Methods, systems, and storage media for judging the uniformity of koji mixing in brewing processes.

Abstract

This application provides a method, system, and storage medium for judging the uniformity of koji mixing in the brewing process. By acquiring an image of the koji mixing area in the brewery, extracting the image features, and judging whether the koji mixing area is uniform based on the image features and a preset koji mixing uniformity judgment model, the method achieves automated recognition of the uniformity of koji mixing in the brewing process. Compared with the existing solution that requires manual visual recognition of koji mixing uniformity, this method improves the accuracy and objectivity of the recognition results and saves labor costs.

Description

Technical Field

[0001] This application relates to the field of brewing technology, and in particular to a method, equipment and storage medium for judging the uniformity of koji mixing in a brewing process. Background Technology

[0002] In the production process of Maotai-flavor liquor, after the mash is distilled and removed from the still, it first needs to undergo processes such as spreading and cooling, shaping, and kicking. After the temperature drops to a certain level, operations such as mixing with yeast and piling up the mash are carried out. The mixing with yeast involves sprinkling a certain amount of yeast onto the mash and manually turning it to ensure that the mash and yeast are fully mixed. Insufficient or inadequate manual turning can lead to uneven mixing of the yeast, affecting the quality of subsequent piling and fermentation.

[0003] Traditional methods involve manual spreading and mixing of the starter culture. Whether the starter culture is evenly mixed depends entirely on human visual inspection. However, human eyes are prone to fatigue and are highly subjective, which inevitably leads to missed or incorrect judgments. Summary of the Invention

[0004] The purpose of this application is to provide a method, equipment, and storage medium for judging the uniformity of koji mixing in the brewing process, so as to solve the above-mentioned technical problems.

[0005] On the one hand, a method for judging the uniformity of koji mixing in a brewing process is provided, the method comprising:

[0006] Obtain the image corresponding to the koji mixing area in the brewery;

[0007] Extract the image features of the image;

[0008] The image features and a preset mixing uniformity discrimination model are used to determine whether the mixing region is uniformly mixed. The mixing uniformity discrimination model is a model trained based on the image features of multiple first sample images and multiple second sample images. The first sample images are the images corresponding to the uniformly mixed mixing region, and the second sample images are the images corresponding to the unevenly mixed mixing region.

[0009] In one embodiment, acquiring the image corresponding to the koji mixing area in the brewery includes:

[0010] A panoramic image is obtained by capturing images of the ground of the brewery using cameras installed in the brewery building.

[0011] The mixing area is determined from the panoramic image, and the image corresponding to the mixing area is obtained.

[0012] In one embodiment, determining the mixing area from the panoramic image includes:

[0013] The mixing region is determined from the panoramic image based on the panoramic image and the preset mixing region recognition model; the mixing region recognition model is a model trained based on multiple sample panoramic images and multiple mixing region mask images, and each mixing region mask image is an image obtained by annotating the corresponding sample panoramic image with mixing regions.

[0014] In one embodiment, the step of acquiring a panoramic image of the ground of the brewery by using a camera installed in the brewery includes:

[0015] By using cameras installed in the brewery to capture images of the ground of the brewery, a series of panoramic images in time can be obtained.

[0016] The step of determining the mixing region from the panoramic image based on the panoramic image and a preset mixing region recognition model includes:

[0017] Inputting a frame of the panoramic image into the preset mixing region recognition model yields the first potential mixing region;

[0018] Based on multiple frames of the panoramic image, the average value of image feature change of each pixel in the panoramic image is determined, and the region composed of pixels whose average value of image feature change is greater than or equal to a preset change threshold is taken as the second potential mixing region.

[0019] The intersection area between the first potential mixing region and the second potential mixing region is taken as the mixing region.

[0020] In one embodiment, determining the average image feature change of each pixel in the panoramic image based on multiple frames of the panoramic image includes:

[0021] For each pixel in the panoramic image at interval r frames, calculate its corresponding image feature change value;

[0022] Calculate the average value of the image feature changes corresponding to multiple frames of the panoramic image, and use the average value as the average value of the image feature changes of the corresponding pixel; wherein, r is an integer greater than 0.

[0023] In one embodiment, the image features include the color distribution features of each pixel;

[0024] The step of determining whether the mixing region is uniformly mixed based on the image features and a preset mixing uniformity discrimination model includes:

[0025] The uniformity value of mixing for each pixel in the image is determined based on the color distribution characteristics of each pixel and the preset uniformity discrimination model.

[0026] Calculate the overall koji mixing uniformity value corresponding to the image based on the koji mixing uniformity value;

[0027] The overall uniformity value of the koji mixing is compared with a preset uniformity threshold.

[0028] The comparison results determine whether the koji mixing area is uniformly mixed.

[0029] In one embodiment, after determining the uniformity value of each pixel in the image based on the color distribution characteristics of each pixel and a preset uniformity discrimination model, the process includes:

[0030] A heatmap corresponding to the image is generated based on the uniformity value of the mixing at each pixel in the image.

[0031] In one embodiment, the method further includes:

[0032] If it is determined that the mixing of the starter culture is uneven in the mixing area, an alarm message is reported.

[0033] On the other hand, a system for judging the uniformity of koji mixing in a brewing process is provided, including a processor, a memory, and a camera. The memory stores a computer program, and the processor executes the computer program to implement any of the above methods.

[0034] On the other hand, a computer-readable storage medium is provided that stores a computer program, which, when executed by at least one processor, implements any of the methods described above.

[0035] The method, system, and storage medium for judging the uniformity of koji mixing in the brewing process provided in this application acquire images of the koji mixing area in the brewery, extract image features from the images, and judge whether the koji mixing area is uniform based on the image features and a preset koji mixing uniformity judgment model. This achieves automated recognition of the uniformity of koji mixing in the brewing process. Compared with the existing solutions that require manual visual recognition of koji mixing uniformity, this improves the accuracy of the recognition results and saves labor costs. Attached Figure Description

[0036] Figure 1 This is a flowchart illustrating the method for determining the uniformity of koji mixing in the brewing process provided in Embodiment 1 of this application.

[0037] Figure 2 This is a schematic diagram of the camera layout provided in Embodiment 1 of this application;

[0038] Figure 3This is a schematic diagram of the process for determining whether the koji mixing area is uniformly mixed, provided in Embodiment 1 of this application.

[0039] Figure 4 This is a schematic diagram of 5x5 pixel coordinates provided in Embodiment 1 of this application;

[0040] Figure 5 A schematic diagram of a heat map provided in Embodiment 1 of this application;

[0041] Figure 6 This is a schematic diagram of the structure of the brewing process uniformity discrimination system provided in Embodiment 2 of this application. Detailed Implementation

[0042] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0043] Example 1:

[0044] This application provides a method for determining the uniformity of koji mixing in a brewing process. For details, please refer to [link to relevant documentation]. Figure 1 As shown, it includes the following steps:

[0045] S11: Obtain the image corresponding to the mixing area in the brewery.

[0046] S12: Extract image features from the image.

[0047] S13: Determine whether the mixing area is uniformly mixed based on image features and a preset mixing uniformity discrimination model; the mixing uniformity discrimination model is a model trained based on the image features of multiple first sample images and multiple second sample images; the first sample image is the image corresponding to the uniformly mixed mixing area, and the second sample image is the image corresponding to the unevenly mixed mixing area.

[0048] The above steps will be explained in detail below.

[0049] In this embodiment of the application, a camera can be installed inside the brewery.

[0050] In a first optional implementation, the camera image acquisition direction can be adjusted according to the position of the mixing area so that the camera only captures images of the mixing area, thereby obtaining the corresponding image.

[0051] In a second alternative implementation, the camera can take a global picture of the ground of the brewery to obtain a panoramic image, then determine the mixing area from the panoramic image, and obtain the image corresponding to the mixing area.

[0052] Regarding the second implementation method described above, the area for spreading and mixing the yeast in a brewery is typically large, and the location of the spreading and mixing is not fixed. A single camera may not be able to cover the entire mixing area. To ensure that the entire mixing area can be captured, a panoramic image of the entire brewery floor can be captured. Based on the field of view of each camera, several cameras are deployed at certain intervals on the top of the brewery, ensuring that there is an overlapping area between each pair of cameras. For details, please refer to [link to relevant documentation]. Figure 2 As shown. In this embodiment, the images captured by the top-deployed camera can be stitched together in pairs according to the overlapping areas, and then the gaps can be fused to obtain the final panoramic image.

[0053] The process of determining the koji mixing area from the panoramic image is described in detail below.

[0054] In this embodiment of the application, the mixing region can be determined from the panoramic image based on the panoramic image and the preset mixing region recognition model; the mixing region recognition model is a model trained based on multiple sample panoramic images and multiple mixing region mask images, and each mixing region mask image is an image obtained by annotating the mixing region of the corresponding sample panoramic image.

[0055] In the first example, a two-step method can be used to extract the mixing region. Generally, the koji spreading and mixing processes are performed sequentially, with mixing starting shortly after spreading. Since the color difference between the spreading and mixing regions is small, directly extracting the mixing region from a single frame image may introduce errors, easily mistaking the area where koji was just spread for the mixing region. However, the spreading and drying region differs significantly in color from the spreading and mixing region. Therefore, in this example, a deep learning method can be used to extract the first potential mixing region, and a temporal background modeling method can be used to extract the second potential mixing region. The intersection of the first and second potential mixing regions is taken as the final mixing region. The process will be explained in detail below.

[0056] In this example, a camera installed in the brewery can be used to capture images of the ground of the brewery to obtain a series of panoramic images that are continuous in time.

[0057] A first potential mixing region can be obtained by inputting one frame of the acquired multi-frame panoramic image into a preset mixing region recognition model. Specifically, any frame of the multi-frame panoramic image can be input into the preset mixing region recognition model. The average image feature change of each pixel in the panoramic image is determined based on the multi-frame panoramic image, and the region consisting of pixels whose average image feature change is greater than or equal to a preset change threshold is designated as the second potential mixing region. The intersection region between the first and second potential mixing regions is designated as the mixing region.

[0058] In this embodiment, a data-driven method based on deep learning can be used to segment the first potential mixing region, which is also known as the mixing area. It is understood that a model can be trained based on FCN, DeepLabV3, UNet, or U2Net to obtain a mixing region recognition model, and then the first potential mixing region can be determined based on this model. During the model training phase, a semantic segmentation annotation tool can be used to annotate the mixing regions in the sample panoramic image, generating a mixing region mask image. Then, using the original sample panoramic image and the corresponding mixing region mask image as training samples, a well-designed deep learning network is used to train the model. The trained model is stored for subsequent online mixing region extraction.

[0059] In this example, a panoramic image I can be input into a trained distorted region recognition model. The model performs inference and prediction on the panoramic image I to obtain the first potential distorted region. Specifically, the distorted region recognition model can perform binary prediction on each pixel in the panoramic image I. For example, when the predicted pixel value is 255, the pixel is determined to belong to the first potential distorted region; when the predicted pixel value is 0, the pixel is determined not to belong to the first potential distorted region. The binary prediction of each pixel in the panoramic image I by the distorted region recognition model is denoted as:

[0060]

[0061] Where 0≤i≤M, 0≤j≤N, M and N represent the length and width of the panoramic image I, respectively. This method yields the binary image B0 of the first potential mixing region. Since areas where koji has just been added may also be identified as the first potential mixing region by the model, but these areas have koji powder piled up and are generally yellowish-white, interfering with the uniformity count of the mixing region, it is necessary to further exclude areas where koji has just been added after obtaining the first potential mixing region, thus obtaining an interference-free mixing region.

[0062] The process of obtaining the first potential mixing region has been described in detail above. The process of obtaining the second potential mixing region will be described below.

[0063] Due to continuous manual stirring, the visual characteristics of the koji-mixing area change significantly, while the surface characteristics of the freshly sprinkled koji area do not change much over a certain period of time. Borrowing from background modeling methods, we can perform statistical modeling on each pixel over time. If the pixel brightness value of a certain pixel changes significantly over time, then that pixel belongs to the second potential koji-mixing area. Specifically, the average image feature change of each pixel in the panoramic image can be determined as follows:

[0064] For each pixel in the panoramic image at interval r frames, calculate its corresponding image feature change value;

[0065] Calculate the average value of the image feature changes corresponding to multiple frames of panoramic images, and use this average value as the average value of the image feature changes of the corresponding pixels; where r is an integer greater than 0.

[0066] In this embodiment, the image feature change value of each pixel can be any change value that can characterize the feature change of that pixel in the image. For ease of understanding, a specific example is provided below:

[0067] First, the panoramic image I at time t... t Each pixel (i,j) in the image is converted to the Lab color space, and the a-channel image features are extracted. The panoramic image at time tr I t-r Convert the pixel (i,j) in the image to the Lab color space and extract the feature image of channel a. The absolute frame difference at interval r frames is calculated as follows:

[0068]

[0069] D t (i,j) represents the image feature change value of pixel (i,j) in the panoramic image at interval r frames.

[0070] Secondly, for pixel (i,j), the average absolute frame difference over the past T frames is calculated, expressed as:

[0071]

[0072] F t (i,j) represents the average change of image features of pixel (i,j) in the panoramic image over the past T frames. It can be understood that the above process can be performed once for each T frames of panoramic image to obtain the average change of image features of each pixel over T frames.

[0073] It is understood that the specific sizes of T and r in the embodiments of this application can be flexibly set by the developers.

[0074] Finally, based on the preset threshold ε, it is calculated whether each pixel is located within the second potential mixing region. The binary image B1 of the second potential mixing region at the current time can be represented as:

[0075]

[0076] In the binary image, pixels with a value greater than 0 in B1 belong to the second potential mixing region.

[0077] After obtaining the binary image B0 of the first potential mixing region and the binary image B1 of the second potential mixing region, the intersection of the two binary images is calculated to obtain the final binary image of the interference-free mixing region, represented as:

[0078] B(i,j)=B0(i,j)∩B1(i,j).

[0079] The above describes in detail the specific process of extracting the koji mixing area using the two-step method. It is understood that in other embodiments, the koji mixing area can also be determined by other means.

[0080] It should be noted that the mixing region is essentially composed of multiple pixels. For ease of description, the pixels within the mixing region are referred to as target pixels. The image features corresponding to each target pixel may change at different times. In step S11, after determining the mixing region (i.e., the target pixels) based on the single-frame panoramic image A and multiple panoramic images, the image content corresponding to the target pixel in any one of these panoramic images can be used as the image corresponding to the mixing region in step S11. For example, the image content corresponding to the target pixel in the single-frame panoramic image A can be used as the image corresponding to the mixing region in step S11.

[0081] In the second example, the first potential mixing region obtained in the above manner can be directly used as the mixing region in step S11.

[0082] In the third example, the second potential mixing region obtained in the above manner can be directly used as the mixing region in step S11.

[0083] During the mixing process, if the mixing is insufficient or infrequent, the sprinkled koji will clump together and become darker in color, indicating uneven mixing. Conversely, areas with even mixing will be lighter in color. Therefore, exemplarily, in this embodiment, local color distribution features are used as statistical features; that is, the image features in this embodiment can be the color distribution features of pixels. Thus, in step S12, the color distribution features of each pixel in the image can be extracted.

[0084] The following is a detailed explanation of step S13.

[0085] In some embodiments, the image corresponding to the mixing area can be input into the mixing uniformity discrimination model. This model can be a binary classification model that outputs information indicating whether the mixing area is uniform or uneven. In this case, the binary classification model can be trained as follows:

[0086] The model is trained using multiple first sample images labeled "uniform mixing" and multiple second sample images labeled "uneven mixing". The resulting model is a binary classification model that can be used to classify images corresponding to mixing regions.

[0087] In other embodiments, the mixing uniformity discrimination model is a model trained based on the color distribution features of each pixel in the first sample image and the color distribution features of each pixel in the second sample image. In this embodiment, a classification model can be trained using the Logistic Regression machine learning method to determine the mixing uniformity value of each pixel, thereby distinguishing between uniform and non-uniform pixels.

[0088] At this time, please see Figure 3 As shown, step S13 may include the following sub-steps:

[0089] S121: Determine the uniformity value of each pixel in the image based on the color distribution characteristics of each pixel and the preset uniformity discrimination model.

[0090] S122: Calculate the overall koji uniformity value corresponding to the image based on the koji uniformity value.

[0091] S123: Compare the overall uniformity value of the koji mixing with the preset uniformity threshold.

[0092] S124: Determine whether the koji mixing area is uniform based on the comparison results.

[0093] During the calculation of color distribution features, the panoramic image I can be converted to the Lab color space, and the Lab channel feature images are represented as I0, I1, I2, I3, I4, I5, I6, I7, I8, I9, I10< / L I a and I b ,like Figure 4 As shown, assuming a local image patch of size 5x5 receptive field is used as the basic unit, the color distribution features of the Lab three channels of pixel coordinate (i,j) can be expressed as follows:

[0094] f L =[I L (i-2,j-2),I L (i-2,j-1),…I L (i,j),…,I L (i+2,j+1),I L (i+2,j+2)]

[0095] f a =[I a (i-2,j-2),I a (i-2,j-1),…Ia (i,j),…,I a (i+2,j+1),I a (i+2,j+2)]

[0096] f b =[I b (i-2,j-2),I b (i-2,j-1),…I b (i,j),…,I b (i+2,j+1),I b (i+2,j+2)]

[0097] Channel feature f L f a and f b By cascading, we obtain the final color distribution feature description of pixel coordinates (i,j), that is:

[0098] f = [f L ,f a ,f b ];

[0099] The length of the color distribution feature f is 75.

[0100] During the model training phase, multiple images of uniformly mixed and non-uniformly mixed koji are collected. Pixels in the uniformly mixed images are designated as uniformly mixed pixels, and pixels in the non-uniformly mixed images are designated as non-uniformly mixed pixels. The color distribution features of uniformly mixed pixels and non-uniformly mixed pixels are obtained and calculated. The color distribution feature set of uniformly mixed pixels is denoted as Ω0, the total number of uniformly mixed pixels is m, the color distribution feature set of non-uniformly mixed pixels is denoted as Ω1, and the total number of non-uniformly mixed pixels is n. In this embodiment, m = 1500 and n = 1500 are used. The Logistic regression method is used to train classification models on Ω0 and Ω1, and the models are stored.

[0101] When calculating uniformity online, specifically in step S12, the color distribution features of each pixel p(i,j) in the image can be calculated. Then, in step S13, the trained Logistic regression model, i.e., the uniformity discrimination model, is used to perform binary prediction on this feature, outputting the current uniformity value of the pixel. The uniformity distribution map corresponding to the current image can be represented as follows:

[0102]

[0103] Γ(i,j) represents the predicted value of pixel p(i,j) in the image by the mixing uniformity discrimination model. When the predicted value of a pixel is 0, it indicates that the pixel is a non-uniform pixel. When the predicted value of a pixel is 1, it indicates that the pixel is a uniform pixel. B represents the mixing region. The overall mixing uniformity value of the image is calculated based on the mixing uniformity value, which is to take the average value of Γ(i,j). The value range is [0,1]. The larger the value, the more uniform it is.

[0104] In one optional implementation, when the overall uniformity value of the koji mixing is greater than or equal to a preset uniformity threshold, the koji mixing area can be determined to be a uniformly mixed area; when the overall uniformity value of the koji mixing is less than the preset uniformity threshold, the koji mixing area can be determined to be an unevenly mixed area.

[0105] It is understood that the uniformity threshold in the embodiments of this application can be flexibly set by the developers based on experience or actual application scenarios. Assuming that the overall uniformity value of the current mixing image is 0.8958 and the uniformity threshold is 0.5, it indicates that the mixing area is currently uniform.

[0106] In some embodiments, after determining the uniformity value of each pixel in the image based on the color distribution characteristics of each pixel and a preset uniformity discrimination model, a heatmap corresponding to the image can be generated based on the uniformity value of each pixel. The color depth of the heatmap represents different degrees of uniformity. For details, please refer to [link to relevant documentation]. Figure 5 As shown. For example, when it is determined that the koji mixing area is uneven, a heat map can be generated to prompt the user.

[0107] In some embodiments, when it is determined that the mixing of koji in the mixing area is uneven, an alarm message can be automatically reported so that staff can take timely measures to reduce the decline in fermentation quality of the mash caused by uneven mixing.

[0108] The method for determining the uniformity of koji mixing in the brewing process provided in this application embodiment can automatically determine whether the koji mixing area is uniform. If it is not uniform, an alarm message is reported, greatly reducing the problem of reduced fermentation quality caused by uneven mixing. Compared with traditional methods relying on human visual identification, it avoids the disadvantages of visual fatigue and strong subjectivity, and improves the accuracy of judgment. In addition, it can also output a uniformity heat map, which can quickly locate uneven areas, making it convenient for process personnel to remix until the overall uniformity meets the requirements.

[0109] In summary, the method for judging the uniformity of koji mixing in the brewing process provided in this application has the characteristics of high automation, strong objectivity, and high accuracy. In the koji mixing process of brewing, this solution can replace the traditional manual monitoring solution and reduce manual intervention.

[0110] It should be understood that although the steps in the flowchart above are shown sequentially as indicated by the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowchart above may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these sub-steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the sub-steps or stages of other steps.

[0111] Example 2:

[0112] This embodiment provides a system for judging the uniformity of koji mixing in the brewing process. Please refer to [link to relevant documentation]. Figure 6 As shown, the system for judging the uniformity of koji mixing in the brewing process includes a processor 601, a memory 602, and a camera 603. The memory 602 stores a computer program. The processor 601 and the memory 602 communicate through a communication bus. The processor 601 executes the computer program to implement the steps of the method in the above embodiments, which will not be described in detail here.

[0113] Understandable. Figure 6 The structure shown is for illustrative purposes only. The system for determining the uniformity of koji mixing in the brewing process may also include a comparison... Figure 6 The more or fewer components shown, or having the same Figure 6 The different configurations shown.

[0114] Processor 601 can be an integrated circuit chip with signal processing capabilities. The processor 601 can be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), etc.; it can also be a digital signal processor (DSP), an application-specific integrated circuit (ASIC), an off-the-shelf programmable gate array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components. It can implement or execute the various methods, steps, and logic block diagrams disclosed in the embodiments of this application. The general-purpose processor can be a microprocessor or any conventional processor.

[0115] The memory 602 may include, but is not limited to, random access memory (RAM), read-only memory (ROM), programmable read-only memory (PROM), erasable read-only memory (EPROM), electrically erasable read-only memory (EEPROM), etc.

[0116] This embodiment also provides a computer-readable storage medium, such as a floppy disk, optical disk, hard disk, flash memory, USB flash drive, SD card, MMC card, etc., in which one or more programs implementing the above steps are stored. These one or more programs can be executed by one or more processors 601 to implement the steps of the methods in the above embodiments, which will not be described again here.

[0117] It should be noted that the illustrations provided in this embodiment are only schematic representations of the basic concept of the present invention. Therefore, the drawings only show components relevant to the present invention and are not drawn according to the actual number, shape, and size of components in implementation. In actual implementation, the form, quantity, and proportion of each component can be arbitrarily changed, and the component layout may be more complex. The structures, proportions, sizes, etc., depicted in the accompanying drawings are only for illustrative purposes to aid those skilled in the art and are not intended to limit the implementation conditions of the present invention. Therefore, they have no substantial technical significance. Any modifications to the structure, changes in proportions, or adjustments to size, without affecting the effects and objectives of the present invention, should still fall within the scope of the technical content disclosed in the present invention. Furthermore, the terms such as "upper," "lower," "left," "right," "middle," and "one" used in this specification are only for clarity of description and are not intended to limit the scope of the present invention. Changes or adjustments to their relative relationships, without substantially altering the technical content, should also be considered within the scope of the present invention.

[0118] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0119] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the invention patent. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this patent application should be determined by the appended claims.

Claims

1. A method for judging the uniformity of koji mixing in a brewing process, characterized in that, The method includes: Obtain the image corresponding to the koji mixing area in the brewery; Extract the image features of the image; The image features and a preset mixing uniformity discrimination model are used to determine whether the mixing region is uniformly mixed. The mixing uniformity discrimination model is a model trained based on the image features of multiple first sample images and multiple second sample images. The first sample images are the images corresponding to the uniformly mixed mixing region, and the second sample images are the images corresponding to the unevenly mixed mixing region. The acquisition of the image corresponding to the koji mixing area in the brewery includes: By using cameras installed in the brewery to capture images of the ground of the brewery, a series of panoramic images in time can be obtained. A frame of the panoramic image is input into a preset mixing region recognition model to obtain a first potential mixing region; the mixing region recognition model is a model trained based on multiple sample panoramic images and multiple mixing region mask images, and each mixing region mask image is an image obtained by annotating the corresponding sample panoramic image with mixing regions. Based on multiple frames of the panoramic image, the average value of image feature change of each pixel in the panoramic image is determined, and the region composed of pixels whose average value of image feature change is greater than or equal to a preset change threshold is taken as the second potential mixing region. The intersection area between the first potential mixing region and the second potential mixing region is taken as the mixing region, and the image corresponding to the mixing region is obtained.

2. The method for determining the uniformity of koji mixing in the brewing process as described in claim 1, characterized in that, Determining the average change of image features for each pixel in the panoramic image based on multiple frames of the panoramic image includes: For each pixel in the panoramic image at interval r frames, calculate its corresponding image feature change value; Calculate the average value of the image feature changes corresponding to multiple frames of the panoramic image, and use the average value as the average value of the image feature changes of the corresponding pixel; wherein, r is an integer greater than 0.

3. The method for determining the uniformity of koji mixing in the brewing process as described in claim 1 or 2, characterized in that, The image features include the color distribution features of each pixel; The step of determining whether the mixing region is uniformly mixed based on the image features and a preset mixing uniformity discrimination model includes: The uniformity value of mixing for each pixel in the image is determined based on the color distribution characteristics of each pixel and the preset uniformity discrimination model. Calculate the overall koji mixing uniformity value corresponding to the image based on the koji mixing uniformity value; The overall uniformity value of the koji mixing is compared with a preset uniformity threshold. The comparison results determine whether the koji mixing area is uniformly mixed.

4. The method for judging the uniformity of koji mixing in the brewing process as described in claim 3, characterized in that, After determining the uniformity value of mixing for each pixel in the image based on the color distribution characteristics of each pixel and a preset uniformity discrimination model, the process includes: A heatmap corresponding to the image is generated based on the uniformity value of the mixing at each pixel in the image.

5. The method for determining the uniformity of koji mixing in the brewing process as described in claim 1 or 2, characterized in that, The method further includes: If it is determined that the mixing of the starter culture is uneven in the mixing area, an alarm message is reported.

6. A system for judging the uniformity of koji mixing in a brewing process, characterized in that, The device includes a processor, a memory, and a camera, wherein the memory stores a computer program, and the processor executes the computer program to implement the method as described in any one of claims 1-5.

7. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by at least one processor, implements the method as described in any one of claims 1-5.

Images