Dual-Band Thermal Fire Detection Algorithm
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing automated fire detection systems using thermal imaging struggle to accurately determine the boundaries of both large and small fires, especially when they are obscured by smoke and particulates, leading to incomplete fire perimeter definition and missed isolated fires.
Innovation Solution
The system employs a dual-band thermal imaging approach, applying different thresholds to different regions based on the presence of isolated or large fires, using a threshold curve to distinguish fire pixels in the 3-5 um band and confirming fire status with the 8-12 um band, while eliminating false positives by analyzing adjacent pixels.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If the 3-5 um band is used for fire detection, then small isolated fires can be detected, but large intense fires produce scattered radiation that creates false positives and obscures true fire boundaries
Solution Approach 1:
The patent combines data from two thermal imaging bands (3-5 um and 8-12 um) to achieve both detection sensitivity and boundary accuracy. The 3-5 um band detects small fires while the 8-12 um band provides accurate fire perimeter definition, and the system merges these complementary data sources to resolve the contradiction between detection precision and measurement reliability.
2Reliability
If the 8-12 um band is used for fire detection, then accurate fire boundaries can be determined, but small isolated fires are lost due to lower sensitivity
Solution Approach 1:
The system merges the complementary strengths of both thermal bands by processing images from the 8-12 um band for boundary accuracy and the 3-5 um band for small fire detection, then integrating these results to achieve both reliable fire perimeter definition and sensitive small fire detection simultaneously.
Solution Approach 2:
The patent adds a temporal dimension by capturing and comparing images at different times, allowing the system to distinguish between persistent fire signatures and transient scattered radiation artifacts, thereby improving both boundary accuracy and small fire detection capability.
3Adaptability or versatility
If a single fire threshold is applied to all regions, then the system cannot simultaneously detect small fires and reject large warm bodies
Solution Approach 1:
The patent applies different fire thresholds to different spatial regions of the thermal image. By analyzing the spatial distribution of thermal signatures and applying region-specific thresholds, the system can detect small fires in cool regions while rejecting large warm bodies such as parking lots and rooftops, thereby achieving both adaptability and measurement precision.
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
This method enhances the detection of small fires while accurately defining fire boundaries, reducing the need for human interpretation and minimizing missed fires, allowing for more effective firefighting efforts.
Implementation Method 1
images an area containing fire, or which is suspected of containing fire, in one or more infrared bands, often in the 3-5 um and 8-12 um thermal infrared bands
Implementation Method 2
The 3-5 um band corresponds to peak spectral emissions from objects at temperatures of 900 K (roughly the temperature of smoldering combustion), making the 3-5 um band particularly useful for detecting the radiant energy emitted from fire
Implementation Method 3
longer wavelengths of radiant energy experience less scattering from smoke and particulates, and thus these bands—which have longer wavelengths than visible light—allow imaging of fires that are completely obscured by smoke
Data Source
AI summary
One or more thermal images containing pixel-by-pixel radiation intensity data in two or more wavelength bands can be processed to accurately determine the location of fire within the image(s), regardless of the radiation-scattering effects of smoke. First, all pixels having radiation intensities above a fire threshold in the longer-wavelength band are classified as being aflame. Second, a threshold curve defining a fire threshold in the shorter-wavelength band is applied to the pixels, and those having radiation intensities above the fire threshold in the shorter-wavelength band are classified as being aflame. Third, at least the second group of pixels above is tested to see if each pixel classified as being aflame is part of a group of adjoining pixels which were all classified as being aflame, and if so, each such pixel is reclassified as not being aflame unless it also falls within the first group of pixels above.