A tea rolling degree classification method based on a machine vision system
By establishing a tea rolling degree classification model using machine vision system and CNN-SVM algorithm, the problems of subjectivity and unevenness in judging the rolling degree of tea leaves are solved, and precise control of tea processing and quality stability are achieved.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- RES INST OF ZHEJIANG UNIV TAIZHOU
- Filing Date
- 2026-02-09
- Publication Date
- 2026-05-29
Smart Images
Figure CN122115956A_ABST
Abstract
Description
Technical Field
[0001] This invention belongs to the field of tea preparation and relates to a method for classifying the degree of tea rolling based on a machine vision system. Background Technology
[0002] Tea is made through traditional processing techniques, which can be simply divided into four steps: withering, rolling, fermentation, and drying. Rolling, as the second step in tea processing, is a key process in shaping the appearance of tea and forming its unique flavor and quality. However, proper rolling is an important condition for subsequent production and processing. Light rolling will result in insufficient damage to the tea cell structure, leading to an unsuccessful fermentation process, while excessive rolling will cause the tea juice to evaporate.
[0003] Currently, experienced tea masters determine the degree of rolling in Tiantai Yunwu tea by sensing changes in the color and aroma of the rolled leaves. However, this sensory evaluation method is subjective and easily influenced by personal factors and environmental factors such as ambient light. Furthermore, different tea masters have varying evaluation standards. At present, there is still a lack of objective evaluation methods to differentiate the degree of rolling in industrial production. This leads to uneven manual judgment of rolling degree in actual production, potentially resulting in inconsistent tea quality. Summary of the Invention
[0004] In order to overcome at least one deficiency of the prior art, the present invention provides a method for classifying the degree of tea kneading based on a machine vision system.
[0005] To achieve the above objectives, the present invention adopts the following technical solution: a method for classifying the degree of tea rolling based on a machine vision system, comprising the following steps: Step 1: Acquire tea leaf sample images at different rolling times using a machine vision system; Step 2: Extract and analyze the color feature information of the tea leaf sample images; Step 3: Collect the color appearance index of tea leaf samples under different rolling times using a portable colorimeter; Step 4: Based on the color appearance index under different kneading times, obtain the variation law of color appearance index with kneading time; Step 5: Based on color feature information and the variation law of color appearance index with kneading time, establish a classification model of tea kneading degree using CNN-SVM algorithm.
[0006] Furthermore, the machine vision system includes an image sensor, an arc light source, a camera, and a GUI software processing system.
[0007] Furthermore, the method for extracting and analyzing the color feature information of the tea leaf sample image in step 2 is to automatically select a 1080*1080 pixel area in the shooting area of the tea leaf using a machine vision system, and take this area as the region of interest (ROI). The color feature variables in the ROI are extracted; the average value of each color feature variable in the ROI is calculated, and these values are used to represent the feature information of the corresponding image color of each sample.
[0008] Furthermore, step 2 extracts and calculates each image into 9 color feature variables, encompassing 3 color models: RGB, HSV, and La*b*. After dataset partitioning, the training set contains 120 sets of data, and the prediction set contains 80 sets of data. The image color feature variable data, after filtering and integration, includes 50 sets of average color feature variable data in each label category, resulting in a total of 500 sets of average color feature variable data.
[0009] Furthermore, the portable colorimeter is model CM-600d, with a wavelength range of 400-700nm, a wavelength interval of 10nm, and a standard observation angle set at 2. 0 Viewpoint, observation light source set to D65.
[0010] Furthermore, the CNN-SVM algorithm consists of a convolutional neural network and a support vector machine.
[0011] Furthermore, the convolutional neural network is composed of convolutional layers, pooling layers, and fully connected layers, and the spatial features of the input data are extracted by using convolutional layers and pooling layers.
[0012] Furthermore, the convolution operation can capture local features in the image. The convolution kernel slides across the input data, extracting features from different locations through weight sharing. The pooling layer performs feature selection after the convolutional layer, reducing the dimensionality of the convolutional layer output by downsampling the input data, thus reducing training parameters, extracting the most significant features, alleviating overfitting, and increasing the model's robustness to positional changes. The fully connected layer follows the convolutional and pooling layers. In the fully connected layer, each neuron is connected to all neurons in the previous layer, achieving the integration of differential local feature information between different categories.
[0013] Furthermore, the fully connected layer uses ReLU as the activation function, which is then combined with the Softmax activation function to complete the output of the multi-classification task, transforming the original output of the network into a probability distribution.
[0014] Furthermore, the support vector machine is used to find a hyperplane to maximize the boundary spacing between tea leaf samples under the rolling time.
[0015] In summary, the advantages of this invention are: This invention establishes an objective and quantitative classification method for the degree of tea kneading based on a machine vision system, thereby achieving precise control over the tea processing and stabilizing product quality. Attached Figure Description
[0016] Figure 1 This is a flowchart of the tea rolling degree classification method of the present invention.
[0017] Figure 2 This is a schematic diagram of the convolutional neural network structure of the present invention. Detailed Implementation
[0018] The following specific examples illustrate the implementation of the present invention. Those skilled in the art can easily understand other advantages and effects of the present invention from the content disclosed in this specification. The present invention can also be implemented or applied through other different specific embodiments, and various details in this specification can also be modified or changed based on different viewpoints and applications without departing from the spirit of the present invention. It should be noted that, unless otherwise specified, the following embodiments and features described therein can be combined with each other.
[0019] Example: like Figures 1-2 As shown, a method for classifying the degree of tea rolling based on a machine vision system includes the following steps: Step 1: Acquire tea leaf sample images at different rolling times using a machine vision system; The machine vision system includes an image sensor, an arc light source, a camera, and a GUI software processing system. The camera (FI-S200C-G) includes an image sensor with a 4nm low-distortion lens and a CMOS image sensor with a resolution of (1080*1080).
[0020] Step 2: Extract and analyze the color feature information of the tea leaf sample images; The machine vision system automatically selects a 1080*1080 pixel area in the shooting area of tea leaves and uses this area as the region of interest. At the same time, the software extracts the color feature variables in the ROI; calculates the average value of each color feature variable in the ROI area, and uses these values to represent the feature information of the corresponding image color of each sample. Through extraction and calculation, each image is converted into nine color feature variables, encompassing three color models: RGB, HSV, and La*b*. The dataset is divided into a training set containing 120 data sets and a prediction set containing 80 data sets.
[0021] After filtering and integration, the image color feature variable data contained 50 sets of average color feature variable data in each label category (label 1 to label 10), resulting in a total of 500 sets of average color feature variable data.
[0022] Step 3: Collect the color appearance index of tea leaf samples under different rolling times using a portable colorimeter; The portable colorimeter, model CM-600d, can measure wavelengths from 400 to 700 nm with a wavelength interval of 10 nm. The standard observation angle of the instrument is set to 2. 0 Viewpoint, observation light source set to D65.
[0023] Step 4: Based on the color appearance index under different kneading times, obtain the variation law of color appearance index with kneading time; Step 5: Based on color feature information and the variation law of color appearance index with kneading time, establish a classification model of tea kneading degree using CNN-SVM algorithm.
[0024] The CNN-SVM algorithm consists of a convolutional neural network and a support vector machine. A convolutional neural network consists of convolutional layers, pooling layers, and fully connected layers. It extracts spatial features from the input data using convolutional and pooling layers. The structure of a convolutional neural network is as follows: Figure 2 As shown.
[0025] Convolution operations can effectively capture local features in an image. The convolution kernel slides across the input data, extracting features from different locations through weight sharing.
[0026] Pooling layers perform feature selection after convolutional layers. By downsampling the input data, they reduce the dimensionality of the convolutional layer output, reduce training parameters, extract the most significant features, alleviate overfitting, and increase the model's robustness to positional changes.
[0027] Fully connected layers follow convolutional and pooling layers. In a fully connected layer, each neuron is connected to all neurons in the previous layer, enabling the integration of differential local feature information between different categories.
[0028] The fully connected layer uses ReLU as the activation function, which helps to introduce non-linearity and enhance the expressive power of the network. The fully connected layer is then combined with the Softmax activation function to complete the output of multi-classification tasks, transforming the network's original output into a probability distribution, making it more suitable for classification problems.
[0029] Support vector machines are used to find a hyperplane that maximizes the boundary spacing between tea leaf samples under different rolling times, thereby achieving more accurate classification.
[0030] Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without inventive effort should fall within the scope of protection of the present invention.
Claims
1. A method for classifying the degree of tea rolling based on a machine vision system, characterized in that: Includes the following steps: Step 1: Acquire tea leaf sample images at different rolling times using a machine vision system; Step 2: Extract and analyze the color feature information of the tea leaf sample images; Step 3: Collect the color appearance index of tea leaf samples under different rolling times using a portable colorimeter; Step 4: Based on the color appearance index under different kneading times, obtain the variation law of color appearance index with kneading time; Step 5: Based on color feature information and the variation law of color appearance index with kneading time, establish a classification model of tea kneading degree using CNN-SVM algorithm.
2. The tea rolling degree classification method based on a machine vision system according to claim 1, characterized in that: The machine vision system includes an image sensor, an arc light source, a camera, and a GUI software processing system.
3. The tea rolling degree classification method based on a machine vision system according to claim 2, characterized in that: The method for extracting and analyzing the color feature information of the tea leaf sample image in step 2 is as follows: automatically select a 1080*1080 pixel area in the shooting area of the tea leaf using a machine vision system, and take this area as the region of interest (ROI). Extract the color feature variables in the ROI; calculate the average value of each color feature variable in the ROI area, and use these values to represent the feature information of the corresponding image color of each sample.
4. The tea rolling degree classification method based on a machine vision system according to claim 3, characterized in that: Step 2 involves extracting and calculating nine color feature variables for each image, encompassing three color models: RGB, HSV, and La*b*. After dataset partitioning, the training set contains 120 data sets, and the prediction set contains 80 data sets. The image color feature variable data, after filtering and integration, includes 50 sets of average color feature variable data for each label category, resulting in a total of 500 sets of average color feature variable data.
5. The tea rolling degree classification method based on a machine vision system according to claim 1, characterized in that: The portable colorimeter is model CM-600d, with a wavelength range of 400-700nm, a wavelength interval of 10nm, and a standard observation angle set at 2. 0 Viewpoint, observation light source set to D65.
6. The tea rolling degree classification method based on a machine vision system according to claim 1, characterized in that: The CNN-SVM algorithm consists of a convolutional neural network and a support vector machine.
7. The tea rolling degree classification method based on a machine vision system according to claim 6, characterized in that: The convolutional neural network consists of convolutional layers, pooling layers, and fully connected layers. It extracts spatial features of the input data by using convolutional and pooling layers.
8. The tea rolling degree classification method based on a machine vision system according to claim 7, characterized in that: The convolution operation captures local features in the image. The convolution kernel slides across the input data, extracting features from different locations through weight sharing. The pooling layer performs feature selection after the convolutional layer, reducing the dimensionality of the convolutional layer's output by downsampling the input data, thus reducing training parameters, extracting the most significant features, mitigating overfitting, and increasing the model's robustness to positional changes. The fully connected layer follows the convolutional and pooling layers. In the fully connected layer, each neuron is connected to all neurons in the previous layer, integrating differential local feature information between different categories.
9. The tea rolling degree classification method based on a machine vision system according to claim 8, characterized in that: The fully connected layer uses ReLU as the activation function, which is then combined with the Softmax activation function to complete the output of the multi-classification task, transforming the network's original output into a probability distribution.
10. A method for classifying the degree of tea rolling based on a machine vision system according to claim 6, characterized in that: The support vector machine is used to find a hyperplane to maximize the boundary spacing between tea leaf samples under the specified rolling time.