Wine-brewing sorghum grain quality grading method based on deep learning

Abstract

The invention discloses a wine brewing sorghum grain quality grading method based on deep learning, and relates to the technical field of wine brewing, and the method comprises the following steps: S1, establishing a sorghum grain selection module based on deep learning; s2, performing preliminary grading on the sorghum grains according to particle sizes; according to the method, the sorghum grains are graded in a screening mode, after grading is completed, the separated sorghum grains are selected and recognized through the selecting module, whether defective sorghum grains exist in the sorghum grains or not is judged, and after the defective sorghum grains are recognized, the defective sorghum grains in the sorghum grains are removed; compared with a traditional mode, through combination of deep learning and mechanical separation, defective grains in the sorghum grains are removed as much as possible, so that separation of the sorghum grains is completed, sorghum grain defect selection is more accurate, and improvement of the quality of the sorghum grains is facilitated.

Description

Technical Field

[0001] The present invention relates to the field of brewing technology, and in particular to a brewing sorghum grain quality grading method based on deep learning. Background Art

[0002] Winemaking is the process of using microbial fermentation to produce alcoholic beverages with a certain concentration. Winemaking raw materials and winemaking containers are two prerequisites for grain winemaking. Among the grain raw materials used for winemaking, sorghum is the most commonly used grain. When sorghum grains are used as raw materials for winemaking, the quality of sorghum grains directly affects the quality of the wine. In other words, the better the quality of sorghum, the higher the quality of the wine.

[0003] In the wine industry, sorghum is a common and key raw material for winemaking. The quality of its grains directly determines the quality of the wine. Therefore, it is very important to grade the quality of sorghum grains before winemaking. However, the traditional grading of sorghum grain quality mainly relies on human visual observation and sorting, which has serious defects: on the one hand, faced with a large number of sorghum grains with various defect types, the human eye is easily fatigued, resulting in low accuracy in identifying defective grains such as shriveled, incomplete, and moldy, and a large number of defective grains are easily missed, especially if the moldy sorghum grains are not removed, which will seriously affect the quality of the wine and even endanger the health of consumers; on the other hand, manual sorting is inefficient, which is difficult to meet the needs of large-scale winemaking production and will increase production costs; the existing grading method cannot learn and analyze the data of a large number of sorghum grains, automatically extract feature information, and grade a large number of sorghum grains; and it is impossible to quickly and accurately identify sorghum grains with defects in a large number of sorghum grains, and the efficiency of sorghum grain quality grading is low. Therefore, a quality grading method for winemaking sorghum grains based on deep learning is proposed to solve the above problems. Summary of the invention

[0004] The purpose of this application is to provide a deep learning-based method for grading the quality of brewing sorghum grains to solve the problems raised in the above-mentioned background technology.

[0005] To achieve the above objectives, the present application provides the following technical solution: a method for grading the quality of brewing sorghum grains based on deep learning, comprising the following steps:

[0006] S1: Establish a sorghum grain selection module based on deep learning;

[0007] S2: Preliminarily classify the sorghum grains according to their particle size. Prepare multiple sieves with different apertures and stack them so that the apertures gradually decrease from top to bottom. Pour the sorghum grains into the sieves and let the grains fall naturally under the action of gravity to achieve separation. Classify the sorghum grains according to their particle size to preliminarily classify their quality grades.

[0008] S3: Defect identification of sorghum grains at all levels, selecting all sorghum grains of a certain level, passing them through the selection module, and identifying defective grains in the sorghum grains of this level with the help of the recognition module in the module; when the recognition module detects that the number of defective grains exceeds the preset threshold of the total number of sorghum grains of this level, the prompt module issues a quality failure prompt and enters the next step; if the number of defective grains is lower than the preset threshold of the total number of sorghum grains of this level, the prompt module prompts that the quality is qualified; after completing the inspection of this level, the sorghum grains of other levels are inspected in the same way;

[0009] S4: Remove defective grains from the unqualified grade. For sorghum grains of unqualified quality grade, first use the soaking method to separate the defective grains, fish out the remaining grains and dry them, and then use the recognition module in the selection module to detect; if the number of defective grains is lower than the preset threshold of the total number of sorghum grains of this grade, the prompt module prompts that the quality is qualified and the processing is terminated; if it exceeds the preset threshold, the fan separation method is used for processing; thereafter, the recognition module is used for detection again. If the number of defective grains is lower than the preset threshold, the prompt module prompts that the quality is qualified and the processing is terminated; if it still exceeds the preset threshold, the soaking method and the fan separation method are used alternately in a cycle, and the recognition module is used for detection after each processing until the number of defective grains detected by the recognition module is lower than the preset threshold of the total number of sorghum grains of this grade, the prompt module prompts that the quality is qualified, and the processing of the sorghum grains of this grade is terminated;

[0010] The establishment method of the selection module is:

[0011] Step 1: Establish a data model based on sorghum seed quality parameters. After demonstration and verification, establish a selection module containing an identification module and a prompt module based on the model;

[0012] Step 2: Select hundreds of good and bad sorghum grains, input model data with discontinuous curvature and missing data, use the recognition module to identify, and prompt the module when it meets the requirements. Re-identify after manual sorting, and verify again if unqualified grains are mixed in. If it can be identified, the module is normal.

[0013] The steps to establish a data model based on sorghum seed quality parameters are:

[0014] Step 1: Collect sorghum seed quality parameters, including more than 10,000 appearance image data of different growth environments and varieties, showing full, shriveled, broken, and moldy states, and more than 8,000 sets of physical and chemical index data of starch content, protein content, and moisture content;

[0015] Step 2: Build a convolutional neural network model;

[0016] First, the data was preprocessed: the appearance images were cropped and unified into 224×224 pixels, and the pixel values were normalized to [0,1]; the physical and chemical index data were standardized to have a mean of 0 and a standard deviation of 1;

[0017] Then, the ResNet-50 architecture is adopted, and convolutional layers with 3×3 convolutional kernels, a stride of 1, and "same" padding, max pooling layers with 2×2 pooling kernels and a stride of 2, and three fully connected layers with 512, 256 neurons and the number of corresponding quality grade classifications are set;

[0018] Then, the training set, validation set, and test set are divided in a ratio of 7:2:1, and the model is trained using the Stochastic Gradient Descent (SGD) algorithm with a learning rate of 0.001, 50 iteration times, and a momentum of 0.9. The weights are adjusted through backpropagation. If the loss function value of the validation set does not decrease for 5 consecutive iterations, the training stops;

[0019] Finally, the model is evaluated using the test set, and the accuracy, recall rate, and F1 value are calculated. If the accuracy is lower than 90%, the learning rate is adjusted, the data volume is increased, or the architecture is improved to optimize the model;

[0020] Step 3: Apply the optimized CNN model as a data model for identifying the quality of sorghum seeds to the selection module for the identification module to accurately and efficiently identify the quality of sorghum seeds.

[0021] Preferably, when separating sorghum grains by the soaking method, the sorghum grains are immersed in water. The hollow or shriveled sorghum grains will float on the water surface, while the intact and good-quality grains will sink to the bottom, thus completing the quality separation of sorghum grains;

[0022] When separating sorghum grains by a blower, the sorghum grains are poured from a high place, and at the same time, the blower is aimed at the falling position of the sorghum grains for blowing; during the falling process, the incomplete sorghum grains will be blown away, while the intact and heavier grains will fall vertically under the action of their own gravity, thus completing the separation of sorghum grains.

[0023] Preferably, when initially separating sorghum grains, the minimum number of sieve meshes for screening is three groups, and the maximum is six groups. The sieve pore size is selected according to the size of sorghum grains, and the pore size and number of sieve meshes are adjusted according to the quality of sorghum grains in the detection batch.

[0024] Preferably, the identification module of the selection module includes a scanner and an image analysis processor, the prompt module includes a voice alarm prompt, the image analysis processor is connected to the voice alarm prompt through a wireless network, and the scanner is connected to the image analysis processor through a wireless network.

[0025] Preferably, when establishing the data model of the selection module, the relevant parameters of sorghum grains are used as the basis. The sorghum grain parameters used as the basis are the plumpness of sorghum grains. If it is a complete sorghum grain, its surface arc is complete. If the sorghum grain is incomplete and has defects, its surface arc must also be defective.

[0026] Preferably, when the selection module conducts tests, the number of sorghum grains used for testing shall not be less than one hundred, and the number of tests shall not be less than three. The selected sorghum grains shall be a mixture of sorghum grains with intact appearance but different sizes, defective sorghum grains, shriveled sorghum grains, and spoiled and deteriorated sorghum grains.

[0027] Preferably, the selection module shall include a conveyor belt adapted to it. When the sorghum grains are selected and separated, the sorghum grains shall be evenly spread on the conveyor belt for transportation. When spreading the sorghum grains, attention shall be paid not to stack them. While transporting, the recognition module recognizes the transported sorghum grains.

[0028] Preferably, after the sorghum grains are separated by the soaking method, the intact sorghum grains sinking to the lower layer are fished out, drained, and then dried by a drying device. During drying, attention shall be paid to the drying degree of the sorghum grains at all times, and the moisture content of the sorghum grains shall be detected. When the moisture content of the sorghum grains reaches the requirements for brewing, the drying is stopped.

[0029] Preferably, when the sorghum grains are selected and graded for quality, the sorghum grains are first separated according to their particle sizes, and then the graded sorghum grains are screened by the selection module. When the sorghum grains are initially separated and graded, the selection module does not participate in the selection of the sorghum grains. After the sorghum grains are separated by particle size, the selection module intervenes to select the sorghum grains.

[0030] In summary, the technical effects and advantages of the present invention are as follows:

[0031] In the present invention, before brewing, a selection module for sorghum grains is first established based on deep learning, and then the sorghum grains are graded by screening. After grading, the separated sorghum grains are selected and recognized by the selection module to determine whether there are defective sorghum grains among the sorghum grains. After identifying the defective sorghum grains, the defective, shriveled or spoiled and deteriorated sorghum grains in the sorghum grains are removed by the cooperation of immersion separation and the wind power of the fan. Finally, the sorghum grains are verified by the selection module to check whether there are defective sorghum grains among them. Compared with the traditional method, by combining deep learning and mechanical separation, the defective grains in the sorghum grains are removed as much as possible to complete the separation of the sorghum grains. The selection of defective sorghum grains is more accurate, which is beneficial to improving the quality of the sorghum grains. Description of the Drawings

[0032] To more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the accompanying drawings required in the description of the embodiments or the prior art. Obviously, the accompanying drawings in the following description are only some embodiments of the present application. For those skilled in the art, without creative efforts, other accompanying drawings can also be obtained based on these drawings.

[0033] Figure 1 It is a schematic diagram of the operation process of a method for grading the quality of brewing sorghum grains based on deep learning proposed in the present application;

[0034] Figure 2 It is a schematic diagram of the verification result of the selection module. Detailed implementation manners

[0035] The following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, rather than all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts belong to the scope of protection of the present invention.

[0036] Embodiment: Refer to Figure 1 A method for grading the quality of brewing sorghum grains based on deep learning shown in the figure includes the following steps:

[0037] S1: Establish a selection module for sorghum grains based on deep learning;

[0038] S2: Preliminarily grade the sorghum grains according to the particle size. Prepare multiple sieves with different pore sizes, stack them, and make the pore sizes gradually decrease from top to bottom; pour the sorghum grains into the sieve and let the grains fall naturally under the action of gravity to achieve separation; grade the sorghum grains according to the particle size to initially divide their quality grades;

[0039] S3: Identify the defects of sorghum grains at each level. Select all sorghum grains at a certain level and let them pass through the selection module. Use the recognition module in the module to identify the defective grains in this level of sorghum grains; when the recognition module detects that the number of defective grains exceeds the preset threshold of the total number of sorghum grains at this level, the prompt module issues a quality unqualified prompt and enters the next step; if the number of defective grains is lower than the preset threshold of the total number of sorghum grains at this level, the prompt module prompts that the quality is qualified; after completing the detection of this level, detect other levels of sorghum grains in the same way;

[0040] S4: Remove defective grains from the unqualified grade. For sorghum grains of unqualified quality grade, first use the soaking method to separate the defective grains, fish out the remaining grains and dry them, and then use the recognition module in the selection module to detect; if the number of defective grains is lower than the preset threshold of the total number of sorghum grains of this grade, the prompt module prompts that the quality is qualified and the processing is terminated; if it exceeds the preset threshold, the fan separation method is used for processing; thereafter, the recognition module is used for detection again. If the number of defective grains is lower than the preset threshold, the prompt module prompts that the quality is qualified and the processing is terminated; if it still exceeds the preset threshold, the soaking method and the fan separation method are used alternately in a cycle, and the recognition module is used for detection after each processing until the number of defective grains detected by the recognition module is lower than the preset threshold of the total number of sorghum grains of this grade, the prompt module prompts that the quality is qualified, and the processing of the sorghum grains of this grade is terminated;

[0041] The creation method of the selection module is:

[0042] Step 1: Establish a data model based on sorghum seed quality parameters. After demonstration and verification, establish a selection module containing an identification module and a prompt module based on the model;

[0043] Step 2: Select hundreds of good and bad sorghum grains, input model data with discontinuous curvature and missing data, use the recognition module to identify, and prompt the module when it meets the requirements. Re-identify after manual sorting, and verify again if unqualified grains are mixed in. If it can be identified, the module is normal.

[0044] The steps to establish a data model based on sorghum seed quality parameters are:

[0045] Step 1: Collect sorghum seed quality parameters, including more than 10,000 appearance image data of different growth environments and varieties, showing full, shriveled, broken, and moldy states, and more than 8,000 sets of physical and chemical index data of starch content, protein content, and moisture content;

[0046] Step 2: Build a convolutional neural network model;

[0047] First, the data was preprocessed: the appearance images were cropped and unified into 224×224 pixels, and the pixel values were normalized to [0,1]; the physical and chemical index data were standardized to have a mean of 0 and a standard deviation of 1;

[0048] Then, the ResNet-50 architecture was used, with a 3×3 convolution kernel, stride 1, and padding “same” convolutional layer, a 2×2 pooling kernel, a stride 2 max pooling layer, and 3 fully connected layers with 512 and 256 neurons and the corresponding number of quality level categories;

[0049] Then divide the training set, validation set, and test set in a ratio of 7:2:1, and train the model using the Stochastic Gradient Descent (SGD) algorithm with a learning rate of 0.001, 50 iterations, and a momentum of 0.9. Adjust the weights through backpropagation. If the loss function value of the validation set does not decrease for 5 consecutive iterations, stop training;

[0050] Finally, evaluate the model using the test set, calculate the accuracy, recall rate, and F1 value. If the accuracy is lower than 90%, adjust the learning rate, increase the data volume, or improve the architecture to optimize the model;

[0051] Step 3: Apply the optimized CNN model as a data model for identifying the quality of sorghum seeds to the selection module for the identification module to accurately and efficiently identify the quality of sorghum seeds.

[0052] As a preferred implementation manner of this embodiment, when separating sorghum grains by the soaking method, by immersing the sorghum grains in water, the hollow or shriveled sorghum grains will float on the water surface, while the intact and good-quality grains will sink to the bottom, thereby completing the quality separation of sorghum grains;

[0053] When separating sorghum grains by a blower, pour the sorghum grains from a high place, and at the same time aim the blower at the falling position of the sorghum grains for blowing; during the falling process, the incomplete sorghum grains will be blown away, while the intact and heavier grains will fall vertically under the action of their own gravity, thus completing the separation of sorghum grains.

[0054] As a preferred implementation manner of this embodiment, when initially separating sorghum grains, the minimum number of sieve meshes for screening is three groups, and the maximum is six groups. The sieve aperture is selected according to the size of sorghum grains, and the aperture and number of sieve meshes are adjusted according to the quality of sorghum grains in the detection batch.

[0055] As a preferred implementation manner of this embodiment, the identification module of the selection module includes a scanner and an image analysis processor, the prompt module includes a voice alarm prompt device, the image analysis processor is connected to the voice alarm prompt device through a wireless network, and the scanner is connected to the image analysis processor through a wireless network.

[0056] As a preferred implementation manner of this embodiment, when establishing the data model of the selection module, based on the relevant parameters of sorghum grains, the sorghum grain parameters relied on are the plumpness of sorghum grains. If it is a complete sorghum grain, its surface curvature is complete. If the sorghum grain is incomplete and has defects, its surface curvature is definitely defective.

[0057] As a preferred implementation manner of this embodiment, when the selection module conducts tests, the number of sorghum grains used for the tests shall not be less than one hundred, and the number of test times shall not be less than three. The selected sorghum grains need to be mixed with sorghum grains that are in good condition but have different sizes, defective sorghum grains, shriveled sorghum grains, and spoiled and deteriorated sorghum grains.

[0058] As a preferred implementation manner of this embodiment, the selection module shall include a conveyor belt adapted thereto. When the sorghum grains are selected and separated, the sorghum grains need to be evenly spread on the conveyor belt for conveyance. Pay attention not to stack the sorghum grains during spreading. While conveying, the recognition module recognizes the conveyed sorghum grains.

[0059] As a preferred implementation manner of this embodiment, after the sorghum grains are separated by the soaking method, the sorghum grains in good condition that sink to the lower layer are fished out, drained, and then dried by a drying device. During drying, it is necessary to always pay attention to the drying degree of the sorghum grains and at the same time detect the water content of the sorghum grains. Stop drying when the water content of the sorghum grains reaches the requirements for brewing.

[0060] As a preferred implementation manner of this embodiment, when the sorghum grains are selected and graded for quality, first separate the sorghum grains by particle size, and then screen the graded sorghum grains through the selection module. When the sorghum grains are initially separated and graded, the selection module does not participate in the selection of the sorghum grains. After the sorghum grains are separated by particle size, the selection module intervenes to select the sorghum grains.

[0061] Working principle of the present invention:

[0062] In the present invention, before brewing, first establish a selection module for sorghum grains based on deep learning, and then classify the sorghum grains by screening. After classification is completed, the separated sorghum grains are selected and recognized through the selection module to determine whether there are defective sorghum grains in the sorghum grains. After identifying the defective sorghum grains, cooperate with the soaking separation and the wind power of the fan to remove the defective, shriveled or spoiled and deteriorated sorghum grains in the sorghum grains. Finally, the selection module is used to verify the sorghum grains to check whether there are defective sorghum grains. Compared with the traditional method, by combining deep learning and mechanical separation, as many defective grains in the sorghum grains as possible are removed to complete the separation of the sorghum grains. The selection of defective sorghum grains is more accurate, which is beneficial to improving the quality of sorghum grains.

[0063] Finally, it should be noted that the above are only preferred embodiments of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or perform equivalent replacements for some of the technical features. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the present invention shall be included within the protection scope of the present invention.

Claims

1. A method for grading the quality of sorghum grains for brewing based on deep learning, characterized in that, The following steps are involved: S1: Establish a sorghum grain selection module based on deep learning; S2: Preliminarily classify the sorghum grains according to their particle size. Prepare multiple sieves with different apertures and stack them so that the apertures gradually decrease from top to bottom. Pour the sorghum grains into the sieves and let the grains fall naturally under the action of gravity to achieve separation. Classify the sorghum grains according to their particle size to preliminarily classify their quality grades. S3: Defect identification of sorghum grains at all levels, selecting all sorghum grains of a certain level, passing them through a selection module, and identifying defective grains in the sorghum grains of that level with the help of a recognition module in the module; when the recognition module detects that the number of defective grains exceeds a preset threshold of the total number of sorghum grains of that level, the prompt module issues a quality failure prompt and proceeds to the next step; If the number of defective kernels is lower than the preset threshold of the total number of kernels of this grade of sorghum, the prompt module will indicate that the quality is qualified; After completing this level of testing, the other levels of sorghum grains were tested in the same way; S4: removing defective grains from the unqualified grade. For sorghum grains of unqualified quality, first use the soaking method to separate the defective grains, fish out the remaining grains and dry them, and then use the recognition module in the selection module to detect them; if the number of defective grains is lower than the preset threshold of the total number of sorghum grains of this grade, the prompt module prompts that the quality is qualified and the processing ends; If it exceeds the preset threshold, the fan separation method is used; After that, the recognition module is used for detection again. If the number of defective kernels is lower than the preset threshold, the prompt module prompts that the quality is qualified and the processing ends; If it still exceeds the preset threshold, the immersion method and the fan separation method are used alternately in a cycle. After each treatment, the recognition module is used for detection until the number of defective grains detected by the recognition module is lower than the preset threshold of the total number of sorghum grains of this grade. The prompt module prompts that the quality is qualified, and the processing of this grade of sorghum grains is ended.

2. The method for grading the quality of sorghum grains for brewing based on deep learning according to claim 1, wherein, The establishment method of the selection module is: Step 1: Establish a data model based on sorghum seed quality parameters. After demonstration and verification, establish a selection module containing an identification module and a prompt module based on the model; Step 2: Select hundreds of good and bad sorghum grains, input model data with discontinuous curvature and missing data, use the recognition module to identify, and prompt the module when it meets the requirements. Re-identify after manual sorting, and verify again if unqualified grains are mixed in. If it can be identified, the module is normal.

3. A method for grading the quality of sorghum grains for brewing based on deep learning according to claim 2, characterized in that, The steps to establish a data model based on sorghum seed quality parameters are: Step 1: Collect sorghum seed quality parameters, including more than 10,000 appearance image data of different growth environments and varieties, showing full, shriveled, broken, and moldy states, and more than 8,000 sets of physical and chemical index data of starch content, protein content, and moisture content; Step 2: Build a convolutional neural network model; First, the data was preprocessed: the appearance images were cropped and unified into 224×224 pixels, and the pixel values were normalized to [0,1]; the physical and chemical index data were standardized to have a mean of 0 and a standard deviation of 1; Then, the ResNet-50 architecture is adopted, setting up a convolutional layer with a 3×3 convolutional kernel, a stride of 1, and "same" padding, a max pooling layer with a 2×2 pooling kernel and a stride of 2, and three fully connected layers with 512, 256 neurons and the number of categories corresponding to the quality grades. Then, the training set, validation set, and test set are divided in the ratio of 7:2:1, and the model is trained with the Stochastic Gradient Descent (SGD) algorithm with a learning rate of 0.001, 50 iterations, and a momentum of 0.

9. The weights are adjusted through backpropagation. If the loss function value of the validation set does not decrease for 5 consecutive iterations, the training stops. Finally, the model is evaluated using the test set, and the accuracy, recall rate, and F1 value are calculated. If the accuracy is lower than 90%, the learning rate is adjusted, the data volume is increased, or the architecture is improved to optimize the model. Step 3: Apply the optimized CNN model as a data model for identifying the quality of sorghum seeds to the selection module for the identification module to accurately and efficiently identify the quality of sorghum seeds.

4. A method for grading the quality of sorghum grains for brewing based on deep learning according to claim 1, characterized in that: When separating sorghum grains by the soaking method, the sorghum grains are immersed in water. The hollow or shriveled sorghum grains will float on the water surface, while the intact and high-quality grains will sink to the bottom, thus completing the quality separation of sorghum grains. When separating sorghum grains by a blower, the sorghum grains are poured from a height, and at the same time, the blower is aimed at the falling position of the sorghum grains for blowing. During the falling process, the incomplete sorghum grains will be blown away, while the intact and heavier grains will fall vertically under the action of their own gravity, thus completing the separation of sorghum grains. When initially separating sorghum grains, the minimum number of sieve meshes for screening is three groups, and the maximum is six groups. The sieve aperture is selected according to the size of sorghum grains, and the aperture and number of sieve meshes are adjusted according to the quality of sorghum grains in the detection batch.

5. A method for grading the quality of sorghum grains for brewing based on deep learning according to claim 1, characterized in that: The identification module of the selection module includes a scanner and an image analysis processor, the prompt module includes a voice alarm prompt, the image analysis processor is connected to the voice alarm prompt through a wireless network, and the scanner is connected to the image analysis processor through a wireless network.

6. The method for grading the quality of sorghum grains for brewing based on deep learning according to claim 1, characterized in that: When establishing the data model of the selection module, the relevant parameters of sorghum grains are used as the basis. The sorghum grain parameters used as the basis are the plumpness of sorghum grains. If the sorghum grains are complete, the surface arc is complete. If the sorghum grains are incomplete and have defects, the surface arc is definitely defective.

7. A method for grading the quality of sorghum grains for brewing based on deep learning according to claim 1, characterized in that: When testing the selection module, the number of sorghum grains used for testing shall not be less than one hundred, and the number of tests shall not be less than three times. The selected sorghum grains need to be mixed with sorghum grains with intact appearance but different sizes, incomplete sorghum grains, shriveled sorghum grains, and spoiled and deteriorated sorghum grains.

8. A method for grading the quality of sorghum grains for brewing based on deep learning according to claim 5, characterized in that: The selection module should include a conveyor belt adapted to it. When separating and selecting sorghum grains, the sorghum grains need to be evenly spread on the conveyor belt for conveying. When spreading the sorghum grains, pay attention not to stack them, and the identification module identifies the conveyed sorghum grains while conveying.

9. A method for grading the quality of sorghum grains for brewing based on deep learning according to claim 1, characterized in that: After the sorghum grains are separated by the soaking method, the sorghum grains with intact appearance that sink to the lower layer are fished out, drained, and then dried by a drying device. During drying, it is necessary to constantly pay attention to the drying degree of the sorghum grains and at the same time detect the water content of the sorghum grains. When the water content of the sorghum grains reaches the requirements for brewing, the drying is stopped.

10. A method for grading the quality of sorghum grains for brewing based on deep learning according to claim 1, characterized in that: When the sorghum grains are graded and selected for quality, first separate the sorghum grains according to their particle sizes, and then screen the graded sorghum grains through a selection module. When the sorghum grains are initially separated and graded, the selection module does not participate in the selection of the sorghum grains. After the sorghum grains are separated by particle size, the selection module then intervenes to select the sorghum grains.

Images