A method and device for detecting and alarming a large-scale fire in a wooden structure building group

Abstract

The application discloses a kind of wood structure building group wide range fire detection alarm method and device, specifically related to fire detection alarm technical field, for solving the problem of directional error when fire hazard grade evaluation in multi-band spectrum fire detection;By obtaining the multi-band spectrum data covering ultraviolet band to far infrared band and extracting spectral feature parameters, it is judged whether the symmetry of full-band spectrum energy distribution breaks down to identify optical path interference, and the distortion offset is stripped when there is optical path interference and a full-band spectrum extinction vector is constructed, then it is analyzed whether the asymmetric direction of the full-band spectrum extinction vector within the preset time window is reversed to determine the time-varying evolution mode, the pre-stored characteristic extinction vector is selected according to the time-varying evolution mode to compare and identify the type of burning material, and the remaining spectral feature parameters after stripping the distortion offset output the fire identification result containing the fire hazard level.

Description

Technical Field

[0001] This invention relates to the field of fire detection and alarm technology, and in particular to a method and device for large-scale fire detection and alarm in wooden building complexes. Background Technology

[0002] In fire monitoring of contiguous historical wooden building complexes, early warning is crucial for minimizing losses. These complexes are characterized by their vast area, narrow and winding internal streets and alleys, and numerous semi-enclosed spaces such as eaves and overhangs constructed of wood and brick, creating a complex fire detection environment. To protect the architectural style, non-contact optical radiation detection methods are typically employed. Multi-band sensors are deployed to collect spectral data ranging from ultraviolet to far-infrared. After data preprocessing, spectral characteristic parameters related to combustion are extracted, and fire identification results are output based on parameter variation patterns, thus providing early warning when anomalies are detected.

[0003] Current fire identification technologies primarily rely on extracting characteristic indicators from the spectral energy distribution obtained by sensor receivers. However, in the narrow alleyways and eaves spaces unique to wooden building complexes, the smoke generated at the onset of a fire often contains a large number of particles from building repair materials or wood pyrolysis. As these particles slowly dissipate from the building surface, they cause non-uniform absorption and scattering of light radiation passing through them. Because the attenuation is inconsistent across different wavelengths, the spectral energy distribution received by the sensor will be distorted. For existing methods that rely on spectral characteristic parameters such as multi-band energy ratios for fire classification, this interference during optical path transmission is equivalent to adding an unidentified offset to the true fire characteristics, leading to directional errors in the assessment of fire hazard levels. This can easily result in rapidly developing high-risk fires being mistakenly classified as low-risk, delaying the confirmation of critical fires. Summary of the Invention

[0004] This invention addresses the technical problems existing in the prior art by providing a method and device for large-scale fire detection and alarm in wooden building complexes.

[0005] The technical solution of the present invention to solve the above-mentioned technical problems is as follows: A method for large-scale fire detection and alarm in a wooden building complex, comprising: S1: Acquire multi-band spectral data within the monitoring area synchronously collected by distributed multi-band detection units, with the multi-band spectral data covering the ultraviolet to far-infrared bands; S2: Preprocess multi-band spectral data and extract spectral feature parameters; S3: Determine whether the symmetry of the full-band spectral energy distribution characterized by the spectral characteristic parameters is broken. If it is broken, it is identified as optical path interference. S4: When there is optical path interference, the distortion offset that characterizes non-uniform attenuation distortion is extracted from the spectral characteristic parameters, and a full-band spectral extinction vector is constructed based on the distortion offset. S5: Extract the spectral attenuation distribution contour of the full-band spectral extinction vector at the start and end of the preset time window, determine whether the asymmetric direction of the attenuation on the short-wavelength side and the attenuation on the long-wavelength side has reversed within the preset time window, determine the time-varying evolution mode of the full-band spectral extinction vector based on whether the reversal has occurred, and calibrate the evolution state of the full-band spectral extinction vector according to the time-varying evolution mode. S6: Select a matching pre-stored feature extinction vector based on the evolution state calibration result, compare it with the full-band spectral extinction vector to identify the type of burning material, and combine it with the remaining spectral feature parameters after stripping the distortion offset to output the fire identification result including the fire hazard level.

[0006] Furthermore, S1 includes: The multi-band detection unit can simultaneously acquire ultraviolet, visible, near-infrared, mid-infrared, and far-infrared spectral data of the same field of view within the monitoring area using a synchronous triggering method. Band registration is performed on the spectral data of different bands collected at the same time, so that the spectral data of different bands are aligned pixel by pixel in spatial position, forming registered multi-band spectral data covering the ultraviolet band to the far-infrared band.

[0007] Furthermore, S2 includes: Noise removal is performed on the registered multi-band spectral data to obtain denoised multi-band spectral data. Environmental baseline correction was performed on the spectral response values ​​of each band in the denoised multi-band spectral data to eliminate the non-uniform background shift caused by water vapor absorption band and solar scattered light in each band, thus obtaining environmentally compensated multi-band spectral data. From the environmentally compensated multi-band spectral data, the radiative energy values ​​of each band at the preset characteristic spectral lines are extracted, and the spectral energy distribution sequence composed of the radiative energy values ​​of each band is used as the spectral characteristic parameter.

[0008] Furthermore, environmental baseline correction is performed on the spectral response values ​​of each band in the denoised multi-band spectral data to eliminate the non-uniform background shifts caused by water vapor absorption bands and solar scattered light in each band, including: Under the condition that there is no fire in the monitoring area, the water vapor absorption spectrum response curve and solar scattering spectrum response curve of each band under clear sky conditions are obtained in advance. For the spectral response values ​​of each band in the denoised multi-band spectral data, the absorption component of the corresponding band on the water vapor absorption spectral response curve and the scattering component of the corresponding band on the solar scattering spectral response curve are subtracted band by band to obtain the environmentally compensated multi-band spectral data.

[0009] Furthermore, S3 includes: Obtain the wavelength points corresponding to the peak positions in the spectral energy distribution sequence, and divide the full-band spectral energy distribution into short-wavelength side energy distribution curves and long-wavelength side energy distribution curves using the wavelength points as boundaries; By integrating the energy distribution curves on the shortwave side and the longwave side respectively, the integrated areas on the shortwave side and the longwave side are obtained. Calculate the ratio of the short-wavelength side integral area to the long-wavelength side integral area. If the ratio exceeds the preset symmetry tolerance range, it is determined that the symmetry of the full-band spectral energy distribution has been broken, and it is identified as the presence of optical path interference.

[0010] Furthermore, S4 includes: Using the long-wavelength integrated area as a reference, the deviation of the ratio of the short-wavelength integrated area to the long-wavelength integrated area from the center value of the preset symmetry tolerance interval is taken as the distortion offset. The remaining spectral characteristic parameters after removing the distortion offset are obtained by subtracting the attenuation component of the distortion offset in each band from the spectral energy distribution sequence. The extinction components of each band are obtained by subtracting the spectral energy distribution sequence from the remaining spectral characteristic parameters after removing the distortion offset, and the extinction components of each band are used to form the full-band spectral extinction vector.

[0011] Furthermore, subtracting the attenuation component corresponding to the distortion shift in each band from the spectral energy distribution sequence, including: Obtain the preset attenuation coefficient curve of Mie scattering attenuation coefficient of flue gas particles as a function of wavelength; Multiply the distortion offset by the attenuation coefficient of each band on the attenuation coefficient curve to obtain the attenuation component of the distortion offset in each band. Subtract the attenuation component of the corresponding band from the radiant energy value of each band in the spectral energy distribution sequence.

[0012] Furthermore, S5 includes: Extract the extinction components of each band in the full-band spectral extinction vector at the start of the preset time window, and arrange them in band order to form the starting spectral attenuation distribution profile; and extract the extinction components of each band in the full-band spectral extinction vector at the end of the preset time window, and arrange them in band order to form the ending spectral attenuation distribution profile. For the spectral attenuation distribution profile at the starting point, compare the sum of the extinction components from the ultraviolet band to the visible band with the sum of the extinction components from the mid-infrared band to the far-infrared band. If the sum of the extinction components from the ultraviolet band to the visible band is greater than the sum of the extinction components from the mid-infrared band to the far-infrared band, then mark the direction of the asymmetry at the starting point as the short-wavelength side attenuation being dominant, and vice versa as the long-wavelength side attenuation being dominant. Compare and mark the direction of asymmetry at the endpoint for the spectral attenuation distribution profile in the same manner. If the asymmetric direction at the starting point is different from the asymmetric direction at the ending point, it is determined that the asymmetric direction has reversed, the time-varying evolution mode of the full-band spectral extinction vector is determined to be the compositional evolution mode, and the full-band spectral extinction vector is calibrated as the compositional evolution state. If the asymmetric direction at the starting point is the same as the asymmetric direction at the ending point, it is determined that the asymmetric direction has not been reversed, the time-varying evolution mode of the full-band spectral extinction vector is determined to be the diffusion-dilution mode, and the full-band spectral extinction vector is calibrated as the diffusion-dilution state.

[0013] Furthermore, S6 includes: Based on the evolution state calibration results, retrieve the first pre-stored feature extinction vector sub-library corresponding to the composition evolution state, or the second pre-stored feature extinction vector sub-library corresponding to the diffusion and dilution state, from the pre-stored feature extinction vector library. The full-band spectral extinction vector is compared with each of the pre-stored feature extinction vectors in the retrieved pre-stored feature extinction vector sub-library for similarity. The type of combustion substance corresponding to the pre-stored feature extinction vector with the highest similarity is selected as the type of combustion substance identified. The identified type of burning material is jointly judged with the remaining spectral feature parameters after stripping the distortion offset. When the remaining spectral feature parameters fall into the preset fire hazard level range and the type of burning material corresponds to a high-risk combustible, the fire situation is confirmed and marked as a high fire hazard level. When the remaining spectral characteristic parameters fall within the preset fire hazard level range but the type of burning material corresponds to a non-high-risk combustible, output confirmation of fire and mark it as a medium fire hazard level. When the remaining spectral characteristic parameters do not fall within the preset fire hazard level range, but the type of burning material corresponds to a high-risk combustible, a suspected fire is output and marked as a high fire hazard level. When the remaining spectral characteristic parameters do not fall within the preset fire hazard level range and the type of burning material corresponds to a non-high-risk combustible, the output is "No fire".

[0014] On the other hand, the present invention provides a large-scale fire detection and alarm device for wooden building complexes, comprising: The spectral data module is used to acquire multi-band spectral data within the monitoring area synchronously collected by distributed multi-band detection units. The multi-band spectral data covers the ultraviolet band to the far-infrared band. The data preprocessing module is used to preprocess multi-band spectral data and extract spectral feature parameters; The interference identification module is used to determine whether the symmetry of the full-band spectral energy distribution characterized by the spectral feature parameters is broken. If it is broken, it is identified as optical path interference. The extinction vector module is used to extract the distortion offset, which represents non-uniform attenuation distortion, from the spectral characteristic parameters when there is optical path interference, and to construct a full-band spectral extinction vector based on the distortion offset. The evolution analysis module is used to extract the spectral attenuation distribution profile of the full-band spectral extinction vector at the start and end of a preset time window, determine whether the asymmetric direction of the attenuation on the short-wavelength side and the attenuation on the long-wavelength side has reversed within the preset time window, determine the time-varying evolution mode of the full-band spectral extinction vector based on whether the reversal has occurred, and calibrate the evolution state of the full-band spectral extinction vector according to the time-varying evolution mode. The fire identification module is used to select a matching pre-stored feature extinction vector based on the evolution state calibration result, compare it with the full-band spectral extinction vector to identify the type of burning material, and combine the remaining spectral feature parameters after stripping the distortion offset to output the fire identification result including the fire hazard level.

[0015] The beneficial effects of this invention are: 1. Smoke interference is identified by analyzing the morphological changes in the full-band spectral energy distribution, rather than relying directly on distorted spectral values. When smoke particles in narrow alleys and eaves of wooden building complexes cause non-uniform attenuation of light radiation, the symmetry of the spectral energy distribution is broken. This morphological distortion is more stable and reliable than single-band intensity changes and is less susceptible to fluctuations in lighting conditions and overall temperature drift. After identifying optical path interference, the distortion information is extracted from the spectral feature parameters and independently constructed as a full-band spectral extinction vector. This separates the radiation characteristics of the fire source itself from the extinction effect of the smoke propagation process. Subsequent fire severity assessments can be based on these two different types of information, avoiding the assessment bias caused by superimposing smoke interference on fire source characteristics.

[0016] 2. By analyzing whether the attenuation asymmetry between the short-wavelength and long-wavelength sides of the full-band spectral extinction vector reverses within a preset time window, the evolution process of flue gas is divided into two categories: compositional evolution mode and diffusion-dilution mode. The type of flue gas source material is identified. Combining the real radiation characteristics of the fire source recovered after stripping the distortion offset with the identification results of the combustion material type, the judgment of fire hazard level is based on a comprehensive assessment of the dual dimensions of fire source activity and potential hazard of combustion material. This allows high-risk fast-burning fires caused by building repair materials such as tung oil and raw lacquer to be accurately identified and marked as high fire hazard level, making up for the deficiency of existing methods that misjudge such fires as low-risk due to flue gas optical path distortion. Attached Figure Description

[0017] Figure 1 This is a flowchart of a method for large-scale fire detection and alarm in a wooden building complex according to the present invention; Figure 2 This is a schematic diagram of the structure of a large-scale fire detection and alarm device for wooden building complexes according to the present invention. Detailed Implementation

[0018] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0019] Example 1: Figure 1 This invention provides a method for large-scale fire detection and alarm in wooden building complexes, comprising: S1: Acquire multi-band spectral data within the monitoring area synchronously collected by distributed multi-band detection units, with the multi-band spectral data covering the ultraviolet to far-infrared bands; S2: Preprocess multi-band spectral data and extract spectral feature parameters; S3: Determine whether the symmetry of the full-band spectral energy distribution characterized by the spectral characteristic parameters is broken. If it is broken, it is identified as optical path interference. S4: When there is optical path interference, the distortion offset that characterizes non-uniform attenuation distortion is extracted from the spectral characteristic parameters, and a full-band spectral extinction vector is constructed based on the distortion offset. S5: Extract the spectral attenuation distribution contour of the full-band spectral extinction vector at the start and end of the preset time window, determine whether the asymmetric direction of the attenuation on the short-wavelength side and the attenuation on the long-wavelength side has reversed within the preset time window, determine the time-varying evolution mode of the full-band spectral extinction vector based on whether the reversal has occurred, and calibrate the evolution state of the full-band spectral extinction vector according to the time-varying evolution mode. S6: Select a matching pre-stored feature extinction vector based on the evolution state calibration result, compare it with the full-band spectral extinction vector to identify the type of burning material, and combine it with the remaining spectral feature parameters after stripping the distortion offset to output the fire identification result including the fire hazard level.

[0020] In a specific implementation of S1, the multi-band detection unit acquires ultraviolet, visible, near-infrared, mid-infrared and far-infrared spectral data of the same field of view within the monitoring area at the same time in a synchronous triggering manner.

[0021] The multi-band detection unit is a common-path multi-band detection device that integrates ultraviolet band detectors, visible light band detectors, near-infrared band detectors, mid-infrared band detectors, and far-infrared band detectors. The response band range of the ultraviolet band detector is, for example, 0.25μm to 0.38μm, the response band range of the visible light band detector is, for example, 0.38μm to 0.78μm, the response band range of the near-infrared band detector is, for example, 0.78μm to 2.5μm, the response band range of the mid-infrared band detector is, for example, 2.5μm to 8μm, and the response band range of the far-infrared band detector is, for example, 8μm to 14μm.

[0022] The synchronous triggering method is achieved through an external synchronous triggering signal source. The synchronous triggering signal source simultaneously sends trigger pulses to the ultraviolet band detector, visible light band detector, near-infrared band detector, mid-infrared band detector, and far-infrared band detector. The five band detectors start integration exposure at the same moment they receive the trigger pulses. The integration exposure time is preset according to the radiation intensity of each band on the surface of a typical wooden structure in the monitoring area under fire-free conditions. Specifically, during the fire-free period of the deployment phase, the integration exposure time of the ultraviolet band detector, visible light band detector, near-infrared band detector, mid-infrared band detector, and far-infrared band detector is adjusted so that the signal collected by each band detector within the integration exposure time reaches 60% to 80% of the detector's full-sink capacity. The integration exposure time of each band detector at this time is used as the fixed integration exposure time parameter for subsequent synchronous acquisition to ensure that the spectral data of each band has consistent effective bit depth and signal-to-noise ratio.

[0023] The same field of view is achieved by designing the optical receiving paths of the ultraviolet, visible, near-infrared, mid-infrared, and far-infrared detectors as a common principal mirror beam splitting structure. After the incident light radiation is collected by the same principal mirror, it is sequentially split into ultraviolet, visible, near-infrared, mid-infrared, and far-infrared beams by a beam splitter, and then guided to the focal plane of the detectors of the corresponding bands, so that the spatial ranges observed by the ultraviolet, visible, near-infrared, mid-infrared, and far-infrared detectors completely overlap.

[0024] Band registration is performed on ultraviolet, visible, near-infrared, mid-infrared, and far-infrared spectral data collected at the same time, so that the spectral data of different bands are aligned pixel by pixel in spatial location, forming registered multi-band spectral data covering the ultraviolet to far-infrared bands.

[0025] Band registration uses visible light spectral data as the registration reference. Feature point targets with known spatial coordinates are deployed within the monitoring area. These feature point targets exhibit identifiable spectral response differences in the ultraviolet, visible, near-infrared, mid-infrared, and far-infrared bands. The pixel coordinates of the feature point targets in the ultraviolet, visible, near-infrared, mid-infrared, and far-infrared spectral data are obtained respectively. Using the pixel coordinates of the feature point targets in the visible light spectral data as a reference, the spatial transformation relationship of the ultraviolet, near-infrared, mid-infrared, and far-infrared spectral data relative to the visible light spectral data is calculated. The spatial transformation relationship includes translation, rotation, and scaling.

[0026] Pixel-by-pixel alignment is based on the calculated spatial transformation relationship. The ultraviolet, near-infrared, mid-infrared, and far-infrared spectral data are spatially resampled respectively. This ensures that the same spatial location in the resampled ultraviolet, near-infrared, mid-infrared, and far-infrared spectral data corresponds to the same physical point within the monitoring area, forming registered multi-band spectral data. Each spatial location in the registered multi-band spectral data contains ultraviolet, visible, near-infrared, mid-infrared, and far-infrared radiance values.

[0027] In a specific implementation of S2, noise removal is performed on the registered multi-band spectral data to obtain denoised multi-band spectral data.

[0028] Noise removal is achieved through a combination of temporal smoothing filtering and spatial median filtering. Temporal smoothing filtering performs moving average processing on the ultraviolet, visible, near-infrared, mid-infrared, and far-infrared radiance values ​​at each spatial location in the registered multi-band spectral data, along with the continuously acquired time series. The window width for moving average processing is determined by the frame rate of the multi-band detection unit and is calculated as: W = floor(F / 10); where W represents the window width in frames, F represents the frame rate in frames per second, and floor represents rounding down; the frame rate is the number of frames acquired per second by the multi-band detection unit, determined by the trigger pulse frequency of the synchronization trigger signal source. Spatial median filtering, for each band in the registered multi-band spectral data at the same time, takes a neighborhood window centered on each pixel, sorts the radiance values ​​of all pixels within the neighborhood window, and replaces the radiance value of the center pixel with the median value. The neighborhood window size is, for example, 3 pixels × 3 pixels. After time-domain smoothing and spatial-domain median filtering, the registered multi-band spectral data is the denoised multi-band spectral data. Each spatial location in the denoised multi-band spectral data contains the denoised ultraviolet radiance, the denoised visible radiance, the denoised near-infrared radiance, the denoised mid-infrared radiance, and the denoised far-infrared radiance.

[0029] Environmental baseline correction is performed on the spectral response values ​​of each band in the denoised multi-band spectral data to eliminate the non-uniform background shifts caused by water vapor absorption bands and solar scattered light in each band, resulting in environmentally compensated multi-band spectral data. The spectral response values ​​of each band in the denoised multi-band spectral data refer to the denoised ultraviolet, visible, near-infrared, mid-infrared, and far-infrared radiance values ​​at each spatial location. The specific method of environmental baseline correction is as follows: Under fire-free conditions within the monitoring area, the water vapor absorption spectral response curves and solar scattering spectral response curves for each band under clear sky conditions are pre-acquired; for each band in the denoised multi-band spectral data, the absorption component of the corresponding band on the water vapor absorption spectral response curve and the scattering component of the corresponding band on the solar scattering spectral response curve are subtracted band by band from the spectral response values ​​of each band, resulting in environmentally compensated multi-band spectral data.

[0030] Under fire-free conditions within the monitoring area, the water vapor absorption spectral response curves and solar scattering spectral response curves for each band under clear sky conditions are pre-acquired. Specifically, after deployment in the monitoring area and confirmation of fire-free conditions, clear, cloudless weather conditions are selected. Registered multi-band spectral data are continuously collected at multiple times using a multi-band detection unit. The average spectral data is obtained by averaging the registered multi-band spectral data pixel-by-pixel according to spatial location. The water vapor absorption spectral response curve is obtained as follows: In the average spectral data, the band radiation value with a center wavelength at the known absorption peak wavelength of water vapor is taken as the water vapor absorption peak value, and the band radiation value with a center wavelength at the known non-absorption wavelength of water vapor is taken as the water vapor absorption baseline value. The difference between the water vapor absorption peak value and the water vapor absorption baseline value is taken as the water vapor absorption component. Arranging the water vapor absorption components of each band according to wavelength yields the water vapor absorption spectral response curve. The known absorption peak wavelengths of water vapor are, for example, 1.38 μm, 1.87 μm, 2.7 μm, and 6.3 μm, and the known non-absorption wavelengths of water vapor are, for example, 1.0 μm, 1.6 μm, 2.2 μm, and 4.0 μm. The solar scattering spectral response curve is obtained as follows: In the average spectral data, the band radiation value within the shaded area where direct solar radiation is blocked by buildings is selected as the scattering reference value, and the band radiation value within the directly irradiated area where direct solar radiation is not blocked is selected as the scattering peak value. The difference between the scattering peak value and the scattering reference value is taken as the solar scattering component. Arranging the solar scattering components of each band according to wavelength yields the solar scattering spectral response curve. Subtracting the absorption component of the corresponding band on the water vapor absorption spectral response curve and the scattering component of the corresponding band on the solar scattering spectral response curve for each band at each spatial location in the denoised multi-band spectral data means subtracting the absorption component of the same band in the water vapor absorption spectral response curve and the scattering component of the same band in the solar scattering spectral response curve from the spectral response value of each band at each spatial location.

[0031] After environmental baseline correction, the non-uniform background shift caused by water vapor absorption and solar scattering in the spectral response values ​​of each band at each spatial location has been eliminated, resulting in environmentally compensated multi-band spectral data. Each spatial location in the environmentally compensated multi-band spectral data includes environmentally compensated ultraviolet band radiation values, environmentally compensated visible band radiation values, environmentally compensated near-infrared band radiation values, environmentally compensated mid-infrared band radiation values, and environmentally compensated far-infrared band radiation values.

[0032] From the environmentally compensated multi-band spectral data, the radiant energy values ​​of each band at the preset characteristic spectral lines are extracted, and the spectral energy distribution sequence composed of the radiant energy values ​​of each band is used as the spectral characteristic parameter. The preset characteristic spectral lines are determined in advance based on the characteristic emission spectral lines of typical combustion products in timber-framed building complexes. Specifically, under laboratory conditions, common building timber and building repair materials from timber-framed building complexes are burned separately, and the emission spectra of each band during the combustion process are collected using a spectroradiometer. The wavelength positions corresponding to the peak radiant energy of each band are extracted from the emission spectra, and the wavelength positions corresponding to one peak radiant energy in each of the ultraviolet, visible, near-infrared, mid-infrared, and far-infrared bands are selected as the preset characteristic spectral lines. For example, the preset characteristic spectral line in the ultraviolet band is taken as the wavelength of the emission spectrum of hydroxyl radicals during combustion (0.308 μm), the preset characteristic spectral line in the visible light band is taken as the wavelength of the emission spectrum of hydrocarbon radicals during combustion (0.431 μm), the preset characteristic spectral line in the near-infrared band is taken as the wavelength of the emission spectrum of water vapor during combustion (2.7 μm), the preset characteristic spectral line in the mid-infrared band is taken as the wavelength of the emission spectrum of carbon dioxide during combustion (4.3 μm), and the preset characteristic spectral line in the far-infrared band is taken as the wavelength of the emission spectrum of carbon monoxide during combustion (10.6 μm). For each spatial location, the radiant energy values ​​at 0.308 μm in the ultraviolet band, 0.431 μm in the visible band, 2.7 μm in the near-infrared band, 4.3 μm in the mid-infrared band, and 10.6 μm in the far-infrared band are extracted from the environmentally compensated multi-band spectral data. The five radiant energy values ​​are arranged in ascending order of wavelength to form the spectral energy distribution sequence for that spatial location. The spectral energy distribution sequence serves as the spectral characteristic parameter for that spatial location.

[0033] In a specific implementation of S3, the wavelength point corresponding to the peak position in the spectral energy distribution sequence is obtained, and the full-band spectral energy distribution is divided into short-wavelength side energy distribution curves and long-wavelength side energy distribution curves with the wavelength point as the boundary.

[0034] The spectral energy distribution sequence is composed of ultraviolet, visible, near-infrared, mid-infrared, and far-infrared radiant energy values ​​arranged in ascending order of wavelength. The peak position refers to the wavelength point corresponding to the band with the highest radiant energy value in the spectral energy distribution sequence. The peak position is obtained by comparing the radiant energy values ​​of each band in the spectral energy distribution sequence one by one, selecting the band with the highest radiant energy value, and taking the wavelength point of that band as the wavelength point corresponding to the peak position. Using the wavelength point as a boundary, bands with wavelengths shorter than the wavelength point corresponding to the peak position form the short-wavelength side energy distribution curve, while bands with wavelengths longer than the wavelength point corresponding to the peak position form the long-wavelength side energy distribution curve. For example, when the wavelength point corresponding to the peak position is 2.7 μm in the near-infrared band, the energy distribution curve on the short-wave side is composed of the wavelength points and radiant energy values ​​corresponding to the radiant energy values ​​in the ultraviolet band and the visible light band, while the energy distribution curve on the long-wave side is composed of the wavelength points and radiant energy values ​​corresponding to the radiant energy values ​​in the mid-infrared band and the far-infrared band.

[0035] Integrating the energy distribution curves on the shortwave side and the longwave side respectively yields the integrated area on the shortwave side and the integrated area on the longwave side.

[0036] The shortwave side integral area is calculated as: Ashort = Σ(Ei × Δλi), where i traverses all bands in the shortwave side energy distribution curve; where Ashort represents the shortwave side integral area, in units of radiant energy value multiplied by wavelength, Ei represents the radiant energy value of the i-th band in the shortwave side energy distribution curve, and Δλi represents the wavelength width of the i-th band in the shortwave side energy distribution curve, which is the difference between the upper and lower wavelength limits of the band's response wavelength range. The longwave side integral area is calculated as: Along = Σ(Ej × Δλj), where j traverses all bands in the longwave side energy distribution curve; where Along represents the longwave side integral area, in units of radiant energy value multiplied by wavelength, Ej represents the radiant energy value of the j-th band in the longwave side energy distribution curve, and Δλj represents the wavelength width of the j-th band in the longwave side energy distribution curve. For example, the wavelength width in the ultraviolet band is 0.13 μm, the wavelength width in the visible light band is 0.4 μm, the wavelength width in the near-infrared band is 1.72 μm, the wavelength width in the mid-infrared band is 5.5 μm, and the wavelength width in the far-infrared band is 6 μm.

[0037] Calculate the ratio of the short-wavelength side integral area to the long-wavelength side integral area. If the ratio exceeds the preset symmetry tolerance range, it is determined that the symmetry of the full-band spectral energy distribution has been broken, and it is identified as the presence of optical path interference.

[0038] The ratio of the shortwave integral area to the longwave integral area is calculated as: R = Ashort / Along; where R represents the ratio of the shortwave integral area to the longwave integral area, which is dimensionless. The preset symmetry tolerance range is pre-set based on the natural symmetry fluctuation range of the full-band spectral energy distribution under conditions of no fire and no smoke interference within the monitoring area. Specifically, after the monitoring area is deployed and confirmed to be free of fire and smoke interference, spectral characteristic parameters are continuously collected at multiple times. For each time point, the ratio of the shortwave integral area to the longwave integral area is calculated using the above method. The average of these ratios is taken as the center of symmetry. A preset percentage above and below the center of symmetry is used as the upper and lower limits of the preset symmetry tolerance range, for example, 10%. When the calculated ratio R is greater than the upper limit of the preset symmetry tolerance range or less than the lower limit, i.e., R exceeds the preset symmetry tolerance range, the symmetry of the full-band spectral energy distribution is determined to be broken, indicating the presence of optical path interference. Under clean optical path conditions free from smoke particle interference, the full-band spectral energy distribution exhibits an approximately symmetrical shape about the peak position. The integrated area on the short-wavelength side is nearly equal to that on the long-wavelength side, with a ratio R close to 1 and a limited fluctuation range. When smoke particles are present in the optical path, they selectively attenuate light radiation at different wavelengths. The scattering attenuation at the short-wavelength band is significantly stronger than that at the long-wavelength band, resulting in a significant decrease in the integrated area on the short-wavelength side relative to the long-wavelength side. The ratio R deviates significantly, exceeding the natural symmetry fluctuation range, which indicates the presence of optical path interference.

[0039] In a specific implementation of S4, when optical path interference is identified, the deviation of the ratio of the short-wavelength side integral area to the long-wavelength side integral area relative to the center value of the preset symmetry tolerance interval is taken as the distortion offset, with the long-wavelength side integral area as the reference benchmark.

[0040] The distortion offset is calculated as: D = R - Rcenter; where D represents the dimensionless distortion offset, R represents the ratio of the short-wavelength integrated area to the long-wavelength integrated area, and Rcenter represents the center of symmetry of the preset symmetry tolerance interval. The distortion offset D reflects the overall intensity of the non-uniform attenuation distortion caused by smoke particles on the full-band spectral energy distribution in the optical path. A positive D indicates that the energy on the short-wavelength side has excessive attenuation relative to the energy on the long-wavelength side, and a negative D indicates that the energy on the long-wavelength side has excessive attenuation relative to the energy on the short-wavelength side. In the narrow alleys and eaves of wooden building complexes, the smoke particles generated at the onset of a fire are mainly fine particles, and the scattering attenuation of the short-wavelength band is significantly stronger than that of the long-wavelength band. At this time, the distortion offset D is usually positive, indicating that the integrated area on the short-wavelength side has decreased significantly relative to the integrated area on the long-wavelength side.

[0041] The remaining spectral characteristic parameters after removing the distortion offset are obtained by subtracting the attenuation component of the distortion offset in each band from the spectral energy distribution sequence.

[0042] The attenuation coefficient curve of the Mie scattering attenuation coefficient of flue gas particles as a function of wavelength was obtained. This curve was pre-calculated using Mie scattering theory based on the particle size distribution and complex refractive index of flue gas particles generated from the combustion of typical building timber and building repair materials in a timber-framed building complex. Specifically, under laboratory conditions, flue gas particle samples generated from the combustion of building timber and building repair materials were collected. The particle size distribution of the flue gas particle samples was measured using a particle size analyzer, and the complex refractive index of the flue gas particle samples at each wavelength was measured using a spectroscopic ellipsometry. The particle size distribution and complex refractive index were input into the Mie scattering calculation program, which output the scattering attenuation coefficient at each wavelength. The scattering attenuation coefficients at each wavelength were then calculated according to... The wavelengths are arranged and normalized. Normalization is performed by dividing the scattering attenuation coefficient of each band by the sum of the scattering attenuation coefficients of all bands, so that the sum of the attenuation coefficients of each band on the attenuation coefficient curve equals 1. This results in a normalized attenuation coefficient curve, which shows the attenuation coefficients for the ultraviolet, visible, near-infrared, mid-infrared, and far-infrared bands. For example, the attenuation coefficient for the ultraviolet band is αuv, for the visible band it is αvis, for the near-infrared band it is αnir, for the mid-infrared band it is αmir, and for the far-infrared band it is αfir, and αuv + αvis + αnir + αmir + αfir = 1. Since the particle size of flue gas is usually comparable to that of short-wavelength wavelengths, the scattering efficiency for short-wavelength bands is much higher than that for long-wavelength bands. Therefore, the attenuation coefficient for short-wavelength bands on the attenuation coefficient curve is greater than that for long-wavelength bands.

[0043] Multiply the distortion offset by the attenuation coefficient of each band on the attenuation coefficient curve to obtain the attenuation component of the distortion offset in each band. The attenuation component corresponding to the ultraviolet band is calculated as follows: ΔEuv = D × αuv; the attenuation component corresponding to the visible band is calculated as follows: ΔEvis = D × αvis; the attenuation component corresponding to the near-infrared band is calculated as follows: ΔEnir = D × αnir; the attenuation component corresponding to the mid-infrared band is calculated as follows: ΔEmir = D × αmir; the attenuation component corresponding to the far-infrared band is calculated as follows: ΔEfir = D × αfir; where ΔEuv represents the attenuation component corresponding to the ultraviolet band, ΔEvis represents the attenuation component corresponding to the visible band, ΔEnir represents the attenuation component corresponding to the near-infrared band, ΔEmir represents the attenuation component corresponding to the mid-infrared band, and ΔEfir represents the attenuation component corresponding to the far-infrared band; D represents the distortion offset, which is dimensionless; and αuv, αvis, αnir, αmir, and αfir represent the attenuation coefficients of the ultraviolet, visible, near-infrared, mid-infrared, and far-infrared bands on the attenuation coefficient curve, respectively, which are dimensionless.

[0044] Subtracting the corresponding attenuation component from the radiant energy value of each band in the spectral energy distribution sequence, i.e., subtracting ΔEuv from the radiant energy value of the ultraviolet band, yields the radiant energy value of the ultraviolet band after distortion offset removal. Subtracting ΔEvis from the radiant energy value of the visible light band, yields the radiant energy value of the visible light band after distortion offset removal. Subtracting ΔEnir from the radiant energy value of the near-infrared band, yields the radiant energy value of the near-infrared band after distortion offset removal. Subtracting ΔEmir from the radiant energy value of the mid-infrared band, yields the radiant energy value of the mid-infrared band after distortion offset removal. Subtracting ΔEfir from the radiant energy value of the far-infrared band, yields the radiant energy value of the far-infrared band after distortion offset removal. The above five band radiant energy values ​​after distortion offset removal are arranged in ascending order of wavelength to form the remaining spectral characteristic parameters after distortion offset removal. The remaining spectral feature parameters after removing the distortion offset represent the characteristics of the fire source's radiation across the entire spectral energy distribution after eliminating the influence of non-uniform attenuation distortion of smoke particles. This stripping method separates the optical path interference of smoke particles from the fire source's own radiation characteristics, enabling subsequent fire identification to be based on the true spectral characteristics of the fire source that are not contaminated by smoke distortion.

[0045] The extinction components of each band are obtained by subtracting the spectral energy distribution sequence from the remaining spectral characteristic parameters after removing the distortion offset, and the extinction components of each band are used to form the full-band spectral extinction vector.

[0046] The band-by-band subtraction method is as follows: Subtract the ultraviolet band radiant energy value (after removing the distortion offset) from the remaining spectral characteristic parameters in the spectral energy distribution sequence to obtain the ultraviolet band extinction component; subtract the visible light band radiant energy value (after removing the distortion offset) from the remaining spectral characteristic parameters in the spectral energy distribution sequence to obtain the visible light band extinction component; subtract the near-infrared band radiant energy value (after removing the distortion offset) from the near-infrared band radiant energy value to obtain the near-infrared band extinction component; subtract the mid-infrared band radiant energy value (after removing the distortion offset) from the mid-infrared band radiant energy value to obtain the mid-infrared band extinction component; and subtract the far-infrared band radiant energy value (after removing the distortion offset) from the far-infrared band radiant energy value to obtain the far-infrared band extinction component. The vector composed of the ultraviolet, visible, near-infrared, mid-infrared, and far-infrared extinction components arranged in ascending order of wavelength is the full-band spectral extinction vector. The full-band spectral extinction vector completely describes the extinction intensity distribution of the light radiation of each band by the flue gas particles in the optical path, and the extinction component of each band is the absolute amount of light radiation of that band attenuated by the flue gas particles.

[0047] In a specific implementation of S5, the extinction components of each band in the full-band spectral extinction vector at the start of the preset time window are extracted and arranged in band order to form the starting point spectral attenuation distribution profile; and the extinction components of each band in the full-band spectral extinction vector at the end of the preset time window are extracted and arranged in band order to form the end point spectral attenuation distribution profile.

[0048] The preset time window is pre-set based on the frame rate of the multi-band detection unit and the response time requirements of fire monitoring. Specifically, the preset time window duration is set as an integer multiple of the time interval between two consecutive frames of data acquired by the multi-band detection unit. The start point of the preset time window is the moment corresponding to the preset time window duration preceding the current moment, and the end point is the current moment. For example, if the frame rate of the multi-band detection unit is 10 frames per second, and the time interval between two consecutive frames is 0.1 seconds, the preset time window duration could be set to 2 seconds, encompassing 20 consecutive frames of full-band spectral extinction vectors. The start point of the preset time window would be the moment corresponding to 2 seconds preceding the current moment. The full-band spectral extinction vector is composed of extinction components in the ultraviolet, visible, near-infrared, mid-infrared, and far-infrared bands, arranged in ascending order of wavelength. Arranged sequentially by band, i.e., in the order of ultraviolet, visible, near-infrared, mid-infrared, and far-infrared, the extinction components of each band are arranged in that order, forming a curve with the band as the horizontal axis and the extinction component as the vertical axis. This curve is the spectral attenuation distribution profile, which visually describes the distribution of the attenuation intensity of the smoke particles to the light radiation of each band along the spectral dimension. The starting point spectral attenuation distribution profile reflects the extinction characteristics of the smoke particles at the beginning of the preset time window, while the ending point spectral attenuation distribution profile reflects the extinction characteristics of the smoke particles at the end of the preset time window.

[0049] For the spectral attenuation distribution profile at the starting point, compare the sum of the extinction components from the ultraviolet to the visible band with the sum of the extinction components from the mid-infrared to the far-infrared band. If the sum of the extinction components from the ultraviolet to the visible band is greater than the sum of the extinction components from the mid-infrared to the far-infrared band, then mark the direction of asymmetry at the starting point as having predominance on the short-wavelength side, and vice versa.

[0050] The sum of extinction components from the ultraviolet to the visible light band is calculated as: Eshorttotal = Euv + Evis; where Eshorttotal represents the sum of extinction components from the ultraviolet to the visible light band, Euv represents the ultraviolet extinction component, and Evis represents the visible light extinction component. The sum of extinction components from the mid-infrared to the far-infrared band is calculated as: Elongtotal = Emir + Efir; where Elongtotal represents the sum of extinction components from the mid-infrared to the far-infrared band, Emir represents the mid-infrared extinction component, and Efir represents the far-infrared extinction component. Short-wavelength attenuation dominance indicates that the attenuation intensity of flue gas particles in the ultraviolet and visible light bands is greater than that in the mid-infrared and far-infrared bands, which is consistent with the Mie scattering characteristics of flue gas particles, which are mainly fine particles. Long-wavelength attenuation dominance indicates that the attenuation intensity of flue gas particles in the mid-infrared and far-infrared bands is greater than that in the ultraviolet and visible light bands. This usually occurs when the particle size of flue gas particles increases or when the content of absorbent gases such as water vapor and carbon dioxide in the flue gas increases.

[0051] For the spectral attenuation distribution profile at the endpoint, compare the sum of the extinction components from the ultraviolet to the visible band with the sum of the extinction components from the mid-infrared to the far-infrared band in the same way. If the sum of the extinction components from the ultraviolet to the visible band is greater than the sum of the extinction components from the mid-infrared to the far-infrared band, then mark the direction of asymmetry at the endpoint as having predominance on the short-wavelength side, and vice versa as having predominance on the long-wavelength side.

[0052] If the asymmetric direction at the starting point differs from that at the ending point, the asymmetric direction is determined to have reversed. The time-varying evolution mode of the full-band spectral extinction vector is then identified as the compositional evolution mode, and the full-band spectral extinction vector is calibrated as the compositional evolution state. A reversal of the asymmetric direction indicates a qualitative change in the extinction characteristics of the smoke particles within a preset time window, shifting from predominantly attenuated on the short-wavelength side to predominantly attenuated on the long-wavelength side, or vice versa. The physical meaning of this reversal is that the composition or particle size distribution of the smoke particles has significantly changed within the preset time window. For example, it could be a shift from initial combustion smoke dominated by fine particle scattering to pyrolysis products dominated by water vapor and carbon dioxide absorption, or from old smoke dominated by large particle settling to newly generated fine particle smoke. The compositional evolution mode differs from a simple concentration dilution process, reflecting the dynamic evolution of the smoke at the chemical and physical levels. This evolutionary information is valuable for distinguishing the type of smoke source material and determining the development stage of a fire.

[0053] If the asymmetric direction at the starting point is the same as the asymmetric direction at the ending point, it is determined that the asymmetric direction has not reversed. The time-varying evolution mode of the full-band spectral extinction vector is then identified as a diffusion-dilution mode, and the full-band spectral extinction vector is calibrated as a diffusion-dilution state. The absence of asymmetric direction reversal indicates that within a preset time window, the extinction characteristics of the flue gas particles have not undergone a qualitative change; the dominance of short-wavelength attenuation or long-wavelength attenuation remains unchanged. The flue gas only undergoes concentration dilution and diffusion, and the composition and particle size distribution of the particles remain relatively stable. The diffusion-dilution mode corresponds to the physical process of flue gas being gradually transported and diluted under the action of airflow. At this time, the extinction components of each band decrease synchronously, but their proportional relationship remains unchanged, and the asymmetric direction remains stable. By judging whether the asymmetric direction has reversed, the flue gas evolution process is divided into two time-varying evolution modes: compositional evolution mode and diffusion-dilution mode. Based on this, the evolution state of the full-band spectral extinction vector is calibrated, providing a basis for subsequently selecting a matching pre-stored characteristic extinction vector based on the evolution state calibration results.

[0054] In a specific implementation of S6, based on the evolution state calibration result, the first pre-stored feature extinction vector sub-library corresponding to the component evolution state or the second pre-stored feature extinction vector sub-library corresponding to the diffusion and dilution state is retrieved from the pre-stored feature extinction vector library.

[0055] The pre-stored feature extinction vector library is established in advance during the deployment phase. Specifically, it is established as follows: Under laboratory conditions, various types of building timber and various building repair materials commonly found in wooden building complexes are burned. Building timber includes pine, fir, and elm, while building repair materials include tung oil, raw lacquer, and linseed oil. Multi-band detection units are used to collect the extinction components of the flue gas produced by each material at different combustion stages in each band. This forms a first pre-stored feature extinction vector for each material under the compositional evolution state and a second pre-stored feature extinction vector under the diffusion and dilution state. The first pre-stored feature extinction vectors of all materials are collected to form a first pre-stored feature extinction vector sub-library, and the second pre-stored feature extinction vectors of all materials are collected to form a second pre-stored feature extinction vector sub-library. Each pre-stored feature extinction vector is accompanied by a corresponding combustion substance type label. The compositional evolution state corresponds to a stage where the composition or particle size distribution of the flue gas is undergoing significant changes. At this stage, the asymmetric direction of the extinction characteristics in the spectral dimension reverses over time. The diffusion and dilution state corresponds to a stage where the composition and particle size distribution of the flue gas remain relatively stable, with only concentration dilution occurring. At this stage, the asymmetric direction of the extinction characteristics in the spectral dimension remains stable. If the time-varying evolution mode determined in S5 is the compositional evolution mode, then the first pre-stored feature extinction vector sub-library is retrieved; if the time-varying evolution mode determined in S5 is the diffusion and dilution mode, then the second pre-stored feature extinction vector sub-library is retrieved. By retrieving data according to the evolutionary state calibration results, the scope of comparison and matching is limited to the pre-stored feature extinction vector sub-library consistent with the current flue gas evolution state, avoiding matching errors caused by cross-state comparisons.

[0056] The full-band spectral extinction vector is compared with each of the pre-stored feature extinction vectors in the retrieved pre-stored feature extinction vector sub-library for similarity. The type of combustion substance corresponding to the pre-stored feature extinction vector with the highest similarity is selected as the type of combustion substance identified.

[0057] Similarity comparison is calculated using the cosine similarity of the vector angle. The specific formula for calculating similarity is as follows: cosθ=(Vcurrent·Vpre) / (|Vcurrent|×|Vpre|); where cosθ represents the cosine similarity, which is dimensionless and ranges from -1 to 1. The closer the value is to 1, the more similar the two vectors are. Vcurrent represents the full-band spectral extinction vector, Vpre represents a pre-stored feature extinction vector from the pre-stored feature extinction vector sub-library, · represents the vector dot product operation, |Vcurrent| represents the magnitude of the full-band spectral extinction vector, and |Vpre| represents the magnitude of the pre-stored feature extinction vector. The cosine similarity of the vectors measures the similarity in spectral extinction distribution between the full-band spectral extinction vector and the pre-stored feature extinction vector. It is unaffected by the absolute magnitude of the vectors, reflecting only the matching degree of the relative proportions between the extinction components of each band. This matching method ensures that even with different flue gas concentrations, as long as the flue gas source substances and evolution states are the same, the cosine similarity between the full-band spectral extinction vector and the corresponding pre-stored feature extinction vector remains close to 1, thus accurately identifying the type of combustion substance. The cosine similarity between the full-band spectral extinction vector and each pre-stored feature extinction vector in the retrieved pre-stored feature extinction vector sub-library is calculated one by one. The pre-stored feature extinction vector corresponding to the maximum cosine similarity is selected, and the combustion substance type label attached to this pre-stored feature extinction vector is used as the identified combustion substance type.

[0058] The identified combustion material type is jointly determined with the remaining spectral characteristic parameters after removing the distortion offset. The remaining spectral characteristic parameters after removing the distortion offset consist of the ultraviolet radiant energy value, visible light radiant energy value, near-infrared radiant energy value, mid-infrared radiant energy value, and far-infrared radiant energy value after removing the distortion offset, arranged in ascending order of wavelength. These parameters represent the characteristics of the fire source's radiation across the entire spectral energy distribution after eliminating the influence of non-uniform attenuation distortion of smoke particles.

[0059] The preset fire hazard level range is pre-set based on the baseline fluctuation range of spectral characteristic parameters under fire-free conditions within the monitoring area. Specifically, after the monitoring area is deployed and fire-free conditions are confirmed, the remaining spectral characteristic parameters after removing distortion offsets are continuously collected at multiple times. The sum of the radiation energy values ​​of each band in the remaining spectral characteristic parameters after removing distortion offsets at each time is calculated. The average of the sums at multiple times is taken as the fire-free baseline value. The lower limit of the preset fire hazard level range is set by a preset multiple above the fire-free baseline value. The preset multiple is determined based on the maximum natural fluctuation range of radiation energy caused by historical temperature fluctuations and solar radiation changes within the monitoring area, for example, 1.5 times the fire-free baseline value. There is no upper limit to the preset fire hazard level range. When the sum of the radiant energy values ​​of each band in the remaining spectral characteristic parameters after removing the distortion offset is greater than the lower limit of the preset fire hazard level range, the remaining spectral characteristic parameters are determined to fall within the preset fire hazard level range, indicating that the radiant energy of the fire source itself is significantly higher than that of a fire-free state, and there is active combustion. When the sum of the radiant energy values ​​of each band in the remaining spectral characteristic parameters after removing the distortion offset is less than or equal to the lower limit of the preset fire hazard level range, the remaining spectral characteristic parameters are determined not to fall within the preset fire hazard level range, indicating that the radiant energy of the fire source itself has not significantly deviated from the fire-free state, and may be in the very early stage of smoldering or only environmental fluctuations.

[0060] High-risk combustibles refer to building and repair materials in timber-framed buildings whose combustion spread rate or heat release rate exceeds a preset threshold, including tung oil, raw lacquer, and linseed oil. Non-high-risk combustibles refer to building timber in timber-framed buildings whose combustion spread rate and heat release rate are both below or equal to the preset threshold, including pine, fir, and elm. The preset combustion spread rate threshold and preset heat release rate threshold are pre-set based on the difference in combustion characteristic parameters between building timber and building repair materials in standard combustion tests. Specifically, under standard laboratory combustion test conditions, the combustion spread rate and heat release rate of various building timbers and building repair materials are measured separately, and the maximum combustion spread rate and maximum heat release rate of the building timber are used as the preset combustion spread rate threshold and preset heat release rate threshold, respectively. The classification of high-risk combustibles and non-high-risk combustibles is pre-defined based on the combustion characteristic parameters of building repair materials or building timber corresponding to each type of combustible material when the pre-stored characteristic extinction vector library is established. The combustion characteristic parameters are obtained from laboratory standard combustion tests.

[0061] The specific method of joint judgment is as follows: when the remaining spectral characteristic parameters fall within the preset fire hazard level range and the identified combustible material type corresponds to a high-risk combustible, a fire is confirmed and marked as a high fire hazard level; when the remaining spectral characteristic parameters fall within the preset fire hazard level range but the identified combustible material type corresponds to a non-high-risk combustible, a fire is confirmed and marked as a medium fire hazard level; when the remaining spectral characteristic parameters do not fall within the preset fire hazard level range but the identified combustible material type corresponds to a high-risk combustible, a suspected fire is output and marked as a high fire hazard level; when the remaining spectral characteristic parameters do not fall within the preset fire hazard level range and the identified combustible material type corresponds to a non-high-risk combustible, no fire is output. This joint judgment method combines the combustion hazard characteristics of the smoke source material with the radiation intensity of the fire source itself for classification, considering both the current activity level of the fire and its potential development hazard, thus solving the prominent problem in existing technologies where high-risk combustible fires are misjudged as low-risk due to smoke optical path interference.

[0062] Example 2: Figure 2 A schematic diagram of a large-scale fire detection and alarm device for wooden building complexes according to the present invention is provided. The device includes: The spectral data module is used to acquire multi-band spectral data within the monitoring area synchronously collected by distributed multi-band detection units. The multi-band spectral data covers the ultraviolet band to the far-infrared band. The data preprocessing module is used to preprocess multi-band spectral data and extract spectral feature parameters; The interference identification module is used to determine whether the symmetry of the full-band spectral energy distribution characterized by the spectral feature parameters is broken. If it is broken, it is identified as optical path interference. The extinction vector module is used to extract the distortion offset, which represents non-uniform attenuation distortion, from the spectral characteristic parameters when there is optical path interference, and to construct a full-band spectral extinction vector based on the distortion offset. The evolution analysis module is used to extract the spectral attenuation distribution profile of the full-band spectral extinction vector at the start and end of a preset time window, determine whether the asymmetric direction of the attenuation on the short-wavelength side and the attenuation on the long-wavelength side has reversed within the preset time window, determine the time-varying evolution mode of the full-band spectral extinction vector based on whether the reversal has occurred, and calibrate the evolution state of the full-band spectral extinction vector according to the time-varying evolution mode. The fire identification module is used to select a matching pre-stored feature extinction vector based on the evolution state calibration result, compare it with the full-band spectral extinction vector to identify the type of burning material, and combine the remaining spectral feature parameters after stripping the distortion offset to output the fire identification result including the fire hazard level.

[0063] All calculations involved in the embodiments are dimensionless numerical calculations, and the preset parameters and thresholds in the calculations are set by those skilled in the art according to the actual situation.

[0064] It should be noted that this invention can be deployed on the device itself to realize embedded applications, or it can run on a PC or other terminal with a user interface, thereby meeting various hardware environments and usage requirements.

[0065] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented, in whole or in part, as a computer program product. A computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, all or part of the processes or functions according to the embodiments of this application are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. Computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wireless or wired transmission; wired transmission methods include optical fiber, twisted pair, coaxial cable, etc.; wireless transmission includes infrared, microwave, etc. Computer-readable storage media can be any available medium that a computer can access or a data storage device such as a server or data center that contains one or more sets of available media. Available media can be magnetic media (e.g., floppy disks, hard disks, magnetic tapes), optical media (e.g., DVDs), or semiconductor media. Semiconductor media can be solid-state drives.

[0066] Those skilled in the art will understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and modules described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0067] In the several embodiments provided in this application, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of modules is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple modules or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between apparatuses or modules may be electrical, mechanical, or other forms.

[0068] The modules described as separate components may or may not be physically separate. The components shown as modules may or may not be physical modules; they may be located in one place or distributed across multiple network modules. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs.

[0069] In addition, the functional modules in the various embodiments of this application can be integrated into one processing module, or each module can exist physically separately, or two or more modules can be integrated into one module.

[0070] If a function is implemented as a software module and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0071] The above are merely specific embodiments of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

[0072] In conclusion, the above are merely preferred embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for large-scale fire detection and alarm in a wooden building complex, characterized in that, include: S1: Acquire multi-band spectral data within the monitoring area synchronously collected by distributed multi-band detection units, with the multi-band spectral data covering the ultraviolet to far-infrared bands; S2: Preprocess multi-band spectral data and extract spectral feature parameters; S3: Determine whether the symmetry of the full-band spectral energy distribution characterized by the spectral characteristic parameters is broken. If it is broken, it is identified as optical path interference. S4: When there is optical path interference, the distortion offset that characterizes non-uniform attenuation distortion is extracted from the spectral characteristic parameters, and a full-band spectral extinction vector is constructed based on the distortion offset. S5: Extract the spectral attenuation distribution contour of the full-band spectral extinction vector at the start and end of the preset time window, determine whether the asymmetric direction of the attenuation on the short-wavelength side and the attenuation on the long-wavelength side has reversed within the preset time window, determine the time-varying evolution mode of the full-band spectral extinction vector based on whether the reversal has occurred, and calibrate the evolution state of the full-band spectral extinction vector according to the time-varying evolution mode. S6: Select a matching pre-stored feature extinction vector based on the evolution state calibration result, compare it with the full-band spectral extinction vector to identify the type of burning material, and combine it with the remaining spectral feature parameters after stripping the distortion offset to output the fire identification result including the fire hazard level.

2. The method for large-scale fire detection and alarm of a wooden building complex according to claim 1, characterized in that, S1 includes: The multi-band detection unit can simultaneously acquire ultraviolet, visible, near-infrared, mid-infrared, and far-infrared spectral data of the same field of view within the monitoring area using a synchronous triggering method. Band registration is performed on the spectral data of different bands collected at the same time, so that the spectral data of different bands are aligned pixel by pixel in spatial position, forming registered multi-band spectral data covering the ultraviolet band to the far-infrared band.

3. The method for large-scale fire detection and alarm of a wooden building complex according to claim 1, characterized in that, S2 include: Noise removal is performed on the registered multi-band spectral data to obtain denoised multi-band spectral data. Environmental baseline correction was performed on the spectral response values ​​of each band in the denoised multi-band spectral data to eliminate the non-uniform background shift caused by water vapor absorption band and solar scattered light in each band, thus obtaining environmentally compensated multi-band spectral data. From the environmentally compensated multi-band spectral data, the radiative energy values ​​of each band at the preset characteristic spectral lines are extracted, and the spectral energy distribution sequence composed of the radiative energy values ​​of each band is used as the spectral characteristic parameter.

4. A method for large-scale fire detection and alarm in a wooden building complex according to claim 3, characterized in that, Environmental baseline correction is performed on the spectral response values ​​of each band in the denoised multi-band spectral data to eliminate the non-uniform background shifts caused by water vapor absorption bands and solar scattered light in each band, including: Under the condition that there is no fire in the monitoring area, the water vapor absorption spectrum response curve and solar scattering spectrum response curve of each band under clear sky conditions are obtained in advance. For the spectral response values ​​of each band in the denoised multi-band spectral data, the absorption component of the corresponding band on the water vapor absorption spectral response curve and the scattering component of the corresponding band on the solar scattering spectral response curve are subtracted band by band to obtain the environmentally compensated multi-band spectral data.

5. A method for large-scale fire detection and alarm of a wooden building complex according to claim 1, characterized in that, S3 include: Obtain the wavelength points corresponding to the peak positions in the spectral energy distribution sequence, and divide the full-band spectral energy distribution into short-wavelength side energy distribution curves and long-wavelength side energy distribution curves using the wavelength points as boundaries; By integrating the energy distribution curves on the shortwave side and the longwave side respectively, the integrated areas on the shortwave side and the longwave side are obtained. Calculate the ratio of the short-wavelength side integral area to the long-wavelength side integral area. If the ratio exceeds the preset symmetry tolerance range, it is determined that the symmetry of the full-band spectral energy distribution has been broken, and it is identified as the presence of optical path interference.

6. A method for large-scale fire detection and alarm in a wooden building complex according to claim 1, characterized in that, S4 include: Using the long-wavelength integrated area as a reference, the deviation of the ratio of the short-wavelength integrated area to the long-wavelength integrated area from the center value of the preset symmetry tolerance interval is taken as the distortion offset. The remaining spectral characteristic parameters after removing the distortion offset are obtained by subtracting the attenuation component of the distortion offset in each band from the spectral energy distribution sequence. The extinction components of each band are obtained by subtracting the spectral energy distribution sequence from the remaining spectral characteristic parameters after removing the distortion offset, and the extinction components of each band are used to form the full-band spectral extinction vector.

7. A method for large-scale fire detection and alarm in a wooden building complex according to claim 6, characterized in that, Subtract the attenuation component of the distortion offset in each band from the spectral energy distribution sequence, including: Obtain the preset attenuation coefficient curve of Mie scattering attenuation coefficient of flue gas particles as a function of wavelength; Multiply the distortion offset by the attenuation coefficient of each band on the attenuation coefficient curve to obtain the attenuation component of the distortion offset in each band. Subtract the attenuation component of the corresponding band from the radiant energy value of each band in the spectral energy distribution sequence.

8. A method for large-scale fire detection and alarm of a wooden building complex according to claim 1, characterized in that, S5 include: Extract the extinction components of each band in the full-band spectral extinction vector at the start of the preset time window, and arrange them in band order to form the starting spectral attenuation distribution profile; and extract the extinction components of each band in the full-band spectral extinction vector at the end of the preset time window, and arrange them in band order to form the ending spectral attenuation distribution profile. For the spectral attenuation distribution profile at the starting point, compare the sum of the extinction components from the ultraviolet band to the visible band with the sum of the extinction components from the mid-infrared band to the far-infrared band. If the sum of the extinction components from the ultraviolet band to the visible band is greater than the sum of the extinction components from the mid-infrared band to the far-infrared band, then mark the direction of the asymmetry at the starting point as the short-wavelength side attenuation being dominant, and vice versa as the long-wavelength side attenuation being dominant. Compare and mark the direction of asymmetry at the endpoint for the spectral attenuation distribution profile in the same manner. If the asymmetric direction at the starting point is different from the asymmetric direction at the ending point, it is determined that the asymmetric direction has reversed, the time-varying evolution mode of the full-band spectral extinction vector is determined to be the compositional evolution mode, and the full-band spectral extinction vector is calibrated as the compositional evolution state. If the asymmetric direction at the starting point is the same as the asymmetric direction at the ending point, it is determined that the asymmetric direction has not been reversed, the time-varying evolution mode of the full-band spectral extinction vector is determined to be the diffusion-dilution mode, and the full-band spectral extinction vector is calibrated as the diffusion-dilution state.

9. A method for large-scale fire detection and alarm of a wooden building complex according to claim 1, characterized in that, S6 include: Based on the evolution state calibration results, retrieve the first pre-stored feature extinction vector sub-library corresponding to the composition evolution state, or the second pre-stored feature extinction vector sub-library corresponding to the diffusion and dilution state, from the pre-stored feature extinction vector library. The full-band spectral extinction vector is compared with each of the pre-stored feature extinction vectors in the retrieved pre-stored feature extinction vector sub-library for similarity. The type of combustion substance corresponding to the pre-stored feature extinction vector with the highest similarity is selected as the type of combustion substance identified. The identified type of burning material is jointly judged with the remaining spectral feature parameters after stripping the distortion offset. When the remaining spectral feature parameters fall into the preset fire hazard level range and the type of burning material corresponds to a high-risk combustible, the fire situation is confirmed and marked as a high fire hazard level. When the remaining spectral characteristic parameters fall within the preset fire hazard level range but the type of burning material corresponds to a non-high-risk combustible, output confirmation of fire and mark it as a medium fire hazard level. When the remaining spectral characteristic parameters do not fall within the preset fire hazard level range, but the type of burning material corresponds to a high-risk combustible, a suspected fire is output and marked as a high fire hazard level. When the remaining spectral characteristic parameters do not fall within the preset fire hazard level range and the type of burning material corresponds to a non-high-risk combustible, the output is "No fire".

10. A large-scale fire detection and alarm device for a wooden building complex, used to implement the large-scale fire detection and alarm method for a wooden building complex as described in any one of claims 1-9, characterized in that, include: The spectral data module is used to acquire multi-band spectral data within the monitoring area synchronously collected by distributed multi-band detection units. The multi-band spectral data covers the ultraviolet band to the far-infrared band. The data preprocessing module is used to preprocess multi-band spectral data and extract spectral feature parameters; The interference identification module is used to determine whether the symmetry of the full-band spectral energy distribution characterized by the spectral feature parameters is broken. If it is broken, it is identified as optical path interference. The extinction vector module is used to extract the distortion offset, which represents non-uniform attenuation distortion, from the spectral characteristic parameters when there is optical path interference, and to construct a full-band spectral extinction vector based on the distortion offset. The evolution analysis module is used to extract the spectral attenuation distribution profile of the full-band spectral extinction vector at the start and end of a preset time window, determine whether the asymmetric direction of the attenuation on the short-wavelength side and the attenuation on the long-wavelength side has reversed within the preset time window, determine the time-varying evolution mode of the full-band spectral extinction vector based on whether the reversal has occurred, and calibrate the evolution state of the full-band spectral extinction vector according to the time-varying evolution mode. The fire identification module is used to select a matching pre-stored feature extinction vector based on the evolution state calibration result, compare it with the full-band spectral extinction vector to identify the type of burning material, and combine the remaining spectral feature parameters after stripping the distortion offset to output the fire identification result including the fire hazard level.

Images