A flame recognition method and flame detector based on multi-band infrared signal fusion

Abstract

This invention relates to the field of infrared flame detector technology, specifically disclosing a flame recognition method based on multi-band infrared signal fusion: S1, the main detection channel periodically acquires real-time main infrared signals, and three reference channels periodically acquire real-time reference infrared signals; S2, when the peak value of the real-time main infrared signal is greater than the fire threshold, spectral analysis is performed on the real-time main infrared signal to extract the peak main frequency, which is the frequency corresponding to the frequency band with the highest power spectral density; S3, it is determined whether the peak main frequency is within the warning frequency range. If so, spectral analysis is performed on the three real-time reference infrared signals acquired by the three reference channels to extract the peak main frequency of each real-time reference infrared signal; S4, it is determined whether at least one real-time reference infrared signal has a peak main frequency within the warning frequency range. If so, a fire is determined to have occurred. The above method can realize flame recognition in scenarios such as kitchens where artificial fire sources are used.

Description

Technical Field

[0001] This invention relates to the field of infrared flame detector technology, specifically to a flame identification method and flame detector based on multi-band infrared signal fusion. Background Technology

[0002] Liquefied petroleum gas (LPG) is a widely used fuel in the catering industry. However, due to improper management and other reasons, accidents such as LPG tank fires and explosions occur frequently. In order to improve fire safety in the catering industry and reduce fire risks during operation, it is necessary to conduct fire monitoring in areas where LPG is used to detect unexpected fires in a timely manner and reduce fire losses.

[0003] When a flame burns, the presence of carbon dioxide creates a peak in the 4.4μm infrared band, a distinctive characteristic of flames. Currently available point-type infrared flame detectors rely on this principle for flame identification. However, considering the presence of interfering heat sources in everyday environments, some of which also emit infrared radiation in the 4.4μm band and interfere with flame detection, some point-type infrared flame detectors employ dual-band or tri-band detection to simultaneously identify infrared light intensities of other wavelengths. The presence of interfering heat sources is then inferred based on the intensity of these other wavelengths.

[0004] CN115762042B discloses a detector that uses three narrow-band detection channels and one wide-band detection channel to identify flames. The detector determines the presence of a flame by comparing the ratio of the sum of the energy of the three narrow-band detection channels within a preset time window with a threshold value, and eliminates the influence of interfering heat sources.

[0005] However, the flame of a gas stove during normal use will also produce a peak in the 4.4μm infrared band. It is difficult to distinguish the normal use of a gas stove from the accidental combustion of other non-fuel substances by comparing the intensity of infrared light alone. Summary of the Invention

[0006] To address the technical problem of inaccurate identification of interfering heat sources in existing technologies, this application provides a flame identification method and a flame detector based on multi-band infrared signal fusion. The flame identification method based on multi-band infrared signal fusion includes the following steps:

[0007] S1. The main detection channel periodically acquires real-time main infrared signals, and the three reference channels periodically acquire real-time reference infrared signals.

[0008] S2. When the peak value of the real-time main infrared signal is greater than the fire threshold, perform spectrum analysis on the real-time main infrared signal to extract the peak main frequency, which is the frequency corresponding to the frequency band with the largest power spectral density.

[0009] S3. Determine whether the peak main frequency is within the warning frequency range. If so, perform spectrum analysis on the three real-time reference infrared signals and extract the peak main frequency of each real-time reference infrared signal.

[0010] By following the steps above, normal gas usage can be distinguished from accidental fires during flame detection, enabling the identification of accidental fires in environments with artificial flames, such as kitchens.

[0011] Specifically, the following steps are used to determine whether the peak value of the real-time main infrared signal is greater than the fire threshold:

[0012] S11. Preprocess the real-time main infrared signal, including bandpass filtering and baseline correction;

[0013] S12. Perform peak detection on the preprocessed real-time main infrared signal to obtain all peak values ​​within the detection period;

[0014] S13. Filter out peak values ​​that are greater than the fire threshold, calculate the proportion of the number of peak values ​​that are greater than the fire threshold to the total number of peak values. If the proportion exceeds the warning proportion, then determine that the peak value of the real-time main infrared signal is greater than the fire threshold.

[0015] By comparing peak ratios, the impact of interference factors such as sampling noise on the judgment results can be reduced, thereby improving the accuracy of the judgment.

[0016] Furthermore, when the peak frequency of the real-time main infrared signal is not within the fire warning range, the following steps are performed:

[0017] S31. Calculate the main frequency matching degree, kurtosis and harmonic intensity of the real-time main infrared signal;

[0018] S32. Perform Hilbert transform on the real-time main infrared signal to obtain the envelope signal of the real-time main infrared signal, and calculate the ratio of the standard deviation to the mean of the envelope signal.

[0019] S33. Construct a weighted decision model based on the main frequency matching degree, kurtosis, harmonic intensity, and the ratio of the standard deviation to the mean of the envelope signal;

[0020] S34. Calculate the fire matching degree through the weighted decision model. When the fire matching degree is greater than the first threshold, it is determined that a fire has occurred.

[0021] Envelope signal analysis can effectively extract the amplitude characteristics of flame flicker and quantify the amplitude. By combining it with the main frequency matching degree, kurtosis, and harmonic intensity, the characteristics of the acquired real-time main infrared signal are comprehensively evaluated, avoiding misjudgment caused by a single indicator and improving the accuracy of flame identification.

[0022] Furthermore, when the peak main frequencies of the three real-time reference infrared signals are all outside the warning frequency range, the following steps are performed:

[0023] S41. Find a matching combination of interfering heat source models based on the real-time main infrared signal and the three real-time reference infrared signals. The combination of interfering heat source models is composed of infrared feature models of multiple interfering heat sources. The infrared feature models are the infrared light intensity information radiated by a single interfering heat source measured by the main detection channel and the three reference channels.

[0024] S42. If there is a combination of interfering heat source models, no fire warning will be issued; if there is no such combination, a fire will be determined to have occurred.

[0025] By utilizing the principle of light intensity superposition, it can be determined whether the infrared light intensity in the current environment is composed of multiple interfering heat source models, thereby determining whether there is an accidental fire and improving the sensitivity of flame recognition.

[0026] Specifically, the following steps are used to find a matching combination of interfering heat source models:

[0027] S411. Construct a real-time infrared vector based on the real-time main infrared signal and the real-time reference infrared signal;

[0028] S412. Construct the feature vector of each interfering heat source for its infrared feature model;

[0029] S413. Construct a matching equation based on the real-time infrared vector and the feature vectors of multiple interference heat sources;

[0030] S414. Determine whether the matching equation has a non-zero real solution. If it does, then there is a matching combination of interfering heat source models.

[0031] Specifically, each component of the real-time infrared vector represents the proportion of the light intensity of the main detection channel and the three reference channels in the total light intensity, and the total light intensity is the sum of the light intensity collected by the main detection channel and the three reference channels.

[0032] Specifically, the feature vector of the interfering heat source is obtained through the following steps:

[0033] S355. Place an interfering heat source in a laboratory environment, wherein the laboratory environment is configured such that there are no other heat sources that can be detected by the main detection channel and the three reference channels except for the interfering heat source.

[0034] S356. The infrared light intensity signal radiated by the interfering heat source is collected using the main detection channel and three reference channels. The infrared light intensity signal is preprocessed and subjected to spectrum analysis to obtain an infrared feature model.

[0035] S357. Extract the peak main frequency of the main detection channel and the light intensity ratio of the main detection channel and the three reference channels from the infrared feature model to construct the feature vector of the interference heat source.

[0036] The present invention also provides a flame detector, employing the above-mentioned flame identification method, comprising: an infrared acquisition module and a data processing module. The infrared acquisition module includes a main detection channel and three reference channels. The main detection channel and the three reference channels are used to acquire real-time infrared light intensity data radiated by various heat sources in the environment. The data processing module is used to determine whether a fire has occurred based on the real-time multi-band infrared light intensity data acquired by the main detection channel and the three reference channels.

[0037] Specifically, the main detection channel collects infrared light intensity in the 4.4μm band, and the three reference channels collect infrared light intensity in the 5.3μm, 3.8μm, and 2.2μm bands, respectively.

[0038] The technical effects and advantages of this invention are as follows: By performing flame identification through the above method, performing spectrum analysis on infrared signals of four bands, and determining whether there is combustion of other non-fuel substances by using the peak main frequency of multiple bands, the normal use of the gas stove can be distinguished from accidental fires, thus realizing flame identification in scenarios such as kitchens where artificial fire sources are used. Attached Figure Description

[0039] Figure 1 This is a schematic diagram of the overall process of the flame recognition method of the present invention.

[0040] Figure 2 This is a flowchart illustrating the method of the present invention for determining the fire situation based on a weighted decision model.

[0041] Figure 3 This is a flowchart illustrating the process of finding a matching combination of interfering heat source models in the method of this invention. Detailed Implementation

[0042] 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.

[0043] Example 1

[0044] refer to Figure 1 Embodiment 1 of the present invention provides a flame identification method based on multi-band infrared signal fusion, which distinguishes flames generated by normal gas combustion from flames generated by accidental fires during the flame identification process, including the following steps:

[0045] S1. The main detection channel periodically acquires real-time main infrared signals, and the three reference channels periodically acquire real-time reference infrared signals.

[0046] S2. When the peak value of the real-time main infrared signal is greater than the fire threshold, perform spectrum analysis on the real-time main infrared signal and extract the peak main frequency. The peak main frequency is the frequency corresponding to the frequency band with the largest power spectral density.

[0047] S3. Determine whether the peak main frequency is within the warning frequency range. If so, perform spectrum analysis on the three real-time reference infrared signals obtained from the three reference channels and extract the peak main frequency of each real-time reference infrared signal.

[0048] S4. Determine whether at least one real-time reference infrared signal has a peak main frequency within the warning frequency range. If so, determine that a fire has occurred and issue a fire warning.

[0049] When a flame burns, it not only produces a peak in the 4.4μm wavelength band, but also often flickers due to factors such as eddies. The flicker frequency is typically between 1-30Hz, which distinguishes flames from other heat sources. Therefore, when the peak value of the real-time main infrared signal exceeds the fire threshold, it is necessary to determine whether the real-time main infrared signal flickers. If it does, it indicates the presence of a flame in the environment. Further investigation is needed to determine whether the flame is an accidental fire, in order to distinguish between normal gas usage and an accidental fire.

[0050] During normal gas usage, the gas is almost completely burned, resulting in a peak value in the 4.4μm wavelength band. Although high-temperature fumes or steam are often generated during gas use, the infrared peak value radiated by these fumes and steam is around 2.9μm, but without flickering, and the infrared fluctuations are relatively slow. In the event of an accidental fire, non-fuel materials often burn. Different materials emit different infrared light during combustion. Therefore, in addition to the peak value around 4.4μm, other wavelengths may also produce peak values ​​during an accidental fire. For example, the high-temperature carbon black produced when polyethylene melts emits an infrared peak value of 3.4μm, polyvinyl chloride emits an infrared peak value of around 5.7μm, and materials like cotton and linen emit an infrared peak value of around 2.9μm. Therefore, it is necessary to analyze the light intensity signals at other wavelengths to determine if flickering is also present. If flickering is also observed in the real-time infrared signals acquired in other reference channels, it indicates the combustion of other non-fuel materials, thus confirming an accidental fire.

[0051] Therefore, by following the steps above, normal gas usage can be distinguished from accidental fires during flame detection, enabling the identification of accidental fires in environments with artificial flames, such as kitchens.

[0052] Point-type infrared flame detectors sample infrared light using pyroelectric sensors. During the sampling process, due to fluctuations in the infrared signal itself and interference from noise in the circuit, the real-time infrared signal obtained is a fluctuating curve. Therefore, directly relying on the peak value in the curve for flame identification results in a large error. It is necessary to process the original infrared signal to obtain the peak value that can be used for flame identification.

[0053] Specifically, in step S1, the peak value of the real-time main infrared signal is determined to be greater than the fire threshold through the following steps:

[0054] S11. Preprocess the real-time main infrared signal, including bandpass filtering and baseline correction, to suppress high-frequency noise and low-frequency drift, and remove the DC component in the ambient thermal radiation, retaining only the variable part.

[0055] S12. Perform peak detection on the preprocessed real-time main infrared signal to obtain all peak values ​​within the detection period;

[0056] S13. Filter out peak values ​​that are greater than the fire threshold, calculate the proportion of the number of peak values ​​that are greater than the fire threshold to the total number of peak values. If the proportion exceeds the warning proportion, then it is determined that the peak value of the real-time main infrared signal is greater than the fire threshold.

[0057] Through experiments and practical applications, the ratio can be set to 60%, which can achieve a good balance between accuracy and real-time performance.

[0058] Specifically, as a preferred implementation, the main detection channel of the present invention collects infrared light intensity in the 4.4μm band, and the three reference channels collect infrared light intensity in the 5.3μm, 3.8μm and 2.2μm bands, respectively.

[0059] Considering that other interfering heat sources will also generate infrared radiation in the 4.4μm main band, and that there will likely be fluctuations of different frequencies when radiating infrared radiation (not flickering, and the fluctuation frequency is not between 1-30Hz), it may have a certain impact on the spectrum analysis results of the real-time main infrared signal, causing misjudgment of the fire situation.

[0060] refer to Figure 2 When the peak frequency of the real-time main infrared signal is not within the fire warning range, the following steps are executed:

[0061] S31. Calculate the main frequency matching degree, kurtosis and harmonic intensity of the real-time main infrared signal;

[0062] S32. Perform Hilbert transform on the real-time main infrared signal to obtain the envelope signal of the real-time main infrared signal, and calculate the ratio of the standard deviation to the mean of the envelope signal.

[0063] S33. Construct a weighted decision model based on the main frequency matching degree, kurtosis, harmonic intensity, and the ratio of the standard deviation to the mean of the envelope signal;

[0064] S34. Calculate the fire matching degree through a weighted decision model. When the fire matching degree is greater than the first threshold, it is determined that a fire has occurred.

[0065] Envelope signal analysis can effectively extract the amplitude characteristics of flame flicker and quantify the amplitude. By combining it with the main frequency matching degree, kurtosis, and harmonic intensity, the characteristics of the acquired real-time main infrared signal are comprehensively evaluated, avoiding misjudgment caused by a single indicator and improving the accuracy of flame identification.

[0066] As a preferred implementation method, the clock frequency matching degree can be obtained through the following steps:

[0067] S311. Perform spectrum analysis on the real-time main infrared signal to determine the peak main frequency, which can be done according to the following formula:

[0068]

[0069] In the formula, N is the number of sampling points, X[n] is the number of envelope signal sampling points, which is obtained by performing Hilbert transform on the real-time main infrared signal, W[n] is the Hanning window function, P[k] is the power spectral density at frequency index k, and j is the imaginary unit;

[0070] Peak main frequency f dom Calculated using the following formula:

[0071]

[0072] That is, the frequency at which the power spectral density is at its maximum is the peak dominant frequency f. dom Ω represents the characteristic frequency band of the flame, which can generally be defined as [1Hz, 30Hz]. This means searching for possible flame flicker frequency components within this range and excluding interference signals. Similarly, the peak main frequency of the real-time reference infrared signal can also be obtained through this step.

[0073] S312. Calculate the frequency matching degree based on the peak frequency. Specifically, it can be calculated using the following formula:

[0074]

[0075] In the formula, [f L f H [ ] represents the target frequency band, i.e., the range of scintillation frequencies during ideal flame combustion, typically taken as f.L =5Hz, f H =15Hz, where S is the Gaussian attenuation scale parameter, which can be set to 5Hz in most application scenarios.

[0076] Specifically, the kurtosis of the real-time main infrared signal can be calculated using the following formula:

[0077]

[0078] In the formula, X is the envelope signal, μ is the mean of the envelope signal, σ is the standard deviation of the envelope signal, and E[(X-μ)] is the standard deviation of the envelope signal. 4 [x-μ] 4 The expected value, the ratio C between the two v The calculation method is as follows:

[0079]

[0080] In practical applications, light intensity signals are often acquired through discrete sampling, therefore the kurtosis K... S The calculation can be performed using the following formula:

[0081]

[0082] In the formula, Let N be the mean of the envelope signal, and N be the number of sampling points.

[0083] Generally, the light intensity that the detector can receive decreases as the distance between the heat source and the detector increases. To eliminate the influence of distance on the judgment result, the kurtosis is normalized to obtain a unit kurtosis K. F :

[0084]

[0085] K L The lower limit of typical flame kurtosis is generally taken as K. L =3,K H The upper limit of typical flame kurtosis is generally taken as K. H =8.

[0086] Specifically, harmonic intensity H S Calculated using the following formula:

[0087]

[0088] In the formula, ∈ is a small constant to prevent division by zero, and H2 is the second harmonic of the main frequency (i.e., 2×f). dom The harmonic intensity of the component, H3 (i.e., 3×f) dom The harmonic intensity of the third harmonic component of the fundamental frequency, where F is the fundamental frequency energy, is calculated using the following formula:

[0089]

[0090] In the formula, f s Δf is the sampling frequency, and Δf is the search bandwidth.

[0091] The harmonic intensities of the second and third harmonic components are calculated using the following formula:

[0092]

[0093] In the formula, when m=2, it is the second harmonic component, and when m=3, it is the third harmonic component.

[0094] Based on the above analysis, the weighted decision model can be constructed in the following form:

[0095] S F =αC v +βM f +γK F +δH s

[0096] In the formula, α, β, γ, and δ are the weight coefficients of each item. Generally, α = 0.4, β = 0.3, γ = 0.2, and δ = 0.1 are taken.

[0097] S F This refers to the fire situation matching degree. Generally, when the fire situation matching degree S... F A fire can be identified when the value is greater than 0.65 (the first threshold).

[0098] In step S4, determining whether a fire has occurred by checking if at least one real-time reference infrared signal's peak frequency falls within the warning frequency range carries a risk of misjudgment. In some special cases (such as the burning of tiny flames), only the real-time main infrared signal can detect obvious flickering characteristics; other bands cannot effectively identify clear flickering features. Therefore, further flame identification is required.

[0099] Specifically, refer to Figure 3 When the peak main frequencies of the three real-time reference infrared signals are all outside the warning frequency range, the following steps are executed:

[0100] S41. Find a matching combination of interfering heat source models based on the real-time main infrared signal and three real-time reference infrared signals. The combination of interfering heat source models consists of infrared feature models of multiple interfering heat sources. The infrared feature models are the infrared light intensity information radiated by a single interfering heat source measured by the main detection channel and the three reference channels.

[0101] S42. If there is a combination of interfering heat source models, no fire warning will be issued; if there is no such combination, a fire will be determined to have occurred.

[0102] The increase in light intensity conforms to the principle of linear superposition; the light intensities generated by each light source are linearly superimposed in the sensor and do not affect each other. Since the infrared signals radiated by different interfering heat sources are distributed differently in the four bands, when identifying flames, an infrared feature model can be established separately for the infrared signals generated by each common interfering heat source. When there is no unexpected fire, the real-time light intensity distribution should be obtainable by linearly combining the infrared feature models of multiple interfering heat sources.

[0103] Specifically, refer to Figure 3 Finding a matching combination of interfering heat source models can be done using vector operations:

[0104] S411. Construct a real-time infrared vector based on the real-time main infrared signal and the real-time reference infrared signal;

[0105] S412. Construct the feature vector of each interfering heat source for its infrared feature model;

[0106] S413. Construct a matching equation based on the real-time infrared vector and the feature vectors of multiple interfering heat sources;

[0107] S414. Determine whether the matching equation has a non-zero real solution. If it does, then there is a matching combination of interfering heat source models.

[0108] Specifically, considering that the infrared signals radiated by different interfering heat sources have different phases, and that the effect of phase difference on the final light intensity should also be considered when superimposing light intensities, the following method can be used to construct the real-time infrared vector:

[0109] F S =(a1) min a1 max a2 min a2 max a3 min a3 max a4 min a4 max f1)

[0110] Among them, a1 min a1 max Let f1 and a2 represent the minimum and maximum proportions of the real-time main infrared signal intensity in the total light intensity collected from all channels, respectively. Let f1 represent the peak main frequency of the real-time main infrared signal. min a2 max a3 min a3 max a4 min a4 max These represent the maximum and minimum percentages of the light intensity of the three reference channels in the total light intensity collected from all channels, respectively.

[0111] The construction format of the interference heat source feature vector and the data represented by each component should be consistent with the real-time infrared vector. In practical applications, interference heat sources can include ovens, induction cookers, lighting fixtures, sunlight, people, gas stoves, etc. Each interference heat source can establish the aforementioned interference heat source feature vector. Based on the above heat source feature vectors, an interference heat source model combination M can be constructed. S :

[0112]

[0113] In the formula, F n Let φ be the feature vector of the nth interfering heat source. n This represents the combination coefficient corresponding to the interfering heat source.

[0114] Based on the above combination of interference heat source models and real-time infrared vectors, the following matching equation can be constructed:

[0115] F S -M S =0

[0116] When there exists a matching combination of interfering heat source models, that is, when the homogeneous linear equation system has non-zero real solutions.

[0117] Specifically, the feature vector of the interfering heat source is obtained through the following steps:

[0118] S355. Place an interfering heat source in a laboratory environment where there are no other heat sources that can be detected by the main detection channel and the three reference channels except for the interfering heat source.

[0119] S356. The infrared light intensity signal radiated by the interfering heat source is collected using the main detection channel and three reference channels. The infrared light intensity signal is preprocessed and subjected to spectrum analysis to obtain an infrared feature model.

[0120] S357. Extract the peak main frequency of the main detection channel and the light intensity ratio of the main detection channel and the three reference channels from the infrared feature model to construct the feature vector of the interference heat source.

[0121] Flame identification is performed using the above method. The spectrum of infrared signals in four bands is analyzed, and the peak main frequency of multiple bands is used to determine whether there is combustion of other non-fuel substances. This distinguishes normal use of gas stoves from accidental fires, enabling flame identification in scenarios such as kitchens where artificial fire sources are used.

[0122] Example 2

[0123] Based on the flame recognition method provided in Embodiment 1, the present invention also provides a flame detector, including: an infrared acquisition module and a data processing module;

[0124] The infrared acquisition module includes a main detection channel and three reference channels. The main detection channel and the three reference channels are used to collect real-time infrared light intensity data radiated by various heat sources in the environment. The data processing module is used to determine whether a fire has occurred based on the real-time multi-band infrared light intensity data collected by the main detection channel and the three reference channels.

[0125] Specifically, the main detection channel collects infrared light intensity in the 4.4μm band, while the three reference channels collect infrared light intensity in the 5.3μm, 3.8μm, and 2.2μm bands, respectively.

[0126] In conclusion, the above description is only a preferred embodiment of the present invention and is 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 flame recognition method based on multi-band infrared signal fusion, characterized in that, The method comprises the following steps: S1, periodically acquiring real-time main infrared signals by a main detection channel, and periodically acquiring real-time reference infrared signals by three reference channels; S2, when a peak value of the real-time main infrared signals is greater than a fire threshold, performing spectral analysis on the real-time main infrared signals to extract a peak main frequency, the peak main frequency being a frequency corresponding to a frequency band with the maximum power spectral density; S3, judging whether the peak main frequency is within a pre-warning frequency interval, if yes, performing spectral analysis on the three real-time reference infrared signals to extract a peak main frequency of each of the real-time reference infrared signals; S4, judging whether at least one peak main frequency of the real-time reference infrared signals is within the pre-warning frequency interval, if yes, determining that a fire occurs, otherwise, searching for a matched interference heat source model combination according to the real-time main infrared signals and the three real-time reference infrared signals, the interference heat source model combination being composed of infrared characteristic models of multiple interference heat sources, and the infrared characteristic model being infrared intensity information radiated by a single interference heat source measured by the main detection channel and the three reference channels; The matched interference heat source model combination is searched for through the following steps: S411, constructing a real-time infrared vector according to the real-time main infrared signals and the real-time reference infrared signals, each component of the real-time infrared vector representing a proportion of light intensity of the main detection channel and the three reference channels in total light intensity, the total light intensity being a sum of light intensity collected by the main detection channel and the three reference channels; S412, constructing an interference heat source characteristic vector for each infrared characteristic model of the interference heat source; S413, constructing a matching equation according to the real-time infrared vector and the multiple interference heat source characteristic vectors; S414, judging whether the matching equation has a non-zero real solution, if yes, there is a matched interference heat source model combination.

2. The method of claim 1, wherein, The peak value of the real-time main infrared signals is judged through the following steps: S11, pre-processing the real-time main infrared signals, including band-pass filtering and baseline correction; S12, performing peak value detection on the pre-processed real-time main infrared signals to obtain all peak values in a detection period; S13, screening peak values greater than the fire threshold, calculating a proportion of the number of peak values greater than the fire threshold in the total number of peak values, and if the proportion exceeds a pre-warning proportion, determining that the peak value of the real-time main infrared signals is greater than the fire threshold.

3. The method of claim 1, wherein, When the peak main frequency of the real-time main infrared signals is not within the fire pre-warning interval, the following steps are performed: S31, calculating a main frequency matching degree, a kurtosis, and a harmonic intensity of the real-time main infrared signals; S32, performing Hilbert transform on the real-time main infrared signals to obtain an envelope signal of the real-time main infrared signals, and calculating a ratio of a standard deviation to a mean value of the envelope signal; S33, constructing a weighted decision model according to the main frequency matching degree, the kurtosis, the harmonic intensity, and the ratio of the standard deviation to the mean value of the envelope signal; S34, calculating a fire matching degree through the weighted decision model, and when the fire matching degree is greater than a first threshold, determining that a fire occurs.

4. The method of claim 1, wherein, The interference heat source characteristic vector is obtained through the following steps: S355, placing an interference heat source in a laboratory environment, the laboratory environment being configured to have no heat source other than the interference heat source that can be detected by the main detection channel and the three reference channels; S356, collecting infrared light intensity signals radiated by the interference heat source by using the main detection channel and the three reference channels, pre-processing and spectrum analyzing the infrared light intensity signals to obtain an infrared characteristic model; S357, extracting a peak main frequency of the main detection channel and light intensity proportions of the main detection channel and the three reference channels from the infrared characteristic model to construct an interference heat source characteristic vector.

5. A flame detector employing the flame recognition method according to any one of claims 1 to 4, characterized in that Comprise: an infrared ray collecting module and a data processing module; the infrared ray collecting module comprises a main detection channel and three reference channels, the main detection channel and the three reference channels being used to collect real-time infrared light intensity data radiated by various heat sources in an environment; the data processing module is used to determine whether a fire occurs according to the real-time multi-band infrared light intensity data collected by the main detection channel and the three reference channels.

6. The flame detector of claim 5, wherein, the main detection channel collects infrared light intensity of a 4.4 μm band, and the three reference channels collect infrared light intensity of three bands of 5.3 μm, 3.8 μm and 2.2 μm respectively.

Images