Near-infrared moisture measurement method and device for on-line moving material

Abstract

The invention discloses a near-infrared moisture measurement method and a near-infrared moisture measurement device for an online moving material, belongs to the technical field of near-infrared moisture measurement, and aims to solve the problems of poor adaptability, low measurement precision under a dynamic working condition and large spatial difference interference in traditional near-infrared moisture measurement. The method comprises the following steps: generating an adaptive target wavelength combination based on specific attributes of an online moving material; a transmission trigger signal is dynamically generated in combination with a real-time transmission state, time-sharing transmission of a light source is controlled, a reflection signal is collected at a dynamic frequency in an exclusive collection window, and a standardized data set is generated through adaptive gain correction and binding of coordinates of a detection area; and extracting multi-dimensional correlation spectral features based on the set, generating an iterative initial feature difference value through space consistency calibration, correlation index table construction and detection unit division, and calculating the moisture content through a difference value iterative calibration method. According to the invention, accurate and stable measurement of the moisture of the on-line moving material under the dynamic working condition is realized, and the measurement suitability and reliability are obviously improved.

Description

Technical Field

[0001] This invention relates to the field of near-infrared moisture measurement technology, and more specifically to a near-infrared moisture measurement method and apparatus for online moving materials. Background Technology

[0002] Near-infrared spectroscopy moisture measurement technology, with its outstanding advantages of being non-contact, rapid, and non-destructive, has become a core technology for online detection of moisture content in moving materials in various fields such as grain, chemical, and building materials, playing a crucial role in quality control and process optimization in industrial production. In actual industrial testing scenarios, moving materials often exhibit complex and variable characteristics. Different batches of materials have significant differences in inherent properties such as matrix composition, morphology, moisture content range, and distribution. At the same time, operating parameters such as the speed and bulk density of the material fluctuate in real time during transportation, and the spatial distribution of the material within the detection area is difficult to maintain uniformly. These objectively existing variables place stringent requirements on the accuracy and stability of near-infrared moisture measurement.

[0003] However, existing online near-infrared moisture measurement technologies for moving materials have significant limitations when dealing with the aforementioned complex scenarios. Current technologies generally use pre-set fixed near-infrared wavelength combinations for measurement. These wavelength combinations are mostly developed and determined for materials in specific states, and cannot flexibly adapt to the differences in properties between different batches of materials. This easily leads to confusion between moisture characteristic signals and matrix interference signals, directly affecting the accurate identification of moisture content. At the same time, the signal acquisition parameters of existing technologies are mostly static preset values, which cannot be dynamically adjusted according to changes in operating conditions during material transportation. This results in a mismatch between the timing and frequency of signal acquisition and the actual transportation state of the material. The acquired reflected signals are difficult to accurately reflect the moisture characteristics of the material, further reducing the reliability of subsequent data processing. In addition, when calculating moisture content, existing measurement methods often do not fully consider the spatial distribution differences of the material within the detection area, and lack effective means to suppress common interferences such as matrix component interference, light source intensity fluctuations, and changes in material thickness during transportation. This allows interference signals to mix into the moisture characteristic signals, ultimately causing the measurement results to fail to accurately reflect the true moisture state of the material. Summary of the Invention

[0004] To address the shortcomings of existing technologies, the present invention aims to provide a near-infrared moisture measurement method and apparatus for online moving materials, solving the problem of how to improve the adaptability of near-infrared moisture measurement of online moving materials to differences in material-specific properties, fluctuations in dynamic transmission conditions, and spatial distribution differences, and to reduce the influence of various interferences in order to achieve accurate acquisition of material moisture content.

[0005] To achieve the above objectives, the present invention provides the following technical solution:

[0006] A near-infrared moisture measurement method for online moving materials, comprising:

[0007] Based on the unique properties of online moving materials, a target wavelength combination adapted to online moving material moisture measurement is generated;

[0008] Based on the target wavelength combination and the real-time transmission status of the online moving material, the emission trigger signal of the near-infrared light source is dynamically calculated and output. The near-infrared light source is controlled to emit near-infrared light according to the time-division emission sequence. In the dedicated acquisition window of each wavelength, the original reflection signal data is acquired based on the dynamic acquisition frequency. Through adaptive gain correction, the coordinates of the detection area are bound to generate a standardized reflection signal data set.

[0009] Based on a standardized set of reflection signal data, multi-dimensional correlated spectral features are extracted, and spatial consistency calibration is performed in combination with the coordinates of the detection area. Using the standard benchmark features of spatial matching as a reference, an index table of correlation between multi-dimensional correlated spectral features and the coordinates of the detection area is constructed, and detection units are divided. Then, the initial feature difference of each detection unit is generated, and the difference iterative calibration method is used to calculate the moisture content of the material.

[0010] Specifically, the steps for generating a target wavelength combination suitable for online mobile material moisture measurement include:

[0011] Acquire the unique properties of online moving materials, including material matrix composition, material morphology and structure, moisture content range, and moisture distribution state;

[0012] Based on the preset wavelength adaptation rule library, the feasible range of near-infrared light wavelengths corresponding to the unique attributes of online moving materials is matched.

[0013] The intersection of the feasible regions of near-infrared light wavelengths corresponding to each unique attribute obtained by matching is solved to obtain the wavelength range that simultaneously meets the constraints of all unique attributes of online moving materials, which is used as the candidate wavelength pool.

[0014] Obtain the wavelength interval length of the candidate wavelength pool. If the wavelength interval length of the candidate wavelength pool is not less than the preset effective length threshold, the candidate wavelength pool is determined to be valid; otherwise, the candidate wavelength pool is determined to be invalid.

[0015] The minimum interval and maximum number of wavelengths that can be selected are calculated based on the preset signal independence threshold and the wavelength range length of the candidate wavelength pool.

[0016] Traverse all wavelengths in the candidate wavelength pool, filter out wavelengths whose adjacent wavelength interval is not less than the minimum interval, and construct multiple initial wavelength combinations; wherein, the upper limit of the number of wavelengths in the multiple initial wavelength combinations is the maximum number of combinations.

[0017] Calculate the uniformity of wavelength distribution within the candidate wavelength pool range for each initial wavelength combination, and select the target wavelength combination from multiple initial wavelength combinations.

[0018] Specifically, the steps for dynamically calculating and outputting the emission trigger signal of the near-infrared light source include:

[0019] The system acquires real-time data on the online moving material transmission density and speed, retrieves the length of the material detection unit, and quantitatively calculates the basic triggering cycle of the light source emission based on the ratio of the material detection unit length to the real-time transmission speed.

[0020] Based on the attenuation coefficient threshold of near-infrared light in materials with different transmission densities, the transmission density is classified into different levels to obtain the real-time transmission density classification interval of the online moving material, and the gain coefficient corresponding to the transmission density classification interval is matched.

[0021] The number of wavelengths included in the target wavelength combination is counted, and a time slot allocation matrix for time-division transmission is constructed. The time slot allocation matrix is ​​used to record a structured data set of transmission timing parameters of each wavelength in the target wavelength combination, including a one-to-one mapping relationship between wavelength identifiers and corresponding transmission time slot intervals.

[0022] Based on the basic triggering period and time slot allocation matrix, the transmission triggering time corresponding to each wavelength in a single triggering process of the target wavelength combination is quantitatively calculated, and a transmission triggering signal containing wavelength identification field, triggering time field, attenuation coefficient field, and time slot interval field is constructed.

[0023] Specifically, within a dedicated acquisition window for each wavelength, the steps for acquiring raw reflected signal data based on the dynamic acquisition frequency include:

[0024] The system collects real-time data on the transmission location and speed fluctuation of online moving materials to determine whether the material has entered the near-infrared light irradiation detection area. If so, it calculates the trigger compensation delay based on the transmission speed fluctuation and a preset compensation coefficient to determine the actual output time of the trigger signal.

[0025] The basic acquisition frequency of the near-infrared detector is calculated based on the ratio of the material detection unit length to the transmission speed; the transmission speed is continuously statistically analyzed using a sliding window to obtain the transmission speed fluctuation amplitude within the sliding window; the corresponding acquisition frequency correction coefficient is matched according to the graded interval of the transmission speed fluctuation amplitude; the basic acquisition frequency is multiplied by the acquisition frequency correction coefficient to obtain the dynamic acquisition frequency of the near-infrared detector.

[0026] Retrieve the time slot interval information from the transmission trigger signal and configure a dedicated acquisition window for each wavelength;

[0027] Within a dedicated acquisition window for each wavelength, based on the dynamic acquisition frequency, the near-infrared light signal reflected from the material surface is received by a near-infrared detector to obtain the original reflected signal data corresponding to each wavelength.

[0028] Specifically, the steps for generating a standardized set of reflected signal data include:

[0029] Retrieve the material transport density at the time of acquisition corresponding to the original reflection signal data of each wavelength, and perform noise reduction operation on the original reflection signal data according to the preset correlation rule between transport density and noise type;

[0030] The preset transmission density and gain compensation coefficient mapping table is retrieved, the signal gain compensation coefficient corresponding to the current transmission density is extracted, the intensity value of the denoised reflected signal data is multiplied with the gain compensation coefficient to complete the adaptive gain correction of the reflected signal data; and the baseline correction and extreme value normalization are performed on the adaptive gain corrected reflected signal data to obtain the preprocessed reflected signal data.

[0031] Retrieve the time slot allocation matrix and trigger time information from the transmission trigger signal, sort the preprocessed reflection signal data using the wavelength transmission time sequence as an index, and calibrate and align the acquisition timestamps of each wavelength reflection signal data.

[0032] Based on the real-time transmission speed of each wavelength reflection signal data within the corresponding acquisition period, and combined with the duration of the acquisition period, the material movement distance within the acquisition period is calculated, and the material detection area coordinates corresponding to each wavelength reflection signal data segment are corrected to obtain the material detection area coordinates.

[0033] Each wavelength reflection signal data segment is bound to its corresponding wavelength identifier, trigger time, and corrected detection area coordinates to construct a multi-dimensional parameter association table;

[0034] The sorted wavelength reflection signal data segments and the corresponding material transmission parameters at the acquisition time are entered one by one into a multi-dimensional parameter association table to form a reflection signal data set with material transmission parameter identifiers. Then, the format standardization process is performed to form a standardized reflection signal data set.

[0035] Specifically, the multi-dimensional correlated spectral features include: core moisture response features, interference cancellation features, and operating condition correlated features;

[0036] The steps for extracting multidimensional correlated spectral features include:

[0037] Retrieve the absorbance time-series data, acquisition time period identifier, and associated material transport parameters for each wavelength from the standardized reflection signal data set;

[0038] The core characteristic wavelengths and reference wavelengths for classifying target wavelength combinations are based on a rule base for near-infrared spectral absorption characteristics.

[0039] For the absorbance time series data corresponding to the core feature wavelengths, core moisture response features are extracted. The core moisture response features include the absorbance mean feature, the first-order difference feature of absorbance between adjacent wavelengths, and the absorbance time series extreme value difference feature.

[0040] Based on the absorbance time series data of the core feature wavelength and the reference wavelength, the mean of the absorbance time series data of the reference wavelength is selected as the dynamic interference reference value. By calculating the ratio of the mean absorbance of each core feature wavelength to the dynamic interference reference value, interference cancellation features are constructed.

[0041] By combining the wavelength absorbance time series data of the core characteristic wavelength with the associated material transport parameters, the operating condition correlation features are extracted. The operating condition correlation features include the coupling feature between the core wavelength absorbance and the transport density, and the correlation feature between the first-order difference of the core wavelength absorbance and the transport speed.

[0042] The three feature subsets—core moisture response features, interference cancellation features, and operating condition correlation features—are integrated and sorted in a structured manner to form multi-dimensional correlation spectral features.

[0043] Specifically, the steps for generating the initial feature difference values ​​for each detection unit iteration include:

[0044] The detection area coordinates bound to the multi-dimensional correlated spectral features and standardized reflectance signal data are retrieved to construct an association index table between the multi-dimensional correlated spectral features and the detection area coordinates; and the detection units are divided according to the spatial distribution pattern of the detection area coordinates; spatial consistency preprocessing is performed on the multi-dimensional correlated spectral features in each detection unit.

[0045] Retrieve the preset standard material spectral database, filter the standard multidimensional correlation spectral feature data of standard materials that are consistent with the unique properties of the currently online moving materials, and perform detection unit division on the standard material data;

[0046] The core moisture response features within each detection unit are extracted as spatial matching standard benchmarks and associated spectral features, which are then used to calibrate the standard moisture content of standard material data.

[0047] Based on the association index table of multi-dimensional correlation spectral features and detection area coordinates, the core moisture response features of online moving materials and standard materials are matched one by one according to the detection unit, and the difference between the core moisture response features of the two is calculated.

[0048] The interference cancellation features corresponding to each detection unit of the online moving material are retrieved, and the interference stripping process is performed on the core moisture response feature difference to obtain the core moisture response feature difference after purification.

[0049] The working condition association features bound to each detection unit are retrieved, and the difference between the current transmission working condition and the standard material test working condition is calculated as the working condition constraint compensation amount. Combined with the spatial attenuation coefficient corresponding to the detection area coordinates, the initial compensation amount of the iteration is obtained. Substituted into the difference of the core moisture response features after purification, the initial feature difference of each detection unit is generated.

[0050] Specifically, the steps for calculating the moisture content of materials using the difference iterative calibration method include:

[0051] Based on the inherent correspondence between the standard moisture content calibration data of standard materials and the corresponding standard reference spectral characteristics, and combined with the correction effect of the iterative initial feature difference on the standard correlation, the initial moisture content of each detection unit of the online moving material is derived.

[0052] The initial moisture content is back-mapped to the core moisture response feature space to generate a back-mapped core moisture response feature value corresponding to the initial moisture content. The online core moisture response feature data of the current detection unit is retrieved, and the deviation between the back-mapped core moisture response feature value and the online core moisture response feature value is calculated.

[0053] A preset convergence threshold is set. If the deviation is less than the convergence threshold, the current iteration is determined to have converged, and the moisture content value at this time is taken as the converged moisture content value of the detection unit. Otherwise, the current iteration is determined not to have reached the convergence state, and an iterative correction operation is performed until the calculated deviation is less than the preset convergence threshold.

[0054] Based on the pre-calibrated weights of the spatial representativeness of each detection unit, the converged moisture content values ​​of each detection unit are spatially weighted and integrated to generate the real-time moisture content of the online moving material.

[0055] Specifically, the standard material spectral database refers to a structured database built based on the near-infrared spectral characteristics, moisture content calibration results, and test material transmission parameters of similar materials. It integrates the unique attributes of standard materials with the correspondence between multi-dimensional related spectral features and consists of standard material unique attribute data, standard multi-dimensional related spectral feature data, standard moisture content calibration data, and standard test material transmission parameter data.

[0056] A near-infrared moisture measuring device for online moving materials, used to implement the above-mentioned near-infrared moisture measuring method for online moving materials, includes: a target wavelength combination generation module, a signal acquisition and processing module, a feature extraction module, and a moisture calculation module;

[0057] The target wavelength combination generation module is used to generate a target wavelength combination that is suitable for online moving material moisture measurement based on the unique properties of the material.

[0058] The signal acquisition and processing module is used to dynamically calculate and output the emission trigger signal of the near-infrared light source based on the target wavelength combination and the real-time transmission status of the online moving material. It controls the near-infrared light source to emit near-infrared light according to the time-division emission sequence. In the dedicated acquisition window of each wavelength, it acquires reflected signal data based on the dynamic acquisition frequency. Through adaptive gain correction, it binds the coordinates of the detection area and generates a standardized set of reflected signal data.

[0059] The feature extraction module is used to extract multi-dimensional correlated spectral features based on a standardized set of reflection signal data;

[0060] The moisture calculation module is used to perform spatial consistency calibration by combining the coordinates of the detection area. It uses the standard benchmark features of spatial matching as a reference to construct a correlation index table between multi-dimensional correlated spectral features and the coordinates of the detection area, divides the detection units, generates the initial feature difference of each detection unit, and uses the difference iterative calibration method to calculate the moisture content of the material.

[0061] The beneficial effects of this invention are:

[0062] This invention generates a suitable target wavelength combination based on the unique properties of online mobile materials, such as matrix composition and morphological structure. It dynamically calculates the emission trigger signal and controls the time-division emission of the light source in conjunction with the real-time transmission status of the material. Reflected signals are acquired at a dynamic frequency in a dedicated acquisition window, and after adaptive gain correction and binding to the coordinates of the detection area, a standardized reflected signal data set is generated. From this set, multi-dimensional correlated spectral features of core moisture response, interference cancellation, and operating condition correlation are extracted. Then, spatial consistency calibration is performed using the coordinates of the detection area, and detection units are divided to generate iterative initial feature differences. Finally, the difference iterative calibration method is used to calculate the synergistic effect of core technical features such as material moisture content. This solves the technical problems of poor adaptability, low measurement accuracy under dynamic conditions, and large spatial difference interference in traditional online mobile material near-infrared moisture measurement. The customized target wavelength combination breaks the limitations of fixed wavelengths, achieving precise matching with the unique properties of the material. This invention enhances the adaptability to differences in material properties across different batches and reduces matrix component interference at the source. Dynamic signal acquisition and standardized processing ensure precise synchronization between light emission and signal acquisition and real-time material transmission conditions, guaranteeing the authenticity and integrity of reflected signals and improving adaptability to fluctuations in operating conditions. The extraction of multi-dimensional correlated spectral features effectively removes common interferences such as matrix components, light intensity fluctuations, and material transmission thickness changes. The calculation method combining spatial consistency calibration with differential iterative calibration not only compensates for the shortcomings of traditional technologies that ignore differences in the spatial distribution of materials but also accurately captures the overall moisture distribution of materials. Ultimately, this invention achieves accurate and stable measurement of moisture in online moving materials under dynamic operating conditions, significantly improving the adaptability of measurement technology to different batches of materials and real-time operating condition changes, as well as the reliability of measurement results. It fully meets the high-precision and high-reliability requirements for online moving material moisture detection in industrial production processes. Attached Figure Description

[0063] Figure 1 This is a flowchart of a near-infrared moisture measurement method for online moving materials according to the present invention;

[0064] Figure 2 A flowchart for generating a target wavelength combination adapted to the current online mobile material moisture measurement for this invention;

[0065] Figure 3 A flowchart for generating a standardized set of reflected signal data for this invention;

[0066] Figure 4 This is a flowchart for generating the initial feature difference values ​​for each detection unit in this invention. Detailed Implementation

[0067] Example 1

[0068] Please see Figure 1This embodiment describes a near-infrared moisture measurement method for online moving materials, including:

[0069] Step S1: Based on the unique properties of online moving materials, define the feasible wavelength range of near-infrared light used for moisture measurement, and filter the candidate wavelength pool through intersection operation. After dynamic adaptation optimization, generate the target wavelength combination that is suitable for the current online moving material moisture measurement.

[0070] In this embodiment, due to the dynamic changes in the transmission status of online moving materials and the differences in unique properties between batches of materials, traditional fixed wavelength combinations are difficult to balance measurement accuracy and working condition adaptability. They are easily affected by matrix interference or signal instability, leading to excessive measurement errors. Therefore, it is necessary to define the feasible wavelength domain by the unique properties of the material, and combine intersection screening and dynamic optimization to generate target wavelength combinations to achieve accurate matching between wavelength and material characteristics and measurement conditions. Among them, unique properties refer to the core set of inherent properties of online moving materials that directly affect the near-infrared light absorption and propagation characteristics and moisture measurement effect, including the material matrix composition, material morphology and structure, moisture content range, and moisture distribution state.

[0071] Please see Figure 2 Preferably, the specific steps for generating a target wavelength combination adapted to the current online mobile material moisture measurement include:

[0072] The unique properties of online moving materials are obtained, including the material matrix composition, material morphology and structure, moisture content range, and moisture distribution state. The material matrix composition refers to the components other than moisture in the online moving material, including starch, protein, resin, fiber, and additives. The material morphology and structure refer to the external physical form and internal structural characteristics of the online moving material, including external forms such as granules, powder, flakes, and lumps, and internal structural characteristics such as porosity, bulk density, surface smoothness, and internal density. The moisture content range refers to the possible moisture concentration range of the current batch of online moving material, including low moisture, medium moisture, and high moisture ranges. The moisture distribution state refers to the distribution characteristics of moisture inside and on the surface of the online moving material, including surface-dominated water distribution, internal water-dominated water distribution, and uniform distribution.

[0073] Based on a pre-defined wavelength adaptation rule library, the feasible near-infrared wavelength range corresponding to each unique attribute of the online moving material is matched. The pre-defined wavelength adaptation rule library is a pre-constructed set of structured rules storing the correspondence between different unique attributes and their corresponding near-infrared wavelength ranges. These rules are established based on near-infrared spectral propagation characteristics, the characteristic absorption law of hydroxyl groups in water molecules, and the influence mechanism of different material unique attributes on measurement. This is used to match each unique attribute of the online moving material with the attribute type in the rule library, invoke the corresponding wavelength constraints, and define the wavelength range that meets the measurement requirements of that attribute. For example, if the material matrix is ​​starch, then the wavelength adaptation rule library is matched with the starch... The feasible wavelength range for powdery material matrix is ​​a wavelength range that avoids the strong near-infrared absorption band of starch. If the material is granular, the feasible wavelength range for granular material morphology in the rule library is matched. This feasible range is the wavelength range for detecting internal moisture of granules with near-infrared light penetration depth. If the moisture content range is low, the feasible wavelength range for low moisture content in the rule library is matched. This feasible range is the high-sensitivity wavelength range corresponding to the strong absorption of water molecule hydroxyl groups. If the moisture distribution is dominated by surface water, the feasible wavelength range for surface water detection in the rule library is matched. This feasible range is the wavelength range for surface signal acquisition with shallow near-infrared light penetration depth.

[0074] The intersection of the feasible regions of near-infrared light wavelengths corresponding to each unique attribute obtained by matching is solved to obtain the wavelength range that simultaneously meets the constraints of all unique attributes of online moving materials, which is used as the candidate wavelength pool.

[0075] Obtain the wavelength interval length of the candidate wavelength pool. If the wavelength interval length of the candidate wavelength pool is not less than the preset effective length threshold, the candidate wavelength pool is determined to be valid and can be directly used for subsequent wavelength combination construction; otherwise, the candidate wavelength pool is determined to be invalid and the candidate wavelength pool correction process needs to be started.

[0076] The preset effective length threshold is determined based on the number of wavelengths required to construct a basic qualified wavelength combination. For example, it includes two core feature wavelengths and one reference wavelength, as well as the minimum wavelength interval required to avoid signal interference between adjacent wavelengths. This ensures that the threshold setting meets the basic requirements for combination construction while also guaranteeing the independence and recognizability of each wavelength signal.

[0077] The candidate wavelength pool correction process includes:

[0078] Based on the importance priority of unique attributes, the gradient is relaxed according to the preset constraint threshold. First, the wavelength feasible region constraint threshold corresponding to the secondary priority unique attributes is relaxed, and then the intersection solution is performed again on the near-infrared light wavelength feasible region corresponding to each unique attribute to obtain the preliminary corrected candidate wavelength pool. Then, it is determined whether the corrected candidate wavelength pool is effective. If not, the gradient is relaxed again according to the preset constraint threshold, and then the wavelength feasible region constraint threshold corresponding to the secondary priority unique attributes is relaxed again, until the candidate wavelength pool is effective.

[0079] Among them, the importance priority of unique attributes refers to the attribute hierarchy divided according to the degree of influence of each unique attribute on the near-infrared moisture measurement results. The material matrix composition and moisture content range are core priority attributes, and their corresponding wavelength feasible domain constraints directly determine the recognition degree and measurement sensitivity of moisture characteristic signals. Material morphology and structure, and moisture distribution state are secondary priority attributes, and the impact of their constraint adjustments on measurement accuracy is relatively low. The influence weight of unique attributes is set according to the near-infrared spectral propagation characteristics, the characteristic absorption law of water molecule hydroxyl groups, and experimental verification. The preset constraint threshold relaxation gradient refers to the upper limit of the amplitude when relaxing the wavelength feasible domain constraints corresponding to the secondary priority attributes each time. It is set through wavelength screening experiments of a large number of materials with different attributes and measurement accuracy verification results.

[0080] A signal-independent threshold is configured, and then the minimum interval and maximum number of selectable wavelengths are calculated based on the signal-independent threshold and the wavelength interval length of the candidate wavelength pool. Subsequently, the minimum interval is used as a hard constraint on the spacing between adjacent wavelengths, and the maximum number of combinations is used as an upper limit constraint on the number of wavelengths in a single combination. All wavelength points in the candidate wavelength pool are traversed, and the spacing and number of wavelengths are checked one by one. Only wavelengths with an adjacent wavelength interval not less than the minimum interval are retained, and the number of wavelengths in a single combination does not exceed the maximum number of combinations. Multiple initial wavelength combinations that meet the dual constraints are constructed by traversing all wavelengths. Then, for each initial wavelength combination, the distribution uniformity of the wavelengths in the combination is quantitatively calculated from the dimensions of wavelength spacing consistency and interval coverage in the candidate wavelength pool interval, and the wavelength distribution status of each initial wavelength combination is quantitatively characterized. Finally, the distribution uniformity is used as the core screening index, and the group with the highest distribution uniformity is selected from multiple initial wavelength combinations as the target wavelength combination for online mobile material moisture measurement. The higher the wavelength distribution uniformity, the better it can ensure the full-range response of the material moisture characteristics in the subsequent near-infrared spectral signal acquisition, effectively improving the acquisition stability of the spectral signal and the effectiveness of subsequent feature extraction.

[0081] Step S2: Based on the target wavelength combination and the real-time transmission status of the online moving material, execute the complete control process for near-infrared light emission and reflection signal data acquisition, specifically as follows:

[0082] The system collects real-time data on the transmission density and speed of moving materials online, dynamically generates and outputs emission trigger signals for near-infrared light sources, calls up near-infrared light sources that are one-to-one matched with each wavelength in the target wavelength combination, and controls each light source to start emitting sequentially according to a preset time-division emission sequence. The emission time interval between adjacent wavelength light sources is strictly set to a preset minimum threshold for anti-crosstalk. A linkage control mechanism is simultaneously constructed between the light source emission action and the near-infrared detector acquisition action. The signal acquisition frequency of the near-infrared detector is dynamically adjusted in combination with the real-time material transmission speed to ensure that the near-infrared detector's acquisition of material surface reflection signal data covers the complete detection area during the material movement process. Finally, the collected reflection signal data of different wavelengths are systematically integrated according to the emission sequence of each wavelength, and associated with the material transmission parameter information at the corresponding acquisition time to generate a standardized reflection signal data set for subsequent quantitative analysis of material moisture content.

[0083] Preferably, the specific steps for dynamically calculating and outputting the emission trigger signal of the near-infrared light source include:

[0084] The system acquires real-time online moving material transmission density and speed, and retrieves the length of the material detection unit. This material detection unit length parameter is quantitatively preset based on the near-infrared light irradiation cross-sectional size and moisture measurement resolution. The system quantitatively calculates the basic trigger cycle of the light source emission according to the ratio of the material detection unit length to the real-time transmission speed, ensuring that the material movement distance is exactly equal to the material detection unit length within the interval between two adjacent emission triggers, so that the material detection unit completely covers the near-infrared light irradiation area.

[0085] Among them, transmission density is the core quantitative parameter characterizing the density of distribution of online moving materials during transmission. It refers to the mass of online moving materials within a unit effective cross-sectional area or unit transmission volume. Its core is to describe the real-time accumulation or distribution state of materials in the transmission channel.

[0086] Based on the attenuation coefficient threshold of near-infrared light in materials with different transmission densities, the transmission density is classified into different levels to obtain the real-time transmission density classification interval of the online moving material, and the gain coefficient corresponding to the transmission density classification interval is matched.

[0087] The attenuation coefficient threshold is quantitatively configured through near-infrared light attenuation experiments on multiple sets of gradient transmission density material samples. The gain coefficient is a quantitative parameter used to adjust the intensity of the near-infrared light source driving signal. Its value has a preset linear or non-linear quantitative correlation with the output light power of the light source. It is used to ensure that the near-infrared light, after attenuation by the current density material, forms a light signal intensity that meets the signal acquisition threshold requirements of the near-infrared detector.

[0088] The number of wavelengths included in the target wavelength combination is counted, and a time slot allocation matrix for time-division transmission is constructed. The time slot allocation matrix is ​​used to record a structured data set of transmission timing parameters of each wavelength in the target wavelength combination, including a one-to-one mapping relationship between wavelength identifiers and corresponding transmission time slot intervals.

[0089] It should be noted that the row dimension of the time slot allocation matrix corresponds to the wavelength identifier in the target wavelength combination, and the column dimension corresponds to the order of wavelength emission timing. According to the control requirements of time-division multiplexing, continuous and non-overlapping time slot intervals are allocated to each wavelength in the time slot allocation matrix. The start time interval of time slot intervals corresponding to adjacent wavelengths is strictly controlled to ensure that this interval is equal to the preset minimum anti-crosstalk threshold. The minimum anti-crosstalk threshold refers to the minimum time interval that can avoid interference from the superposition of near-infrared light signals of different wavelengths in the time domain. It is determined through comprehensive experimental calibration of the near-infrared light source emission response time, the near-infrared detector signal acquisition delay, and the wavelength signal discrimination accuracy.

[0090] Based on the basic triggering period and time slot allocation matrix, the emission triggering time corresponding to each wavelength in a single triggering process of the target wavelength combination is quantitatively calculated, and a one-to-one correspondence between the triggering time and the wavelength identifier is established. Combined with the pre-calibrated attenuation coefficient parameter, a transmission triggering signal containing wavelength identifier field, triggering time field, gain coefficient field, and time slot interval field is constructed. The transmission triggering signal is transmitted to the near-infrared light source driver end through a preset bus interface. After receiving the transmission triggering signal, the driver end performs quantitative analysis of each field information in the data packet, extracts the emission timing and drive-related parameters corresponding to each wavelength, and forms a control parameter set for near-infrared light source emission.

[0091] The system collects the actual operating parameters of each wavelength light source in real time, including the emission start time, emission duration, and emission power, at a preset frequency. It then performs a field-by-field quantitative comparison between the actual operating parameters and the command parameters in the emission trigger signal, and calculates the deviation value based on the difference between the actual and command parameters. When the deviation value exceeds the preset allowable deviation range, it quantitatively corrects the basic trigger period and gain coefficient values ​​according to the magnitude of the deviation value, while simultaneously updating the time slot allocation matrix. Finally, it reintegrates the corrected parameters with the updated time slot allocation scheme to generate a corrected composite emission trigger signal and outputs it quantitatively, completing the dynamic closed-loop adjustment of the trigger signal.

[0092] Please see Figure 3 Preferably, the specific steps for generating a standardized set of reflected signal data include:

[0093] The system collects real-time data on the transmission location and speed fluctuations of moving materials online. It then quantitatively compares the transmission location data with a preset detection area boundary threshold to determine whether the material has entered the near-infrared light irradiation detection area. If so, it calculates the trigger compensation delay based on the transmission speed fluctuation and a preset compensation coefficient through multiplication. The compensation delay is then added to the theoretical trigger time corresponding to the basic trigger cycle to determine the actual output time of the trigger signal. At this time, the system outputs the trigger signal to the near-infrared light source driver to ensure that the near-infrared light irradiation action and the material's arrival status are accurately matched.

[0094] The compensation coefficient is a quantitative parameter used to correct the deviation of the launch triggering time caused by the fluctuation of material transmission speed. Its value is determined based on experimental data of near-infrared light irradiation area misalignment under different transmission speed fluctuation amplitudes, and is positively correlated with the degree of influence of speed fluctuation amplitude on the triggering time.

[0095] The system retrieves real-time data on material transport speed and density, along with a preset material detection unit length. It calculates the base acquisition frequency of the near-infrared detector based on the ratio of the material detection unit length to the transport speed. A sliding window is used to continuously statistically analyze the transport speed, obtaining the transport speed fluctuation amplitude within the sliding window. Corresponding acquisition frequency correction coefficients are extracted according to the graded intervals of the transport speed fluctuation amplitude. The base acquisition frequency is then quantitatively multiplied by the acquisition frequency correction coefficients to obtain the dynamic acquisition frequency of the near-infrared detector. The sliding window refers to a fixed-length data segment used for segmented statistical analysis of continuously acquired transport speed data; the sliding window length is preset according to the transport speed acquisition frequency and speed fluctuation response requirements.

[0096] Simultaneously, the time slot interval information in the transmission trigger signal is retrieved, and a dedicated acquisition window is configured for each wavelength to ensure that the start and end times of the acquisition window completely coincide with the start and end times of the corresponding wavelength's transmission time slot.

[0097] Within the dedicated acquisition window for each wavelength, based on the dynamic acquisition frequency, the near-infrared light signal reflected from the material surface is received by the near-infrared detector, the light signal is converted into an analog electrical signal, and then the analog electrical signal is converted into a digital signal by the analog-to-digital conversion unit to obtain the original reflected signal data corresponding to each wavelength.

[0098] The material transport density at the corresponding acquisition time is retrieved from the original reflection signal data of each wavelength. According to the preset association rule between transport density and noise type, the original reflection signal data is denoised. For example: if the transport density is in the high-density range, the characteristic noise is determined to be scattering noise. The db4 wavelet basis function is selected, and the wavelet threshold denoising is performed on the original reflection signal data using an adaptive threshold calculation method based on noise estimation. If the transport density is in the medium-low density range, the characteristic noise is determined to be environmental stray light noise and electromagnetic interference noise. The sym5 wavelet basis function is selected, and the wavelet threshold denoising is performed on the original reflection signal data using a threshold calculation method based on signal amplitude distribution.

[0099] To address the differences in reflected signal intensity caused by different transmission densities, a preset mapping table of transmission density and gain compensation coefficients is retrieved. The signal gain compensation coefficient corresponding to the current transmission density is extracted, and the intensity value of the denoised reflected signal data is multiplied by the gain compensation coefficient to complete the adaptive gain correction of the reflected signal data. The mapping table of transmission density and gain compensation coefficients is configured as follows: First, gradient density material samples covering the entire transmission density range of the measurement scenario are prepared. Using the same type of near-infrared light source and near-infrared detector as the online measurement system, each gradient sample is irradiated and reflected signal data is collected at a fixed optical power and a fixed irradiation distance. The transmission density and attenuation amplitude of the reflected signal data corresponding to each sample are recorded. The gain compensation coefficient corresponding to each transmission density is calculated according to the quantitative correspondence between the attenuation amplitude and the gain compensation amount. The transmission density interval is associated with the corresponding gain compensation coefficient one by one. After multiple rounds of repeated experiments to verify and remove abnormal data, the mapping table is solidified.

[0100] A polynomial fitting algorithm is used to perform baseline correction on the adaptive gain-corrected reflection signal data to eliminate the signal baseline offset caused by light source power fluctuations and near-infrared detector response drift. Then, the extreme value normalization method is used to map the intensity values ​​of the reflection signal data after wavelength correction to a uniform dimension range to eliminate the influence of intensity differences between signals of different wavelengths and obtain the preprocessed reflection signal data.

[0101] Retrieve the time slot allocation matrix and trigger time information from the transmission trigger signal, and sort the preprocessed reflected signal data by wavelength transmission time sequence; retrieve the standard time data of the system clock, and calibrate and align the acquisition timestamps of each wavelength reflected signal data with the standard time data;

[0102] Based on the real-time transmission speed within the acquisition period corresponding to each wavelength reflection signal data, the material movement distance within that acquisition period is calculated in conjunction with the acquisition period duration; the initial position parameters of the detection area are retrieved, and the initial position parameters are superimposed with the calculated movement distance to correct and obtain the material detection area coordinates corresponding to each wavelength reflection signal data segment.

[0103] Each wavelength reflection signal data segment is individually bound to its corresponding wavelength identifier, trigger time, and corrected detection area coordinates. Based on the binding results, a multi-dimensional parameter association table is constructed. The sorted wavelength reflection signal data segments and their corresponding material transmission parameters at the acquisition time are then entered into the association table, forming a set of reflection signal data with material transmission parameter identifiers. This enables traceable association between signal data and material transmission status. Material transmission parameters refer to the core quantitative parameter set characterizing the real-time motion and distribution status of online moving materials within the transmission channel, including transmission density and transmission speed.

[0104] Based on the preset standard data format specifications, the set of reflected signal data with material transfer parameter identifiers is subjected to format standardization processing to form a standardized set of reflected signal data: the number of sampling points, data storage structure and intensity value dimensions of each reflected signal data segment are unified, and the intensity values ​​of all reflected signal data are converted into light absorption dimensions.

[0105] The standardized set of reflected signal data is integrated and encapsulated with the material transmission parameters in the association table to form a standard data structure containing a file header, a data area, and a verification area: the file header stores wavelength combination information, a summary of material-specific attributes, and the data generation time; the data area stores sorted signal segments and corresponding material transmission parameters in the order of transmission time; and the verification area stores data integrity check codes to ensure the integrity of data transmission and storage.

[0106] Step S3: Based on the standardized reflectance signal data set, extract and integrate multi-dimensional correlated spectral features. Reduce signal interference caused by transmission density and velocity fluctuations through condition-adaptive feature compensation. Perform spatial consistency calibration of features in conjunction with the detection area coordinates. Employ the difference iterative calibration method, constructing a three-dimensional benchmark library based on the standard material spectral database, including detection area coordinates, standard benchmark correlated spectral features, and standard moisture content. Achieve precise spatial matching and difference calculation between online and standard core moisture response features. Use interference cancellation features to remove common interference and obtain the purified difference. Integrate condition constraints and spatial constraints to generate compensation. The method generates initial differences and derives initial moisture content through iteration. It then uses reverse mapping verification, deviation calculation, and iterative convergence determination to dynamically optimize the compensation amount until the converged moisture content value of each detection unit is obtained. Finally, it generates the real-time moisture content of the moving material online through spatial weighted integration. The core of this method is to solve the technical problems in traditional near-infrared moisture measurement, such as feature distortion caused by spatial location differences, susceptibility to interference in difference calculation, insufficient adaptability to working conditions, dilution of effective signals by interference features, and high cost of model fitting. This method ensures the adaptability and reliability of measurement results to differences in material characteristics, dynamic transmission conditions, and spatial distribution differences, thereby achieving accurate acquisition of the real-time moisture content of the material.

[0107] In this embodiment, step S3 overcomes the technical bottlenecks of traditional near-infrared moisture measurement, which relies on model fitting, ignores spatial differences, and has poor interference suppression. Its core design logic is to first divide the core feature wavelengths and reference wavelengths based on the near-infrared spectral absorption characteristic rule library, and then selectively extract and structurally integrate three types of multi-dimensional correlated spectral features: core moisture response, interference cancellation, and operating condition correlation. Then, it performs feature spatial consistency calibration in combination with the detection area coordinates to eliminate feature-position matching deviations caused by material transport offsets. Subsequently, it constructs a three-dimensional benchmark library based on the standard material spectral database to achieve accurate spatial matching between online and standard features. It removes common interferences by using the dynamic benchmark ratio of interference cancellation features and integrates the dual compensation amounts of operating conditions and space to generate the initial difference for iteration. Finally, guided by the deviation amount verified by reverse mapping, it completes the iterative convergence calculation by dynamically adjusting the compensation amount. The real-time moisture content is obtained through spatial weighted integration. The entire process does not require complex model training, which not only amplifies the role of core moisture features but also achieves comprehensive adaptation to differences in material characteristics, dynamic transport conditions, and spatial distribution differences, significantly improving the accuracy and reliability of online mobile material moisture measurement.

[0108] Preferably, the specific steps for extracting multi-dimensional correlated spectral features include:

[0109] Retrieve the absorbance time-series data, acquisition time period identifier, and associated material transport parameters for each wavelength from the standardized reflection signal data set;

[0110] Based on a near-infrared spectral absorption characteristic rule library, target wavelength combinations are divided into core characteristic wavelengths and reference wavelengths. The core characteristic wavelengths correspond to wavelengths that match the absorption peak of hydroxyl groups in water molecules and avoid strong absorption bands in the material matrix. Reference wavelengths correspond to wavelengths with no moisture absorption characteristics that only respond to matrix composition and light intensity fluctuations. By classifying and constructing the correlation between wavelength and absorbance time-series data, a data classification foundation is provided for subsequent targeted feature extraction. The near-infrared spectral absorption characteristic rule library is a structured set of rules integrating various wavelength spectral characteristic judgment rules, constructed based on near-infrared spectral propagation characteristics, the characteristic absorption law of hydroxyl groups in water molecules, and the spectral response mechanism of different material matrix components. Its core content covers wavelength judgment rules for the characteristic absorption range of hydroxyl groups in water molecules, wavelength exclusion rules for strong absorption bands of different material matrix components, and screening rules for wavelengths with no moisture absorption characteristics. It also associates the adaptation relationship between the unique properties of different materials and spectral characteristic rules. By calling the corresponding rules in the rule library, it is possible to accurately determine whether each wavelength in the target wavelength combination matches the moisture response requirements and whether it is affected by matrix interference, providing a unified and traceable technical basis for the division of core characteristic wavelengths and reference wavelengths.

[0111] For the absorbance time-series data corresponding to the core characteristic wavelengths, core moisture response features are extracted. These core moisture response features include the mean absorbance, the first-order difference of absorbance between adjacent wavelengths, and the difference in absorbance time-series extreme values. Specifically, by calculating the mean absorbance of a single core characteristic wavelength within the corresponding acquisition period, the moisture absorption intensity of the material in the corresponding band is characterized. By calculating the first-order difference of absorbance between adjacent core characteristic wavelengths, the absorbance gradient variation characteristics between different moisture absorption peak intervals are characterized. By calculating the extreme value difference of absorbance time-series data for a single core characteristic wavelength, the uniformity of moisture distribution of the material within the acquisition period is characterized. Through the synergistic extraction of multi-dimensional core features, a comprehensive capture of moisture absorption characteristics is achieved, avoiding the deficiency that a single feature cannot reflect the dynamic changes in moisture distribution.

[0112] Based on the absorbance time-series data of the core characteristic wavelength and the reference wavelength, the mean of the absorbance time-series data of the reference wavelength is selected as the dynamic interference benchmark value. By calculating the ratio of the mean absorbance of each core characteristic wavelength to this dynamic interference benchmark value, an interference cancellation feature is constructed to cancel the common interference caused by the material matrix composition, light intensity fluctuations, and material transport thickness changes, thereby improving the purity of the core moisture response feature. By utilizing the common response characteristics of the reference wavelength and the core characteristic wavelength to the light intensity fluctuations of the matrix composition and the material transport thickness changes, the precise cancellation of such common interference is achieved. This differs from the traditional fixed benchmark interference cancellation method and improves the dynamic adaptability of interference cancellation.

[0113] By combining the time-series data of wavelength absorbance of core characteristic wavelengths with associated material transport parameters, operating condition-related features are extracted. These features include the coupling feature between core wavelength absorbance and transport density, and the correlation feature between the first-order difference of core wavelength absorbance and transport speed. The coupling value between the mean absorbance of a single core characteristic wavelength and the transport density during the corresponding acquisition period is calculated as a density-related scattering interference characterization feature, thus reflecting the impact of scattering interference caused by transport density on the absorbance signal. Furthermore, the correlation value between the first-order difference of absorbance of a single core characteristic wavelength and the transport speed during the corresponding acquisition period is calculated as a speed-related acquisition offset interference characterization feature, reflecting the interference of acquisition window offset caused by transmission speed fluctuations on absorbance gradient changes. This achieves deep binding between absorbance features and material transport parameters, improving the adaptability of features to dynamic changes in operating conditions.

[0114] The three feature subsets—core moisture response features, interference cancellation features, and operating condition correlation features—are integrated. Duplicate features are removed through feature repeatability verification, and invalid features are removed through feature validity determination. The features are then structured and sorted in the order of core moisture response features, interference cancellation features, and operating condition correlation features to form a regular multi-dimensional correlation spectral feature set. This ensures that subsequent operating condition-adaptive feature compensation can be carried out in an orderly manner based on structured features, thereby improving the effectiveness and relevance of feature applications.

[0115] Preferably, the specific steps for calculating the moisture content of materials include:

[0116] Please see Figure 4 The system retrieves the detection area coordinates bound to the multi-dimensional correlated spectral feature set and standardized reflectance signal data, constructs an association index table between the multi-dimensional correlated spectral features and the detection area coordinates, and divides the detection units according to the spatial distribution of the detection area coordinates to ensure that the multi-dimensional correlated spectral feature data in each detection unit corresponds to a unique spatial location identifier. Spatial consistency preprocessing is performed on the core moisture response features, interference cancellation features, and operating condition correlated features within each detection unit, eliminating multi-dimensional correlated spectral feature data corresponding to abnormal spatial locations or overlapping coordinates. The system corrects the mismatch between the multi-dimensional correlated spectral features and the detection area coordinates caused by material transport offsets. A spatial alignment algorithm ensures the accurate correspondence between the multi-dimensional correlated spectral feature data and the actual spatial location of the material, laying the foundation for subsequent spatial matching difference calculations.

[0117] The system retrieves a pre-defined standard material spectral database and filters out standard multi-dimensional correlated spectral feature data of standard materials that match the unique properties of the currently moving online material. Based on the detection area coordinate division rules for the moving online material, the same detection unit division process is applied to the standard multi-dimensional correlated spectral feature data. The standard material spectral database is a structured database built upon the near-infrared spectral characteristics, moisture content calibration results, and test material transmission parameters of similar materials. It integrates the correspondence between the unique properties of standard materials and multi-dimensional correlated spectral features, and consists of standard material unique attribute data, standard multi-dimensional correlated spectral feature data, standard moisture content calibration data, and standard test material transmission parameter data.

[0118] The core moisture response features within each detection unit are extracted as spatial matching standard reference spectral features. These are then linked to the standard moisture content calibration data of the standard material data to construct a three-dimensional reference library consisting of detection area coordinates, standard reference reference spectral features, and standard moisture content. This ensures strict spatial matching between the standard reference reference spectral features and the online multi-dimensional reference spectral features, avoiding the difference calculation deviation caused by spatial misalignment between traditional reference reference spectral features and online multi-dimensional reference spectral features.

[0119] Based on a correlation index table of multi-dimensional correlated spectral features and detection area coordinates, the core moisture response characteristics of online moving materials and standard materials are matched one by one according to the detection unit, and the difference between the core moisture response characteristics of the two is calculated. The interference cancellation characteristics corresponding to each detection unit are retrieved. By establishing a correlation mapping between the dynamic interference benchmark ratio and common interference, splitting the components of multiple types of interference ratios and determining their influence ratios, and then quantitatively stripping interference by category and verifying the purification effect, the interference stripping processing is performed on the difference of core moisture response characteristics. The common interference caused by matrix composition, light intensity fluctuations, and material transport thickness changes is accurately removed, and the difference of core moisture response characteristics after purification is obtained. The interference purification correction amount of each detection unit is recorded simultaneously to provide data support for subsequent anomaly tracing.

[0120] The system retrieves the operational condition-related features bound to each detection unit of the online moving material, extracts density-related scattering interference information and velocity-related acquisition offset information, and calculates the quantitative difference between the current transmission conditions and the standard material testing conditions as the operational condition constraint compensation amount. Combined with the spatial attenuation coefficient corresponding to the detection area coordinates (pre-calibrated based on the propagation characteristics of near-infrared light at different spatial locations), the spatial constraint compensation amount is determined. The operational condition constraint compensation amount and the spatial constraint compensation amount are synergistically integrated to obtain the initial compensation amount for iteration. This initial compensation amount is then substituted into the difference in the core moisture response characteristics after purification to generate the initial feature difference values ​​for each detection unit, achieving dual adaptation of the difference calculation to operational condition fluctuations and spatial differences.

[0121] Based on the inherent correspondence between the standard moisture content calibration data of standard materials and the corresponding standard reference spectral characteristics, and combined with the correction effect of the iterative initial feature difference on the standard correlation, the initial moisture content of each detection unit of the online moving material is derived, ensuring that the initial moisture content is directly correlated with the standard calibration data and the real-time feature difference, thus guaranteeing the consistency of the calculation results with the benchmark.

[0122] The initial moisture content is back-mapped to the core moisture response feature space to generate a back-mapped core moisture response feature value corresponding to the initial moisture content. The online core moisture response feature data of the current detection unit is retrieved, and the deviation between the back-mapped core moisture response feature value and the online core moisture response feature value is calculated. This deviation directly reflects the degree of fit between the initial moisture content and the actual moisture state of the material.

[0123] A convergence threshold is preset. The calculated deviation is compared with the convergence threshold. If the deviation is less than the convergence threshold, the current iteration is considered to have converged. The moisture content value at this point is determined as the converged moisture content value of the detection unit. The difference in the core moisture response characteristics of the detection unit, the data of the entire process of iterative compensation adjustment, and the final value of the deviation are recorded simultaneously to complete the iterative calculation process of the detection unit. Otherwise, the current iteration is considered not to have reached the convergence state, and an iterative correction operation is performed. At this time, the real-time change of the working condition associated features of each detection unit is retrieved. Combined with the spatial deviation correction coefficient corresponding to the coordinates of the detection area, the value of the initial compensation amount of the iteration is dynamically adjusted to ensure that the adjusted compensation amount can adapt to the real-time working condition fluctuations and spatial position differences.

[0124] The adjusted compensation amount is substituted back into the core moisture response characteristic difference after purification to generate an updated iterative characteristic difference. Based on this updated difference, the moisture content is recalculated, and the reverse mapping feature verification, deviation calculation, and convergence determination steps are repeated. This iterative process continues until the calculated deviation is less than the preset convergence determination threshold. The iteration stops, and the moisture content calculated at this point is determined as the converged moisture content value for each detection unit.

[0125] Based on the pre-calibrated weights of the spatial representativeness of each detection unit, the converged moisture content values ​​of each detection unit are spatially weighted and integrated. The weight allocation is positively correlated with the representativeness of the material region corresponding to the detection unit, ensuring that the integration result can comprehensively reflect the overall moisture distribution of the current batch of online moving materials, and finally generate the real-time moisture content of the online moving materials.

[0126] Example 2

[0127] This embodiment introduces a near-infrared moisture measurement device for online moving materials, including a target wavelength combination generation module, a signal acquisition and processing module, a feature extraction module, and a moisture calculation module;

[0128] The target wavelength combination generation module is used to generate a target wavelength combination that is adapted to the moisture measurement of the online moving material based on the unique properties of the material.

[0129] In one embodiment, it includes:

[0130] Acquire unique attribute data of online moving materials, such as material matrix composition, material morphology and structure, moisture content range, and moisture distribution status, to ensure that the acquired attribute data fully covers the core dimensions that affect near-infrared light absorption and propagation characteristics and moisture measurement results.

[0131] The preset wavelength adaptation rule library is invoked to accurately match the unique attributes of the online moving materials with the attribute types in the rule library, and the corresponding wavelength constraint conditions are invoked to define the feasible domain of near-infrared light wavelengths for each attribute adaptation.

[0132] Perform an intersection operation on each feasible wavelength region to obtain a candidate wavelength pool. Determine the validity of the candidate wavelength pool by judging the relationship between the length of the candidate wavelength pool interval and the preset effective length threshold. Start the correction process for invalid candidate wavelength pools, that is, relax the wavelength constraints of secondary priority attributes in order of priority of specific attributes and preset constraint threshold relaxation gradient until the candidate wavelength pool is valid.

[0133] Configure an independent threshold for the signal, calculate the minimum interval and maximum number of wavelengths that can be selected based on the threshold, traverse the candidate wavelength pool to construct multiple initial wavelength combinations, and select the optimal combination as the target wavelength combination by calculating the uniformity of wavelength distribution within each combination.

[0134] The core of the target wavelength combination generation module lies in achieving precise matching between the target wavelength combination and the unique properties of the online moving material and the measurement conditions, providing a highly adaptable wavelength foundation for subsequent measurement processes and avoiding measurement error problems caused by insufficient adaptability of traditional fixed wavelength combinations.

[0135] The signal acquisition and processing module is used to acquire reflected signal data and generate a standardized set of reflected signal data.

[0136] In one embodiment, it includes:

[0137] The system collects material transmission parameters such as transmission density, transmission speed, and transmission position of online moving materials in real time, retrieves the preset material detection unit length, and calculates the basic triggering cycle of light source emission based on the ratio of material detection unit length to real-time transmission speed.

[0138] Based on the attenuation coefficient threshold of near-infrared light in materials with different transmission densities, the transmission density is classified, the corresponding gain coefficient is matched, and the time slot allocation matrix for time-division emission is constructed in combination with the number of wavelengths of the target wavelength combination. The emission trigger time of each wavelength is quantitatively calculated, and an emission trigger signal containing fields such as wavelength identifier, trigger time, and attenuation coefficient is generated and transmitted to the light source driver.

[0139] The actual operating parameters of the light source are collected in real time, compared with the command parameters to calculate the deviation value, and the basic triggering cycle, gain coefficient and time slot allocation matrix are dynamically corrected when they exceed the allowable range, so as to realize the closed-loop adjustment of the triggering signal.

[0140] A linkage mechanism is established between light source emission and near-infrared detector acquisition. The acquisition frequency is dynamically adjusted based on the transmission speed, and a dedicated acquisition window is configured for each wavelength to ensure precise synchronization between the acquisition window and the emission time slot.

[0141] The system receives reflected light signals and converts them into digital signals. It performs denoising processing according to the corresponding denoising strategy matched to the transmission density, and completes signal preprocessing through gain compensation, baseline correction, and extreme value normalization.

[0142] A multi-dimensional parameter association table is constructed by associating wavelength identifiers, detection area coordinates, and transmission parameters. The signal data is then processed to standardize its format and integrated into a standardized set of reflected signal data, including a file header, data area, and verification area. This ensures the integrity, consistency, and traceability of the signal data.

[0143] The feature extraction module is responsible for extracting multi-dimensional correlated spectral features.

[0144] In one embodiment, it includes:

[0145] Retrieve the absorbance time-series data, acquisition time period identifier, and associated material transport parameters for each wavelength from the standardized reflection signal data set;

[0146] By calling the near-infrared spectral absorption characteristic rule library, the core characteristic wavelengths and reference wavelengths in the target wavelength combination are divided, and the corresponding correlation between wavelength and absorbance time series data is constructed.

[0147] Based on the absorbance time series data of the core characteristic wavelengths, core moisture response features such as absorbance mean features, first-order difference features of absorbance between adjacent wavelengths, and absorbance time series extreme value difference features are extracted to comprehensively capture the moisture absorption intensity, gradient change and distribution uniformity features of the material.

[0148] Based on the absorbance time series data of the core feature wavelength and the reference wavelength, interference cancellation features are constructed to cancel common interferences such as matrix composition and light intensity fluctuations of the light source;

[0149] By combining core characteristic wavelength data with material transport parameters, we extract working condition-related features such as the coupling feature between core wavelength absorbance and transport density, and the correlation feature between the first-order difference of core wavelength absorbance and transport speed.

[0150] Repeatability checks and validity determinations are performed on the three types of feature subsets. After removing duplicate and invalid features, the features are sorted in a pre-defined order to generate a regular multi-dimensional correlated spectral feature set, providing a high-quality and targeted feature foundation for subsequent moisture content calculation.

[0151] The moisture calculation module is responsible for calculating the moisture content of materials.

[0152] In one embodiment, it includes:

[0153] The detection area coordinates bound in the multi-dimensional correlated spectral feature set and standardized reflectance signal data are retrieved, an associated index table is constructed and the detection units are divided according to the spatial distribution law, spatial consistency preprocessing is performed on the feature data in each detection unit, and the accurate correspondence between the feature data and the actual spatial position of the material is ensured by the spatial alignment algorithm.

[0154] The standard material spectral database is retrieved, and standard multidimensional associated spectral feature data that are consistent with the unique properties of online moving materials are selected. After dividing the detection units according to the same rules, standard benchmark associated spectral features are extracted and associated with standard moisture content calibration data to construct a three-dimensional benchmark library of detection area coordinates, standard benchmark associated spectral features, and standard moisture content.

[0155] The online and standard core moisture response characteristics of the detection unit are matched and the difference is calculated. The interference cancellation feature is called and the common interference is removed through step-by-step processing logic to obtain the purified difference.

[0156] The working condition associated features are retrieved to calculate the working condition constraint compensation amount. The spatial constraint compensation amount is determined by combining the preset spatial attenuation coefficient. The initial compensation amount of the iteration is obtained by integrating the results and the initial feature difference of the iteration is generated.

[0157] The initial moisture content is derived based on the standard correlation and the initial difference. The deviation is verified by reverse mapping and compared with the preset convergence judgment threshold to perform convergence judgment. If convergence is not achieved, the compensation amount is dynamically adjusted and the iterative process is repeated until the converged moisture content value of each detection unit is obtained.

[0158] Based on the spatial representativeness of the detection unit space, a spatial weighted integration is performed to generate real-time moisture content that can comprehensively reflect the overall moisture distribution of the current batch of online moving materials.

[0159] Working principle and its effects:

[0160] This invention is based on customizing and adapting wavelength combinations according to the unique properties of materials, and dynamically controlling the light emission and signal acquisition process in real time transmission status. Through multi-dimensional feature extraction and spatialized iterative calculation, it finally achieves accurate measurement of moisture content, taking into account the adaptability to material property differences, operating condition fluctuations and uneven spatial distribution throughout the process.

[0161] In the core implementation phase, the unique properties of the material matrix, such as composition and morphology, are first acquired, and the feasible wavelength domain is matched and a target wavelength combination is generated. This design breaks the limitations of traditional fixed wavelengths, enabling the wavelength combination to be precisely matched with the material characteristics, improving the identification of moisture characteristic signals from the source, and effectively reducing matrix component interference. Subsequently, the transmission trigger signal is dynamically calculated by combining real-time transmission density, speed, and other states, and a time-division transmission, dedicated acquisition window, and dynamic acquisition frequency control method is adopted. Simultaneously, signal standardization is completed through adaptive gain correction and spatial coordinate binding. This ensures precise synchronization between light emission and material arrival status, and improves the integrity and reliability of reflected signal data, laying a high-quality data foundation for subsequent analysis. Finally, three types of features—core moisture response, interference cancellation, and operating condition correlation—are extracted from the standardized data. After spatial consistency calibration and detection unit division, an iterative initial difference is generated based on the standard benchmark features. The compensation amount is dynamically optimized through the difference iterative calibration method, and finally, the real-time moisture content is obtained by weighted integration. This process not only effectively removes common interferences but also makes up for the shortcomings of traditional measurements that ignore spatial distribution differences, enabling the measurement results to truly reflect the overall moisture distribution of the material.

[0162] In summary, this invention achieves multiple key effects through the coordinated adaptation of each stage: First, the precise matching of wavelength combinations with the unique properties of materials enhances adaptability to different batches of materials; second, the dynamic control mechanism adapts to real-time fluctuations in transmission conditions, ensuring signal acquisition quality; and third, spatial iterative calculation weakens the influence of various interferences, accurately capturing differences in the spatial moisture distribution of materials. The end-to-end design effectively solves the problems of existing technologies in simultaneously addressing differences in material properties, dynamic operating conditions, and uneven spatial distribution, significantly improving the accuracy and stability of online mobile material moisture measurement, fully meeting the high-precision and high-reliability detection requirements in industrial production.

[0163] The above description is merely a preferred embodiment of the present invention. The scope of protection of the present invention is not limited to the above embodiments. All technical solutions falling within the scope of the present invention's concept are within the scope of protection of the present invention. It should be noted that for those skilled in the art, any improvements and modifications made without departing from the principles of the present invention should also be considered within the scope of protection of the present invention.

Claims

1. A near infrared moisture measurement method for on-line moving material, characterized in that, The method comprises the following steps: Based on the unique properties of the online moving material, a target wavelength combination suitable for the online moving material moisture measurement is generated; Based on the target wavelength combination and the real-time transmission state of the online moving material, the emission trigger signal of the near-infrared light source is dynamically calculated and output, the near-infrared light source is controlled to emit near-infrared light according to the time-sharing emission timing, in the exclusive collection window of each wavelength, the original reflection signal data is collected based on the dynamic collection frequency; and through adaptive gain correction, the detection area coordinates are bound to generate a standardized reflection signal data set; Based on the standardized reflection signal data set, multi-dimensional correlation spectral features are extracted; Combined with the spatial consistency calibration of the detection area coordinates, the correlation index table of the multi-dimensional correlation spectral features and the detection area coordinates is constructed by taking the spatially matched standard reference features as the reference, the detection units are divided, and then the iterative initial feature difference of each detection unit is generated, and the difference value iterative calibration method is used to calculate the moisture content of the material.

2. A near infrared moisture measurement method for on-line movement of material as claimed in claim 1, wherein, The step of generating a target wavelength combination suitable for the online moving material moisture measurement comprises: Obtaining the unique properties of the online moving material, including the material matrix composition, the material morphology and structure, the moisture content range, and the moisture distribution state; According to the preset wavelength adaptation rule library, the near-infrared light wavelength feasible region corresponding to each unique property of the online moving material is matched; The intersection of the near-infrared light wavelength feasible region corresponding to each unique property obtained by matching is solved to obtain a wavelength interval that simultaneously satisfies the constraint requirements of all unique properties of the online moving material, which is used as a candidate wavelength pool; The wavelength interval length of the candidate wavelength pool is obtained, if the wavelength interval length of the candidate wavelength pool is not less than the preset effective length threshold, it is determined that the candidate wavelength pool is effective; otherwise, it is determined that the candidate wavelength pool is invalid; Based on the preset signal independence threshold and the wavelength interval length of the candidate wavelength pool, the minimum interval and the maximum combinable number of the selectable wavelengths are calculated; All wavelengths in the candidate wavelength pool are traversed, and the wavelengths with a wavelength interval not less than the minimum interval are selected to construct multiple initial wavelength combinations; wherein the upper limit of the number of wavelengths in the multiple initial wavelength combinations is the maximum combinable number; The distribution uniformity of the wavelengths in each initial wavelength combination in the candidate wavelength pool interval is calculated, and the target wavelength combination is selected from the multiple initial wavelength combinations.

3. A near infrared moisture measurement method for on-line movement of material as claimed in claim 1 wherein, The step of dynamically calculating and outputting the emission trigger signal of the near-infrared light source comprises: The real-time collected online moving material transmission density and transmission speed data are obtained, and the material detection unit length is called, the basic trigger period of the light source emission is quantitatively calculated according to the ratio of the material detection unit length to the real-time transmission speed; Based on the attenuation coefficient threshold of near-infrared light in different transmission density materials, the transmission density is classified and divided, the real-time transmission density classification interval of the online moving material is obtained, and the gain coefficient corresponding to the transmission density classification interval is matched; The number of wavelengths contained in the target wavelength combination is counted, and a time-sharing emission time slot allocation matrix is constructed; the time slot allocation matrix is used to record the structured data set of the emission timing parameter of each wavelength in the target wavelength combination, and contains a one-to-one mapping relationship between the wavelength identifier and the corresponding emission time slot interval; Based on the basic trigger cycle and the time slot allocation matrix, the emission trigger time of each wavelength in a single trigger process of the target wavelength combination is quantitatively calculated, and an emission trigger signal containing a wavelength identification field, a trigger time field, a gain coefficient field and a time slot interval field is constructed.

4. A near infrared moisture measurement method for on-line movement of material as claimed in claim 3 wherein, The step of collecting original reflection signal data in the dedicated acquisition window of each wavelength based on the dynamic acquisition frequency comprises: Real-time acquisition of transmission position data and transmission speed fluctuation of the online moving material, determination of whether the material enters the near-infrared light irradiation detection area, calculation of a trigger compensation delay based on the transmission speed fluctuation and a preset compensation coefficient if the material enters the near-infrared light irradiation detection area, and determination of the actual output time of the emission trigger signal; The basic acquisition frequency of the near-infrared detector is calculated according to the ratio of the length of the material detection unit to the transmission speed, and the transmission speed is continuously counted by using a sliding window to obtain the transmission speed fluctuation amplitude in the sliding window; the acquisition frequency correction coefficient corresponding to the transmission speed fluctuation amplitude is matched according to the hierarchical interval of the transmission speed fluctuation amplitude, the basic acquisition frequency is multiplied by the acquisition frequency correction coefficient to obtain the dynamic acquisition frequency of the near-infrared detector; The time slot interval information in the emission trigger signal is called to configure a dedicated acquisition window for each wavelength; In the dedicated acquisition window of each wavelength, the near-infrared light signal reflected by the material surface is received by the near-infrared detector based on the dynamic acquisition frequency to obtain the original reflection signal data corresponding to each wavelength.

5. A near infrared moisture measurement method for on-line movement of material as claimed in claim 4 wherein, The step of generating a standardized reflection signal data set comprises: The material transmission density at the acquisition time of the original reflection signal data of each wavelength is called, and a denoising operation is performed on the original reflection signal data according to a preset association rule of the transmission density and the noise type; A preset transmission density and gain compensation coefficient mapping table is called to extract the signal gain compensation coefficient corresponding to the current transmission density, the intensity value of the denoised reflection signal data is multiplied by the gain compensation coefficient to complete the adaptive gain correction of the reflection signal data, and the reflection signal data after adaptive gain correction is subjected to baseline correction and extreme value normalization to obtain preprocessed reflection signal data; The time slot allocation matrix and trigger time information in the emission trigger signal are called to sort the preprocessed reflection signal data with the wavelength emission time sequence as the index; and the acquisition time stamps of the reflection signal data of each wavelength are calibrated and aligned; Based on the real-time transmission speed in the acquisition period corresponding to each wavelength reflection signal data, the moving distance of the material in the acquisition period is calculated in combination with the acquisition period length to correct the material detection area coordinates corresponding to the reflection signal data segments of each wavelength; Each wavelength reflection signal data segment is bound with the corresponding wavelength identification, trigger time and corrected detection area coordinates to construct a multi-dimensional parameter association table; The sorted reflection signal data segments of each wavelength and the material transmission parameters at the corresponding acquisition time are sequentially entered into the multi-dimensional parameter association table to form a reflection signal data set with material transmission parameter identification, and a format standardization process is performed to form a standardized reflection signal data set.

6. A near infrared moisture measurement method for on-line movement of material as claimed in claim 1 wherein, The multi-dimensional associated spectral features include: core moisture response features, interference cancellation features and working condition association features. The step of extracting the multi-dimensional correlation spectral feature includes: Call the absorbance time series data corresponding to each wavelength in the standardized reflectance signal data set, the acquisition time period identifier and the associated material conveying parameters; Divide the core characteristic wavelength and the reference wavelength of the target wavelength combination based on the near-infrared spectral absorption characteristic rule library; For the absorbance time series data corresponding to the core characteristic wavelength, perform core moisture response feature extraction, and the core moisture response feature includes an absorbance mean value feature, an adjacent wavelength absorbance first-order difference feature, and an absorbance time series extreme value difference feature; Based on the absorbance time series data of the core characteristic wavelength and the reference wavelength, select the mean value of the reference wavelength absorbance time series data as a dynamic interference reference value, calculate the ratio of the absorbance mean value of each core characteristic wavelength to the dynamic interference reference value, and construct an interference cancellation feature; Combine the wavelength absorbance time series data of the core characteristic wavelength and the associated material conveying parameters to extract a working condition correlation feature, and the working condition correlation feature includes a core wavelength absorbance and conveying density coupling feature and a core wavelength absorbance first-order difference and conveying speed correlation feature; Integrate and structure the three types of feature subsets, i.e., the core moisture response feature, the interference cancellation feature and the working condition correlation feature, and perform sorting to form a multi-dimensional correlation spectral feature.

7. A near infrared moisture measurement method for on-line movement of material as claimed in claim 6 wherein, The step of generating the iterative initial feature difference value of each detection unit includes: Call the detection area coordinates bound in the multi-dimensional correlation spectral feature and the standardized reflectance signal data, construct an association index table of the multi-dimensional correlation spectral feature and the detection area coordinates, divide the detection units according to the spatial distribution law of the detection area coordinates, and perform spatial consistency preprocessing on the multi-dimensional correlation spectral feature in each detection unit; Call the preset standard material spectral database, select the standard multi-dimensional correlation spectral feature data of the standard material consistent with the unique properties of the current online mobile material, and perform detection unit division on the standard material data; Extract the core moisture response feature in each detection unit as a spatial matching standard reference correlation spectral feature, and associate the standard moisture content calibration data of the standard material data; Based on the association index table of the multi-dimensional correlation spectral feature and the detection area coordinates, match the core moisture response features of the online mobile material and the standard material one by one according to the detection unit, and calculate the core moisture response feature difference value of the two; Call the interference cancellation feature corresponding to each detection unit of the online mobile material, perform interference stripping processing on the core moisture response feature difference value, and obtain the purified core moisture response feature difference value; Call the working condition correlation feature bound in each detection unit, calculate the difference quantization value of the current conveying working condition and the standard material test working condition as a working condition constraint compensation amount, combine the spatial attenuation coefficient corresponding to the detection area coordinates, obtain the iterative initial compensation amount, and substitute it into the purified core moisture response feature difference value to generate the iterative initial feature difference value of each detection unit.

8. A near infrared moisture measurement method for on-line movement of material as claimed in claim 7 wherein, The step of calculating the moisture content of the material by using the difference value iterative calibration method includes: According to the standard moisture content calibration data of the standard material and the inherent correspondence relationship of the corresponding standard reference correlation spectrum characteristics, combined with the correction effect of the initial feature difference on the standard correlation, the initial moisture content of each detection unit of the online moving material is derived; The initial moisture content is reversely mapped to the core moisture response feature space to generate a reverse mapping core moisture response feature value corresponding to the initial moisture content, the online core moisture response feature data of the current detection unit is called, and the deviation of the reverse mapping core moisture response feature value and the online core moisture response feature value is calculated; A preset convergence judgment threshold is set, if the deviation is less than the convergence judgment threshold, it is judged that the current iteration has converged, and the moisture content value at this time is taken as the converged moisture content value of the detection unit, otherwise, it is judged that the current iteration has not reached the convergence state, and the iteration correction operation is performed; until the calculated deviation is less than the preset convergence judgment threshold; Based on the spatial representative pre-calibration weight of each detection unit, the spatial weighted integration calculation is performed on the converged moisture content value of each detection unit to generate the real-time moisture content of the online moving material.

9. A near infrared moisture measurement method for on-line movement of material as claimed in claim 7 wherein, The standard material spectrum database is constructed based on the near-infrared spectrum characteristics, moisture content calibration results and test material transmission parameters of the same type of material, is a structured database integrating the corresponding relationship of the unique attributes of the standard material and the multi-dimensional correlation spectrum characteristics, and is composed of standard material unique attribute data, standard multi-dimensional correlation spectrum characteristic data, standard moisture content calibration data and standard test material transmission parameter data.

10. A near infrared moisture measuring device for on-line moving material for carrying out a near infrared moisture measuring method for on-line moving material according to any one of claims 1 to 9, characterized in that It includes a target wavelength combination generation module, a signal acquisition and processing module, a feature extraction module and a moisture content calculation module. The target wavelength combination generation module is used to generate a target wavelength combination suitable for the moisture measurement of the online moving material based on the unique attributes of the online moving material; The signal acquisition and processing module is used to dynamically calculate and output the emission trigger signal of the near-infrared light source based on the target wavelength combination and the real-time transmission state of the online moving material, control the near-infrared light source to emit near-infrared light according to the time-sharing emission timing, and acquire reflection signal data based on the dynamic acquisition frequency in the exclusive acquisition window of each wavelength; And through adaptive gain correction, the detection area coordinates are bound to generate a standardized reflection signal data set; The feature extraction module is used to extract multi-dimensional correlation spectrum characteristics based on the standardized reflection signal data set; The moisture content calculation module is used to perform spatial consistency calibration combined with the detection area coordinates, take the spatially matched standard reference features as a reference, construct the correlation index table of the multi-dimensional correlation spectrum characteristics and the detection area coordinates, divide the detection units, and then generate the initial feature difference of each detection unit, and calculate the moisture content of the material by using the difference value iteration calibration method.

Images