Fresh medicine osmotic pressure equipment detection system and method based on multi-data fusion

Abstract

The application discloses a fresh medicine osmotic pressure equipment detection system and method based on multi-data fusion, and the system comprises a multi-source acquisition unit, which is used for acquiring osmotic pressure time series data, spectral response data and electrochemical response data of a fresh medicine sample to be detected within a preset detection period, and forming original multi-source time series data sets with sampling time markers; and an event extraction unit, which is used for extracting key response events from various original detection data and forming event feature sequences of corresponding modes, and the application relates to the technical field of fresh medicine detection. The fresh medicine osmotic pressure equipment detection system and method based on multi-data fusion can establish a correlation model between the functional state of fresh medicine and the osmotic pressure change process through key steps such as event extraction, feature fusion and interval backstepping, and can output individualized functional osmotic pressure intervals, which provides a scientific basis for the quality control, activity evaluation and storage and transportation management of fresh medicine.

Description

Technical Field

[0001] This invention relates to the field of fresh herb detection, and more specifically, to a fresh herb osmotic pressure testing system and method based on multi-data fusion. Background Technology

[0002] In the research and application of fresh medicinal herbs, osmotic pressure is one of the key indicators reflecting the physiological state and efficacy of fresh herbs. Accurately measuring the osmotic pressure of fresh medicinal herbs is of vital importance for assessing their quality, determining storage and transportation conditions, guiding pre-processing, and ensuring the stability of their efficacy.

[0003] Traditional methods for detecting the osmotic pressure of fresh herbs often rely on a single type of data acquisition and analysis, such as obtaining osmotic pressure values ​​solely through an osmotic pressure sensor. However, the physiological processes of fresh herbs are complex, and a single data source cannot comprehensively and accurately reflect the intrinsic relationship between osmotic pressure changes and efficacy. Different physiological processes may have varying effects on osmotic pressure, and different detection methods emphasize different aspects of information. A single data source cannot comprehensively consider the effects of multiple factors on osmotic pressure, resulting in limited detection results that cannot accurately reflect the true functional state of fresh herbs.

[0004] Furthermore, in actual testing, data from different testing modalities may exhibit differences in response time, i.e., a lag phenomenon. This lag leads to asynchrony of data on the time axis, making cross-modal data analysis difficult and hindering the accurate establishment of a correlation model between the functional state of fresh herbs and changes in osmotic pressure. Consequently, it affects the accurate determination of the functional osmotic pressure range of fresh herbs, and fails to provide a scientific and reasonable basis for the storage, transportation, and processing of fresh herbs. Summary of the Invention

[0005] The purpose of this invention is to provide a fresh herb osmotic pressure detection system and method based on multi-data fusion. This invention solves the problem that existing fresh herb osmotic pressure detection methods often rely on only a single type of data acquisition and analysis. Fresh herbs have complex physiological processes, and a single data source cannot comprehensively and accurately reflect the intrinsic relationship between osmotic pressure changes and efficacy. Different physiological processes may have different effects on osmotic pressure, and different detection methods reflect different information focuses. A single data source cannot comprehensively consider the effects of multiple factors on osmotic pressure, resulting in limited detection results that cannot meet the application requirements.

[0006] This invention achieves the above objective through the following technical solution: a fresh herb osmotic pressure testing system based on multi-data fusion, the system comprising: Multi-source acquisition unit, event extraction unit, hysteresis modeling unit, dynamic compensation unit, feature fusion unit, interval back-inference unit, and result output unit; The multi-source acquisition unit is used to acquire osmotic pressure time-series data, spectral response data and electrochemical response data of the fresh herb sample to be tested within a preset detection period, and form an original multi-source time-series dataset with sampling time markers. The event extraction unit is used to extract key response events from various types of raw detection data and form event feature sequences of corresponding modalities; The hysteresis modeling unit is used to establish a multimodal response hysteresis relationship model based on the event feature sequence of each modality; The dynamic compensation unit is used to perform dynamic time-shift compensation and window reconstruction processing on the original multi-source time-series data according to the multimodal response hysteresis relationship model to form a unified and effective response sequence; The feature fusion unit is used to construct a fresh drug function detection feature vector based on the unified effective response sequence; The interval back-inference unit is used to establish a correlation model between the functional state of fresh medicine and the osmotic pressure change process based on the fresh medicine functional detection feature vector, and to screen the osmotic pressure time series segments that meet the preset drug efficacy response conditions, stability conditions, residual threshold conditions and continuity conditions, and to determine the osmotic pressure time series segments as the individualized functional osmotic pressure intervals corresponding to the fresh medicine samples. The result output unit is used to output one or more of the following based on the individualized functional osmotic pressure range: fresh drug quality grade, activity status evaluation result, storage and transportation osmotic pressure control range, pre-processing range, and detection anomaly warning result.

[0007] Furthermore, the multi-source acquisition unit includes: Osmotic pressure detection subunit, spectral detection subunit, and electrochemical detection subunit; The osmotic pressure detection subunit is used to continuously collect the osmotic pressure value sequence of fresh herb samples at a sampling frequency adapted to the physiological metabolic rate of the fresh herb variety. The spectral detection subunit is used to collect characteristic spectral signal sequences around the characteristic absorption range corresponding to the core active ingredient of fresh medicine. The electrochemical detection subunit is used to collect electrochemical response sequences that characterize changes in cell fluid ion migration state and cell activity. The multi-source acquisition unit is also used to perform unified time stamping, time alignment preprocessing, outlier screening, data storage, and sample identity association on the three types of detection data, so as to ensure that subsequent time series correlation analysis has a unified time benchmark, consistent data format, and traceable sample source.

[0008] Furthermore, the event extraction unit includes the following steps: The initial response time, local peak time, peak intensity, rise rate, fall rate, steady-state duration, and fluctuation width are extracted from the osmotic pressure time series data, spectral response data, and electrochemical response data, respectively, and event-level feature labels are generated according to the detection mode, event type, and stage of change. The event extraction unit is also used to perform validity screening based on the signal-to-noise ratio, sampling stability and consistency of changes of adjacent sampling points of candidate event points, so as to eliminate abnormal event points and invalid event points, and organize the remaining event points into event feature sequences and staged event segments in chronological order.

[0009] Furthermore, the hysteresis modeling unit includes the following steps: The time offset of the osmotic pressure signal relative to the spectral response signal and the time delay of the spectral response signal relative to the electrochemical response signal were calculated respectively. The corresponding time sequence relationship was weighted according to the degree of correlation between different modes and the efficacy characterization of fresh drug, and a multimodal response hysteresis relationship model was obtained to characterize the sequential relationship between local mutation, peak response and steady state formation. The lag modeling unit is also used to distinguish between direct pharmacodynamic correlation modal relationships and indirect pharmacodynamic correlation modal relationships, so that the multimodal response lag relationship model can simultaneously reflect the sequence, lag degree and stage coupling relationship of key response events.

[0010] Furthermore, the dynamic compensation unit includes the following steps: To address the asynchronous response characteristics of multimodal detection data, response extension processing is performed on the preceding modes, time offset correction processing is performed on the lagging modes, and window removal, window compression, and cross-valid interval reconstruction processing are performed on locally invalid sampling segments. The dynamic compensation unit is also used to allocate time offset weights according to the feature stage and feature type to which the sampling points belong, and to determine the window reconstruction range according to the fluctuation degree of the sampling segment, the duration of stability and the density of events, so that the compensated modal feature points can maintain temporal synchronization on a unified time axis and form a continuous response segment that can be used for cross-modal joint analysis.

[0011] Furthermore, the feature fusion unit includes the following steps: Normalization, correlation evaluation, and coupling mapping are performed on various features in the unified and effective response sequence to obtain the fresh drug function detection feature vector; The fresh herb function detection feature vector includes osmotic pressure change rate, spectral peak response intensity, peak occurrence time of active ingredient, electrochemical activity stability range, peak synchronization between modes, and residual hysteresis compensation between modes. The feature fusion unit is also used to assign feature weights based on the degree of correlation between each feature and the efficacy characterization of fresh medicine, and to uniformly convert features of different dimensions into standardized vectors that can be used as model inputs, so as to enhance the comprehensive characterization ability of the functional state of fresh medicine, cross-sample comparability and interval back-inference stability.

[0012] Furthermore, the interval back-calculation unit includes: Model construction sub-units and interval solution sub-units; The model building subunit is used to construct a ternary coupling relationship between the rate of change of osmotic pressure, the peak response of active ingredients, and the stable cell viability region based on the fresh drug function detection feature vector, and to form an association model for predicting the functional osmotic pressure range using sample training, parameter tuning, and performance verification, wherein the sample data is grouped according to training data, verification data, and test data; The interval solution subunit is used to traverse the osmotic pressure time series data in a sliding window manner, select continuous windows that meet the active ingredient response conditions, electrochemical activity stability conditions, intermodal compensation residual threshold conditions, and osmotic pressure change continuity conditions, and merge adjacent effective windows into a complete individualized functional osmotic pressure interval, which is used to characterize the target functional osmotic pressure range of fresh drug under the current physiological state.

[0013] Furthermore, the intermodal hysteresis compensation residual is used to characterize the time synchronization error of key feature points between the compensated reference mode and the target mode; The key feature points include the initial response point, local peak points, and steady-state starting points; The feature fusion unit or the interval back-inference unit is also used to weight the time synchronization error according to the detection importance corresponding to different key feature points to obtain the compensation effect evaluation result, and to determine whether to include the candidate key feature point in the residual calculation based on the signal credibility, peak significance and stage integrity, so as to use the compensation effect evaluation result as a constraint condition for effective segment selection.

[0014] Furthermore, the result output unit includes the following steps: The individualized functional osmotic pressure range is fed back to the osmotic pressure detection device or osmotic pressure control device to determine the osmotic pressure control setting value based on the upper and lower limits of the individualized functional osmotic pressure range, and to trigger parameter adjustment when the actual detected osmotic pressure deviates from the allowable range corresponding to the osmotic pressure control setting value. The result output unit is also used to generate fresh drug quality grade results, storage and transportation control suggestions, pre-processing suggestions, and detection and early warning results and result call interface information for sampling abnormalities, compensation abnormalities, interval abnormalities or equipment control deviations based on the matching relationship between the individualized functional osmotic pressure range and the standard osmotic pressure range.

[0015] A method for detecting the osmotic pressure of fresh herbs based on multi-data fusion, applied to the aforementioned detection system for fresh herbs based on multi-data fusion, the method comprising the following steps: S1. Collect osmotic pressure time-series data, spectral response data and electrochemical response data of the fresh herb sample to be tested within the preset detection period, and form an original multi-source time-series dataset with sampling time markers; S2. Extract key response events from various types of raw detection data and form event feature sequences for the corresponding modalities; S3. Establish a multimodal response hysteresis model based on the event characteristic sequences of each mode; S4. Based on the multimodal response hysteresis relationship model, perform dynamic time shift compensation and window reconstruction processing on the original multi-source time series data to form a unified and effective response sequence; S5. Construct a fresh herb function detection feature vector based on the unified and effective response sequence; S6. Based on the fresh drug function detection feature vector, establish a correlation model between the fresh drug function state and the osmotic pressure change process, and screen the osmotic pressure time series segments that meet the preset drug efficacy response conditions, stability conditions, residual threshold conditions and continuity conditions, and determine the osmotic pressure time series segments as the individualized functional osmotic pressure intervals corresponding to the fresh drug samples. S7. Based on the individualized functional osmotic pressure range, output one or more of the following: fresh drug quality grade, activity status evaluation result, storage and transportation osmotic pressure control range, pre-processing range, and detection abnormality early warning result.

[0016] The beneficial effects of this invention are as follows: 1. By combining osmotic pressure, spectral response, and electrochemical response data, and utilizing multi-source data fusion technology, comprehensive monitoring of the osmotic pressure state of fresh herbs was achieved. This fusion method breaks through the limitations of traditional single data sources and can more accurately reflect the physiological state and activity level of fresh herbs.

[0017] 2. By using hysteresis modeling units and dynamic compensation units, the asynchronous problem of multimodal data acquisition can be effectively solved. By performing time shift compensation and window reconstruction on different modal signals, the temporal consistency of the data is ensured, thus providing a reliable data foundation for subsequent analysis.

[0018] 3. Through key steps such as event extraction, feature fusion, and interval back-inference, the system can establish a correlation model between the functional state of fresh herbs and the osmotic pressure change process, and output individualized functional osmotic pressure ranges, which provides a scientific basis for the quality control, activity evaluation, and storage and transportation management of fresh herbs.

[0019] 4. The system can automatically process large-scale multimodal data, which greatly improves the efficiency of fresh herb osmotic pressure detection. Through accurate models and algorithms, it ensures the stability and reliability of the detection results. The system can provide real-time feedback of the detection results, providing strong support for the pre-processing, storage and transportation control of fresh herbs.

[0020] 5. The feature fusion unit performs unified standardization processing on multimodal data, making the data between different samples more comparable, thereby improving the system's generalization ability and the reliability of cross-sample analysis. Attached Figure Description

[0021] The accompanying drawings, which are included to provide a further understanding of this application and form part of this application, illustrate exemplary embodiments and are used to explain this application, but do not constitute an undue limitation of this application. In the drawings: Figure 1 This is a system block diagram of the present invention; Figure 2 This is a detailed flowchart of the multi-source acquisition unit of the present invention; Figure 3 This is a detailed flowchart of the interval back-calculation unit of the present invention; Figure 4 This is a flowchart of the overall method of the present invention. Detailed Implementation

[0022] The present application will now be described in further detail with reference to the accompanying drawings. It should be noted that the following specific embodiments are only used to further illustrate the present application and should not be construed as limiting the scope of protection of the present application. Those skilled in the art can make some non-essential improvements and adjustments to the present application based on the above application content.

[0023] Example 1: Please see Figures 1-3 This invention provides a technical solution: a fresh herb osmotic pressure testing system based on multi-data fusion, the system comprising: Multi-source acquisition unit, event extraction unit, hysteresis modeling unit, dynamic compensation unit, feature fusion unit, interval back-inference unit, and result output unit; The multi-source acquisition unit is used to place the fresh herb sample to be tested in the detection device and perform multi-source detection on the fresh herb sample within a preset detection period to obtain three types of raw detection data: osmotic pressure time series data, spectral response data, and electrochemical response data. Each type of detection data is accompanied by a corresponding sampling time mark to form a raw multi-source time series dataset. Among them, the fresh medicinal sample to be tested refers to a fresh Chinese medicinal material sample that needs to be tested for osmotic pressure and other related tests; the detection device is a device used to carry and test the fresh medicinal sample to realize multi-source detection function; the preset detection cycle is a pre-set time range for testing the fresh medicinal sample; multi-source detection means testing the fresh medicinal sample from multiple different sources or angles, including osmotic pressure, spectral, and electrochemical detection; osmotic pressure time series data is a data sequence composed of the osmotic pressure values ​​of the fresh medicinal sample recorded at different sampling times, reflecting the change of osmotic pressure over time; spectral response data is the spectral characteristic data of the fresh medicinal sample at different sampling times during spectral detection, reflecting the response characteristics of the sample to different wavelengths of light; electrochemical response data is the electrochemical signal data generated by the fresh medicinal sample at different sampling times during electrochemical detection, reflecting the electrochemical properties of the sample; and the original multi-source time series dataset is a data set formed by integrating the three types of original detection data—osmotic pressure time series data, spectral response data, and electrochemical response data—according to their respective sampling time markers. The event extraction unit is used to extract key response events from three types of raw detection data, namely, the start response time, local peak time, peak intensity, rise rate, fall rate, steady state duration, and fluctuation width, and to form event feature sequences corresponding to each mode based on the key response events. Among them, the initial response time is the moment when the fresh drug sample begins to produce a significant response in various types of raw detection data; the local peak time is the moment when the local maximum value appears in the raw detection data; the peak intensity is the magnitude of the local peak value in the raw detection data; the rate of ascent is the magnitude of the change in the response value per unit time during the ascent from the initial point to the peak point in the raw detection data; the rate of descent is the magnitude of the change in the response value per unit time during the descent from the peak point in the raw detection data; the steady-state duration is the duration during which the response value remains in a relatively stable state in the raw detection data; the fluctuation width is the difference between the maximum and minimum values ​​when the response value fluctuates near the steady state in the raw detection data; the key response events are important events extracted from the raw detection data, such as the initial response time and the local peak time, that can reflect the response characteristics of the fresh drug sample; and the event feature sequence is a sequence formed by arranging the key response events in a certain order to describe the response characteristics of each mode, including osmotic pressure, spectral, and electrochemical responses. The hysteresis modeling unit is used to perform time-series correlation analysis on event feature sequences of different modes, identify the sequential response patterns between osmotic pressure signals, spectral signals and electrochemical signals, and construct a multimodal response hysteresis relationship model; Among them, time-series correlation analysis analyzes the temporal relationship of event characteristic sequences of different modes, osmotic pressure, spectral, and electrochemical signals to find the sequential connections between them; sequential response law analyzes the order and characteristics of the temporal occurrence of signals of different modes; and multimodal response hysteresis model is a model constructed through time-series correlation analysis that can describe the sequential response law between osmotic pressure signals, spectral signals, and electrochemical signals. The dynamic compensation unit is used to perform dynamic time-shift compensation and window reconstruction on the original multi-source time series data according to the multimodal response hysteresis relationship model, so as to form a unified and effective response sequence that reflects the real physiological changes of fresh medicine. Among them, dynamic time-shift compensation, based on the multimodal response lag relationship model, performs corresponding time-shift compensation on the data of different modes in the original multi-source time series data to eliminate the time difference between different modal signals; window reconstruction processing, after dynamic time-shift compensation, recombines and divides the data to form a new data window in order to better reflect the real physiological changes of fresh medicine; unified effective response sequence, after dynamic time-shift compensation and window reconstruction processing, forms a data sequence that can comprehensively reflect the real physiological changes of fresh medicine. The interval back-inference unit is used to construct a fusion feature set based on a unified effective response sequence. This set includes features such as osmotic pressure change rate, spectral peak response intensity, peak occurrence time of active ingredient, electrochemical activity stability region, intermodal peak synchronization, and intermodal hysteresis compensation residual. The features in the fusion feature set are normalized and coupled to form a fresh drug function detection feature vector. A ternary coupling model of osmotic pressure change rate, active ingredient peak response, and cell viability stability region is constructed. The model is then trained on samples, hyperparameters are tuned, and performance is verified. Among them, the following features are considered: osmotic pressure change rate characteristics (the correlation between the rate of osmotic pressure change over time in the unified effective response sequence); spectral peak response intensity characteristics (the correlation between the magnitude of spectral peaks in the unified effective response sequence); active ingredient peak occurrence time characteristics (the correlation between the time when the active ingredient in the fresh drug reaches its peak in the unified effective response sequence); electrochemical activity stability range characteristics (the correlation between the time interval in which electrochemical activity remains relatively stable in the unified effective response sequence); intermodal peak synchronization characteristics (the characteristics of the temporal synchronization of peak values ​​of osmotic pressure, spectral, and electrochemical activity in different modes); intermodal hysteresis compensation residual characteristics (the correlation between the differences that still exist between different modes after compensation by the multimodal response hysteresis relationship model); fusion feature set (the feature set formed by integrating the above features); and normalization (the transformation of different feature data in the fusion feature set according to certain rules). To ensure the numerical range is within a unified standard range, facilitating subsequent processing and analysis; coupling mapping involves associating and transforming the normalized features to establish their intrinsic connections, forming a feature vector for fresh drug function detection; a ternary coupling model based on the fresh drug function detection feature vector reflects the relationship between the osmotic pressure change rate, the peak response of the active ingredient, and the cell viability stable zone; sample training uses sample data with known results to learn and adjust the ternary coupling model, enabling it to better fit the data and predict results; hyperparameter tuning adjusts and optimizes some pre-set parameters in the model to improve its performance and accuracy; performance validation uses independent test datasets to evaluate the trained and tuned model, verifying its effectiveness and accuracy in practical applications. The reverse calculation unit is used to use the trained ternary coupling model to screen out the osmotic pressure time series segment that meets the requirements of the active ingredient response reaching or near the peak value, the electrochemical activity being in a stable state, the multimodal compensation residual being lower than the preset threshold, and the osmotic pressure change being in a continuous and interpretable state. This osmotic pressure time series segment is determined as the individualized functional osmotic pressure interval corresponding to the fresh drug sample. Among these, the active ingredient response reaches its peak or near its peak, meaning the response value of the active ingredient in the unified effective response sequence reaches its maximum or near its maximum value; electrochemical activity is maintained in a stable state, meaning the electrochemical activity in the unified effective response sequence remains relatively stable; multimodal compensation residuals are below a preset threshold, meaning the difference between different modes after compensation is less than a preset standard value; osmotic pressure changes are in a continuous and explainable state, meaning the osmotic pressure change process conforms to a certain law and can be explained by reasonable theory; osmotic pressure time series segment, a range on the time axis that satisfies the above conditions of active ingredient response, electrochemical activity, multimodal compensation residuals, and osmotic pressure changes; and individualized functional osmotic pressure range, an osmotic pressure range determined for a specific fresh drug sample that reflects its functional characteristics. The result output unit is used to output at least one of the following results based on the individualized functional osmotic pressure range: fresh drug quality grade, activity status evaluation result, suitable storage and transportation osmotic pressure control range, suitable pre-processing range, and detection abnormality warning result. Among them, the fresh herb quality grade is a grade that classifies and evaluates the quality of fresh herb samples based on indicators such as individualized functional osmotic pressure range; the activity status evaluation result is the result obtained by assessing the activity status of active ingredients in fresh herb samples; the suitable storage and transportation osmotic pressure control range is the appropriate range of osmotic pressure to control in order to ensure the quality of fresh herbs during storage and transportation; the suitable pre-processing range is the appropriate range of osmotic pressure before processing fresh herbs to ensure processing effect and fresh herb quality; and the detection anomaly warning result is the warning information issued by the system when an abnormality occurs during the detection process.

[0024] It should be noted that during use, the multi-source acquisition unit acquires multiple types of raw data and marks the time, comprehensively and accurately recording fresh herb information. The event extraction unit extracts key response events to form feature sequences, which can accurately capture the response characteristics of fresh herbs. The hysteresis modeling unit constructs a multimodal response hysteresis relationship model to clarify the sequential rules between signals. The dynamic compensation unit processes the data to form a unified and effective response sequence, which more realistically reflects physiological changes. The interval back-inference unit constructs a fusion feature set and a ternary coupling model, which, after training and optimization, can comprehensively consider multiple factors. The back-inference solution unit determines the individualized functional osmotic pressure interval, which is highly targeted. The result output unit provides a variety of practical results, providing a scientific basis for fresh herb quality assessment, storage, transportation and processing, which helps to ensure the quality of fresh herbs and improve production and processing efficiency and quality control.

[0025] In one embodiment, the osmotic pressure time series data is a sequence of osmotic pressure values ​​of fresh herb samples continuously collected at a preset sampling frequency within the detection period. The sampling frequency is determined based on the physiological metabolic rate of the fresh herb variety, and the sampling frequency for herbaceous fresh herbs is set to... The sampling frequency for fresh woody medicinal herbs was set to It is used to accurately characterize the dynamic changes in osmotic pressure of fresh drug samples during the detection period; The spectral response data is a sequence of characteristic spectral signals of active ingredients in fresh herbal samples collected by a spectral detection module. The characteristic spectral bands are determined based on the characteristic absorption peaks of the core active ingredients in fresh herbs and are used to characterize the changes in the concentration of active ingredients in fresh herbal samples over time and their peak value changes. Electrochemical response data is a sequence of electrical signal changes in fresh drug samples collected by an electrochemical sensor. The sensor detection parameters are calibrated based on the ion concentration range of fresh drug cell fluid and are used to characterize the real-time features of cell activity, ion migration status, or metabolic response of fresh drug samples. The original multi-source time series dataset was divided into training, validation, and test sets in a 7:2:1 ratio. The training set was used for model parameter fitting, the validation set was used for model hyperparameter tuning, and the test set was used for final performance verification of the model. The dataset partitioning process adopted stratified sampling to ensure the consistency of the distribution of fresh herb varieties and physiological states in each set.

[0026] This design determines the sampling frequency based on the physiological metabolic rate of the fresh medicinal herbs, setting different ranges for herbaceous and woody fresh herbs. It also clearly defines the data collection methods and uses for each type of data. Different fresh herbs have different physiological characteristics, and setting targeted sampling frequencies can accurately capture their dynamic changes in osmotic pressure, avoiding insufficient or excessive sampling. Spectral response data is determined based on the absorption peaks of the core active ingredients, accurately reflecting changes in active ingredients. Electrochemical response data is calibrated according to the ion concentration in the cell sap, enabling real-time characterization of cell activity and other features. Reasonable division of the dataset and the use of stratified sampling ensure consistent distribution across sets, providing a reliable data foundation for model training, optimization, and validation, thereby improving model accuracy and generalization ability.

[0027] In one embodiment, the multimodal response hysteresis model is used to quantitatively characterize the time shift of osmotic pressure change relative to the peak value change of the active ingredient, the time delay of spectral response relative to the electrochemical activity change, local abrupt changes between different modes, and the correspondence between peak response and steady-state formation. This model serves as the core basis for subsequent asynchronous data dynamic compensation. The time shift of osmotic pressure change relative to the peak value change of the active ingredient is a key characteristic of this model. Calculate using the following formula:

[0028] In the formula, This refers to the local peak moment of the osmotic pressure signal, that is, the sampling moment when the value in the osmotic pressure time series data reaches a local maximum value; This refers to the local peak moment of the spectral response signal, that is, the sampling moment when the intensity of the characteristic peak of the active ingredient in the spectral response data reaches its local maximum value; The regularity indicates that the peak value of the osmotic pressure signal lags behind the peak value of the spectral response signal. A negative value indicates that the peak value of the osmotic pressure signal leads the peak value of the spectral response signal; The time delay of the spectral response relative to the change in electrochemical activity Calculate using the following formula:

[0029] In the formula, This refers to the local peak moment of the electrochemical response signal, that is, the sampling moment when the electrical signal intensity in the electrochemical response data reaches its local maximum value; The canonical form indicates that the peak value of the spectral response signal lags behind the peak value of the electrochemical response signal. A negative value indicates that the peak value of the spectral response signal leads the peak value of the electrochemical response signal; The weighting of time offset and time delay relationships is determined based on the correlation between fresh drug efficacy and the weighting of intermodal temporal relationships directly related to the core efficacy is set to... The weights of the temporal relationships between indirectly related modes are set as follows: ; The multimodal response hysteresis model is constructed using the gradient boosting tree algorithm, and the number of iterations for model training is set to... Next, the learning rate is set to The decision tree depth is set to Layer, minimum number of sample splits set to .

[0030] This design allows the multimodal response lag model to quantitatively characterize the time offset and delay relationship between different modalities, clearly define the basis for weight allocation, and use the gradient boosting tree algorithm to construct and set relevant parameters to accurately quantify the time relationship between each modality. This provides a core basis for dynamic compensation of asynchronous data. The weight allocation is based on the correlation between drug efficacy and highlights the key modal relationships. The gradient boosting tree algorithm can handle complex nonlinear relationships. By reasonably setting parameters such as the number of iterations and the learning rate, the model can achieve a balance between training efficiency and accuracy, better explore the intrinsic connections between multimodal data, and improve the accuracy of subsequent data processing and analysis.

[0031] In one embodiment, dynamic time-shift compensation and window reconstruction are the core correction steps for the asynchronicity of multi-source data. Specifically, they include response extension of the preceding mode, time offset correction of the lagging mode, window removal or compression of locally invalid sampling segments, and reconstruction of the effective interval of multi-modal crossover. The time offset correction amount of the lagging mode is calculated according to the following formula:

[0032] In the formula, For the lag mode, the first The original time of each sampling point, that is, the actual time stamp when the detection device collects the data at that sampling point; The time offset corresponding to this sampling point is quantitatively determined by the multimodal response hysteresis model based on the characteristic stage to which this sampling point belongs. The intermodal time delay value corresponding to this stage is also given. The offset weight is determined based on the characteristic type to which the sampling point belongs, and the offset weight of the peak point is set to [value missing]. The weights for the offsets of the initial response point and the steady-state point are set to... ; The corrected sampling time is the time stamp of the lag mode sampling point that is synchronized with the time stamp of the reference mode. The threshold for determining invalid sampling segments is that the coefficient of variation of the sampling data is greater than 0.3. The coefficient of variation is calculated as the ratio of the standard deviation to the mean of the sampling data. Sampling segments exceeding this threshold are determined to be invalid segments and are removed.

[0033] This design, with its dynamic time-shift compensation and window reconstruction processing, addresses the asynchronicity of multi-source data by clearly defining specific correction steps, calculation methods for lag modal time offset corrections, and thresholds for identifying locally invalid sampling segments. Through operations such as response extension and time offset correction, it effectively solves the problem of asynchronous multi-source data, synchronizing different modal data in time and ensuring data consistency and validity. Determining offset weights based on feature stages and types enables more precise time correction. Ignoring and removing invalid sampling segments using the coefficient of variation removes noisy data, improves data quality, and provides more accurate data support for subsequent feature extraction and model training.

[0034] In one embodiment, the Z-score normalization method is used to normalize the features in the fused feature set. This method can eliminate the influence of differences in the dimensions and orders of magnitude of different features on subsequent modeling, achieving equal-weighted fusion of each feature, and individual features... The normalized calculation formula is:

[0035] In the formula, This is the mean of the feature in the sample set, reflecting the overall level of the feature in the sample set; The standard deviation of this feature in the sample set reflects the degree of dispersion of this feature in the sample set; The normalized eigenvalues ​​are dimensionless and follow a standard normal distribution. The weighting of normalized features is determined based on the correlation coefficient between the feature and the efficacy of the fresh herb. Features with a correlation coefficient greater than [value missing] are assigned weights accordingly. The feature weights are set as follows: The correlation coefficient is in The feature weights between are set as The correlation coefficient is less than The feature weights are set as follows: ; The dimension of the feature vector for fresh herb function detection is consistent with the number of features in the fused feature set. After the vector is constructed, data standardization is performed to ensure that the values ​​of each dimension of the vector are distributed within a certain range. Within the interval, it serves as the input data for model training.

[0036] This design employs Z-score normalization on the fused feature set, clearly defining the normalization calculation formula. Weights are assigned based on the correlation coefficient between features and the efficacy of fresh medicinal herbs, constructing and standardizing a feature vector for fresh medicinal herb function detection. Z-score normalization eliminates differences in the dimensions and orders of magnitude of different features, achieving equal-weighted fusion. This ensures that each feature has equal importance in model training. Weights are assigned based on their correlation with efficacy, highlighting features with a significant impact on efficacy and improving the model's targeting for fresh medicinal herb function detection. Constructing a standardized feature vector as model input helps the model better learn data features, improving model training efficiency and prediction accuracy, and providing a reliable basis for fresh medicinal herb quality assessment.

[0037] In one embodiment, the mean and standard deviation These are all global statistics of a single feature in the fusion feature set, calculated based on all sampled samples of that feature. Their calculation formulas are as follows:

[0038]

[0039] In the formula, The number of samples for a feature is the number of valid sampling points for that feature within the detection period. The threshold for determining a valid sampling point is that the absolute value of the sampled value deviating from the mean does not exceed 2 times the standard deviation. Sampling points exceeding this threshold are determined to be outliers and are removed. For the first feature The sampled value, that is, the first sampled value of this feature collected by the detection device. One set of raw data; using The standard deviation is calculated as an unbiased estimate, which can more accurately reflect the actual discrete characteristics of the features and is suitable for small sample scenarios of fresh drug detection.

[0040] With this design, the mean and standard deviation To integrate the global statistics of individual features in a feature set, the calculation formula, the effective sampling point determination threshold, and the methods used are clarified. The reason for calculating the standard deviation is that, based on global statistics, it can comprehensively reflect the overall level and dispersion of features in the sample set, making normalization more accurate and reasonable, removing outlier sampling points to avoid their interference with statistics, and improving data quality. The standard deviation is calculated as an unbiased estimate, which can more accurately reflect the actual discrete features in the scenario of small sample fresh drug detection. This provides a more reliable data foundation for subsequent feature normalization and model training, and enhances the stability and accuracy of the model.

[0041] In one embodiment, the intermodal hysteresis compensation residual characteristic is a core quantitative indicator for evaluating the effect of asynchronous compensation of multi-source data. It reflects the time synchronization error between the reference mode and the target mode feature points after compensation, and its calculation formula is:

[0042] In the formula, The number of feature points participating in compensation refers to the total number of key feature points selected from the feature sequences of each modal event, including initial response points, local peak points, and steady-state starting points. The weight of each type of feature point is determined based on its detection importance, with the weight of peak points set to [value missing]. The weights of the initial response point and the steady-state initial point are set to... ; For reference mode number The time after feature point correction, that is, the feature point time stamp after response extension or time correction of the reference mode; For the target mode, the first The time of each feature point, i.e. the original feature point time stamp of the target mode without additional correction; For intermodal lag compensation residuals, the smaller the value, the better the temporal synchronization of multi-source data after compensation, and the higher the validity of the detection data; The feature point selection threshold for the calculation is a signal-to-noise ratio (SNR) of the feature point greater than 1. The signal-to-noise ratio is calculated as the ratio of the peak signal value to the root mean square noise value. Feature points below this threshold are not included in the residual calculation.

[0043] This design uses the intermodal hysteresis compensation residual characteristic as the core indicator for evaluating the asynchronous compensation effect of multi-source data. It clarifies the calculation formula, the threshold for selecting feature points involved in the calculation, and the weights of each type of feature point. This indicator can quantitatively reflect the temporal synchronization error of the compensated multi-source data; the smaller the value, the better the synchronization and the higher the data validity. This provides an intuitive standard for evaluating the data compensation effect. High signal-to-noise ratio feature points are selected for calculation to ensure the reliability of the results. Feature point weights are allocated based on detection importance, highlighting the impact of key feature points on residual calculation. This makes the evaluation more in line with actual detection needs, helps optimize data compensation methods, and improves the quality of multi-source data fusion.

[0044] In one embodiment, the ternary coupling model is a nonlinear correlation model based on a deep neural network. It uses the peak response of the active ingredient as the core characterization of enhanced efficacy of the fresh herb, the cell viability stability zone as the key characterization of the function maintenance of the fresh herb sample, and the osmotic pressure change process as the dynamic mapping object of the functional state of the fresh herb. The weight allocation of the three is determined based on the fresh herb quality evaluation indicators, with the peak response of the active ingredient set as... The weight of the cell viability stable region is set to The weight of the osmotic pressure change process is set as The nonlinear correlation among the three factors was used to functionally explain the osmotic pressure state of fresh medicinal materials at different time periods. The specific steps for training the ternary coupling model are as follows: Model initialization: Set the input layer dimension to the dimension of the fresh herb function detection feature vector, and set the hidden layer to... Layers, with the number of neurons in each layer set to... The model has one output layer containing the predicted upper and lower limits of the osmotic pressure functional zone. The ReLU activation function is used, and the output layer uses a linear activation function. The model weights are initialized to follow a normal distribution. Random values; Model training uses the fresh herb function detection feature vectors from the training set as input and the optimal osmotic pressure range of the corresponding fresh herb as the label. Batch gradient descent is employed for training, with the batch size set to [value missing]. The number of training rounds is set to In this round, the loss function used is the mean squared error loss function, and the formula for calculating the loss function is:

[0045] in, This represents the batch sample size. This is the actual label value. These are the model's predicted values; Hyperparameter tuning was performed using a grid search method based on the validation set, with the learning rate search range set to [value missing]. The search range for the weight decay coefficient is set to The dropout probability search range is set to The combination of hyperparameters with the highest prediction accuracy on the validation set is selected as the optimal hyperparameters for the model. Regularization is implemented to prevent overfitting of the model. L2 regularization and dropout regularization are introduced, with the L2 regularization coefficient set to [value missing]. The dropout probability is set to That is, randomly discarding during training. Neurons; Performance validation involves inputting the test set into the trained model and calculating the model's prediction accuracy, root mean square error, and coefficient of determination. The performance threshold for the model is: Prediction accuracy Root mean square error Coefficient of determination If the target is not met, return to adjust the training parameters and retrain; The rate of change of osmotic pressure As a core indicator characterizing the dynamic changes in osmotic pressure, its calculation formula is:

[0046] In the formula, This refers to the change in osmotic pressure within a preset time interval, i.e., the difference between the end value and the start value of the osmotic pressure time series data within that time interval. The duration of this time interval is the time difference between the end and start times of the interval. The threshold for dividing the time interval is determined based on the absolute value of the first derivative of the osmotic pressure change. A first derivative absolute value greater than... When, it is a rapidly changing interval, less than At that time, it was a period of slow change; The sign of the symbol indicates the trend of osmotic pressure change; a positive sign indicates an increase in osmotic pressure, and a negative sign indicates a decrease in osmotic pressure. The absolute value of osmotic pressure indicates how fast or slow the osmotic pressure changes.

[0047] This design, with its ternary coupling model being a nonlinear correlation model based on deep neural networks, clearly defines its construction basis, weight allocation, training steps, hyperparameter tuning, regularization, and performance verification methods. The model is constructed using the peak response of active ingredients, the stable cell viability region, and the osmotic pressure change process as its objects, meeting the core requirements for fresh herb functional testing. Weight allocation is based on quality evaluation indicators, highlighting key factors. Detailed training steps and tuning methods optimize model performance. Regularization prevents overfitting and improves the model's generalization ability. Rigorous performance verification ensures that the model's prediction accuracy, error, and coefficient of determination meet the standards, enabling the model to accurately interpret the osmotic pressure state of fresh herbs and providing a scientific basis for fresh herb quality assessment and osmotic pressure regulation.

[0048] In one embodiment, a multimodal compensation residual below a preset threshold is an important criterion for screening effective osmotic pressure time series segments. This ensures that the selected segments have good synchronization of multi-source data and reliable detection results. The specific criterion is as follows:

[0049] in, The intermodal hysteresis compensation residuals, which are calculated in practice, are obtained by the formula. The residual threshold is preset according to the accuracy requirements of fresh herb detection, and its value selection rule is as follows: for high-precision detection scenarios, such as accurate determination of the efficacy of fresh herbs. Pick In routine testing scenarios, such as grading and screening of fresh medicinal herbs, Pick In rudimentary testing scenarios, such as initial testing during the storage and transportation of fresh medicinal herbs, Pick It can be adaptively adjusted according to the variety of fresh medicinal herbs, the accuracy of the testing equipment, and the actual application scenario; The threshold for determining whether an active ingredient's response has reached or is close to its peak value is when the active ingredient's response value reaches its peak value. The threshold for determining whether electrochemical activity is in a stable state is a coefficient of variation of less than 100% for the electrochemical response data. The threshold for determining whether osmotic pressure changes are in a continuous and interpretable state is that there are no abrupt changes in osmotic pressure data at 3 or more consecutive sampling points. Abrupt changes are defined as the absolute value of the sampled value deviating from the mean of the five adjacent sampling points by more than 3 times the standard deviation. The screening process uses a trained ternary coupling model for segment-by-segment prediction and judgment, with the sliding window length set to [value missing]. There are 1 sampling point, and the window step size is set to 1. Each sampling point is used to traverse the osmotic pressure time series data window by window, and all continuous windows that meet the conditions are selected and merged into a complete individualized functional osmotic pressure range.

[0050] This design clearly defines the multimodal compensation residual judgment conditions, other condition judgment thresholds, and the screening process for effective osmotic pressure time series segments. The multimodal compensation residual being lower than the preset threshold ensures good synchronization of multi-source data and reliable detection results. The judgment thresholds for each condition are set according to different scenarios and indicator characteristics, which can accurately screen out segments that meet the requirements of active ingredient response, electrochemical activity, and osmotic pressure change. The use of a sliding window for segment-by-segment prediction and judgment can comprehensively traverse osmotic pressure time series data, and merge continuous windows that meet the conditions to obtain a complete interval, providing an accurate basis for determining individualized functional osmotic pressure intervals and helping to accurately assess the quality and functional status of fresh drugs.

[0051] In one embodiment, the individualized functional osmotic pressure range is fed back to the osmotic pressure detection or control device in real time to achieve adaptive closed-loop adjustment of the device parameters, ensuring that the osmotic pressure remains within the optimal functional range during the detection, storage, and processing of fresh herbs. The osmotic pressure control parameters of the device are specified in the provided text. The midpoint of the individualized functional osmotic pressure range is calculated using the following formula:

[0052] In the formula, This is the upper limit of the individualized functional osmotic pressure range, i.e., the maximum value among the selected osmotic pressure time series segments; The lower limit of the individualized functional osmotic pressure range is the minimum value in the selected osmotic pressure time series. Using this intermediate value as the control setpoint can reserve sufficient upward and downward fluctuation space for the control of fresh drug osmotic pressure, and adapt to the slight dynamic changes in the physiological state of fresh drug. The error threshold for equipment regulation is determined based on the osmotic pressure tolerance range of the fresh herb variety. The error threshold for regulating fresh herb varieties is set as follows: The threshold for regulating the error of fresh woody medicinal herbs is set to When the actual measured osmotic pressure exceeds When the error threshold is reached, the device automatically triggers parameter adjustment; The quality grade of fresh herbs is determined based on the degree of matching between the individualized functional osmotic pressure range and the standard osmotic pressure range of fresh herbs. A matching degree greater than [missing value] indicates a quality grade of [missing value]. It is a first-class product, and the matching degree is... The products in question are grade II, with a matching degree of less than [missing information]. For grade three products, the matching degree is calculated as the ratio of the overlapping area of ​​the two intervals to the area of ​​the standard interval; The model supports online updates after training. This applies when the number of newly added fresh drug detection samples reaches the original training set sample size. When the new samples are 10 or higher, they are processed and added to the original training set, and the model is retrained for fine-tuning. The learning rate for fine-tuning is set to 1 / 3 of the original training learning rate. The number of training rounds is set to 1 / 3 of the original number of training rounds. This ensures the model's adaptability to different varieties of fresh herbs and their physiological states.

[0053] This design feeds back individualized functional osmotic pressure ranges to the equipment for adaptive closed-loop adjustment. It clarifies the calculation methods for equipment control parameters, error threshold settings, quality grade determination criteria, and online model update methods. Using the midpoint of the range as the control setpoint reserves space for osmotic pressure regulation, adapting to the dynamic physiological changes of fresh herbs. Error thresholds are set according to the variety to ensure control accuracy. The quality grade is determined based on the matching degree between the range and the standard range, intuitively reflecting the quality of fresh herbs. The online model update mechanism allows the model to learn from newly added sample data in a timely manner. By reasonably setting and fine-tuning training parameters, the model's adaptability to different fresh herb varieties and physiological states is ensured, continuously improving the model's accuracy and reliability, and providing strong support for fresh herb testing, storage, transportation, and processing.

[0054] Example 2: Please see Figure 4 A method for detecting the osmotic pressure of fresh herbs based on multi-data fusion, applied to the aforementioned detection system for fresh herbs based on multi-data fusion, includes the following steps: S1. Collect osmotic pressure time-series data, spectral response data and electrochemical response data of the fresh herb sample to be tested within the preset detection period, and form an original multi-source time-series dataset with sampling time markers; S2. Extract key response events from various types of raw detection data and form event feature sequences for the corresponding modalities; S3. Establish a multimodal response hysteresis model based on the event characteristic sequences of each mode; S4. Based on the multimodal response hysteresis relationship model, perform dynamic time shift compensation and window reconstruction processing on the original multi-source time series data to form a unified and effective response sequence; S5. Construct a feature vector for fresh herb function detection based on a unified and effective response sequence; S6. Based on the feature vector of fresh drug function detection, establish a correlation model between the functional state of fresh drug and the osmotic pressure change process, and screen the osmotic pressure time series segments that meet the preset drug efficacy response conditions, stability conditions, residual threshold conditions and continuity conditions, and determine the osmotic pressure time series segments as the individualized functional osmotic pressure intervals corresponding to the fresh drug samples. S7. Based on individualized functional osmotic pressure range, output one or more of the following: fresh drug quality grade, activity status evaluation results, storage and transportation osmotic pressure control range, pre-processing range, and detection abnormality early warning results.

[0055] Those skilled in the art will understand that all or part of the steps in the methods of the above embodiments can be implemented by a program instructing related hardware. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Moreover, this application can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0056] The above embodiments provide a detailed description of the present invention. Specific examples have been used to illustrate the principles and implementation methods of the present invention. The descriptions of the above embodiments are only for the purpose of helping to understand the method and core ideas of the present invention. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of the present invention. Therefore, the content of this specification should not be construed as a limitation of the present invention.

Claims

1. A fresh herb osmotic pressure testing system based on multi-data fusion, characterized in that, The system includes: Multi-source acquisition unit, event extraction unit, hysteresis modeling unit, dynamic compensation unit, feature fusion unit, interval back-inference unit, and result output unit; The multi-source acquisition unit is used to acquire osmotic pressure time-series data, spectral response data and electrochemical response data of the fresh herb sample to be tested within a preset detection period, and form an original multi-source time-series dataset with sampling time markers. The event extraction unit is used to extract key response events from various types of raw detection data and form event feature sequences of corresponding modalities; The hysteresis modeling unit is used to establish a multimodal response hysteresis relationship model based on the event feature sequence of each modality; The dynamic compensation unit is used to perform dynamic time-shift compensation and window reconstruction processing on the original multi-source time-series data according to the multimodal response hysteresis relationship model to form a unified and effective response sequence; The feature fusion unit is used to construct a fresh drug function detection feature vector based on the unified effective response sequence; The interval back-inference unit is used to establish a correlation model between the functional state of fresh medicine and the osmotic pressure change process based on the fresh medicine functional detection feature vector, and to screen the osmotic pressure time series segments that meet the preset drug efficacy response conditions, stability conditions, residual threshold conditions and continuity conditions, and to determine the osmotic pressure time series segments as the individualized functional osmotic pressure intervals corresponding to the fresh medicine samples. The result output unit is used to output one or more of the following based on the individualized functional osmotic pressure range: fresh drug quality grade, activity status evaluation result, storage and transportation osmotic pressure control range, pre-processing range, and detection anomaly warning result.

2. The fresh herb osmotic pressure testing system based on multi-data fusion according to claim 1, characterized in that, The multi-source acquisition unit includes: Osmotic pressure detection subunit, spectral detection subunit, and electrochemical detection subunit; The osmotic pressure detection subunit is used to continuously collect the osmotic pressure value sequence of fresh herb samples at a sampling frequency adapted to the physiological metabolic rate of the fresh herb variety. The spectral detection subunit is used to collect characteristic spectral signal sequences around the characteristic absorption range corresponding to the core active ingredient of fresh medicine. The electrochemical detection subunit is used to collect electrochemical response sequences that characterize changes in cell fluid ion migration state and cell activity. The multi-source acquisition unit is also used to perform unified time stamping, time alignment preprocessing, outlier screening, data storage, and sample identity association on the three types of detection data, so as to ensure that subsequent time series correlation analysis has a unified time benchmark, consistent data format, and traceable sample source.

3. The fresh herb osmotic pressure testing system based on multi-data fusion according to claim 1, characterized in that, The event extraction unit includes the following steps: The initial response time, local peak time, peak intensity, rise rate, fall rate, steady-state duration, and fluctuation width are extracted from the osmotic pressure time series data, spectral response data, and electrochemical response data, respectively, and event-level feature labels are generated according to the detection mode, event type, and stage of change. The event extraction unit is also used to perform validity screening based on the signal-to-noise ratio, sampling stability and consistency of changes of adjacent sampling points of candidate event points, so as to eliminate abnormal event points and invalid event points, and organize the remaining event points into event feature sequences and staged event segments in chronological order.

4. The fresh herb osmotic pressure testing system based on multi-data fusion according to claim 2, characterized in that, The hysteresis modeling unit includes the following steps: The time offset of the osmotic pressure signal relative to the spectral response signal and the time delay of the spectral response signal relative to the electrochemical response signal were calculated respectively. The corresponding time sequence relationship was weighted according to the degree of correlation between different modes and the efficacy characterization of fresh drug, and a multimodal response hysteresis relationship model was obtained to characterize the sequential relationship between local mutation, peak response and steady state formation. The lag modeling unit is also used to distinguish between direct pharmacodynamic correlation modal relationships and indirect pharmacodynamic correlation modal relationships, so that the multimodal response lag relationship model can simultaneously reflect the sequence, lag degree and stage coupling relationship of key response events.

5. The fresh herb osmotic pressure testing system based on multi-data fusion according to claim 4, characterized in that, The dynamic compensation unit includes the following steps: To address the asynchronous response characteristics of multimodal detection data, response extension processing is performed on the preceding modes, time offset correction processing is performed on the lagging modes, and window removal, window compression, and cross-valid interval reconstruction processing are performed on locally invalid sampling segments. The dynamic compensation unit is also used to allocate time offset weights according to the feature stage and feature type to which the sampling points belong, and to determine the window reconstruction range according to the fluctuation degree of the sampling segment, the duration of stability and the density of events, so that the compensated modal feature points can maintain temporal synchronization on a unified time axis and form a continuous response segment that can be used for cross-modal joint analysis.

6. The fresh herb osmotic pressure testing system based on multi-data fusion according to claim 1, characterized in that, The feature fusion unit includes the following steps: Normalization, correlation evaluation, and coupling mapping are performed on various features in the unified and effective response sequence to obtain the fresh drug function detection feature vector; The fresh herb function detection feature vector includes osmotic pressure change rate, spectral peak response intensity, peak occurrence time of active ingredient, electrochemical activity stability range, peak synchronization between modes, and residual hysteresis compensation between modes. The feature fusion unit is also used to assign feature weights based on the degree of correlation between each feature and the efficacy characterization of fresh medicine, and to uniformly convert features of different dimensions into standardized vectors that can be used as model inputs, so as to enhance the comprehensive characterization ability of the functional state of fresh medicine, cross-sample comparability and interval back-inference stability.

7. The fresh herb osmotic pressure testing system based on multi-data fusion according to claim 6, characterized in that, The interval reverse calculation unit includes: Model construction sub-units and interval solution sub-units; The model building subunit is used to construct a ternary coupling relationship between the rate of change of osmotic pressure, the peak response of active ingredients, and the stable cell viability region based on the fresh drug function detection feature vector, and to form an association model for predicting the functional osmotic pressure range using sample training, parameter tuning, and performance verification, wherein the sample data is grouped according to training data, verification data, and test data; The interval solution subunit is used to traverse the osmotic pressure time series data in a sliding window manner, select continuous windows that meet the active ingredient response conditions, electrochemical activity stability conditions, intermodal compensation residual threshold conditions, and osmotic pressure change continuity conditions, and merge adjacent effective windows into a complete individualized functional osmotic pressure interval, which is used to characterize the target functional osmotic pressure range of fresh drug under the current physiological state.

8. The fresh herb osmotic pressure testing system based on multi-data fusion according to claim 5, characterized in that: The intermodal hysteresis compensation residual is used to characterize the time synchronization error of key feature points between the reference mode and the target mode after compensation; The key feature points include the initial response point, local peak points, and steady-state starting points; The feature fusion unit or the interval back-inference unit is also used to weight the time synchronization error according to the detection importance corresponding to different key feature points to obtain the compensation effect evaluation result, and to determine whether to include the candidate key feature point in the residual calculation based on the signal credibility, peak significance and stage integrity, so as to use the compensation effect evaluation result as a constraint condition for effective segment selection.

9. The fresh herb osmotic pressure testing system based on multi-data fusion according to claim 7, characterized in that, The result output unit includes the following steps: The individualized functional osmotic pressure range is fed back to the osmotic pressure detection device or osmotic pressure control device to determine the osmotic pressure control setting value based on the upper and lower limits of the individualized functional osmotic pressure range, and to trigger parameter adjustment when the actual detected osmotic pressure deviates from the allowable range corresponding to the osmotic pressure control setting value. The result output unit is also used to generate fresh drug quality grade results, storage and transportation control suggestions, pre-processing suggestions, and detection and early warning results and result call interface information for sampling abnormalities, compensation abnormalities, interval abnormalities or equipment control deviations based on the matching relationship between the individualized functional osmotic pressure range and the standard osmotic pressure range.

10. A method for detecting the osmotic pressure of fresh medicinal herbs based on multi-data fusion, characterized in that, The method, applied to the fresh herb osmotic pressure testing system based on multi-data fusion as described in any one of claims 1-9, comprises the following steps: S1. Collect osmotic pressure time-series data, spectral response data and electrochemical response data of the fresh herb sample to be tested within the preset detection period, and form an original multi-source time-series dataset with sampling time markers; S2. Extract key response events from various types of raw detection data and form event feature sequences for the corresponding modalities; S3. Establish a multimodal response hysteresis model based on the event characteristic sequences of each mode; S4. Based on the multimodal response hysteresis relationship model, perform dynamic time shift compensation and window reconstruction processing on the original multi-source time series data to form a unified and effective response sequence; S5. Construct a fresh herb function detection feature vector based on the unified and effective response sequence; S6. Based on the fresh drug function detection feature vector, establish a correlation model between the fresh drug function state and the osmotic pressure change process, and screen the osmotic pressure time series segments that meet the preset drug efficacy response conditions, stability conditions, residual threshold conditions and continuity conditions, and determine the osmotic pressure time series segments as the individualized functional osmotic pressure intervals corresponding to the fresh drug samples. S7. Based on the individualized functional osmotic pressure range, output one or more of the following: fresh drug quality grade, activity status evaluation result, storage and transportation osmotic pressure control range, pre-processing range, and detection abnormality early warning result.

Images