupancies, some of which are undoubtedly owned by utilities, manufacturing, and industrial facilities had 31,000 reported fires resulting in $670mm in loss. The loss values stated above do not include business interruption, which can be quite substantial. 

Looking further into the statistics in the report entitled US Structure Fires in Industrial and Manufacturing Properties by Jennifer Flynn, the average number of fires per year between 2000 and 2004 was 12,000 with a total average loss of $747mm per year. The nature of the work done in this occupancy type increases the probability of a fire when compared to more docile occupancies. 

In many cases flammable liquids, hazardous chemicals, dust, and/or other combustible materials occupy the same area as heat source such as a lamp, heater, or generator. Hot work is also very common in many areas and statistics show that shop tools are the number one cause of fires at 15% and heating equipment is the number two cause at 12%. The efficiency of fire detection in these environments is challenged by the sheer volume of the spaces. Smoke particles and heat may simply never make it to a smoke detector or sprinkler system, reducing their effectiveness. The early warning capabilities of VID systems allow for quicker detection while the fire is still at its incipient stage, followed by faster response with the situational awareness providing for better coordination of resources and public notification.  

VID cameras detect the presence of fire and smoke within the field of view at greater distances, covering larger areas, and providing faster detection times than conventional smoke detection methods. This is due to the fact that these cameras do not depend on the movement of smoke resulting in the physical presence of the combustion products at the location of the sensor. VID cameras can work in environments where spot detectors are inefficient or non-applicable, namely large volume structures which include warehouses, industrial facilities, utilities, and manufacturing facilities. VID cameras can also play multifunction roles in providing both fire protection, oversight of day to day operations, and general security for a facility.

Real-time data

As an input, VID cameras receive massive streams of real-time data.  For example, the axonX SigniFire IP camera captures images at a 640 x 480 pixel resolution that are processed at the rate of 16.7 frames per second. That amounts to over 5 million data readings (pixels) per second. Processing one pixel takes approximately 500 machine instructions so it will require processing power to fulfill 2.5 billion instructions per second. 

These numbers demonstrate the limitation of today’s state of the art processing devices, so in order to implement a VID system the processing needs to be done in stages where each stage reduces the amount of data that needs to be addressed. Techniques used in different systems may vary; the SigniFire system uses a three-stage process. At stage 1, proprietary cascading Digital Signal Processing (DSP) filters are applied to transform high frame rate monochrome video (colours are simply ignored) into low frame-rate colour coded images that provide very distinctive patterns for flame and smoke. At stage 2, specific patterns are extracted at the low frame rate and normalised. Stage 3, patterns are verified by a feed forward neural network to determine if the image contains smoke and/or fire.

For illustration purposes, let’s consider the effect of the patented DSP filter as if it is generating an artificially colour-coded image for an ongoing smoke event.  It changes the color of each pixel when pixels experience minor but consistent increases or decreases in its intensity against the background. The first time a pixel changes, it is considered excited and changes to blue. If pixel remains in the excited state, it undergoes an “aging” process, gradually turning the pixel color to RED. When pixels are no longer excited they quickly loose “colour” or simply blend within the background. This creates a pattern that takes into account the shape of a smoke plume, and it’s evolution over time that can be identified by pattern matching techniques (Fig 1, page 14). The image produced by this filter is used only internally.  At Stage 2, the camera software extracts and normalizes the patterns to a standard size to be processed by a neural network decision making engine. At Stage 3, the neural network decides if the real time pattern matches the criteria identified for smoke during training.  

Neural networks were developed by training the system to recognise smoke patterns within large number of sample videos (over 15,000) for a greater number of different scenarios and are pre-loaded in to the cameras firmware. Those scenarios include a great variety of actual smoke: black and white, small and large, fast and slow as well as a variety of nuisances such as light changes, moving objects, human activities, etc.  A similar approach is used in implementing the SigniFire flame detection algorithm with the difference being in the characteristics of the DSP filter and the samples the neural network was trained on.  

In cases where fire or smoke has been positively identified, the camera signals an alarm via built-in dry contacts and over the computer network. The second option provides a more sophisticated means of communication since it combines video transmission, alarm data including the cause of the alarm and coordinates of the source within the image and can be used to supply power to the camera via Power over Ethernet (PoE). 

Approval

The SigniFire IP Camera is FM approved as both a fire and smoke detector.  The camera has been tested to the ANSI/UL 268 smoke room test for smoke detectors and the FM 3260 test standard for optical flame detectors. Presently the approval process for VID systems requires adherence to one of the existing standards: UL268 or FM 3260, depending on capabilities. These standards were created around legacy detection technologies namely the spot smoke detector and optical flame detector.  

Because these standards and their metrics do not incorporate some of the advantages provided by VID systems, the standards are modified and adapted to suit VID systems.  For example in the UL268 test, as an independent reference for sensitivity of a particular detector, the standard relies on light obscuration measured at the vicinity of the spot detector.  This means that the standard is based on an assumption that all smoke detectors detect the smoke concentration at the location of the detector.  For VID systems, such a reference presents confusion, since at the time the VID system triggers an alarm, the smoke may not have developed and propagated to the optical density meter (ODM) and the camera itself to cause any obscuration value.  Therefore a testing procedure is modified so a time to detection from ignition is recorded and compared to spot detectors. 

The UL268 standard also limits the compartment size (36 feet by 22 ft by 10ft).  The compartment size limitation is overcome by testing to the FM 3260 standard that allows for longer distances. The point of all this is that VID systems present a significant shift in technology and VID system applications and implementation is significantly different from previous fire and smoke detection technologies. The move from analog state devices to intelligent IP video systems that combine the benefits of networking, digital video, and camera intelligence results in a far more effective means of security surveillance, fire detection, and monitoring than has ever been available before in the fire industry. The three applications and how these systems were implemented can be seen on page 16.

These case studies demonstrate a small range of the applications this technology has been applied to. There are many additional areas that can use the benefits of VID that did not have detection because of the limitations of conventional detectors. This technology fills a new segment of fire protection and offers benefits not seen before. Unlike previous detection technologies, the implementation of a VID system results in a positive cost benefit for the customer. As acceptance and knowledge of this technology grows so will the use and applications. Areas considered problematic for current detection methods due to size, ceiling height, aesthetics, functionality, or the necessity for security, early detection and situational awareness are now capable of being protected. 

Printer friendly version

Email this article to a friend

Bookmark this

---