A rotary near-infrared moisture detection device and an angle domain compensation method

Abstract

This invention discloses a rotary near-infrared moisture detection device and an angle domain compensation method, belonging to the field of near-infrared optical detection technology. It is used to detect the moisture content of materials through a dual-wavelength measurement system, including a light source assembly, a photoelectric detection assembly, and a controller. The light source assembly includes a rotatable lampshade, a motor for driving the lampshade's rotation, and several LED light sources. The photoelectric detection assembly includes a photodetector, a transimpedance amplifier, a lock-in amplifier, and a high-speed acquisition card. The controller includes a central processing unit, a light source driving circuit, and a motor control module. The central processing unit includes a signal modulation module and a data processing module, and the motor control module includes a motor driving circuit and an angle acquisition module. This invention constructs a moisture measurement system based on dual-wavelength compensation, introducing angle domain processing and common-mode correction mechanisms during moisture detection. Based on joint modeling of spatial and spectral features, it enhances the device's anti-interference capability, thereby improving the accuracy of moisture detection.

Description

Technical Field

[0001] This invention relates to the field of near-infrared optical detection technology, and more particularly to the field of moisture detection technology based on near-infrared optical detection, specifically to a rotating near-infrared moisture detection device and an angle domain compensation method. Background Technology

[0002] Near-infrared (NIR) moisture detection typically utilizes the absorption characteristics of water molecules in the near-infrared band. It involves illuminating the sample with a near-infrared light source and collecting the intensity signals of the scattered or reflected light. This data is then combined with a calibration model to estimate the moisture content of the sample. To reduce the influence of factors such as light source intensity drift, changes in detection distance, fluctuations in overall sample reflectance, and ambient light, engineering practices often employ multi-wavelength compensation methods or spectrometer-based moisture detection methods. Spectrometer-based methods acquire full-spectrum data of the sample using a spectrometer and establish a regression model through machine learning and other modeling methods to perform non-destructive testing of the sample's moisture content. Multi-wavelength compensation methods calculate the measurement wavelength and reference wavelength to compensate for common noise caused by minor fluctuations in detection distance, environmental dust interference, and changes in the overall reflectance of the material, thereby improving the accuracy of moisture detection.

[0003] However, in applications involving grains, particulate matter, and rough-surface materials, the micro-geometry, particle orientation, particle size distribution, and local packing state of the sample surface can significantly alter the scattering path of incident light and the effective sampling volume, causing the light intensity received at the receiver to fluctuate significantly with the observation angle. Even under the same moisture conditions, different angles, different sample spreading states, or different batches of samples may produce significantly different light intensity responses, forming "inconsistent" perturbations that do not satisfy the strict common-mode assumption. This results in prediction errors that increase with increasing surface roughness of the sample when relying solely on wavelength domain ratios, conventional multivariate correction, or single sampling, leading to insufficient robustness of moisture detection methods.

[0004] To mitigate the angular sensitivity caused by surface roughness, existing solutions have employed methods such as expanding the calibration sample set to cover more surface states or using more complex multivariate calibration models. However, this significantly increases the cost of model building and maintenance, and the model's predictive performance drops sharply when encountering entirely new surface states not included in the model, lacking true robustness.

[0005] Therefore, there is an urgent need for a high-precision online moisture detection device that can address common interferences while specifically targeting and effectively suppressing the problem of "inconsistent" optical disturbances caused by changes in surface roughness, so as to improve the accuracy and robustness of moisture detection for rough surface materials. Summary of the Invention

[0006] The purpose of this invention is to provide a rotating near-infrared moisture detection device and an angle domain compensation method to solve the problems of low accuracy and insufficient robustness in the non-destructive detection of moisture in rough surface materials in the prior art.

[0007] To achieve the above objectives, the present invention provides a rotary near-infrared moisture detection device, comprising an optical acquisition component, a photoelectric detection component, and a controller; the optical acquisition component includes a lampshade, a motor for driving the lampshade to rotate, and several LED light sources; The photoelectric detection component includes a photodetector for photoelectric signal conversion and a high-speed acquisition card for acquiring the output signal of the photodetector and obtaining raw time-series data; The controller includes a central processing unit, a light source driving circuit, and a motor control module. The central processing unit includes a signal modulation module and a data processing module. The signal modulation module is used to modulate the emitted light signal and control the LED light source through the light source driving circuit. The data processing module is used to process the raw time-series data obtained by the high-speed acquisition card and output stable optical quantities for moisture characterization. The motor control module includes a motor driving circuit for controlling the motor and an angle acquisition module for acquiring the rotation angle information of the lampshade.

[0008] Furthermore, the motor is a hollow motor, and a conductive slip ring is provided in the hollow cavity of the hollow motor. The conductive slip ring is used to transmit a drive signal from the controller to the LED light source when the lamp cover rotates relative to the controller, and to transmit the rotation data of the lamp cover to the controller.

[0009] Furthermore, the LED light source includes a reference light source and a measurement light source. The center wavelength of the reference light source and the center wavelength of the measurement light source are different. The reference light source and the measurement light source are alternately arranged along the circumference of the lamp cover, and together they form a near-infrared light-emitting array.

[0010] Furthermore, the angle acquisition module is equipped with an encoder, which is integrated inside the motor. The encoder is used to collect the rotation information of the motor shaft and obtain the rotation angle information of the lampshade based on the rotation information of the motor shaft.

[0011] Furthermore, a sample tray for placing the sample to be tested is disposed below the lamp cover, and a photodetector is disposed on the lamp cover, with the light signal receiving end of the photodetector disposed directly above the sample tray.

[0012] Furthermore, a transimpedance amplifier and a lock-in amplifier are provided between the photodetector and the high-speed acquisition card. The input terminal of the transimpedance amplifier is connected to the electrical signal output terminal of the photodetector, the output terminal of the transimpedance amplifier is connected to the input terminal of the lock-in amplifier, the output terminal of the lock-in amplifier is connected to the input terminal of the high-speed acquisition card, and the output terminal of the high-speed acquisition card is connected to the central processing unit.

[0013] Furthermore, a motor mounting platform is provided at the top of the lampshade, and the motor is mounted on the top surface of the motor mounting platform. The rotor end of the motor is located on the central axis of the lampshade and is connected to the lampshade through the motor mounting platform. The stator end of the motor is connected to a device bracket to support the optical acquisition components.

[0014] To achieve the above objectives, the present invention also provides a method for angular domain compensation in a rotating near-infrared moisture detection system, using the aforementioned rotating near-infrared moisture detection device, comprising: S1. A rotatable lampshade is used to carry the reference light source and the measurement light source. The two light sources are driven by different modulation methods. The sample under test is irradiated from multiple angles by rotating the lampshade to obtain the original light intensity sequence, including the original reference channel sequence and the original measurement channel sequence. S2. Based on the rotation angle information of the lampshade, the original light intensity sequence is mapped from the time domain to the angle domain, and the original light intensity sequence is resampled at equal angles to obtain the angle domain light intensity sequence with equal angle intervals, including the angle domain reference light intensity sequence and the angle domain measured light intensity sequence. S3. In the angular domain, the common-mode perturbation component is estimated using the angular domain light intensity sequence with equal angular intervals. Based on the common-mode perturbation component, the angular domain light intensity sequence with equal angular intervals is subjected to common-mode correction to obtain the compensated angular domain light intensity sequence. S4. Perform angle domain statistical convergence on the compensated angle domain light intensity sequence to obtain converged optical quantities. Extract the spatial statistical features and spectral analytical features of the angle domain light intensity sequence from the converged optical quantities. Input the spatial statistical features and spectral analytical features into the multivariate correction model to output stable optical quantities used to characterize the moisture content of the sample.

[0015] S1 includes: S1.1 A rotatable lampshade is set above the sample tray on which the sample to be tested is placed. A reference light source and a measurement light source are alternately installed along the circumference of the lampshade. The lampshade is driven to rotate at a constant angular velocity, and modulation drives are applied to the reference light source and the measurement light source respectively. The center wavelength of the reference light source and the center wavelength of the measurement light source are different, and the modulation methods of the reference light source and the measurement light source are different. By rotating the lampshade, the two light sources illuminate the sample to be tested in sequence, so as to realize multi-angle illumination of the sample to be tested. S1.2, A photodetector is set directly above the sample plate. The photodetector receives the reference light reflection signal and the measurement light reflection signal after being reflected by the sample to be tested, and converts the light reflection signal into an electrical signal. The electrical signal output by the photodetector is processed sequentially by a transimpedance amplifier, a lock-in amplifier and a high-speed acquisition card to obtain the original light intensity sequence in the time domain.

[0016] In S3, the common-mode perturbation components include additive common-mode perturbation components and multiplicative common-mode perturbation components. The additive common-mode perturbation components are obtained by performing median filtering on the angle domain reference light intensity sequence in the angle domain. The multiplicative common-mode perturbation components are obtained based on the additive common-mode perturbation components using the ratio method. In S4, spatial statistical features include at least one of the following: mean, standard deviation, and quantiles; spectral analytical features include at least one of the following: the difference between the measurement channel and the reference channel, the normalized difference, and the ratio.

[0017] Compared with the prior art, the present invention has the following beneficial effects: This invention uses a rotating lampshade to drive the synchronous rotation of dual-wavelength LEDs, constructing a multi-angle dynamic sampling structure to acquire circumferential optical response information of the sample and reduce the randomness of single-point measurement. Combined with heterodyne modulation, it achieves effective separation of the measurement channel and the reference channel, and obtains a more stable moisture characterization through dual-wavelength compensation. It suppresses the non-uniform disturbances caused by surface roughness differences from both the signal acquisition and processing levels.

[0018] This invention introduces an angle domain processing and common-mode correction mechanism, mapping time-series data to the angle domain and estimating angle-related common-mode disturbance components. Combined with robust fitting / filtering and angle domain statistical convergence, it can suppress interference caused by speed fluctuations, mechanism eccentricity, jitter, and environmental noise. Furthermore, by using a compensation quality index to remove or reduce the weight of abnormal loop data, it reduces the contamination of results by abnormal data, making the near-infrared moisture detection device highly resistant to interference and improving its adaptability in industrial settings.

[0019] This invention overcomes the limitations of traditional single-spectral modeling by employing a combined modeling approach using spatial statistical features and spectral analytical features. Spatial statistical features characterize the geometric morphology and uniformity of the material surface, while spectral analytical features represent differences in moisture absorption. These two features complement each other and are combined with regression or machine learning algorithms for joint modeling, improving the model's generalization ability and predictive stability when dealing with different batches and sample preparation conditions. Attached Figure Description

[0020] Figure 1 This is the overall design principle diagram of the present invention; Figure 2 This is a schematic diagram of the overall mechanical structure of the present invention; Figure 3 This is a top view of the lampshade provided by the present invention; Figure 4 The training results of the angle domain compensation method provided by this invention in a random forest model are shown in the figure.

[0021] In the picture: 1. Controller; 11. Central Processing Unit; 12. Motor Control Module; 13. Light Source Drive Circuit; 2. Photoelectric Detection Components; 21. Photodetector; 22. Transimpedance Amplifier; 23. Lock-in Amplifier; 24. High-Speed ​​Acquisition Card; 3. Optical Acquisition Components; 31. Motor; 32. LED Light Source; 33. Lampshade; 34. LED Placement Hole; 35. Center Hole of Lampshade; 36. Motor Mounting Platform; 4. Sample Tray. Detailed Implementation

[0022] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention are described clearly and completely below. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0023] Example 1 like Figure 1 and Figure 2 The rotating near-infrared moisture detection device shown includes an optical acquisition component 3, a photoelectric detection component 2, and a controller 1.

[0024] The optical acquisition component 3 is a rotary optical acquisition component 3, including a lampshade 33, a motor 31, and several LED light sources 32; the lampshade 33 is trumpet-shaped, and the central axis of the lampshade 33 is collinear with the central axis of the motor 31. Driven by the motor 31, the lampshade 33 can rotate around the central axis of the lampshade 33. Several LED light sources 32 together form a near-infrared light emission array, which can emit near-infrared light of two wavelengths. The near-infrared light emission array includes a reference light source and a measurement light source alternately arranged along the circumference of the lampshade 33. The center wavelength of the reference light source and the center wavelength of the measurement light source are different.

[0025] Specifically, the center wavelength of the reference light source is the non-water absorption peak wavelength, preferably the 1000nm near-infrared light band, and the center wavelength of the measurement light source is the water absorption peak wavelength, preferably the 1450nm near-infrared light band.

[0026] Specifically, such as Figure 3 As shown, LED placement holes 34 are provided on the spherical surface of the lampshade 33 (equivalent to the inner curved surface of the trumpet-shaped lampshade 33) corresponding to the positions of each LED light source 32. Two LED light sources 32 with different wavelengths are arranged alternately around each other to form a rotating light source multi-angle sampling structure. The principal optical axis of each LED light source 32 forms a preset angle with the central axis of the photodetector 21. The preset angle is usually set to 45 degrees, which is equivalent to the incident light direction of the optical acquisition component forming a 45-degree angle with the receiving direction of the detector.

[0027] Specifically, the detector's central axis points vertically downwards, and the extensions of the principal optical axes of all LED light sources intersect at the center point of the sample surface. The angle between the principal optical axis of the LED light source and the vertical direction is 45 degrees, that is, the angle between the principal optical axis of the LED light source and the detector's central axis is 45 degrees. The detector's central axis passes through the center point of the sample surface and vertically receives diffuse reflected light.

[0028] Specifically, eight evenly arranged LED placement holes 34 are provided along the circumference of the lampshade 33, and two different wavelength LED light sources 32 are alternately placed in the corresponding LED placement holes 34.

[0029] Specifically, a motor mounting platform 36 is provided at the top of the lamp cover 33, and a motor 31 is provided on the top surface of the motor mounting platform 36. The rotor end of the motor 31 is connected to the motor mounting platform 36, and the stator end of the motor 31 is connected to an equipment bracket for supporting the rotating optical acquisition component 3. The equipment bracket and the motor mounting platform 36 together constitute the supporting rotation structure of the optical acquisition component 3.

[0030] In this embodiment, the motor adopts a hollow motor structure with a hollow cavity along the axial direction. A hollow servo motor 31 is preferred. A conductive slip ring is coaxially mounted within the hollow cavity of the hollow servo motor 31. This slip ring is used to transmit power and signals between the fixed structure (controller) and the rotating structure (such as the LED light source on the lampshade), solving the power supply and communication problems of the moisture detection device. The stator side (fixed side) of the conductive slip ring is electrically connected to the controller, and the rotor side (rotating side) is electrically connected to the LED light source on the lampshade and the angle acquisition module (such as an encoder) used to detect the rotation angle of the lampshade. The conductive slip ring, also called a slip ring or rotating electrical interface, is a precision device that can stably transmit current or data signals from the fixed structure to the rotating structure while the device rotates continuously at 360 degrees without restriction. It is used to power the motor 31 and transmit the rotational data of the lampshade 33 to the controller 1. When the motor drives the lampshade to rotate continuously, the conductive slip ring can stably transmit the modulated drive signal emitted by the light source drive circuit from the controller on the fixed side to the LED light source on the rotating side, ensuring that the LED light source continuously receives a stable operating current during the 360° rotation. Simultaneously, the conductive slip ring also transmits the real-time rotation angle data collected by the rotation angle acquisition module on the lampshade back to the motor control module in the controller, forming a closed-loop feedback. This allows the controller to accurately grasp the current rotation attitude of the lampshade. By setting up the conductive slip ring, the problem of non-contact transmission of power and data between the rotating and fixed components is solved, avoiding rotation restrictions or signal interruptions caused by cable entanglement. This ensures the reliability and real-time performance of light source driving and angle synchronous acquisition in multi-angle, continuous rotation detection modes.

[0031] Specifically, a sample tray 4 is positioned directly below the lampshade 33 to hold the sample to be tested. The central axis of the rotor end of the motor 31, the central axis of the motor mounting platform 36, and the central axis of the lampshade 33 are collinear. The motor mounting platform 36 connects the motor 31 and the lampshade 33, enabling synchronous rotation of the lampshade 33 and the rotor end of the motor 31. By setting up a device bracket and a motor mounting platform 36 on the optical acquisition component 3, and placing the sample tray 4 directly below the lampshade 33, a stable multi-angle circumferential illumination and scanning physical platform for the sample to be tested is constructed. The distance between the upper surface of the sample tray and the lower end of the lampshade is adjustable. By adjusting the distance between the sample tray and the lampshade, the diameter of the light spot illuminating the sample surface by each LED light source and the overlap of adjacent LED light spots are adjusted to achieve continuous angular sampling. During illumination, it is necessary to ensure that the light spots of adjacent LEDs partially overlap to maintain the continuity of the angular domain light intensity sequence.

[0032] The photoelectric detection assembly 2 includes a photodetector 21 for photoelectric signal conversion and a high-speed acquisition card 24 for acquiring the output signal of the photodetector 21 and obtaining raw timing data. A central hole 35 penetrating the lampshade 33 is provided on the top of the lampshade 33, located on its central axis. The photodetector 21 is disposed within the central hole 35. A transimpedance amplifier 22 and a lock-in amplifier 23 are provided between the photodetector 21 and the high-speed acquisition card 24. The input of the transimpedance amplifier 22 is connected to the photodetector 21, and its output is connected to the input of the lock-in amplifier 23. The output of the lock-in amplifier 23 is connected to the input of the high-speed acquisition card 24, and the output of the high-speed acquisition card 24 is connected to the central processing unit 11.

[0033] Specifically, the optical acquisition component 3 and the photoelectric detection component 2 are both located above the configured sample disk 4, and the main optical axis of each LED light source 32 forms a preset angle with the central axis of the photoelectric detector 21. The preset angle is usually set to 45 degrees.

[0034] Specifically, the photodetector 21 preferably uses an indium gallium arsenide (InGaAs) photodiode to receive light signals reflected or scattered by the sample under test and convert the light signals into electrical signals. InGaAs photodiodes have the characteristics of fast response time, high sensitivity, and stable operation over long periods of time, which can perfectly match the high-speed dynamic photoelectric signal capture requirements of the rotating mechanism. The effective detection wavelength range of InGaAs photodiodes is 800 nm to 1800 nm.

[0035] Specifically, the vertical distance between the center hole 35 of the lampshade and the LED placement hole 34 is 16 mm, and the horizontal distance is 20 mm.

[0036] The controller 1 includes a central processing unit 11, a light source driving circuit 13, and a motor control module 12. The central processing unit 11 includes a signal modulation module and a data processing module. The signal modulation module modulates the emitted light signals of each LED light source 32 and controls the LED light sources 32 through the light source driving circuit 13. The data processing module processes the raw time-series data obtained by the high-speed acquisition card 24 and outputs stable optical quantities for moisture characterization. The motor control module 12 controls the motor 31 driving circuit and the angle acquisition module for acquiring the rotation angle information of the lampshade 33. The central processing unit 11 outputs two control signals of different frequencies through the signal modulation module, and controls the reference light source and the measurement light source to flash synchronously at different frequencies through the light source driving circuit 13. The motor control module 12 of the central processing unit 11 drives the rotor end of the motor 31 to rotate through the motor 31 driving circuit, thereby driving the lampshade 33 to rotate along the central axis of the lampshade 33. After the rotating light source illuminates the sample under test, the diffusely reflected light from the sample is received by the photodetector 21. After processing by the transimpedance amplifier 22 and the lock-in amplifier 23, the data is acquired by the high-speed acquisition card 24. The data acquired by the high-speed acquisition card 24 is transmitted to the central processing unit 11 via a conductive slip ring. An encoder, integrated inside the motor 31, is installed in the angle acquisition module to acquire the rotation angle information of the lampshade 33 based on the rotation information of the motor 31 shaft. Based on the rotation angle information of the lampshade 33 acquired by the encoder, the original time-series data acquired by the photodetector component 2 for each revolution of the lampshade 33 is resampled at equal angles to obtain an angle-domain light intensity sequence with equal angle intervals.

[0037] The central processing unit 11 outputs two square waves of different frequencies through a signal modulation module, which drive the reference light source and the measurement light source to flash synchronously, respectively. A lock-in amplifier sets up two reference channels to lock the frequencies of the two square waves, and simultaneously demodulates the two frequency components, outputting light intensity signals corresponding to the two wavelengths. After processing by the lock-in amplifier, the data acquired by the high-speed acquisition card is already two separated voltage sequences. After synchronous sampling by the high-speed acquisition card, a pair of light intensity values ​​is obtained at each sampling moment.

[0038] When acquiring the angular domain light intensity sequence with equal angular intervals, since the two types of LED light sources are arranged alternately, each angular position corresponds to only one wavelength per rotation. Based on the encoder angle information and the physical arrangement order of the LED light sources, the sequence is split into a reference light intensity angle sequence and a measured light intensity angle sequence. The length of both sequences is 4 (a total of 8 LED light sources in the circumference, distributed alternately) or encrypted by interpolation.

[0039] The data processing module performs angular domain statistical aggregation on the reference light intensity angle sequence and the measured light intensity angle sequence, and performs equal-angle resampling based on the encoder angle information. This includes calculating the mean, variance, autocorrelation coefficient, etc. of both as spatial statistical features, calculating the ratio of the two and the mean and standard deviation of their angle sequences, or constructing a normalized difference index as a spectral resolution feature. The spatial statistical features and spectral resolution features are then input into an existing multivariate correction model, such as a regression model or a machine learning model, to output the moisture content.

[0040] A rotating near-infrared moisture detection device employs rotating frequency modulation technology and utilizes a dual-wavelength LED light source 32 and a photodetector 21 to construct the measurement system, providing the hardware foundation for dual-wavelength compensation. Under the control of the central processing unit 11, the measuring light source and the reference light source emit light at different modulation frequencies to illuminate the sample under test. The light signal reflected or scattered by the sample is received by the photodetector 21 and converted into an electrical signal. This electrical signal is demodulated by the lock-in amplifier 23 to obtain the original light intensity sequence, including the acquisition of the original reference channel sequence and the original measurement channel sequence. The original reference channel sequence corresponds to the demodulation amplitude of the center wavelength of the reference light source band, and the original measurement channel sequence corresponds to the demodulation amplitude of the center wavelength of the measurement light source band. The original light intensity sequence is acquired by a high-speed acquisition card 24 to obtain the original time-series data and sent to the data processing module of the central processing unit 11. The controller 1 controls the reference light source and the receiving light source to flash at different frequencies, and synchronously acquires the intensity signals of the measuring light and the reference light and their corresponding physical angle information through heterogeneous frequency demodulation technology.

[0041] Example 2 A method for angular domain compensation in a rotating near-infrared moisture detection system, using a rotating near-infrared moisture detection device as described in Example 1, includes: S1. A rotatable lampshade is used to carry the reference light source and the measurement light source. The two light sources are driven by different modulation methods. The sample under test is irradiated from multiple angles by rotating the lampshade to obtain the original light intensity sequence, including the original reference channel sequence and the original measurement channel sequence. S2. Based on the rotation angle information of the lampshade, the original light intensity sequence is mapped from the time domain to the angle domain, and the original light intensity sequence is resampled at equal angles to obtain the angle domain light intensity sequence with equal angle intervals, including the angle domain reference light intensity sequence and the angle domain measured light intensity sequence. S3. In the angular domain, the common-mode perturbation component is estimated using the angular domain light intensity sequence with equal angular intervals. Based on the common-mode perturbation component, the angular domain light intensity sequence with equal angular intervals is subjected to common-mode correction to obtain the compensated angular domain light intensity sequence. S4. Perform angle domain statistical convergence on the compensated angle domain light intensity sequence to obtain converged optical quantities. Extract spatial statistical features and spectral analytical features of the angle domain light intensity sequence from the converged optical quantities. Input the spatial statistical features and the spectral analytical features into the multivariate correction model to output stable optical quantities used to characterize the moisture content of the sample.

[0042] S1 includes: S1.1 A rotatable lampshade is set above the sample tray on which the sample to be tested is placed. A reference light source and a measurement light source are alternately installed along the circumference of the lampshade. The lampshade is driven to rotate at a constant angular velocity, and modulation drives are applied to the reference light source and the measurement light source respectively. The center wavelength of the reference light source and the center wavelength of the measurement light source are different, and the modulation methods of the reference light source and the measurement light source are different, so as to distinguish the light signals of the corresponding light sources. By rotating the lampshade, the two light sources illuminate the sample to be tested in sequence, so as to realize multi-angle illumination of the sample to be tested. S1.2, A photodetector is set directly above the sample plate to receive the reference light reflection signal and the measurement light reflection signal after being reflected by the sample under test. After the optical signal is converted into an electrical signal, it is processed by a transimpedance amplifier, a lock-in amplifier and a high-speed acquisition card in sequence to obtain the raw time series data.

[0043] Specifically, a hollow servo motor is installed on the lampshade to drive the lampshade to rotate around a central axis, which is a vertical axis. The central axis of the rotor of the hollow servo motor is collinear with the central axis of the lampshade. A photodetector is installed on the central axis of the lampshade. The controller controls the reference light source and the measurement light source to flash synchronously at different frequencies, and synchronously controls the lampshade to rotate at a fixed angular frequency. After the rotating light source illuminates the object under test, the photodetector receives the light reflected from the sample. After processing by a transimpedance amplifier and a lock-in amplifier, the raw timing data is acquired by a high-speed acquisition card and sent to the central processing unit 11 of the controller 1 through a conductive slip ring.

[0044] Specifically, an LED capable of emitting a near-infrared band with a center wavelength of 1450 nm is selected as the measurement light source, with a modulation frequency of 980 Hz. An LED capable of emitting a near-infrared band with a center wavelength of 1000 nm is selected as the reference light source, with a modulation frequency of 2028 Hz. A dual-wavelength detection basis is constructed by applying distinguishable modulation drives to the reference light source and the measurement light source.

[0045] Specifically, there is a certain harmonic relationship between 980 Hz and 2028 Hz. To avoid frequency crosstalk, a modulation frequency without integer multiples can be selected.

[0046] In S2, the rotation angle information corresponding to the original light intensity sequence is acquired synchronously. Based on the rotation angle information, the original light intensity sequence is mapped from the time domain to the angle domain, and equal-angle resampling is performed to eliminate the sampling unevenness caused by the rotation speed fluctuation, so as to obtain the angle domain light intensity sequence with equal angle intervals.

[0047] Based on encoder-triggered sampling, the original light intensity sequence is sampled at equal angles to obtain an equally spaced angle sequence: ; In the formula, Indicates the first Each sampling angle, Indicates the sampling index. The sampling interval is indicated by resampling the original light intensity sequence at equal angles based on the equally spaced angle sequence to obtain the angle-domain measured light intensity sequence and the angle-domain reference light intensity sequence.

[0048] Equal-angle sampling of the original light intensity sequence based on encoder-triggered sampling includes two sampling methods. If the encoder triggers sampling once every fixed angle increment, that is, the encoder outputs a trigger signal at a fixed angle interval (e.g., every 1°), the high-speed acquisition card synchronously acquires the dual-wavelength light intensity values ​​at the corresponding angle, directly obtaining the equal-angle-interval angle-domain light intensity sequence. If the encoder only provides an angle reference (e.g., Z-phase pulse), then the encoder needs to record the angle at each sampling moment, and then linear interpolation is used to map the non-uniform time-domain sequence into an equal-angle-interval sequence.

[0049] In S3, common-mode disturbance components caused by rotation mechanism jitter, eccentricity, or changes in material surface morphology are estimated using an angle-domain light intensity sequence with equal angular intervals within the angle domain. The equally spaced light intensity sequence is then corrected using robust regression or filtering methods to obtain a compensated angle-domain light intensity sequence. The common-mode disturbance components include additive and multiplicative common-mode disturbance components. The additive common-mode disturbance component is obtained by median filtering of the angle-domain reference light intensity sequence within the angle domain. The multiplicative common-mode disturbance component is obtained based on the additive common-mode disturbance component using a ratio method.

[0050] Since the optical paths of the two wavelengths almost overlap and are acquired synchronously, the effects of the rotating mechanism jitter, eccentricity, or changes in the surface morphology of the sample on the two channels are common and are called common-mode disturbance components.

[0051] In a light intensity sequence with equal angular intervals, Measuring light intensity for: ; In the formula, express The corresponding ideal measured light intensity, express Multiplicative common-mode perturbation components, express The additive common-mode perturbation component, This indicates the noise in the measurement channel, which is the transmission channel for the light signal emitted by the measurement light source. In a light intensity sequence with equal angular intervals, Reference light intensity for: ; In the formula, express The corresponding ideal reference light intensity, This represents the noise of the reference channel, which is the transmission channel for the optical signal emitted by the reference light source. At the same sampling angle, the reference channel and the measurement channel have the same common-mode disturbance component.

[0052] The additive common-mode perturbation component is obtained by performing median filtering on the angular domain reference light intensity sequence: ; In the formula, This represents the half-window width of the median filter. Median filtering based on a sliding window calculates the additive common-mode disturbance component. The sliding window is... The total width of the sliding window is , Indicates median filtering. for The corresponding estimated values ​​of additive perturbations. At the boundaries of the angular domain sequence, a symmetrical expansion or contraction of the window width is used to ensure that an estimated value is available for each angular position.

[0053] Additive common-mode perturbation components are removed from the angular domain light intensity sequence with equal angular intervals. Then, the ratio of the measured light intensity after removing the additive perturbation to the reference light intensity is calculated using the ratio method. Finally, the multiplicative common-mode perturbation components are obtained by processing the median filtered comparison value based on a sliding window. ; ; ; ; In the formula, for The estimated value of the corresponding multiplicative perturbation reflects the multiplicative gain deviation caused by changes in scattering path length, etc., at the corresponding angle. for The ratio of the measured light intensity after removing additive perturbations to the reference light intensity. Indicates to The value after median filtering is equivalent to the value of the measured / reference ratio sequence after removing additive perturbations and then applying median filtering. Represents the global median, which is the ideal ratio without multiplicative disturbances; after common-mode correction and before angular domain statistical convergence, a dual-wavelength compensation is added based on the measurement channel and reference channel after common-mode correction. The dual-wavelength compensation includes at least one of the ratio and normalized difference.

[0054] Formula to be executed after compensation: ; ; Based on common-mode disturbance components and Common-mode correction is performed on the angular domain intensity sequences with equal angular intervals to obtain the compensated angular domain intensity sequences, including and .

[0055] In S4, the compensated angle-domain light intensity sequence is statistically aggregated in the angle domain to extract spatial statistical features such as the mean, standard deviation, robust ratio, and quantile of the measured light intensity and the reference light intensity, as well as spectral analytical features such as the difference, normalized difference, reflectance ratio, surface uniformity index, and absorption contrast. These features are then input into an existing regression model or an existing machine learning model to output a stable optical quantity characterizing the moisture content of the sample.

[0056] Specifically, in S4, the compensated angle domain light intensity sequence is statistically averaged or robustly statistically analyzed within an angle window over one loop, and data from multiple loops can be statistically aggregated at the same angle point.

[0057] Specifically, a two-dimensional feature vector is constructed based on the extracted spatial statistical features and spectral analysis features, which serves as the input to the existing multivariate correction model.

[0058] Data feature extraction is performed on the compensated light intensity data. The light intensity data of one circle is directly calculated using formulas, and then extracted based on spatial statistical features and spectral analytical features: Spatial statistical features include mean, standard deviation (std), robust ratio, signal quality index, intensity range, and quantile. The mean feature represents the average level of light intensity, reflecting the overall energy intensity, and includes the reference light intensity mean and the measured light intensity mean. The standard deviation feature represents the spatial dispersion of light intensity, reflecting the surface inhomogeneity or disturbance amplitude, and includes the standard deviation of the reference light intensity sequence and the standard deviation of the measured light intensity sequence. The quantile features include the upper quantile, the median quantile, and the lower quantile.

[0059] The formula for calculating the mean characteristic is: ; In the formula, This represents the average light intensity. Indicates the first The light intensity corresponding to each sampling angle , indicating the sampling index, This indicates the total number of sampling angles.

[0060] The formula for calculating the standard deviation characteristic is: ; In the formula, This represents the standard deviation of the light intensity sequence.

[0061] The formula for calculating the robustness ratio is: ; In the formula, This represents the robust ratio of the median of the measured light intensity angular domain sequence to the median of the reference light intensity angular domain sequence. This represents the median of the angular domain sequence of measured light intensity (1450 nm). This represents the median of the reference light intensity (1000 nm) angular domain sequence.

[0062] The formula for calculating the signal quality index characteristic is: ; In the formula, signal_quality represents the signal quality index, which is equivalent to the ratio of the mean to the standard deviation, and is used to reflect the signal-to-noise ratio level of the signal. The average light intensity. The standard deviation is denoted as .

[0063] The formula for calculating the characteristic of light intensity intervals is: ; In the formula, Indicates the range of light intensity values. This represents the maximum value of the light intensity in the light intensity sequence. This represents the minimum light intensity in the light intensity sequence.

[0064] Based on the magnitude of the light intensity values, the light intensities in the light intensity sequence are arranged from smallest to largest. The lower quantile is the 25th percentile of the light intensity sequence, indicating that 25% of the light intensity values ​​in the light intensity sequence are below the lower quantile. The middle quantile is the median value of the light intensity sequence. The upper quantile is the 75th percentile of the light intensity sequence, indicating that 75% of the light intensity values ​​in the light intensity sequence are below the upper quantile.

[0065] The spectral resolution features include mean diff, normalized difference, reflectance ratio, surface uniformity index, and absorption contrast, with the corresponding formulas as follows: The formula for calculating the difference feature is: ; In the formula, This represents the difference between the reference light and the measurement light. This represents the average reference light intensity. This represents the average measured light intensity.

[0066] The formula for calculating the normalized difference is: ; In the formula, This represents the difference after symmetric normalization.

[0067] Reflectivity ratio The calculation formula is: ; Surface uniformity index The calculation formula is: ; In the formula, The standard deviation of the measured light intensity sequence This represents the standard deviation of the reference light intensity sequence. It is an absolute value function; The formula for calculating absorption contrast is: .

[0068] Specifically, a two-dimensional feature vector is constructed based on the extracted spatial statistical features and spectral analytical features. This vector is then used as input to the random forest model for training. The final training result of the model is as follows: Figure 4 As shown, Figure 4 The green scatter plot represents the moisture content values ​​predicted by the random forest model, the red dashed line represents the ideal prediction line, which represents the actual humidity, the vertical axis represents the actual moisture content (%), and the horizontal axis represents the predicted moisture content (%). The coefficient of determination represents the goodness of fit of the model. The value ranges from 0 to 1; the closer to 1, the better the model fits the data. RMSE, or Root Mean Square Error, measures the average error between the predicted and actual values; a smaller value indicates a more accurate prediction. Figure 4 It can be observed that, The value is 0.9927, very close to 1, indicating that the prediction model can explain 99.27% ​​of the actual humidity changes, showing an excellent fit. The RMSE value is 0.3100 (RMSE = 31% if the moisture content is expressed as a percentage), a relatively small RMSE value, indicating a small average error between the predicted and actual humidity. Overall, Figure 4 of high Both the low RMSE value and the low RMSE value indicate that the prediction model has high accuracy and reliability, and the prediction results can well reflect the actual humidity situation, showing a high degree of consistency between the predicted humidity and the actual humidity. Furthermore... Figure 4 The scatter points closely surround the ideal prediction line, further verifying that the joint features based on spatial statistical information and spectral analysis information have a good effect on training the moisture prediction model.

[0069] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A rotary near-infrared moisture detection device, characterized by comprising: It includes an optical acquisition component, a photoelectric detection component, and a controller; the optical acquisition component includes a lampshade, a motor for driving the lampshade to rotate, and several LED light sources; The photoelectric detection component includes a photoelectric detector for photoelectric signal conversion and a high-speed acquisition card for acquiring the output signal of the photoelectric detector and obtaining raw time-series data. The controller includes a central processing unit, a light source driving circuit, and a motor control module. The central processing unit includes a signal modulation module and a data processing module. The signal modulation module is used to modulate the emitted light signal and control the LED light source through the light source driving circuit. The data processing module is used to process the raw time-series data obtained by the high-speed acquisition card and output stable optical quantities for moisture characterization. The motor control module includes a motor driving circuit for controlling the motor and an angle acquisition module for acquiring the rotation angle information of the lampshade.

2. The rotating near-infrared moisture detection device according to claim 1, wherein The motor is a hollow motor, and a conductive slip ring is provided in the hollow cavity of the hollow motor. The conductive slip ring is used to transmit a drive signal from the controller to the LED light source when the lamp cover rotates relative to the controller, and to transmit the rotation data of the lamp cover to the controller.

3. The rotating near-infrared moisture detection device according to claim 1, wherein The LED light source includes a reference light source and a measurement light source. The center wavelength of the reference light source and the center wavelength of the measurement light source are different. The reference light source and the measurement light source are alternately arranged along the circumference of the lamp cover. The reference light source and the measurement light source together form a near-infrared light-emitting array.

4. The rotating near-infrared moisture detection device according to claim 1, wherein The angle acquisition module is equipped with an encoder, which is integrated inside the motor. The encoder is used to collect the rotation information of the motor shaft and obtain the rotation angle information of the lampshade based on the rotation information of the motor shaft.

5. The rotating near-infrared moisture detection device of claim 1, wherein, A sample tray for placing the sample to be tested is disposed below the lampshade. A photodetector is disposed on the lampshade, and the light signal receiving end of the photodetector is disposed directly above the sample tray.

6. The rotating near-infrared moisture detection device of claim 1, wherein, A transimpedance amplifier and a lock-in amplifier are provided between the photodetector and the high-speed acquisition card. The input terminal of the transimpedance amplifier is connected to the electrical signal output terminal of the photodetector, the output terminal of the transimpedance amplifier is connected to the input terminal of the lock-in amplifier, the output terminal of the lock-in amplifier is connected to the input terminal of the high-speed acquisition card, and the output terminal of the high-speed acquisition card is connected to the central processing unit.

7. The rotating near-infrared moisture detection device of claim 1, wherein, The top of the lampshade is provided with a motor mounting platform, the motor is mounted on the top surface of the motor mounting platform, and the rotor end of the motor is located on the central axis of the lampshade and is connected to the lampshade through the motor mounting platform; The stator end of the motor is connected to a device bracket for supporting the optical acquisition component.

8. A method for compensating for angle domain in a rotary near-infrared moisture detection, characterized in that, Using the rotary near-infrared moisture detection device according to claim 3, comprising: S1. A rotatable lampshade is used to carry the reference light source and the measurement light source. The two light sources are driven by different modulation methods. The sample under test is irradiated from multiple angles by rotating the lampshade to obtain the original light intensity sequence, including the original reference channel sequence and the original measurement channel sequence. S2. Collect the rotation angle information of the lampshade, map the original light intensity sequence from the time domain to the angle domain based on the rotation angle information of the lampshade, and perform equal angle resampling on the original light intensity sequence to obtain the angle domain light intensity sequence with equal angle intervals, including the angle domain reference light intensity sequence and the angle domain measured light intensity sequence. S3. In the angular domain, the common-mode perturbation component is estimated using the angular domain light intensity sequence with equal angular intervals. Based on the common-mode perturbation component, the angular domain light intensity sequence with equal angular intervals is subjected to common-mode correction to obtain the compensated angular domain light intensity sequence. S4. Perform angle domain statistical convergence on the compensated angle domain light intensity sequence to obtain converged optical quantities. Extract spatial statistical features and spectral analytical features of the angle domain light intensity sequence from the converged optical quantities. Input the spatial statistical features and the spectral analytical features into the multivariate correction model to output stable optical quantities used to characterize the moisture content of the sample.

9. The angle domain compensation method for a rotating near-infrared moisture detector according to claim 8, characterized in that, S1 includes: S1.1 A rotatable lampshade is set above the sample tray on which the sample to be tested is placed. A reference light source and a measurement light source are alternately installed along the circumference of the lampshade. The lampshade is driven to rotate at a constant angular velocity, and modulation drives are applied to the reference light source and the measurement light source respectively. The center wavelength of the reference light source and the center wavelength of the measurement light source are different, and the modulation methods of the reference light source and the measurement light source are different. By rotating the lampshade, the two light sources illuminate the sample to be tested in sequence, so as to realize multi-angle illumination of the sample to be tested. S1.2, A photodetector is set directly above the sample plate. The photodetector receives the reference light reflection signal and the measurement light reflection signal after being reflected by the sample to be tested, and converts the light reflection signal into an electrical signal. The electrical signal output by the photodetector is processed sequentially by a transimpedance amplifier, a lock-in amplifier and a high-speed acquisition card to obtain the original light intensity sequence in the time domain.

10. The angle domain compensation method for a rotating near-infrared moisture detector according to claim 8, characterized in that, In S3, the common-mode perturbation components include additive common-mode perturbation components and multiplicative common-mode perturbation components. The additive common-mode perturbation components are obtained by performing median filtering on the angular domain reference light intensity sequence in the angular domain. The multiplicative common-mode perturbation components are obtained based on the additive common-mode perturbation components using the ratio method. In S4, the spatial statistical characteristics include at least one of the following: mean, standard deviation, and quantiles. The spectral resolution characteristics include at least one of the following: the difference between the measurement channel and the reference channel, the normalized difference, and the ratio.

Images