Method and apparatus for detecting smoke and fire based on particle shape characteristics

Abstract

To provide a smoke detection fire detection method and a smoke detection fire detector based on particle shape characteristics for detecting particles in an optical darkroom.SOLUTION: In the optical darkroom, an optical path angle between one of the blue light and the infrared light and the photoelectric conversion module is an acute angle, and an optical path angle between the other blue light and the photoelectric conversion module is an obtuse angle. Starting a light source combination, calculating a change value between a current optical power of scattered light of each light source and a background value, continuously starting the light source combination when the change value exceeds a set threshold, recording obtained change values and constructing a space vector, classifying a plurality of space vectors by using a classifier, to obtain a plurality of classification results, categories of the space vectors including fire smoke and interfering aerosol, and performing statistics on the plurality of classification results, to obtain a correct classification result.SELECTED DRAWING: Figure 2

Description

[Technical Field]

[0001] The present invention relates to a technical field related to fire monitoring, and more particularly to a smoke detection fire detection method and a smoke detection fire detector based on particle shape characteristics. [Background technology]

[0002] Fire smoke is an aerosol produced by the burning of combustible materials. It is not only a product of fire, but is also considered a precursor to fire and is the most noticeable feature in the early stages of a fire. Smoke detectors are the primary fire detector for early detection of fires. Currently, the most widely used type is the spot-type photoelectric smoke detector, whose main advantages are simple structure, low cost, easy maintenance, and no environmental pollution.

[0003] The basic structure of a photoelectric smoke detector is that smoke enters an optical darkroom that blocks surrounding light, inside which a light source and a photoelectric conversion device are installed. This type of detector primarily utilizes the principle of optical scattering by particulate matter. When there is no fire smoke in the detection chamber, the detector remains in normal monitoring mode and does not sound an alarm. When smoke generated by combustion enters the detection chamber, the light scattering effect of the smoke particles increases the scattered light signal collected by the photoelectric conversion device, which gradually increases as the smoke concentration increases. When this signal exceeds a certain threshold, the detector determines it is a fire and generates an alarm signal.

[0004] Conventional smoke-sensing fire detectors are susceptible to interference from non-fire aerosols such as oil smoke, dust, and water vapor. Smoke detectors used in homes are particularly susceptible to interference from oil smoke in the kitchen, causing false alarms. Summary of the Invention [Problem to be solved by the invention]

[0005] In consideration of the above-mentioned shortcomings or improvement requests of the prior art, the present invention provides a smoke detection and fire detection method and a smoke detection and fire detector based on particle shape characteristics, the purpose of which is to more accurately distinguish between interfering aerosols and fire smoke, reduce the false alarm rate of the smoke detection and fire detector, and improve the accuracy of fire alarms. [Means for solving the problem]

[0006] To achieve the above object, one aspect of the present invention provides a smoke and fire detection method based on particle shape characteristics, wherein particles are detected in an optical darkroom, the optical darkroom including a light source combination and a photoelectric conversion module D, the light source combination including a light source B1, a light source I1, and a light source B2, the light sources B1 and B2 are both blue light sources, and the light source I1 is an infrared light source, the center of the optical darkroom is O, the light from the light source B1 is reflected from the center O to the photoelectric conversion module D at an acute angle, the light from the light source I1 is reflected from the center O to the photoelectric conversion module D at an acute angle, and the light from the light source B2 is reflected from the center O to the photoelectric conversion module D at an obtuse angle.

[0007] The sensing method comprises: Step S1: obtaining a background value of the optical power of the scattered light of each light source, the background value of each light source being the optical power of the scattered light received by the photoelectric conversion module from the corresponding light source in an environment without fire smoke or interfering aerosols; Step S2: periodically activate the light source combination to sequentially send light pulse signals, calculate the change value of each light source, determine whether there is a change value exceeding the set threshold, if there is not, repeat step S2, if there is, it is considered that there is an abnormality, and execute step S3, the change value of each light source is the absolute value of the difference between the optical power of the scattered light currently received by the photoelectric conversion module D from the light source and the background value of the optical power of the scattered light of the light source; Step S3: if there is an abnormality, activate the light source combination n times consecutively and record the obtained change value, where n is a positive integer; Step S4: constructing a space vector based on the three change values ​​of the light source combinations at each startup, and using a classifier to classify the n space vectors respectively to obtain n classification results, where the classifier is a trained neural network model, and the classification categories include fire smoke and nuisance aerosols. If the number of one category among the n classification results obtained by statistics is k or more, the category is determined to be the correct classification result, where k>n / 2; Includes.

[0008] Furthermore, in step S4, if there are no categories with a number of k or more in the statistical results of multiple classifications, the process proceeds to step S3, where the change values ​​are acquired again and reclassification is performed until it is determined that the classification results are correct.

[0009] Furthermore, n is 10 or greater and k is 8 or greater.

[0010] Furthermore, the classification categories include white fire smoke, black fire smoke, and nuisance aerosols, and the nuisance aerosols include one or more of dust, water vapor, or oil smoke.

[0011] Furthermore, obtaining the background value of the optical power of the scattered light of each light source in step S1 includes performing initialization, sequentially activating each light source in an environment free of fire smoke or interfering aerosols, and obtaining the optical power of the scattered light received by the photoelectric conversion module from the corresponding light source as the background value of the optical power of the scattered light of the corresponding light source.

[0012] Furthermore, the light source combination is activated to transmit a light pulse signal at a period of one or several seconds, and the pulse width of the light pulse signal ranges from several milliseconds to several tens of milliseconds.

[0013] Furthermore, the classifier includes multiple support vector machine (SVM) binary classification models, each of which is used to realize classification of two different categories. When classifying a space vector using the classifier, each binary classification model performs binary classification on the input space vector, and the most frequent classification result of all the binary classification models is the classification result of the vector by the classifier.

[0014] According to another aspect of the present invention, there is provided a particle shape signature based smoke sensing fire detector that includes an optical darkroom and a control module. The optical darkroom includes a light source combination and a photoelectric conversion module D, the light source combination includes a light source B1, a light source I1, and a light source B2, the light source B1 and the light source B2 are both blue light sources, the light source I1 is an infrared light source, the center of the optical darkroom is O, the light path change angle at which light from the light source B1 is reflected through the center O to the photoelectric conversion module D is an acute angle, the light path change angle at which light from the light source I1 is reflected through the center O to the photoelectric conversion module D is an acute angle, and the light path change angle at which light from the light source B2 is reflected through the center O to the photoelectric conversion module D is an obtuse angle. The control module is used to implement any of the methods described above.

[0015] Furthermore, the acute angle is equal to or greater than 20°, and the obtuse angle is equal to or less than 170°.

[0016] Furthermore, the two acute angles are the same or differ by within 10°. [Effects of the Invention]

[0017] Compared with the prior art, the above technical means of the present invention mainly have the following advantageous effects: 1. In smoke detection environments, there are differences in shape between interfering aerosols and fire smoke. Smoke particles generally have chain-like shape characteristics, while interfering aerosols such as dust, water vapor, and oil smoke are generally spherical. These differences in particle shape manifest as unique fingerprints of particulate matter in light scattering effects. Fire smoke and interfering aerosols can be classified based on their particle shapes, thereby reducing the false alarm rate of conventional smoke and fire detectors. Through research, the applicant discovered that in the design of photoelectric smoke and fire detectors, the scattering characteristics of light sources with the same wavelength but different observation angles can be used to distinguish the shapes of particulate matter. However, this research also found that large interfering aerosol particles affect the scattered light intensity, causing certain deviations in detection. To reduce false alarms when interfering aerosol particles are large, the research and development team conducted further research and found that adding light sources with different wavelengths can correct the deviations caused by particle size. Therefore, in this invention, three light sources are designed, two blue light sources, and a red light source is added to correct the data. Using three light sources allows for the reception of three different types of data, which can then be used to extend the aerosol classification curve into a classification surface. A vector composed of the three types of data is input into a classifier, and a machine learning algorithm is used to determine the coefficients of the classification surface equation, resulting in more accurate classification results. Furthermore, the present invention acquires multiple data sets to obtain multiple classification results, thereby using statistical techniques to improve the accuracy of the classification results. In other words, the present invention can correct the sensing data based on particle shape characteristics, enabling classification of fire smoke and nuisance aerosols, thereby reducing the false alarm rate of smoke and fire detectors and enabling appropriate countermeasures to be taken. 2. This invention distinguishes between fire smoke and nuisance aerosols based on differences in shape. As long as there are differences in shape, this invention can be used to classify and identify them, allowing fire smoke to be distinguished from traditional nuisance aerosols such as dust, water vapor, and oil smoke. Compared to classification based on particle size comparison, this invention has a wider range of application and higher classification accuracy. 3. Furthermore, fire smoke is further subdivided into white fire smoke and black fire smoke, and the classification categories include white fire smoke, black fire smoke, and nuisance aerosols. By subdividing the classification, more detailed classification results can be obtained and timely responses can be made. 4. Furthermore, by selecting the angle range, the shape information characteristics of aerosols can be detected more accurately by the scattered light intensity at different angles. More specifically, the angles of the two acute angles are the same, which simplifies the design of the detector. [Brief explanation of the drawings]

[0018] [Figure 1] FIG. 1 is a schematic diagram illustrating the configuration of an optical darkroom according to an embodiment of the present invention. [Figure 2] 1 is a flowchart of steps of a particle shape characteristic based smoke and fire detection method in accordance with one embodiment of the present invention. [Figure 3] 4 is a flowchart of steps of a particle shape characteristic based smoke and fire detection method in accordance with another embodiment of the present invention. [Figure 4] FIG. 10 is a statistical diagram of classification results in one embodiment of the present invention. DETAILED DESCRIPTION OF THE INVENTION

[0019] In order to clarify the objectives, technical means and advantages of the present invention, the present invention will be described in more detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention. In addition, the technical features according to various embodiments of the present invention described below can be combined with each other unless they are mutually inconsistent.

[0020] The present invention relates to a smoke and fire detection method based on particle shape characteristics, which is used in combination with an optical darkroom to achieve special results. Figure 1 shows a schematic diagram of the optical darkroom in one embodiment of the present invention. The optical darkroom includes a light source combination and a photoelectric conversion module D. The light source combination includes light source B1, light source I1, and light source B2. Light source B1 and light source B2 are both blue light sources, and light source I1 is an infrared light source. The center of the optical darkroom is O. The light from light source B1 is reflected from the center O to the photoelectric conversion module D at an acute angle. The light from light source I1 is reflected from the center O to the photoelectric conversion module D at an acute angle. The light from light source B2 is reflected from the center O to the photoelectric conversion module D at an obtuse angle.

[0021] Specifically, light source B1 is an LED light source capable of emitting 450 nm blue light, light source I1 is an LED light source capable of emitting 950 nm infrared light, and light source B2 is an LED light source capable of emitting 450 nm blue light. The photoelectric conversion module D is a photoelectric conversion device capable of converting the optical functional signal of the received scattered light into an electrical signal, and may be specifically a photodiode. The optical path change angle of the light beam from B1 to D via O is an acute angle, generally 20° or more, specifically 40°. The optical path change angle of the light beam from I1 to D via O is an acute angle, generally 20° or more, specifically 40°. The optical path change angle of the light beam from B2 to D via O is an obtuse angle, generally 170° or less, specifically 140°. Selecting the above angle range allows for more accurate detection of aerosol shape information characteristics based on the scattered light intensity at different angles. More specifically, having the two acute angles be the same simplifies the design of the sensor.

[0022] 2 shows a flowchart of steps of a smoke and fire detection method based on particle shape characteristics in one embodiment of the present invention, and includes at least steps S1 to S4. Each step will be described in detail below.

[0023] Step S1: Obtain the background value of the optical power of the scattered light of each light source. The background value of each light source is the optical power of the scattered light received by the photoelectric conversion module from the corresponding light source in an environment without fire smoke or interfering aerosols.

[0024] Specifically, initialization is performed at the start of detection, i.e., in an environment free of fire smoke or interfering aerosols, each light source is activated sequentially, and the optical power of the scattered light received by the photoelectric conversion module from the corresponding light source is obtained as a background value of the optical power of the scattered light of the corresponding light source.

[0025] In a specific operation, light source B1, light source I1, and light source B2 are turned on in sequence, and the three light sources sequentially transmit optical pulse signals, with a period of 1 second and a pulse width of several milliseconds to several tens of milliseconds. After each light source emits light, the photoelectric conversion module records the optical power of the received scattered light, and the photoelectric conversion module converts the optical power signal into a voltage signal, and the AD sampling value is the background value of the optical power signal of each scattered light. It is written as JPEG2026025834000002.jpg732.

[0026] Step S2: Periodically activate the light source combination to sequentially transmit light pulse signals, calculate the change value of each light source, and determine whether there is a change value exceeding the set threshold value. If there is not, repeat step S2; if there is, it is considered that there is an abnormality, and execute step S3.

[0027] The change value of each light source is the absolute value of the difference between the optical power of the scattered light currently received by the photoelectric conversion module D from the light source and the background value of the optical power of the scattered light from the light source.

[0028] Specifically, the interval time For each JPEG2026025834000003.jpg55, light source B1, light source I1, and light source B2 are turned on in sequence, and the three light sources sequentially transmit optical pulse signals. The photoelectric conversion module subtracts the corresponding background value from the optical power of the scattered light, and the corresponding optical power change value of the scattered light is calculated. We get JPEG2026025834000004.jpg739. The purpose of this step is to determine if there are any anomalies in the current environment. JPEG2026025834000005.jpg520, or JPEG2026025834000006.jpg621, or If JPEG2026025834000007.jpg622, then determine that there is an anomaly in the current environment and perform the following steps to further identify the type of particle causing the anomaly. If not, repeat this procedure. TH is a preset threshold value selected based on experiments, typically 300-500.

[0029] Step S3: If there is an abnormality, activate the light source combination n times in succession and record the obtained change value, where n is a positive integer.

[0030] Specifically, if there is an abnormality in the current environment, a series of change values JPEG2026025834000008.jpg538 starts to be recorded continuously, and the change value of light source B1 is It is written as JPEG2026025834000009.jpg768, JPEG2026025834000010.jpg611 represents the change value due to the activation of light source B1 for the i-th time, and the change value of light source I1 is It is written as JPEG2026025834000011.jpg764, JPEG2026025834000012.jpg710 represents the change value due to the activation of light source I1 for the i-th time, and the change value of light source B2 is It is written as JPEG2026025834000013.jpg867, JPEG2026025834000014.jpg711 represents the change value due to the activation of light source B2 for the i-th time. Specifically, n can be 10, that is, 10 sets of data are collected for category determination.

[0031] Step S4: A spatial vector consisting of the three change values ​​of the light source combination at each startup is obtained, and the multiple spatial vectors are classified using a classifier to obtain multiple classification results. The classifier is a trained neural network model, and the classification categories include fire smoke and nuisance aerosols. If the number of categories among the multiple classification results obtained through statistics is k or more, that category is determined to be the correct classification result, where k>n / 2.

[0032] Specifically, the change value obtained in step S3 is converted into the space vector JPEG2026025834000015.jpg8132JPEG2026025834000016.jpg755 is obtained. Using a classifier to predict each spatial vector individually, multiple spatial vectors can be predicted, resulting in multiple classification results. If the same classification result appears k or more times out of n prediction results, it means that the probability of that classification result is relatively high, and that classification result is considered to be the correct classification result. Specifically, n can be 10 and k can be 8.

[0033] Specifically, white smoke and black smoke have certain morphological differences. White fire smoke is in the form of short chains, while black fire smoke is in the form of long chains. Therefore, fire smoke can be further subdivided into white fire smoke and black fire smoke. That is, the classification categories include white fire smoke, black fire smoke, and nuisance aerosols. If the classification result is white fire smoke, the fire detector will emit an alarm signal indicating that the type of fire smoke is white smoke. If the classification result is black fire smoke, the fire detector will emit an alarm signal indicating that the type of fire smoke is black smoke. If the classification result is nuisance aerosols, the fire detector will emit a nuisance signal.

[0034] The classifier is a pre-trained model, which is trained using historical data before being used for classification. Specifically, the classifier may directly use a multi-classification model or may use multiple binary classification models.

[0035] For example, if the categories are white fire smoke, dark fire smoke, and nuisance aerosols, a multi-classifier model can directly distinguish between white fire smoke, dark fire smoke, and nuisance aerosols. If the model only supports binary classification problems, the fire detector must distinguish between the three types: white fire smoke, dark fire smoke, and nuisance aerosols. Therefore, three binary classifiers are required. Three support vector machine (SVM) binary classification models (SVMModel1, SVMModel2, and SVMModel3) can be used. SVMModel1 is used to classify nuisance aerosols and white fire smoke, SVMModel2 is used to classify white fire smoke and dark fire smoke, and SVMModel3 is used to classify nuisance aerosols and dark fire smoke. Each vector is predicted using the three SVM classifiers, and the most frequent value of the prediction results of the three classifiers is used as the classification result for the corresponding spatial vector. In this embodiment, the SVM binary classification model can map the spatial vectors to a higher-dimensional space, improving the accuracy of the classification results.

[0036] In one embodiment, if there are no categories with a number of k or more in the statistical results of multiple classifications in step S4, proceed to step S3, re-acquire the change value, and re-classify until it is determined that the classification result is correct.

[0037] FIG. 3 shows a flowchart of steps of a smoke detection and fire detection method based on particle shape characteristics in another embodiment of the present invention. Specifically, Initialize the system and measure background values. Record JPEG2026025834000017.jpg732, The three light sources transmit optical pulse signals sequentially, and the photoelectric conversion module subtracts the corresponding background value from the optical power of the scattered light to obtain the corresponding change in the optical power of the scattered light. To obtain JPEG2026025834000018.jpg739, JPEG2026025834000019.jpg520, or JPEG2026025834000020.jpg621, or determining whether JPEG2026025834000021.jpg622 exists, and if not, repeating the steps above; if present, performing the following steps; A series of JPEG2026025834000022.jpg538 starts to be recorded continuously, JPEG2026025834000023.jpg7148JPEG2026025834000024.jpg767, Loading the model parameters of three pre-trained SVM binary classifiers, SVMModel1, SVMModel2, and SVMModel3; Use an SVM classifier to predict and classify, determine whether the prediction result has k or more classification results, if so, determine the classification result, and issue an alarm or disturbance signal according to the classification result, if not, try a series of attempts until the correct classification is achieved. Returning to the step of continuously recording JPEG2026025834000025.jpg538; Includes.

[0038] Figure 4 shows a statistical diagram of classification results in one embodiment of the present invention. In the diagram, circles represent white smoke from fires, triangles represent black smoke from fires, and plus signs represent interfering aerosols. This diagram allows intuitive counting of which classification results are above k.

[0039] In a smoke detection environment, there are differences in shape between interference aerosols and fire smoke. Smoke particle formation is a dynamic process. During the combustion process, many nanometer-sized primary particles are formed. These primary particles collide and combine through agglomeration, forming larger smoke particle aggregates. Smoke particle aggregates have obvious fractal characteristics, irregular shapes, and loose structures. They generally exhibit obvious non-spherical characteristics. In particular, polyurethane and Huangshan tobacco particle aggregates exhibit long-chain morphology. Interference aerosols are typically generated by physical reactions and are generally spherical in shape. Therefore, there are differences in particle shape between fire smoke and interference aerosols. These particle shape differences manifest as unique fingerprints of particulate matter in the light scattering effect. Fire smoke and interference aerosols can be distinguished by their particle shape differences, thereby reducing the false alarm rate of conventional smoke detectors. Through research, the applicant discovered that in the design of photoelectric smoke detectors, the scattering characteristics of light sources with the same wavelength but different observation angles can be used to distinguish the shape of particulate matter. However, this study also found that large interfering aerosol particles affect the scattered light intensity, causing certain deviations in the detection of these scattering characteristics. To reduce false alarms when interfering aerosol particles are large, the research and development team conducted further research and discovered that adding light sources with different wavelengths can correct the deviations caused by particle size. Therefore, in this invention, three light sources are designed, two blue light sources, and a red light source is added to correct the data. Using three light sources allows for three different types of data to be received, which can then be used to extend the aerosol classification curve into a classification surface. A vector composed of the three types of data is input into the classifier, and machine learning algorithms are used to calculate the coefficients of the classification surface equation, resulting in more accurate classification results. Furthermore, this invention acquires multiple data sets and obtains multiple classification results, thereby using statistical methods to improve the accuracy of the classification results. In other words, this invention can correct the detection data based on particle shape characteristics, thereby achieving classification of fire smoke and interfering aerosols, thereby reducing the false alarm rate of smoke detectors and taking reasonable measures.

[0040] The present invention also relates to a smoke and fire detector including an optical darkroom and a control module. The optical darkroom includes a light source combination and a photoelectric conversion module D. The light source combination includes a light source B1, a light source I1, and a light source B2. Light sources B1 and B2 are both blue light sources, and light source I1 is an infrared light source. The center of the optical darkroom is O. The light from light source B1 is reflected from the center O to the photoelectric conversion module D at an acute angle. The light from light source I1 is reflected from the center O to the photoelectric conversion module D at an acute angle. The light from light source B2 is reflected from the center O to the photoelectric conversion module D at an obtuse angle. The control module is used to cooperate with the optical darkroom to implement the above-mentioned smoke and fire detection method based on particle shape characteristics. For details about the optical darkroom and the specific detection method, please refer to the above descriptions, and they will not be repeated here.

[0041] As described above, the present invention uses three light sources, two of which are blue light LED light sources and one is an infrared light source, to form an optical darkroom, and combined with the data processing process by the control module, it can accurately classify fire smoke and interfering aerosols, thereby reducing the false alarm rate of traditional smoke detection fire detectors and improving the detection accuracy of photoelectric smoke detection fire detectors.

[0042] The technical features of the above-described embodiments can be combined in any manner, and for the sake of simplicity, not all possible combinations of the various technical features in the above-described embodiments are described. However, as long as there is no contradiction in the combination of these technical features, it should be considered to be within the scope of the present invention. Note that the terms "in one embodiment," "for example," "also," and the like in the present invention are intended to illustrate the present invention, rather than limiting the present invention.

[0043] The above-described embodiments only represent some embodiments of the present invention, and although the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the claims. It should be noted that a person skilled in the art can make some modifications and improvements without departing from the spirit of the present invention, all of which belong to the protection scope of the present invention.

Claims

1. 1. A smoke detection and fire detection method based on particle shape characteristics, comprising: Particles are detected in an optical darkroom, the optical darkroom including a light source combination and a photoelectric conversion module D, the light source combination including a light source B1, a light source I1, and a light source B2, the light source B1 and the light source B2 are both blue light sources, the light source I1 is an infrared light source, the center of the optical darkroom is O, the light path change angle at which light from the light source B1 is reflected to the photoelectric conversion module D through the center O is an acute angle, the light path change angle at which light from the light source I1 is reflected to the photoelectric conversion module D through the center O is also an acute angle, and the light path change angle at which light from the light source B2 is reflected to the photoelectric conversion module D through the center O is an obtuse angle; The sensing method comprises: Step S1: obtaining a background value of the optical power of the scattered light of each light source, and the background value of each light source is the optical power of the scattered light received by the photoelectric conversion module D from the corresponding light source in an environment without fire smoke or interfering aerosols; Step S2: periodically activating the light source combination to sequentially transmit light pulse signals, calculating the change value of each light source, determining whether there is a change value exceeding a set threshold, if there is not, repeating step S2, if there is, it is considered that there is an abnormality, and executing step S3, the change value of each light source is the absolute value of the difference between the optical power of the scattered light currently received by the photoelectric conversion module D from the light source and the background value of the optical power of the scattered light of the light source; Step S3: if there is an abnormality, activate the light source combination n times consecutively and record the obtained change value, where n is a positive integer; Step S4: constructing the three change values ​​of the light source combinations at each startup into a space vector, and using a classifier to classify the n space vectors respectively to obtain n classification results, the classifier being a trained neural network model, and the classification categories include fire smoke and nuisance aerosols, and among the n classification results obtained by statistics, if the number of one category is k or more, the category is determined to be the correct classification result, where k>n / 2; A method for detecting smoke and fire based on particle shape characteristics, comprising:

2. 2. The method for detecting smoke and fire based on particle shape characteristics according to claim 1, wherein, in step S4, if there are no categories with a number of k or more in the statistical results of the multiple classifications, the method proceeds to step S3, where the change values ​​are reacquired and reclassification is performed until it is determined that the classification result is correct.

3. 2. The method for detecting smoke and fire based on particle shape characteristics according to claim 1, wherein n is 10 or more and k is 8 or more.

4. 2. The smoke detection and fire detection method based on particle shape characteristics according to claim 1, wherein the classification categories include white fire smoke, black fire smoke, and nuisance aerosols, and the nuisance aerosols include one or more of dust, water vapor, and oily smoke.

5. 2. The smoke detection and fire detection method based on particle shape characteristics according to claim 1, wherein obtaining the background value of the optical power of the scattered light of each light source in step S1 includes performing initialization, sequentially activating each light source in an environment free of fire smoke and interfering aerosols, and obtaining the optical power of the scattered light received by the photoelectric conversion module from the corresponding light source as the background value of the optical power of the scattered light of the corresponding light source.

6. 2. The method for detecting smoke and fire based on particle shape characteristics as claimed in claim 1, wherein the period for activating the light source combination to transmit the light pulse signal is one or several seconds, and the pulse width of the light pulse signal is in the range of several milliseconds to several tens of milliseconds.

7. 2. The method for detecting smoke and fire based on particle shape characteristics according to claim 1, wherein the classifier includes a plurality of support vector machine (SVM) binary classification models, each of which is used to realize classification of two different categories. When classifying a spatial vector using the classifier, each binary classification model performs binary classification on the input spatial vector, and the most frequent classification result of all the binary classification models is used as the classification result of the spatial vector by the classifier.

8. 1. A particle shape characteristic based smoke detection and fire detector, comprising an optical darkroom and a control module, the optical darkroom includes a light source combination and a photoelectric conversion module D, the light source combination includes a light source B1, a light source I1, and a light source B2, the light source B1 and the light source B2 are both blue light sources, the light source I1 is an infrared light source, the center of the optical darkroom is O, the light path change angle at which light from the light source B1 is reflected to the photoelectric conversion module D through the center O is an acute angle, the light path change angle at which light from the light source I1 is reflected to the photoelectric conversion module D through the center O is also an acute angle, and the light path change angle at which light from the light source B2 is reflected to the photoelectric conversion module D through the center O is an obtuse angle; A smoke and fire detector based on particle shape characteristics, characterized in that the control module is used to carry out the method according to any one of claims 1 to 7.

9. The particle shape characteristic-based smoke and fire detector according to claim 8, wherein the acute angle is equal to or greater than 20° and the obtuse angle is equal to or less than 170°.

10. 9. The particle shape characteristic-based smoke and fire detector according to claim 8, wherein the two acute angles are the same or have a difference of less than 10 degrees.

Images