Detection device and method based on mid-wave near-infrared non-contact diffuse reflectance spectrum

Abstract

The application relates to the technical field of near-infrared spectrum detection, and discloses a detection device and method based on a middle-wave near-infrared non-contact diffuse reflection spectrum, which adopts a transmitting conical light path and a receiving conical light path, so that an illumination area and a spectrum collection area are completely coincident, the distance between a light source module, a sample disc and a protective window sheet is adjustable, the protective window sheet is arranged in an inclined mode, has a good illumination effect, the sample disc rotates eccentrically, can irradiate different areas of a sample, effectively covers the spatial distribution of the whole batch of samples, avoids evaluation deviation caused by local abnormal particles or vacancy areas, significantly improves the overall component estimation representation, reduces spectrum variation caused by a placement mode, and improves the repeatability and stability of multiple measurements of the same batch of samples. In addition, the preset algorithm adopted by the application can enhance the fault tolerance to abnormal samples and noise interference, significantly improve the prediction accuracy and robustness, and simultaneously improve the efficiency and precision.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of near-infrared spectroscopy detection technology, in particular to a detection device and method based on mid-infrared non-contact diffuse reflectance spectroscopy. BACKGROUND

[0002] At present, the detection of the component content of rapeseed, rice bran and the like mainly adopts near-infrared diffuse reflectance measurement. Usually, the sample is placed in a sample cup or cuvette with optical glass at the bottom, and the bottom of the sample is optically detected by a light source and a receiving unit to obtain the near-infrared diffuse reflectance spectrum of the sample. Through analysis of the absorption spectrum, the component content of the sample can be obtained.

[0003] The existing detection device directly places the sample in the area irradiated by the light source, and the direction of the light source irradiation is fixed. Therefore, only the local area of the sample can be measured, and the detection results of other areas of the sample cannot be obtained. The detection results are easily disturbed by local outliers, resulting in large deviation of the overall component or quality evaluation and poor repeatability. For example, rapeseed, rice bran and the like are randomly placed. The fixed irradiation direction can only capture a few particles in the irradiation area, and it is difficult to represent the overall characteristics of the whole batch of samples. Moreover, the orientations of the particles are different, causing large fluctuations in the reflectance spectrum under different irradiation forms. In addition, the height difference between the particles causes some samples to deviate from the best focal plane of the optical system, resulting in the expansion of the light spot, the decrease of the energy density, and the instability of the signal strength.

[0004] Moreover, in order to isolate the internal environment of the detection device from the outside, there are multiple layers of glass between the light source and the sample. On the one hand, this will cause the light reaching the sample to weaken. On the other hand, the multiple layers of glass may cause optical interference, affecting the accuracy of the measurement.

[0005] Therefore, in the current near-infrared spectroscopy detection, due to the fixed direction of the light source, the random distribution of the particles and the existence of multiple layers of glass, the measurement results are easily affected by local outliers, light intensity attenuation and optical interference. The overall characteristics of the sample cannot be accurately reflected, and the repeatability is poor. SUMMARY

[0006] The purpose of the present application is to provide a detection device and method based on mid-infrared non-contact diffuse reflectance spectroscopy, which can improve the accuracy and reliability of overall detection.

[0007] In order to achieve the above purpose, the present application provides the following technical solutions:

[0008] In a first aspect, the present application provides a detection device based on mid-infrared non-contact diffuse reflectance spectroscopy, comprising:

[0009] A light source module is used for collimating and converging the light source to form a light source conical space with adjustable focal length to irradiate the sample.

[0010] a guide mirror for guiding the detection light emitted by the light source module to the sample and coupling the light beam reflected by the sample to the spectrometer through the receiving module;

[0011] a protective window sheet arranged between the guide mirror and the sample, for isolating the internal environment of the detection device from the outside, the detection light emitted by the light source module irradiates the sample after being reflected by the guide mirror and transmitted by the protective window sheet in turn, wherein the protective window sheet is arranged obliquely to reduce the specular reflection of the light source;

[0012] a calibration module arranged between the guide mirror and the protective window sheet, the calibration module is provided with switchable first and second calibration plates, the first calibration plate has characteristic absorption in the near-infrared mid-wave band for wavelength calibration, and the second calibration plate is used for background calibration;

[0013] a receiving module provided with an optical fiber moving towards or away from the sample, the receiving cone space formed by the light receiving angle of the optical fiber is concentric with the light source cone space formed by the light source module, and the spot area of the light source is larger than the receiving area of the optical fiber by moving the optical fiber;

[0014] a spectrometer connected with the receiving module, the photosensitive element contained in the spectrometer performs photoelectric conversion on the light signal received from the receiving module to generate sample spectrum data;

[0015] a processing and analysis module connected with the spectrometer, for pre-processing the sample spectrum data, and calculating the pre-processed sample spectrum data based on a preset algorithm and outputting the corresponding component content result;

[0016] a sample module including a sample disc and a rotating sample table, the rotating sample table is arranged at the intersection of the light source cone space and the receiving cone space, the sample disc is used to hold the sample, and the rotating sample table drives the sample disc to rotate, and the rotation center of the sample disc deviates from the light source cone space;

[0017] an adjustment module for adjusting the distance between the sample disc and the protective window sheet.

[0018] In a second aspect, the present application provides a non-contact diffuse reflectance spectroscopy detection method based on mid-wave near-infrared, based on the above detection device, comprising:

[0019] S1. Determine the working distance according to the sample type, and adjust the distance between the sample tray and the protective window according to the working distance; adjust the light source module and the receiving module so that the light spot area of ​​the light source is larger than the receiving area of ​​the optical fiber of the receiving module.

[0020] S2. The detection light emitted from the light source module shines on the sample in the sample tray through the protective window. The diffuse reflected light of the sample is collected by the optical fiber of the receiving module and sent to the spectrometer for sample spectral data acquisition.

[0021] S3. Rotate the sample stage to drive the sample disk to rotate. Perform a spectral acquisition once for each preset angle of rotation until all samples to be tested have completed spectral acquisition.

[0022] S4. The calibration module moves the first calibration plate between the light source module and the protective window, and the spectrometer collects the current spectral data of the first calibration plate for wavelength calibration of the spectrometer.

[0023] S5. The calibration module moves the second calibration plate between the light source module and the protective window, and the spectrometer collects the current spectral data of the second calibration plate for background calibration of the spectrometer.

[0024] S6. The processing and analysis module preprocesses the sample spectral data, the first calibration plate spectral data, and the second calibration plate spectral data received from the spectrometer, and then calculates the preprocessed sample spectral data based on a preset algorithm and outputs the corresponding component content results.

[0025] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0026] The receiving cone space formed by the light receiving angle of the optical fiber of the application is concentric with the light source cone space formed by the light source module, so that the illumination area and the spectral acquisition area completely coincide, and the spot area of the light source is larger than the receiving area of the optical fiber, which ensures that the optical fiber always receives stable, reliable and representative reflected light signals, avoiding signal loss or severe fluctuations. The focal length of the light source module can be adjusted, so that the spot size and energy density of the outgoing detection light on the sample surface can be dynamically adjusted. At the same time, the distance between the sample disc and the protective window piece is adjustable, which is used to accurately adjust the sample surface to the best focusing position of the detection light. By adjusting the focal length of the light source module and the distance between the guide mirror and the protective window piece, the optimal illumination condition and focusing state are configured, which significantly improves the signal-to-noise ratio, measurement repeatability and application range of the spectral signal. The protective window piece of the application is inclined, which can reflect the interference light reflected by the protective window piece out of the collection range, preventing errors caused by lens mirror reflection of light, and providing detection accuracy. In addition, the sample disc is driven to rotate by the rotating sample table, and the center of rotation of the sample disc deviates from the light source cone space, so that the sample disc rotates eccentrically, which can irradiate different areas of the sample, effectively cover the spatial distribution of the whole batch of samples, avoid evaluation deviation caused by local abnormal particles or empty areas, significantly improve the overall composition estimation representativeness, and reduce the spectral variation caused by the placement form, improve the repeatability and stability of multiple measurements of the same batch of samples. In addition, the preset algorithm used in the application can enhance the fault tolerance of abnormal samples and noise interference, significantly improve the prediction accuracy and robustness, and improve the efficiency and accuracy. BRIEF DESCRIPTION OF DRAWINGS

[0027] Figure 1 is a structural schematic diagram of a detection device based on mid-infrared non-contact diffuse reflectance spectroscopy according to an embodiment of the application;

[0028] Figure 2 is an optical path diagram of a light source module according to an embodiment of the application;

[0029] Figure 3 is an optical path diagram of a light source module according to another embodiment of the application;

[0030] Figure 4 is a spectral diagram of rapeseed using the detection device and method of the application;

[0031] Figure 5 is a spectral diagram of rice bran using the detection device and method of the application;

[0032] Figure 6 is a spectral diagram of raw soybean powder using the detection device and method of the application;

[0033] Figure 7 is a spectral diagram of wheat using the detection device and method of the application;

[0034] Figure 8 Figure is a flow chart of the detection method based on mid-infrared non-contact diffuse reflectance spectroscopy of the embodiment of the present application.

[0035] In the figure, 1 is a light source module; 101 is a light source; 102 is a first lens; 103 is a second lens; 104 is an extinction lens barrel; 2 is a guide mirror; 3 is a protective window piece; 4 is a calibration module; 5 is a receiving module; 501 is an optical fiber; 6 is a spectrometer; 7 is a sample disc; 8 is a rotating sample stage. DETAILED DESCRIPTION

[0036] The specific embodiments of the present application are described in detail below with reference to the accompanying drawings and examples. The following examples are used to illustrate the present application, but are not used to limit the scope of the present application.

[0037] In the description of the present application, it should be noted that the terms “center”, “longitudinal”, “transverse”, “upper”, “lower”, “front”, “rear”, “left”, “right”, “vertical”, “horizontal”, “top”, “bottom”, “inner”, “outer” and the like indicate the orientation or positional relationship based on the orientation or positional relationship shown in the drawings, and are only for the convenience of describing the present application and simplifying the description, and therefore cannot be understood as indicating or implying that the devices or elements referred to must have a particular orientation, be constructed and operated in a particular orientation, and therefore cannot be understood as limiting the present application. In addition, the terms “first”, “second”, “third” are only for descriptive purposes and cannot be understood as indicating or implying relative importance.

[0038] In the description of the present application, it should be noted that unless otherwise explicitly specified and limited, the terms “mounting”, “connection”, “connection” should be understood broadly, for example, it can be fixedly connected, or it can be detachably connected, or integrally connected; it can be mechanically connected, or it can be electrically connected; it can be directly connected, or it can be indirectly connected through an intermediate medium; it can be the communication inside two elements. For those skilled in the art, the specific meaning of the above terms in the present application can be understood according to the specific circumstances.

[0039] In addition, in the description of the present application, unless otherwise specified, the meaning of “multiple” is two or more.

[0040] Example 1

[0041] As shown in Figure 1 and Figure 2 A detection device based on mid-infrared non-contact diffuse reflectance spectroscopy of the preferred embodiment of the present application comprises:

[0042] The light source module 1 is used for collimating and converging the light source 101 to form a light source 101 cone-shaped space irradiating the sample with adjustable focal length;

[0043] a guide mirror 2 configured to guide the detection light emitted by the light source module 1 to the sample and couple the light reflected by the sample to the spectrometer 6 through the receiving module 5;

[0044] a protective window sheet 3 disposed between the guide mirror 2 and the sample, the protective window sheet 3 being configured to isolate the internal environment of the detection device from the outside, and the detection light emitted by the light source module 1 being sequentially reflected by the guide mirror 2 and transmitted by the protective window sheet 3 to irradiate the sample, wherein the protective window sheet 3 is disposed obliquely to reduce the specular reflection of the light source 101;

[0045] a calibration module 4 disposed between the guide mirror 2 and the protective window sheet 3, the calibration module 4 being provided with switchable first and second calibration plates, the first calibration plate being configured for background calibration, and the second calibration plate being configured for wavelength calibration;

[0046] a receiving module 5 provided with an optical fiber 501 movable along a direction close to or away from the sample, a receiving cone space formed by a light receiving angle of the optical fiber 501 being concentric with a light source 101 cone space formed by the light source module 1, and the area of the light spot of the light source 101 being kept larger than the receiving area of the optical fiber 501 through the movement of the optical fiber 501;

[0047] a spectrometer 6 connected to the receiving module 5, the light signal received from the receiving module 5 being photoelectrically converted by a photosensitive element contained in the spectrometer 6 to generate sample spectrum data;

[0048] a processing and analysis module connected to the spectrometer 6, configured to pre-process the sample spectrum data, and calculate and output the corresponding component content results based on a preset algorithm on the pre-processed sample spectrum data;

[0049] a sample module including a sample disc 7 and a rotating sample table 8, the rotating sample table 8 being disposed at the intersection of the light source 101 cone space and the receiving cone space, the sample disc 7 being configured to hold the sample, and the rotating sample table 8 driving the sample disc 7 to rotate, the rotation center of the sample disc 7 being offset from the light source 101 cone space;

[0050] an adjustment module configured to adjust the distance between the sample disc 7 and the protective window sheet 3.

[0051] The receiving cone space formed by the light receiving angle of the optical fiber 501 is concentric with the light source 101 cone space formed by the light source module 1, so that the illumination area and the spectral acquisition area are completely coincident, and the spot area of the light source 101 is larger than the receiving area of the optical fiber 501, which ensures that the optical fiber 501 always receives stable, reliable and representative reflected light signals, avoiding signal loss or severe fluctuations. The focal length of the light source module 1 can be adjusted, so that the spot size and energy density of the outgoing detection light on the sample surface can be dynamically adjusted; at the same time, the distance between the sample disc 7 and the protective window piece 3 is adjustable, which is used to accurately adjust the sample surface to the best focusing position of the detection light. By adjusting the focal length of the light source module 1 and the distance between the guide mirror 2 and the protective window piece 3, the optimal illumination condition and focusing state are configured, which significantly improves the signal-to-noise ratio, measurement repeatability and application range of the spectral signal. The protective window piece 3 of the present application is inclinedly arranged, which can reflect the interference light reflected by the protective window piece 3 out of the collection range, preventing the error of the collected light spot caused by the mirror reflection light, and improving the detection accuracy. In addition, the sample disc 7 is driven to rotate by the rotating sample table 8, and the rotation center of the sample disc 7 deviates from the light source 101 cone space, so that the sample disc 7 rotates eccentrically, which can irradiate different areas of the sample, effectively cover the spatial distribution of the whole batch of samples, avoid the evaluation deviation caused by local abnormal particles or vacancy areas, significantly improve the overall composition estimation representativeness, and reduce the spectral variation caused by the placement form, improve the repeatability and stability of multiple measurements of the same batch of samples.

[0052] In specific embodiments, first, the working distance is determined according to the type of the sample to be measured, the working distance being the distance between the sample and the protective window sheet 3, the up and down position of the rotating sample is adjusted by the adjusting module to make the distance between the sample disc 7 and the protective window sheet 3 reach the working distance. Then the light source module 1 and the receiving module 5 are adjusted, so that the light spot area of the light source 101 irradiating on the sample disc 7 which has reached the working distance is larger than the receiving area of the optical fiber 501 of the receiving module 5; the detection light emitted by the light source module 1 is reflected by the guide mirror 2, transmitted by the protective window sheet 3 and then irradiates on the sample to be measured, the sample diffuse reflection light is collected by the optical fiber 501 of the receiving module 5 and sent to the spectrometer 6 for sample spectrum data acquisition; the sample disc 7 is rotated by the rotating sample stage 8, and the spectrum is collected once every specified angle until the spectrum collection of all samples to be measured on the sample disc 7 is completed; the first calibration plate is moved to between the guide mirror 2 and the protective window sheet 3 by the calibration module 4, and the current first calibration plate spectrum data is collected by the spectrometer 6 for wavelength calibration of the spectrometer 6, the wavelength calibration is performed periodically, which can be performed once every specified angle or once every three specified angles; the second calibration plate is moved to between the guide mirror 2 and the protective window sheet 3 by the calibration module 4, and the current second calibration plate spectrum data is collected by the spectrometer 6 for background calibration of the spectrometer 6, the background calibration is performed once after every sample spectrum data collection; after the spectrum collection of all samples to be measured on the sample disc 7 is completed, the sample spectrum data, the first calibration plate spectrum data and the second calibration plate spectrum data received from the spectrometer 6 are preprocessed, and the preprocessed sample spectrum data is calculated based on the preset algorithm and the corresponding component content result is output.

[0053] In some embodiments, the inclination angle of the protective window sheet 3 is between 10°-20°, that is, the inclination angle of the protective window sheet 3 can be 10°, 15° or 20°. The inclination of the protective window sheet 3 means that the optical axis (system optical axis) of the light source 101 cone space formed by the light source module 1 has a certain angle with the surface of the protective window sheet 3, and the angle of the angle is the inclination angle. The equipment shell is provided with a window, and the protective window sheet 3 is installed at the window. The protective window sheet 3 is a medium for isolating the internal environment of the equipment from the outside, and is not placed horizontally, but at a certain inclination angle, which can prevent errors in collecting light spots caused by mirror reflection when placed horizontally, and the inclined placement can reflect the interference light reflected by the protective window sheet 3 to the outside of the collection range. The inclination angle of the protective window sheet 3 is determined according to the size of the guide mirror 2 and the light spot range.

[0054] In other embodiments, the light source 101 of the light source module 1 uses a halogen lamp, which has high radiation intensity and good stability, can effectively excite the characteristic absorption of sample components, and significantly improve the spectral signal-to-noise ratio and component prediction accuracy.

[0055] In addition, the calibration module 4 has three switching modes, respectively, a blank area, a first calibration plate and a second calibration plate, and is switched to the blank area when the sample is photographed, at which time there is no object between the guide mirror 2 and the protective window sheet 3, and the light does not irradiate any calibration plate.

[0056] Optionally, the photosensitive element of the spectrometer 6 of the embodiment adopts an array detector.

[0057] Embodiment Two

[0058] The difference between the embodiment and Embodiment Two is that, on the basis of Embodiment Two, the embodiment further describes a detection device based on mid-wave near-infrared non-contact diffuse reflectance spectroscopy.

[0059] The guide mirror 2 of the embodiment is obliquely arranged, the guide mirror 2 is provided with a receiving hole penetrating through, and the optical fiber 501 is arranged in the receiving hole. The guide mirror 2 is obliquely arranged, which can prevent the spot from being distorted. The optical axis of the light source 101 cone space formed by the light source module 1 of the embodiment is at an angle of 45° with the surface of the guide mirror 2. Moreover, the optical fiber 501 is arranged in the receiving hole at the center of the guide mirror 2, which can make the illumination light beam and the collection light beam share the same optical axis, realize true coaxial vertical irradiation and collection, and ensure that the illumination area and the receiving field of view are completely coincident at any working distance.

[0060] In the embodiment, the light source module 1 includes a light source 101, a first lens 102 and a second lens 103. The detection light emitted by the light source 101 passes through the first lens 102, the second lens 103, the guide mirror 2 and the protective window sheet 3 in sequence and irradiates on the sample of the sample disc 7. The first lens 102 is used to convert the divergent light emitted by the light source 101 into parallel light, and the second lens 103 is used to convert the parallel light into convergent light. The first lens 102 and the second lens 103 are wrapped by an extinction lens barrel 104 with an extinction structure and material, which is used to reduce the non-parallel light reaching the second lens 103. The light emitted by the light source 101 is collimated and converged by the lens, so that the spot irradiated on the sample is uniform, which can significantly improve the repeatability, accuracy and spectral representativeness of the detection result, avoid signal distortion and model prediction deviation caused by local overexposure or underexposure, and at the same time, if the lens is not used, the light rays irradiating to the guide mirror 2 are better shaped into high-quality parallel light or convergent light, which requires the guide mirror 2 to be arranged larger, resulting in that the overall volume of the equipment is larger and the utilization of the equipment space is not conducive. Moreover, the spot irradiated on the sample disc 7 is not uniform and there is more stray light. Therefore, the light source module 1 of the embodiment is beneficial to provide a high-quality light source, equipment miniaturization and improve the space utilization.

[0061] In some embodiments, the first lens 102 can adopt one or more, and the second lens 103 can adopt one or more.

[0062] In some embodiments, the first lens 102 and the second lens 103 are installed in a light-eliminating lens barrel 104, which can eliminate the non-parallel light reaching the second lens 103, making the light spot more uniform, while isolating the external light from the internal light, and facilitating the lens installation and removal.

[0063] As shown in the figure, the embodiment adopts a first lens 102 and a second lens 103, and the guide mirror 2 adopts a plane mirror, and the first lens 102 and the second lens 103 are wrapped by a light-eliminating lens barrel 104. Figure 2

[0064] Optionally, for the light-eliminating lens barrel 104 with light-eliminating structure and material, in terms of structure, the light-eliminating structure refers to a geometric structure arranged in the inner wall of the lens barrel or the internal light path, which is used to suppress the reflection and propagation of stray light, by changing the light incidence angle, increasing the light path shielding, or prolonging the non-imaging light path, to reduce the possibility of reaching the imaging surface. For example, the light-eliminating structure can include but is not limited to inner wall grooves, light shielding rings, multi-stage steps, or rough surfaces formed by sandblasting; in terms of material, the light-eliminating material refers to a material with low reflectivity or high light absorption performance, which is used to constitute the lens barrel body or the surface treatment layer thereof, and which suppresses stray light by reducing the reflection of light in the inner wall of the lens barrel; such materials can include but are not limited to: black non-reflective coating (such as black chromium, black nickel, carbon nanotube coating, or Acktar metal black film) coated on the surface of metal substrate (such as aluminum alloy, stainless steel), directly injection molded with engineering plastics (such as carbon black added polyether ether ketone PEEK or polysulfone PSU) with high light absorption performance, or forming a low reflectivity surface layer on metal and plastic substrates through anodizing, chemical deposition, physical vapor deposition (PVD) process, and can also use commercially available high-absorption light-eliminating paint (such as Nextel Velvet Coating, Aeroglaze Z306, etc.) for spraying treatment.

[0065] ​Furthermore, in this embodiment, the optical axes of the first lens 102 and the second lens 103 are located on the same straight line, and the first lens 102 and the second lens 103 are closer to or further away from each other along the direction of the optical axis. The first lens 102 and the second lens 103 are arranged in parallel and spaced apart, with a certain distance between them, which improves the collimation of the light and better eliminates some of the diffused light from the outside of the parallel light converged by the lenses, making the illumination within the conical space of the light source 101 formed by the light source module 1 more uniform, and improving the stability and accuracy of the detection results. By moving the first lens 102 or the second lens 103, the distance between two adjacent lenses can be changed, and the cone angle of the conical space of the light source 101 formed by the light source module 1 can be adjusted. This allows for dynamic adjustment of the divergence angle, focusing position of the illumination beam, and the size and energy density of the light spot on the sample surface, thereby adapting to samples of different sizes, reflection characteristics, or measurement requirements, and optimizing the signal-to-noise ratio and measurement representativeness.

[0066] Optionally, the first lens 102 is an aspherical lens, and the second lens 103 is a plano-convex lens. Aspherical lenses can effectively correct spherical aberration, achieving high-precision collimation or strong convergence; plano-convex lenses are suitable for converging parallel light. Therefore, the combination of an aspherical lens and a plano-convex lens provides effective collimation and convergence performance. Optionally, the first lens 102 is a biconvex lens.

[0067] In other embodiments, such as Figure 3 As shown, the light source module 1 includes a light source 101 and a first lens 102. The guide mirror 2 is an off-axis parabolic reflector. The detection light emitted from the light source 101 passes sequentially through the first lens 102, the guide mirror 2, and the protective window 3 to illuminate the sample on the sample tray 7. The first lens 102 converts the divergent light emitted from the light source 101 into parallel light. The guide mirror 2 converts the parallel light into converging light and guides it onto the sample on the sample tray 7. The first lens 102 is encased in an extinction mirror tube 104 with an extinction structure and material to reduce non-parallel light reaching the guide mirror 2. The first lens 102 and the extinction mirror tube 104 can form high-quality parallel light that illuminates the off-axis parabolic guide mirror 2. The guide mirror 2, using an off-axis parabolic reflector, converges the incident parallel light after a 90° bend, forming a uniform conical light that illuminates the sample. In this embodiment, the first lens 102 is a biconvex lens.

[0068] In addition, the receiving module 5 also includes a lifting device for moving the optical fiber 501 up and down, thereby adjusting the vertical position of the optical fiber 501.

[0069] The implementation process of the detection device of the embodiment is as follows: first, according to the type of the sample to be detected, the working distance is determined, the working distance is the distance between the sample and the protective window sheet 3, the up and down position of the rotating sample is adjusted by the adjusting module, so that the distance between the sample disc 7 and the protective window sheet 3 reaches the working distance. Then adjust the distance between the first lens 102 and the second lens 103, the up and down position of the optical fiber 501, so that the light spot area of the light source 101 irradiated on the sample disc 7 which has reached the working distance is greater than the receiving area of the optical fiber 501 of the receiving module 5, and the light spot area of the light source 101 irradiated on the sample disc 7 which has reached the working distance is greater than the receiving area of the optical fiber 501 of the receiving module 5 to achieve the best proportion. The detection light emitted by the light source 101 is focused into a specific range of light by the first lens 102 and the second lens 103, and after being reflected by the guide mirror 2, the reflected light is irradiated on the sample to be detected after passing through the protective window sheet 3. This light path is a transmission cone-shaped light path, that is, a cone-shaped space of the light source 101 formed by the light source module 1. The light irradiated on the sample is collected by the optical fiber 501 and then transmitted to the photosensitive element inside the spectrometer 6 for induction. This light path is a receiving cone-shaped light path, that is, a receiving cone-shaped space formed by the light collecting angle of the optical fiber 501. The spot size between the transmission cone-shaped light path and the receiving cone-shaped light path tends to be the same, but the spot size of the transmission cone-shaped light path is slightly larger than that of the receiving cone-shaped light path. The sample diffuse reflection light is collected by the optical fiber 501 and sent to the spectrometer 6 for sample spectrum data acquisition. The sample disc 7 is rotated by the rotating sample table 8, and the spectrum is collected once every specified angle of rotation, until the spectrum collection of all samples to be detected on the sample disc 7 is completed; the first calibration plate is moved to between the guide mirror 2 and the protective window sheet 3 by the calibration module 4, and the current first calibration plate spectrum data is collected by the spectrometer 6 for wavelength calibration of the spectrometer 6. The wavelength calibration is performed periodically, which can be performed once every specified angle of rotation, or once every three specified angles of rotation. The second calibration plate is moved to between the guide mirror 2 and the protective window sheet 3 by the calibration module 4, and the current second calibration plate spectrum data is collected by the spectrometer 6 for background calibration of the spectrometer 6. The background calibration is performed once after every sample spectrum data collection. After the spectrum collection of all samples to be detected on the sample disc 7 is completed, the processing and analysis module pre-processes the sample spectrum data according to the sample spectrum data, the first calibration plate spectrum data and the second calibration plate spectrum data received from the spectrometer 6, and then calculates and outputs the corresponding component content result based on the pre-set algorithm.

[0070] In some embodiments, the detection device has two working modes: normal mode, breeding or medicine mode. In step S1, the working mode is selected according to the number of samples, the rarity, and the particle size. The working mode has two modes: breeding or medicine mode and normal mode. If it is breeding or medicine mode, the working distance is 55-60 mm, and the light spot diameter of the light source 101 is 34-36 mm. The working distance can be 55 mm, 57 mm or 60 mm, and the light spot diameter of the light source 101 can be 34 mm, 35 mm or 36 mm. If it is normal mode, the working distance is 90-100 mm, and the light spot diameter is 46-51.4 mm. The working distance can be 90 mm, 95 mm or 100 mm, and the light spot diameter is 46 mm, 50 mm or 51.4 mm.

[0071] In other embodiments, the processing and analysis module pre-processes the sample spectrum data, the first calibration plate spectrum data, and the second calibration plate spectrum data received from the spectrometer 6, and then calculates the pre-processed sample spectrum data based on a preset algorithm and outputs the corresponding component content result. The pre-processing includes calibration, correction, and waveband selection.

[0072] The pre-processing includes calibrating and correcting the sample spectrum data,

[0073] The calibration includes: generating wavelength calibration data based on the first calibration plate spectrum data and its preset wavelength characteristics; performing wavelength calibration on the sample spectrum data and the second calibration plate spectrum data based on the wavelength calibration data, to obtain wavelength-calibrated sample spectrum data and wavelength-calibrated second calibration plate spectrum data; performing background subtraction on the wavelength-calibrated sample spectrum data based on the wavelength-calibrated second calibration plate spectrum data, to obtain the relative spectral response values of the sample at each wavelength; and the calibrated sample spectrum data is determined by the relative spectral response values and their corresponding wavelength positions after wavelength calibration.

[0074] The correction includes: performing first processing and second processing on the calibrated sample spectrum data. The first processing suppresses spectral noise, and the second processing eliminates physical interference caused by particle scattering, optical path difference, or baseline drift. Optionally, the first processing uses a five-point smoothing algorithm, and the second processing uses a standard normal variable transformation SNV.

[0075] The waveband selection is used to retain spectral data within a certain wavelength range, wherein:

[0076] For the quantitative detection of rapeseed components, the spectral data in the wavelength range of 1050-1630 nm are reserved; for the quantitative detection of rice bran components, the spectral data in the wavelength range of 1100-1650 nm are reserved; for the quantitative detection of raw soybean powder components, the spectral data in the wavelength range of 1100-1630 nm are reserved; and for the quantitative detection of wheat components, the spectral data in the wavelength range of 1080-1650 nm are reserved.

[0077] In some other embodiments, the preset algorithm adopts a prediction model, which is obtained by the following method: obtaining spectral data of a plurality of calibration samples, and pre-processing the spectral data to obtain pre-processed calibration sample spectra; obtaining reference values of target components of each calibration sample measured by a standard chemical analysis method; based on the pre-processed calibration sample spectra and the corresponding reference values, a multivariate correction modeling method is used to establish a prediction model of each target component; the pre-processed sample spectral data is calculated based on the preset algorithm and the corresponding component content results are output, including: inputting the pre-processed sample spectral data into the prediction model to obtain the component content results. The multivariate correction modeling method can adopt one or more of multivariate linear regression (MLR), principal component regression (PCR) and partial least squares regression (PLS) methods.

[0078] In this embodiment, the prediction model includes a linear regression sub-model, a nonlinear regression sub-model and a fusion model. The pre-processed sample spectrum is input into the linear regression sub-model to obtain a first component content prediction value, the pre-processed sample spectrum is input into the nonlinear regression sub-model to obtain a second component content prediction value, based on a pre-established model confidence evaluation mechanism, the reliability of the first component content prediction value and the second component content prediction value is screened to determine whether to input both, input only one, or trigger abnormal processing; the screened prediction values are input into the fusion model, and the final component content prediction result is output through the fusion model. The linear regression sub-model adopts a partial least squares regression (PLS) model, the nonlinear regression sub-model adopts a random forest regression (RF) model, and the fusion model adopts a multi-layer perception (MLP) model.

[0079] In an embodiment, rapeseed is selected as the sample to be detected, 310 representative rapeseed samples from different regions are collected, and the detection device of the embodiment is used in the ordinary working mode, the working distance is 90-100 mm, and the light spot diameter is 46-51.4 mm. For the quantitative detection of rapeseed components, the spectral data of the 310 rapeseed samples and the first and second calibration plate spectral data during the collection process are collected, and the spectral graph of the rapeseed is as shown in Figure 4The spectra of the samples are shown in FIG. 1. Standard chemical analysis methods were used to collect each sample three times, and the average value was taken as the reference value of the sample.

[0080] The rapeseed sample spectrum data were calibrated, corrected and band-selected pretreated according to the rapeseed sample spectrum data, the first calibration plate spectrum data and the second calibration plate spectrum data in the rapeseed sampling process, to obtain the pretreated rapeseed sample spectrum data. The prediction model established by using the multivariate correction modeling method was trained and verified using the rapeseed sample spectrum data and the reference value. After repeated training and comparative analysis, the model performance was optimal using the spectrum data with a wavelength range of 1050-1630 nm combined with the above algorithm. The performance of the prediction model of the present embodiment is shown in Table 1.

[0081] Table 1 Detection performance of rapeseed component content

[0082]

[0083] In another embodiment, several rapeseed samples under breeding conditions need to be detected for component content. Since the sample amount is small and valuable, a breeding or drug mode is used to adjust the working distance to 55-60 mm, and the light spot diameter to 34-36 mm. The spectrum of the breeding rapeseed sample is collected, and the rapeseed prediction model established in the previous embodiment is used to obtain the component content value of the breeding rapeseed sample.

[0084] In another embodiment, rice bran is selected as the sample to be detected. 200 rice bran samples from different sources are collected. The detection device of the present embodiment is used in the ordinary working mode, the working distance is 90-100 mm, and the light spot diameter is 46-51.4 mm. For the quantitative detection of the components of the rice bran, the spectrum data of the 200 rice bran samples and the first calibration plate spectrum data and the second calibration plate spectrum data in the sampling process are collected, and the spectrum of the rice bran is shown in FIG. 2. Figure 5 The standard chemical analysis method is used to collect each sample three times, and the average value is taken as the reference value of the sample.

[0085] The rice bran sample spectrum data were calibrated, corrected and band-selected pretreated according to the rice bran sample spectrum data, the first calibration plate spectrum data and the second calibration plate spectrum data in the rice bran sampling process, to obtain the pretreated rice bran sample spectrum data. The prediction model established by using the multivariate correction modeling method was trained and verified using the rice bran sample spectrum data and the reference value. After repeated training and comparative analysis, the model performance was optimal using the spectrum data with a wavelength range of 1100-1650 nm combined with the above algorithm. The performance of the prediction model of the present embodiment is shown in Table 2.

[0086] Table 2 Detection performance of rice bran component content

[0087]

[0088] In another embodiment, raw soybean powder was selected as the sample to be tested. Two hundred raw soybean powder samples from different sources were collected. Using the detection device of this embodiment in normal operating mode, with a working distance of 90-100 mm and a spot diameter of 46-51.4 mm, for quantitative detection of the components of the raw soybean powder, spectral data of these 200 raw soybean powder samples, as well as spectral data from the first calibration plate and the second calibration plate during the acquisition process, were collected. The spectrum of the raw soybean powder is shown below. Figure 6 As shown in the figure. Each sample was collected three times using standard chemical analysis methods, and the average value was used as the reference value for that sample.

[0089] Based on the spectral data of the raw soybean flour samples, as well as the spectral data from the first and second calibration plates during the raw soybean flour collection process, the spectral data of the raw soybean flour samples were calibrated, corrected, and preprocessed with band selection to obtain the preprocessed spectral data of the raw soybean flour samples. The prediction model established using the multivariate calibration modeling method was trained and validated using the raw soybean flour sample spectral data and reference values. After repeated and extensive training and comparative analysis, the model established using spectral data in the wavelength range of 1100~1630nm combined with the above algorithm showed the best performance. The performance of the prediction model in this embodiment is shown in Table 3.

[0090] Table 3. Detection performance of raw soybean flour component content

[0091]

[0092] In another embodiment, wheat was selected as the sample to be tested. 150 wheat samples from different sources were collected. Using the detection device of this embodiment in normal operating mode, with a working distance of 90-100 mm and a spot diameter of 46-51.4 mm, for quantitative detection of wheat components, spectral data of these 150 wheat samples, as well as spectral data from the first calibration plate and the second calibration plate during the acquisition process, were collected. The wheat spectrum is shown below. Figure 7 As shown in the figure. Each sample was collected three times using standard chemical analysis methods, and the average value was used as the reference value for that sample.

[0093] The wheat sample spectrum data is calibrated, corrected and band-selected pretreated according to the wheat sample spectrum data, the first calibration plate spectrum data and the second calibration plate spectrum data in the wheat collecting process, so as to obtain the pretreated wheat sample spectrum data. The wheat sample spectrum data and the reference value are used to train and verify the prediction model established by using the multivariate correction modeling method. After repeated training and comparison and analysis, finally, the model performance is optimal by using the spectrum data with the wavelength range of 1080-1650 nm combined with the above algorithm. The performance of the prediction model of the embodiment is shown in Table 4.

[0094] Table 4: Detection performance of wheat component content

[0095]

[0096] In another embodiment, several wheat sample component contents under breeding conditions need to be detected. Since the sample amount is small and precious, the breeding or drug mode needs to be used, the working distance is adjusted to 55-60 mm, the light spot diameter is 34-36 mm, the spectrum of the breeding wheat sample is collected, and the wheat prediction model established in the last embodiment is used to obtain the component content value of the breeding wheat sample.

[0097] According to Figures 4-7 and Tables 1-4, the detection device of the embodiment obtains a high correlation coefficient and a low error in all sample types, which fully proves that the detection device of the embodiment has high repeatability, high precision and high reliability.

[0098] The other structures of the embodiment are the same as those of Embodiment 1, which will not be described here.

[0099] Embodiment three

[0100] As Figure 8 shown, the embodiment of the application provides a non-contact diffuse reflection spectrum detection method based on medium wave near infrared, based on the detection device of Embodiment 1 or Embodiment 2, which comprises:

[0101] S1, according to the number of samples, the rarity degree and the particle size, the distance between the sample disc 7 and the protective window piece 3 is adjusted according to the working distance; the light source module 1 and the receiving module 5 are adjusted, so that the light spot area of the light source 101 is larger than the receiving area of the optical fiber 501 of the receiving module 5;

[0102] S2, the detection light emitted by the light source module 1 irradiates on the sample of the sample disc 7 through the protective window piece 3, the diffuse reflection light of the sample is collected by the optical fiber 501 of the receiving module 5 and sent to the spectrometer 6 for sample spectrum data collection;

[0103] S3, the sample stage 8 drives the sample disc 7 to rotate, and the spectrum acquisition is performed once per a preset angle until the spectrum acquisition of all samples to be measured is completed;

[0104] S4, the calibration module 4 moves the first calibration plate between the light source module 1 and the protective window sheet 3, and the current first calibration plate spectrum data is acquired by the spectrometer 6 for background calibration of the spectrometer 6;

[0105] S5, the calibration module 4 moves the second calibration plate between the light source module 1 and the protective window sheet 3, and the current second calibration plate spectrum data is acquired by the spectrometer 6 for wavelength calibration of the spectrometer 6;

[0106] S6, the processing and analysis module pre-processes the sample spectrum data, the first calibration plate spectrum data and the second calibration plate spectrum data received from the spectrometer 6, and then calculates the pre-processed sample spectrum data based on a preset algorithm and outputs the corresponding component content result.

[0107] Specifically, in step S1, the working mode required by the sample is determined, if the medicine or breeding mode is selected, the working distance is 55-60mm, and the spot diameter of the light source 101 is 34-36mm; if the normal working mode is selected, the working distance is 90-100mm, and the spot diameter is 46-51.4mm.

[0108] In this embodiment, the pre-processing includes calibration and correction of the sample spectrum data,

[0109] The calibration includes:

[0110] Based on the first calibration plate spectrum data and the preset wavelength characteristics, wavelength calibration data is generated;

[0111] According to the wavelength calibration data, the sample spectrum data and the second calibration plate spectrum data are respectively wavelength calibrated to obtain the wavelength calibrated sample spectrum data and the wavelength calibrated second calibration plate spectrum data;

[0112] Based on the wavelength calibrated second calibration plate spectrum data, the wavelength calibrated sample spectrum data is background-subtracted to obtain the relative spectral response value of the sample at each wavelength;

[0113] The calibrated sample spectrum data is determined by the relative spectral response value and the corresponding wavelength position after wavelength calibration;

[0114] The correction includes:

[0115] The calibrated sample spectrum data is subjected to first processing and second processing, the first processing suppresses spectral noise, and the second processing eliminates physical interference caused by particle scattering, optical path difference or baseline drift.

[0116] In addition, the preprocessing includes waveband screening, which is used to retain spectral data in a certain wavelength range, wherein:

[0117] Spectrum data in different wavelength ranges are selected to establish a prediction model. After repeated and large training and comparative analysis, for quantitative detection of rapeseed components, spectral data in a wavelength range of 1050-1630 nm are retained; for quantitative detection of rice bran components, spectral data in a wavelength range of 1100-1650 nm are retained; for quantitative detection of raw soybean powder components, spectral data in a wavelength range of 1100-1630 nm are retained; and for quantitative detection of wheat components, spectral data in a wavelength range of 1080-1650 nm are retained.

[0118] Optionally, in step S6, the preset algorithm adopts a prediction model, and the prediction model is obtained by the following method:

[0119] Spectrum data of a plurality of calibration samples are acquired, and the spectrum data are preprocessed to obtain preprocessed calibration sample spectrum;

[0120] Reference values of target components of each calibration sample measured by a standard chemical analysis method are acquired;

[0121] Based on the preprocessed calibration sample spectrum and the corresponding reference values, a multivariate correction modeling method is used to establish a prediction model of each target component;

[0122] The preprocessed sample spectrum data are calculated based on the preset algorithm, and corresponding component content results are output, including: the preprocessed sample spectrum data are input into the prediction model to obtain the component content results.

[0123] The multivariate correction modeling method can adopt one or more of a multivariate linear regression (MLR), a principal component regression (PCR), and a partial least squares regression (PLS) method.

[0124] In some embodiments, in the present embodiment, the prediction model comprises a linear regression sub-model, a nonlinear regression sub-model, and a fusion model, the preprocessed sample spectrum is input into the linear regression sub-model to obtain a first component content prediction value, the preprocessed sample spectrum is input into the nonlinear regression sub-model to obtain a second component content prediction value, based on a pre-established model confidence evaluation mechanism, the reliability of the first component content prediction value and the second component content prediction value is screened to determine whether to input both, input only one, or trigger abnormal processing; the screened prediction value is input into the fusion model, and the final component content prediction result is output through the fusion model. The linear regression sub-model adopts a partial least squares regression (PLS) model, the nonlinear regression sub-model adopts a random forest regression (RF) model, and the fusion model adopts a multi-layer perception (MLP) model. The present embodiment utilizes the linear modeling capability of the linear regression sub-model and the anti-noise and nonlinear fitting capability of the nonlinear regression sub-model, effectively overcomes the limitations of a single model, and introduces reliability screening and fusion, which can automatically suppress low-quality prediction results when the model outputs are inconsistent or the reliability is low, and enhances the fault tolerance to abnormal samples and noise interference.

[0125] In the present embodiment, the model confidence evaluation mechanism has a calibration sample spectrum set establishment, which comprises:

[0126] S601, for each calibration sample, respectively run the linear regression sub-model and the nonlinear regression sub-model using leave-one-out cross-validation, that is, each time train the model with the remaining samples, and predict the left-out sample, , to obtain the corresponding prediction values and , and calculate the residual of the true reference value :

[0127] ; ;

[0128] S602, divide the target component content range into intervals, for each interval ( ), respectively count the sample subset falling into the interval, and calculate the residual standard deviations and of the linear regression sub-model and the nonlinear regression sub-model on the subset;

[0129] S603, construct a confidence mapping table: when the model output prediction value falls into the interval, the confidence score is , where is the residual standard deviation of the corresponding model in the interval.

[0130] In detection, the sample spectrum data is input into the linear regression sub-model and the nonlinear regression sub-model to obtain a first component content prediction value and a second component content prediction value According to the first component content prediction value and the second component content prediction value The belonging component content interval is determined According to the confidence mapping table, corresponding If the relative deviation of the two prediction values is less than a preset threshold And the confidence is higher than a threshold Both are input into the fusion model, otherwise, only the model with higher confidence is output, or the confidence weighted fusion is performed to suppress abnormal prediction interference.

[0131] In summary, the receiving cone space formed by the light receiving angle of the optical fiber 501 is concentric with the light source 101 cone space formed by the light source module 1, so that the illumination area and the spectrum acquisition area completely coincide, and the spot area of the light source 101 is larger than the receiving area of the optical fiber 501, which ensures that the optical fiber 501 always receives stable, reliable and representative reflected light signals, avoiding signal loss or severe fluctuations. The focal length of the light source module 1 can be adjusted, and the spot size and energy density of the outgoing detection light on the sample surface can be dynamically adjusted. At the same time, the distance between the sample disc 7 and the protective window piece 3 is adjustable, which is used to accurately adjust the sample surface to the best focusing position of the detection light. By adjusting the focal length of the light source module 1 and the distance between the guide mirror 2 and the protective window piece 3, the optimal illumination condition and focusing state are configured, which significantly improves the signal-to-noise ratio, measurement repeatability and application range of the spectrum signal. The protective window piece 3 of the present application is inclinedly arranged, which can reflect the interference light reflected by the protective window piece 3 out of the collection range, preventing errors caused by mirror reflection light, and providing detection accuracy. In addition, the sample disc 7 is driven to rotate by the rotating sample table 8, and the rotation center of the sample disc 7 deviates from the light source 101 cone space, so that the sample disc 7 rotates eccentrically, which can irradiate different areas of the sample, effectively cover the spatial distribution of the whole batch of samples, avoid evaluation deviation caused by local abnormal particles or empty areas, significantly improve the overall component estimation representativeness, and reduce the spectral variation caused by the placement form, improve the repeatability and stability of multiple measurements of the same batch of samples. In addition, the preset algorithm used in the present application can enhance the fault tolerance to abnormal samples and noise interference, significantly improve the prediction accuracy and robustness, and improve the efficiency and accuracy.

[0132] ​​​The above merely describes the preferred embodiments of the present application, and it should be pointed out that, for those skilled in the art, several improvements and replacements can be made without departing from the technical principles of the present application, and these improvements and replacements should also be considered as the protection scope of the present application.

Claims

1. A detection device based on mid-wave near-infrared non-contact diffuse reflectance spectroscopy, characterized in that, include: The light source module (1) is used to collimate and converge the light source (101) to form a cone-shaped space of the light source (101) with adjustable focal length to irradiate the sample; The guide mirror (2) is used to guide the detection light emitted from the light source module (1) to the sample and couple the light beam reflected by the sample to the spectrometer (6) through the receiving module (5). A protective window (3) is provided between the guide mirror (2) and the sample. The protective window (3) is used to isolate the internal environment of the detection device from the external environment. The detection light emitted from the light source module (1) is reflected by the guide mirror (2) and transmitted through the protective window (3) to the sample. The protective window (3) is tilted to reduce the specular reflection light of the light source (101). The calibration module (4) is located between the guide mirror (2) and the protective window (3). The calibration module (4) is equipped with a switchable first calibration plate and a second calibration plate. The first calibration plate has characteristic absorption in the near-infrared mid-wave band and is used for wavelength calibration. The second calibration plate is used for background calibration. The receiving module (5) is provided with an optical fiber (501) that moves along the direction of approaching or moving away from the sample. The receiving cone space formed by the light receiving angle of the optical fiber (501) is concentric with the cone space of the light source (101) formed by the light source module (1). The movement of the optical fiber (501) keeps the light spot area of ​​the light source (101) larger than the receiving area area of ​​the optical fiber (501). A spectrometer (6) is connected to the receiving module (5). The photosensitive element contained in the spectrometer (6) performs photoelectric conversion on the light signal received from the receiving module (5) to generate sample spectral data. The processing and analysis module is connected to the spectrometer (6) and is used to preprocess the sample spectral data, and then calculate the preprocessed sample spectral data based on a preset algorithm and output the corresponding component content results. The sample module includes a sample tray (7) and a rotating sample stage (8). The rotating sample stage (8) is located at the intersection of the conical space of the light source (101) and the conical space of the receiving device. The sample tray (7) is used to hold the sample. The rotating sample stage (8) drives the sample tray (7) to rotate. The rotation center of the sample tray (7) is offset from the conical space of the light source (101). An adjustment module is provided for adjusting the distance between the sample tray (7) and the protective window (3). The light source module (1) includes a light source (101), a first lens (102), and a second lens (103). The detection light emitted from the light source (101) passes through the first lens (102), the second lens (103), the guide mirror (2), and the protective window (3) in sequence to illuminate the sample on the sample plate (7). The first lens (102) is used to convert the divergent light emitted by the light source (101) into parallel light, and the second lens (103) converts the parallel light into converging light. The first lens (102) and the second lens (103) are wrapped by an extinction lens tube (104) with an extinction structure and material to reduce the non-parallel light reaching the second lens (103). Alternatively, the light source module (1) includes a light source (101) and a first lens (102), and the guide mirror (2) is an off-axis parabolic reflector. The detection light emitted from the light source (101) passes through the first lens (102), the guide mirror (2), and the protective window (3) in sequence to illuminate the sample on the sample plate (7). The first lens (102) is used to convert the divergent light emitted by the light source (101) into parallel light. The guide mirror (2) converts the parallel light into converging light and guides it onto the sample on the sample plate (7). The first lens (102) is wrapped by an extinction mirror tube (104) with an extinction structure and material to reduce the non-parallel light reaching the guide mirror (2).

2. The detection device according to claim 1, characterized in that, The guide mirror (2) is tilted and has a through receiving hole, through which the optical fiber (501) passes.

3. The detection device according to claim 1, characterized in that, The optical axes of the first lens (102) and the second lens (103) are on the same straight line, and the first lens (102) and the second lens (103) are close to or far from each other along the direction of the optical axis.

4. A detection method based on mid-wave near-infrared non-contact diffuse reflectance spectroscopy, based on the detection device according to any one of claims 1-3, characterized in that, include: S1. Determine the working distance based on the quantity, rarity, and particle size of the samples, and adjust the distance between the sample tray (7) and the protective window (3) according to the working distance; adjust the light source module (1) and the receiving module (5) so that the spot area of ​​the light source (101) is larger than the receiving area of ​​the optical fiber (501) of the receiving module (5); in step S1, select the working mode according to the quantity, rarity, and particle size of the samples. There are two working modes: breeding or drug mode and normal mode; if it is breeding or drug mode, the working distance is 55-60mm and the spot diameter of the light source (101) is 34-36mm; if it is normal mode, the working distance is 90-100mm and the spot diameter of the light source (101) is 46-51.4mm. S2. The detection light emitted from the light source module (1) shines on the sample in the sample plate (7) through the protective window (3). The diffuse reflection light of the sample is collected by the optical fiber (501) of the receiving module (5) and sent to the spectrometer (6) for sample spectral data acquisition. S3. Rotate the sample stage (8) to drive the sample disk (7) to rotate. Perform a spectral acquisition once for each preset angle of rotation until all samples to be tested have completed spectral acquisition. S4. The calibration module (4) moves the first calibration plate between the light source module (1) and the protective window (3), and the spectrometer (6) collects the current spectral data of the first calibration plate for wavelength calibration of the spectrometer (6). S5. The calibration module (4) moves the second calibration plate between the light source module (1) and the protective window (3), and the spectrometer (6) collects the current spectral data of the second calibration plate for background calibration of the spectrometer (6). S6. The processing and analysis module preprocesses the sample spectral data, the first calibration plate spectral data, and the second calibration plate spectral data received from the spectrometer (6), and then calculates the preprocessed sample spectral data based on a preset algorithm and outputs the corresponding component content results.

5. The method according to claim 4, characterized in that, The preprocessing includes calibrating and correcting the spectral data of the sample. The calibration includes: Based on the spectral data of the first calibration plate and its preset wavelength characteristics, wavelength calibration data is generated; Based on the wavelength calibration data, wavelength calibration is performed on the sample spectral data and the second calibration plate spectral data respectively to obtain wavelength-calibrated sample spectral data and wavelength-calibrated second calibration plate spectral data. Based on the wavelength-calibrated second calibration plate spectral data, background subtraction is performed on the wavelength-calibrated sample spectral data to obtain the relative spectral response values ​​of the sample at each wavelength. The calibrated sample spectral data is determined by the relative spectral response value and its corresponding wavelength position after wavelength calibration. The correction includes: The calibrated sample spectral data are subjected to a first processing and a second processing. The first processing suppresses spectral noise, and the second processing eliminates physical interference caused by particle scattering, optical path difference, or baseline drift.

6. The method according to claim 4 or 5, characterized in that, The preprocessing includes band filtering, which is used to retain spectral data within a certain wavelength range, wherein: For quantitative analysis of rapeseed components, spectral data in the wavelength range of 1050–1630 nm were retained; for quantitative analysis of rice bran components, spectral data in the wavelength range of 1100–1650 nm were retained; for quantitative analysis of raw soybean flour components, spectral data in the wavelength range of 1100–1630 nm were retained; and for quantitative analysis of wheat components, spectral data in the wavelength range of 1080–1650 nm were retained.

7. The method according to claim 4, characterized in that, In step S6, the preset algorithm employs a prediction model, which is obtained through the following method: Spectral data of multiple calibration samples are acquired, and the spectral data are preprocessed to obtain the preprocessed calibration sample spectra. Obtain the reference values ​​of the target components for each calibration sample as determined by standard chemical analysis methods; Based on the preprocessed calibrated sample spectra and their corresponding reference values, a prediction model for each target component is established using a multivariate calibration modeling method. The preprocessed sample spectral data is calculated based on a preset algorithm, and the corresponding component content results are output, including: inputting the preprocessed sample spectral data into the prediction model to obtain the component content results.

Images