Particulate matter detection system and method

Abstract

The invention discloses a particulate matter detection system and method, and relates to the field of optical detection and environmental monitoring, combined linearly polarized light is generated through a light source module, the combined linearly polarized light comprises light beams with different wavelengths and adjustable linear polarization directions, and irradiation of the linearly polarized light is combined with multi-wavelength response, so that the particle size of the particulate matter is detected. Scattered light has higher information density, after the scattered light with the higher information density is received by the multi-channel photoelectric detection module and converted into an electric signal, the electric signal is processed through the data processing module, and scattered light intensities, linear polarization degrees and linear polarization angles of different wavelengths and different scattering angles can be obtained. The data processing module can obtain accurate particle size distribution through the scattered light intensities, the linear polarization degrees and the linear polarization angles of different wavelengths and different scattering angles and by means of the scattered light intensities, the linear polarization degrees and the linear polarization angles of different particle sizes and different wavelengths and different scattering angles provided by the Mie scattering simulation library, and the particle concentration can be accurately estimated according to the particle size distribution.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of optical detection and environmental monitoring, and particularly relates to a particulate matter detection system and method. BACKGROUND

[0002] Traditional particulate detection equipment is mostly based on laser scattering measurement, and the scattering intensity is received by multiple photodetectors at different angles to inversely calculate the particle size and concentration of particles. The recognition ability of this kind of particulate matter detection system for particle information is limited, and the accuracy of the detection result obtained by detecting particulate matter by using this kind of particulate matter detection system is low, so that the accurate particle size distribution and particle concentration cannot be obtained. Therefore, it has become a problem to be solved by those skilled in the art to develop a particulate matter detection system capable of obtaining accurate particle size distribution and particle concentration. SUMMARY

[0003] The purpose of the present application is to provide a particulate matter detection system and method to obtain accurate particle size distribution and particle concentration.

[0004] To achieve the above-mentioned purpose, the present application provides the following solutions:

[0005] In a first aspect, the present application provides a particulate matter detection system, which comprises a light source module, a detection cavity, a multi-channel photodetector module and a data processing module.

[0006] The light source module is used to generate a combined linearly polarized light; the combined linearly polarized light comprises different wavelength light beams with adjustable linear polarization directions;

[0007] The detection cavity has an incident hole and a gas or particulate sample flow passage; the incident hole is arranged on the side wall of the detection cavity close to the light source module; the combined linearly polarized light generated by the light source module is irradiated on the particles in the gas or particulate sample flow passage through the incident hole, and scattering occurs;

[0008] The multi-channel photodetector module is arranged on the side wall of the detection cavity; the multi-channel photodetector module is used to receive the light signal scattered by the particles after the combined linearly polarized light, convert the light signal into an electrical signal, and send the electrical signal to the data processing module; the multi-channel photodetector module comprises multiple photodetectors;

[0009] The data processing module is configured to obtain scattering light intensity and polarization characteristics at different wavelengths and different scattering angles according to the electrical signals sent by the multi-channel photoelectric detection module, and to obtain a particle size distribution by inverting the particle size distribution using scattering light intensity and polarization characteristics at different wavelengths and different scattering angles for different particle sizes provided by a Mie scattering simulation library, and to estimate a particle concentration according to the particle size distribution; the polarization characteristics include a degree of linear polarization and a linear polarization angle.

[0010] Optionally, the particulate matter detection system further comprises an image acquisition module.

[0011] The image acquisition module is arranged on a side wall of the detection cavity.

[0012] The image acquisition module is configured to capture a scattering region synchronously when the multi-channel photoelectric detection module acquires the optical signals, to acquire a particle image, and to send the particle image to the data processing module.

[0013] The data processing module is further configured to obtain image features from the particle image sent by the image acquisition module, and to obtain a particle category based on a multi-modal feature fusion network using the image features and the scattering light intensity and polarization characteristics at different wavelengths and different scattering angles; the particle category includes a category of particle morphology, a category of particle optical properties, a category of particle dynamic properties, and a category of particle physicochemical properties.

[0014] Optionally, the image acquisition module uses a high-resolution camera.

[0015] Optionally, the particulate matter detection system further comprises a control and synchronization module.

[0016] The control and synchronization module is configured to synchronously control wavelength switching, photoelectric detector sampling, and high-resolution camera exposure.

[0017] Optionally, the light source module comprises a short-wave laser, a medium-wave laser, a long-wave laser, a near-infrared laser, and a light combining device.

[0018] The short-wave laser is configured to generate a short-wave light beam with an adjustable linear polarization direction.

[0019] The medium-wave laser is configured to generate a medium-wave light beam with an adjustable linear polarization direction.

[0020] The long-wave laser is configured to generate a long-wave light beam with an adjustable linear polarization direction.

[0021] The near-infrared laser is configured to generate a near-infrared light beam with an adjustable linear polarization direction.

[0022] The light combining device is used for combining the short-wave light beams, the middle-wave light beams, the long-wave light beams and the near-infrared light beams with the adjustable linear polarization direction to generate the combined linearly polarized light.

[0023] Optionally, when the short-wave laser, the middle-wave laser, the long-wave laser and the near-infrared laser are arranged from short to long or from long to short in wavelength, the light combining device combines the short-wave light beams, the middle-wave light beams, the long-wave light beams and the near-infrared light beams with the adjustable linear polarization direction by using a first dichroic mirror, a second dichroic mirror and a third dichroic mirror.

[0024] When the short-wave laser, the middle-wave laser, the long-wave laser and the near-infrared laser are arranged from short to long in wavelength, the first dichroic mirror is used for combining the short-wave light beams and the middle-wave light beams with the adjustable linear polarization direction, the second dichroic mirror is used for combining the short-wave light beams, the middle-wave light beams and the long-wave light beams with the adjustable linear polarization direction, and the third dichroic mirror is used for combining the short-wave light beams, the middle-wave light beams, the long-wave light beams and the near-infrared light beams with the adjustable linear polarization direction.

[0025] When the short-wave laser, the middle-wave laser, the long-wave laser and the near-infrared laser are arranged from long to short in wavelength, the first dichroic mirror is used for combining the near-infrared light beams and the long-wave light beams with the adjustable linear polarization direction, the second dichroic mirror is used for combining the near-infrared light beams, the long-wave light beams and the middle-wave light beams with the adjustable linear polarization direction, and the third dichroic mirror is used for combining the near-infrared light beams, the long-wave light beams, the middle-wave light beams and the short-wave light beams with the adjustable linear polarization direction.

[0026] Optionally, when the short-wave laser, the middle-wave laser, the long-wave laser and the near-infrared laser are not arranged from short to long or from long to short in wavelength, the light combining device combines the short-wave light beams, the middle-wave light beams, the long-wave light beams and the near-infrared light beams with the adjustable linear polarization direction by using a combination of a light combining prism and a fourth dichroic mirror.

[0027] The light combining prism is used for combining the short-wave light beams, the middle-wave light beams and the long-wave light beams with the adjustable linear polarization direction.

[0028] The fourth dichroic mirror is used for combining the short-wave light beams, the middle-wave light beams, the long-wave light beams and the near-infrared light beams with the adjustable linear polarization direction.

[0029] In a second aspect, the present application provides a particulate matter detection method, which is applied to the particulate matter detection system, and includes the following steps.

[0030] Obtaining the electrical signal sent by the multi-channel photoelectric detection module;

[0031] The scattering light intensity and polarization characteristics at different wavelengths and different scattering angles are obtained according to the electrical signals sent by the multi-channel photoelectric detection module; the polarization characteristics include linear polarization degree and linear polarization angle;

[0032] A Mie scattering simulation library is obtained;

[0033] According to the scattering light intensity and polarization characteristics at different wavelengths and different scattering angles, the Mie scattering simulation library provides the scattering light intensity and polarization characteristics at different wavelengths and different scattering angles of different particle sizes, and the particle size distribution is inversed based on a least square method, Bayesian inference or a neural network model;

[0034] The particle concentration is estimated according to the particle size distribution.

[0035] Optionally, the input of the Mie scattering simulation library is a series of particle sizes, complex refractive indexes, wavelengths and scattering angles; the output of the Mie scattering simulation library is relative concentrations at a series of particle sizes, and scattering light intensity, linear polarization degree and linear polarization angle at different wavelengths and different scattering angles of each particle size; the relative concentrations at a series of particle sizes are used to represent the particle size distribution.

[0036] Optionally, the particulate matter detection method further comprises:

[0037] A particle image sent by an image acquisition module is obtained;

[0038] Image features are obtained according to the particle image sent by the image acquisition module;

[0039] The particle class is obtained based on a multi-modal feature fusion network by using the image features and the scattering light intensity and polarization characteristics at different wavelengths and different scattering angles; the particle class includes a class of particle morphology, a class of optical characteristics of particles, a class of dynamic characteristics of particles and a class of physicochemical characteristics of particles.

[0040] According to the specific embodiments provided in the present application, the present application has the following technical effects:

[0041] The application provides a particulate matter detection system and method, a combined linearly polarized light is generated by a light source module, the combined linearly polarized light includes light beams with different wavelengths and adjustable linear polarization directions, the light beams with different wavelengths are combined, and the light beams with different wavelengths are combined to respond, so that the scattered light has higher information density, the scattered light with higher information density is received by a multi-channel photoelectric detection module and converted into an electric signal, and then the electric signal is processed by a data processing module, so that the scattered light intensity, linear polarization degree and linear polarization angle of different wavelengths and different scattering angles can be obtained, the data processing module obtains the scattered light intensity, linear polarization degree and linear polarization angle of different wavelengths and different scattering angles of different particle diameters provided by a Mie scattering simulation library, and the accurate particle size distribution can be obtained, and the particle concentration can be accurately estimated according to the particle size distribution. BRIEF DESCRIPTION OF DRAWINGS

[0042] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the drawings needed in the embodiments will be briefly introduced below. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor.

[0043] Figure 1 A structural schematic diagram of a particulate matter detection system provided by an embodiment of the present application is shown in the figure.

[0044] Figure 2 A light source module structural schematic diagram when a light combination device of the present application adopts three dichroic mirrors is shown in the figure.

[0045] Figure 3 A light source module structural schematic diagram when a light combination device of the present application adopts a combination of one light combination prism and one dichroic mirror is shown in the figure.

[0046] Figure 4 A flowchart of a particulate matter detection method provided by an embodiment of the present application is shown in the figure. DETAILED DESCRIPTION

[0047] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only some embodiments of the present application, not all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0048] The purpose of the present application is to provide a particulate matter detection system and method to obtain accurate particle size distribution and particle concentration.

[0049] In order to make the above objectives, characteristics and advantages of the present application more apparent, further specific embodiments will be described in detail below with reference to the accompanying drawings and specific embodiments.

[0050] As shown in Figure 1 The particle detection system provided by the present application includes a light source module 101, a detection cavity 102, a multi-channel photoelectric detection module 103, an image acquisition module 104 and a data processing module 105.

[0051] The light source module 101 is used to generate a combined linearly polarized light; the combined linearly polarized light includes different wavelength beams with adjustable linear polarization directions.

[0052] The detection cavity 102 has an incident hole and a gas or particle sample flow passage; the incident hole is arranged on the side wall of the detection cavity 102 close to the light source module 101; the combined linearly polarized light generated by the light source module 101 is irradiated on the particles in the gas or particle sample flow passage through the incident hole, and scattering occurs.

[0053] The multi-channel photoelectric detection module 103 is arranged on the side wall of the detection cavity 102; the multi-channel photoelectric detection module 103 is used to receive the light signal scattered by the particles after the combined linearly polarized light, convert the light signal into an electrical signal, and send the electrical signal to the data processing module 105; the multi-channel photoelectric detection module 103 includes a plurality of photoelectric detectors (only one photoelectric detector is shown as an example, Figure 1 The installation angles of the plurality of photoelectric detectors cover the range of 0°-180°. In particular, the photoelectric detector can be a polarization-sensitive photoelectric detector, or a photoelectric detector with a polarization analyzer (a polarization analyzer is arranged in front of the photoelectric detector);

[0054] The data processing module 105 is used to obtain the scattering light intensity and polarization characteristics of different wavelengths and different scattering angles according to the electrical signal sent by the multi-channel photoelectric detection module 103, and inversely calculate the particle size distribution according to the scattering light intensity and polarization characteristics of different wavelengths and different scattering angles provided by the Mie scattering simulation library for different particle sizes, and estimate the particle concentration according to the particle size distribution; the polarization characteristics include the linear polarization degree and the linear polarization angle.

[0055] The image acquisition module 104 is arranged on the side wall of the detection cavity 102; the image acquisition module 104 is used to synchronously shoot the scattering area when the multi-channel photoelectric detection module 103 acquires the light signal, acquire the particle image, and send the particle image to the data processing module 105.

[0056] The data processing module 105 is further configured to obtain image features from the particle images sent by the image acquisition module 104, and obtain a particle category based on a multi-modal feature fusion network by using the image features and the scattering light intensity and polarization features of different wavelengths and different scattering angles; the particle category includes a particle morphology category, a particle optical property category, a particle dynamic property category, and a particle physicochemical property category.

[0057] The light source module 101 includes a short-wave laser, a middle-wave laser, a long-wave laser, a near-infrared laser, and a light combination device. The short-wave laser is configured to generate a short-wave light beam with an adjustable linear polarization direction. The middle-wave laser is configured to generate a middle-wave light beam with an adjustable linear polarization direction. The long-wave laser is configured to generate a long-wave light beam with an adjustable linear polarization direction. The near-infrared laser is configured to generate a near-infrared light beam with an adjustable linear polarization direction. The light combination device is configured to combine the short-wave light beam, the middle-wave light beam, the long-wave light beam, and the near-infrared light beam with adjustable linear polarization directions to generate a linearly polarized light beam.

[0058] Specifically, the short-wave laser, the middle-wave laser, the long-wave laser, and the near-infrared laser each adopt an edge-emitting laser diode.

[0059] The light combination device can combine the short-wave light beam, the middle-wave light beam, the long-wave light beam, and the near-infrared light beam with adjustable linear polarization directions by using a first dichroic mirror, a second dichroic mirror, and a third dichroic mirror, or by using a combination of a light combination prism and a fourth dichroic mirror.

[0060] When the short-wave laser, the middle-wave laser, the long-wave laser, and the near-infrared laser are arranged in the order of short to long or long to short in terms of wavelength, the light combination device combines the short-wave light beam, the middle-wave light beam, the long-wave light beam, and the near-infrared light beam with adjustable linear polarization directions by using a first dichroic mirror, a second dichroic mirror, and a third dichroic mirror.

[0061] When the short-wave laser, the middle-wave laser, the long-wave laser, and the near-infrared laser are arranged in the order of short to long in terms of wavelength, the first dichroic mirror is configured to combine the short-wave light beam and the middle-wave light beam with adjustable linear polarization directions, the second dichroic mirror is configured to combine the short-wave light beam, the middle-wave light beam, and the long-wave light beam with adjustable linear polarization directions, and the third dichroic mirror is configured to combine the short-wave light beam, the middle-wave light beam, the long-wave light beam, and the near-infrared light beam with adjustable linear polarization directions.

[0062] When the short-wave laser, the medium-wave laser, the long-wave laser and the near-infrared laser are arranged from long to short in wavelength, the first dichroic mirror is used for beam combining of the near-infrared beam and the long-wave beam with adjustable linear polarization direction, the second dichroic mirror is used for beam combining of the near-infrared beam, the long-wave beam and the medium-wave beam with adjustable linear polarization direction, and the third dichroic mirror is used for beam combining of the near-infrared beam, the long-wave beam, the medium-wave beam and the short-wave beam with adjustable linear polarization direction.

[0063] When the short-wave laser, the medium-wave laser, the long-wave laser and the near-infrared laser are not arranged from short to long or from long to short in wavelength, the light combining device adopts a combination of a light combining prism and a fourth dichroic mirror to combine the short-wave beam, the medium-wave beam, the long-wave beam and the near-infrared beam with adjustable linear polarization direction.

[0064] The light combining prism is used for beam combining of the short-wave beam, the medium-wave beam and the long-wave beam with adjustable linear polarization direction.

[0065] The fourth dichroic mirror is used for beam combining of the short-wave beam, the medium-wave beam, the long-wave beam and the near-infrared beam with adjustable linear polarization direction.

[0066] The image acquisition module 104 adopts a high-resolution camera.

[0067] Further, the particulate matter detection system provided in the application further comprises a control and synchronization module (not shown in the figure). The control and synchronization module is used for synchronous control of wavelength switching, photodetector sampling and high-resolution camera exposure.

[0068] The technical scheme of the application is described below with one specific embodiment:

[0069] The particulate matter detection system provided in the application is a particulate matter detection system based on polarized light and multi-wavelength scattering analysis, which combines linearly polarized light, multi-wavelength irradiation, image recognition and multi-channel synchronous photodetection, realizes particulate matter detection, and is suitable for application scenarios such as air quality monitoring, biological aerosol identification and industrial dust detection.

[0070] The particulate matter detection system based on polarized light and multi-wavelength scattering analysis provided in the application is a particulate matter detection system based on linearly polarized light irradiation, multi-wavelength combined illumination, image recognition and multi-channel photodetection fusion. The system can extract multiple characteristic parameters of particulate matter such as size, shape, refractive index, concentration and polarization response, realize more comprehensive, accurate and intelligent detection, and is composed of a polarized light and multi-wavelength scattering detection system. Based on the system, the particle size distribution, particle concentration and particle morphology can be inverted.

[0071] As Figure 1As shown, the particulate matter detection system provided by the present application includes a light source module 101, a detection cavity 102, a multi-channel photoelectric detection module 103, an image acquisition module 104, a control and synchronization module, and a data processing module 105. The particulate matter detection system can adopt a modular structure design and is suitable for portable or industrial embedded deployment environments.

[0072] The light source module 101, the detection cavity 102, the multi-channel photoelectric detection module 103, the image acquisition module 104, the control and synchronization module, and the data processing module 105 will be described in detail as follows:

[0073] The light source module 101 includes a multi-wavelength light source group (including a short-wave laser, a medium-wave laser, a long-wave laser, and a near-infrared laser) and a light combining device, which is used to generate different wavelength light beams with adjustable linear polarization directions, at least including four wave bands of short-wave (about 450 nm), medium-wave (about 530 nm), long-wave (about 630 nm), and near-infrared (805-950 nm). The light source can adopt an edge-emitting laser diode (EELD) with high extinction ratio linearly polarized light output, which can greatly reduce the use of additional polarizing plates or wave plates and other devices, simplify the system structure, and reduce the cost. A vertical cavity surface emitting laser (VCSEL) can also be used to realize rapid switching of the polarization direction in cooperation with a filter, a polarizer, and other elements. Figure 1 The medium incidence combined laser (linear polarization) is the linearly polarized light combined by the light beams, and the incidence direction is usually fixed.

[0074] The light combining device can adopt multiple dichroic mirrors or a combination of one light combining prism and one dichroic mirror. The dichroic mirror is used to transmit light of one wave band in one direction and fully reflect light of another wave band in another direction, and the light in the two directions is combined in the outgoing direction. The light combining prism is used to combine light in three directions to one direction. For example, Figure 2As shown, the light combining device includes three dichroic mirrors, which are sequentially arranged on the outgoing light path of the short-wave laser and form a 45° angle (preferably 45°) with the outgoing light path of the short-wave laser (the dichroic mirror is used for beam combining, and this relative installation position enables the reflected light to be totally reflected), and the remaining lasers correspond to the dichroic mirrors one by one and are arranged in a direction perpendicular to the outgoing light path of the short-wave laser (the angle can not be limited, as long as the light of one waveband can be transmitted and the light of another waveband can be totally reflected, and the transmitted light can be combined, finally all the light can be combined and enter the detection cavity 102 through the entrance hole), and the four wavelength laser beams are combined by the dichroic mirrors to form a coaxial composite light path, that is, the light beams of different wavelengths completely overlap in space. The positions of the different waveband lasers can be changed, but they need to be arranged from short to long or from long to short according to the wavelength, which is convenient for the work of the dichroic mirror. Among them, the three dichroic mirrors are not the same dichroic mirror. The dichroic mirror is usually composed of multiple thin films. The thickness and refractive index of each thin film are precisely designed so that the reflected and transmitted light can meet the specific ratio at a specific wavelength. For example Figure 3 As shown, the light emitted by the short-wave laser, the medium-wave laser and the long-wave laser is combined by the light combining prism and then emitted by the dichroic mirror, and the light emitted by the near-infrared laser is totally reflected by the dichroic mirror and then combined and emitted. The light emitted by the near-infrared laser can also be emitted by the dichroic mirror, and the combined light of the light combining prism is reflected by the dichroic mirror and combined with the near-infrared light.

[0075] Specifically, different lasers can be controlled by a multi-channel constant current driver to realize wavelength switching or sequential irradiation.

[0076] The detection cavity 102 has an entrance hole and a gas or particle sample flow passage.

[0077] The multi-channel photodetector module (i.e., multi-channel photodetector module) 103 includes multiple photodetectors (PDs) arranged at different positions of the detection cavity 102 to realize multi-angle polarized scattered light signal acquisition and conversion into an electrical signal; the multi-angle polarized scattered light signal includes scattered light intensity (I), linear polarization degree (DoLP) and linear polarization angle (AoLP); preferably, a high-sensitivity PD is used, the installation angle is set to cover the range of 0°-180°, and the PD has analog amplification and high-speed analog-to-digital conversion capabilities. The combined linearly polarized light is irradiated on the particle through the entrance hole, and the light signal scattered by the particle is received by the photodetector (PD) at the corresponding position and converted into an electrical signal. A polarization analyzer is arranged in front of each detector to measure the light signal of different polarization directions.

[0078] The image acquisition module 104 includes only one high-resolution camera, which can be positioned anywhere and can capture the scattering area to obtain the particle image.

[0079] Control and synchronization module: used for synchronously controlling light source switching (realizing wavelength switching), detector sampling and camera exposure, and realizing multi-source synchronous data acquisition.

[0080] Data processing module 105: used for receiving electrical signals of the multi-channel photoelectric detection module, extracting particle size, shape, polarization characteristics, scattered light intensity (i.e. scattered light intensity), and the like, and inversely calculating particle size distribution, refractive index and particle concentration in combination with a Mie scattering model, a Rayleigh scattering model or an AI model; the polarization characteristics include a degree of linear polarization (DoLP) and a polarization angle (AoLP), and the polarization angle (AoLP) is a linear polarization angle (AoLP). The data processing module 105 is also used for receiving electrical signals of the image acquisition module 104 and the multi-channel photoelectric detection module 103, and can further identify particle morphology based on computer vision technologies such as edge detection, particle tracking and morphological classification, in cooperation with the Mie scattering model, the Rayleigh scattering model or the deep learning model.

[0081] The comprehensive scattering inversion method of the data processing module is specifically introduced as follows:

[0082] The particle detection system (multi-wavelength scattering) provided in the application realizes particle multi-parameter inversion in combination with multi-wavelength scattering data, polarization information and image recognition features. Specifically, the method of realizing particle multi-parameter inversion in combination with multi-wavelength scattering data, polarization information and image recognition features is as follows:

[0083] (1) Constructing a Mie scattering simulation library:

[0084] Input: a series of particle sizes α i , complex refractive index m, wavelength λ j , scattering angle θ k ;

[0085] Output: I(λ j , θ k ), DoLP(λ j , θ k ), AoLP(λ j , θ k ) of each particle size;

[0086] Wherein, is a discrete variable vector, wherein d i is a relative concentration at the particle size α i , used to represent a particle size distribution function (particle size distribution) D(α). i = 1, 2, …, N, and N is dimension, determines the length of the vector and the number of discrete variables contained. m = m' + iκ, m' is the real part of the complex refractive index (refractive index), iκ is the imaginary part of the complex refractive index, i is the imaginary unit, and κ is the extinction coefficient. I(λ j , θ k ) represents the scattering light intensity at wavelength λ j and scattering angle θ k . DoLP(λ j , θ k ) represents the degree of linear polarization at wavelength λ j and scattering angle θ k . AoLP(λ j , θ k ) represents the angle of linear polarization at wavelength λ j and scattering angle θ k .

[0087] I(λ j , θ k ), DoLP(λ j , θ k ), AoLP(λ j , θ k ) are calculated as follows:

[0088] Calculate the size parameter x:

[0089]

[0090] Calculate Mie coefficients a n , b n :

[0091]

[0092] where ψ n and ξ n are Riccati-Bessel functions, ψ' n and ξ' n are the derivatives of ψ n and ξ n , respectively, a n is the scattering coefficient describing the magnetic multipole component, b n is the scattering coefficient describing the electric multipole component. n represents the order of the multipole.

[0093] Calculate the scattering amplitude functions S V (λ j , θ k ), S H (λ j , θ k ):

[0094]

[0095] where S V (λ j ,θ k ) and S H (λ j ,θ k ) describe the polarization components of the scattered light in the perpendicular and parallel directions to the scattering plane. The angle functions π n (cosθ k ) and τ n (cosθ k ) are important components describing the polarization properties of the scattered light, which are related to the scattering angle θ k and derived from the Legendre polynomials. n denotes the order of the multi-level.

[0096] Calculate I(λ j ,θ k ), DoLP(λ j ,θ k ), AoLP(λ j ,θ k ):

[0097] I(λ j ,θ k ) = |S V (λ j ,θ k )| 2 + |S H (λ j ,θ k )| 2 ;

[0098]

[0099] where is the conjugate complex of S H (λ j ,θ k ), denotes taking the real part of the complex number .

[0100] Construct the tensor library:

[0101] K i ∈ R M×L×3 ;

[0102] For each particle size α i , construct a tensor K i , the tensor structure is the first dimension (M) for the wavelength λ j , the second dimension (L) for the scattering angle θ k , and the third dimension (3) for [I(λ j ,θ k), DoLP (λ j , θ k ), AoLP (λ j , θ k )], the elements of the tensor K i are defined as K i [j, k, 0] = I (λ j , θ k ), K i [j, k, 1] = DoLP (λ j , θ k ), K i [j, k, 2] = AoLP (λ j , θ k ), where M is the number of wavelengths, j ∈ {0, 1, …, M-1} is the wavelength index, L is the number of scattering angles, k ∈ {0, 1, …, L-1} is the scattering angle index. R represents the set of real numbers, indicating that all elements of the tensor K i are real numbers.

[0103] Extended tensor library: for multiple particle sizes α1, α2, …, α N , a larger tensor library K' i ∈ R N ×M×L×3 , where: K[i, j, k, 0:2]: corresponds to particle size α i , wavelength λ j , scattering angle θ k under [I (λ j , θ k ), DoLP (λ j , θ k ), AoLP (λ j , θ k )].

[0104] Weighted superposition calculation of simulation tensor T sim , that is, the response of each particle size is weighted and synthesized according to its distribution, to obtain the simulation tensor T sim :

[0105]

[0106] Where d i is the relative concentration at particle size α i .

[0107] (2) Construct the feature tensor T meas :

[0108] Multi-source data acquisition: based on the particulate matter detection system provided in the present application, multiple angles (scattering angles) θ' k are obtained under different wavelengths λ' j .the scattering intensity I(λ' j ,θ' k ) and its corresponding polarization characteristics DoLP(λ' j ,θ' k ), AoLP(λ' j ,θ' k ). Wherein I(λ' j ,θ' k ) represents the scattering light intensity of wavelength λ' j and scattering angle θ'k. DoLP(λ' j ,θ' k ) represents the degree of linear polarization of wavelength λ' j and scattering angle θ' k . AoLP(λ' j ,θ' k ) represents the linear polarization angle of wavelength λ' j and scattering angle θ' k . j is the wavelength index, j∈{0,1,……,M-1}, k is the scattering angle index, k∈{0,1,……,L-1}, where M is the number of wavelengths, and L is the number of scattering angles.

[0109] Construct a three-dimensional characteristic tensor (i.e. characteristic tensor T meas ):

[0110] A polarization analyzer is arranged in front of each detector to measure the light intensity I H (λ' j ,θ' k ) in the horizontal direction, the light intensity I V (λ' j ,θ' k ) in the vertical direction, the light intensity I +45° (λ' j ,θ' k ) in the +45° direction, and the light intensity I -45° (λ' j ,θ' k ) in the -45° direction.

[0111] In the system of describing the polarization state of light by Stokes vector, the Stokes vector contains four components (S0(λ' j ,θ' k ), S1(λ' j ,θ' k ), S2(λ' j ,θ' k ), S3(λ' j ,θ' k )). S0(λ' j ,θ' k) is the first component in the Stokes parameter system, used to describe the total intensity or total light intensity of light:

[0112] S0(λ′ j ,θ′ k ) = I H (λ′ j ,θ′ k ) + I V (λ′ j ,θ′ k );

[0113] S1(λ′ j ,θ′ k ) = I H (λ′ j ,θ′ k ) - I V (λ′ j ,θ′ k );

[0114] S2(λ′ j ,θ′ k ) = I +45° (λ j ,θ′ k ) - I -45° (λ′ j ,θ′ k ).

[0115] Wherein, S1(λ' j ,θ' k ) is the second component in the Stokes parameter system, S2(λ' j ,θ' k ) is the third component in the Stokes parameter system, S3(λ' j ,θ' k ) is the fourth component in the Stokes parameter system.

[0116] Calculate I(λ' j ,θ' k ), DoLP(λ' j ,θ' k ), AoLP(λ' j ,θ' k ), construct three-dimensional feature tensor T meas = [I(λ′ j ,θ′ k ), DoLP(λ′ j ,θ′ k ), AoLP(λ′ j ,θ′ k )]:

[0117] I(λ′j ,θ′ k )=S0(λ′ j ,θ′ k );

[0118]

[0119]

[0120] (3) Inverted particle size distribution D(α):

[0121] The particle size distribution D(α) is inverted using the least squares method, Bayesian inference, or neural network models (such as MLP):

[0122] Least squares expression:

[0123]

[0124] in For discrete variable vectors, Where d i To achieve a particle size α i The relative concentration at a given point is used to represent the particle size distribution function D(α); Let T be the objective function of the least squares method, representing the measured data (i.e., the feature tensor). meas With the simulated data (i.e., the simulated tensor) T sim The least squares error between them, d i ≥0 is a vector constraint condition, representing a discrete variable vector. Each element d i ≥0 (non-negative concentration constraint); This is a regularization term to prevent overfitting and stabilize the inversion process; T meas For the measured characteristic tensor, T sim For weighted superposition simulation tensors; ε is the regularization parameter, typically ranging from 10 depending on the data and the regularization type. -6 Up to 10 2 The process of inverting particle size distribution using the least squares method is essentially the process of solving the above expression. When the simulation results match the measured data as closely as possible, that is, when the result matches the measured data T... meas The best-matching particle size distribution D(α).

[0125] Bayesian inference expression:

[0126]

[0127] in The posterior distribution refers to the distribution of the measured data T. meas Lower particle size distribution The probability distribution of σ; 2It is the noise variance, ||T meas -T sim || 2 T is the sum of squared errors between the measured data and the simulated data. meas For the measured characteristic tensor, T sim For weighted superposition simulation tensors; Particle size distribution The prior distribution is chosen based on specific issues such as sparsity and smoothness. Bayesian inference is a statistical method used to estimate the posterior distribution of unknown parameters given measurement data and prior knowledge. Finally, the Markov Chain Monte Carlo (MCMC) method is used to sample from the posterior distribution to obtain the particle size distribution D(α).

[0128] The particle size distribution d(α) is inverted using the existing neural network model MLP:

[0129] A multilayer perceptron (MLP), also known as an artificial neural network (ANN), can have multiple hidden layers in addition to the input and output layers. The simplest MLP contains only one hidden layer, i.e., a three-layer structure.

[0130] Use T sim Construct a dataset with the corresponding particle size distribution D(α).

[0131] Use T sim Generate T meas For example, through linear transformation or by adding noise.

[0132] Modeling with MLP sim The relationship with D(α):

[0133] D(α)=MLP ω (T sim );

[0134] Where MLP represents a multilayer perceptron neural network model, ω is the neural network parameter, and T... sim It is a simulated tensor (i.e., a weighted superimposed simulated tensor).

[0135] Train the MLP using the dataset, and then use the trained model on T. meas The particle size distribution D(α) is obtained by prediction.

[0136] (4) Particle concentration estimation:

[0137] Superimposed on λ' j ,θ' k Scattering intensity I of the particle size to be measured meas(λ' j ,θ' k ):

[0138] I meas (λ' j ,θ' k )=∫D(α)·σ sca (λ' j ,α)·p(λ' j ,θ' k ,α)dα;

[0139] σ sca (λ' j ,α)=Q s c a (λ' j ,α)×πα 2 ;

[0140]

[0141] Where D(α) is the particle size distribution function obtained through inversion; σ sca (λ' j p(λ') represents the scattering cross-sectional area, where α represents the particle size to be measured; j ,θ' k Let α be the scattering phase function, a mathematical function describing the distribution of scattered light in different directions. Its expression in Mie scattering theory is: Q s c a (λ' j ,α) represents the scattering efficiency factor.

[0142] The total scattering intensity I at various wavelengths and angles meas (λ,θ):

[0143] I meas (λ,θ)=∫∫I meas (λ' j ,θ' k )dθdλ.

[0144] Estimate the concentration C of particles of the target particle size:

[0145] C∝I meas (λ,θ), given that D(α) is fixed, the particle concentration C and I meas The absolute values ​​of (λ,θ) are directly proportional.

[0146] (5) Particle morphology recognition:

[0147] (a) Data collection and preprocessing:

[0148] The particles are artificially divided into different categories (including categories of particle morphology, categories of optical properties of particles, categories of dynamic properties of particles, and categories of physicochemical properties of particles, wherein the categories of particle morphology include spherical, non-spherical, fibrous, irregular shape, the categories of optical properties of particles include high refractive index, low refractive index, light absorption, non-light absorption, the categories of dynamic properties of particles include stable scattering, dynamic rotation, dynamic Brownian motion, and the categories of physicochemical properties of particles include high toxicity, low toxicity, monodisperse, polydisperse, biological particles), and hierarchical labeling is performed:

[0149] Primary label (based on particle morphology): spherical, non-spherical, fibrous, irregular shape.

[0150] Secondary label 1 (based on optical properties): high refractive index, low refractive index, light absorption, non-light absorption.

[0151] Secondary label 2 (based on dynamic properties): stable scattering, dynamic rotation, dynamic Brownian motion.

[0152] Secondary label 3 (based on physicochemical properties): high toxicity, low toxicity, monodisperse, polydisperse, biological particles.

[0153] Implementation rules: each particle must and only can be assigned one primary label; secondary labels can be multiple and logically compatible with the primary label; if the primary label and the secondary label are contradictory, manual review is triggered; the primary label is used for model training, and the secondary label is used for error analysis.

[0154] Feature extraction:

[0155] The data processing module 105 extracts image features such as aspect ratio, area, perimeter, convexity, circularity, fibrous index, gray level co-occurrence matrix, local binary pattern, motion trajectory, and morphological change from the images collected by the image acquisition module 104.

[0156] The data processing module 105 extracts polarization features such as degree of polarization (DoLP), polarization angle (AoLP), DoLP mean / standard deviation, AoLP angular spectrum distribution, DoLP fluctuation amplitude, AoLP change frequency, wavelength-dependent DoLP, and wavelength-dependent AoLP from the electrical signals converted by the multi-channel photoelectric detection module 103; the degree of polarization (DoLP) is the linear degree of polarization (DoLP).

[0157] Feature dimension reduction: reduce the dimension of high-dimensional features and retain the main variance components; in the case where the category is known, extract features with large inter-class differences and small intra-class differences.

[0158] Collect a sufficient number of particle samples, each sample containing image features, polarization features, and labeled particle category labels. Establish an image-polarization signal space-time synchronization mechanism to ensure that the image features and polarization features of the same particle correspond strictly.

[0159] (b) Constructing the base model (offline training):

[0160] Multi-modal feature fusion network design:

[0161] Dual-branch feature extractor: Lightweight CNN (e.g. MobileNetV3) is used to extract shape features, outputting a 256-dimensional vector; 1D-CNN is designed to process DoLP / AoLP time series signals, outputting a 128-dimensional vector.

[0162] Adaptive fusion layer: Attention mechanism (e.g. Squeeze-and-Excitation module) is used to dynamically adjust the weights of the two branches, formula: w = σ(W2 · δ(W1 · avg(f img ⊕f polar )); where w represents a weight vector, used to quantify the relative importance of f img and f polar two branch features in the fusion process; f img is the feature vector extracted from the image branch, encoding the shape, texture and other visual information of the particles; f polar is the feature vector extracted from the polarization branch, encoding the scattering intensity, polarization state and other physical properties of the particles; ⊕ represents the feature concatenation operation (Concatenation), which merges the two vectors along the dimension direction (e.g. 256+128=384 dimensions); W1 and W2 represent the weight matrix of the first layer of fully connected layer and the weight matrix of the second layer of fully connected layer respectively; avg(·) represents global average pooling on the concatenated feature vector, compressing spatial information and retaining global semantic information in channel dimension. δ(W1·) is used for dimension reduction and introduces nonlinearity. σ(W2·) is used to generate normalized attention weights.

[0163] Mixed loss function training:

[0164] Main loss function:

[0165] For class-balanced data, cross-entropy loss L CE is used:

[0166]

[0167] where C' represents the number of all classes in the data set, used to determine the range of classes considered by the loss function, ensuring probability evaluation for each class. c is the class index, taking values 1, 2, …, C'. y c represents the true label of the cth class; represents the probability of the model predicting that the sample belongs to the cth class. is the logarithm of the predicted probability, used to quantify the uncertainty of the prediction.

[0168] For class imbalance data, focal loss L FL :

[0169]

[0170] where β is the class balance factor, a weight parameter to balance the contribution of different classes to the loss function, with a value range of [0, 1]; γ is the hard example mining factor, a modulation factor to adjust the contribution of easy and hard samples to the loss function, with a value range of [0, 5] in general.

[0171] Auxiliary loss function: add contrastive loss L contrast Strengthen the distinction between similar particles:

[0172]

[0173] where Y is a binary label, indicating whether the sample pair belongs to the same class, when Y = 1, the sample pair is a positive pair (same class samples, such as two spherical particles), when Y = 0, the sample pair is a negative pair (different class samples, such as spherical particles and fibrous particles), by applying different constraints on positive and negative pairs, control the behavior mode of the loss function; D w represents the distance between two samples in the feature space extracted by the model, usually Euclidean distance or cosine distance, used to quantify the similarity of the sample pair, the smaller the distance, the more similar, the larger the distance, the less similar; is a hyperparameter, representing the minimum distance boundary that negative pairs (different class samples) should maintain in the feature space, usually set to 0.5-2.0. is the Hinge Loss term, which is only activated when , represents the punishment for negative pairs with insufficient distance .

[0174] Transfer learning and hyperparameter optimization:

[0175] Pre-training strategy:

[0176] Pre-train on synthetic data sets (e.g. generate regular shape particle images with Blender + simulate polarization signals with a physics engine).

[0177] For real data, use progressive fine-tuning: first freeze the image branch, train the polarization branch; then train jointly.

[0178] Hyperparameter search: use Optuna for Bayesian optimization, search parameters include: learning rate (1×10 -4 ~ 1×10 -2 , fusion layer dimension (64-512), β parameter of Focal Loss (0.1-0.5).

[0179] (c) Real-time detection with continuous learning (online optimization):

[0180] Incremental learning framework:

[0181] Dynamic data buffer pool: maintain the last N=1000 detection samples, stored as (image, polarization feature, prediction confidence) tuples by category. Prioritize low-confidence samples (<0.7) as hard-example sample pool.

[0182] Periodic fine-tuning: trigger model fine-tuning every K=200 new samples collected.

[0183] Real-time feature quality monitoring:

[0184] Feature drift detection: calculate Z-score for key features (e.g. DoLP standard deviation, shape complexity).

[0185] Confidence calibration: use Platt Scaling to temperature scale model output probabilities.

[0186] Active learning feedback loop: uncertainty sampling, model distillation, etc.

[0187] (d) Performance evaluation and iteration:

[0188] Multi-dimensional evaluation of classification performance and system performance, and continuous iterative optimization.

[0189] Through the above steps, a complete closed loop from offline modeling to online optimization can be realized, making the model continuously adapt to new scenarios while maintaining classification accuracy (expected to be improved by 15%~25%) and system stability.

[0190] The application provides a particle detection system with richer detection dimensions and a more compact structure to meet the particle identification requirements in complex environments, based on the limited recognition ability of traditional particle detection equipment on particle shape, material, internal structure and the like, and the complex structure which is not convenient for modular integration and intelligent upgrading and cannot meet the particle identification requirements in complex environments. The particle detection system provided by the application adopts a modular structure design based on a light source module, a multi-channel photoelectric detection module, an image acquisition module and a data processing module, and has a more compact particle detection system structure. Through the light source module, the multi-channel photoelectric detection module, the image acquisition module and the data processing module, linear polarized light, multi-wavelength irradiation, image recognition and multi-channel synchronous photoelectric detection are fused, particle size distribution is inverted, particle concentration is estimated, and particle categories are identified. The particle size distribution contains richer particle system information than a single particle size, and can better reflect key characteristics such as material uniformity and process stability, and is a more comprehensive characterization method. The particle categories include particle morphology categories, particle optical property categories, particle dynamic property categories and particle physicochemical property categories. The particle detection system has richer detection dimensions and can meet the particle identification requirements in complex environments.

[0191] The particle detection system provided by the application combines a polarization signal with an image. Compared with a traditional image method, the polarization image information channel is superimposed, the polarization parameter is increased, surface structure, material characteristics, microscopic orientation and other characteristics can be further identified, the polarization retention difference correction can be performed to reduce the misjudgment rate of aggregates, and the multi-wavelength polarization can also reflect the refractive index change of particles, so that the classification accuracy is further improved.

[0192] The algorithm platform provided by the application can integrate a Mie scattering simulation module (such as BHMIE), a Rayleigh approximation model and an AI model to work together for particle size distribution inversion, concentration quantitative estimation and particle type identification, and is suitable for various applications such as environmental air monitoring, biological particle identification and powder industry control.

[0193] Compared with a traditional particle detection device, the application provides a particle detection device (i.e., a particle detection system) with a more compact structure, richer detection dimensions and more intelligent information extraction, which can meet the particle identification requirements in complex environments.

[0194] Compared with the existing particle measurement method based on laser multi-angle scattering which mainly relies on light intensity ratio analysis, the particle detection system provided by the application not only detects particle species according to the polarization light properties, but also involves image shooting and introduces image recognition technology, so that higher-precision particle classification and composite feature extraction can be realized.

[0195] Compared with the prior art, the application provides a particle detection system combining polarization scattering, multi-wavelength illumination and image recognition technology, which has good innovation, integration and application prospect.

[0196] 1. The application can realize comprehensive analysis of particle size, shape, material (refractive index), concentration and polarization characteristics.

[0197] 2. Using linearly polarized light illumination combined with multi-wavelength response makes the scattered light have higher information density, and can jointly distinguish particle morphology (spherical, ellipsoidal, irregular), surface roughness and material refractive index.

[0198] 3. Compared with the existing system based on pure photodetector, the image acquisition and recognition technology is introduced, the information is more comprehensive, and the recognition ability is stronger.

[0199] 4. Compact structure, only one camera and a small amount of PD are needed to realize multi-channel synchronous acquisition and recognition, and the system integration is high.

[0200] 5. Support AI recognition and remote data transmission, convenient for industrial deployment and intelligent upgrading.

[0201] As shown in Figure 4 The application also provides a particle detection method applied to the particle detection system, comprising:

[0202] Step S1: Obtain the electrical signal sent by the multi-channel photodetector module.

[0203] Step S2: Obtain the scattering light intensity and polarization characteristics of different wavelengths and different scattering angles according to the electrical signal sent by the multi-channel photodetector module; the polarization characteristics include linear polarization degree and linear polarization angle.

[0204] Step S3: Obtain the Mie scattering simulation library.

[0205] The input of the Mie scattering simulation library is a series of particle sizes, complex refractive indexes, wavelengths and scattering angles; the output of the Mie scattering simulation library is a series of relative concentrations at particle sizes, and scattering light intensity, linear polarization degree and linear polarization angle of different wavelengths and different scattering angles of each particle size; the series of relative concentrations at particle sizes are used to represent the particle size distribution.

[0206] Step S4: According to the scattering light intensity and polarization characteristics of different wavelengths and different scattering angles, using the scattering light intensity and polarization characteristics of different wavelengths and different scattering angles of different particle sizes provided by the Mie scattering simulation library, inversing the particle size distribution based on the least square method, Bayesian inference or neural network model.

[0207] Step S5: Estimate the particle concentration according to the particle size distribution.

[0208] Further, the particle detection method provided by the application further comprises:

[0209] Obtaining the particle image sent by the image acquisition module.

[0210] Obtaining the image feature according to the particle image sent by the image acquisition module.

[0211] Obtaining the particle category based on the multi-modal feature fusion network by using the image feature and the scattering light intensity and polarization feature of different wavelengths and different scattering angles; the particle category comprises a particle morphology category, a particle optical property category, a particle dynamic property category and a particle physicochemical property category.

[0212] The principles and implementation manners of the application are described by using specific examples in the present application, and the above example description is only used to help understand the method and core idea of the application; meanwhile, according to the idea of the application, the specific implementation manner and application range will be changed by the person skilled in the art. In conclusion, the content of the present description should not be understood as a limitation of the application.

Claims

1. A particulate matter detection system, characterized by, The particle detection system comprises a light source module, a detection cavity, a multi-channel photoelectric detection module and a data processing module; The light source module is configured to generate combined linearly polarized light; the combined linearly polarized light comprises different wavelength beams with adjustable linear polarization directions; The detection cavity is provided with an incident hole and a gas or particle sample flow passage; the incident hole is arranged on the side wall of the detection cavity close to the light source module; the combined linearly polarized light generated by the light source module is irradiated on particles in the gas or particle sample flow passage through the incident hole, and scattering occurs; The multi-channel photoelectric detection module is arranged on the side wall of the detection cavity; the multi-channel photoelectric detection module is configured to receive light signals scattered by the particles from the combined linearly polarized light, convert the light signals into electrical signals, and send the electrical signals to the data processing module; the multi-channel photoelectric detection module comprises a plurality of photoelectric detectors; The data processing module is configured to obtain scattering light intensities and polarization characteristics of different wavelengths and different scattering angles according to the electrical signals sent by the multi-channel photoelectric detection module, and inversely calculate a particle size distribution according to the scattering light intensities and polarization characteristics of different wavelengths and different scattering angles and scattering light intensities and polarization characteristics of different particle sizes and different wavelengths and different scattering angles provided by a Mie scattering simulation library, and estimate a particle concentration according to the particle size distribution; the polarization characteristics comprise a linear polarization degree and a linear polarization angle.

2. The particulate matter detection system of claim 1, wherein, The particle detection system further comprises an image acquisition module; The image acquisition module is arranged on the side wall of the detection cavity; The image acquisition module is configured to synchronously capture a scattering region, obtain a particle image, and send the particle image to the data processing module when the multi-channel photoelectric detection module obtains the light signals; The data processing module is further configured to obtain image features according to the particle image sent by the image acquisition module, and obtain a particle category based on a multi-modal feature fusion network by using the image features and the scattering light intensities and polarization characteristics of different wavelengths and different scattering angles; the particle category comprises a category of particle morphology, a category of particle optical characteristics, a category of particle dynamic characteristics, and a category of particle physicochemical characteristics.

3. The particulate matter detection system of claim 2, wherein, The image acquisition module adopts a high-resolution camera.

4. The particulate matter detection system of claim 3, wherein, The particle detection system further comprises a control and synchronization module; The control and synchronization module is configured to synchronously control wavelength switching, photoelectric detector sampling and high-resolution camera exposure.

5. The particulate matter detection system of claim 1, wherein, The light source module comprises a short-wave laser, a medium-wave laser, a long-wave laser, a near-infrared laser and a light combination device; The short-wave laser is configured to generate a short-wave light beam with an adjustable linear polarization direction; The medium-wave laser is configured to generate a medium-wave light beam with an adjustable linear polarization direction; The long-wave laser is configured to generate a long-wave light beam with an adjustable linear polarization direction; The near-infrared laser is configured to generate a near-infrared light beam with an adjustable linear polarization direction; The light combination device is configured to combine the short-wave light beam, the medium-wave light beam, the long-wave light beam and the near-infrared light beam with adjustable linear polarization directions to generate combined linearly polarized light.

6. The particulate matter detection system of claim 5, wherein, When the short-wave laser, the middle-wave laser, the long-wave laser and the near-infrared laser are arranged in the order of short to long or long to short in wavelength, the light combining device combines the short-wave beam, the middle-wave beam, the long-wave beam and the near-infrared beam with adjustable linear polarization directions by using a first dichroic mirror, a second dichroic mirror and a third dichroic mirror; When the short-wave laser, the middle-wave laser, the long-wave laser and the near-infrared laser are arranged in the order of short to long in wavelength, the first dichroic mirror is used to combine the short-wave beam and the middle-wave beam with adjustable linear polarization directions, the second dichroic mirror is used to combine the short-wave beam, the middle-wave beam and the long-wave beam with adjustable linear polarization directions, and the third dichroic mirror is used to combine the short-wave beam, the middle-wave beam, the long-wave beam and the near-infrared beam with adjustable linear polarization directions; When the short-wave laser, the middle-wave laser, the long-wave laser and the near-infrared laser are arranged in the order of long to short in wavelength, the first dichroic mirror is used to combine the near-infrared beam and the long-wave beam with adjustable linear polarization directions, the second dichroic mirror is used to combine the near-infrared beam, the long-wave beam and the middle-wave beam with adjustable linear polarization directions, and the third dichroic mirror is used to combine the near-infrared beam, the long-wave beam, the middle-wave beam and the short-wave beam with adjustable linear polarization directions.

7. The particulate matter detection system of claim 5, wherein, When the short-wave laser, the middle-wave laser, the long-wave laser and the near-infrared laser are not arranged in the order of short to long or long to short in wavelength, the light combining device combines the short-wave beam, the middle-wave beam, the long-wave beam and the near-infrared beam with adjustable linear polarization directions by using a combination of a light combining prism and a fourth dichroic mirror; The light combining prism is used to combine the short-wave beam, the middle-wave beam and the long-wave beam with adjustable linear polarization directions; The fourth dichroic mirror is used to combine the short-wave beam, the middle-wave beam, the long-wave beam and the near-infrared beam with adjustable linear polarization directions.

8. A particulate matter detection method characterized by, The particulate matter detection method is applied to the particulate matter detection system in any one of claims 1-7, and the particulate matter detection method comprises: obtaining an electrical signal sent by the multi-channel photoelectric detection module; obtaining scattering light intensity and polarization characteristics at different wavelengths and different scattering angles according to the electrical signal sent by the multi-channel photoelectric detection module; the polarization characteristics include linear polarization degree and linear polarization angle; obtaining a Mie scattering simulation library; inverting a particle size distribution based on a least square method, Bayesian inference or a neural network model according to the scattering light intensity and polarization characteristics at different wavelengths and different scattering angles and the Mie scattering simulation library providing scattering light intensity and polarization characteristics at different wavelengths and different scattering angles of different particle sizes; estimating a particle concentration according to the particle size distribution.

9. The particulate matter detection method of claim 8, wherein, The input of the Mie scattering simulation library is a series of particle sizes, complex refractive indexes, wavelengths and scattering angles; the output of the Mie scattering simulation library is relative concentrations at a series of particle sizes, and scattering light intensity, linear polarization degree and linear polarization angle at different wavelengths and different scattering angles of each particle size; the relative concentrations at a series of particle sizes are used to represent a particle size distribution.

10. The particulate matter detection method of claim 8, wherein, The particulate matter detection method further comprises: Obtaining the particle image sent by the image acquisition module; Obtaining image features according to the particle image sent by the image acquisition module; Using the image features and the scattering light intensity and polarization features of different wavelengths and different scattering angles, obtaining a particle category based on a multi-modal feature fusion network; the particle category includes a particle morphology category, a particle optical property category, a particle dynamic property category and a particle physicochemical property category.

Images