A hyperspectral diffractive airborne disease spore discrimination device and method

Abstract

The application provides a hyperspectral diffraction type airborne disease spore identification device and identification method, and belongs to the technical field of agricultural disease monitoring and optical detection; the application adopts a single folding light path of "incident slit-collimating mirror-concave reflection grating-focusing mirror-micro hole-sample-CMOS", integrates monochromatic illumination and diffraction imaging on the same optical axis, is simple in structure, easy to assemble and adjust, and convenient to package in a small-sized shell, and is suitable for deployment of a weather station and a mobile platform; the device can realize cost controllable, compact and stable imaging hyperspectral diffraction detection, and provides technical support for early warning and on-site rapid diagnosis of crop airborne diseases.

Description

Technical Field

[0001] This invention belongs to the field of agricultural disease monitoring and optical detection technology, and relates to a hyperspectral diffraction-based device and method for identifying airborne disease spores. Background Technology

[0002] Typical airborne fungal diseases such as wheat rust and corn rust release large numbers of pathogenic spores into the atmosphere during their occurrence and spread by wind, leading to significant yield reductions in field crops. Changes in the concentration of airborne pathogen spores in the air often precede the appearance of field symptoms. Therefore, continuous monitoring and accurate identification of spores are crucial for early warning and precise control of crop diseases.

[0003] The current monitoring and identification of spores of airborne crop diseases mainly relies on microscopic and molecular detection techniques under laboratory conditions. For example, airborne spores are collected using volumetric spore traps, then prepared and stained on slides, and finally interpreted by professionals under optical or fluorescence microscopes based on morphological characteristics; or specific nucleic acid sequences are detected using molecular biology methods such as PCR. While these methods are relatively reliable in terms of accuracy, they all suffer from problems such as numerous operational steps, long detection cycles, high dependence on personnel, and difficulty in timely reflecting the real-time dynamics of spores in the field, making them unsuitable for routine monitoring at monitoring stations and in production areas.

[0004] With the development of photoelectric detection and data processing technologies, some methods for identifying pathogenic spores based on imaging and spectral information have gradually emerged. For example, spore morphology images are acquired using ordinary digital imaging or microscopic imaging systems, and then classified using feature extraction and machine learning algorithms. Other studies have introduced microscopic hyperspectral imaging, using gratings or tunable filters to record the spectral response of spores in the visible-near-infrared band to distinguish different pathogens. However, existing microscopic hyperspectral equipment is mostly research-grade, typically employing complex dispersive structures and high-sensitivity dedicated detectors (such as cooled CCDs and InGaAs arrays). These devices are large, have complex optical paths, and are costly, requiring strict environmental stability and installation conditions, making large-scale deployment in pest and disease monitoring stations and mobile monitoring platforms difficult. Such systems primarily focus on the spectral dimension, underutilizing physical information such as diffraction imaging that reflects microstructure and scattering characteristics. Their ability to distinguish airborne pathogenic spores with similar morphology and subtle spectral differences remains limited. Furthermore, existing online optical detection equipment for aerosol particles mostly uses single-wavelength or a few-wavelength scattered light intensity measurement, or obtains the target contour through simple transmission imaging. The information dimensions obtained are limited, making it difficult to simultaneously take into account microscopic morphology, scattering characteristics, and spectral features. Existing systems also rarely consider the synergistic optimization with portable packaging, heat dissipation structure, and general image sensors in their optical path design, resulting in problems such as insufficient imaging stability, high thermal noise, and high maintenance costs during actual long-term operation.

[0005] In summary, existing technologies for the on-site identification of airborne disease spores generally suffer from several drawbacks, including complex optical path structures, high costs, and limitations in portability and engineering application; limited information acquisition dimensions and insufficient ability to distinguish spores with similar morphologies; and a lack of integrated consideration in heat dissipation, structural design, and optical imaging modules. There is an urgent need to develop an optical identification device for airborne disease spores that features a simple and integrable optical path structure, can simultaneously acquire rich spectral information and diffraction characteristics, and balances cost and operational stability, to meet the application needs of early warning and rapid on-site diagnosis of crop diseases. Summary of the Invention

[0006] To address the problems of complex optical path structures, high costs, difficulty in portable deployment, and insufficient ability to distinguish morphologically similar spores in existing airborne disease spore monitoring devices, this invention provides a hyperspectral diffraction-based airborne disease spore identification device and method. This invention employs a single folded optical path of "incident slit – collimating lens – concave reflection grating – focusing lens – micropore – sample – CMOS," integrating monochromatic illumination and diffraction imaging on the same optical axis. The structure is simple, easy to assemble and adjust, and readily packaged in a miniaturized housing, making it suitable for deployment in monitoring stations and mobile platforms. The device described in this invention enables cost-effective, compact, and stable hyperspectral diffraction detection, providing technical support for early warning and rapid on-site diagnosis of crop airborne diseases.

[0007] To achieve the above-mentioned technical objectives, the present invention employs the following technical means:

[0008] The present invention first provides a hyperspectral diffraction-based device for identifying airborne disease spores, the device comprising: a hyperspectral diffraction imaging device housing and a host computer disposed outside the hyperspectral diffraction imaging device housing;

[0009] The hyperspectral diffraction imaging device integrates a light source module, an integrated spectrometer module, a light shield, a CMOS sensor, a nanoscale displacement stage, and a Z-axis displacement stage inside its housing.

[0010] The light source module includes a light source and an optical fiber for emitting light from the light source;

[0011] The integrated spectrometer module includes an optical path integrated housing, on which an entrance slit is arranged coaxially with the light-emitting end of the optical fiber, so that the polychromatic light output from the optical fiber is perpendicularly irradiated by the entrance slit. Inside the optical path integrated housing, along the optical axis, there are sequentially arranged a collimating lens, a concave reflection grating, and a focusing lens to process the polychromatic light input from the entrance slit; the bottom of the optical path integrated housing is provided with a light-emitting hole.

[0012] The top of the light shield has micro-holes, which correspond to the positions of the light-emitting holes; the middle of the interior of the light shield has a sample layer for placing the sample to be tested; and the lower end of the interior of the light shield has a CMOS sensor.

[0013] The light shield is fixed on the nanoscale displacement stage, which is fixed to the moving end of the Z-axis displacement stage; the Z-axis displacement stage is fixed to the bottom of the hyperspectral diffraction-type airborne disease spore identification device.

[0014] The host computer is electrically connected to the CMOS sensor and is used to receive the diffraction image sequence acquired by the CMOS sensor, execute signal processing and identification algorithms, and output the spore category identification result.

[0015] Preferably, the light source of the light source module is a hyperspectral halogen light source with an output spectral range covering 380–1000 nm.

[0016] Preferably, the concave reflective grating has a line density of 600 lines / mm and a radius of curvature of -300 mm;

[0017] The radius of curvature of the collimating lens is -180 mm;

[0018] The radius of curvature of the focusing lens is -210 mm;

[0019] The collimating lens is positioned on the exit side of the entrance slit. The concave reflective grating is positioned behind the collimating lens and receives the collimated polychromatic beam at an incident angle of 30°. The focusing lens is positioned on the dispersive beam path of the concave reflective grating and is used to perform secondary focusing on the dispersive beam to obtain a focused dispersive beam. The focused dispersive beam forms a focused monochromatic spot at the micro-aperture that matches the aperture of the micro-aperture.

[0020] Preferably, the light shield is a closed hollow structure, and its upper end is provided with a micro-hole adjustment block for fixing the micro-holes, and the micro-holes are arranged on the micro-hole adjustment block;

[0021] The inner wall of the light shield is treated with a matte finish.

[0022] A metal heat sink is fixedly connected to the bottom of the CMOS sensor;

[0023] The distance between the micropore and the upper surface of the sample layer is maintained at 50-70 mm, and the distance between the lower surface of the sample layer and the CMOS sensor is 2-6 mm, preferably 3 mm.

[0024] Preferably, a heat sink is also provided on the lower side of the CMOS sensor;

[0025] The heat sink includes a thermoelectric cooler, a finned heat sink, and a fan assembly; the cold end of the thermoelectric cooler is in close contact with the metal heat sink base plate, and the hot end is in close contact with the finned heat sink; the fan assembly is located at the airflow channel of the finned heat sink.

[0026] Preferably, an auxiliary heat sink is also provided on the outer side of the metal heat sink base plate.

[0027] Preferably, the host computer includes a signal processing and identification module; the signal processing and identification module includes:

[0028] Hyperspectral diffraction fingerprint building unit: used to combine diffraction images acquired at different wavelengths to form a multidimensional data cube of "wavelength-space diffraction image";

[0029] Feature extraction unit: used for band selection and spatial feature extraction of the data cube;

[0030] The classification and identification unit uses a machine learning classification model to identify different airborne disease spores and outputs the spore category and corresponding confidence level.

[0031] Preferably, the machine learning classification model is trained using data of different categories of airborne disease spores extracted by the feature extraction unit, and the machine learning classification model includes an SVM classification model.

[0032] The present invention also provides an identification method for the above-mentioned hyperspectral diffraction-based airborne disease spore identification device, the identification method comprising:

[0033] S1. Prepare the airborne spore sample to be tested on a glass slide and place the glass slide in the sample layer in the middle of the light shield so that the area to be tested is directly below the micropore;

[0034] S2. Coarse adjustment is performed using a Z-axis displacement stage, and fine adjustment is performed using a nanometer-scale displacement stage. The relative height between the micropore and the sample layer, as well as the relative distance between the sample layer and the CMOS sensor, are adjusted so that the focused spot corresponding to the spectral focusing line can stably enter the micropore and form a clear diffraction pattern on the CMOS sensor.

[0035] S3. Control the integrated spectrometer module to perform wavelength scanning within the preset working band range of 380-1000nm, so that the dispersive beam generated by the concave reflection grating is focused twice by the focusing lens to form a focused monochromatic light spot at the micro-hole. The micro-hole spatially confines the light spot and outputs a narrow-band monochromatic illumination beam to irradiate the sample layer.

[0036] S4. Under monochromatic illumination conditions corresponding to each wavelength, the corresponding diffraction images are acquired by the CMOS sensor, and a diffraction image sequence is formed in wavelength order. Then, the obtained diffraction image sequence is transmitted to the host computer.

[0037] S5. The signal processing and identification module in the host computer preprocesses the diffraction image sequence. After preprocessing, the diffraction image sequence is combined in the hyperspectral diffraction fingerprint construction unit to form a multidimensional data cube of "wavelength-space diffraction image". Then, the feature vector for classification and recognition is extracted in the feature extraction unit.

[0038] S6. Use the feature vectors obtained after processing different types of airborne disease spores in steps S1-S5 to train a classification and recognition model;

[0039] The feature vector of the sample to be tested is input into the trained classification and recognition model, which outputs the spore category and the corresponding confidence level.

[0040] Preferably, in step S5, the preprocessing includes denoising, background subtraction, normalization, and smoothing in sequence.

[0041] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0042] (1) The hyperspectral diffraction-type airborne disease spore identification device of the present invention adopts a single folded optical path of “incident slit – collimating lens – concave reflection grating – focusing lens – micro-hole – sample – CMOS”, which integrates monochromatic illumination and diffraction imaging on the same optical axis. It has a simple structure, is easy to assemble and adjust, and is easy to encapsulate in a miniaturized housing, making it suitable for deployment in monitoring stations and mobile platforms.

[0043] (2) This invention uses Zemax software for optical simulation, employs a concave reflection grating with 600 lines / mm, and ensures that the dispersive beam forms a focused spot matching the micropore diameter within the preset working wavelength range by adjusting the curvature radius of the collimating lens and the focusing lens. Simulation results show that when the micropore diameter is 300 μm, it can provide sufficient light intensity and effectively avoid diffraction pattern overlap, ensuring the accuracy of spore identification.

[0044] (3) This invention utilizes a more cost-effective CMOS sensor in conjunction with a metal heat sink base and a simple heat dissipation structure of semiconductor cooling chip-fan. Without introducing complex cooling and temperature control circuits, it effectively reduces thermal noise, achieves a balance between imaging quality and system cost, and makes the overall cost of the device significantly lower than that of traditional microscopic hyperspectral systems, which is convenient for large-scale promotion and application. Attached Figure Description

[0045] Figure 1 This is a schematic diagram of the overall structure of a hyperspectral diffraction-based airborne spore identification device proposed in this invention.

[0046] Figure 2 This is a schematic diagram of the integrated spectrometer module casing of a hyperspectral diffraction-based airborne spore identification device proposed in this invention.

[0047] Figure 3 This is a schematic diagram of the light shield shell of a hyperspectral diffraction-based airborne spore identification device proposed in this invention.

[0048] Figure label:

[0049] 1-Z-axis displacement stage; 2-Nanoscale displacement stage; 3-Heat dissipation housing support; 4-Heat dissipator; 5-Metal heat dissipation base plate; 6-CMOS sensor; 7-Sample layer; 8-Light shield housing; 9-Micro-orifice adjustment block; 10-Micro-orifice; 11-Hyperspectral diffraction imaging device housing; 12-Focused dispersive beam; 13-Optical path integrated housing; 14-Dispersive beam; 15-Focusing lens; 16-Collimating lens; 17-Collimated polychromatic beam; 18-Diverging polychromatic beam; 19-Concave reflection grating; 20-Incident slit; 21-Fiber optic cable; 22-Light source; 23-Host computer.

[0050] Figure 4 This is a schematic diagram of the optical path simulation of a hyperspectral diffraction-based airborne spore identification device.

[0051] Figure 5 This is a schematic diagram of the simulated spectral resolution results of a hyperspectral diffraction-based airborne spore identification device proposed in this invention.

[0052] Figure 6This is a schematic diagram of the spectral data of wheat leaf rust spores and corn leaf rust spores after smoothing.

[0053] Figure 7 This is a schematic diagram of the spectral data of wheat leaf rust spores after smoothing.

[0054] Figure 8 This is a schematic diagram of the spectral data of corn leaf rust spores after smoothing.

[0055] Figure 9 This is a schematic diagram showing the relationship between the microaperture size and light intensity changes on the diffraction imaging module.

[0056] Figure 10 This is a schematic diagram of the data on the sample-micropore distance, showing the first minimum value of the radial spectrum for different micropore sizes on the diffraction imaging module.

[0057] Figure 11 This is a schematic diagram of the data on the first minimum value of the radial spectrum and the sample-CMOS distance on the diffraction imaging module when the sample-micropore distance Z1 is limited to 62 mm.

[0058] Figure 12 This is a schematic diagram of the confusion matrix results of the classification model of this invention on the test set for the validation of maize leaf rust spores, wheat leaf rust spores and microsphere samples.

[0059] Figure 13 This is the host computer software interface of the present invention. Detailed Implementation

[0060] The present invention will be further described below with reference to the accompanying drawings and specific embodiments, but the scope of protection of the present invention is not limited thereto.

[0061] Example 1:

[0062] like Figure 1-3 As shown, this embodiment provides a hyperspectral diffraction-based airborne disease spore identification device, which includes: a hyperspectral diffraction imaging device housing 11 and a host computer 23 disposed outside the hyperspectral diffraction imaging device housing 11;

[0063] The hyperspectral diffraction imaging device housing 11 integrates a light source module, an integrated spectrometer module, a light shield, a CMOS sensor 6, a nanoscale displacement stage 2, and a Z-axis displacement stage 1.

[0064] The light source module includes a light source 22 and an optical fiber 21 for emitting light output from the light source 22;

[0065] The integrated spectrometer module includes an optical path integrated housing 13. An entrance slit 20 is provided on the optical path integrated housing 13 and arranged coaxially with the light-emitting end of the optical fiber 21, so that the polychromatic light output from the optical fiber 21 is perpendicularly irradiated into the entrance slit 20. Inside the optical path integrated housing 13, along the optical axis, there are sequentially arranged a collimating lens 16, a concave reflection grating 19, and a focusing lens 15 to process the diverging polychromatic beam 18 input from the entrance slit 20. A light-emitting hole is provided at the bottom of the optical path integrated housing 13.

[0066] The top of the light shield 8 is provided with a micro-hole 10, which corresponds to the position of the light-emitting hole; the middle part of the interior of the light shield 8 is provided with a sample layer 7 for placing the sample to be tested; the lower end of the interior of the light shield 8 is provided with a CMOS sensor 6.

[0067] The light shield 8 is fixed on the nanoscale displacement stage 2, which is fixed to the moving end of the Z-axis displacement stage 1; the Z-axis displacement stage 1 is fixed to the bottom of the hyperspectral diffraction-type airborne disease spore identification device.

[0068] The host computer 23 is electrically connected to the CMOS sensor 6 and is used to receive the diffraction image sequence acquired by the CMOS sensor 6, execute signal processing and identification algorithms, and output the spore category identification result.

[0069] In the specific implementation process, the light source 22 of the light source module is a hyperspectral halogen light source with an output spectral range covering 380–1000 nm; the curvature radii and installation positions of the collimating lens 16, the concave reflection grating 19, and the focusing lens 15 are customized based on optical simulation results. Simulation results show that when the grating density of the concave reflection grating 19 is 600 lines / mm and the curvature radius is -300 mm, the curvature radius of the collimating lens 16 is -180 mm, and the curvature radius of the focusing lens 15 is -210 mm, the desired curvature is achieved. When the diameter is mm, the collimating lens 16 is positioned on the exit side of the entrance slit 20, the concave reflective grating 19 is positioned behind the collimating lens 16 and receives the collimated polychromatic beam 17 at an incident angle of 30°, and the focusing lens 15 is positioned on the path of the dispersive beam 14 of the concave reflective grating 19 to perform secondary focusing on the dispersive beam 14 to obtain a focused dispersive beam 12. The focused dispersive beam 12 forms a focused monochromatic light spot at the micro-aperture 10 and matches the aperture of the micro-aperture 10.

[0070] In the specific implementation process, the light shield 8 is a closed hollow structure, with a micro-hole adjustment block 9 at its upper end for fixing the micro-hole 10. The micro-hole 10 is set on the micro-hole adjustment block 9. The inner wall of the light shield 8 is treated with an anti-glare coating. A metal heat sink base plate 5 is fixedly connected to the CMOS sensor 6 at the lower end inside the light shield 8. The distance between the micro-hole 10 and the upper surface of the sample layer 7 is maintained at 50-70 mm, and the distance between the lower surface of the sample layer 7 and the CMOS sensor is 2-6 mm, preferably 3 mm. The lower part of the light shield 8 is also provided with a heat sink housing support 3, which is fixedly connected to the Z-axis displacement stage 1 or the housing 11 of the hyperspectral diffraction imaging device. It is used to install, support, and fix the components where the heat sink 4, the metal heat sink base plate 5, and the CMOS sensor 6 are located, so as to ensure the relative positional stability of the heat dissipation structure and the imaging structure.

[0071] In the specific implementation process, a heat sink 4 is provided on the lower side of the CMOS sensor 6; the heat sink 4 includes a thermoelectric cooler, a finned heat sink, and a fan assembly. The CMOS sensor 6 is tightly attached to the metal heat sink base 5 with thermal grease to achieve efficient heat conduction. The thermoelectric cooler is disposed between the metal heat sink base 5 and the finned heat sink, with its cold end in close contact with the metal heat sink base 5 and its hot end in close contact with the finned heat sink, used to pump the heat conducted from the metal heat sink base to the finned heat sink side. The fan assembly is disposed at the airflow channel of the finned heat sink to perform forced convection cooling of the finned heat sink, dissipating the heat from the hot end in a timely manner, thereby stabilizing the operating temperature of the CMOS sensor 6 within a preset range, reducing dark current and thermal noise, and improving the signal-to-noise ratio and long-term imaging stability of the diffraction image. An auxiliary heat sink is also provided on the outer side of the metal heat sink base 5 to further dissipate heat through natural convection without the need for complex temperature control circuitry, thereby reducing thermal noise and improving the quality of the diffraction image.

[0072] In the specific implementation process, the host computer 23 is equipped with a signal processing and identification module; the signal processing and identification module includes: a hyperspectral diffraction fingerprint construction unit: used to combine diffraction images acquired at different wavelengths to form a multi-dimensional data cube of "wavelength-spatial diffraction image"; a feature extraction unit: used to perform band selection and spatial feature extraction on the data cube; and a classification and identification unit, which uses a machine learning classification model to identify different airborne disease spores and outputs the spore category and corresponding confidence level. The machine learning classification model is trained using data of different airborne disease spores extracted by the feature extraction unit, and the machine learning classification model is preferably an SVM classification model.

[0073] In specific embodiments, the micropores 10 are circular holes machined on metal or stainless steel sheets, with a diameter selectable from 200 to 300 μm. Combined with... Figures 9-11 The simulation and experimental data shown demonstrate that when the micropore diameter is approximately 300 μm, sufficient light transmission can be ensured while achieving a partially coherent illumination field under narrow wavelength bandwidth conditions. This reduces overlap and blurring between different spore diffraction patterns, improving the contrast and stability of the diffraction pattern. Through this structural design, the 380–1000 nm polychromatic light from the light source 22, after dispersion and focusing by the fiber optic cable 21, the entrance slit 20, the collimating lens 16, the concave diffraction grating 19, and the focusing lens 15, forms a wavelength-tunable focused spot at the micropore. The micropore spatially confines this spot, outputting narrowband monochromatic illumination to irradiate the airborne pathogenic spores on the sample layer 7 below, forming diffraction patterns. The diffraction patterns are then acquired by the CMOS sensor 6 below the sample layer, resulting in a diffraction image sequence. With this structure, the device of this invention can provide narrowband monochromatic illumination of spore samples at different wavelengths and acquire diffraction image sequences, providing a data foundation for subsequent hyperspectral diffraction fingerprint construction and spore category identification.

[0074] Example 2:

[0075] To verify the focusing capability and spectral resolution of the optical path system of the hyperspectral diffraction-type airborne disease spore identification device described in Example 1 within the working wavelength band, this example simulates and tests the folded optical path of "incident slit – collimating lens – concave reflection grating – focusing lens – micro-aperture".

[0076] First, the optical path of "entry slit—collimating lens—concave reflective grating—focusing lens—microaperture" was modeled and simulated using Zemax optical design software. The light source wavelength range was set to 380–1000 nm, and the concave reflective grating line density was set to 600 lines / mm. Based on the simulation, the curvature radius and installation position parameters of the collimating lens and focusing lens were optimized to ensure that the dispersive beam forms a focused spot at the microaperture. The simulation results are as follows. Figure 4 As shown in the figure, a single folded optical path can achieve stable wavelength dispersion and secondary focusing within a preset wavelength band, ensuring that the focused spot can effectively enter the micro-aperture, providing an optical basis for subsequent narrowband monochromatic illumination and diffraction imaging.

[0077] This embodiment also calculates spectral line broadening at different wavelengths based on slit width, grating parameters, and image plane position in Zemax simulation, and uses software to measure the emitted monochromatic light after actual assembly to obtain the spectral resolution range of the system within the operating wavelength band. The results are as follows: Figure 5 As shown in the figure, the hyperspectral diffraction-based airborne disease spore identification device described in Example 1 can achieve an effective spectral resolution of approximately 2–4 nm in the 380–1000 nm wavelength range, which meets the narrowband illumination requirements for constructing spore diffraction fingerprints.

[0078] Example 3:

[0079] To balance light transmittance, partial coherence, and diffraction fringe contrast, this embodiment examines and optimizes the micropore size, sample-micropore distance Z1, and sample-CMOS distance Z2 of the hyperspectral diffraction-based airborne spore identification device described in Example 1. The specific steps are as follows:

[0080] (1) Relationship between micropore size and light transmission intensity:

[0081] Microporous sheets with different aperture sizes (e.g., 200 μm, 300 μm, etc.) were selected, and diffraction images were acquired under the same light source power and assembly conditions. The light transmission intensity was characterized by the average gray level / total intensity of the diffraction images. The results are as follows: Figure 9 As shown. The results indicate that increasing the micropore size can improve the light transmission, but an excessively large pore size will weaken the spatial confinement effect and may increase the risk of diffraction pattern overlap; considering both light transmission and fringe contrast, the preferred micropore size in this invention is approximately 300 μm.

[0082] (2) Relationship between the first minimum value of the radial spectrum and Z1 under different pore sizes: Adjust the sample-micropore distance Z1 (e.g., within the range of 58–62 mm) under different micropore sizes, acquire diffraction images and calculate their radial intensity spectra, extract the position R1 (pixels) of the first minimum value of the radial spectrum, and obtain the results. Figure 10 As shown. The results indicate that the device can achieve imaging within the range of 50–70 mm. Among them, the diffraction fringe sharpness and contrast are better when the sample-microaperture distance is about 62 mm. Therefore, 62 mm can be regarded as the preferred working distance in this embodiment. R1 responds regularly to the change of Z1, indicating that the diffraction fringe scale can be changed by adjusting Z1 and affecting the pattern contrast. It can be used to determine the preferred working distance of the device. The preferred value is 62 mm.

[0083] (3) Relationship between the first minimum and Z2 under the constraint of Z1: Under the constraint of Z1 being 62 mm, the sample-CMOS distance Z2 was adjusted (e.g., within the range of 2.0–3.4 mm), diffraction images were acquired and the change of R1 was extracted. The results are as follows: Figure 11 As shown in the figure. The results indicate that, under the condition of fixed Z1, optimizing Z2 to approximately 3 mm can yield the optimal imaging position with clearer diffraction fringes and higher contrast, thereby improving the signal-to-noise ratio of the diffraction image and enhancing recognition stability.

[0084] Example 4:

[0085] This embodiment examines the identification capability of Embodiment 1 in practical applications. The identification method includes:

[0086] S1. Prepare the airborne spore sample to be tested on a glass slide and place the glass slide in the sample layer in the middle of the light shield so that the area to be tested is directly below the micropore;

[0087] S2. Coarse adjustment is performed using a Z-axis displacement stage, and fine adjustment is performed using a nanometer-scale displacement stage. The relative height between the micropore and the sample layer, as well as the relative distance between the sample layer and the CMOS sensor, are adjusted so that the focused spot corresponding to the spectral focusing line can stably enter the micropore and form a clear diffraction pattern on the CMOS sensor.

[0088] S3. Control the integrated spectrometer module to perform wavelength scanning within the preset working band range of 380-1000nm, so that the dispersive beam generated by the concave reflection grating is focused twice by the focusing lens to form a focused monochromatic light spot at the micro-hole. The micro-hole spatially confines the light spot and outputs a narrow-band monochromatic illumination beam to irradiate the sample layer.

[0089] S4. Under monochromatic illumination conditions corresponding to each wavelength, the corresponding diffraction images are acquired by the CMOS sensor, and a diffraction image sequence is formed in wavelength order. Then, the obtained diffraction image sequence is transmitted to the host computer.

[0090] S5. The signal processing and identification module in the host computer preprocesses the diffraction image sequence. After preprocessing, the diffraction image sequence is combined in the hyperspectral diffraction fingerprint construction unit to form a multidimensional data cube of "wavelength-space diffraction image". Then, the feature vector for classification and recognition is extracted in the feature extraction unit.

[0091] S6. Use the feature vectors obtained after processing airborne spores of different categories in steps S1-S5 to train a classification and recognition model; input the feature vectors of the test samples into the trained classification and recognition model, and output the spore category and corresponding confidence level.

[0092] Based on the above identification methods, this embodiment uses maize leaf rust spores, wheat leaf rust spores (purchased from the Chinese Academy of Agricultural Sciences), and microsphere samples as classification objects. Figure 6-8 Only examples of the spectral responses of two types of disease spores are shown, while Figure 12 This corresponds to the classification verification results of three types of samples. The device of this invention is used to acquire diffraction image sequences at different wavelengths, and diffraction intensity curves varying with wavelength are extracted to construct a hyperspectral diffraction fingerprint.

[0093] The data acquisition process is as follows:

[0094] The control optical path integrated housing 13 performs wavelength scanning in the range of 380–1000 nm, and acquires corresponding diffraction images at each wavelength using a CMOS sensor. The diffraction images are transmitted to a host computer, where intensity is extracted from the target region of interest (ROI) to obtain a curve showing the diffraction intensity versus wavelength. The curve is then smoothed and normalized. The processing result is as follows: Figure 6-8As shown in the figure, the two types of spores exhibit distinguishable intensity differences and trends across multiple spectral bands, indicating that the hyperspectral diffraction fingerprints obtained by the device of this invention have the potential to distinguish between different airborne disease spores and can be used for subsequent feature extraction and classification.

[0095] The signal processing and discrimination methods are as follows:

[0096] S1. Perform wavelength scanning within a preset working band range, preferably 380–1000 nm; under monochromatic illumination conditions corresponding to each wavelength, acquire the corresponding spore diffraction images by a CMOS sensor and form a diffraction image sequence according to wavelength order; then transmit the diffraction image sequence to a host computer for subsequent signal processing and classification recognition.

[0097] S2. Image Preprocessing:

[0098] The host computer performs preprocessing on the diffraction image sequence sequentially, the preprocessing including:

[0099] The diffraction images acquired at each wavelength were denoised to reduce the impact of random noise on the diffraction fringes. Background subtraction was performed on the denoised diffraction images to reduce the interference of system background signals on subsequent analysis. Then, the gray values ​​of the diffraction images after background subtraction or the diffraction intensity of the target area were normalized to eliminate the influence of overall light intensity fluctuations at different wavelengths. Finally, the diffraction intensity curves formed in wavelength order were smoothed to improve curve stability and facilitate subsequent feature extraction.

[0100] After the above processing, a preprocessed diffraction image sequence for constructing a hyperspectral diffraction fingerprint is obtained;

[0101] S3. Diffraction fingerprint construction and feature extraction:

[0102] The preprocessed diffraction image sequence is combined according to wavelength order to form a multidimensional data cube of "wavelength-spatial diffraction image". In the multidimensional data cube, for the diffraction image at each wavelength, diffraction intensity information is extracted within a pre-defined target region (ROI) to obtain a characteristic curve of diffraction intensity changing with wavelength, so as to construct the hyperspectral diffraction fingerprint of the sample under test. Based on the differences in diffraction intensity response of different types of samples at each wavelength, the band in the range of 380 to 1000 nm is screened, and multiple wavelengths with stable signal response, low noise and obvious differences between different types of samples are selected as characteristic wavelengths.

[0103] In the diffraction image corresponding to the characteristic band, spatial features are further extracted. These spatial features include gray-scale distribution features of the target area, diffraction fringe intensity features, radial intensity spectrum features, and the location of the first minimum value. The spectral features corresponding to the characteristic wavelength are then combined with the spatial features to form a feature vector for classification and recognition.

[0104] S4. Model Training and Validation:

[0105] Maize leaf rust spores (purchased from the Chinese Academy of Agricultural Sciences), wheat leaf rust spores (purchased from the Chinese Academy of Agricultural Sciences), and microspheres (purchased from Tianjin Dage Technology Co., Ltd.) were selected as samples for model training. The samples were divided into training and test sets at a 5:1 ratio. The training set contained 180, 85, and 95 samples of maize leaf rust spores, wheat leaf rust spores, and microspheres, respectively, totaling 360 samples. The test set contained 36, 17, and 19 samples of maize leaf rust spores, wheat leaf rust spores, and microspheres, respectively, totaling 72 samples.

[0106] The training set samples are sequentially subjected to diffraction image acquisition, image preprocessing, diffraction fingerprint construction, band selection and spatial feature extraction according to steps S1) to S3) to obtain the feature vectors corresponding to each training set sample; then the feature vectors and class labels of the training set samples are input into the SVM classification model for training to obtain a classification model for the classification and recognition of three types of samples.

[0107] Then, the feature vectors of the test set samples are input into the trained classification model to verify the model's performance. A classification evaluation index is calculated based on the consistency between the predicted results and the true labels. The results are as follows: Figure 12 As shown. From Figure 12 As can be seen, the classification model constructed in this step has an overall accuracy of 0.9444 and an overall precision of 0.9497 for the three types of samples; specifically, the precision for maize leaf rust spores, wheat leaf rust spores, and microspheres are 0.9444, 1.0000, and 0.9048, respectively. The verification results demonstrate that the classification model established in this invention can effectively distinguish between maize leaf rust spores, wheat leaf rust spores, and microspheres.

[0108] In summary, this invention provides a hyperspectral diffraction-based device and method for identifying airborne disease spores. The invention employs a single folded optical path—"incident slit – collimating lens – concave reflection grating – focusing lens – micropore – sample – CMOS"—integrating monochromatic illumination and diffraction imaging onto the same optical axis. This results in a simple structure, easy assembly and adjustment, and convenient packaging within a miniaturized housing, making it suitable for deployment in monitoring stations and mobile platforms. The device described in this invention enables cost-effective, compact, and stable hyperspectral diffraction detection, providing technical support for early warning and rapid on-site diagnosis of airborne crop diseases.

[0109] The embodiments described above are preferred embodiments of the present invention, but the present invention is not limited to the above embodiments. Any obvious improvements, substitutions or modifications that can be made by those skilled in the art without departing from the essence of the present invention shall fall within the protection scope of the present invention.

Claims

1. A hyperspectral diffraction-based device for identifying airborne pathogenic spores, characterized in that, The identification device includes: a hyperspectral diffraction imaging device housing (11) and a host computer (23) disposed outside the hyperspectral diffraction imaging device housing (11). The hyperspectral diffraction imaging device housing (11) integrates a light source module, an integrated spectrometer module, a light shield, a CMOS sensor (6), a nanoscale displacement stage (2), and a Z-axis displacement stage (1). The light source module includes a light source (22) and an optical fiber (21) for the light source (22) to emit light output. The integrated spectrometer module includes an optical path integrated housing (13), on which an entrance slit (20) is arranged coaxially with the light-emitting end of the optical fiber (21), so that the polychromatic light output from the optical fiber (21) is perpendicularly irradiated by the entrance slit (20). Inside the optical path integrated housing (13), along the optical axis, there are a collimating lens (16), a concave reflection grating (19), and a focusing lens (15) for processing the polychromatic light input into the entrance slit (20); the bottom of the optical path integrated housing (13) is provided with a light-emitting hole. The top of the light shield is provided with a micro-hole (10), and the position of the micro-hole (10) corresponds to the position of the light-emitting hole; the middle part of the interior of the light shield is provided with a sample layer (7) for placing the sample to be tested; the lower end of the interior of the light shield is provided with a CMOS sensor (6). The light shield is fixed on the nanoscale displacement stage (2), which is fixed to the moving end of the Z-axis displacement stage (1); the Z-axis displacement stage (1) is fixed at the bottom of the hyperspectral diffraction-type airborne disease spore identification device. The host computer (23) is electrically connected to the CMOS sensor (6) and is used to receive the diffraction image sequence collected by the CMOS sensor (6), execute signal processing and identification algorithms, and output the spore category identification result.

2. The hyperspectral diffraction-based airborne disease spore identification device according to claim 1, characterized in that, The light source (22) of the light source module is a hyperspectral halogen light source with an output spectral range covering 380 to 1000 nm.

3. The hyperspectral diffraction-based airborne disease spore identification device according to claim 1, characterized in that, The concave reflective grating (19) has a line density of 600 lines / mm and a radius of curvature of -300 mm; The radius of curvature of the collimating lens (16) is -180 mm; The radius of curvature of the focusing lens (15) is -210 mm; The collimating lens (16) is positioned on the exit side of the entrance slit (20). The concave reflective grating (19) is positioned behind the collimating lens (16) and receives the collimated polychromatic beam (17) at an incident angle of 30°. The focusing lens (15) is positioned on the path of the dispersive beam (14) of the concave reflective grating (19) and is used to perform secondary focusing on the dispersive beam (14) to obtain a focused dispersive beam (12). The focused dispersive beam (12) forms a focused monochromatic spot at the micro-aperture (10) that matches the aperture of the micro-aperture (10).

4. The hyperspectral diffraction-based airborne disease spore identification device according to claim 1, characterized in that, The light shield is a closed hollow structure, and its upper end is provided with a micro-hole adjustment block (9) for fixing the micro-hole (10), and the micro-hole is set on the micro-hole adjustment block (9); The inner wall of the light shield is treated with a matte finish. A metal heat sink plate (5) is fixedly connected below the CMOS sensor (6). The distance between the micropore and the upper surface of the sample layer (7) is maintained at 50-70 mm, and the distance between the lower surface of the sample layer (7) and the CMOS sensor (6) is 2-6 mm, preferably 3 mm.

5. The hyperspectral diffraction-based airborne disease spore identification device according to claim 1, characterized in that, A heat sink (4) is also provided on the lower side of the CMOS sensor (6). The heat sink (4) includes a semiconductor cooling chip, a finned heat sink and a fan assembly; the cold end of the semiconductor cooling chip is in close contact with the metal heat sink base plate (5) and the hot end is in close contact with the finned heat sink; the fan assembly is located at the airflow channel of the finned heat sink.

6. The hyperspectral diffraction-based airborne disease spore identification device according to claim 5, characterized in that, An auxiliary heat sink is also provided on the outside of the metal heat sink base plate (5).

7. The hyperspectral diffraction-based airborne disease spore identification device according to claim 1, characterized in that, The host computer (23) is equipped with a signal processing and identification module; the signal processing and identification module includes: Hyperspectral diffraction fingerprint building unit: used to combine diffraction images acquired at different wavelengths to form a multidimensional data cube of "wavelength-space diffraction image"; Feature extraction unit: used for band selection and spatial feature extraction of the data cube; The classification and identification unit uses a machine learning classification model to identify different airborne disease spores and outputs the spore category and corresponding confidence level.

8. The hyperspectral diffraction-based airborne disease spore identification device according to claim 7, characterized in that, The machine learning classification model is trained using data of different categories of airborne disease spores extracted by the feature extraction unit, and the machine learning classification model includes an SVM classification model.

9. An identification method using the hyperspectral diffraction-based airborne spore identification device according to any one of claims 1 to 8, characterized in that, The identification method includes: S1. Prepare the airborne spore sample to be tested on a glass slide and place the glass slide in the sample layer (7) in the middle of the light shield so that the area to be tested is directly below the micropore; S2. Coarse adjustment is performed using the Z-axis displacement stage (1) and fine adjustment is performed using the nanoscale displacement stage (2). The relative height between the micropore and the sample layer (7) and the relative distance between the sample layer (7) and the CMOS sensor (6) are adjusted so that the focused spot corresponding to the spectral focusing line can stably enter the micropore and form a clear diffraction pattern on the CMOS sensor (6). S3. Control the integrated spectrometer module to perform wavelength scanning within the preset working band range of 380-1000nm, so that the dispersive beam (14) generated by the concave reflective grating (19) is focused twice by the focusing lens (15) to form a focused monochromatic light spot at the micro-hole. The micro-hole spatially confines the light spot and outputs a narrow-band monochromatic illumination beam to irradiate the sample layer (7). S4. Under the monochromatic illumination conditions corresponding to each wavelength, the corresponding diffraction images are acquired by the CMOS sensor (6), and a diffraction image sequence is formed in wavelength order. Then, the obtained diffraction image sequence is transmitted to the host computer (23). S5. The signal processing and identification module in the host computer (23) preprocesses the diffraction image sequence. After the preprocessing, the diffraction image sequence is combined in the hyperspectral diffraction fingerprint construction unit to form a multidimensional data cube of "wavelength-space diffraction image". Then, the feature vector for classification and identification is extracted in the feature extraction unit. S6. Use the feature vectors obtained after processing different types of airborne disease spores in steps S1-S5 to train a classification and recognition model; The feature vector of the sample to be tested is input into the trained classification and recognition model, which outputs the spore category and the corresponding confidence level.

10. The identification method according to claim 9, characterized in that, In step S5, the preprocessing includes noise reduction, background subtraction, normalization, and smoothing in sequence.

Images