An automated bird's nest component detection system and method

The automated bird's nest component detection system enables the coordinated processing of sample introduction, homogenization, quantitative splitting, and compaction. Combined with multi-dimensional state data acquisition and calibration models, it solves the problems of insufficient sample representativeness and poor consistency in traditional bird's nest testing, thereby improving the accuracy and reliability of the detection.

CN123112349APending Publication Date: 2026-11-17ANKANG GREEN POINT AGRI TECH CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202611413279.5
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-09-11
Publication Date
2026-11-17

AI Technical Summary

Technical Problem

Traditional bird's nest testing methods rely on manual sampling and single testing, resulting in insufficient sample representativeness, large deviations in single sampling, and a lack of linkage control over state variables such as temperature, thickness, pressure, and impedance, leading to unstable testing conditions and poor consistency of batch results.

Method used

An automated bird's nest component detection system is adopted, which integrates sample injection, homogenization, quantitative splitting and compaction through the linkage of mechanical actuators, sensing devices and controllers. It simultaneously collects multi-dimensional state data, uses near-infrared spectroscopy and electrode impedance detection, and combines a calibration model to improve detection accuracy and consistency.

Benefits of technology

It improves the overall representativeness of samples and the consistency of batch results, reduces single sampling deviation, enhances the reliability and accuracy of detection, and can promptly identify inhomogeneities within a batch.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN123112349A_ABST
    Figure CN123112349A_ABST
Patent Text Reader

Abstract

This invention relates to the field of food component detection technology, specifically to an automated bird's nest component detection system and method. The system includes interconnected mechanical actuators, sensing devices, and a controller. The mechanical actuators sequentially include a sample injection module, a homogenization module, a quantitative flow splitting module, and a compaction detection module. The sensing devices are used to collect data on mass, temperature, near-infrared reflectance spectra, multi-frequency impedance, platen displacement, and detection layer pressure. The controller determines whether the sample meets the measurement conditions based on impedance, thickness, and pressure. It inputs the near-infrared reflectance spectrum into a spectral component analysis model to obtain initial component detection results. Combining these with mass, thickness, temperature, and impedance data, a calibration model is used to obtain corrected component detection results for each sample. After statistical processing of multiple samples, batch detection results are output. This invention provides more morphologically consistent and data-comparable sample samples for subsequent component detection without destructive pretreatment.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention relates to the field of food component detection technology, specifically to an automated bird's nest component detection system and method. Background Technology

[0002] With the increasing demand for bird's nest product testing, the rapid determination of bird's nest component content, batch consistency, and water content has gradually become an important issue in the production and distribution process. In particular, dried bird's nest and soaked bird's nest exhibit heterogeneous characteristics in terms of morphology, fiber structure, and surface free water. If traditional manual sampling, manual pressing, and single testing methods are still used, it is impossible to simultaneously meet the requirements of testing efficiency and result stability. Traditional methods for testing bird's nests mainly rely on manual sampling, weighing and dividing the samples, near-infrared spectroscopy, or physicochemical analysis. However, these methods all have certain drawbacks. For example, manual sampling is easily influenced by experience, resulting in insufficient sample representativeness. A single sampling is difficult to reflect the true state of the entire batch of samples, and local deviations are likely to occur. Soaked samples are also susceptible to insufficient leaching and dehydration, inconsistent compaction, and interference from free water, which in turn affect the spectral results and composition determination. At the same time, conventional testing often lacks coordinated control of state variables such as temperature, thickness, pressure, and impedance, resulting in unstable testing conditions and poor consistency of batch results. Summary of the Invention

[0003] The purpose of this invention is to provide an automated bird's nest component detection system and method. Traditional bird's nest detection technologies, when faced with differences in bird's nest morphology, fiber structure, and surface free water, often rely on manual sampling and a single detection method. This results in insufficient sample representativeness, large deviations in single sampling, and significant shortcomings such as unstable detection conditions and poor batch-to-batch consistency due to the lack of coordinated control over state variables such as temperature, thickness, pressure, and impedance. Therefore, an automated bird's nest component detection system and method is needed that can integrate the sample introduction, homogenization, splitting, and compaction processes, and simultaneously collect multi-dimensional state data for feedback judgment and model correction, thereby effectively improving the overall representativeness of the sample and the accuracy of component determination. To solve the above technical problems, the technical solution of this invention includes: An automated bird's nest component detection system includes: Mechanical actuators, sensing and detection devices, and controllers that are interconnected and communicate with each other; The mechanical actuator includes a sample injection module, a homogenization module, a quantitative flow splitting module, and a compaction detection module connected in sequence, wherein the compaction detection module includes a detection box and a movable pressure plate; The sensing and detection device includes a weighing sensor, a temperature sensor, a near-infrared spectroscopy detection module, an electrode impedance detection module, a displacement sensor, and a pressure sensor, which are used to collect the mass of the bird's nest sample, the injection temperature and the sample temperature during the homogenization process, the near-infrared reflectance spectrum, the impedance at multiple frequencies, the displacement of the movable pressure plate, and thereby determine the thickness of the detection layer and the pressure on the detection layer. The controller is used to determine whether the detection samples obtained by the quantitative shunting module meet the measurement conditions based on the impedance, thickness and pressure. It inputs the near-infrared reflectance spectrum into a pre-established spectral composition analysis model to obtain the initial composition detection results. It inputs the mass, thickness, temperature and impedance into a correction model to correct the initial composition detection results to obtain the corrected composition detection results of each detection sample. It performs statistical processing on the corrected composition detection results of multiple detection samples and outputs the batch detection results.

[0004] Preferably, the homogenization module includes a rotating drum and a flexible feeding plate. A material hopper is formed inside the rotating drum. The flexible feeding plate is disposed inside the rotating drum and rotates with the rotating drum. It is made of silicone or polytetrafluoroethylene and maintains a 2-5 mm gap with the bottom of the material hopper of the rotating drum. The rotation speed of the rotating drum is 20-80 r / min, and the running time is 15-60 s. When the increase in sample temperature relative to the injection temperature is less than or equal to a preset temperature threshold, the homogenization module continues to operate. The preset temperature threshold is 3℃-5℃.

[0005] Preferably, the quantitative diversion module is provided with multiple diversion channels, each of which is equipped with a mechanical baffle. The sampling frequency of the weighing sensor of the sensing and detection device is 100-500Hz. The target mass of a single test sample of dried bird's nest is 2-5g, and the target mass of a single test sample of soaked bird's nest is 5-15g. When the mass of the test sample is 95%-105% of the target mass, the mechanical baffle of the corresponding diversion channel is closed, and 3-7 test samples are formed in each batch.

[0006] Preferably, the descent speed of the movable pressure plate is 2-10 mm / s, the target thickness of the detection layer is 3-10 mm, and the upper limit of the pressure on the detection layer is 5-30 kPa; the probe of the near-infrared spectroscopy detection module is set below the detection box and collects the near-infrared diffuse reflectance spectrum of the detection layer through the quartz window at the bottom of the detection box; the four electrodes of the electrode impedance detection module are embedded in the side wall of the detection box and are located at the same height position as the probe inside the detection box corresponding to the detection layer.

[0007] Preferably, the electrode impedance detection module is used to acquire resistance and reactance components at frequencies of 1 kHz, 10 kHz, and 100 kHz, and calculate the frequency according to the following formula. impedance amplitude : in, For frequency The resistive component below, For frequency The reactance component below.

[0008] The present invention also provides an automated method for detecting bird's nest components, which is based on the above-mentioned automated bird's nest component detection system. The method includes: collecting the sample injection temperature, homogenizing the bird's nest sample by dispersing it through a homogenization module, collecting the sample temperature during the homogenization process and determining the sample temperature rise based on the difference between the injection temperature and the sample temperature, and continuing homogenization when the sample temperature rise is less than or equal to a preset temperature rise threshold. The homogenized bird's nest sample is divided into multiple test sub-samples based on the mass collected by the weighing sensor using the quantitative flow separation module. The soaked bird's nest is filtered and dehydrated using a food-grade mesh plate in the quantitative flow separation module. The test sub-sample is then fed into the test box, and the movable pressure plate is lowered. The thickness is determined based on the displacement detected by the displacement sensor, and the movable pressure plate is controlled based on the pressure detected by the pressure sensor to press the test sub-sample into a test layer. When the thickness of the detection layer reaches the target thickness and the pressure does not exceed the upper pressure limit, near-infrared reflectance spectrum and impedance at multiple frequencies are collected simultaneously; the impedance, thickness and pressure are used to determine whether the test sample meets the measurement conditions; the near-infrared reflectance spectrum of the test sample that meets the measurement conditions is input into the pre-established spectral composition analysis model to obtain the initial composition detection results, and the mass, temperature, thickness and impedance obtained by the weighing sensor and temperature sensor are input into the correction model to obtain the composition content of each test sample; Based on the component content of multiple test samples within the same batch and the statistical distribution of impedance amplitude of multiple test samples at the multiple frequencies, the batch test results or the prompt information of non-uniformity within the batch are output.

[0009] Preferably, the step of determining whether the test sample meets the measurement conditions based on impedance, thickness, and pressure includes: determining the thickness of the test layer. Does it meet the requirements? ,pressure Does it meet the requirements? ,in This is the preset lower limit of the detection layer thickness. This is the preset upper limit of the detection layer thickness. The preset lower limit of pressure, The preset pressure limit; and whether the impedance amplitude at 10kHz and 100kHz is satisfied: in, for Impedance amplitude at frequency, for Impedance amplitude at a given frequency; when the above conditions are simultaneously met, the sample being tested is deemed to meet the measurement conditions; among which, According to The average of the impedance amplitude difference ratios of standard samples of the same grade and moisture content and standard deviation According to the formula A defined threshold.

[0010] Preferably, the near-infrared reflectance spectrum is sequentially subjected to dark current correction, white board reference correction, median filtering, smoothing, and standard normal variable transformation to obtain the first... Corrected spectral values ​​at each wavelength point And using multiple linear regression or partial least squares regression, the first step is to calculate the second step using the following formula. Content of each component : in, To test the quality of the subsamples, To detect layer thickness, The temperature of the sample being detected during spectral acquisition. , , , , , , These are the regression coefficients obtained through calibration.

[0011] Preferably, when the pressure reaches the upper limit but the thickness of the detection layer does not reach the target thickness, the movable pressure plate is released, causing the detection box to vibrate back and forth in the horizontal direction, and the descent speed of the movable pressure plate is reduced before re-compacting; after re-compacting, the thickness, pressure and impedance are detected again. If the measurement conditions are still not met after two consecutive re-compactings, the detection sample is isolated. For soaked bird's nest, before compaction, it is filtered and dehydrated through the food-grade mesh plate. When the absolute value of the mass change rate calculated based on the continuous sampling value of the weighing sensor within 2 seconds is greater than or equal to the preset mass change rate threshold, it waits. When the impedance amplitude at 1kHz is lower than the preset impedance lower limit, the sample is automatically flipped and replaced at the preset flipping speed before detection or a manual verification prompt is output. The preset impedance lower limit is the applicable lower limit of the model training sample.

[0012] Preferably, the test sample that meets the measurement conditions and obtains the component content is defined as a valid test sample, and the median and interquartile range of the component content of each valid test sample in the same batch are calculated respectively; When there are at least 3 valid detection samples and their proportion of the total number of detection samples obtained by shunting is not less than 80%, the coefficient of variation determined by the ratio of interquartile range to median does not exceed the preset upper limit of coefficient of variation, and the statistical distribution of the impedance amplitude of each valid detection sample at 10kHz is determined by the preset peak number determination algorithm not to show a double peak, the median is output as the batch detection result; if the above conditions are not met, a prompt message indicating unevenness within the batch is output.

[0013] Compared with the prior art, the present invention has the following beneficial effects: 1. By integrating sample injection, homogenization, quantitative splitting, and compaction testing, and simultaneously collecting mass, temperature, thickness, pressure, near-infrared reflectance spectrum, and multi-frequency impedance data at each stage, the problems of insufficient representativeness of traditional manual sampling, high statistical dispersion of single sampling deviation, and unstable detection conditions can be transformed into statistical judgment of multiple test samples, thereby improving the overall representativeness of the sample and the consistency of batch results. 2. By employing a low-shear homogenization method and constraining the process with a temperature rise threshold, the sample state can be deagglomerated while avoiding excessive heating and shearing, thus balancing homogenization and preserving the original state of the sample. Furthermore, by using a quantitative diversion method based on continuous weighing feedback, and by setting up a leaching dehydration structure and stable quality judgment for soaked bird's nest, the interference of surface free water on the sample quality can be reduced, resulting in more consistent sample quantities for each test subsample. 3. By collecting spectral and electrical data only after the thickness, pressure, and multi-frequency impedance of the detection layer meet the measurement conditions, and by incorporating thickness, mass, temperature, and impedance into the calibration model for correction, the influence of changes in compaction, contact, and moisture content on the spectral results can be effectively reduced, thus improving the accuracy of component detection. By statistically outputting multiple test samples using the median and interquartile range, and combining this with impedance distribution to determine whether there is inhomogeneity within the batch, the robustness of batch determination can be further improved. Furthermore, timely alerts can be provided for samples that are not fully expanded, have localized clumping, or exhibit significant internal state differences, thereby enhancing the reliability of automated detection. Attached Figure Description

[0014] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. The drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0015] Figure 1 This is a block diagram of the module structure of an automated bird's nest component detection system provided in an embodiment of the present invention.

[0016] Figure 2 This is a flowchart illustrating an automated method for detecting bird's nest components provided in an embodiment of the present invention. Detailed Implementation

[0017] To make the objectives, technical solutions, and advantages of the present invention clearer, the technical solutions of the present invention will be described in detail below with reference to the accompanying drawings and specific embodiments.

[0018] Please see the appendix Figure 1 An automated bird's nest component detection system includes: Mechanical actuators, sensing and detection devices, and controllers that are interconnected and communicate with each other; The mechanical actuator includes a sample injection module, a homogenization module, a quantitative splitting module, and a compaction detection module connected in sequence, wherein the compaction detection module includes a detection box and a movable pressure plate; The sensing and detection device includes a weighing sensor, a temperature sensor, a near-infrared spectroscopy detection module, an electrode impedance detection module, a displacement sensor, and a pressure sensor, which are used to collect the mass of the bird's nest sample, the sample injection temperature and the sample temperature during the homogenization process, the near-infrared reflectance spectrum, the impedance at multiple frequencies, the displacement of the movable pressure plate, and thereby determine the thickness of the detection layer and the pressure on the detection layer. The controller is used to determine whether the test samples obtained by the quantitative shunting module meet the measurement conditions based on impedance, thickness and pressure. It inputs the near-infrared reflectance spectrum into the pre-established spectral composition analysis model to obtain the initial composition detection results. It inputs the mass, thickness, temperature and impedance into the correction model to correct the initial composition detection results to obtain the corrected composition detection results of each test sample. It performs statistical processing on the corrected composition detection results of multiple test samples and outputs the batch detection results. In this embodiment, the homogenization module includes a rotating drum and a flexible feeding plate. A material hopper is formed inside the rotating drum. The flexible feeding plate is disposed inside the rotating drum and rotates with the drum. It is made of silicone or polytetrafluoroethylene and maintains a 2-5mm gap with the bottom of the material hopper of the rotating drum. The rotation speed of the rotating drum is 20-80 r / min, and the running time is 15-60s. When the increase in sample temperature relative to the injection temperature is less than or equal to a preset temperature threshold, the homogenization module continues to run. The preset temperature threshold is 3℃-5℃. In this embodiment, the quantitative diversion module is equipped with multiple diversion channels, each of which is equipped with a mechanical baffle. The sampling frequency of the weighing sensor of the sensing and detection device is 100-500Hz. The target mass of a single test sample of dried bird's nest is 2-5g, and the target mass of a single test sample of soaked bird's nest is 5-15g. When the mass of the test sample is 95%-105% of the target mass, the mechanical baffle of the corresponding diversion channel is closed, and 3-7 test samples are formed in each batch. In this embodiment, the basic process of the system is as follows: the whole batch of bird's nest samples are sequentially sent into the injection, homogenization, quantitative splitting and compaction detection process. At each stage, mass, temperature, thickness, pressure, near-infrared reflectance spectrum and multi-frequency impedance data are collected simultaneously. The controller then judges whether the measurement conditions are met, corrects the spectral results and outputs the batch results. Among them, the homogenization module is used to deagglomerate bird's nest fiber agglomerates, such as locally clustered and slightly adhered fiber bundles; the quantitative diversion module is used to divide the same batch of samples into multiple test sub-samples; and the compaction test module is used to compact the sub-samples into a test layer that is easy to measure consistently. Unlike conventional testing methods that involve independent processing of samples without quantitative standards, this implementation method transforms the problems of insufficient sample representativeness and large single-point sampling deviations into statistical judgment of multiple subsamples through the coordinated arrangement of homogenization, splitting, and compaction. This implementation method can output test subsamples with more consistent morphology and more comparable data for subsequent component detection without destructive pretreatment. Based on the above implementation method, the homogenization module uses a rotating drum to drive a flexible feeding plate to perform low-shear feeding on the sample. The flexible feeding plate inside the rotating drum maintains a gap of 2-5mm with the bottom of the hopper to avoid excessive shearing of the bird's nest fibers, while still gradually breaking up the loose agglomerates. The rotating drum speed is set to 20-80r / min and the running time is set to 15-60s to achieve a balance between the dispersing effect and the stability of the sample state. Among them, the injection temperature is used as the reference temperature before homogenization, the sample temperature during the homogenization process is used to reflect the temperature rise caused by material feeding, and the difference between the two is used to determine whether the sample has generated an undesirable temperature rise due to mechanical friction. Unlike conventional homogenization which relies solely on high-speed stirring, this implementation method captures the implicit patterns of sample heating and fiber structure changes through the combined constraints of controlled rotation speed, limited gap, and temperature rise threshold. It simultaneously incorporates two contradictory issues—insufficient dispersion and excessive heating—into the control conditions. This implementation method avoids sample deviation from its original state due to excessive homogenization, while also avoiding insufficient representativeness of subsequent sampling due to insufficient homogenization. Based on the above implementation method, the quantitative diversion module cuts off the outflowing sample through a mechanical baffle, and sets up a food-grade mesh plate for leaching and dehydration for soaked bird's nest to reduce the interference of surface free water on the judgment of diversion quality; the weighing sensor continuously acquires the mass change at a sampling frequency of 100-500Hz, and dry bird's nest and soaked bird's nest correspond to different target masses of individual detection subsamples, and the corresponding mechanical baffle is closed when the current mass is 95%-105% of the target mass; Here, target quality refers to the control quality of a single test sample. For example, 2-5g of dried bird's nest and 5-15g of soaked bird's nest are used. Its purpose is to ensure that different samples have comparable sample quantities during subsequent compaction and spectral measurement. Unlike the experience-based judgment without quantitative standard constraints in manual sampling, this implementation method limits the sampling deviation within a preset range through continuous feedback from the weighing sensor and timely interception by the mechanical baffle. This implementation method can make the morphology of different samples more consistent, thereby providing sufficient samples for subsequent statistical judgment.

[0019] In this embodiment, the descent speed of the movable pressure plate is 2-10 mm / s, the target thickness of the detection layer is 3-10 mm, and the upper limit of the pressure on the detection layer is 5-30 kPa. The probe of the near-infrared spectroscopy detection module is set below the detection box and collects the near-infrared diffuse reflectance spectrum of the detection layer through the quartz window at the bottom of the detection box. The four electrodes of the electrode impedance detection module are embedded in the side wall of the detection box and are located at the same height position as the probe inside the detection box corresponding to the detection layer. In this embodiment, the electrode impedance detection module is used to acquire the resistance and reactance components at frequencies of 1kHz, 10kHz, and 100kHz, and calculate the frequency according to the following formula. impedance amplitude : in, For frequency The resistive component below, For frequency The reactance component below; In this embodiment, after the detection layer is formed, the sample is compacted by a movable pressure plate at a controlled speed, and near-infrared reflectance spectrum and multi-frequency impedance data are collected simultaneously when the thickness and pressure meet the conditions; the near-infrared probe obtains diffuse reflectance spectrum through quartz window, and the four electrodes collect impedance information from the side wall of the detection box. The two types of information jointly characterize the composition and physical state of the sample. The same height position is referenced to the geometric mid-surface of the detection layer after it has been stably compacted, so that the probe acquisition area and the electrode acquisition area correspond to the same cross section of the detection layer; the target thickness refers to the morphological thickness of the detection layer that is suitable for repeated measurements, such as 3-10mm; the upper limit of pressure is used to avoid excessive compaction that would cause significant changes in the pore structure. Unlike conventional methods that rely solely on spectra, this implementation incorporates compaction state and electrical response into the measurement conditions, avoiding deviations in spectral results caused by changes in optical path and contact state; this implementation can improve sampling stability while ensuring relatively consistent detection layer morphology. Based on the above implementation method, the electrode impedance detection module acquires the resistance and reactance components at three frequency points: 1kHz, 10kHz, and 100kHz, respectively, and calculates the impedance amplitude at each frequency point according to the impedance amplitude formula; here... and These respectively reflect the resistive and capacitive responses of the sample to alternating current, and the impedance amplitude. This is used to unify the electrical responses at different frequencies to a comparable numerical scale; Among them, low-frequency points are more likely to reflect the influence of moisture, ion migration and contact state, while high-frequency points are more likely to reflect changes in the dielectric response inside the sample. Therefore, multi-frequency acquisition is better able to distinguish between the differences in the sample itself and the unstable electrode contact than single-frequency acquisition. Unlike impedance processing methods that only look at single-point values, this implementation method captures the law of impedance change with frequency through the coordinated acquisition of multiple frequency points, and quantifies factors that affect spectral measurements in advance, such as surface water film and porosity changes. This implementation method can output input quantities that are closer to the actual measured state of the sample for subsequent calibration models.

[0020] This embodiment also provides an automated method for detecting bird's nest components, implemented based on the automated bird's nest component detection system of the present invention. The method includes: The sample injection temperature is collected, and the bird's nest sample is dispersed and homogenized through the homogenization module. The sample temperature during the homogenization process is collected, and the sample temperature rise is determined based on the difference between the injection temperature and the sample temperature. Homogenization continues when the sample temperature rise is less than or equal to the preset temperature rise threshold. The homogenized bird's nest sample is divided into multiple test sub-samples based on the mass collected by the weighing sensor using the quantitative flow separation module. Among them, the soaked bird's nest is dehydrated by leaching through the food-grade mesh plate in the quantitative flow separation module. The test sample is fed into the test box, the movable pressure plate is controlled to descend, the thickness is determined according to the displacement detected by the displacement sensor, and the movable pressure plate is controlled according to the pressure detected by the pressure sensor to press the test sample into a test layer. When the thickness of the detection layer reaches the target thickness and the pressure does not exceed the upper limit of the pressure, near-infrared reflectance spectrum and impedance at multiple frequencies are collected simultaneously. The impedance, thickness, and pressure are used to determine whether the test sample meets the measurement conditions. The near-infrared reflectance spectrum of the test sample that meets the measurement conditions is input into the pre-established spectral composition analysis model to obtain the initial composition detection results. The mass, temperature, thickness, and impedance obtained by the weighing sensor and temperature sensor are input into the correction model to obtain the composition content of each test sample. Based on the component content of multiple test samples within the same batch and the statistical distribution of impedance amplitude of multiple test samples at multiple frequencies, the batch test results or prompts for unevenness within the batch are output. This embodiment also provides an automated method for detecting bird's nest components. The step of determining whether a sample meets the measurement conditions based on impedance, thickness, and pressure includes: determining the thickness of the detection layer. Does it meet the requirements? ,pressure Does it meet the requirements? ,in This is the preset lower limit of the detection layer thickness. This is the preset upper limit of the detection layer thickness. The preset lower limit of pressure, The preset pressure limit; and whether the impedance amplitude at 10kHz and 100kHz is satisfied: in, for Impedance amplitude at frequency, for Impedance amplitude at a given frequency; when the above conditions are simultaneously met, the sample being tested is deemed to meet the measurement conditions; among which, According to The average of the impedance amplitude difference ratios of standard samples of the same grade and moisture content and standard deviation According to the formula A defined threshold; This embodiment also provides an automated method for detecting bird's nest components, which sequentially performs dark current correction, white plate reference correction, median filtering, smoothing, and standard normal variable transformation on the near-infrared reflectance spectrum to obtain the first... Corrected spectral values ​​at each wavelength point And using multiple linear regression or partial least squares regression, the first step is to calculate the second step using the following formula. Content of each component : Among them, the The first component refers to the first component in the calibration set of the same model version. The analyte corresponding to the reference label of the analyte must have a name that matches the reference label. The serial number of the component to be tested. , The preset number of component reference labels within the same model version. It is a positive integer; during the same batch of testing, and each The corresponding ingredient items remain unchanged. The wavelength point number, , This represents the number of wavelength points in the corrected spectrum. For the first Corrected spectral values ​​at each wavelength point; To test the quality of the subsamples, To detect layer thickness, The temperature of the sample being detected during spectral acquisition; , , , , , , The regression coefficients are obtained through calibration, and their values ​​are determined according to the calibration method of standard samples with known component contents, and remain unchanged during the same model version and the same batch of testing; This embodiment also provides an automated method for detecting bird's nest components. When the pressure reaches the upper limit but the thickness of the detection layer does not reach the target thickness, the movable pressure plate is released, causing the detection box to vibrate back and forth in the horizontal direction. The descent speed of the movable pressure plate is reduced and then it is re-compacted. After re-compacting, the thickness, pressure and impedance are detected again. If the measurement conditions are still not met after two consecutive re-compactings, the detection sample is isolated. For soaked bird's nest, it is dehydrated by filtration through a food-grade mesh before compaction. When the absolute value of the mass change rate calculated based on the continuous sampling value of the weighing sensor within 2 seconds is greater than or equal to the preset mass change rate threshold, it waits. When the impedance amplitude at 1kHz is lower than the preset impedance lower limit, the sample is automatically flipped and replaced at the preset flipping speed before detection or a manual verification prompt is output. The preset impedance lower limit is the applicable lower limit for model training samples. This embodiment also provides an automated method for detecting bird's nest components, defining a test sample that meets the measurement conditions and obtains the component content as a valid test sample, and calculating the median and interquartile range of the component content of each valid test sample in the same batch; When there are at least 3 valid detection samples and their proportion of the total number of detection samples obtained by current splitting is not less than 80%, the coefficient of variation determined by the ratio of interquartile range to median does not exceed the preset upper limit of coefficient of variation, and the statistical distribution of the impedance amplitude of each valid detection sample at 10kHz does not show a double peak as determined by the preset peak number determination algorithm, the median is output as the batch detection result; if the above conditions are not met, a prompt message indicating unevenness within the batch is output. In this embodiment, the core process is to first stabilize and control the state parameters of the sample, and then perform spectral acquisition and composition calculation. After the sample is injected, the fiber clumps are first broken up by a homogenization module, and the difference between the injection temperature and the homogenization process temperature is used to determine whether excessive heating has occurred. A weighing sensor is used to control the split mass so that the sample enters multiple detection subsamples. Then, the compaction process is controlled by displacement and pressure to ensure that the detection layer reaches the target thickness and then near-infrared reflectance spectrum and multi-frequency impedance are acquired simultaneously. Among them, meeting the measurement conditions does not depend on a single index, but simultaneously checks the thickness, pressure, and the change ratio of impedance amplitude at 10kHz and 100kHz; the acquisition of impedance, thickness, and pressure is based on the data of the stable period after the pressure plate stops; when the pressure or impedance fluctuates continuously within a short time window, the controller maintains the current sample and delays the judgment until the value enters the stable range before performing the judgment on whether the measurement conditions are met. During the stable period, after the pressure plate stops, a continuous sampling interval of 0.5s-2s is performed. The controller calculates the arithmetic mean of the thickness, pressure, and impedance sampling values ​​within this interval as representative values. When the number of sampling points is less than the preset lower limit, the sampling period is extended until the preset lower limit is reached. The allowable fluctuation range of thickness, pressure, and impedance is determined by the equipment calibration. The allowable fluctuation range of thickness is 1%-5% of the target thickness, the allowable fluctuation range of pressure is 1%-5% of the current pressure reading, and the allowable fluctuation range of impedance amplitude is 1%-5% of the corresponding representative value. If any parameter exceeds its allowable fluctuation range, the current time period is considered unstable. The sample temperature here refers to the same temperature parameter at different stages, including the injection temperature and the sample temperature during homogenization. The injection temperature is used to determine the baseline state before homogenization, and the sample temperature during homogenization is used to determine the temperature rise during the material feeding process. When the subsample completes compaction and enters the spectral acquisition stage, the calibration model... The temperature value of the sample obtained most recently from the temperature sensor at the time of spectral acquisition is used. In the calibration model The most recent stable weighing value of the sample before it enters the compaction testing module is used. The thickness representative value of the same test sample within a stable period is used. and The impedance amplitude at the corresponding frequency point within the same stable time period is used as a representative value to ensure that all input parameters come from the same detection sample and the same measurement process; Based on the above implementation method, the measurement conditions are determined by combining the thickness range, pressure range, and impedance difference ratio. and These are the lower and upper limits of the allowable thickness of the detection layer, respectively, and their values ​​correspond to the equipment calibration range of the detection box structure, the optical path of the near-infrared probe, and the target thickness of the detection layer. and These are the lower and upper limits of the allowable pressure for the detection layer, respectively. The compaction pressure shall not exceed the aforementioned upper limit; in specific implementation, and Set within the target thickness range of 3-10mm. and The pressure limit is set within the range of 5-30 kPa and remains unchanged under the same detection box structure and the same model version; In order to make With a clearly identifiable source, 3-5 standard samples were selected from samples of the same grade and moisture content range as the batch to be tested. Homogenization, splitting, compaction, and stabilization measurements were performed on each standard sample. The impedance amplitude of each standard sample was collected at 10kHz and 100kHz, and the corresponding difference ratio was calculated. The controller then averaged each difference ratio. and standard deviation Then, determine according to the above difference ratio formula. ; When changing the sample grade, moisture content range, or test kit structure, recollect the corresponding standard sample and update the test kit. ;when When the impedance is zero or lower than the minimum effective impedance value preset by the equipment, the difference ratio division calculation is not performed. Instead, the test sample is determined to be unsatisfied with the current round of measurement conditions, and is processed according to the above-mentioned re-compaction process, sample flipping and replacement for re-testing, or isolation process. when Impedance difference ratio judgment is only performed when the minimum effective impedance value is reached; this avoids invalid judgments caused by a denominator of zero or below the preset model detection lower limit; the difference ratio is used to characterize the frequency response consistency of the detection layer at 10kHz and 100kHz, and when it exceeds... When the situation is as described, the primary cause is uneven distribution of pores within the detection layer, abnormal electrode contact, or unstable free water distribution, rather than directly determining that the bird's nest components are abnormal. The controller performs the judgment in a fixed order: first, it reads the representative thickness value within the stable period and determines whether it falls within the range. to After the thickness is deemed acceptable, read the pressure value and determine if it falls into the required range. to Both thickness and pressure are qualified and After reaching the minimum effective impedance value, calculate the impedance difference ratio and compare it with... Comparison; only when all of the above conditions are met will the sample be included in the spectral composition calculation; if any condition is not met, the sample will be marked as not meeting the measurement conditions of the current round, and the spectral and impedance data collected in that round will not be used to calculate the composition content. Based on the above implementation method, the near-infrared reflectance spectrum is first corrected for dark current and white plate reference to eliminate instrument offset and reflectance reference difference. Then, local peaks and baseline fluctuations are weakened by median filtering, smoothing and standard normal variable transformation to obtain the corrected spectral value. Median filtering is used to remove isolated outliers, smoothing is used to reduce random fluctuations at adjacent wavelengths, and standard normal transformation is used to reduce the impact of particle size and scattering differences on the spectral baseline. Each process is executed in the order described above; the spectral composition analysis model and the calibration model use fixed regression coefficients from the same model version; the spectral composition analysis model first determines the initial component detection results based on the preprocessed spectral sequence, where the initial component detection results correspond to the constant term and spectral term in the regression expression; the calibration model then uses the impedance amplitude, thickness, mass, and temperature of the same sample as state correction inputs to obtain the final result. ; The above calculation formula is a unique calculation expression that expands the initial spectral results and each state correction term. The controller no longer uses another set of independent coefficients to apply the already obtained... Repeated calculations are performed to avoid two undefined calculation paths between the initial results and the final component content; in, As the benchmark, Used to characterize the The wavelength point pair of the first The influence of various components, and Used to correct the effects of 10kHz and 100kHz impedance on moisture, porosity, and electrode contact conditions. Used to correct the effect of the detection layer thickness on the optical path. Used to correct the effect of sample quality on packing density. Used to compensate for the effect of sample temperature on spectral response; Each coefficient was obtained by establishing a calibration set using standard samples with known component content. The calibration set consisted of no less than 30 sets and covered both dried and rehydrated bird's nests. For each set of standard samples, spectral, impedance, thickness, mass, and temperature data were obtained by following the same injection, homogenization, splitting, compaction, and collection process as the test sample. The corresponding component content obtained by manual or chemical determination was used as a reference label for fitting. When the number of calibration wavelength points is no greater than the number of calibration samples and the correlation between wavelength points does not exceed the preset allowable range, multiple linear regression is used; when the number of wavelength points is greater than the number of calibration samples or the correlation between adjacent wavelength points exceeds the preset allowable range, partial least squares regression is used. The number of latent variables for partial least squares regression was determined through cross-validation on the calibration set, with the final number being the latent variables whose cross-validation error was the lowest and did not exceed a preset error upper limit. The regression coefficients were obtained using the mean squared error between the predicted component content and the reference label as the fitting criterion, and remained constant during the same batch of testing. For each model version, the controller stored and... Corresponding coefficient group Each coefficient group is determined by the same calibration set according to the above fitting criteria and serves as a fixed value for calculating the content of the corresponding component under this model version; when the sample grade, water content or detector structure changes, a new model version is established. Based on the above implementation method, when the pressure reaches the upper limit first during the compaction process but the thickness still does not reach the target value, it indicates that there may be agglomerates, local hard blocks or uneven distribution inside the sample. At this time, release the movable pressure plate and make the detection box vibrate back and forth in the horizontal direction, and then reduce the descent speed of the pressure plate to re-compact. When recompacting, first raise the movable pressure plate until it is detached from the surface of the test sample, then start the test box to vibrate back and forth in the horizontal direction for a preset time, adjust the descent speed of the movable pressure plate to 50%-90% of the previous compaction descent speed and press it down again; after recompacting, the measurement conditions are still judged in the order of thickness, pressure and impedance. If the conditions are still not met after two consecutive re-compactions, the test sample is isolated and no longer included in the batch results. For soaked bird's nest, it is first dehydrated by filtration through a food-grade mesh, and then the mass change rate of continuous sampling is used to determine whether it can be compacted. The mass change rate within 2 seconds is determined according to the mass difference and time interval between adjacent sampling times, and the maximum value of the absolute value of the mass change rate within the 2-second period is used as the judgment value. When the judgment value is less than the preset mass change rate threshold, compaction is initiated; the preset mass change rate threshold is calibrated according to the equipment no-load noise and the mass change level at the end of the normal leaching and dewatering stage of the screen, and is set to 0.01g / s-0.10g / s, and remains unchanged under the same equipment and the same screen structure; if the impedance amplitude at 1kHz is lower than the preset impedance lower limit, it indicates that the sample state deviates from the applicable range of the model training sample; the preset impedance lower limit is the applicable lower boundary of the impedance amplitude at 1kHz in the model training sample, and is updated synchronously when the training sample range is changed; At this point, the current sample is removed from the detection position, and the flipping mechanism completes the sample flipping and replacement at the preset flipping speed and returns to the detection position. The compaction, stability judgment, impedance acquisition, spectrum acquisition and composition calculation are then performed again. If the impedance amplitude at 1 kHz is lower than the preset impedance lower limit again after the sample flipping and replacement, a manual review prompt is output and the sample flipping and replacement is not repeated. Based on the above implementation method, the batch result is not a simple average value, but the median of the effective detection samples is used as the output result, and consistency judgment is made by combining the interquartile range and impedance distribution; the controller first counts the number of effective detection samples that meet the measurement conditions and have completed the component calculation. and the total number of samples obtained from the diversion ; when If the proportion of valid test samples to total test samples is less than 80%, a prompt message indicating unevenness within the batch will be output directly; after the quantity condition is met, the component content of each valid test sample will be sorted from smallest to largest and the value of the middle position will be taken as the median; when the number of valid test samples is even, the arithmetic mean of the two middle values ​​will be taken as the median. Determine the upper and lower quartiles, and use the difference between them as the interquartile range. Then compare the ratio of the interquartile range to the median with the preset upper limit of the coefficient of variation. The preset upper limit of the coefficient of variation is determined according to the ratio of the interquartile range to the median of the same grade standard batch under stable testing conditions, and is taken as 0.05-0.30. When the test object is changed to a different grade or a different moisture content, the upper limit is updated again according to the corresponding standard batch. When the median of a certain component to be tested is zero or lower than the preset model detection limit, such that the ratio of interquartile range to median cannot be calculated stably, the batch consistency of the component is judged as unsatisfactory, and a prompt message of non-uniformity within the batch is output. The batch is not directly judged as qualified. When the coefficient of variation does not exceed the preset upper limit, the preset peak number determination algorithm is then executed on the statistical distribution of the impedance amplitude of each valid detection sample at 10kHz. The peak number determination algorithm first sorts the impedance amplitudes, and then identifies the concentrated distribution based on local density changes. When the high impedance region and the low impedance region form mutually separated concentrated distributions, and there is a continuous separation region between the two concentrated distributions, it is determined that there are two peaks. When no bimodal distribution is observed, the median is output as the batch test result. When a bimodal distribution is observed, an indication of unevenness within the batch is output. Thus, the batch output is executed in a fixed order according to the quantity conditions, the degree of dispersion of components, and the impedance distribution state, avoiding misjudgment of batch results when occasional abnormal samples or samples with different water content are mixed in. Based on the above implementation, the controller establishes five states for each sample: to be tested, stable, retest, isolated, and output. All sensor data first enter the to be tested window. Only after the thickness, pressure, and impedance conditions are simultaneously stable will the controller switch to the stable state and trigger spectral acquisition. If any condition exceeds the limit, the current window is directly marked as retest and will not participate in the regression calculation. The stable window is organized using timestamp alignment. The spectrum, thickness, pressure, mass, and temperature must correspond to the representative values ​​within the same window. If any channel is missing, drifting, or timed out, it is only allowed to enter the retest or isolation state, and is not allowed to directly enter the output state. The preset peak number determination algorithm prioritizes the normalized impedance histogram distribution and combines it with the local density changes of adjacent sub-boxes for identification. When the local density shows a continuous rise in both intervals and a continuous trough in the middle, it is determined to be a bimodal distribution. The local density is calculated by counting the number of samples falling into each bin with a preset impedance amplitude as the step size. When the peak number of samples in both the high impedance region and the low impedance region is greater than the preset number threshold, and the number of samples in the bin between the two continuously decreases to the minimum value and the minimum value is lower than a certain percentage of the peak value, it is determined to be a continuous trough in the middle. Otherwise, it is determined to be a single peak. The specific percentage is 30%-50% lower than the peak value, and continuous decrease means that the number of samples in at least three consecutive bins shows a decreasing trend. For test samples that still cannot form a stable window after retesting, the controller marks them as invalid samples and removes them from the statistical set determined by median, interquartile range and peak number, thereby avoiding invalid samples from interfering with batch results.

[0021] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. An automated bird's nest component detection system, characterized in that, include: Mechanical actuators, sensing and detection devices, and controllers that are interconnected and communicate with each other; The mechanical actuator includes a sample injection module, a homogenization module, a quantitative flow splitting module, and a compaction detection module connected in sequence, wherein the compaction detection module includes a detection box and a movable pressure plate; The sensing and detection device includes a weighing sensor, a temperature sensor, a near-infrared spectroscopy detection module, an electrode impedance detection module, a displacement sensor, and a pressure sensor, which are used to collect the mass of the bird's nest sample, the injection temperature and the sample temperature during the homogenization process, the near-infrared reflectance spectrum, the impedance at multiple frequencies, the displacement of the movable pressure plate, and thereby determine the thickness of the detection layer and the pressure on the detection layer. The controller is used to determine whether the detection samples obtained by the quantitative shunting module meet the measurement conditions based on the impedance, thickness and pressure. It inputs the near-infrared reflectance spectrum into a pre-established spectral composition analysis model to obtain the initial composition detection results. It inputs the mass, thickness, temperature and impedance into a correction model to correct the initial composition detection results to obtain the corrected composition detection results of each detection sample. It performs statistical processing on the corrected composition detection results of multiple detection samples and outputs the batch detection results.

2. The automated bird's nest component detection system according to claim 1, characterized in that, The homogenization module includes a rotating drum and a flexible feeding plate. A material hopper is formed inside the rotating drum. The flexible feeding plate is disposed inside the rotating drum and rotates with the rotating drum. It is made of silicone or polytetrafluoroethylene and maintains a gap of 2-5mm with the bottom of the material hopper of the rotating drum. The rotating drum rotates at a speed of 20-80 r / min and runs for 15-60 s. The homogenization module continues to operate when the increase in sample temperature relative to the injection temperature is less than or equal to a preset temperature threshold, which is 3℃-5℃.

3. The automated bird's nest component detection system according to claim 1, characterized in that, The quantitative diversion module is equipped with multiple diversion channels, each of which is equipped with a mechanical baffle. The sampling frequency of the weighing sensor of the sensing and detection device is 100-500Hz. The target mass of a single test sample of dried bird's nest is 2-5g, and the target mass of a single test sample of soaked bird's nest is 5-15g. When the mass of the test sample is 95%-105% of the target mass, the mechanical baffle of the corresponding diversion channel is closed, and 3-7 test samples are formed in each batch.

4. The automated bird's nest component detection system according to claim 1, characterized in that, The movable pressure plate descends at a speed of 2-10 mm / s, the target thickness of the detection layer is 3-10 mm, and the upper limit of the pressure on the detection layer is 5-30 kPa. The probe of the near-infrared spectroscopy detection module is located below the detection box and collects the near-infrared diffuse reflectance spectrum of the detection layer through the quartz window at the bottom of the detection box. The four electrodes of the electrode impedance detection module are embedded in the side wall of the detection box and are located at the same height as the probe inside the detection box corresponding to the detection layer.

5. The automated bird's nest component detection system according to claim 1, characterized in that, The electrode impedance detection module is used to acquire resistance and reactance components at frequencies of 1kHz, 10kHz, and 100kHz, and calculate the frequency according to the following formula. impedance amplitude : in, For frequency The resistive component below, For frequency The reactance component below.

6. An automated method for detecting the components of bird's nest, characterized in that, Based on the automated bird's nest component detection system according to claim 1, the method includes: The sample injection temperature is collected, and the bird's nest sample is dispersed and homogenized through the homogenization module. The sample temperature during the homogenization process is collected, and the sample temperature rise is determined based on the difference between the injection temperature and the sample temperature. Homogenization continues when the sample temperature rise is less than or equal to the preset temperature rise threshold. The homogenized bird's nest sample is divided into multiple test sub-samples based on the mass collected by the weighing sensor using the quantitative flow separation module. Among them, the soaked bird's nest is dehydrated by leaching through the food-grade mesh plate in the quantitative flow separation module. The test sample is fed into the test box, the movable pressure plate is controlled to descend, the thickness is determined according to the displacement detected by the displacement sensor, and the movable pressure plate is controlled according to the pressure detected by the pressure sensor to press the test sample into a test layer. When the thickness of the detection layer reaches the target thickness and the pressure does not exceed the upper limit of the pressure, near-infrared reflectance spectrum and impedance at multiple frequencies are collected simultaneously. The impedance, thickness, and pressure are used to determine whether the test sample meets the measurement conditions. The near-infrared reflectance spectrum of the test sample that meets the measurement conditions is input into the pre-established spectral composition analysis model to obtain the initial composition detection results. The mass, temperature, thickness, and impedance obtained by the weighing sensor and temperature sensor are input into the correction model to obtain the composition content of each test sample. Based on the component content of multiple test samples within the same batch and the statistical distribution of impedance amplitude of multiple test samples at the multiple frequencies, the batch test results or the prompt information of non-uniformity within the batch are output.

7. The automated method for detecting bird's nest components according to claim 6, characterized in that, The steps for determining whether a test sample meets the measurement conditions based on impedance, thickness, and pressure include: determining the thickness of the test layer. Does it meet the requirements? ,pressure Does it meet the requirements? ,in This is the preset lower limit of the detection layer thickness. This is the preset upper limit of the detection layer thickness. The preset lower limit of pressure, The preset pressure limit; and whether the impedance amplitude at 10kHz and 100kHz is satisfied: in, for Impedance amplitude at frequency, for Impedance amplitude at a given frequency; when the above conditions are simultaneously met, the sample being tested is deemed to meet the measurement conditions; among which, According to The average of the impedance amplitude difference ratios of standard samples of the same grade and moisture content and standard deviation According to the formula A defined threshold.

8. The automated method for detecting bird's nest components according to claim 6, characterized in that, The near-infrared reflectance spectrum was sequentially subjected to dark current correction, white board reference correction, median filtering, smoothing, and standard normal variable transformation to obtain the first... Corrected spectral values ​​at each wavelength point And using multiple linear regression or partial least squares regression, the first step is to calculate the second step using the following formula. Content of each component : ;in, To test the quality of the subsamples, To detect layer thickness, The temperature of the sample being detected during spectral acquisition. , , , , , , These are the regression coefficients obtained through calibration.

9. The automated method for detecting bird's nest components according to claim 6, characterized in that, When the pressure reaches the upper limit but the thickness of the detection layer does not reach the target thickness, the movable pressure plate is released, causing the detection box to vibrate back and forth in the horizontal direction. After reducing the descent speed of the movable pressure plate, it is re-compacted. After recompaction, the thickness, pressure and impedance are tested again. If the measurement conditions are still not met after two consecutive recompactions, the test sample is isolated. For soaked bird's nest, before compaction, it is filtered and dehydrated through the food-grade mesh plate. When the absolute value of the mass change rate calculated based on the continuous sampling value of the weighing sensor within 2 seconds is greater than or equal to the preset mass change rate threshold, it waits. When the impedance amplitude at 1kHz is lower than the preset impedance lower limit, the sample is automatically flipped and replaced at the preset flipping speed before detection or a manual verification prompt is output. The preset impedance lower limit is the applicable lower limit of the model training sample.

10. The automated method for detecting bird's nest components according to claim 6, characterized in that, The test sample that meets the measurement conditions and obtains the component content is defined as a valid test sample. The median and interquartile range of the component content of each valid test sample in the same batch are calculated respectively. When there are at least 3 valid detection samples and their proportion of the total number of detection samples obtained by shunting is not less than 80%, the coefficient of variation determined by the ratio of interquartile range to median does not exceed the preset upper limit of coefficient of variation, and the statistical distribution of the impedance amplitude of each valid detection sample at 10kHz is determined by the preset peak number determination algorithm not to show a double peak, the median is output as the batch detection result; if the above conditions are not met, a prompt message indicating unevenness within the batch is output.