A Chinese herbal medicine extraction monitoring system and method based on image recognition

Abstract

The present invention relates to the technical field of object recognition, and specifically to a monitoring system and method for Chinese medicinal materials extraction based on image recognition. The system includes: a dynamic data acquisition module, a feature extraction and comparison module, a dynamic feature optimization module, a state prediction module, and a process monitoring module. In the present invention, through comprehensive analysis of data and images, the state changes in the process of Chinese medicinal materials extraction are dynamically monitored, and the multi-dimensional grayscale, texture features and pressure gradients are combined to achieve accurate identification of the extraction state. Through the analysis of the fluctuation trend of key parameters and the optimization of the correlation between features, abnormal nodes are quickly identified and the extraction parameters are adjusted to improve process stability and product quality consistency. Dynamically analyze the data characteristics at different stages, strengthen the ability to accurately control the extraction process, significantly improve efficiency and resource utilization, and reduce abnormal risks.

Description

Technical Field

[0001] The present invention relates to the technical field of object recognition, and in particular to a Chinese medicinal material extraction monitoring system and method based on image recognition. Background Art

[0002] The field of object recognition technology includes the classification, positioning, tracking and other processing of target objects by analyzing and detecting objects in images or videos. The core content of this technical field is to use image processing and computer vision technology to extract key features in images, and to identify and analyze targets through pattern matching, deep learning and other methods. The overall technical field includes but is not limited to image preprocessing, feature extraction, object detection and recognition, three-dimensional object analysis and other aspects, and is widely used in industrial inspection, autonomous driving, medical image analysis, intelligent monitoring and other scenarios.

[0003] Among them, the Chinese herbal medicine extraction monitoring system based on image recognition refers to the use of object recognition technology to monitor and manage the key links in the Chinese herbal medicine extraction process. The system focuses on technical matters such as material identification and extraction status judgment in the Chinese herbal medicine extraction process. It obtains real-time image information of materials in the extraction process through image acquisition equipment, uses feature extraction and pattern matching technology to identify different types of Chinese herbal medicines, and analyzes key extraction parameters in the image through deep learning models to achieve accurate monitoring and real-time judgment of the Chinese herbal medicine extraction process. The system completes the identification of key materials and accurate distinction of extraction status based on image segmentation, edge detection and multi-dimensional feature matching methods to ensure that the monitoring and management of the entire process meet the established goals.

[0004] Existing technologies only focus on a single data dimension, ignoring the correlation between physical parameters and image features. They are unable to cope with sudden changes in state during the extraction process, resulting in delayed dynamic adjustment and insufficient recognition accuracy. They fail to deeply explore the complex correlation between features, limiting the accurate analysis of multi-stage extraction processes. There is a lack of a rapid identification mechanism for abnormal nodes, and the process control efficiency is low, which easily leads to problems such as energy consumption and resource waste in the extraction process. Summary of the invention

[0005] The purpose of the present invention is to solve the shortcomings of the prior art and to propose a Chinese medicinal material extraction monitoring system and method based on image recognition.

[0006] In order to achieve the above-mentioned purpose, the present invention adopts the following technical scheme: A Chinese herbal medicine extraction monitoring system based on image recognition comprises:

[0007] The dynamic data acquisition module collects image data and temperature sensor and pressure sensor data based on the operating status of the Chinese medicinal material extraction and distillation device, describes the morphological characteristics of the Chinese medicinal materials by grayscale changes and texture differences, and analyzes the relationship between the morphological characteristics of the Chinese medicinal materials and the temperature distribution and pressure changes of the distillation device in combination with the sensor data, and establishes a phased data matrix;

[0008] The feature extraction and comparison module extracts the grayscale changes and texture differences of the morphological features of the Chinese medicinal materials in the image based on the stage data matrix, analyzes the pressure gradient and temperature distribution of the distillation device in the temperature sensor and pressure sensor data, matches the data features of the same stage, and generates the feature distribution value of the extraction stage;

[0009] The dynamic feature optimization module calculates the amplitude of grayscale change and pressure gradient change in multiple stages based on the feature distribution value in the extraction stage, selects the feature items with the most concentrated changes, analyzes the correlation of multiple features, and generates dynamic key feature values;

[0010] The state prediction module analyzes the image texture change and pressure gradient trend during the extraction of Chinese medicinal materials based on the dynamic key feature values, locates the key change nodes, analyzes the characteristic fluctuation range of the key change nodes, and generates key state prediction values;

[0011] The process monitoring module analyzes the difference between the current distillation state and the target state based on the key state prediction value, calculates the distillation temperature and operating time parameters that need to be adjusted, and generates the extraction process monitoring parameters.

[0012] The phased data matrix specifically includes the reorganization analysis and pressure change of the image morphological features; the extraction phase feature distribution value includes the grayscale change, texture difference, pressure gradient, and temperature distribution of the morphological features of the Chinese medicinal materials in the image; the dynamic key feature value specifically includes the amplitude of grayscale change, the amplitude of pressure gradient change, feature item concentration, and multi-feature correlation; the key state prediction value includes image texture change, pressure gradient trend, and feature fluctuation range of key change nodes; the extraction process monitoring parameters specifically include distillation temperature adjustment and running time parameters.

[0013] As a further solution of the present invention, the step of obtaining the phased data matrix is ​​specifically as follows:

[0014] Collect the temperature sensor and pressure sensor data of the running status of the Chinese herbal medicine extraction and distillation device, call the sensor interface of the device to collect temperature and pressure signals in real time, record the temperature change and pressure fluctuation data in the collected time series in time series segments, record the mean and fluctuation range of the segmented feature data, organize them into phased data sets, and generate temperature and pressure data pairs;

[0015] Calling the segmented records of the temperature and pressure data pairs, analyzing the change characteristics of temperature and pressure item by item through the segmented records of the time series, calculating the mean and standard deviation of each segmented data, setting the relative ratio of the mean to the fluctuation range as a screening basis, screening and marking the data pairs whose relative ratio exceeds the threshold, and generating key stage data records;

[0016] The temperature and pressure distribution characteristics corresponding to the key stage data records are reorganized into a matrix, and the weighted matrix of the key stage data is calculated using the formula:

[0017] ;

[0018] Normalize the matrix, calculate the multi-region data ratio, and generate a phased data matrix;

[0019] in, Represents the weighted value of the corresponding partition in the stage data matrix, Represents the temperature value, Represents the pressure value, Represents the fluctuation weight of key stage data, Represents the total number of key stage data.

[0020] As a further solution of the present invention, the step of obtaining the characteristic distribution value in the extraction stage is specifically as follows:

[0021] Utilizing the phased data matrix, extracting grayscale changes and texture differences in the image data of Chinese medicinal materials, calling the grayscale histogram to calculate the distribution characteristics of pixel values ​​in the region, using the local change gradient and direction distribution of the image to perform texture feature analysis, and obtaining image feature data;

[0022] Combined with the data records of temperature and pressure sensors, the time series data is called to extract the change rate of pressure and temperature values, the change amplitude of the pressure gradient in each stage is calculated based on the stage time nodes, and the mean and standard deviation of the temperature distribution in multiple stages are analyzed to generate pressure and temperature characteristic data;

[0023] The image feature data is correlated with the pressure and temperature feature data, and a phase feature distribution relationship is established through a feature matching algorithm, using the formula:

[0024] ;

[0025] Calculate the feature distribution value of each stage and generate the feature distribution value of the extraction stage;

[0026] in, represents the characteristic distribution value of the i-th stage, is the image feature value of the kth region, and its absolute value indicates the change of texture or grayscale. is the maximum value of the image eigenvalues, , are the characteristic values ​​corresponding to pressure and temperature, respectively, , is the maximum recorded value of the characteristic values ​​of pressure and temperature, It is the coefficient for adjusting the weight of image feature data, which is used to balance the influence of image data in each region in the calculation of total feature distribution value. It is the coefficient for adjusting the weight of pressure and temperature sensor data, which helps determine the criticality of pressure and temperature data in each stage. Represents the number of regions.

[0027] As a further solution of the present invention, the step of obtaining the dynamic key feature value is specifically:

[0028] Based on the extracted stage characteristic distribution value, grayscale change and pressure gradient change records in the stage data are called, grayscale change values ​​and pressure gradient change amplitudes are calculated, and the difference in change characteristics of each stage is analyzed to generate grayscale and pressure change amplitude data;

[0029] Performing statistical analysis on the grayscale and pressure change amplitude data, calling the Gaussian mixture model to detect the distribution characteristics of the data change amplitude, screening and marking the stages with concentrated and significant change amplitudes, and organizing them into screened key stage data;

[0030] The multi-feature correlation analysis is performed on the key stage data after the screening, and the correlation between the feature values ​​between the stages is calculated by matching the difference between the grayscale change value and the pressure gradient value, using the formula:

[0031] ;

[0032] Generate dynamic key characteristic values;

[0033] in, represents the dynamic key characteristic value, represents the grayscale change value of the i-th stage, represents the pressure gradient value of the i-th stage, Represents the standard deviation of grayscale change value, represents the standard deviation of the pressure gradient values, Indicates the number of key stages.

[0034] As a further solution of the present invention, the step of obtaining the key state prediction value is specifically as follows:

[0035] Based on the dynamic key feature value, the image texture change and pressure gradient data in the time series are called, the segmented trend of the time series characteristic curve is extracted, the change amplitude of the key node is calculated, and the change feature is segmented and marked to generate the key change node data;

[0036] Perform characteristic fluctuation range analysis on the key change node data, call the adjacent data before and after the key node, calculate the interval fluctuation range of the characteristic value, analyze the change rate and amplitude of the node data, quantify the characteristics of each node by building a fluctuation range model, and generate node fluctuation characteristic data;

[0037] Combined with the node fluctuation characteristic data, the model optimization and numerical calculation are performed by adjusting the characteristic node weight distribution and trend influencing factors, using the formula:

[0038] ;

[0039] Generate key status prediction values;

[0040] in, represents the key state prediction value, represents the characteristic fluctuation value of the i-th node, is the weight factor of the corresponding node, which is used to describe the criticality of the node. is a normalization parameter used to control the dynamic range of the predicted values, Indicates the number of nodes.

[0041] As a further solution of the present invention, the steps of obtaining the monitoring parameters of the extraction process are specifically as follows:

[0042] Based on the key state prediction value, the real-time data monitoring system is called to obtain the temperature and time data of the current distillation process, the current state is compared with the target state, the degree of deviation of the temperature and time in the current distillation state is analyzed, and the current state data is generated by difference calculation;

[0043] Calculate the adjustment amount of temperature and time parameters using the current state data, analyze the optimal range of temperature and time in the target state, extract the temperature difference and time difference that need to be adjusted by performing difference calculation between the current data and the target parameter range, and generate adjustment parameter data;

[0044] Combined with the adjustment parameter data, an extraction process monitoring model is designed to adjust the temperature and time through a parameterized dynamic adjustment mechanism, using the formula:

[0045] ;

[0046] Calculate the extraction process monitoring parameters;

[0047] in, represents the monitoring parameters of the extraction process, Indicates the difference between the current temperature and the target temperature. Indicates the difference between the current running time and the target time. , are the target temperature and target time respectively, Represents the adjustment factor, which is used to balance the effects of temperature and time on the monitoring parameters.

[0048] A method for monitoring the extraction of Chinese medicinal materials based on image recognition, wherein the method is performed based on the above-mentioned Chinese medicinal materials extraction monitoring system based on image recognition, and comprises the following steps:

[0049] S1: Based on the operating status of the Chinese herbal medicine extraction and distillation device, the data sets of the temperature sensor and the pressure sensor are called, and the morphological features of the Chinese herbal medicine in the image data collected during the distillation process are reorganized and processed, and the morphological features of the Chinese herbal medicine, temperature distribution and pressure change data are integrated to generate a phased data matrix;

[0050] S2: Based on the phased data matrix, the grayscale difference in the image data is calculated, the texture variation range is analyzed, the distribution values ​​in the pressure and temperature data are integrated, the data items are matched and the difference is compared, and the feature distribution value of the extraction phase is generated;

[0051] S3: Based on the feature distribution value in the extraction stage, multi-stage amplitude calculation is performed on the grayscale change and the pressure gradient change, characteristic items in the change concentration are selected, and multi-parameter analysis and calculation are performed on the selected features to obtain dynamic key characteristic values;

[0052] S4: Based on the dynamic key feature values, analyze the change amplitude and pressure gradient trend of the image texture during the extraction of traditional Chinese medicine, locate the key change nodes, analyze the node feature fluctuation range, and generate the key state prediction value;

[0053] S5: Based on the key state prediction value, the deviation between the current distillation state and the target state is analyzed, and the extraction process monitoring parameters are obtained in combination with the temperature control and time parameters that need to be adjusted.

[0054] Compared with the prior art, the advantages and positive effects of the present invention are:

[0055] In the present invention, the state changes during the extraction of Chinese medicinal materials are dynamically monitored through comprehensive analysis of data and images, and the extraction state is accurately identified by combining multi-dimensional grayscale, texture features and pressure gradients. By analyzing the fluctuation trend of key parameters and optimizing the correlation between features, abnormal nodes can be quickly identified and extraction parameters can be adjusted to improve process stability and product quality consistency. Dynamic analysis of data characteristics at different stages can enhance the ability to accurately control the extraction process, significantly improve efficiency and resource utilization, and reduce abnormal risks. BRIEF DESCRIPTION OF THE DRAWINGS

[0056] Figure 1 is a system flow chart of the present invention;

[0057] Figure 2 It is a flow chart of the steps of obtaining the phased data matrix of the present invention;

[0058] Figure 3 This is a flow chart of the steps for obtaining characteristic distribution values ​​in the extraction stage of the present invention;

[0059] Figure 4 A flowchart of the steps for obtaining the dynamic key characteristic value of the present invention;

[0060] Figure 5 A flowchart of the steps for obtaining the key state prediction value of the present invention;

[0061] Figure 6 The figure is a flow chart of the steps for obtaining monitoring parameters of the extraction process of the present invention. DETAILED DESCRIPTION

[0062] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0063] In the description of the present invention, it should be understood that the terms "length", "width", "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside", "outside" and the like indicate positions or positional relationships based on the positions or positional relationships shown in the drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as limiting the present invention. In addition, in the description of the present invention, "multiple" means two or more, unless otherwise clearly and specifically defined.

[0064] Example 1: Please refer to Figure 1 The present invention provides a technical solution: a Chinese herbal medicine extraction monitoring system based on image recognition comprises:

[0065] The dynamic data acquisition module collects image data and temperature sensor and pressure sensor data based on the operating status of the Chinese medicinal material extraction and distillation device, describes the morphological characteristics of the Chinese medicinal materials by grayscale changes and texture differences, and analyzes the relationship between the morphological characteristics of the Chinese medicinal materials and the temperature distribution and pressure changes of the distillation device in combination with the sensor data, and establishes a phased data matrix;

[0066] The feature extraction and comparison module extracts the grayscale changes and texture differences of the morphological features of Chinese medicinal materials in the image based on the stage data matrix, analyzes the pressure gradient and temperature distribution of the distillation device in the temperature sensor and pressure sensor data, matches the data features of the same stage, and generates the feature distribution value of the extraction stage;

[0067] The dynamic feature optimization module calculates the magnitude of grayscale changes and pressure gradient changes in multiple stages based on the feature distribution values ​​in the extraction stage, selects the feature items with the most concentrated changes, analyzes the correlation of multiple features, and generates dynamic key feature values;

[0068] The state prediction module analyzes the image texture changes and pressure gradient trends during the extraction of Chinese medicinal materials based on dynamic key feature values, locates key change nodes, analyzes the characteristic fluctuation range of key change nodes, and generates key state prediction values;

[0069] The process monitoring module analyzes the difference between the current distillation state and the target state based on the key state prediction value, calculates the distillation temperature and operating time parameters that need to be adjusted, and generates the extraction process monitoring parameters.

[0070] The phased data matrix specifically includes the reorganization analysis and pressure changes of the image morphological features. The feature distribution values ​​of the extraction phase include the grayscale changes, texture differences, pressure gradients, and temperature distributions of the morphological features of the Chinese medicinal materials in the image. The dynamic key feature values ​​specifically refer to the amplitude of grayscale changes, the amplitude of pressure gradient changes, the concentration of feature items, and the correlation of multiple features. The key state prediction values ​​include image texture changes, pressure gradient trends, and the feature fluctuation range of key change nodes. The monitoring parameters of the extraction process specifically include distillation temperature adjustment and running time parameters.

[0071] See also Figure 2 , the specific steps for obtaining the phased data matrix are:

[0072] Collect the temperature sensor and pressure sensor data of the running status of the Chinese herbal medicine extraction and distillation device, call the sensor interface of the device to collect temperature and pressure signals in real time, record the temperature change and pressure fluctuation data in the collected time series in time series segments, record the mean and fluctuation range of the segmented feature data, organize them into phased data sets, and generate temperature and pressure data pairs;

[0073] The temperature and pressure signals are collected in real time, and the time series data of the current operating status is obtained through the sensor interface call. According to the equipment collection cycle, the data in the time series are segmented for processing, and the maximum, minimum, mean and fluctuation range of the temperature and pressure signals of each segment are calculated respectively. The changes in the data in the segment are characterized and analyzed in combination with these statistics. During the analysis process, it is necessary to determine whether each segment data is abnormal based on a specific range threshold, and record the location and distribution frequency of the abnormal feature points. In this way, the feature points of the key stages of the distillation unit operation are marked, and the statistical characteristics of the temperature and pressure in the segment are accumulated segment by segment to generate the summary feature data of each stage. These data are combined and processed to form a phased data set to generate temperature and pressure data pairs.

[0074] Call the segmented records of temperature and pressure data pairs, analyze the change characteristics of temperature and pressure item by item through the segmented records of the time series, calculate the mean and standard deviation of each segmented data, set the relative ratio of the mean to the fluctuation range as the screening basis, screen and mark the data pairs whose relative ratio exceeds the threshold, and generate key stage data records;

[0075] Mathematical operations are performed on the temperature and pressure characteristic data in each segment, including calculating statistics such as mean, variance and range of variation. The characteristics in the segmented data are judged in combination with preset thresholds to distinguish data segments with obvious fluctuations. For segments where the ratio of the temperature and pressure mean to the fluctuation range is greater than the preset ratio threshold, the data characteristics within the segment are used to further verify the degree of abnormality, mark it as an abnormal segment, and record its time position and affected interval. The key stage data are screened and sorted in combination with the characteristics within all segments to generate key stage data records.

[0076] The corresponding temperature and pressure distribution characteristics in the key stage data records are reorganized into a matrix, and the weighted matrix of the key stage data is calculated using the formula:

[0077] ;

[0078] Normalize the matrix, calculate the multi-region data ratio, and generate a phased data matrix;

[0079] in, Represents the weighted value of the corresponding partition in the stage data matrix, Represents the temperature value, Represents the pressure value, Represents the fluctuation weight of key stage data, Represents the total number of key stage data.

[0080] formula:

[0081] ;

[0082] The benefit of the formula is that it combines the temperature and pressure data of the key stages with weight parameters for weighted calculation and adds normalization processing to ensure that the data features of different stages can be accurately integrated according to their importance and reflect the relative relationship between the stages.

[0083] Detailed explanation of the formula and the process of formula calculation and derivation:

[0084] Represents the weighted value of the corresponding partition in the stage data matrix, Represents the temperature value, which is obtained by collecting the real-time temperature sensor and calculating the segment mean. Represents the pressure value, which is obtained by obtaining real-time pressure data from the pressure sensor and calculating the segmented mean. The fluctuation weight representing the key stage data is calculated by the ratio of the fluctuation range to the standard fluctuation range. Represents the total number of data in the key stage, obtained by counting the number of data markers in the segment.

[0085] Assign values ​​in the formula:

[0086] Assume that the average temperature in a certain stage is The mean pressure is , volatility weight .

[0087] First calculate the numerator:

[0088] ;

[0089] Calculate the denominator again:

[0090] ;

[0091] Calculate the weighted value:

[0092] ;

[0093] The results show that the comprehensive weighted value of this stage is 5437.59, which reflects the comprehensive characteristics of the device operating status after combining the temperature and pressure data with the weights, and provides a basis for subsequent matrix normalization processing and distribution ratio calculation.

[0094] See also Figure 3 , the steps for obtaining the feature distribution value in the extraction stage are as follows:

[0095] The grayscale changes and texture differences in the image data of Chinese medicinal materials are extracted by using the phased data matrix, the distribution characteristics of the pixel values ​​in the region are calculated by using the grayscale histogram, and the texture feature analysis is performed using the local change gradient and direction distribution of the image to obtain the image feature data;

[0096] The grayscale histogram is called to calculate the pixel value distribution of each partition in the image one by one, and the grayscale mean, grayscale variance and gradient change rate of each area are recorded. The texture feature quantization tool is used to count the texture features of each area, analyze the local gradient direction distribution, calculate the contrast and uniformity of the texture features, match and integrate the grayscale histogram results with the texture feature results, and perform unified normalization processing. The normalization step includes standardizing the grayscale data of each area based on the maximum grayscale value in the grayscale histogram of the entire image, adjusting the weight ratio of grayscale data and texture features using the standardized results, and combining the results of normalized grayscale and texture features with specific coefficients to complete the extraction of Chinese medicinal materials image feature data and obtain image feature data.

[0097] Combined with the data records of temperature and pressure sensors, the time series data is called to extract the change rate of pressure and temperature values, the change amplitude of the pressure gradient in each stage is calculated based on the stage time nodes, and the mean and standard deviation of the temperature distribution in multiple stages are analyzed to generate pressure and temperature characteristic data;

[0098] The original data output by the sensor is called, and the pressure and temperature values ​​of each stage in the time series are analyzed in sections. The pressure gradient value of each stage is obtained by calculating the change rate of the pressure value in each stage. The temperature value of each stage is called, and the mean of the temperature distribution is obtained by using the section-by-section mean calculation method. The distribution variance of the temperature value of each stage is calculated by calling the section-by-section variance formula. The pressure gradient value and the temperature distribution value are plotted as a stage distribution curve with time as the horizontal axis, and the stage change pattern is analyzed. Normalization operation is performed according to the distribution characteristics of the curve. During normalization, the data is normalized based on the maximum value of the pressure gradient and temperature distribution of the entire stage to generate pressure and temperature characteristic data.

[0099] The image feature data is correlated with the pressure and temperature feature data, and the stage feature distribution relationship is established through the feature matching algorithm. The formula is:

[0100] ;

[0101] Calculate the feature distribution value of each stage and generate the feature distribution value of the extraction stage;

[0102] in, represents the characteristic distribution value of the i-th stage, is the image feature value of the kth region, and its absolute value indicates the change of texture or grayscale. is the maximum value of the image eigenvalues, , are the characteristic values ​​corresponding to pressure and temperature, respectively, , is the maximum recorded value of the characteristic values ​​of pressure and temperature, It is the coefficient for adjusting the weight of image feature data, which is used to balance the influence of image data in each region in the calculation of total feature distribution value. It is the coefficient for adjusting the weight of pressure and temperature sensor data, which helps determine the criticality of pressure and temperature data in each stage. Represents the number of regions.

[0103] formula:

[0104] ;

[0105] The benefit of the formula is that it optimizes the accuracy and comprehensiveness of the stage-by-stage feature distribution value by combining the feature weights of image grayscale and texture and the normalized ratio of pressure and temperature of sensor data.

[0106] Detailed explanation of the formula and the process of formula calculation and derivation:

[0107] Assume the number of regions is 5, where the image feature value , the maximum value of image features , pressure value , maximum pressure , temperature value , maximum temperature , adjustment coefficient and .

[0108] Calculate the normalized image eigenvalues:

[0109] ;

[0110] Calculate the normalized ratio of pressure and temperature:

[0111] ;

[0112] Substitute the normalized result into the formula to calculate the feature distribution value at each stage:

[0113] ;

[0114] Compute expansion:

[0115] ;

[0116] ;

[0117] ;

[0118] The results show that the comprehensive characteristics of the feature distribution value at each stage can be quantified as 2.081, which can be used for feature matching and data analysis in subsequent stages to ensure the uniformity and relevance of the distribution value.

[0119] See also Figure 4 , the specific steps for obtaining dynamic key feature values ​​are:

[0120] Based on the extracted stage characteristic distribution value, the grayscale change and pressure gradient change records in the stage data are called, and the grayscale change value and pressure gradient change amplitude are calculated to analyze the difference in the change characteristics of each stage and generate grayscale and pressure change amplitude data;

[0121] The grayscale change value and the pressure gradient change amplitude are divided into stage records by time segmentation, and the variance of the grayscale change value in each stage is calculated. Calculate the grayscale change variance, where is the mean value of grayscale change, using standard deviation Indicates the amplitude of change, and then the pressure gradient change value is also segmented and its mean is calculated With standard deviation , use the above formula for calculation, and classify these standard deviation values ​​according to the quantization interval. The stage with lower standard deviation (such as below a certain threshold) is classified as the stable change stage, and the stage with higher standard deviation (above a certain threshold) is classified as the fluctuating change stage. Combined with the classification of the two standard deviation values, the difference in feature changes in each stage is calculated. By matching the classification of grayscale changes and the classification of pressure gradient changes, the stage data records with the largest difference between the two are screened out to generate grayscale and pressure change amplitude data.

[0122] Perform statistical analysis on the grayscale and pressure change amplitude data, call the Gaussian mixture model to detect the distribution characteristics of the data change amplitude, screen and mark the stages with concentrated and significant change amplitudes, and organize them into the screened key stage data;

[0123] All recorded data points are classified into multiple mixed distribution groups, and the standard deviation range and mean range of the data in each distribution group are counted. The mean difference within the distribution group is used as the classification basis, and a concentration threshold based on the distribution group mean is set to screen out data points that meet the concentration threshold conditions in each distribution group. The distribution ratio of the distribution group data points within the standard deviation range is calculated, and the range of data points belonging to the significant distribution is calculated in combination with the probability density curve of the Gaussian mixture distribution. By grouping and aggregating the characteristics of these data points, the stages with concentrated and significant changes are screened out to generate the screened key stage data.

[0124] The multi-feature correlation analysis is performed on the selected key stage data. By matching the difference between the grayscale change value and the pressure gradient value, the correlation between the feature values ​​of the stages is calculated using the formula:

[0125] ;

[0126] Generate dynamic key characteristic values;

[0127] in, represents the dynamic key characteristic value, represents the grayscale change value of the i-th stage, represents the pressure gradient value of the i-th stage, Represents the standard deviation of grayscale change value, represents the standard deviation of the pressure gradient values, Indicates the number of key stages.

[0128] formula:

[0129] ;

[0130] The benefit of the formula is that by introducing the normalization factor and It can dynamically balance the influence ratio between the grayscale change value and the pressure gradient change value, and amplify the sensitivity of the feature difference through square and square root operations, thereby improving the calculation accuracy of the dynamic key feature values.

[0131] Detailed explanation of the formula and the process of formula calculation and derivation:

[0132] make , , , , the calculation process is as follows:

[0133] 1. Normalize the parameter values: , .

[0134] 2. Calculate the difference between normalized values: .

[0135] 3. Sum and calculate square root: .

[0136] 4. Calculate the average: .

[0137] The result shows that the dynamic key eigenvalue is 1.67, which reflects the dynamic correlation between the grayscale change value and the pressure gradient change value. The numerical results are further used for the aggregation and analysis of dynamic features to obtain the overall pattern of key feature distribution.

[0138] See also Figure 5 , the specific steps for obtaining the key state prediction value are:

[0139] Based on the dynamic key feature value, the image texture change and pressure gradient data in the time series are called, the segmented trend of the time series characteristic curve is extracted, the change amplitude of the key node is calculated, and the change characteristics are segmented and marked to generate the key change node data;

[0140] Extract image texture change and pressure gradient data in time series, and analyze the segmented trend of characteristic curves. First, call the time series data of characteristic change values, divide the data into multiple continuous stage nodes, each node is represented by a set of characteristic values ​​in a specific time period, and then calculate the characteristic change amplitude of each node, including the calculation of the change rate of texture characteristic gradient value and the mean analysis of pressure gradient difference, through the formula: ,in, represents the feature change value of the i-th stage, It represents the time interval. After completing the calculation of the change range in each stage, the result is brought into the fluctuation range analysis model to calculate the node feature change range in each stage. The feature nodes with significant change range are found through point-by-point comparison. Finally, the feature nodes are marked according to the calculation results to generate key change node data.

[0141] Analyze the characteristic fluctuation range of key change node data, call the adjacent data before and after the key node, calculate the interval fluctuation range of the characteristic value, analyze the change rate and amplitude of the node data, quantify the characteristics of each node by building a fluctuation range model, and generate node fluctuation characteristic data;

[0142] By obtaining the characteristic values ​​of each node in the adjacent time period, calling the time series curve smoothing method, the characteristic fluctuation amplitude near the node is calculated through the formula: , to conduct quantitative analysis of the fluctuation range, among which, represents the fluctuation range of the i-th node, and They represent the adjacent eigenvalues ​​before and after the current node respectively. The severity of the feature change is further evaluated by calculating the mean and standard deviation of the fluctuation range. After the feature fluctuation of each node is quantified, the fluctuation data is organized into structured eigenvalues ​​and used as input for subsequent operations to generate node fluctuation feature data.

[0143] Combined with the node fluctuation characteristic data, the model optimization and numerical calculation are carried out by adjusting the characteristic node weight distribution and trend influencing factors, using the formula:

[0144] ;

[0145] Generate key status prediction values;

[0146] in, represents the key state prediction value, represents the characteristic fluctuation value of the i-th node, is the weight factor of the corresponding node, which is used to describe the criticality of the node. is a normalization parameter used to control the dynamic range of the predicted values, Indicates the number of nodes.

[0147] formula:

[0148] ;

[0149] The benefit of the formula is that by introducing the weight factor and normalization parameters , dynamically balances the impact of each node’s characteristic fluctuations on the key state prediction value, and at the same time smoothes the cumulative changes through a logarithmic function to avoid excessive sensitivity of the prediction value to extreme fluctuation data.

[0150] Detailed explanation of the formula and the process of formula calculation and derivation:

[0151] Substitute the values ​​in the node fluctuation characteristic data into the formula, assuming that the parameters obtained through the above analysis are:

[0152] (The number of nodes is 3);

[0153] , , (node ​​characteristic fluctuation value);

[0154] , , (weight factor);

[0155] (Normalization parameter).

[0156] First, calculate the cumulative weighted feature fluctuation value:

[0157] ;

[0158] Then substitute the formula to calculate the key state prediction value:

[0159] ;

[0160] The result shows that the calculated key state prediction value is 2.286, which means that under the current system node characteristic fluctuation data, the predicted state value shows moderate changes within the set range. Through further analysis of the result, it can be used as an input parameter for subsequent optimization and adjustment.

[0161] See also Figure 6 , the specific steps for obtaining the monitoring parameters of the extraction process are:

[0162] Based on the key state prediction value, the real-time data monitoring system is called to obtain the temperature and time data of the current distillation process, the current state is compared with the target state, the deviation degree of temperature and time in the current distillation state is analyzed, and the current state data is generated through difference calculation;

[0163] The sensor collects temperature and time change data in real time, records the average value of the current distillation temperature curve every minute and the increment of the distillation duration, and compares the change range of the current distillation temperature and the increase or decrease of the running time in combination with the optimal distillation temperature and running time set in the target state. and The difference between the actual temperature and the target temperature, and the deviation between the actual running time and the target running time are calculated respectively, and these differences are organized into a monitoring matrix including temperature and time deviations. Further analysis is performed on whether the temperature fluctuation exceeds the target range and whether the running time is overdue or insufficient. By setting the upper and lower thresholds of the temperature fluctuation and the threshold of the time difference, the deviation is classified and labeled to generate the current status data.

[0164] Calculate the adjustment amount of temperature and time parameters using the current state data, analyze the optimal range of temperature and time in the target state, extract the temperature difference and time difference that need to be adjusted by performing difference calculation between the current data and the target parameter range, and generate adjustment parameter data;

[0165] Analyze the deviation matrix between the current distillation state and the target state, extract the abnormal items that exceed the threshold in the temperature and time deviation, call the difference matrix to partition and classify the difference between each temperature and time, and use the formula and Calculate the minimum correction amount of temperature and time that needs to be adjusted respectively, where: and The minimum adjustment benchmark value of temperature and time is used to further analyze whether global or segmented adjustment parameters are needed. The adjustment amount is normalized based on the median value of the target temperature range and the midpoint value of the time interval, and the adjustment result is converted into a dynamic adjustment parameter to generate adjustment parameter data.

[0166] Combined with the adjustment parameter data, the extraction process monitoring model is designed to adjust the temperature and time through a parameterized dynamic adjustment mechanism, using the formula:

[0167] ;

[0168] Calculate the extraction process monitoring parameters;

[0169] in, represents the monitoring parameters of the extraction process, Indicates the difference between the current temperature and the target temperature. Indicates the difference between the current running time and the target time. , are the target temperature and target time respectively, Represents the adjustment factor, which is used to balance the effects of temperature and time on the monitoring parameters.

[0170] formula:

[0171] ;

[0172] The benefit of the formula is that, by introducing normalization parameters and dynamic adjustment weights, coordinated optimization of temperature and time adjustments in monitoring parameters is achieved, so that the overall system can dynamically adapt to changes in temperature and time during the distillation process.

[0173] Detailed explanation of the formula and the process of formula calculation and derivation:

[0174] The current actual temperature is 120℃ and the target temperature is 115℃. ; The current actual running time is 62 minutes and the target time is 60 minutes, then ; The target temperature and target time are and ; Substitute into the formula:

[0175] ;

[0176] First calculate the fractional part:

[0177] ;

[0178] ;

[0179] Then calculate the square root part:

[0180] ;

[0181] Final calculation:

[0182] ;

[0183] The results show that the monitoring parameters of the extraction process are directly related to the current adjusted temperature and time deviations, and the calculated values ​​will be used to dynamically adjust the core parameter inputs of the model to ensure real-time adjustment and optimization of the distillation process.

[0184] A method for monitoring the extraction of Chinese medicinal materials based on image recognition, which is performed based on the above-mentioned Chinese medicinal materials extraction monitoring system based on image recognition, comprises the following steps:

[0185] S1: Based on the operating status of the Chinese herbal medicine extraction and distillation device, the data sets of the temperature sensor and the pressure sensor are called, and the morphological features of the Chinese herbal medicine in the image data collected during the distillation process are reorganized and processed, and the morphological features of the Chinese herbal medicine, temperature distribution and pressure change data are integrated to generate a phased data matrix;

[0186] S2: Based on the stage data matrix, calculate the grayscale difference in the image data, analyze the texture change range, integrate the distribution values ​​in the pressure and temperature data, perform data item matching and difference comparison, and generate the feature distribution value of the extraction stage;

[0187] S3: Based on the feature distribution value in the extraction stage, multi-stage amplitude calculation is performed on the grayscale change and the pressure gradient change, and the feature items in the change concentration are selected, and multi-parameter analysis and calculation are performed on the selected features to obtain dynamic key feature values;

[0188] S4: Based on dynamic key feature values, analyze the change amplitude and pressure gradient trend of image texture during Chinese herbal medicine extraction, locate key change nodes, analyze the node feature fluctuation range, and generate key state prediction values;

[0189] S5: Based on the key state prediction value, analyze the deviation between the current distillation state and the target state, and obtain the extraction process monitoring parameters in combination with the temperature control and time parameters that need to be adjusted.

[0190] The above are only preferred embodiments of the present invention and are not intended to limit the present invention in other forms. Any technician familiar with the profession may use the technical contents disclosed above to change or modify them into equivalent embodiments with equivalent changes and apply them to other fields. However, any simple modification, equivalent change and modification made to the above embodiments based on the technical essence of the present invention without departing from the technical solution of the present invention still falls within the protection scope of the technical solution of the present invention.

Claims

1. A Chinese herbal medicine extraction monitoring system based on image recognition, characterized in that: The system comprises: The dynamic data acquisition module collects image data and temperature sensor and pressure sensor data based on the operating status of the Chinese medicinal material extraction and distillation device, describes the morphological characteristics of the Chinese medicinal materials by grayscale changes and texture differences, and analyzes the relationship between the morphological characteristics of the Chinese medicinal materials and the temperature distribution and pressure changes of the distillation device in combination with the sensor data, and establishes a phased data matrix; The feature extraction and comparison module extracts the grayscale changes and texture differences of the morphological features of the Chinese medicinal materials in the image based on the stage data matrix, analyzes the pressure gradient and temperature distribution of the distillation device in the temperature sensor and pressure sensor data, matches the data features of the same stage, and generates the feature distribution value of the extraction stage; The dynamic feature optimization module calculates the amplitude of grayscale change and pressure gradient change in multiple stages based on the feature distribution value in the extraction stage, selects the feature items with the most concentrated changes, analyzes the correlation of multiple features, and generates dynamic key feature values; The state prediction module analyzes the image texture change and pressure gradient trend during the extraction of Chinese medicinal materials based on the dynamic key feature values, locates the key change nodes, analyzes the characteristic fluctuation range of the key change nodes, and generates key state prediction values; The process monitoring module analyzes the difference between the current distillation state and the target state based on the key state prediction value, calculates the distillation temperature and operating time parameters that need to be adjusted, and generates the extraction process monitoring parameters.

2. The Chinese medicinal material extraction monitoring system based on image recognition according to claim 1 is characterized in that: The phased data matrix specifically includes the reorganization analysis and pressure change of the image morphological features; the extraction phase feature distribution value includes the grayscale change, texture difference, pressure gradient, and temperature distribution of the morphological features of the Chinese medicinal materials in the image; the dynamic key feature value specifically includes the amplitude of grayscale change, the amplitude of pressure gradient change, feature item concentration, and multi-feature correlation; the key state prediction value includes image texture change, pressure gradient trend, and feature fluctuation range of key change nodes; the extraction process monitoring parameters specifically include distillation temperature adjustment and running time parameters.

3. The Chinese medicinal material extraction monitoring system based on image recognition according to claim 2 is characterized in that: The steps for obtaining the phased data matrix are specifically as follows: Collect the temperature sensor and pressure sensor data of the running status of the Chinese herbal medicine extraction and distillation device, call the sensor interface of the device to collect temperature and pressure signals in real time, record the temperature change and pressure fluctuation data in the collected time series in time series segments, record the mean and fluctuation range of the segmented feature data, organize them into phased data sets, and generate temperature and pressure data pairs; Calling the segmented records of the temperature and pressure data pairs, analyzing the change characteristics of temperature and pressure item by item through the segmented records of the time series, calculating the mean and standard deviation of each segmented data, setting the relative ratio of the mean to the fluctuation range as a screening basis, screening and marking the data pairs whose relative ratio exceeds the threshold, and generating key stage data records; The temperature and pressure distribution characteristics corresponding to the key stage data records are reorganized into a matrix, and the weighted matrix of the key stage data is calculated using the formula: ； Normalize the matrix, calculate the multi-region data ratio, and generate a phased data matrix; in, Represents the weighted value of the corresponding partition in the stage data matrix, Represents the temperature value, Represents the pressure value, Represents the fluctuation weight of key stage data, Represents the total number of key stage data.

4. The Chinese medicinal material extraction monitoring system based on image recognition according to claim 3 is characterized in that: The steps for obtaining the feature distribution value in the extraction stage are specifically as follows: Utilizing the phased data matrix, extracting grayscale changes and texture differences in the image data of Chinese medicinal materials, calling the grayscale histogram to calculate the distribution characteristics of pixel values ​​in the region, using the local change gradient and direction distribution of the image to perform texture feature analysis, and obtaining image feature data; Combined with the data records of temperature and pressure sensors, the time series data is called to extract the change rate of pressure and temperature values, the change amplitude of the pressure gradient in each stage is calculated based on the stage time nodes, and the mean and standard deviation of the temperature distribution in multiple stages are analyzed to generate pressure and temperature characteristic data; The image feature data is correlated with the pressure and temperature feature data, and a phase feature distribution relationship is established through a feature matching algorithm, using the formula: ； Calculate the feature distribution value of each stage and generate the feature distribution value of the extraction stage; in, represents the characteristic distribution value of the i-th stage, is the image feature value of the kth region, and its absolute value indicates the change of texture or grayscale. is the maximum value of the image eigenvalues, , are the characteristic values ​​corresponding to pressure and temperature, , is the maximum recorded value of the characteristic values ​​of pressure and temperature, It is the coefficient for adjusting the weight of image feature data, which is used to balance the influence of image data in each region in the calculation of total feature distribution value. It is the coefficient for adjusting the weight of pressure and temperature sensor data, which helps determine the criticality of pressure and temperature data in each stage. Represents the number of regions.

5. The Chinese medicinal material extraction monitoring system based on image recognition according to claim 4 is characterized in that: The steps for obtaining the dynamic key feature value are specifically as follows: Based on the extracted stage characteristic distribution value, grayscale change and pressure gradient change records in the stage data are called, grayscale change values ​​and pressure gradient change amplitudes are calculated, and the difference in change characteristics of each stage is analyzed to generate grayscale and pressure change amplitude data; Performing statistical analysis on the grayscale and pressure change amplitude data, calling the Gaussian mixture model to detect the distribution characteristics of the data change amplitude, screening and marking the stages with concentrated and significant change amplitudes, and organizing them into screened key stage data; The multi-feature correlation analysis is performed on the selected key stage data. By matching the difference between the grayscale change value and the pressure gradient value, the correlation between the feature values ​​of the stages is calculated using the formula: ； Generate dynamic key characteristic values; in, represents the dynamic key characteristic value, represents the grayscale change value of the i-th stage, represents the pressure gradient value of the i-th stage, Represents the standard deviation of grayscale change value, represents the standard deviation of the pressure gradient values, Indicates the number of key stages.

6. The Chinese medicinal material extraction monitoring system based on image recognition according to claim 5 is characterized in that: The steps for obtaining the key state prediction value are specifically as follows: Based on the dynamic key feature value, the image texture change and pressure gradient data in the time series are called, the segmented trend of the time series characteristic curve is extracted, the change amplitude of the key node is calculated, and the change feature is segmented and marked to generate the key change node data; Perform characteristic fluctuation range analysis on the key change node data, call the adjacent data before and after the key node, calculate the interval fluctuation range of the characteristic value, analyze the change rate and amplitude of the node data, quantify the characteristics of each node by building a fluctuation range model, and generate node fluctuation characteristic data; Combined with the node fluctuation characteristic data, the model optimization and numerical calculation are performed by adjusting the characteristic node weight distribution and trend influencing factors, using the formula: ； Generate key status prediction values; in, represents the key state prediction value, represents the characteristic fluctuation value of the i-th node, is the weight factor of the corresponding node, which is used to describe the criticality of the node. is a normalization parameter used to control the dynamic range of the predicted values, Indicates the number of nodes.

7. The Chinese medicinal material extraction monitoring system based on image recognition according to claim 6 is characterized in that: The steps for obtaining the monitoring parameters of the extraction process are specifically as follows: Based on the key state prediction value, the real-time data monitoring system is called to obtain the temperature and time data of the current distillation process, the current state is compared with the target state, the degree of deviation of the temperature and time in the current distillation state is analyzed, and the current state data is generated by difference calculation; Calculate the adjustment amount of temperature and time parameters using the current state data, analyze the optimal range of temperature and time in the target state, extract the temperature difference and time difference that need to be adjusted by performing difference calculation between the current data and the target parameter range, and generate adjustment parameter data; Combined with the adjustment parameter data, an extraction process monitoring model is designed to adjust the temperature and time through a parameterized dynamic adjustment mechanism, using the formula: ； Calculate the extraction process monitoring parameters; in, represents the monitoring parameters of the extraction process, Indicates the difference between the current temperature and the target temperature. Indicates the difference between the current running time and the target time. , are the target temperature and target time respectively, Represents the adjustment factor, which is used to balance the effects of temperature and time on the monitoring parameters.

8. A method for monitoring the extraction of Chinese medicinal materials based on image recognition, characterized in that: The Chinese medicinal material extraction monitoring system based on image recognition according to any one of claims 1 to 7 comprises the following steps: S1: Based on the operating status of the Chinese herbal medicine extraction and distillation device, the data sets of the temperature sensor and the pressure sensor are called, and the morphological features of the Chinese herbal medicine in the image data collected during the distillation process are reorganized and processed, and the morphological features of the Chinese herbal medicine, temperature distribution and pressure change data are integrated to generate a phased data matrix; S2: Based on the phased data matrix, the grayscale difference in the image data is calculated, the texture variation range is analyzed, the distribution values ​​in the pressure and temperature data are integrated, the data items are matched and the difference is compared, and the feature distribution value of the extraction phase is generated; S3: Based on the feature distribution value in the extraction stage, multi-stage amplitude calculation is performed on the grayscale change and the pressure gradient change, characteristic items in the change concentration are selected, and multi-parameter analysis and calculation are performed on the selected features to obtain dynamic key characteristic values; S4: Based on the dynamic key feature values, analyze the change amplitude and pressure gradient trend of the image texture during the extraction of traditional Chinese medicine, locate the key change nodes, analyze the node feature fluctuation range, and generate the key state prediction value; S5: Based on the key state prediction value, the deviation between the current distillation state and the target state is analyzed, and the extraction process monitoring parameters are obtained in combination with the temperature control and time parameters that need to be adjusted.

Images