top
logo logo logo

Relevants Links

Chat Online Users

  • No users online

Login


Firesense The Project Technologies
Technologies PDF Print E-mail

 

FIRESENSE will use various sensing technologies and will develop novel algorithms and systems:


Sensors


CCTV cameras

Fixed CCTV cameras will be used for short range fire and smoke detection or for monitoring especially sensitive places (e.g. archaeological monuments) or places with high risk of fire. A historical building may have already a camera-based surveillance system. In such places, there is no need to install any new equipment. Only a PC (or PCs) with video capture cards will be placed to process the output of CCTV cameras for smoke and flame detection. Another important advantage of the regular cameras is that they deter people from "grave-digging", scratching the walls of ancient buildings, inscribing their names and giving other harms to the cultural site.

PTZ cameras

Whenever there is unusual motion activity or turbulent motion due to flame flicker or smoke motion on the scene, the Pan-Tilt-Zoom (PTZ) cameras will focus on the motion. The video captured by the fixed cameras and the PTZ cameras will be transmitted to a remote center by a wireless link. It will be possible to control the PTZ cameras from the remote center and transmit the camera images to other centers using the Internet. A video motion activity analysis software package will be developed during the project.

IR Cameras

Infrared cameras have the ability to capture rises in temperature and therefore they are suitable for an early warning fire detection system. Mainly uncooled Long-wavelength infrared (LWIR) cameras will be used, as they are cost effective compared with MWIR (Mid-wavelength infrared) cameras. It is hard to visualize smoke using LWIR cameras, because their thermal imaging region is 8–15 µm. However, they can obtain a completely passive picture of the outside world based on thermal emissions only and without requiring an external light or thermal source. Such cameras are especially suitable for monitoring an archaeological site covering a few hectares of land. For long range imaging, cooled LWIR cameras can be used, however, they very are expensive. The SME partner XENICS is an IR camera producer and cost-effective solutions will be assessed through its expertise.

Wireless sensors 

Expandable wireless sensors will be used to provide temperature information. These sensors can be constructed that would guarantee reliable fire detection by thresholding temperature values. These sensors will use a wireless master-slave telecommunication paradigm to transfer high temperature readings and humidity measurements to an information depot. In this way, a dense network of autonomous fire-sensing components can cover the terrain of an archaeological site, forming an appropriate grid of information.

PIR sensors

Passive infrared (PIR) sensors will be used for indoor fire detection, e.g., a museum near the archaeological site, or a monumental building. PIR sensor signals may provide valuable feedback for camera-based systems.

 

Local weather stations

Prototype meteorological stations will be installed near the archaeological sites. These stations will collect meteorological data (air temperature, relevant humidity, barometric pressure, wind velocity, direction and rain gauge) of these areas and will transmit them on-line to the sensor data processing centre.


Algorithms and Systems

Image and Video Processing Algorithms for Fire and Smoke Detection

The main research and technological objective is the application of new techniques for outdoor fire and smoke detection using both regular video and IR cameras. The final objective is the implementation of a real-time fire and smoke detection system, which is robust to false alarms, yet capable of detecting fire in its early stages. Specific algorithms will be developed for short and long distances. Methods such as motion characterization for turbulent behavior detection, color detection (in regular cameras), shape/geometry/contour characteristics extraction, temporal and spatial analysis will be considered. Furthermore, static and dynamic (flame tips flickering) characteristics of flame will be analysed to increase the overall reliability of the algorithm. Present vision-based fire detection algorithms developed lack an automatic verification stage. Only human observers provide feedback. On the other hand, wireless sensors can provide temperature and humidity information and this is an important feedback to verify the results of the camera-based system. A system with an active learning capability will be developed with the help of wireless temperature sensors.

Wireless Sensor Networks

Besides using video cameras and computer vision techniques to detect and track fires, wireless sensor networks (WSNs) will be utilized as well. As briefly described above, a WSN can be used in conjunction with a video camera system. Hence, one of the main objectives of FIRESENSE project is to develop a novel WSN, specifically designed for the protection of archaeological sites from wildfires. As mentioned earlier, a WSN is a collection of tiny sensor nodes networking together using wireless links to sense and monitor a region and report the sensed data to one or more data collection centers (base stations or sinks) where they can be stored, processed and analyzed. However, since the data collection centers may also be affected by the fire, reporting events in a reliable manner requires routing information packets to multiple sinks simultaneously. For that purpose, event based queuing and routing algorithms and novel energy efficient cross-layer mechanisms will be proposed.

Algorithms for Estimating and Visualizing Fire Propagation

Early intervention is a very important issue in wildfires. If the propagation of fire could be known beforehand, the deployment of fire fighters and the equipment could be more efficient. The estimation of fire propagation feature of the proposed system will build upon the system that was recently developed by BILKENT and ITI-CERTH. Terrain data (i.e. DTM data) of the test sites will also be integrated. Fuel models will be extended to be suitable for the modelling of the vegetation (fuel model) of FIRESENSE test sites and will be integrated with the current fire propagation estimation system. The system will be integrated with detection data from the sensors as well as weather data from the local meteorological stations. The estimated fire propagation data will be visualized on a 3D-GIS environment.


Data Fusion Techniques

The integration of the heterogeneous sensors that will be used in the proposed system requires the development of a novel and innovative data fusion mechanism. Data coming from sensors and meteorological information will be fused to achieve a robust site monitoring system. Existing multi-sensor data fusion methods such as Bayesian integration, fuzzy inference, Dempster-Shafer methods and rank-statistics will be considered for this purpose. Fast and robust detection of events under evidence from different modalities is a difficult problem, and Bayesian techniques require the estimation of the probability distributions for each of the modalities. When these are not available, or under presence of novel patterns that are not necessarily negative evidence for a particular model, rank-based methods have been shown to give robust decisions.

 

bottom