A machine vision-based intelligent identification and grading method and system for tea

Abstract

This invention discloses a machine vision-based intelligent tea identification and grading method and system, belonging to the field of intelligent grading technology. The system first photographs the tea leaves, including dry tea and brewed tea, using an image acquisition device. Then, image processing technology is used to preprocess the images to extract tea leaf features. Next, a tea classification model is constructed using machine learning algorithms, and the extracted features are correlated with the tea's quality grade to achieve automatic tea classification. Finally, a comprehensive dry tea coefficient (Gcxs) and a comprehensive wet tea coefficient (Scxs) are calculated, and then correlated with these coefficients to obtain a comprehensive grade coefficient (Pjxs). This allows for multi-level grading of the tea to determine its quality grade. This system not only improves the objectivity and accuracy of tea grading but also increases the efficiency of the assessment, bringing new opportunities for the development of the tea industry.

Description

Technical Field

[0001] This invention relates to the field of intelligent grading technology, specifically to a machine vision-based intelligent identification and grading method and system for tea. Background Technology

[0002] Tea is a beverage made from the leaves of the tea plant through processing. It has a long history and rich cultural significance worldwide and is widely enjoyed as a beverage. There are many types of tea, mainly including green tea, black tea, white tea, oolong tea, and yellow tea. Each type of tea has its own unique flavor and aroma, as well as different processing methods. The tea-making process generally includes picking, withering, rolling, fermentation, and drying. Different tea categories use different processing methods in these steps to produce different flavors and characteristics. Tea grading is usually based on multiple factors such as appearance, aroma, taste, and origin. Different regions and types of tea may have different grading standards. Below are some common tea grading methods.

[0003] At present, there are still some challenges and shortcomings in the field of tea quality assessment. Traditional tea quality assessment usually relies on experienced professionals, which has problems such as strong subjectivity and low efficiency. It is easily affected by subjective factors, and the consistency and accuracy of the assessment results are limited. In addition, traditional methods are less efficient and do not easily meet the needs of the rapid development of the modern tea industry. Summary of the Invention

[0004] To address the shortcomings of existing technologies, this invention provides a machine vision-based intelligent identification and grading method and system for tea, solving the problems mentioned in the background section.

[0005] To achieve the above objectives, the present invention is implemented through the following technical solution: including an image acquisition and preprocessing module, a feature extraction module, a model training and optimization module, a feature calculation module, and a tea grading and evaluation module;

[0006] The image acquisition and preprocessing module is used to acquire images of dry tea using an image acquisition device, and at the same time acquire images of tea leaves in the cup after brewing, and preprocess the acquired images of dry tea and brewed tea.

[0007] The feature extraction module is used to extract the dry tea feature set and wet tea feature set from the preprocessed dry tea image and the brewed image, and convert the image into a data form that can be understood and processed by the computer.

[0008] The model training and optimization module uses machine vision technology to build a tea classification model and associates the extracted features with the quality grade of the tea.

[0009] The feature calculation module is used to perform summary calculations based on the extracted dry tea feature set and wet tea feature set to obtain the comprehensive dry tea coefficient Gcxs and comprehensive wet tea coefficient Scxs. At the same time, the obtained comprehensive dry tea coefficient Gcxs and comprehensive wet tea coefficient Scxs are processed without dimensions and then correlated to obtain the comprehensive grade coefficient Pjxs.

[0010] The tea grading and evaluation module is used to perform a preliminary comparison and evaluation between the preset dry tea evaluation threshold G and the obtained comprehensive dry tea coefficient Gcxs, generate corresponding evaluation results, and activate the second evaluation mechanism. The second evaluation mechanism is used to perform a secondary comparison and evaluation between the preset tea grade threshold C and the obtained comprehensive grade coefficient Pjxs, generate corresponding evaluation results, and finally generate relevant grade reports.

[0011] Preferably, the image acquisition and preprocessing module includes an image acquisition unit and an image preprocessing unit;

[0012] The image acquisition unit is used to take pictures of the tea leaves from five directions using a high-definition camera while they are in their dry state, and to take microscopic pictures of the tea hairs using a microscope camera to obtain a set of dry tea leaf images. After the dry tea images are acquired, the dry tea is placed in a transparent container of a specific capacity for brewing. After brewing, the tea is left to stand for one minute. Then, the brewed wet tea is photographed under both artificial and natural light to obtain a set of wet tea leaf images.

[0013] The image preprocessing unit performs quality checks on the acquired images, selects and retains clear and uniformly lit images, adjusts the size of the retained images to a uniform size, converts color images to grayscale images, removes the background of the images using image segmentation technology and retains only the tea leaves, and uses filters or other denoising techniques to remove noise from the images and performs histogram equalization on the images.

[0014] Preferably, the feature extraction module includes a dry tea feature extraction unit and a wet tea feature extraction unit;

[0015] The dry tea feature extraction unit uses a shape feature extraction method to perform edge detection, contour analysis, shape description and overall geometric structure of the appearance features and tea hairs of the dry tea, and classifies and summarizes them to generate a dry tea feature set, and then performs feature acquisition.

[0016] The dry tea feature set includes an appearance feature set and a tea hair feature set;

[0017] The set of appearance features includes blade length Wcd, blade area Wmj, blade defect area Wqs, blade volume Wtj, and blade width Wkd;

[0018] The tea hair feature set includes tea hair distribution density Hmd, tea hair length Hcd, tea hair area Hmj, tea hair color factor Hys, and tea hair volume Htj;

[0019] The wet tea feature extraction unit uses color feature extraction technology, texture feature extraction technology, and local feature extraction technology to extract features from the brewed tea image, classify and summarize them to generate a wet tea feature set, and then collect features.

[0020] The wet tea feature set includes the tea infusion feature set and the tea leaf infusion feature set;

[0021] The tea infusion feature set includes tea infusion transparency Ttm, tea infusion bubble count Tqp, tea infusion gloss Tgz, tea infusion sediment concentration Tcd, and tea infusion hue angle Tsx;

[0022] The leaf base feature set includes the tea leaf floating height Dsf, leaf base defect area Dqs, leaf base color uniformity Djy, leaf base volume uniformity Dtj, and leaf base cleanliness Dqj;

[0023] Finally, the features obtained from each feature extraction method are combined into a feature vector.

[0024] Preferably, the model training and optimization module includes a model training unit and a model optimization unit;

[0025] The model training unit is used to collect a large number of relevant feature images from the Internet as a training set to train the tea classification model. The extracted dry tea feature set and wet tea feature set are normalized and set as validation set and test set, respectively, to adjust model parameters, monitor model performance, and evaluate the generalization ability and accuracy of the tea classification model. Support vector machine and random forest are selected from machine learning models as the framework for building the tea classification model. The selected model is trained using the data in the training set. By optimizing the loss function and adjusting the model parameters, the trained model is evaluated using the validation set to monitor the model's performance indicators. Finally, the optimized model is tested using the test set to evaluate the model's generalization ability and accuracy.

[0026] The model optimization unit enhances the training data, expands the dataset size, and optimizes the generalization ability and robustness of the tea classification model. It uses L1 and L2 regularization to control the model's complexity, employs ensemble learning techniques to combine the prediction results of multiple models, uses weighted averaging and voting methods for model fusion, and uses cross-validation to evaluate the model's performance. The unit then selects the best model, merges the prediction results of the best model, continuously monitors the model's performance, and adjusts and optimizes it according to actual conditions.

[0027] Preferably, the feature calculation module includes a dry tea calculation unit, a wet tea calculation unit, and a comprehensive calculation unit;

[0028] The dry tea calculation unit includes an appearance feature calculation unit and a tea hair feature calculation unit;

[0029] The appearance feature calculation unit is used to perform dimensionless processing on the appearance feature set in the dry tea feature set, and then summarize and calculate to obtain the appearance feature coefficient Wgxs.

[0030] The tea hair feature calculation unit is used to perform dimensionless processing on the tea hair feature set in the dry tea feature set, and then summarize and calculate to obtain the tea hair feature coefficient Chxs.

[0031] The appearance feature coefficient Wgxs and the tea hair feature coefficient Chxs are obtained by the following formula;

[0032] ;

[0033] .

[0034] Preferably, the wet tea calculation unit includes a tea infusion characteristic calculation unit and a leaf residue characteristic calculation unit;

[0035] The tea infusion feature calculation unit is used to analyze and calculate the tea infusion feature coefficient Ctxs based on the tea infusion feature set after dimensionless processing.

[0036] The leaf base feature calculation unit is used to analyze and calculate the leaf base feature coefficient Ydxs based on the leaf base feature set after dimensionless processing.

[0037] The tea infusion characteristic coefficient Ctxs and the leaf infusion characteristic coefficient Ydxs are obtained by the following formula;

[0038] ;

[0039] .

[0040] Preferably, the integrated calculation unit includes a dry tea correlation unit, a wet tea correlation calculation unit, and an integrated correlation calculation unit;

[0041] The dry tea associated unit obtains the comprehensive dry tea coefficient Gcxs by dimensionlessly processing the appearance feature coefficient Wgxs and tea hair feature coefficient Chxs obtained by the dry tea calculation unit and summarizing them.

[0042] The wet tea related calculation unit obtains the comprehensive wet tea coefficient Scxs by performing dimensionless processing on the tea soup characteristic coefficient Ctxs and leaf bottom characteristic coefficient Ydxs obtained by the dry tea calculation unit and summarizing them.

[0043] The comprehensive dry tea coefficient Gcxs and comprehensive wet tea coefficient Scxs are obtained by the following formulas;

[0044] Gcxs = [ ( Wgxs*a1 ) + ( Chxs*a2 ) ] + A ;

[0045] Scxs = [ ( Ctxs*b1 ) + ( Ydxs*b2 ) ] + B ;

[0046] In the formula, a1 and a2 represent the preset proportional coefficients of appearance feature coefficient Wgxs and tea hair feature coefficient Chxs, and a1≠a2, 0<a1<0.47, 0<a2<0.53. Their specific values ​​are adjusted and set by the user, and A represents the first correction constant.

[0047] b1 and b2 represent the preset ratio coefficients of the tea soup characteristic coefficient Ctxs and the leaf bottom characteristic coefficient Ydxs, and b1≠b2, 0<b1<0.74, 0<b2<0.64. Their specific values ​​are adjusted and set by the user. B represents the second correction constant.

[0048] The comprehensive correlation calculation unit is used to summarize and analyze the obtained comprehensive dry tea coefficient Gcxs and comprehensive wet tea coefficient Scxs after dimensionless processing, and generate comprehensive grade coefficient Pjxs.

[0049] The comprehensive grade coefficient Pjxs is obtained by the following formula;

[0050] .

[0051] Preferably, the tea grading and evaluation module includes a dry tea evaluation unit and a comprehensive evaluation unit;

[0052] The dry tea evaluation unit is used to perform a preliminary comparative evaluation of the comprehensive dry tea coefficient Gcxs obtained by the dry tea correlation unit analysis and calculation, based on the preset dry tea evaluation threshold G, and to obtain the corresponding evaluation results. The specific evaluation scheme is as follows:

[0053] When the comprehensive dry tea coefficient Gcxs > the dry tea evaluation threshold G, it indicates that the tea leaves detected are irregular in shape and have torn and damaged. At this time, the current tea is directly defined as "Grade 3 tea", and the contents are summarized to generate the first report.

[0054] When the comprehensive dry tea coefficient Gcxs = dry tea evaluation threshold G, it means that the tea leaves detected are regular in shape and uniform in volume, with intact and undamaged tea hairs, plump buds without damage, and uniform tea hairs. At this time, the second evaluation mechanism is activated.

[0055] When the comprehensive dry tea coefficient Gcxs < the dry tea evaluation threshold G, it indicates that the currently detected tea leaves are fragmented, the tea leaves are not complete as a whole, the tea hairs are unevenly distributed, and the density of tea hairs is less than 20% of the tea leaf volume. At this time, the current tea is directly defined as "Grade 4 tea", and the contents are summarized to generate a second report.

[0056] Preferably, the comprehensive evaluation unit triggers the second evaluation mechanism due to the tea evaluation unit. The comprehensive evaluation unit is used to compare and evaluate the preset tea grade threshold C with the comprehensive grade coefficient Pjxs obtained by the correlation analysis of the comprehensive correlation calculation unit, and generate the corresponding grade. The specific evaluation scheme is as follows:

[0057] When the comprehensive grade coefficient Pjxs > the tea grade threshold C, it means that the color of the tea soup that is brewed is usually bright and clear without sediment, and the tea leaves sink to the bottom of the water and have a uniform color and volume. At this time, the current tea is defined as "Grade 1 tea", and the contents are summarized to generate a third report.

[0058] When the comprehensive grade coefficient Pjxs ≤ the tea grade threshold C, it means that the color of the tea soup brewed is dull and there is sediment. At the same time, the tea leaves are damaged and uneven when fully brewed. In this case, the current tea is defined as "second-level tea", and the contents are summarized to generate the fourth report.

[0059] A machine vision-based intelligent tea identification and grading method includes the following steps:

[0060] S1. First, use an image acquisition device to take pictures of the dry tea and the brewed tea. Then, perform quality checks on the acquired dry tea images and the brewed tea images, retain clear and evenly lit images, adjust the image size, convert the color images to grayscale images, use image segmentation technology to remove the background, remove noise, and perform histogram equalization.

[0061] S2. Extract features from the preprocessed dry tea image and the brewed wet tea image, including the appearance features and tea hair features of the dry tea, as well as the tea soup features and leaf base features of the wet tea image. Perform dimensionless processing on the features and convert them into a data form that can be understood and processed by computers.

[0062] S3. By constructing a tea classification model and associating the extracted features with the quality grade of tea, the model is trained and optimized using labeled tea image data. Appropriate machine learning algorithms are selected, including support vector machines and random forests. The model performance is evaluated through cross-validation techniques, and the model parameters are adjusted and the model is optimized.

[0063] S4. Based on the extracted dry tea characteristics and wet tea characteristics, calculate the comprehensive dry tea coefficient Gcxs and the comprehensive wet tea coefficient Scxs, perform dimensionless processing, and perform correlation calculation to obtain the comprehensive grade coefficient Pjxs.

[0064] S5. Finally, a preliminary evaluation of the dry tea is conducted. The evaluation threshold G for dry tea is set and compared with the comprehensive dry tea coefficient Gcxs to determine the preliminary grade of the tea. If several teas pass the preliminary evaluation, the second evaluation mechanism is activated. The grade threshold C for tea is set and compared with the comprehensive grade coefficient Pjxs to determine the final grade of the tea.

[0065] The beneficial effects of this invention are:

[0066] (1) This system first takes pictures of tea leaves, including dry tea and brewed tea, using an image acquisition device. Then, it preprocesses the images using image processing technology to extract the characteristics of the tea leaves. Next, it uses machine learning algorithms to build a tea classification model and associates the extracted features with the quality grade of the tea leaves to achieve automatic tea classification. Finally, it calculates the comprehensive dry tea coefficient Gcxs and the comprehensive wet tea coefficient Scxs, and then correlates the comprehensive dry tea coefficient Gcxs and the comprehensive wet tea coefficient Scxs to obtain the comprehensive grade coefficient Pjxs. This allows for multi-level grading of the tea leaves to determine their quality grade. This system not only improves the objectivity and accuracy of tea grading but also increases the efficiency of the assessment, bringing new opportunities for the development of the tea industry. At the same time, the system uses machine vision technology to automatically acquire tea leaf images and perform preprocessing, feature extraction, model training, and optimization, ultimately achieving intelligent identification and grading of tea leaves. Compared with traditional manual grading methods, this system greatly improves the efficiency and accuracy of identification and reduces labor costs and time consumption.

[0067] (2) The system has continuous monitoring and optimization functions, and can continuously collect data and adjust the model to adapt to the characteristics of tea produced and processed under different environments and conditions. By monitoring the identification and grading process of tea, the system can promptly detect the performance decline or misclassification of the model and take corresponding measures to optimize and improve it. Real-time feedback is crucial for tea producers and tea buyers. By understanding the quality of tea in a timely manner, they can react quickly. The system also provides broader value to the tea industry. By collecting and analyzing a large amount of tea data, the system can help tea producers and buyers better understand market trends and consumer preferences, thereby optimizing product mix and improving market competitiveness. In short, the system's continuous monitoring, optimization and real-time feedback functions can not only improve the efficiency and quality of tea production and processing, but also contribute to the sustainable development of the tea industry and the enhancement of market competitiveness. Attached Figure Description

[0068] Figure 1 This is a schematic diagram of the process of a machine vision-based intelligent tea identification and grading system according to the present invention.

[0069] Figure 2 This is a schematic diagram illustrating the steps of a machine vision-based intelligent tea identification and grading method according to the present invention. Detailed Implementation

[0070] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0071] Example 1

[0072] Please see Figure 1 This invention provides a machine vision-based intelligent tea identification and grading system. To achieve the above objectives, this invention is implemented through the following technical solutions: including an image acquisition and preprocessing module, a feature extraction module, a model training and optimization module, a feature calculation module, and a tea grading and evaluation module.

[0073] The image acquisition and preprocessing module is used to acquire images of dry tea using an image acquisition device, and at the same time acquire images of tea leaves in the cup after brewing, and preprocess the acquired images of dry tea and brewed tea.

[0074] The feature extraction module is used to extract the dry tea feature set and wet tea feature set from the preprocessed dry tea image and the brewed image, and convert the image into a data form that can be understood and processed by the computer.

[0075] The model training and optimization module uses machine vision technology to build a tea classification model and associates the extracted features with the quality grade of the tea.

[0076] The feature calculation module is used to perform summary calculations based on the extracted dry tea feature set and wet tea feature set to obtain the comprehensive dry tea coefficient Gcxs and comprehensive wet tea coefficient Scxs. At the same time, the obtained comprehensive dry tea coefficient Gcxs and comprehensive wet tea coefficient Scxs are processed without dimensions and then correlated to obtain the comprehensive grade coefficient Pjxs.

[0077] The tea grading and evaluation module is used to perform a preliminary comparison and evaluation between the preset dry tea evaluation threshold G and the obtained comprehensive dry tea coefficient Gcxs, generate corresponding evaluation results, and activate the second evaluation mechanism. The second evaluation mechanism is used to perform a secondary comparison and evaluation between the preset tea grade threshold C and the obtained comprehensive grade coefficient Pjxs, generate corresponding evaluation results, and finally generate relevant grade reports.

[0078] In this embodiment, the system effectively captures images of dry tea and brewed wet tea through an image acquisition and preprocessing module, and preprocesses them to lay a solid foundation for subsequent feature extraction and model training. The feature extraction module uses advanced algorithms and technologies to extract dry tea feature sets and wet tea feature sets from the preprocessed images, providing important data support for subsequent model training and evaluation. The model training and optimization module constructs an efficient and accurate tea classification model through machine vision technology, and associates the extracted features with the quality grade of the tea, thereby realizing automatic classification and grading of tea. The completion of these tasks effectively improves the accuracy and efficiency of tea identification and grading, and significantly improves work efficiency and accuracy compared with traditional manual grading methods.

[0079] Furthermore, the feature calculation module comprehensively calculates the extracted dry tea feature sets and wet tea feature sets to obtain the comprehensive dry tea coefficient Gcxs and the comprehensive wet tea coefficient Scxs. After dimensionless processing, the comprehensive grade coefficient Pjxs is obtained, providing a scientific basis for accurate tea grading. Meanwhile, the tea grading and evaluation module performs preliminary and secondary quality assessments of the tea through preset evaluation thresholds and comparison mechanisms, ensuring the accuracy and reliability of the grading results. The completion of these tasks makes tea quality evaluation more objective and standardized, providing a more reliable reference for tea production and sales.

[0080] This system fully utilizes advanced image processing technology and machine learning algorithms to achieve automated identification, grading, and evaluation of tea. Compared with traditional manual methods, this system greatly improves the efficiency and quality of tea grade identification, reduces labor costs and human error, and brings significant improvements and enhancements to the development of the tea industry.

[0081] Example 2

[0082] This embodiment is an explanation based on Embodiment 1. Please refer to it. Figure 1 Specifically: the image acquisition and preprocessing module includes an image acquisition unit and an image preprocessing unit;

[0083] The image acquisition unit is used to take pictures of the tea leaves from five directions using a high-definition camera while they are in their dry state, and to take microscopic pictures of the tea hairs using a microscope camera to obtain a set of dry tea leaf images. After the dry tea images are acquired, the dry tea is placed in a transparent container of a specific capacity for brewing. After brewing, the tea is left to stand for one minute. Then, the brewed wet tea is photographed under both artificial and natural light to obtain a set of wet tea leaf images.

[0084] The image preprocessing unit performs quality checks on the acquired images, selects and retains clear and uniformly lit images, adjusts the size of the retained images to a uniform size, converts color images to grayscale images, removes the background of the images using image segmentation technology and retains only the tea leaves, and uses filters or other denoising techniques to remove noise from the images and performs histogram equalization on the images.

[0085] The feature extraction module includes a dry tea feature extraction unit and a wet tea feature extraction unit;

[0086] The dry tea feature extraction unit uses a shape feature extraction method to perform edge detection, contour analysis, shape description and overall geometric structure of the appearance features and tea hairs of the dry tea, and classifies and summarizes them to generate a dry tea feature set, and then performs feature acquisition.

[0087] The dry tea feature set includes an appearance feature set and a tea hair feature set;

[0088] The set of appearance features includes blade length Wcd, blade area Wmj, blade defect area Wqs, blade volume Wtj, and blade width Wkd;

[0089] The tea hair feature set includes tea hair distribution density Hmd, tea hair length Hcd, tea hair area Hmj, tea hair color factor Hys, and tea hair volume Htj;

[0090] The wet tea feature extraction unit uses color feature extraction technology, texture feature extraction technology, and local feature extraction technology to extract features from the brewed tea image, classify and summarize them to generate a wet tea feature set, and then collect features.

[0091] The wet tea feature set includes the tea infusion feature set and the tea leaf infusion feature set;

[0092] The tea infusion feature set includes tea infusion transparency Ttm, tea infusion bubble count Tqp, tea infusion gloss Tgz, tea infusion sediment concentration Tcd, and tea infusion hue angle Tsx;

[0093] The leaf base feature set includes the tea leaf floating height Dsf, leaf base defect area Dqs, leaf base color uniformity Djy, leaf base volume uniformity Dtj, and leaf base cleanliness Dqj;

[0094] Finally, the features obtained from each feature extraction method are combined into a feature vector.

[0095] In this embodiment, the system achieves comprehensive feature capture of dry and wet tea through an image acquisition and preprocessing module and a feature extraction module. The image acquisition unit uses a high-definition camera and a microscope camera to capture tea leaves from multiple angles, ensuring the comprehensiveness and accuracy of the images. The image preprocessing unit effectively improves the clarity and quality of the images through quality checks, size adjustments, background removal, and noise removal, providing a good data foundation for subsequent feature extraction. The feature extraction module uses shape feature extraction methods and various feature extraction techniques to accurately extract and analyze the appearance features, tea hair features, tea soup features, and leaf base features of dry and wet tea, forming a comprehensive feature set. The extraction of these features lays a solid foundation for subsequent model training and tea grading, significantly improving the system's recognition accuracy and grading efficiency.

[0096] The advantages of this system lie in its feature extraction module, which employs various advanced feature extraction technologies, such as shape feature extraction, color feature extraction, and texture feature extraction. This allows for the comprehensive and multi-faceted capture of tea's characteristic information. Compared to traditional manual extraction methods, this automated feature extraction method is more efficient and accurate, significantly saving time and labor costs. Furthermore, it ensures the consistency and objectivity of features. In addition, the use of the image acquisition and preprocessing module makes the image acquisition and processing process more standardized and regulated, further improving the system's stability and reliability. This provides more reliable technical support for the application of intelligent tea identification and grading systems.

[0097] Example 3

[0098] This embodiment is an explanation based on Embodiment 2. Please refer to it. Figure 1 Specifically: the model training and optimization module includes a model training unit and a model optimization unit;

[0099] The model training unit is used to collect a large number of relevant feature images from the Internet as a training set to train the tea classification model. The extracted dry tea feature set and wet tea feature set are normalized and set as validation set and test set, respectively, to adjust model parameters, monitor model performance, and evaluate the generalization ability and accuracy of the tea classification model. Support vector machine and random forest are selected from machine learning models as the framework for building the tea classification model. The selected model is trained using the data in the training set. By optimizing the loss function and adjusting the model parameters, the trained model is evaluated using the validation set to monitor the model's performance indicators. Finally, the optimized model is tested using the test set to evaluate the model's generalization ability and accuracy.

[0100] The model optimization unit enhances the training data, expands the dataset size, and optimizes the generalization ability and robustness of the tea classification model. It uses L1 and L2 regularization to control the model's complexity, employs ensemble learning techniques to combine the prediction results of multiple models, uses weighted averaging and voting methods for model fusion, and uses cross-validation to evaluate the model's performance. The unit then selects the best model, merges the prediction results of the best model, continuously monitors the model's performance, and adjusts and optimizes it according to actual conditions.

[0101] In this embodiment, the system comprehensively utilizes modern machine learning techniques for model training and optimization, resulting in significant improvements in the accuracy, generalization ability, and robustness of the intelligent tea identification and grading system. Simultaneously, continuous monitoring and optimization ensure the system's stability and reliability under different environments, providing tea producers and processors with more reliable data for tea quality assessment and production process improvement, further promoting the development and advancement of the tea industry.

[0102] Example 4

[0103] This embodiment is an explanation based on Embodiment 2. Please refer to it. Figure 1 Specifically: the feature calculation module includes a dry tea calculation unit, a wet tea calculation unit, and a comprehensive calculation unit;

[0104] The dry tea calculation unit includes an appearance feature calculation unit and a tea hair feature calculation unit;

[0105] The appearance feature calculation unit is used to perform dimensionless processing on the appearance feature set in the dry tea feature set, and then summarize and calculate to obtain the appearance feature coefficient Wgxs.

[0106] The tea hair feature calculation unit is used to perform dimensionless processing on the tea hair feature set in the dry tea feature set, and then summarize and calculate to obtain the tea hair feature coefficient Chxs.

[0107] The appearance feature coefficient Wgxs and the tea hair feature coefficient Chxs are obtained by the following formula;

[0108] ;

[0109] .

[0110] The wet tea calculation unit includes a tea infusion characteristic calculation unit and a leaf bottom characteristic calculation unit;

[0111] The tea infusion feature calculation unit is used to analyze and calculate the tea infusion feature coefficient Ctxs based on the tea infusion feature set after dimensionless processing.

[0112] The leaf base feature calculation unit is used to analyze and calculate the leaf base feature coefficient Ydxs based on the leaf base feature set after dimensionless processing.

[0113] The tea infusion characteristic coefficient Ctxs and the leaf infusion characteristic coefficient Ydxs are obtained by the following formula;

[0114] ;

[0115] .

[0116] In this embodiment, the feature calculation module, through the combined action of the dry tea calculation unit, the wet tea calculation unit, and the comprehensive calculation unit, provides a comprehensive and accurate basis for tea quality evaluation. First, the dry tea calculation unit, based on appearance characteristics and tea hair characteristics, calculates and obtains the appearance characteristic coefficient Wgxs and the tea hair characteristic coefficient Chxs, effectively quantifying the appearance quality and tea hair condition of the tea, providing basic data for subsequent comprehensive quality evaluation. Second, the wet tea calculation unit, based on the tea soup characteristics and leaf bottom characteristics, calculates and obtains the tea soup characteristic coefficient Ctxs and the leaf bottom characteristic coefficient Ydxs, fully considering the appearance and bottom condition of the tea after brewing, providing important parameters for comprehensive evaluation of tea quality.

[0117] Example 5

[0118] This embodiment is an explanation based on Embodiment 4. Please refer to it. Figure 1 Specifically: the integrated calculation unit includes a dry tea correlation unit, a wet tea correlation calculation unit, and an integrated correlation calculation unit;

[0119] The dry tea associated unit obtains the comprehensive dry tea coefficient Gcxs by dimensionlessly processing the appearance feature coefficient Wgxs and tea hair feature coefficient Chxs obtained by the dry tea calculation unit and summarizing them.

[0120] The wet tea related calculation unit obtains the comprehensive wet tea coefficient Scxs by performing dimensionless processing on the tea soup characteristic coefficient Ctxs and leaf bottom characteristic coefficient Ydxs obtained by the dry tea calculation unit and summarizing them.

[0121] The comprehensive dry tea coefficient Gcxs and comprehensive wet tea coefficient Scxs are obtained by the following formulas;

[0122] Gcxs = [ ( Wgxs*a1 ) + ( Chxs*a2 ) ] + A ;

[0123] Scxs = [ ( Ctxs*b1 ) + ( Ydxs*b2 ) ] + B ;

[0124] In the formula, a1 and a2 represent the preset proportional coefficients of appearance feature coefficient Wgxs and tea hair feature coefficient Chxs, and a1≠a2, 0<a1<0.47, 0<a2<0.53. Their specific values ​​are adjusted and set by the user, and A represents the first correction constant.

[0125] b1 and b2 represent the preset ratio coefficients of the tea soup characteristic coefficient Ctxs and the leaf bottom characteristic coefficient Ydxs, and b1≠b2, 0<b1<0.74, 0<b2<0.64. Their specific values ​​are adjusted and set by the user. B represents the second correction constant.

[0126] The comprehensive correlation calculation unit is used to summarize and analyze the obtained comprehensive dry tea coefficient Gcxs and comprehensive wet tea coefficient Scxs after dimensionless processing, and generate comprehensive grade coefficient Pjxs.

[0127] The comprehensive grade coefficient Pjxs is obtained by the following formula;

[0128] .

[0129] In this embodiment, the introduction of the comprehensive calculation unit further enhances the accuracy and comprehensiveness of tea quality assessment. First, the dry tea correlation unit calculates the comprehensive dry tea coefficient Gcxs by summing the appearance characteristic coefficient Wgxs and the tea hair characteristic coefficient Chxs. This coefficient comprehensively considers the appearance and tea hair characteristics of the dry tea, and fully reflects the comprehensive evaluation of dry tea quality through preset proportional coefficients and correction constants. Second, the wet tea correlation calculation unit calculates the comprehensive wet tea coefficient Scxs by summing the tea soup characteristic coefficient Ctxs and the leaf residue characteristic coefficient Ydxs. This takes into account the tea soup and leaf residue after brewing, providing an important basis for the evaluation of wet tea quality. Finally, the comprehensive correlation calculation unit summarizes and analyzes the comprehensive dry tea coefficient Gcxs and the comprehensive wet tea coefficient Scxs to obtain the comprehensive grade coefficient Pjxs. This comprehensively considers various characteristics of both dry and wet tea, providing a comprehensive evaluation of the overall quality of the tea, making the evaluation results more comprehensive and objective. This comprehensive evaluation system not only helps tea producers and processors understand the quality of tea in a timely manner and make adjustments and improvements to the production process, but also provides consumers with a more scientific and objective purchasing reference, further enhancing the overall competitiveness of the tea industry.

[0130] Example 6

[0131] This embodiment is an explanation based on Embodiment 1. Please refer to it. Figure 1 Specifically: the tea grading and evaluation module includes a dry tea evaluation unit and a comprehensive evaluation unit;

[0132] The dry tea evaluation unit is used to perform a preliminary comparative evaluation of the comprehensive dry tea coefficient Gcxs obtained by the dry tea correlation unit analysis and calculation, based on the preset dry tea evaluation threshold G, and to obtain the corresponding evaluation results. The specific evaluation scheme is as follows:

[0133] When the comprehensive dry tea coefficient Gcxs > the dry tea evaluation threshold G, it indicates that the tea leaves detected are irregular in shape and have torn and damaged. At this time, the current tea is directly defined as "Grade 3 tea", and the contents are summarized to generate the first report.

[0134] When the comprehensive dry tea coefficient Gcxs = dry tea evaluation threshold G, it means that the tea leaves detected are regular in shape and uniform in volume, with intact and undamaged tea hairs, plump buds without damage, and uniform tea hairs. At this time, the second evaluation mechanism is activated.

[0135] When the comprehensive dry tea coefficient Gcxs < the dry tea evaluation threshold G, it indicates that the currently detected tea leaves are fragmented, the tea leaves are not complete as a whole, the tea hairs are unevenly distributed, and the density of tea hairs is less than 20% of the tea leaf volume. At this time, the current tea is directly defined as "Grade 4 tea", and the contents are summarized to generate a second report.

[0136] The comprehensive evaluation unit triggers the second evaluation mechanism due to the tea evaluation unit. The comprehensive evaluation unit is used to compare and evaluate the preset tea grade threshold C with the comprehensive grade coefficient Pjxs obtained by the comprehensive correlation calculation unit, and generate the corresponding grade. The specific evaluation scheme is as follows.

[0137] When the comprehensive grade coefficient Pjxs > the tea grade threshold C, it means that the color of the tea soup that is brewed is usually bright and clear without sediment, and the tea leaves sink to the bottom of the water and have a uniform color and volume. At this time, the current tea is defined as "Grade 1 tea", and the contents are summarized to generate a third report.

[0138] When the comprehensive grade coefficient Pjxs ≤ the tea grade threshold C, it means that the color of the tea soup brewed is dull and there is sediment. At the same time, the tea leaves are damaged and uneven when fully brewed. In this case, the current tea is defined as "second-level tea", and the contents are summarized to generate the fourth report.

[0139] In this embodiment, the introduction of a tea grading and evaluation module provides a more accurate and comprehensive means for evaluating tea quality. First, the dry tea evaluation unit performs a preliminary evaluation by comparing a preset dry tea evaluation threshold G with a comprehensive dry tea coefficient Gcxs, classifying the tea into different grades. This evaluation scheme considers the shape and integrity of the tea leaves, as well as the condition of the tea hairs, enabling rapid assessment of tea quality. This helps producers adjust their processes in a timely manner during production, improving the overall quality of the tea. Second, the comprehensive evaluation unit, through a second evaluation mechanism, compares a preset tea grade threshold C with a comprehensive grade coefficient Pjxs, further refining the tea grading. This evaluation scheme considers the color and clarity of the brewed tea liquor, as well as the condition of the tea leaves, providing consumers with a more objective and scientific purchasing reference, and improving the transparency and fairness of the tea market. In summary, the introduction of the tea grading and evaluation module not only improves the efficiency and accuracy of the tea production process but also enhances consumers' confidence and recognition of tea quality, promoting the healthy development of the tea industry.

[0140] Example 7

[0141] Please see Figure 1 and Figure 2 A machine vision-based intelligent tea identification and grading method includes the following steps:

[0142] S1. First, use an image acquisition device to take pictures of the dry tea and the brewed tea. Then, perform quality checks on the acquired dry tea images and the brewed tea images, retain clear and evenly lit images, adjust the image size, convert the color images to grayscale images, use image segmentation technology to remove the background, remove noise, and perform histogram equalization.

[0143] S2. Extract features from the preprocessed dry tea image and the brewed wet tea image, including the appearance features and tea hair features of the dry tea, as well as the tea soup features and leaf base features of the wet tea image. Perform dimensionless processing on the features and convert them into a data form that can be understood and processed by computers.

[0144] S3. By constructing a tea classification model and associating the extracted features with the quality grade of tea, the model is trained and optimized using labeled tea image data. Appropriate machine learning algorithms are selected, including support vector machines and random forests. The model performance is evaluated through cross-validation techniques, and the model parameters are adjusted and the model is optimized.

[0145] S4. Based on the extracted dry tea characteristics and wet tea characteristics, calculate the comprehensive dry tea coefficient Gcxs and the comprehensive wet tea coefficient Scxs, perform dimensionless processing, and perform correlation calculation to obtain the comprehensive grade coefficient Pjxs.

[0146] S5. Finally, a preliminary evaluation of the dry tea is conducted. The evaluation threshold G for dry tea is set and compared with the comprehensive dry tea coefficient Gcxs to determine the preliminary grade of the tea. If several teas pass the preliminary evaluation, the second evaluation mechanism is activated. The grade threshold C for tea is set and compared with the comprehensive grade coefficient Pjxs to determine the final grade of the tea.

[0147] Specific examples:

[0148] Blade length Wcd: 7.94, blade area Wmj: 5.27, blade defect area Wqs: 2.11, blade volume Wtj: 8.83, blade width Wkd: 3.76;

[0149] Tea hair distribution density Hmd: 9.45, tea hair length Hcd: 1.59, tea hair area Hmj: 4.38, tea hair color factor Hys: 6.72, tea hair volume Htj: 8.15;

[0150] Tea liquor transparency Ttm: 5.69, tea liquor bubble count Tqp: 3.21, tea liquor gloss Tgz: 6.99, tea liquor sediment concentration Tcd: 4.72, tea liquor hue angle Tsx: 2.33;

[0151] Tea leaf floating height Dsf: 1.94, leaf base defect area Dqs: 3.98, leaf base color uniformity Djy: 2.83, leaf base volume uniformity Dtj: 4.56, leaf base cleanliness Dqj: 5.84;

[0152] All the above data are dimensionless. The collected data will be substituted into the following formula for calculation:

[0153] Appearance characteristic coefficient Wgxs and tea hair characteristic coefficient Chxs:

[0154] ;

[0155] ;

[0156] Tea infusion characteristic coefficient Ctxs and leaf residue characteristic coefficient Ydxs:

[0157] ;

[0158] ;

[0159] Overall dry tea coefficient Gcxs and overall wet tea coefficient Scxs:

[0160] Gcxs = [ ( 0.58*0.45 ) + ( 0.29*0.41 ) ] + 0.04 = 0.3 ;

[0161] Scxs = [ ( 7.28*0.16 ) + ( 4.28*0.09 ) ] + 0.1 = 1.5 ;

[0162] The proportionality constants are a1=0.45, a2=0.41, b1=0.16, b2=0.09; the first correction constant A is 0.04; and the second correction constant B is 0.1.

[0163] Comprehensive grade coefficient Pjxs:

[0164] ;

[0165] All calculation results are rounded to two decimal places.

[0166] Set the dry tea evaluation threshold G to 0.3 and the tea grade threshold C to 1. At this time, the comprehensive dry tea coefficient Gcxs = dry tea evaluation threshold G, which means that the tea leaves detected are regular in shape and uniform in volume. At the same time, the tea hairs are intact and the buds are plump without damage, and the tea hairs are uniform. At this time, the second evaluation mechanism is activated.

[0167] If the comprehensive grade coefficient Pjxs > the tea grade threshold C, it means that the color of the tea soup that is brewed is usually bright and clear without sediment, and the tea leaves sink to the bottom of the water and have a uniform color and volume. In this case, the current tea is defined as "Grade 1 tea".

[0168] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A machine vision-based intelligent tea identification and grading system, characterized in that: It includes an image acquisition and preprocessing module, a feature extraction module, a model training and optimization module, a feature calculation module, and a tea grading and evaluation module; The image acquisition and preprocessing module is used to acquire images of dry tea using an image acquisition device, and at the same time acquire images of tea leaves in the cup after brewing, and preprocess the acquired images of dry tea and brewed tea. The feature extraction module is used to extract the dry tea feature set and wet tea feature set from the preprocessed dry tea image and the brewed image, and convert the image into a data form that can be understood and processed by the computer. The feature extraction module includes a dry tea feature extraction unit and a wet tea feature extraction unit; The dry tea feature extraction unit uses a shape feature extraction method to perform edge detection, contour analysis, shape description and overall geometric structure of the appearance features and tea hairs of the dry tea, and classifies and summarizes them to generate a dry tea feature set, and then performs feature acquisition. The dry tea feature set includes an appearance feature set and a tea hair feature set; The set of appearance features includes blade length Wcd, blade area Wmj, blade defect area Wqs, blade volume Wtj, and blade width Wkd; The tea hair feature set includes tea hair distribution density Hmd, tea hair length Hcd, tea hair area Hmj, tea hair color factor Hys, and tea hair volume Htj; The wet tea feature extraction unit uses color feature extraction technology, texture feature extraction technology, and local feature extraction technology to extract features from the brewed tea image, classify and summarize them to generate a wet tea feature set, and then collect features. The wet tea feature set includes the tea infusion feature set and the tea leaf infusion feature set; The tea infusion feature set includes tea infusion transparency Ttm, tea infusion bubble count Tqp, tea infusion gloss Tgz, tea infusion sediment concentration Tcd, and tea infusion hue angle Tsx; The leaf base feature set includes the tea leaf floating height Dsf, leaf base defect area Dqs, leaf base color uniformity Djy, leaf base volume uniformity Dtj, and leaf base cleanliness Dqj; Finally, the features obtained from each feature extraction method are combined into a feature vector; The model training and optimization module uses machine vision technology to build a tea classification model and associates the extracted features with the quality grade of the tea. The feature calculation module is used to perform summary calculations based on the extracted dry tea feature set and wet tea feature set to obtain the comprehensive dry tea coefficient Gcxs and comprehensive wet tea coefficient Scxs. At the same time, the obtained comprehensive dry tea coefficient Gcxs and comprehensive wet tea coefficient Scxs are processed without dimensions and then correlated to obtain the comprehensive grade coefficient Pjxs. The feature calculation module includes a dry tea calculation unit, a wet tea calculation unit, and a comprehensive calculation unit; The dry tea calculation unit includes an appearance feature calculation unit and a tea hair feature calculation unit; The appearance feature calculation unit is used to perform dimensionless processing on the appearance feature set in the dry tea feature set, and then summarize and calculate to obtain the appearance feature coefficient Wgxs. The tea hair feature calculation unit is used to perform dimensionless processing on the tea hair feature set in the dry tea feature set, and then summarize and calculate to obtain the tea hair feature coefficient Chxs. The appearance feature coefficient Wgxs and the tea hair feature coefficient Chxs are obtained by the following formula; ； ； The wet tea calculation unit includes a tea infusion characteristic calculation unit and a leaf bottom characteristic calculation unit; The tea infusion feature calculation unit is used to analyze and calculate the tea infusion feature coefficient Ctxs based on the tea infusion feature set after dimensionless processing. The leaf base feature calculation unit is used to analyze and calculate the leaf base feature coefficient Ydxs based on the leaf base feature set after dimensionless processing. The tea infusion characteristic coefficient Ctxs and the leaf infusion characteristic coefficient Ydxs are obtained by the following formula; ； ； The tea grading and evaluation module is used to perform a preliminary comparison and evaluation between the preset dry tea evaluation threshold G and the obtained comprehensive dry tea coefficient Gcxs, generate corresponding evaluation results, and activate the second evaluation mechanism. The second evaluation mechanism is used to perform a secondary comparison and evaluation between the preset tea grade threshold C and the obtained comprehensive grade coefficient Pjxs, generate corresponding evaluation results, and finally generate relevant grade reports.

2. The intelligent tea identification and grading system based on machine vision according to claim 1, characterized in that: The image acquisition and preprocessing module includes an image acquisition unit and an image preprocessing unit; The image acquisition unit is used to take pictures of the tea leaves from five directions using a high-definition camera and to take microscopic pictures of the tea hairs using a microscope camera, thereby obtaining a set of images of the dried tea leaves. After the dry tea images are acquired, the dry tea is placed in a transparent container and brewed. After brewing, it is left to stand for one minute. Then, the brewed wet tea is photographed under both artificial and natural light to obtain a set of wet tea images. The image preprocessing unit performs quality checks on the acquired images, selects and retains clear and uniformly lit images, adjusts the size of the retained images to a uniform size, converts color images to grayscale images, removes the background of the images using image segmentation technology and retains only the tea leaves, and uses filters or other denoising techniques to remove noise from the images and performs histogram equalization on the images.

3. The intelligent tea identification and grading system based on machine vision according to claim 2, characterized in that: The model training and optimization module includes a model training unit and a model optimization unit; The model training unit is used to collect a large number of relevant feature images from the Internet as a training set to train the tea classification model. The extracted dry tea feature set and wet tea feature set are normalized and set as validation set and test set, respectively, to adjust model parameters, monitor model performance, and evaluate the generalization ability and accuracy of the tea classification model. Support vector machine and random forest are selected from machine learning models as the framework for building the tea classification model. The selected model is trained using the data in the training set. By optimizing the loss function and adjusting the model parameters, the trained model is evaluated using the validation set to monitor the model's performance indicators. Finally, the optimized model is tested using the test set to evaluate the model's generalization ability and accuracy. The model optimization unit enhances the training data, expands the dataset size, and optimizes the generalization ability and robustness of the tea classification model. It uses L1 and L2 regularization to control the model's complexity, employs ensemble learning techniques to combine the prediction results of multiple models, uses weighted averaging and voting methods for model fusion, and uses cross-validation to evaluate the model's performance. The unit then selects the best model, merges the prediction results of the best model, continuously monitors the model's performance, and adjusts and optimizes it according to actual conditions.

4. The intelligent tea identification and grading system based on machine vision according to claim 3, characterized in that: The integrated calculation unit includes a dry tea related calculation unit, a wet tea related calculation unit, and an integrated related calculation unit; The dry tea associated unit obtains the comprehensive dry tea coefficient Gcxs by dimensionlessly processing the appearance feature coefficient Wgxs and tea hair feature coefficient Chxs obtained by the dry tea calculation unit and summarizing them. The wet tea related calculation unit obtains the comprehensive wet tea coefficient Scxs by performing dimensionless processing on the tea soup characteristic coefficient Ctxs and leaf bottom characteristic coefficient Ydxs obtained by the dry tea calculation unit and summarizing them. The comprehensive dry tea coefficient Gcxs and comprehensive wet tea coefficient Scxs are obtained by the following formulas; ； ； In the formula, a1 and a2 represent the preset proportional coefficients of the appearance feature coefficient Wgxs and the tea hair feature coefficient Chxs, and their specific values ​​are adjusted and set by the user. A represents the first correction constant. b1 and b2 represent the preset ratio coefficients of the tea soup characteristic coefficient Ctxs and the leaf bottom characteristic coefficient Ydxs, and their specific values ​​are adjusted and set by the user. B represents the second correction constant. The comprehensive correlation calculation unit is used to summarize and analyze the obtained comprehensive dry tea coefficient Gcxs and comprehensive wet tea coefficient Scxs after dimensionless processing, and generate comprehensive grade coefficient Pjxs. The comprehensive grade coefficient Pjxs is obtained by the following formula; 。 5. The intelligent tea identification and grading system based on machine vision according to claim 4, characterized in that: The tea grading and evaluation module includes a dry tea evaluation unit and a comprehensive evaluation unit; The dry tea evaluation unit is used to perform a preliminary comparative evaluation of the comprehensive dry tea coefficient Gcxs obtained by the dry tea correlation unit analysis and calculation, based on the preset dry tea evaluation threshold G, and to obtain the corresponding evaluation results. The specific evaluation scheme is as follows: When the comprehensive dry tea coefficient Gcxs > the dry tea evaluation threshold G, the first detection report is generated, indicating that the detected tea leaves are irregular in shape and have torn and damaged. The report also notes "The current tea is grade three tea" at the end. When the comprehensive dry tea coefficient Gcxs = dry tea evaluation threshold G, it means that the tea leaves detected are regular in shape and uniform in volume, with intact and undamaged tea hairs, plump buds without damage, and uniform tea hairs. At this time, the second evaluation mechanism is activated. When the comprehensive dry tea coefficient Gcxs < the dry tea evaluation threshold G, a second detection report is generated, indicating that the detected tea leaves are broken, the tea leaves are not whole, the tea hairs are unevenly distributed, and the density of tea hairs is less than 20% of the tea leaf volume. The report also notes at the end that "the current tea is grade four tea".

6. The intelligent tea identification and grading system based on machine vision according to claim 5, characterized in that: The comprehensive evaluation unit triggers the second evaluation mechanism due to the tea evaluation unit. The comprehensive evaluation unit is used to compare and evaluate the preset tea grade threshold C with the comprehensive grade coefficient Pjxs obtained by the comprehensive correlation calculation unit, and generate the corresponding grade. The specific evaluation scheme is as follows. When the comprehensive grade coefficient Pjxs > the tea grade threshold C, a third test report is generated, indicating that the tea soup is bright and clear with no sediment, and the tea leaves sink to the bottom in the water with uniform color and volume. The report also notes "The current tea is grade one tea" at the end. When the comprehensive grade coefficient Pjxs ≤ tea grade threshold C, a third test report is generated, indicating that the tea soup is dark in color and contains sediment, and that the tea leaves are damaged and uneven when fully steeped. The report also notes at the end that "the current tea is grade two tea".

7. A machine vision-based intelligent tea identification and grading method, comprising the machine vision-based intelligent tea identification and grading system described in any one of claims 1 to 6, characterized in that: Includes the following steps: S1. First, use an image acquisition device to take pictures of the dry tea and the brewed tea. Then, perform quality checks on the acquired dry tea images and the brewed tea images, retain clear and evenly lit images, adjust the image size, convert the color images to grayscale images, use image segmentation technology to remove the background, remove noise, and perform histogram equalization. S2. Extract features from the preprocessed dry tea image and the brewed wet tea image, including the appearance features and tea hair features of the dry tea, as well as the tea soup features and leaf base features of the wet tea image. Perform dimensionless processing on the features and convert them into a data form that can be understood and processed by computers. S3. By constructing a tea classification model and associating the extracted features with the quality grade of tea, the model is trained and optimized using labeled tea image data. Appropriate machine learning algorithms are selected, including support vector machines and random forests. The model performance is evaluated through cross-validation techniques, and the model parameters are adjusted and the model is optimized. S4. Based on the extracted dry tea characteristics and wet tea characteristics, calculate the comprehensive dry tea coefficient Gcxs and the comprehensive wet tea coefficient Scxs, perform dimensionless processing, and perform correlation calculation to obtain the comprehensive grade coefficient Pjxs. S5. Finally, a preliminary evaluation of the dry tea is conducted. The evaluation threshold G for dry tea is set and compared with the comprehensive dry tea coefficient Gcxs to determine the preliminary grade of the tea. If several teas pass the preliminary evaluation, the second evaluation mechanism is activated. The grade threshold C for tea is set and compared with the comprehensive grade coefficient Pjxs to determine the final grade of the tea.

Images