Bioaerosol monitoring device based on intrinsic fluorescence and polarized scattered light

Abstract

A bioaerosol monitoring device based on intrinsic fluorescence and polarized scattered light is disclosed. The device comprises a pump unit, a particle concentration unit, an optical detection unit, and a depth calculation model unit. It employs a single 405nm laser light source and, through an integrated optical path design, utilizes a dichroic mirror and a semi-transparent, semi-reflective beam splitter to divide the excitation light path into intrinsic fluorescence and scattered light paths. Horizontal and vertical polarizers are introduced into the scattered light path, enabling simultaneous measurement of the intrinsic fluorescence and horizontal and vertical polarized scattered light of individual aerosol particles. This method fuses multi-dimensional optical information to generate a unique digital fingerprint for each particle and employs a convolutional neural network (CNN) deep learning algorithm for identification and classification. This invention overcomes the errors of multi-path time-division detection, improves the classification accuracy and anti-interference capability of bioaerosols in complex environments, and also has the advantages of compact structure and low cost.

Description

Technical Field

[0001] This utility model relates to the field of real-time automatic optical monitoring technology for bioaerosols, and in particular to a bioaerosol monitoring device and method based on intrinsic fluorescence and polarized scattered light. Background Technology

[0002] Bioaerosol particles are ubiquitous in the atmospheric environment and play a crucial role in the Earth system, particularly in the interactions between the atmosphere, biosphere, and climate. Currently, among real-time bioaerosol monitoring technologies, ultraviolet-induced fluorescence bioaerosol monitors have achieved good application results in certain scenarios due to their advantages such as high signal-to-noise ratio and compact structure.

[0003] However, existing laser-induced intrinsic fluorescence detection methods can only preliminarily determine the type of bioaerosol and have weak anti-interference capabilities, failing to effectively eliminate interfering substances. While existing high-precision bioaerosol monitoring equipment, such as the Swiss SwisensPoleno pollen / bioaerosol automatic monitor, can achieve 24-hour continuous monitoring and identification, outputting data such as pollen concentration, type, area, perimeter, fluorescence intensity, and lifetime, it is expensive, particularly its digital holographic measurement unit camera. Furthermore, because it acquires data through multiple optical paths, errors can easily occur when particles pass through different optical paths.

[0004] Therefore, there is an urgent need to develop a bioaerosol monitoring device and method that is highly integrated, low-cost, has strong anti-interference capabilities, and can simultaneously acquire multi-dimensional optical information to improve identification accuracy and practicality. Summary of the Invention

[0005] To address the limitations of existing bioaerosol monitoring technologies, such as limited identification capabilities, system complexity, high costs, and susceptibility to interference, this invention provides an automated bioaerosol monitoring device. This device uses a particle concentrator to compress aerosol particle samples and employs ultraviolet-induced fluorescence technology to detect the intrinsic fluorescence, scattered light, and polarized light of biological particles. A proprietary database is built and expanded through machine learning algorithms. This invention can detect multiple information about biological particles, including intrinsic fluorescence, shape, surface roughness, and chemical composition. Based on this, each particle can form a unique digital fingerprint, thus creating a bioaerosol fingerprint database for the classification of bioaerosol particles.

[0006] The technical solution of this utility model is as follows:

[0007] A bioaerosol monitoring device based on intrinsic fluorescence and polarized scattered light, characterized in that it includes: an air pump unit for providing airflow power to inhale and exhale aerosols;

[0008] The particle concentration unit has its inlet connected to the sample air inlet and its outlet connected to the first air pump unit, and is used for pre-concentration and particle size screening of aerosol particles.

[0009] An optical detection unit, the inlet of which is connected to the collection hole of the particle concentration unit, is used to receive the concentrated aerosol particles and perform optical detection.

[0010] A depth computing model unit, connected to the optical detection unit, is used to process optical signals and identify and classify aerosol particles;

[0011] The optical detection unit includes:

[0012] A laser light source is used to emit excitation light to the optical detection area, irradiating individual aerosol particles flowing through this area;

[0013] A light trap is set within the optical detection area at a position opposite to the laser source to absorb transmitted laser light.

[0014] A light-collecting element is disposed next to the optical detection area to collect intrinsic fluorescence and scattered light excited by the aerosol particles;

[0015] A beam splitting assembly is disposed on the reflected light path of the light collecting element and is used to separate the collected composite light signal according to wavelength and light path. It includes a semi-transparent and semi-reflective beam splitter for beam splitting and a dichroic mirror for beam splitting.

[0016] A polarization component includes a first polarizer and a second polarizer, which are respectively disposed on different scattered light paths split by the beam splitter to generate scattered light signals with different polarization directions.

[0017] At least three photodetectors, including a first photodetector, a third photodetector, and a second photodetector, are used to receive the intrinsic fluorescence signal after beam splitting and polarization processing, the scattered light signal in the first polarization direction, and the scattered light signal in the second polarization direction, respectively, and synchronously convert the optical signal into an electrical signal.

[0018] The particle concentration unit includes a virtual impactor and a flow channel. The channel outlet is connected to the air pump unit, and the collection port is connected to the optical monitoring unit.

[0019] The optical detection unit includes a 405nm laser diode, a light trap, a parabolic mirror, a dichroic mirror, a semi-transparent beam splitter, a focusing lens, a horizontal polarizer, a vertical polarizer, an aperture, and a photodetector (PMT). Aerosol particles in the detection area are excited by the 405nm laser diode, generating intrinsic fluorescence and scattered light. The excited light is collected by the parabolic mirror and then incident parallel to the dichroic mirror. The intrinsic fluorescence passes through the dichroic mirror and then sequentially through the focusing lens and the aperture before reaching the first photomultiplier tube. The scattered light is reflected by the dichroic mirror and then passes through the semi-transparent beam splitter. Part of the scattered light passes through the semi-transparent beam splitter, is polarized by the horizontal polarizer, and then sequentially through the focusing lens and the aperture before reaching the second photomultiplier tube; another part of the scattered light is reflected by the semi-transparent beam splitter, is polarized by the vertical polarizer, and then sequentially through the focusing lens and the aperture before reaching the third photomultiplier tube.

[0020] The aforementioned deep computing model unit can combine the received detection results to form a proprietary fingerprint of aerosol particles and store it in a database, thereby classifying bioaerosols by comparing the data in the database.

[0021] This utility model also provides a method for monitoring bioaerosols based on the above-mentioned device, characterized by comprising the following steps:

[0022] The aerosols are concentrated by the particle concentration unit, and particles larger than the target particle size are focused into a beam of particles, which are then introduced into the optical detection unit through the collection hole to solve the error problem caused by inconsistent particle positions in multi-path detection.

[0023] A single laser light source is used to simultaneously irradiate a single aerosol particle, and its excited intrinsic fluorescence, horizontally polarized scattered light, and vertically polarized scattered light are simultaneously detected through a single integrated optical path.

[0024] The three synchronously detected optical signals are converted into electrical signals and transmitted to the depth calculation model unit;

[0025] The signal is processed using the depth computing model unit to generate a digital fingerprint that integrates multi-dimensional optical features;

[0026] Based on machine learning algorithms, the digital fingerprint is compared with a preset database to accurately classify bioaerosol particles, and the database is continuously expanded to improve recognition capabilities.

[0027] Compared with the prior art, the present invention has the following technical effects:

[0028] 1. The detection functions of scattered light, intrinsic fluorescence, polarized light and other information are integrated into the same optical path unit, eliminating the need for multiple light sources, detectors and optical components. This results in low cost, small size, easy integration and strong anti-interference ability.

[0029] 2. This device uses a single optical unit, which can simultaneously collect information such as scattered light, intrinsic fluorescence, and polarized light of a single particle under test. It does not require time-division acquisition of multiple optical information of the particle under test through trigger signals. The control system is simple and avoids the errors caused by using multiple optical path units to detect the flow of aerosol particles, thus preventing the particle under test from being missed or miscounted. Attached Figure Description

[0030] Figure 1 This is a schematic diagram illustrating the composition and working principle of the automatic bioaerosol monitoring device of this utility model.

[0031] Figure 2 This is a schematic diagram illustrating the composition and principle of the optical detection unit of the automatic bioaerosol monitoring device of this utility model. Detailed Implementation

[0032] The present invention will be further described below with reference to the accompanying drawings and embodiments, but this should not be construed as limiting the scope of protection of the present invention.

[0033] like Figure 1 and Figure 2 As shown, this embodiment provides an automatic bioaerosol monitoring device based on intrinsic fluorescence and polarized scattered light. The device includes an air pump unit, a particle concentration unit 2, an optical detection unit 4, and a depth calculation model unit, specifically configured as follows:

[0034] 1. Gas path and particulate concentration system:

[0035] Aerosol particles in the ambient air are drawn into the device through the sample inlet 1 under the negative pressure generated by the air pump unit (main air pump 5 and air pump 3). The airflow first enters the particle concentration unit 2, which in this embodiment adopts a virtual impactor structure. Its working process is as follows: the airflow carrying particles is ejected at high speed inside the impactor. Fine particles with an aerodynamic diameter smaller than the cutting particle size (which can be designed according to requirements, for example, 2.5 micrometers) have low inertia and are easily deflected by the airflow, ultimately being drawn away from the outlet by air pump 3 as waste air. Target particles larger than the cutting particle size (such as pollen, fungal spores, bacteria, etc.) have high inertia and are difficult to change direction, thus maintaining their original trajectory. They are concentrated and focused into a narrow beam, which is ejected through the collection hole at the end of the virtual impactor and directly sent to the detection area of ​​the optical detection unit 4. This process achieves particle concentration and preliminary screening, improving the target concentration and signal-to-noise ratio of subsequent optical detection.

[0036] 2. Integrated optical inspection system:

[0037] The optical detection unit 4 is the core of this utility model, such as Figure 2 As shown, the internal optical path connections and positional relationships are as follows:

[0038] Excitation Optical Path: In this embodiment, a laser diode 401 with a center wavelength of 405 nm is used as the excitation source. This wavelength can effectively excite most bioaerosol particles (such as tryptophan, tyrosine, riboflavin, etc.) to produce intrinsic fluorescence. The laser beam emitted by the laser is collimated, guided, and focused onto the center of the optical detection area. A light trap 402 is placed directly opposite the detection area to completely absorb the laser light transmitted through the particles, preventing it from being reflected back into the detection area and forming background interference.

[0039] Signal Collection and Spectroscopic Path: When concentrated individual aerosol particles pass sequentially through the detection area, they are irradiated by a 405nm laser, simultaneously emitting intrinsic fluorescence and elastic scattered light. A parabolic mirror 403, with its focal point located at the center of the detection area, efficiently collects the fluorescence and scattered light emitted by the particles at a larger solid angle and reflects them to form a parallel beam. This parallel beam first strikes a dichroic mirror 404. In this embodiment, the dichroic mirror is highly reflective of the 405nm laser and highly transparent to fluorescence with wavelengths greater than 420nm. Therefore, the intrinsic fluorescence component will pass through the dichroic mirror 404, then sequentially through the focusing lens 405 and the aperture 406, and finally be detected by the first photodetector 407. The aperture 406 limits the depth of field, ensuring that only light from near the focal point is received, thereby effectively eliminating stray light interference. The excited scattered light is reflected by the dichroic mirror 404, changing the direction of the optical path.

[0040] Polarization scattering analysis optical path: The scattered light reflected by the dichroic mirror 404 is then incident on the semi-transparent and semi-reflective beam splitter 408, which splits the incident beam into transmission and reflection paths in a certain ratio (e.g., 50:50).

[0041] One reflected light path: The scattered light reflected by the semi-transparent beam splitter 408 first passes through a vertical polarizer 409 to filter out the vertical polarization component (P-polarization). Then, this polarized light passes sequentially through the first focusing lens 410 and the first aperture 411, and is finally detected by the third photodetector 412. The other transmitted light path: The scattered light passing through the semi-transparent beam splitter 408 first passes through a horizontal polarizer 413 to filter out the horizontal polarization component (S-polarization). Then, this polarized light passes sequentially through the second focusing lens 414 and the second aperture 415, and is finally detected by the second photodetector 416.

[0042] By simultaneously reading the signals from the second and third photodetectors, the degree of polarization and polarization state of the light scattered by the particles can be calculated. This information is closely related to the shape, surface roughness, and internal structure of the particles.

[0043] 3. Signal Processing and Intelligent Recognition System:

[0044] The first, second, and third photodetectors (PMT1, PMT2, and PMT3) convert the synchronously acquired optical signals into electrical signals (such as voltage pulses). The peak value, integral area, and width of these pulses correspond to the fluorescence intensity, scattered light intensity, and approximate particle size of the particles, respectively.

[0045] These electrical signals are digitized by the data acquisition card and transmitted to the deep learning model unit (usually an embedded computer or server). This unit performs the following core operations:

[0046] Feature extraction and fingerprint construction: The pulse signal of each particle is analyzed to extract feature parameters, such as fluorescence intensity (from PMT1), total scattered light intensity (the sum of PMT2 and PMT3 signals), and polarization degree (calculated from the PMT2 and PMT3 signals). These features are then fused with particle size information to form a unique multi-dimensional "digital fingerprint" for each particle.

[0047] Database and Classification: The built-in machine learning model (preferably a convolutional neural network CNN in this embodiment) calls upon a pre-trained model and database. This database stores a large number of "digital fingerprint" features of biological particles of known categories (such as pollen from different tree species, fungal spores, etc.).

[0048] Network structure example:

[0049] Input layer: Receives an M×K feature matrix (M is the number of angles, K is the number of polarization directions);

[0050] Convolutional layers: using 3×3 or 5×5 convolutional kernels to extract local features (such as the polarization intensity variation pattern within a certain angle range, corresponding to the protrusions or depressions of particles);

[0051] Pooling layer: Compresses features using max pooling or average pooling while preserving key information such as peak positions;

[0052] Fully connected layer: maps pooled features to a preset shape category;

[0053] Output layer: Using the Softmax activation function, outputs the probability of each shape category.

[0054] Model Algorithm Flow: The "angle-polarization" scattering information of each particle is constructed into a two-dimensional feature matrix and input into a CNN model. The model extracts local features (such as the polarization signal variation pattern with angle) through convolutional layers, compresses data and retains key information through pooling layers, and finally outputs the probability of the particle belonging to each preset category through fully connected layers and a Softmax classifier. During training, the model uses the Adam optimizer and Dropout to prevent overfitting, ensuring good generalization ability on unknown samples. Normal testing uses cross-entropy loss to measure the difference between the predicted and true categories; mean squared error (MSE) or mean absolute error (MAE) is used to minimize the deviation between the predicted shape parameters and the true values. The model is divided into training, validation, and test sets in a 7:2:1 ratio. The training set is used for model learning, the validation set for parameter tuning, and the test set for evaluating the final performance. An adaptive moment estimation (Adam) or stochastic gradient descent (SGD) optimizer is used to dynamically adjust the learning rate; random deactivation (Dropout) and weight decay (L2 regularization) are used to avoid overfitting and ensure the model's generalization ability on new samples.

[0055] Results Output and Self-Learning: Classification results (such as particle type, concentration, size distribution, fluorescence intensity distribution, etc.) are displayed and stored in real time. Simultaneously, the system can set "unknown particles" tags and archive the feature data of these particles. After manual verification, these data can be used for incremental learning, continuously expanding and optimizing the identification database, and improving the device's adaptability to complex environments.

[0056] This embodiment integrates all optical components onto a single optical platform, arranged around a common detection point. When a particle passes through, all its excited optical signals (fluorescence, scattered light of different polarization states) are captured instantaneously and synchronously. Using only a single laser source and a set of optical components significantly reduces system cost and size, while enhancing system stability and anti-interference capabilities. It enables the synchronous acquisition of multiple optical parameters from a single aerosol particle, avoiding signal asynchrony and errors caused by time-division multiplexing or multi-path measurements, effectively preventing missed or misjudged particles. All acquired optical features are strictly derived from the physical properties of the particle at the same time, spatial location, and orientation, with highly accurate correlation between features, laying a reliable data foundation for subsequent accurate classification. It eliminates the need for complex multi-path calibration and timing synchronization control systems, reducing the complexity of mechanical and electronic design.

[0057] The monitoring method of the aforementioned bioaerosol monitoring device is as follows:

[0058] 1. Calibrate the particle size information of the scattered light path;

[0059] 2. Collect a large amount of information on particle size range, fluorescence information, and polarized light of different types of bioaerosol particles to form a database.

[0060] 3. Machine deep learning is employed to transform the "angle-polarization" distribution of scattered light into a two-dimensional matrix. The local feature extraction capability of a convolutional neural network (CNN) is then utilized to capture shape-related patterns. The process involves receiving an M×K feature matrix (M being the number of angles and K being the number of polarization directions); extracting local features using convolutional kernels; compressing features through max pooling or average pooling while preserving key information; mapping the pooled features to preset shape categories; and outputting the probability of each shape category using a normalized exponential function (Softmax activation function).

[0061] 4. Conduct pattern recognition and normal testing. Use cross-entropy loss to measure the difference between the predicted and true classes. Use mean squared error (MSE) or mean absolute error (MAE) to minimize the deviation between the predicted shape parameters and the true values. Divide the model into training, validation, and test sets. The training set is used for model learning, the validation set is used to adjust parameters, and the test set is used to evaluate the final performance. Use an adaptive moment estimation (Adam) or stochastic gradient descent (SGD) optimizer to dynamically adjust the learning rate. Avoid overfitting through dropout and weight decay (L2 regularization) to ensure the model's generalization ability on new samples.

[0062] The working process of the aforementioned automatic bioaerosol monitoring device is as follows:

[0063] 1) Bioaerosol particles enter inlet 1 and pass through the virtual impactor of the particle concentration unit 2. A portion of the separated fine aerosol particles with a cutting particle size below the cutting particle size are extracted by the first air pump unit 3, while aerosol particles with a cutting particle size above the cutting particle size maintain their original direction and enter the collection hole, and then enter the optical detection unit 4.

[0064] 2) The optical detection unit starts working, and the laser emitted by the 405nm laser diode 401 excites the aerosol particles in the detection area. The first photomultiplier tube 407 obtains the intrinsic fluorescence signal of the aerosol particles, the second photomultiplier tube 416 obtains the state of horizontally polarized light, and the third photomultiplier tube 412 obtains the state of vertically polarized light. By detecting the changes in the degree of polarization and polarization direction of the scattered light in different directions, the microstructure and overall shape information of the aerosol particle surface are obtained.

[0065] 3) The fluorescence information, microstructure and overall shape information of the aerosol particle surface are sent to the depth computing model unit to form a unique digital fingerprint, which includes particle size, morphology data and data related to particle fluorescence. Then, the fingerprint is processed and classified by machine learning algorithm.

[0066] This invention utilizes the same laser beam to simultaneously excite biofluorescence and as a probe light for polarization scattering measurement. Through an optical beam splitter and filter architecture (dichroic mirror + semi-transparent mirror), a single optical flow is separated into three independent signals, and horizontal and vertical polarizers are introduced into the two scattered light paths. This achieves multi-parameter, high-precision, and high-reliability detection performance—previously only achievable with large, expensive equipment—on a low-cost, compact platform. It effectively addresses the core pain points of current real-time bioaerosol monitoring in terms of cost, error, and anti-interference capabilities, demonstrating significant technological advancement and market application potential.

[0067] The above description is merely a specific embodiment of this utility model, but the protection scope of this utility model is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in this utility model should be included within the protection scope of this utility model.

Claims

1. A bioaerosol monitoring device based on intrinsic fluorescence and polarized scattered light, characterized in that, include: An air pump unit is used to provide airflow power to draw in and expel aerosols; The particle concentration unit (2) has its inlet connected to the sample air inlet (1) and its outlet connected to the first air pump unit (3), and is used for pre-concentration and particle size screening of aerosol particles. An optical detection unit (4) is connected to the collection hole of the particle concentration unit (2) for receiving concentrated aerosol particles and performing optical detection. The depth calculation model unit is connected to the optical detection unit (4) and is used to process optical signals and identify and classify aerosol particles; The optical detection unit (4) includes: A laser source (401) is used to emit excitation light to the optical detection area to irradiate individual aerosol particles flowing through the area; A light trap (402) is disposed in the optical detection area opposite to the laser source (401) for absorbing transmitted laser light; A light-collecting element is disposed next to the optical detection area to collect intrinsic fluorescence and scattered light excited by the aerosol particles; A beam splitting assembly is disposed on the reflected light path of the light collecting element and is used to separate the collected composite light signal according to wavelength and light path. It includes a semi-transparent and semi-reflective beam splitter (408) for beam splitting and a dichroic mirror (404) for beam splitting. The polarization component includes a first polarizer (409) and a second polarizer (413), which are respectively disposed on different scattered light paths split by the beam splitting component to generate scattered light signals with different polarization directions; At least three photodetectors, including a first photodetector (407), a third photodetector (412), and a second photodetector (416), are used to receive the intrinsic fluorescence signal after beam splitting and polarization processing, the scattered light signal in the first polarization direction, and the scattered light signal in the second polarization direction, respectively, and synchronously convert the optical signal into an electrical signal.

2. The bioaerosol monitoring device based on intrinsic fluorescence and polarization scattering light of claim 1, wherein, The light-collecting element is a parabolic mirror (403), whose focal point is located in the optical detection area, used to reflect scattered light and fluorescence as parallel light for emission.

3. The bioaerosol monitoring device based on intrinsic fluorescence and polarization scattering light of claim 2, wherein, The dichroic mirror (404) is disposed on the parallel light path reflected by the parabolic mirror (403) and is configured to transmit intrinsic fluorescence of a specific wavelength range to the first photodetector (407) and reflect scattered light of other wavelengths; the semi-transparent and semi-reflective beam splitter (408) is disposed on the scattered light path reflected by the dichroic mirror (404) and is configured to split the incident scattered light beam into a transmitted light beam and a reflected light beam.

4. The bioaerosol monitoring device based on intrinsic fluorescence and polarization scattering light of claim 3, wherein, The first polarizer is a vertical polarizer (409), which is disposed on the path of the scattered light reflected by the semi-transparent and semi-reflective beam splitter (408). The light signal is received by the third photodetector (412) after passing through the focusing lens (410) and the aperture (411). The second polarizer is a horizontal polarizer (413), which is disposed on the path of the scattered light transmitted through the semi-transparent and semi-reflective beam splitter (408). The light signal is received by the second photodetector (416) after passing through the second focusing lens (414) and the second aperture (415).

5. The bioaerosol monitoring device based on intrinsic fluorescence and polarization scattering light of claim 4, wherein, The intrinsic fluorescence passes through the dichroic mirror (404), then through a first focusing lens (405) and a first diaphragm (406), and is finally received by the first photodetector (407).

6. The bioaerosol monitoring device based on intrinsic fluorescence and polarization scattering light of claim 1, wherein, The particle concentration unit (2) is a virtual impactor, and the collection hole is used to make the concentrated particle flow keep the original movement direction to enter the optical detection unit (4).

Images