Accumulation fermentation temperature measurement digitalization method and device based on infrared perception and modeling reasoning

Abstract

The invention discloses a heap fermentation temperature measurement digital method and device based on infrared sensing and modeling reasoning, which can perform self-adaptation according to the position, number and layout change of fermented grain heaps without fixed installation in a workshop by adopting dynamic identification and mobile temperature measurement strategies. The problems that a traditional temperature measurement means is poor in adaptability and high in layout cost in a flexible production environment are effectively solved, and the universality and deployability of the system are remarkably improved. Moreover, a non-contact optical sensing means is adopted in the whole process, so that the damage to the fermented grain pile structure is avoided, the integrity and stability of the fermentation process are ensured, and the final wine quality is favorably improved.

Description

Technical Field

[0001] This invention relates to the field of digital technology in the production process of Maotai liquor, specifically to a digital method and device for temperature measurement during stacked fermentation based on infrared sensing and modeling reasoning. Background Technology

[0002] The production process of Maotai-flavor baijiu is extremely unique and complex. Among them, "piling fermentation" (also known as "open fermentation" or "aerobic fermentation") is the key step in forming its rich and elegant Maotai-flavor. In this process, the yeast obtained through "high-temperature yeast making" is mixed with steamed sorghum (i.e., "fermentation mash") and piled on the floor of the brewing workshop into conical or hemispherical mounds (known in the industry as "fermentation mash piles"). Without the addition of any foreign microorganisms, the fermentation mash relies entirely on the natural microbial system in the yeast for open secondary fermentation.

[0003] Temperature is the soul of pile fermentation. The process requires the temperature of the fermented mash pile to undergo a complete cycle of "slow rise - rapid rise - steady rise - slow fall," with the temperature needing to increase from an initial temperature of around 30℃ to 45-50℃, or even higher. During pile fermentation, temperature not only affects the microbial community structure and fermentation activity but also directly relates to the synthesis efficiency and proportion of specific aroma compounds (such as ethyl hexanoate and ethyl butyrate). A temperature curve of "slow at the beginning, rapid in the middle, and slow at the end" is considered the ideal state. Therefore, monitoring and analyzing the temperature throughout the entire process is crucial for achieving the digitalization and standardization of the fermentation process.

[0004] However, the temperature measurement methods used in Maotai-flavor liquor production workshops remain very traditional. Distillers primarily rely on experience, manually measuring the temperature by inserting a long, pointed metal thermometer into the mash pile at different depths and positions based on their sense of touch. This method suffers from a series of insurmountable drawbacks: 1. Difficulty in ensuring process compliance: Low frequency and few measurement points for manual temperature measurement make it impossible to capture the complete dynamic process of temperature changes, making it difficult to ensure that the actual temperature curve matches the "golden curve" required by the process. Once the internal temperature of the pile is abnormal, such as excessively rapid local heating or uneven heat distribution, manual detection usually has a response lag, making it difficult to adjust the turning time or ventilation strategy in time, which can easily lead to abnormal fermentation. 2. Incomplete data with blind spots: The temperature difference between the inside and outside of the mash pile is huge, with the core temperature and surface temperature potentially differing by more than ten degrees. Manual measurement can only obtain data from a few points, failing to reflect the temperature field distribution of the entire pile. This makes it very easy for localized excessively high temperatures ("piling up") or excessively low temperatures ("insufficient fermentation") to go undetected in time. 3. Flexible production layout, high monitoring difficulty: In traditional brewing workshops, the location and quantity of mash piles for each production cycle are not completely fixed and are adjusted according to the production schedule. At the same time, the geometry and size of the mash piles also vary, and parameters such as pile height and density vary greatly, making it impossible for wired temperature sensors deployed at fixed points to adapt to the diverse spatial structures; 4. High labor intensity and strong reliance on experience: The workshop is characterized by high temperature and humidity, making manual temperature measurement extremely labor-intensive. More importantly, the accuracy and representativeness of temperature measurements, as well as subsequent decisions regarding turning the compost, heavily depend on the personal experience of the brewers, making it difficult to establish a standardized and replicable production model. 5. Lack of digitization and traceability: Handwritten temperature data is fragmented, making it difficult to conduct systematic data analysis, quality traceability, and process optimization.

[0005] Therefore, this invention proposes a new digital method and device for temperature measurement of stacked fermentation based on infrared sensing and modeling reasoning. Summary of the Invention

[0006] (I) Technical Solution To achieve the above objectives, the present invention provides the following technical solution: a digital method for temperature measurement in stacked fermentation based on infrared sensing and modeling reasoning, comprising the following steps: S1. Dynamic recognition and positioning: Visible light and infrared dual-mode vision system mounted on a mobile platform is used to acquire visible light and infrared images of mash piles in the brewing workshop; computer vision algorithms are used to process the visible light images to automatically recognize, locate and extract the outline of the mash piles, and automatically number the multiple mash piles that are recognized. S2. Non-contact multi-dimensional perception and modeling: The mobile platform is controlled to inspect the numbered mash piles according to a preset path or a self-planned trajectory; during the inspection, a high-resolution infrared thermal image of the surface of the mash pile is acquired simultaneously using an infrared thermal imager, and a three-dimensional geometric model of the mash pile is acquired simultaneously using a three-dimensional imaging device. S3. Temperature Field Inversion and Fermentation State Reasoning: Spatial registration and data fusion are performed between the infrared thermogram and the three-dimensional geometric model; using the fused data as input, and combining the heat generation mechanism of Maotai-flavor liquor fermentation, a heat conduction model including heat source terms is constructed. By solving the heat conduction model, the three-dimensional temperature field distribution inside the mash pile is inverted; based on the evolution characteristics of the three-dimensional temperature field distribution, at least one of the following indicators is extracted: temperature rise rate, high temperature zone proportion, and temperature gradient distribution, and compared with a preset "golden temperature curve" to evaluate the current fermentation state, and generate early warning information and process optimization suggestions when temperature anomalies are detected; S4. Digital Management and Visualization: The numbering, location information, three-dimensional geometric model, three-dimensional temperature field, fermentation status assessment results, early warning information and process optimization suggestions of the fermentation mash pile are integrated and visualized on the digital management platform, and the temperature change history is automatically generated to support historical data traceability and process analysis.

[0007] Furthermore, in step S1, the mobile platform is a gimbal (PTZ) or a track-mounted mobile platform; the computer vision algorithm includes traditional algorithms based on contour recognition and color analysis, or deep learning object detection and instance segmentation algorithms based on YOLO and Mask R-CNN.

[0008] Furthermore, step S1 also includes: based on the identified location and outline of the mash pile, establishing a digital file for each mash pile containing its location, size, and number, and autonomously planning an orderly inspection and measurement path for multiple mash piles based on this file.

[0009] Furthermore, in step S2, the three-dimensional imaging device is a lidar or a structured light camera; the three-dimensional geometric model includes spatial structural parameters such as the volume, height, and surface area of ​​the fermentation mash pile.

[0010] Furthermore, in step S3, the construction of the heat conduction model including the heat source term specifically includes: based on the microbial metabolic activities of the fermentation of soy sauce-flavored liquor, establishing a typical heat production rate curve that varies with time and space, and substituting the heat production rate curve as the heat source term into the three-dimensional unsteady heat conduction differential equation established based on Fourier's law and the law of conservation of energy.

[0011] Furthermore, the method for solving the heat conduction model is the finite element method or the finite volume method, and the surface temperature distribution of the stack provided by the infrared thermogram is used as the boundary condition for solving the heat conduction model.

[0012] Furthermore, in step S3, the assessment of the current fermentation state includes: determining whether the fermentation process is in the heating phase, high-temperature phase, or cooling phase; the process optimization suggestions include specific time points for operations such as turning the pile, mixing the starter culture, and covering, for example, "It is recommended to turn the pile after 6 hours".

[0013] This invention also provides a digital device for measuring temperature during stacked fermentation based on infrared sensing and modeling reasoning, comprising: Infrared sensing module, used for non-contact acquisition of infrared thermal images and three-dimensional geometric structure data of fermentation mash pile; The edge computing and preprocessing unit is communicatively connected to the infrared sensing module. It is used to receive the infrared thermal image and three-dimensional geometric structure data, and to perform data cleaning, registration and coordinate transformation to generate thermal image blocks with three-dimensional coordinates. The model computing server is communicatively connected to the edge computing and preprocessing unit. It is used to receive the heat map with three-dimensional coordinates, and based on the heat conduction model and the preset heat generation mechanism, it inverts and constructs the three-dimensional temperature field inside the mash pile, and then extracts key process indicators, and combines the process knowledge base to perform fermentation state reasoning and prediction. The process knowledge base, which communicates with the model computing server, is used to store temperature windows, historical reference data, and process expert rules for the fermentation stage. The user interaction terminal is connected to the model computing server and is used to receive and visualize the three-dimensional temperature field, key process indicators, fermentation state reasoning and prediction results, and early warning information.

[0014] (II) Beneficial Effects This invention provides a digital method and apparatus for temperature measurement in stacked fermentation based on infrared sensing and modeling reasoning. Compared with the prior art, the advantages of this invention are: 1. By employing a dynamic identification and mobile temperature measurement strategy, the system can autonomously adapt to changes in the location, quantity, and layout of the fermentation mash piles without requiring fixed installation within the workshop. This effectively solves the problems of poor adaptability and high deployment costs associated with traditional temperature measurement methods in flexible production environments, significantly improving the system's versatility and deployability. Furthermore, the use of non-contact optical sensing throughout the process avoids damage to the fermentation mash pile structure, ensuring the integrity and stability of the fermentation process and contributing to improved final product quality.

[0015] 2. By introducing the fusion technology of infrared thermal imaging and 3D geometric modeling, non-contact 3D reconstruction of the internal temperature field of the fermentation mash pile is achieved, enabling visualization and comprehensive perception of the 3D thermal field during fermentation. This overcomes the limitations of traditional manual "point measurement" methods and can be visualized as a "CT-level" thermal imaging diagnosis of the fermentation mash pile. Not only is precise surface temperature obtained, but more importantly, through mechanistic model inversion, "indirect perception" of the temperature in the core area inside the fermentation mash pile is achieved, solving the technical problem that traditional methods cannot reach the interior. The temperature measurement data is more representative and comprehensive. Brewing personnel can fully grasp the temperature status of each area inside the pile, greatly improving the transparency and controllability of the fermentation process.

[0016] 3. The system features automatic inspection, modeling, and intelligent judgment of fermentation status. Through an autonomously moving vision system and path planning, it achieves fully automated identification, positioning, and orderly inspection of all mash piles in the workshop, significantly reducing the labor intensity of manual temperature measurement, replacing experience-based judgment, improving monitoring efficiency, and enabling 24-hour uninterrupted monitoring. By transforming the temperature control standards and judgment experience of master brewers into algorithmic models, it achieves structured expression and replicable scalability of process knowledge, effectively improving product consistency and yield.

[0017] 4. The system automatically generates an independent temperature change history file for each fermentation mash pile, recording key fermentation parameters and forming a complete temperature control data chain. All data is visualized on the management platform, providing a reliable foundation for subsequent quality tracking, process optimization, data analysis, and problem tracing. Attached Figure Description

[0018] Figure 1 This is a schematic diagram of the overall structure of the device provided in Embodiment 2 of the present invention; Figure 2 This is a flowchart of infrared image acquisition and processing provided in Embodiment 1 of the present invention. Detailed Implementation

[0019] 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.

[0020] Example 1 like Figure 2 As shown, this embodiment provides a digital method for temperature measurement in stacked fermentation based on infrared sensing and modeling reasoning, including the following steps: S1. Dynamic Identification and Positioning: A dual-mode vision system with visible light and infrared capabilities, mounted on a pan-tilt-zoom (PTZ) or track-mounted mobile platform, collects environmental data from the brewing workshop, such as visible light and infrared images of the mash piles. Computer vision algorithms (such as contour recognition, color analysis, or deep learning models based on YOLO, Mask R-CNN, etc.) are used to process the visible light images, automatically identifying all mash piles in the workshop, accurately obtaining their two-dimensional position coordinates and contours, automatically numbering the identified mash piles, and establishing digital archives containing information such as location and size. This allows for the autonomous planning of efficient and comprehensive inspection and measurement paths. S2. Non-contact multi-dimensional perception and modeling: The mobile platform is controlled to perform close-range, multi-angle fine-grained inspections of the numbered mash piles according to a preset path or a self-planned trajectory. During the inspection, a high-resolution infrared thermal imager is used to simultaneously acquire high-resolution infrared thermal images of the mash pile surface, reflecting its surface temperature distribution. A three-dimensional imaging device (such as a lidar or structured light camera) is used to simultaneously acquire spatial point cloud data of the pile and reconstruct a high-precision three-dimensional geometric model of the mash pile containing spatial structural parameters such as volume, height, and surface area. S3. Temperature Field Inversion and Fermentation State Inference: The infrared thermogram (surface temperature field) and the three-dimensional geometric model (spatial structure) are spatially registered and fused. Using the fused data as input, and combining the microbial heat production mechanism of Maotai-flavor liquor fermentation (such as the oxygen consumption and metabolic heat production rate of different bacterial groups), a three-dimensional unsteady heat conduction model (such as a model based on finite element analysis) containing dynamic heat source terms is constructed. This model uses the surface temperature provided by the infrared thermogram as the boundary condition. By solving the heat conduction model (i.e., based on the microbial metabolic activities of Maotai-flavor liquor fermentation, a typical heat production rate curve that varies with time and space is established, and this heat production rate curve is substituted as a heat source term into the three-dimensional unsteady heat conduction differential equation established based on Fourier's law and the law of conservation of energy), the interior of the mash pile, especially the fermentation core, is inverted. The system analyzes the three-dimensional temperature field distribution of the region. Based on the evolution characteristics of the three-dimensional temperature field distribution, it extracts key characteristic indicators, such as the overall temperature rise rate, the proportion of high-temperature zones exceeding the threshold, and at least one indicator (vertical and horizontal) in the temperature gradient distribution. These indicators are then compared with a "golden temperature curve" (i.e., the temperature change standard under ideal fermentation conditions) summarized from preset expert experience or historical data to assess the current fermentation state (e.g., heating period, high-temperature period, cooling period). When an abnormal temperature is detected, the system generates early warning information and process optimization suggestions. When a temperature indicator deviates from the normal range, the system automatically triggers an early warning and generates specific process optimization suggestions based on the mechanism model and knowledge base. These process optimization suggestions include specific time node prompts for operations such as turning, mixing, and covering, such as "It is recommended to turn the pile after 6 hours to dissipate heat and equalize the temperature." S4. Digital Management and Visualization: The numbering, location information, 3D geometric model, 3D temperature field, fermentation status assessment results, early warning information, and process optimization suggestions of the fermentation mash piles are uniformly integrated into a digital temperature control management platform. This platform provides a visual display of the overall macroscopic status of all fermentation mash piles in the workshop in the form of an electronic map. Clicking on any pile allows viewing its detailed multi-dimensional information. The platform automatically generates a complete temperature change history curve for each fermentation mash pile, supporting historical data query, comparative analysis, and process optimization effect evaluation, providing production managers with powerful data-driven decision support.

[0021] Example 2 like Figure 1 As shown, the present invention also provides a digital device for measuring temperature in stacked fermentation based on infrared sensing and modeling reasoning, including an infrared sensing module 100, an edge computing and preprocessing unit 200, a model computing server 300, a process knowledge base 400, and a user interaction terminal 500.

[0022] The infrared sensing module, serving as the system's data acquisition front-end, is used for non-contact acquisition of infrared thermal images and three-dimensional geometric structure data of the fermentation mash pile. It consists of an infrared thermal imager 110 and a three-dimensional scanning device 120. The infrared thermal imager is responsible for acquiring the temperature distribution on the surface of the pile and collecting infrared thermal images of the fermentation mash pile. It can be a pan-tilt-zoom camera with self-rotation capabilities, achieving multi-angle scanning through horizontal and vertical rotation, or a fixed module mounted on a moving platform such as a track or mobile robot, scanning different fermentation mash piles through movement or rotation. The three-dimensional scanning device, such as a lidar, structured light camera, or depth camera, is responsible for acquiring the precise three-dimensional contour of the pile and collecting three-dimensional point cloud data or depth images of the fermentation mash pile. Both work synchronously to ensure a one-to-one correspondence between temperature data and spatial structure data.

[0023] The edge computing and preprocessing unit, acting as a bridge between the perception and analysis layers, is deployed on-site in the workshop. It communicates with the infrared sensing module to receive the infrared thermal images and 3D geometric data, performing real-time preprocessing. This includes: correcting distortion and temperature drift in the thermal images, removing abnormal pixels to ensure data accuracy; and mapping the 2D infrared thermal image pixels to a 3D coordinate system constructed by the 3D scanning device using pre-calibrated parameters, generating "thermal image blocks with 3D coordinates," giving each temperature data point physical location information. This step significantly reduces the computational burden on subsequent servers and ensures data quality.

[0024] The model computing server is communicatively connected to the edge computing and preprocessing unit. It is used to receive the heat map with three-dimensional coordinates, and based on the heat conduction model and the preset heat generation mechanism, it inverts and constructs the three-dimensional temperature field inside the mash pile, and then extracts key process indicators, and combines the process knowledge base to perform fermentation state reasoning and prediction. Specifically, (1) Three-dimensional temperature field reconstruction: receiving the pre-processed data, using the surface temperature of the pile as the known boundary condition, combining the heat generation mechanism model of the fermentation of soy sauce-flavored liquor (such as the heat generation rate function at different stages), and using numerical methods such as finite element analysis to solve the partial differential equation of heat conduction, thereby reversing the three-dimensional temperature field distribution of the invisible area inside the mash pile.

[0025] (2) Extraction of key indicators: From the reconstructed three-dimensional temperature field, a series of core process indicators are automatically calculated and extracted, such as the highest temperature and location of the core of the reactor, the overall average temperature, the temperature rise rate, the volume ratio of the high-temperature zone exceeding a specific threshold, and the temperature gradient vector field that characterizes the direction and speed of heat transfer.

[0026] (3) State reasoning and prediction: The extracted real-time indicators and time series are input into the reasoning model and compared with the "golden temperature range" and historical high-quality batch templates stored in the process knowledge base. Through this process, the system can automatically determine whether the current fermentation is in the heating period, high temperature period or cooling period, predict the temperature trend in the next few hours, and immediately generate early warning information once an abnormal temperature is detected (such as excessive heating or core temperature exceeding the standard). It also provides specific process operation suggestions in combination with the knowledge base (such as "it is recommended to turn the pile after 6 hours").

[0027] The process knowledge base, communicating with the model computation server, is a structured database storing the experience and knowledge of brewing experts, historical production data, and process standards. This includes ideal temperature windows for different fermentation stages, reference curves for different raw material ratios, and key parameter templates for historically successful batches. The knowledge base supports dynamic updates, enabling the system to continuously learn and optimize.

[0028] The user interaction terminal 500 serves as the window for human-computer interaction, and can be a PC client, a mobile app, or a large screen in the workshop. It communicates with the model computation server to receive and present the server's analysis results to the user in an intuitive and visual manner. The visualized content includes: a linked display of real-time video and infrared thermograms; interactive, sliceable 3D temperature cloud maps; historical line graphs of key process indicators; fermentation stage prompts, process suggestions, and automatic early warning windows; historical batch comparisons; and data export functions.

[0029] Working Process: In a typical brewing workshop application scenario, a top-mounted track-type slide can be used as a moving platform. A non-rotating infrared thermal imager 110 and a lidar as a 3D scanning device 120 are mounted together on the slide's trolley. The slide moves along a track laid on the workshop ceiling, covering all the mash piles below.

[0030] When the system starts working, the slide moves to the top of the first fermentation mash pile (pile-A), and the infrared thermal imager 110 and lidar 120 are activated simultaneously to collect infrared thermal images and 3D point cloud data of the surface of pile-A, respectively. After the data acquisition is completed, the slide moves to the next fermentation mash pile (pile-B) and repeats the above process until all piles are scanned.

[0031] The edge computing and preprocessing unit 200 is deployed in the control cabinet of the workshop, using industrial-grade edge computing equipment equipped with NVIDIA Jetson series chips. After the infrared sensing module 100 completes the acquisition of data from the stack-A, the raw data is immediately transmitted to this unit via the network. The software within the unit first performs distortion correction on the infrared thermal image to eliminate image distortion caused by the lens; then, it performs temperature normalization using a preset blackbody reference source to correct temperature drift caused by environmental changes. The most important step is coordinate fusion: the software uses the extrinsic parameter matrix between the infrared camera and the lidar, which is obtained in advance through a calibration board, to accurately map each pixel (including the temperature value T) in the infrared thermal image to the three-dimensional point cloud coordinate system generated by the lidar, forming a dataset containing a (x, y, z, T) quadruple, i.e., a "thermal image patch with three-dimensional coordinates". After processing, this data patch is sent to the model computing server 300.

[0032] The model computation server 300 is a high-performance server that hosts all the core algorithms. Taking the data from pile-A as an example, after the server receives the data block, the three-dimensional temperature field reconstruction model 310 begins operation. This model, based on Fourier's law of heat conduction, constructs a physical model with the three-dimensional geometry of pile-A as its domain. It uses the temperature data with coordinates as the first type of boundary condition (known surface temperature). Simultaneously, it calls the typical heat production rate function of the current fermentation stage (such as the heating period) from the process knowledge base 400 as the internal heat source term. Then, the finite element method is used to mesh the entire domain, and the heat conduction equation is solved iteratively to finally obtain the temperature value of each grid point inside pile-A, forming a complete three-dimensional temperature field dataset.

[0033] Next, the key indicator extraction algorithm 320 analyzes the three-dimensional temperature field dataset. For example, by traversing all grid points, the highest temperature value and its coordinates are found and identified as the core temperature point; the average temperature of all grid points is calculated to obtain the overall average temperature; compared with the temperature field at the previous moment, the overall temperature rise rate is calculated; and the proportion of grid volume with temperature higher than 55℃ to the total pile volume is statistically analyzed to obtain the proportion of high-temperature zone.

[0034] Finally, the state prediction and inference module 330 compares the extracted indicators (such as core temperature 58℃, temperature rise rate 2.5℃ / h, and high-temperature zone percentage 30%) with the "golden temperature range during the heating period" (such as core temperature 50-56℃, temperature rise rate 1-2℃ / h) defined in the process knowledge base 400. It finds that both the current core temperature and temperature rise rate have exceeded the limits, and the system determines that pile-A has a risk of "fermentation overheating". At this point, the inference module 330 not only marks the current state as "abnormal", but also generates a specific process recommendation based on expert rules in the knowledge base: "Fermentation overheating of pile-A has been detected. It is recommended to perform a turning operation after 6 hours to reduce the core temperature and promote uniform fermentation." The user interaction terminal 500 receives and displays all analysis results. On the large PC screen in the workshop management center, the icon for pile-A on the electronic map begins to flash and turns red, indicating an anomaly. The operator clicks on the pile-A icon, and the main interface immediately displays a 3D temperature cloud map of pile-A. The operator can rotate and zoom the map using the mouse and use the virtual slice tool to view the temperature distribution of any internal cross-section. The adjacent dashboard displays key indicators such as core temperature and temperature rise rate in real time, highlighting any exceeding limits in red. Simultaneously, a warning window pops up on the screen, clearly displaying the judgment of "fermentation overheating" and the suggestion to "turn the pile in 6 hours." All data is automatically recorded, forming the temperature history of pile-A, which can be queried and analyzed later.

[0035] The core technical concept of this invention lies in combining "dynamic identification and positioning" with "non-contact infrared modeling" to achieve automated, precise, and visual monitoring of the temperature of the mash pile during the fermentation process of Maotai-flavor liquor. Through the coordinated work of the above modules, the device of this invention achieves full-process automation from data acquisition, preprocessing, intelligent analysis to decision support, completely changing the traditional temperature measurement mode that relies on manual methods, and providing strong technical support for the refined production of Maotai-flavor liquor.

[0036] It should be noted that the above examples are merely specific embodiments of the present invention, and the present invention is obviously not limited to the above embodiments, with many similar variations. All modifications that can be directly derived or conceived by those skilled in the art from the content disclosed in this invention should fall within the protection scope of this invention.

[0037] The above are merely preferred embodiments of the present invention and are not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A digital method for temperature measurement in stacked fermentation based on infrared sensing and modeling reasoning, characterized in that, Includes the following steps: S1. Dynamic recognition and positioning: Visible light and infrared dual-mode vision system mounted on a mobile platform is used to acquire visible light and infrared images of mash piles in the brewing workshop; computer vision algorithms are used to process the visible light images to automatically recognize, locate and extract the outline of the mash piles, and automatically number the multiple mash piles that are recognized. S2. Non-contact multi-dimensional perception and modeling: The mobile platform is controlled to inspect the numbered mash piles according to a preset path or a self-planned trajectory; during the inspection, a high-resolution infrared thermal image of the surface of the mash pile is acquired simultaneously using an infrared thermal imager, and a three-dimensional geometric model of the mash pile is acquired simultaneously using a three-dimensional imaging device. S3. Temperature Field Inversion and Fermentation State Reasoning: Spatial registration and data fusion are performed between the infrared thermogram and the three-dimensional geometric model; using the fused data as input, and combining the heat generation mechanism of Maotai-flavor liquor fermentation, a heat conduction model including heat source terms is constructed. By solving the heat conduction model, the three-dimensional temperature field distribution inside the mash pile is inverted; based on the evolution characteristics of the three-dimensional temperature field distribution, at least one of the following indicators is extracted: temperature rise rate, high temperature zone ratio, and temperature gradient distribution, and compared with a preset "golden temperature curve" to evaluate the current fermentation state, and generate early warning information and process optimization suggestions when temperature anomalies are detected; S4. Digital Management and Visualization: The numbering, location information, three-dimensional geometric model, three-dimensional temperature field, fermentation status assessment results, early warning information and process optimization suggestions of the fermentation mash pile are integrated and visualized on the digital management platform, and the temperature change history is automatically generated to support historical data traceability and process analysis.

2. The digital method for temperature measurement in stacked fermentation based on infrared sensing and modeling reasoning according to claim 1, characterized in that, In step S1, the mobile platform is a gimbal (PTZ) or a track-mounted mobile platform; the computer vision algorithm includes traditional algorithms based on contour recognition and color analysis, or deep learning object detection and instance segmentation algorithms based on YOLO and Mask R-CNN.

3. The digital method for temperature measurement in stacked fermentation based on infrared sensing and modeling reasoning according to claim 2, characterized in that, Step S1 also includes: based on the identified location and outline of the mash pile, establishing a digital file for each mash pile containing its location, size and number, and autonomously planning an orderly inspection and measurement path for multiple mash piles based on this file.

4. The digital method for temperature measurement in stacked fermentation based on infrared sensing and modeling reasoning according to claim 2, characterized in that, In step S2, the three-dimensional imaging device is a lidar or a structured light camera; the three-dimensional geometric model includes spatial structural parameters such as the volume, height, and surface area of ​​the fermentation mash pile.

5. The digital method for temperature measurement in stacked fermentation based on infrared sensing and modeling reasoning according to claim 4, characterized in that, In step S3, the construction of the heat conduction model including the heat source term specifically includes: based on the microbial metabolic activities of the fermentation of soy sauce-flavored liquor, establishing a typical heat production rate curve that varies with time and space, and substituting the heat production rate curve as the heat source term into the three-dimensional unsteady heat conduction differential equation established based on Fourier's law and the law of conservation of energy.

6. The digital method for temperature measurement in stacked fermentation based on infrared sensing and modeling reasoning according to claim 5, characterized in that, The method for solving the heat conduction model is the finite element method or the finite volume method, and the surface temperature distribution of the stack provided by the infrared thermogram is used as the boundary condition for solving the heat conduction model.

7. The digital method for temperature measurement in stacked fermentation based on infrared sensing and modeling reasoning according to claim 6, characterized in that, In step S3, the assessment of the current fermentation status includes: determining whether the fermentation process is in the heating phase, high-temperature phase, or cooling phase; the process optimization suggestions include specific time points for operations such as turning the pile, mixing the starter culture, and covering, for example, "It is recommended to turn the pile after 6 hours".

8. A digital device for measuring temperature in stacked fermentation based on infrared sensing and modeling reasoning according to any one of claims 1-7, characterized in that, include: Infrared sensing module (100) is used for non-contact acquisition of infrared thermal images and three-dimensional geometric structure data of mash pile; The edge computing and preprocessing unit (200) is communicatively connected to the infrared sensing module (100) and is used to receive the infrared thermal image and three-dimensional geometric structure data, and to perform data cleaning, registration and coordinate transformation to generate thermal image blocks with three-dimensional coordinates. The model computing server (300) is connected to the edge computing and preprocessing unit (200) to receive the heat map with three-dimensional coordinates, and based on the heat conduction model and the preset heat generation mechanism, it inverts and constructs the three-dimensional temperature field inside the mash pile, and then extracts key process indicators, and combines the process knowledge base (400) to perform fermentation state reasoning and prediction. The process knowledge base (400) is communicatively connected to the model computing server (300) and is used to store the temperature window, historical reference data and process expert rules for the fermentation stage; The user interaction terminal (500) is communicatively connected to the model calculation server (300) and is used to receive and visualize the three-dimensional temperature field, key process indicators, fermentation state reasoning and prediction results, and early warning information.

Images