White tea fermentation quality evaluation method, system and device based on image analysis

Abstract

The invention discloses a method, system and device for evaluating the fermentation quality of white tea based on image analysis, and specifically relates to the technical field of food processing image analysis, and is used to solve the problem that the existing method fails to effectively monitor and quantify the precipitation state of oil substances on the surface of white tea leaves; image data of the entire surface of the leaf is acquired by multispectral image acquisition, a local brightness enhancement algorithm and a brightness threshold segmentation algorithm are combined to identify the reflective area, the reflection spectrum characteristics are extracted and compared with a feature database, the oil precipitation risk area is screened out, the grayscale co-occurrence matrix algorithm is further used to extract texture features, and the dynamic entropy analysis is combined to evaluate the roughness change, and finally the abnormal indicators of the oil precipitation risk area are quantified. The invention realizes the accurate monitoring and evaluation of the fermentation quality of white tea, and provides a scientific basis for fermentation process optimization and quality improvement.

Description

Technical Field

[0001] The present invention relates to the technical field of food processing image analysis, and more specifically, to a method, system and device for evaluating the fermentation quality of white tea based on image analysis. Background Art

[0002] White tea is a type of tea with natural fermentation as its core process. The precipitation effect of oil substances on the surface of the leaves during the fermentation process has an important impact on the quality of the tea. Appropriate and uniform precipitation may give white tea a unique aroma and flavor, but excessive or uneven precipitation may destroy the fermentation uniformity and reduce the quality of the tea. In the prior art, the monitoring and evaluation methods for the fermentation quality of white tea mainly rely on manual sensory identification or single physical parameter detection. However, these methods are highly subjective and lack the accuracy and comprehensiveness of the data, and cannot achieve a systematic analysis and evaluation of the precipitation effect of oil substances on the surface of white tea leaves.

[0003] In the prior art, the precipitation state of oil substances on the surface of white tea leaves has not been effectively monitored and quantified, which may lead to uneven distribution of oil or excessive precipitation during the fermentation process, making it difficult to detect and adjust abnormal changes in the fermentation process in time, further affecting the uniformity of fermentation and the final quality of the tea.

[0004] In order to solve the above problems, a technical solution is now provided. Summary of the invention

[0005] In order to overcome the above-mentioned defects of the prior art, the embodiments of the present invention provide a method, system and device for evaluating the fermentation quality of white tea based on image analysis to solve the problems raised in the above-mentioned background technology.

[0006] To achieve the above object, the present invention provides the following technical solutions:

[0007] The method for evaluating the fermentation quality of white tea based on image analysis comprises the following steps:

[0008] Multispectral image acquisition of white tea leaves during fermentation to obtain image data covering the entire leaf;

[0009] The local brightness enhancement algorithm is used to identify the highlight area on the leaf surface, and the brightness threshold segmentation algorithm is used to separate the reflective area.

[0010] Extract the reflection spectrum features in the reflective area, compare the reflection spectrum features with the oil and fat material feature database, and screen out the oil and fat precipitation risk areas;

[0011] The gray-level co-occurrence matrix algorithm is used to extract and analyze the surface texture characteristics of the oil precipitation risk area, and to evaluate the abnormal impact of oil precipitation on the surface texture characteristics of the blade;

[0012] By analyzing the dynamic entropy characteristics of the surface roughness in the grease precipitation risk area, the degree of change in surface roughness caused by the precipitation of grease substances is evaluated;

[0013] Based on the comprehensive analysis of the abnormal influence of grease precipitation on the surface texture characteristics of the blade and the degree of change of surface roughness caused by grease precipitation, the abnormal indicators of grease precipitation in the grease precipitation risk area are obtained.

[0014] In a preferred embodiment, multispectral image acquisition is performed on white tea leaves during fermentation to obtain image data covering the entire leaf, specifically:

[0015] Scan the surface of white tea leaves one by one during the fermentation process to capture multispectral image data of white tea leaves in the visible light band, near infrared light band and short-wave infrared light band;

[0016] The multispectral image data is stored as a multi-channel format file to ensure that the spectral data of each band is recorded independently;

[0017] The collected multispectral image data is preprocessed to eliminate the influence of ambient light changes and equipment noise on data accuracy during the collection process.

[0018] In a preferred embodiment, a local brightness enhancement algorithm is used to identify the highlight area on the blade surface, and a brightness threshold segmentation algorithm is used to separate the reflective area, specifically:

[0019] For the collected multispectral image data, the local brightness enhancement algorithm is used to calculate the brightness value pixel by pixel, and the brightness contrast of the local area is enhanced to highlight and enhance the recognizability of the highlight area;

[0020] Based on the image data after local brightness enhancement processing, a brightness segmentation threshold is set, and the pixel area with a brightness higher than the brightness segmentation threshold is extracted through the brightness threshold segmentation algorithm and marked as a reflective area;

[0021] Generate a binary image from the reflective area, where the pixel value of the highlight area is defined as 1 and the pixel value of the non-highlight area is defined as 0, in order to distinguish between the highlight area and the non-highlight area.

[0022] In a preferred embodiment, the reflection spectrum characteristics are extracted from the reflective area, and the reflection spectrum characteristics are compared with the oil substance characteristic database to screen out the oil precipitation risk area, specifically:

[0023] Perform spectral decomposition processing on the multispectral image data of the reflective area to extract the reflective spectrum characteristics of each pixel in the reflective area;

[0024] Compare the reflection spectrum characteristics with the spectrum data in the pre-established oil and fat material characteristic database, and calculate the similarity between the reflection spectrum characteristics of the reflective area and the oil and fat material characteristics;

[0025] Based on the comparison between the similarity between the reflective spectrum characteristics of the reflective area and the characteristics of the grease substance and the preset similarity threshold, the risk area of ​​grease precipitation is marked.

[0026] In a preferred embodiment, the gray level co-occurrence matrix algorithm is used to extract and uniformly analyze the surface texture features of the oil precipitation risk area to evaluate the abnormal impact of oil precipitation on the surface texture features of the blade, specifically:

[0027] Perform grayscale quantization on the image data of the oil precipitation risk area and evenly distribute the image grayscale value within the preset grayscale range;

[0028] Generate a gray-level co-occurrence matrix for the oil precipitation risk area and quantify the texture features in the area by calculating the gray-level value relationship between pixel pairs;

[0029] Analyze the uniformity of the gray-level co-occurrence matrix, calculate the distribution law of gray-level values ​​in the region, and extract the characteristic parameters related to texture uniformity. The characteristic parameters related to texture uniformity include consistency characteristics, contrast characteristics, and energy characteristics.

[0030] The characteristic parameters related to texture uniformity were comprehensively analyzed and the surface texture abnormality impact value was calculated. Based on the surface texture abnormality impact value, the abnormal impact of oil precipitation on the blade surface texture characteristics was evaluated.

[0031] In a preferred embodiment, the dynamic entropy characteristics of the surface roughness in the grease precipitation risk area are analyzed to evaluate the degree of change in surface roughness caused by the precipitation of grease substances, specifically:

[0032] Obtaining time series image data of the oil precipitation risk area, extracting surface roughness data at each moment in the time series image, and representing the surface roughness data as a two-dimensional height distribution matrix;

[0033] Dynamic entropy analysis is performed on the extracted surface roughness data, and the height distribution changes in the time series are expressed as dynamic entropy characteristic values;

[0034] The trend analysis of the dynamic entropy characteristic value was carried out by time series analysis method to quantify the change degree of surface roughness caused by the precipitation of grease substances.

[0035] In a preferred embodiment, the extracted surface roughness data is subjected to dynamic entropy analysis, and the height distribution change in the time series is expressed as a dynamic entropy characteristic value, specifically:

[0036] Define a time window and calculate the change entropy of the height distribution matrix within the time window. The formula is as follows: ;in, Indicates time point The dynamic entropy characteristic value of A discrete series representing height values, Indicates time point The altitude value at the moment is The probability of level Refers to the index of the discrete series after the height value is discretized.

[0037] In a preferred embodiment, based on a comprehensive analysis of the abnormal effect of grease precipitation on the surface texture characteristics of the blade and the degree of change in surface roughness caused by grease precipitation, an abnormal index of grease precipitation in the grease precipitation risk area is obtained, specifically:

[0038] The influence value of surface texture anomaly and the absolute value of dynamic entropy change rate are dimensionless and then weighted summed to calculate the influence coefficient of grease precipitation.

[0039] Among them, the calculation of the influence coefficient of grease precipitation can be realized by the following formula: ;in, It represents the influence coefficient of oil precipitation, represents the abnormal impact value of surface texture after dimensionless processing, represents the absolute value of the dynamic entropy change rate after dimensionless processing, and are the weights of the absolute value of the surface texture anomaly impact value and the dynamic entropy change rate, respectively, and and All are greater than 0;

[0040] Set the grease precipitation impact threshold and compare the grease precipitation impact coefficient with the grease precipitation impact threshold:

[0041] When the oil precipitation influence coefficient is greater than the oil precipitation influence threshold, it is judged that the precipitation of oil substances in the oil precipitation risk area has a greater adverse effect on the fermentation quality of white tea; otherwise, it is judged that the adverse effect of the precipitation of oil substances in the oil precipitation risk area on the fermentation quality of white tea is normal.

[0042] On the other hand, the present invention provides a white tea fermentation quality evaluation system based on image analysis, including a multi-spectral image acquisition module, a reflective area recognition module, a spectral feature comparison module, a texture feature analysis module, a roughness change evaluation module and an abnormal index analysis module;

[0043] Multi-spectral image acquisition module: collects multi-spectral images of white tea leaves during the fermentation process to obtain image data covering the entire leaf;

[0044] Reflective area recognition module: uses the local brightness enhancement algorithm to identify the highlight area on the leaf surface, and separates the reflective area through the brightness threshold segmentation algorithm;

[0045] Spectral feature comparison module: extracts the reflection spectrum features in the reflective area, compares the reflection spectrum features with the oil and fat material feature database, and screens out the oil and fat precipitation risk areas;

[0046] Texture feature analysis module: Use the gray-level co-occurrence matrix algorithm to extract and analyze the surface texture features of the oil precipitation risk area, and evaluate the abnormal impact of oil precipitation on the surface texture features of the blade;

[0047] Roughness change assessment module: By analyzing the dynamic entropy characteristics of the surface roughness in the grease precipitation risk area, the degree of surface roughness change caused by the precipitation of grease substances is assessed;

[0048] Abnormal index analysis module: Based on the comprehensive analysis of the abnormal impact of grease precipitation on the surface texture characteristics of the blade and the degree of change in surface roughness caused by grease precipitation, the abnormal index of grease precipitation in the grease precipitation risk area is obtained.

[0049] In another aspect, the present invention provides a device for evaluating the fermentation quality of white tea based on image analysis, comprising:

[0050] one or more processors;

[0051] A storage device for storing one or more programs;

[0052] When one or more programs are executed by one or more processors, the one or more processors implement a method for evaluating the fermentation quality of white tea based on image analysis.

[0053] The technical effects and advantages of the white tea fermentation quality evaluation method, system and device based on image analysis of the present invention are as follows:

[0054] 1. The present invention realizes accurate monitoring and quantitative evaluation of the precipitation effect of oily substances on the surface of white tea leaves. Through multi-spectral image acquisition and processing, the surface area of ​​white tea leaves is fully covered. Combined with the local brightness enhancement algorithm and the brightness threshold segmentation algorithm, it can effectively identify and separate the reflective areas where oily substances may be precipitated, laying a data foundation for subsequent feature analysis. At the same time, by extracting the reflection spectrum characteristics of the reflective area and comparing them with the oily substance feature database, the oil precipitation risk area is screened out, which significantly improves the positioning accuracy of the abnormal precipitation area and solves the problem that the existing technology cannot systematically monitor the oil precipitation status.

[0055] 2. The present invention extracts the texture features of the oil precipitation risk area through the grayscale co-occurrence matrix algorithm, combines the dynamic entropy analysis method to evaluate the rate of change of surface roughness, combines the static texture features with the dynamic roughness changes, and realizes a multi-dimensional comprehensive analysis of the surface texture and roughness characteristics of white tea leaves. Based on these analysis results, the present invention can quantify and output abnormal indicators of the oil precipitation risk area, so as to judge the degree of influence of the precipitation of oil substances on the fermentation quality of white tea. This evaluation method can timely discover and adjust the uneven or excessive precipitation phenomena that may occur during the fermentation process, optimize the white tea fermentation process, and provide a scientific basis for the standardized production and quality control of white tea. BRIEF DESCRIPTION OF THE DRAWINGS

[0056] Figure 1 Schematic diagram of the white tea fermentation quality evaluation method based on image analysis of the present invention;

[0057] Figure 2 It is a structural schematic diagram of the white tea fermentation quality evaluation system based on image analysis of the present invention. DETAILED DESCRIPTION

[0058] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0059] Example 1

[0060] Figure 1 The present invention provides a method for evaluating the fermentation quality of white tea based on image analysis, which comprises the following steps:

[0061] Multispectral images of white tea leaves during the fermentation process are collected to obtain image data covering the entire leaf.

[0062] The local brightness enhancement algorithm is used to identify the highlight area on the leaf surface, and the brightness threshold segmentation algorithm is used to separate the reflective area.

[0063] The reflection spectrum features are extracted in the reflective area, compared with the oil substance feature database, and the risk areas of oil precipitation are screened out.

[0064] The gray-level co-occurrence matrix algorithm was used to extract and uniformly analyze the surface texture features of the oil precipitation risk area, and to evaluate the abnormal impact of oil precipitation on the surface texture characteristics of the blade.

[0065] By analyzing the dynamic entropy characteristics of the surface roughness in the grease precipitation risk area, the degree of change in surface roughness caused by the precipitation of grease substances is evaluated.

[0066] Based on the comprehensive analysis of the abnormal influence of grease precipitation on the surface texture characteristics of the blade and the degree of change of surface roughness caused by grease precipitation, the abnormal indicators of grease precipitation in the grease precipitation risk area are obtained.

[0067] Multispectral image acquisition is performed on white tea leaves during the fermentation process to obtain image data covering the entire leaf, specifically:

[0068] The surface of white tea leaves during the fermentation process is scanned piece by piece to capture multispectral image data of white tea leaves in the visible light band, near infrared light band, and short-wave infrared light band:

[0069] White tea leaves were placed one by one on a standardized acquisition platform, and the platform surface was made of matte black material to avoid interference of background reflection on image acquisition.

[0070] Configure the multispectral camera and set the acquisition angle to be perpendicular to the leaf surface and the distance to be 50 cm to ensure that the camera field of view covers the entire leaf area.

[0071] Enable the automatic slice-by-slice scanning mode, move the sampling blade step by step to the acquisition area, and use the stepper motor to accurately control the blade position to avoid the influence of blade displacement on the consistency of image acquisition.

[0072] By scanning each leaf one by one, the complete multispectral data of each white tea leaf during the fermentation process can be captured, avoiding data loss caused by local occlusion or uneven light.

[0073] Store the multispectral image data as a multi-channel format file to ensure that the spectral data of each band is recorded independently:

[0074] The collected multispectral images are transmitted to the image processing computer through the real-time data transmission module.

[0075] Using the spectral separation function in the acquisition software, the spectral data of visible light, near-infrared light, and short-wave infrared light were extracted separately, and stored as multi-channel format files according to the band arrangement.

[0076] In each file, ensure that the spectral data of each band is represented by an independent pixel matrix, and record the file's metadata (including acquisition time, band range, equipment calibration information, etc.).

[0077] Preprocess the collected multispectral image data to eliminate the impact of ambient light changes and equipment noise on data accuracy during the collection process:

[0078] The collected multispectral image data is normalized to eliminate errors caused by ambient light changes and equipment noise, ensuring the authenticity and availability of the image data. The median filter algorithm is used to remove random noise in each band image to ensure the continuity of pixel data. The image background light signal is separated by spectral analysis software, and background light subtraction is used to remove the influence of ambient light sources on multispectral data.

[0079] The local brightness enhancement algorithm is used to identify the highlight area on the leaf surface, and the brightness threshold segmentation algorithm is used to separate the reflective area. Specifically:

[0080] For the collected multispectral image data, the local brightness enhancement algorithm is used to calculate the brightness value pixel by pixel. By enhancing the brightness contrast of the local area, the recognizability of the highlight area is highlighted and enhanced:

[0081] For the previously collected multispectral image data, the visible light band image is selected as the main analysis object. The local brightness enhancement algorithm is used to calculate the brightness value pixel by pixel. The brightness value is calculated based on the weighted average of the original brightness value of each pixel and the brightness of its neighboring pixels.

[0082] The specific operation is to convert the collected visible light band image into a grayscale image. The brightness of each pixel in the grayscale image represents its light intensity. For each pixel point, calculate the average brightness value of its neighborhood (such as a 3×3 or 5×5 pixel area as a window), and adjust its brightness value in combination with the pixel point's own brightness to enhance the contrast between pixels.

[0083] The brightness value of each pixel is adjusted according to the preset enhancement factor. The enhancement factor is used to amplify the pixel brightness value, making the pixels with higher brightness values ​​brighter.

[0084] During the adjustment process, for each pixel, its brightness value is compared with the average brightness of its neighborhood. For pixels whose brightness value is significantly higher than the neighborhood average, its brightness is appropriately increased. After brightness enhancement, the edge protection filtering method is applied to avoid excessive blurring of image edge details during the enhancement process.

[0085] By enhancing the brightness characteristics of highlight areas, these areas are ensured to be more prominent in the subsequent segmentation process, facilitating accurate identification.

[0086] Based on the image data after local brightness enhancement processing, the brightness segmentation threshold is set, and the pixel area above the brightness segmentation threshold is extracted through the brightness threshold segmentation algorithm and marked as the reflective area:

[0087] Analyze the enhanced image brightness histogram, count the distribution range of all pixel brightness values, and select a brightness segmentation threshold. The brightness segmentation threshold is usually set as the lower limit of the highlight area in the pixel brightness distribution.

[0088] The brightness threshold segmentation method is used to extract the pixel area with brightness values ​​higher than the brightness segmentation threshold and mark it as the reflective area.

[0089] Morphological optimization operations are performed on the extracted reflective areas, such as filling small holes through dilation operations and removing noise points through erosion operations to ensure the continuity of the reflective area boundaries.

[0090] Generate a binary image from the reflective area, where the pixel value of the highlight area is defined as 1 and the pixel value of the non-highlight area is defined as 0, to distinguish between the highlight area and the non-highlight area:

[0091] The pixel value of the reflective area is defined as 1, and the value of the remaining pixel area is defined as 0 to generate a binary image.

[0092] Check the connectivity of the highlight area in the binary image to ensure the accuracy of the segmentation result and avoid the loss of highlight areas due to segmentation errors. Save the generated binary image and store it in a standardized format to provide input data for subsequent texture analysis and spectral feature extraction.

[0093] Binary images can intuitively show the distribution of highlight areas and non-highlight areas, which is convenient for subsequent algorithm processing and analysis.

[0094] Extract the reflection spectrum features in the reflective area, compare the reflection spectrum features with the oil substance feature database, and screen out the oil precipitation risk areas, specifically:

[0095] Perform spectral decomposition on the multispectral image data of the reflective area to extract the reflective spectrum characteristics of each pixel in the reflective area:

[0096] For the separated reflective area, the pixel data in the area is extracted according to its position index in the multispectral image, and each pixel in the multispectral image is represented as spectral reflectance data of multiple bands, where each band corresponds to a specific spectral range (for example, 400-700 nanometers is the visible light band, and 700-1400 nanometers is the near-infrared band).

[0097] The reflectance spectrum characteristics of each pixel are represented as a set of spectral reflectance vectors, in which each component corresponds to the reflectance value of a band.

[0098] Compare the reflection spectrum characteristics with the spectrum data in the pre-established grease material characteristic database, and calculate the similarity between the reflection spectrum characteristics of the reflective area and the grease material characteristics:

[0099] The multi-spectral reflectance spectrum characteristics of oily substances are collected in advance through experiments to build a characteristic database, and each record in the database corresponds to a characteristic spectrum of an oily substance.

[0100] Use a spectral similarity calculation method (such as Euclidean distance method, cosine similarity method or correlation coefficient method) to calculate the similarity between the reflectance spectral characteristics of each pixel in the reflective area and each record in the database.

[0101] The spectral similarity of all pixels in the reflective area is calculated one by one, and the calculation results are stored as a similarity matrix, where each row corresponds to a pixel and each column corresponds to a characteristic spectrum.

[0102] Based on the comparison of the similarity between the reflective spectrum characteristics of the reflective area and the characteristics of the grease substance and the preset similarity threshold, the grease precipitation risk area is marked:

[0103] Based on experimental verification, the similarity threshold is determined. The similarity threshold is a standard for judging the degree of match between the spectral characteristics of pixels in the reflective area and the characteristics of oily substances. It is set to a value between 0 and 1 based on the experimental results. For example, 0.8 means that pixels with a similarity value higher than the threshold are identified as oil precipitation risk areas, ensuring the accuracy of the screening results.

[0104] Each pixel value in the similarity matrix is ​​compared. If the similarity value of a pixel is higher than the similarity threshold, the corresponding pixel position is marked as a grease precipitation risk area.

[0105] The marked grease precipitation risk area may be converted into a binary image, wherein the pixel value of the grease precipitation risk area is set to 1, and the pixel value of the non-grease precipitation risk area is set to 0.

[0106] The gray-level co-occurrence matrix algorithm is used to extract and uniformly analyze the surface texture features of the oil precipitation risk area, and to evaluate the abnormal impact of oil precipitation on the surface texture characteristics of the blades. Specifically:

[0107] Perform grayscale quantization on the image data of the oil precipitation risk area and evenly distribute the image grayscale value within the preset grayscale range:

[0108] Grayscale quantization is performed on the image data of the oil precipitation risk area to reduce data complexity and improve the efficiency of grayscale co-occurrence matrix calculation. The goal of grayscale quantization is to evenly distribute the original grayscale value range (usually 0 to 255) to a preset grayscale range.

[0109] Generate a grayscale co-occurrence matrix for the oil precipitation risk area and quantify the texture features in the area by calculating the grayscale value relationship between pixel pairs:

[0110] Define the direction and distance of the pixel pair. For example, the horizontal distance is 1 pixel. Calculate the gray value at this distance and direction. and The frequency of occurrence of pixel pairs generates the matrix elements in the gray-level co-occurrence matrix: ;in, represents the matrix elements in the gray-level co-occurrence matrix, Indicates the gray value is and The number of occurrences of pixel pairs, Represents the total number of pixel pairs.

[0111] Represents the gray value of the first pixel in the gray level co-occurrence matrix, ranging from 1 to the target gray level, used to describe the gray level of the first pixel; Represents the grayscale value of the second pixel in the gray level co-occurrence matrix, ranging from 1 to the target grayscale level, which is used to describe the grayscale level that forms a pixel pair with the first pixel.

[0112] Analyze the uniformity of the gray-level co-occurrence matrix, calculate the distribution law of gray values ​​in the region, and extract the characteristic parameters related to texture uniformity:

[0113] The characteristic parameters related to texture uniformity include consistency feature, contrast feature and energy feature.

[0114] The gray-level co-occurrence matrix is ​​used to calculate the consistency features and evaluate the uniformity of the gray-level distribution.

[0115] Calculate the consistency feature, the formula is as follows: ;in, Represents the consistency feature, which is used to measure the uniformity of grayscale distribution. The higher the value, the more uniform the grayscale distribution. is the target grayscale level.

[0116] Contrast is used to quantify the difference between grayscale values ​​and reflects the strength of the grayscale change of pixels in the area. The contrast feature is calculated as follows: ;in, Represents contrast features, which are used to indicate the intensity of grayscale value changes in an area.

[0117] Energy is used to evaluate the regularity of the gray-level co-occurrence matrix and reflects the stability of texture distribution. The energy feature is calculated as follows: ;in, Represents energy characteristics and is used to indicate the regularity and uniformity of grayscale distribution.

[0118] Among them, the larger the consistency feature, the more uniform the distribution of grayscale values ​​and the smoother the texture. Too large a consistency value may indicate that the surface texture tends to be overly smooth, which may be related to the loss of texture feature details caused by grease coverage.

[0119] The larger the contrast feature, the more significant the difference in grayscale value changes and the stronger the texture contrast. Too large a contrast may reflect uneven distribution of oil in a local area, resulting in significant changes in surface reflection characteristics, indicating an abnormality.

[0120] The larger the energy feature, the more concentrated the grayscale value distribution and the more regular the texture. An energy feature that is too large may indicate that the surface texture tends to be monotonous and unnatural, which may be related to the loss of texture features caused by excessive coverage of grease substances.

[0121] The characteristic parameters related to texture uniformity were comprehensively analyzed, and the abnormal impact value of surface texture was calculated. Based on the abnormal impact value of surface texture, the abnormal impact of oil precipitation on the surface texture characteristics of the blade was evaluated:

[0122] Each characteristic parameter related to texture uniformity is assigned a weight coefficient, and the weight coefficients are all set to be greater than 0.

[0123] Among them, the weight coefficient is set through experimental verification and statistical analysis, mainly based on the actual contribution of each feature to the abnormal impact. In the experiment, texture feature data under different oil coverage conditions are collected, and the variation range of consistency, contrast and energy is calculated. Combined with expert experience and model optimization results, weight values ​​are assigned. The weight coefficients are all greater than 0 and adjusted according to actual needs to ensure that the comprehensive calculation results can accurately reflect the degree of abnormality.

[0124] The characteristic parameters related to texture uniformity are normalized and then multiplied by their corresponding weight coefficients and summed to obtain the surface texture anomaly impact value.

[0125] The larger the surface texture abnormality impact value, the greater the abnormal impact of oil and fat on the leaf surface texture characteristics; the surface texture abnormality impact value comprehensively reflects the degree of change in texture characteristics by combining characteristics such as consistency, contrast and energy. The larger the value, the more significant the impact on the uniformity of grayscale distribution, the intensity of grayscale change or the regularity of texture, indicating that the abnormal impact of oil and fat coverage on the leaf surface texture characteristics is more significant, and can more accurately reflect the actual situation of surface texture abnormality.

[0126] By analyzing the dynamic entropy characteristics of the surface roughness in the grease precipitation risk area, the degree of change in surface roughness caused by the precipitation of grease substances is evaluated, specifically:

[0127] Obtain the time series image data of the oil precipitation risk area, extract the surface roughness data at each moment in the time series image, and represent the surface roughness data as a two-dimensional height distribution matrix:

[0128] Using multispectral imaging equipment, the target area is imaged at fixed time intervals to generate time series image data covering the entire oil precipitation risk area.

[0129] Time series images are stored in three-dimensional form. Each frame of the image corresponds to a time point and is stored as a three-dimensional data matrix. The three-dimensional data matrix is ​​represented as , Indicates time point At the coordinates Gray value of and Respectively represent the abscissa and ordinate of the plane coordinates of the image, used to describe the two-dimensional spatial position; Represents the time index, corresponding to each frame image in the time series.

[0130] The surface roughness data of each frame is extracted from the time series images and converted into a two-dimensional height distribution matrix for subsequent dynamic entropy calculation.

[0131] By performing texture analysis on the image data, the gray value of each pixel is mapped to the corresponding height value to generate a height distribution matrix, which is expressed as , Indicates time point At the coordinates The height value of .

[0132] The mapping formula is as follows: ;in, represents the grayscale value (ranging from 0 to 1), Represents the height conversion scale factor (used to map grayscale values ​​to physical height).

[0133] Dynamic entropy analysis is performed on the extracted surface roughness data, and the height distribution changes in the time series are expressed as dynamic entropy characteristic values:

[0134] By calculating the dynamic entropy of the height distribution matrix data, the change of height distribution in the time series is quantified as a dynamic entropy characteristic value, which can be achieved as follows:

[0135] Define a time window and calculate the change entropy of the height distribution matrix within the time window. The formula is as follows: ;in, Indicates time point The dynamic entropy characteristic value of A discrete series representing height values, Indicates time point The altitude value at the moment is The probability of level Refers to the index of the discrete series after the height value is discretized.

[0136] in, ; Indicates time point The altitude value at the moment is The number of pixels of the level, Indicates the total number of pixels.

[0137] The dynamic entropy characteristic values ​​are stored as a time series to obtain a dynamic entropy change sequence.

[0138] The trend analysis of the dynamic entropy characteristic value is carried out by time series analysis method to quantify the degree of change in surface roughness caused by the precipitation of grease substances:

[0139] Fitting analysis is performed on the dynamic entropy change sequence using a linear or nonlinear trend model. The fitting formula is: ;in, represents the rate of change of dynamic entropy, Represents the initial entropy value.

[0140] If the dynamic entropy change rate of the fitting result is positive and the larger the absolute value, it indicates that the randomness of the surface roughness caused by the precipitation of oil substances has increased significantly, and the roughness characteristics tend to be more complex and unstable, which may be related to the abnormal protrusions or uneven distribution of the leaf surface texture. If the dynamic entropy change rate is negative and the larger the absolute value, it indicates that the randomness of the surface roughness decreases rapidly, and the roughness characteristics tend to be over-smoothed or single, which may be related to the loss of surface texture details caused by excessive oil coverage. Regardless of whether it is positive or negative, the larger the absolute value of the dynamic entropy change rate, the higher the degree of quality abnormality of the white tea fermentation process caused by the precipitation of oil substances, and attention should be paid to abnormal conditions in fermentation control.

[0141] Based on the comprehensive analysis of the abnormal influence of grease precipitation on the blade surface texture characteristics and the degree of change in surface roughness caused by grease precipitation, the abnormal indicators of grease precipitation in the grease precipitation risk area are obtained, specifically:

[0142] The surface texture anomaly impact value and the dynamic entropy change rate reflect the impact of oil and fat precipitation from different dimensions, one focusing on the spatial texture characteristics and the other focusing on the time series changes. The comprehensive analysis results are more comprehensive.

[0143] The influence value of surface texture anomaly and the absolute value of the dynamic entropy change rate were dimensionlessly processed and then weighted summed to calculate the influence coefficient of grease precipitation.

[0144] Among them, the calculation of the influence coefficient of grease precipitation can be realized by the following formula: ;in, It represents the influence coefficient of oil precipitation, represents the abnormal impact value of surface texture after dimensionless processing, represents the absolute value of the dynamic entropy change rate after dimensionless processing, and are the weights of the absolute value of the surface texture anomaly impact value and the dynamic entropy change rate, respectively, and and Both are greater than 0.

[0145] Among them, the weights of the surface texture abnormality impact value and the absolute value of the dynamic entropy change rate are set according to the contribution of experimental data and characteristics to the fermentation quality. Through a large amount of texture characteristics and roughness change data during the fermentation process of white tea, the correlation between the surface texture abnormality impact value and the dynamic entropy change rate on the abnormal index is analyzed. The weights are usually obtained through multivariate linear regression or machine learning model optimization to ensure that the combined effect of the two can accurately reflect the abnormal degree of oil precipitation. At the same time, the weights are adjusted in combination with expert experience to adapt to different fermentation conditions and quality evaluation requirements.

[0146] The greater the oil precipitation influence coefficient, the higher the abnormal index of oil precipitation in the oil precipitation risk area. This means that the surface texture characteristics (such as consistency, contrast and energy) and the temporal change characteristics of surface roughness (dynamic entropy change rate) in the area are more significantly affected, indicating that the more serious the abnormality of oil precipitation on the leaf surface texture characteristics and roughness changes, the greater the adverse effect on the fermentation quality of white tea.

[0147] Set the grease precipitation impact threshold and compare the grease precipitation impact coefficient with the grease precipitation impact threshold:

[0148] When the oil precipitation influence coefficient is greater than the oil precipitation influence threshold, it is judged that the precipitation of oil substances in the oil precipitation risk area has a greater adverse effect on the fermentation quality of white tea; otherwise, it is judged that the adverse effect of the precipitation of oil substances in the oil precipitation risk area on the fermentation quality of white tea is normal.

[0149] Among them, the threshold of oil precipitation is a reference standard set based on experimental data and actual fermentation experience, which is used to distinguish the normal and abnormal effects of oil precipitation on the fermentation quality of white tea. This threshold is usually determined by statistically analyzing the range of changes in quality characteristics during the fermentation process of white tea to ensure that the key abnormal characteristics of fermentation quality can be accurately reflected.

[0150] Based on the abnormal indicators of oil precipitation in the oil precipitation risk area, the abnormal degree of surface texture and roughness changes during the fermentation of white tea can be accurately evaluated, providing data support for real-time monitoring and optimization of fermentation quality. Abnormal indicators reflect the abnormal effects of oil precipitation on leaf surface characteristics, including texture regularity, roughness change rate, etc., and can identify uneven oil distribution or abnormal coverage during the fermentation process. By analyzing this indicator, it is possible to determine whether there is a risk of quality abnormality or loss of control in the white tea fermentation process, and help to adjust the fermentation conditions (such as temperature and humidity) in a timely manner to ensure that the fermentation process is carried out within the optimal range, thereby improving the consistency of white tea fermentation quality and the quality stability of the final product.

[0151] Example 2

[0152] The difference between Example 2 of the present invention and Example 1 is that this example introduces a white tea fermentation quality evaluation system based on image analysis.

[0153] Figure 2 A structural schematic diagram of a white tea fermentation quality evaluation system based on image analysis of the present invention is given. The white tea fermentation quality evaluation system based on image analysis includes a multi-spectral image acquisition module, a reflective area recognition module, a spectral feature comparison module, a texture feature analysis module, a roughness change assessment module and an abnormal index analysis module.

[0154] Multi-spectral image acquisition module: multi-spectral image acquisition of white tea leaves during the fermentation process to obtain image data covering the entire leaf.

[0155] Reflective area recognition module: Use the local brightness enhancement algorithm to identify the highlight area on the leaf surface, and separate the reflective area through the brightness threshold segmentation algorithm.

[0156] Spectral feature comparison module: extracts the reflection spectrum features in the reflective area, compares the reflection spectrum features with the oil and fat material feature database, and screens out the areas at risk of oil and fat precipitation.

[0157] Texture feature analysis module: The gray-level co-occurrence matrix algorithm is used to extract and uniformly analyze the surface texture features of the oil precipitation risk area, and to evaluate the abnormal impact of oil precipitation on the surface texture features of the blades.

[0158] Roughness change assessment module: By analyzing the dynamic entropy characteristics of the surface roughness in the grease precipitation risk area, the degree of surface roughness change caused by the precipitation of grease substances is evaluated.

[0159] Abnormal index analysis module: Based on the comprehensive analysis of the abnormal impact of grease precipitation on the surface texture characteristics of the blade and the degree of change in surface roughness caused by grease precipitation, the abnormal index of grease precipitation in the grease precipitation risk area is obtained.

[0160] Example 3

[0161] This embodiment introduces a white tea fermentation quality evaluation device based on image analysis.

[0162] The device for evaluating the quality of white tea fermentation based on image analysis comprises one or more processors; and also comprises a storage device for storing one or more programs; when the one or more programs are executed by the one or more processors, the one or more processors implement a method for evaluating the quality of white tea fermentation based on image analysis.

[0163] The above formulas are all dimensionless and numerical calculations. The formula is a formula for the most recent real situation obtained by collecting a large amount of data and performing software simulation. The preset parameters and thresholds in the formula are set by technicians in this field according to actual conditions.

[0164] The above embodiments may be implemented in whole or in part by software, hardware, firmware or any other combination thereof. When implemented by software, the above embodiments may be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, the process or function described in the embodiment of the present application is generated in whole or in part. The computer may be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions may be stored in a computer-readable storage medium, or may be transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, the computer instructions may be transmitted from one website, computer, server or data center to another website, computer, server or data center by wired (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium may be any available medium that a computer can access or a data storage device such as a server or data center that contains one or more available media sets. The available medium may be a magnetic medium (e.g., a floppy disk, a hard disk, a tape), an optical medium (e.g., a DVD), or a semiconductor medium. The semiconductor medium may be a solid-state hard disk.

[0165] Those of ordinary skill in the art will appreciate that the modules and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.

[0166] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and modules described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0167] In the several embodiments provided in the present application, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of the modules is only a logical function division. There may be other division methods in actual implementation, such as multiple modules or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or modules, which can be electrical, mechanical or other forms.

[0168] The modules described as separate components may or may not be physically separated, and the components shown as modules may or may not be physical modules, and may be located in one place or distributed on multiple network modules. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0169] In addition, each functional module in each embodiment of the present application may be integrated into one processing module, or each module may exist physically separately, or two or more modules may be integrated into one module.

[0170] If the functions are implemented in the form of software function modules and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application can be essentially or partly embodied in the form of a software product that contributes to the prior art. The computer software product is stored in a storage medium, including several instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the methods described in the various embodiments of the present application. The aforementioned storage media include: U disk, mobile hard disk, read-only memory (ROM), random access memory (RAM), disk or optical disk, and other media that can store program codes.

[0171] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any technician familiar with the technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

[0172] Finally: The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for evaluating the fermentation quality of white tea based on image analysis, characterized in that: The steps include: Multispectral image acquisition of white tea leaves during fermentation to obtain image data covering the entire leaf; The local brightness enhancement algorithm is used to identify the highlight area on the leaf surface, and the brightness threshold segmentation algorithm is used to separate the reflective area. Extract the reflection spectrum features in the reflective area, compare the reflection spectrum features with the oil and fat material feature database, and screen out the oil and fat precipitation risk areas; The gray-level co-occurrence matrix algorithm is used to extract and analyze the surface texture characteristics of the oil precipitation risk area, and to evaluate the abnormal impact of oil precipitation on the surface texture characteristics of the blade; By analyzing the dynamic entropy characteristics of the surface roughness in the grease precipitation risk area, the degree of change in surface roughness caused by the precipitation of grease substances is evaluated; Based on the comprehensive analysis of the abnormal influence of grease precipitation on the blade surface texture characteristics and the degree of change in surface roughness caused by grease precipitation, the abnormal indicators of grease precipitation in the grease precipitation risk area are obtained; The gray-level co-occurrence matrix algorithm is used to extract and uniformly analyze the surface texture features of the oil precipitation risk area, and to evaluate the abnormal impact of oil precipitation on the surface texture characteristics of the blades. Specifically: Perform grayscale quantization on the image data of the oil precipitation risk area and evenly distribute the image grayscale value within the preset grayscale range; Generate a gray-level co-occurrence matrix for the oil precipitation risk area and quantify the texture features in the area by calculating the gray-level value relationship between pixel pairs; Analyze the uniformity of the gray-level co-occurrence matrix, calculate the distribution law of gray-level values ​​in the region, and extract the characteristic parameters related to texture uniformity. The characteristic parameters related to texture uniformity include consistency characteristics, contrast characteristics, and energy characteristics. The characteristic parameters related to texture uniformity were comprehensively analyzed and the surface texture abnormality impact value was calculated. Based on the surface texture abnormality impact value, the abnormal impact of oil precipitation on the blade surface texture characteristics was evaluated.

2. The method for evaluating the fermentation quality of white tea based on image analysis according to claim 1, characterized in that: Multispectral image acquisition is performed on white tea leaves during the fermentation process to obtain image data covering the entire leaf, specifically: Scan the surface of white tea leaves one by one during the fermentation process to capture multispectral image data of white tea leaves in the visible light band, near infrared light band and short-wave infrared light band; The multispectral image data is stored as a multi-channel format file to ensure that the spectral data of each band is recorded independently; The collected multispectral image data is preprocessed to eliminate the influence of ambient light changes and equipment noise on data accuracy during the collection process.

3. The method for evaluating the fermentation quality of white tea based on image analysis according to claim 2, characterized in that: The local brightness enhancement algorithm is used to identify the highlight area on the leaf surface, and the brightness threshold segmentation algorithm is used to separate the reflective area. Specifically: For the collected multispectral image data, the local brightness enhancement algorithm is used to calculate the brightness value pixel by pixel, and the brightness contrast of the local area is enhanced to highlight and enhance the recognizability of the highlight area; Based on the image data after local brightness enhancement processing, a brightness segmentation threshold is set, and the pixel area with a brightness higher than the brightness segmentation threshold is extracted through the brightness threshold segmentation algorithm and marked as a reflective area; Generate a binary image from the reflective area, where the pixel value of the highlight area is defined as 1 and the pixel value of the non-highlight area is defined as 0, in order to distinguish between the highlight area and the non-highlight area.

4. The method for evaluating the fermentation quality of white tea based on image analysis according to claim 3, characterized in that: Extract the reflection spectrum features in the reflective area, compare the reflection spectrum features with the oil substance feature database, and screen out the oil precipitation risk areas, specifically: Perform spectral decomposition processing on the multispectral image data of the reflective area to extract the reflective spectrum characteristics of each pixel in the reflective area; Compare the reflection spectrum characteristics with the spectrum data in the pre-established oil and fat material characteristic database, and calculate the similarity between the reflection spectrum characteristics of the reflective area and the oil and fat material characteristics; Based on the comparison between the similarity between the reflective spectrum characteristics of the reflective area and the characteristics of the grease substance and the preset similarity threshold, the risk area of ​​grease precipitation is marked.

5. The method for evaluating the fermentation quality of white tea based on image analysis according to claim 4, characterized in that: By analyzing the dynamic entropy characteristics of the surface roughness in the grease precipitation risk area, the degree of change in surface roughness caused by the precipitation of grease substances is evaluated, specifically: Obtaining time series image data of the oil precipitation risk area, extracting surface roughness data at each moment in the time series image, and representing the surface roughness data as a two-dimensional height distribution matrix; Dynamic entropy analysis is performed on the extracted surface roughness data, and the height distribution changes in the time series are expressed as dynamic entropy characteristic values; The trend analysis of the dynamic entropy characteristic value was carried out by time series analysis method to quantify the change degree of surface roughness caused by the precipitation of grease substances.

6. The method for evaluating the fermentation quality of white tea based on image analysis according to claim 5, characterized in that: Dynamic entropy analysis is performed on the extracted surface roughness data, and the height distribution changes in the time series are expressed as dynamic entropy characteristic values, specifically: Define a time window and calculate the change entropy of the height distribution matrix within the time window. The formula is as follows: ;in, Indicates time point The dynamic entropy characteristic value of A discrete series representing height values, Indicates time point The altitude value at the moment is The probability of level Refers to the index of the discrete series after the height value is discretized.

7. The method for evaluating the fermentation quality of white tea based on image analysis according to claim 6, characterized in that: Based on the comprehensive analysis of the abnormal influence of grease precipitation on the blade surface texture characteristics and the degree of change in surface roughness caused by grease precipitation, the abnormal indicators of grease precipitation in the grease precipitation risk area are obtained, specifically: The influence value of surface texture anomaly and the absolute value of dynamic entropy change rate are dimensionless and then weighted summed to calculate the influence coefficient of grease precipitation. Among them, the calculation of the influence coefficient of grease precipitation can be realized by the following formula: ;in, It represents the influence coefficient of oil precipitation, represents the abnormal impact value of surface texture after dimensionless processing, represents the absolute value of the dynamic entropy change rate after dimensionless processing, and are the weights of the absolute value of the surface texture anomaly impact value and the dynamic entropy change rate, respectively, and and All are greater than 0; Set the grease precipitation impact threshold and compare the grease precipitation impact coefficient with the grease precipitation impact threshold: When the oil precipitation influence coefficient is greater than the oil precipitation influence threshold, it is judged that the oil precipitation of oil substances in the oil precipitation risk area has a greater adverse effect on the fermentation quality of white tea; otherwise, it is judged that the adverse effect of oil precipitation of oil substances in the oil precipitation risk area on the fermentation quality of white tea is normal.

8. A white tea fermentation quality evaluation system based on image analysis, used to implement the white tea fermentation quality evaluation method based on image analysis according to any one of claims 1 to 7, characterized in that: It includes a multi-spectral image acquisition module, a reflective area recognition module, a spectral feature comparison module, a texture feature analysis module, a roughness change assessment module, and an abnormal index analysis module; Multi-spectral image acquisition module: collects multi-spectral images of white tea leaves during the fermentation process to obtain image data covering the entire leaf; Reflective area recognition module: uses the local brightness enhancement algorithm to identify the highlight area on the leaf surface, and separates the reflective area through the brightness threshold segmentation algorithm; Spectral feature comparison module: extracts the reflection spectrum features in the reflective area, compares the reflection spectrum features with the oil and fat material feature database, and screens out the oil and fat precipitation risk areas; Texture feature analysis module: Use the gray-level co-occurrence matrix algorithm to extract and analyze the surface texture features of the oil precipitation risk area, and evaluate the abnormal impact of oil precipitation on the surface texture features of the blade; Roughness change assessment module: By analyzing the dynamic entropy characteristics of the surface roughness in the grease precipitation risk area, the degree of surface roughness change caused by the precipitation of grease substances is assessed; Abnormal index analysis module: Based on the comprehensive analysis of the abnormal impact of grease precipitation on the surface texture characteristics of the blade and the degree of change in surface roughness caused by grease precipitation, the abnormal index of grease precipitation in the grease precipitation risk area is obtained.

9. A device for evaluating the fermentation quality of white tea based on image analysis, characterized in that: include: one or more processors; A storage device for storing one or more programs; When one or more programs are executed by one or more processors, the one or more processors implement the method for evaluating the fermentation quality of white tea based on image analysis as described in any one of claims 1 to 7.

Images