Method, device and equipment for predicting microbial growth curve in fermentation process and storage medium

Abstract

The application discloses a fermentation process microorganism growth curve prediction method and device, equipment and a storage medium. The scheme provided by the application determines a maximum concentration estimation range based on non-initial time point data, and solves Gompertz model parameters in combination with a candidate value method, effectively processes parameter identification problems under a sparse data condition, realizes accurate prediction of a growth curve, and has the advantages that the microorganism growth curve can be accurately predicted based on sparse data points, the problem of insufficient parameter identification of a traditional method under a data sparse condition is avoided, real-time prediction in an experimental process is supported, and experimental operation can be adjusted in time.

Description

Technical Field

[0001] This application relates to the fields of teaching informatization and image processing technology, and in particular to a method, apparatus, equipment and storage medium for predicting the microbial growth curve of a fermentation process. Background Technology

[0002] In junior high school biology fermentation experiments, dynamically monitoring the number of microorganisms and predicting their growth trends is a crucial step in cultivating students' quantitative analysis and scientific inquiry abilities. To help students understand the core principle of "dynamic changes in microbial numbers over time," it is usually necessary to plot growth curves based on time-series counting data collected during the experiment, and then analyze key phases such as the lag phase, logarithmic phase, and stationary phase based on these curves.

[0003] Current practices mainly rely on two types of technical means: one is to perform retrospective curve fitting using research-grade software (such as GraphPad Prism) after obtaining complete experimental data; the other is to use a large number of miniature sensors to continuously collect high-density time-series data in industrial or scientific research scenarios and then perform model fitting using conventional least squares methods. However, these solutions generally have the following problems: In teaching scenarios, due to time constraints and students' operational abilities, only 2 to 3 effective count data points can usually be obtained, resulting in extremely sparse data. Under these conditions, the traditional nonlinear least squares method has poor parameter identifiability, making it difficult for the model to converge or easily get trapped in local optima, and unable to stably fit a reliable growth curve. On the other hand, research-grade software is complex to operate and unsuitable for junior high school students. Furthermore, its post-experimental analysis mode prevents students from knowing in advance when the logarithmic phase will arrive during the experiment, which cannot guide orderly sampling and makes it difficult to achieve the inquiry-based learning goal of "experimenting, predicting, and verifying simultaneously". In addition, although commercial cell counters (such as the Countess series) can automatically identify microorganisms in the field of view and output the current concentration, they only output the value at the current time point, do not have the function of growth curve prediction, and are expensive, making them unsuitable for batch configuration in teaching scenarios. Summary of the Invention

[0004] This application provides a method, apparatus, equipment, and storage medium for predicting microbial growth curves during fermentation processes, which aims to achieve accurate prediction of microbial growth curves based on sparse data points and is applicable to junior high school biological fermentation experimental teaching scenarios.

[0005] To achieve the above-mentioned objectives, the first aspect of this application provides a method for predicting the microbial growth curve of a fermentation process, comprising: Microscopic images of the fermentation broth experimental samples were collected at preset sampling time points. Based on the microscopic images, image recognition and target object statistics are performed to obtain the number of microorganisms corresponding to different sampling time points, and the concentration data corresponding to different sampling time points is calculated based on the number of microorganisms. Based on the concentration data corresponding to two non-initial time points, and combined with prior knowledge of microbial growth, the estimated range of the maximum concentration is determined. Based on the estimated range, and according to the preset step size parameters, multiple candidate values ​​for maximum concentration are determined; Based on each candidate maximum concentration value, the deviation between the candidate maximum concentration value and the concentration data at the initial time point is calculated. Then, based on the deviation value, the maximum specific growth rate and the duration of the lag phase are obtained by solving the inverse function of the Gompertz function model. For each candidate value of maximum concentration, the candidate value of maximum concentration and the corresponding maximum specific growth rate and lag phase duration are divided into a candidate parameter group. Combined with the Gompertz function model, the prediction error of each candidate parameter group is calculated, and the final growth curve prediction parameters are determined according to the candidate parameter group with the smallest prediction error. Based on the growth curve prediction parameters, the key time points of the growth period of the fermentation broth experimental sample are calculated, so as to generate a growth prediction curve of the fermentation process based on the key growth period nodes.

[0006] Preferably, after periodically acquiring microscopic images of the fermentation broth experimental sample, the method further includes: If the carrier of the fermentation broth experimental sample is a hemocytometer, then the grid lines in the microscopic image are identified by image recognition and Hough transform processing. Based on the grid lines, several counting regions are divided in the microscopic image.

[0007] Preferably, after identifying the grid lines in the microscopic image, the process further includes: The mask corresponding to the grid lines is removed from the original microscopic image, and then the remaining area is segmented into microbial images.

[0008] Preferably, after periodically acquiring microscopic images of the fermentation broth experimental sample, and before identifying the grid lines in the microscopic images using image recognition and Hough transform processing, the method further includes: The microscopic image is preprocessed, including median filtering for noise removal and adaptive histogram equalization for enhancing the contrast between grid lines and microorganisms.

[0009] Preferably, the formula for the Gompertz function model is:

[0010] In the formula, Let be the microbial concentration at time t. The initial microbial concentration. For maximum concentration, For the maximum specific growth rate, The duration of the lag period is given by ε, and e is a natural constant.

[0011] Preferably, the formula for calculating the key time points of the growth period is:

[0012] In the formula, Represents the starting point of the logarithmic period. Represents the moment of maximum growth rate. Represents the end of the logarithmic period. The initial microbial concentration. For maximum concentration, For the maximum specific growth rate, The duration of the lag period is given by ε, and e is a natural constant.

[0013] Preferably, the sampling time points include: an initial time point and two non-initial time points, wherein the non-initial time points are 6 hours and 12 hours from the initial time point, respectively.

[0014] A second aspect of this application provides a device for predicting the microbial growth curve of a fermentation process, comprising: The image acquisition unit is used to acquire microscopic images of the fermentation broth experimental samples at preset sampling time points. The microbial concentration conversion unit is used to perform image recognition and target object statistics based on the microscopic image to obtain the number of microorganisms corresponding to different sampling time points, and to convert the concentration data corresponding to different sampling time points based on the number of microorganisms. The maximum concentration range prediction unit is used to determine the predicted range of the maximum concentration based on the concentration data corresponding to two non-initial time points and combined with prior knowledge of microbial growth. The maximum concentration candidate value determination unit is used to determine multiple maximum concentration candidate values ​​according to the estimated range and a preset step size parameter. The inverse function solving unit is used to calculate the deviation between the candidate maximum concentration and the concentration data at the initial time point based on each candidate maximum concentration, and then use the deviation to solve the inverse function of the Gompertz function model to obtain the maximum specific growth rate and the duration of the lag phase. The growth curve prediction parameter determination unit is used to divide the maximum concentration candidate value, the corresponding maximum specific growth rate and lag phase duration into a candidate parameter group for each maximum concentration candidate value, calculate the prediction error of each candidate parameter group in combination with the Gompertz function model, and determine the final growth curve prediction parameters based on the candidate parameter group with the smallest prediction error. The growth prediction curve generation unit is used to calculate the key time points of the growth period of the fermentation broth experimental sample based on the growth curve prediction parameters, so as to generate a growth prediction curve of the fermentation process based on the key time points of the growth period.

[0015] A third aspect of this application provides an intelligent microscope device, including: a memory and a processor; The memory is used to store program code, which corresponds to the method for predicting microbial growth curves in fermentation processes as provided in the first aspect of this application. The processor is used to read and execute the program code to implement the method for predicting the microbial growth curve of the fermentation process. The computer-readable storage medium stores program code, which is read and executed by the processor to implement the method for predicting the microbial growth curve of the fermentation process as provided in the first aspect of this application.

[0016] As can be seen from the above technical solutions, this application has the following advantages: The solution provided in this application determines the maximum concentration prediction range based on non-initial time point data and solves the Gompertz model parameters by combining the candidate value method. It effectively handles the parameter identification problem under sparse data conditions and achieves accurate prediction of growth curves. It has the advantages of being able to accurately predict microbial growth curves based on sparse data points, avoiding the problem of insufficient parameter identification under sparse data conditions in traditional methods, supporting real-time prediction during the experiment, and facilitating timely adjustment of experimental operations. Attached Figure Description

[0017] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 This is a schematic flowchart illustrating an embodiment of a method for predicting microbial growth curves during fermentation provided in this application.

[0019] Figure 2 This is a schematic diagram of the architecture of an embodiment of a microbial growth curve prediction device for a fermentation process provided in this application. Detailed Implementation

[0020] In traditional junior high school biology fermentation experiments, when dynamically monitoring the number of microorganisms and predicting their growth trends, only sparse time-series count data points can be obtained due to time constraints and students' operational abilities. Under these sparse data conditions, the traditional nonlinear least squares method lacks sufficient parameter identifiability, making it difficult for the model to converge or easily get trapped in local optima, and unable to generate stable growth curve prediction results. At the same time, research-grade software tools have complex operation processes and lack real-time interactive capabilities. Their post-analysis mode prevents students from dynamically predicting growth trends during the experiment, which cannot guide the selection of subsequent sampling times or support the inquiry-based learning process of experimenting, predicting, and verifying simultaneously.

[0021] For example, in a yeast fermentation experiment in a junior high school biology class, microscopic images were collected at initial time points, 6 hours, and 12 hours. The number of microorganisms was counted manually and converted into concentration data. When conventional curve fitting methods were applied, the parameter solution process was highly unstable due to insufficient effective data points. Different initial parameter settings led to significant differences in the shape of the predicted curve, and the start time of the logarithmic phase could not be accurately identified. As a result, students could not adjust the pace of the experiment based on the prediction results, the observation window of the critical growth period was missed, and the process of understanding the dynamic changes in the number of microorganisms was hindered.

[0022] In view of this, this application provides a method, apparatus, device, and storage medium for predicting microbial growth curves during fermentation processes, which is used to accurately predict microbial growth curves based on sparse data points and is applicable to junior high school biology fermentation experimental teaching scenarios.

[0023] To make the inventive objectives, features, and advantages of this application more apparent and understandable, the technical solutions in the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the embodiments described below are only some embodiments of this application, and not all embodiments. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0024] First, a detailed description of an embodiment of the method for predicting microbial growth curves in a fermentation process provided in this application is as follows: Please see Figure 1 An embodiment of a method for predicting microbial growth curves in a fermentation process provided in this application includes the following steps: Step 101: Collect microscopic images of the fermentation broth experimental samples at preset sampling time points; Step 102: Perform image recognition and target object statistics based on the microscopic images to obtain the number of microorganisms corresponding to different sampling time points, and calculate the concentration data corresponding to different sampling time points based on the number of microorganisms. Step 103: Based on the concentration data corresponding to the two non-initial time points and combined with prior knowledge of microbial growth, determine the estimated range of the maximum concentration. Step 104: Based on the estimated range, determine multiple candidate values ​​for maximum concentration according to the preset step size parameters; Step 105: Based on each candidate maximum concentration value, calculate the deviation between the candidate maximum concentration value and the concentration data at the initial time point, and then, based on the deviation value, solve the maximum specific growth rate and the duration of the lag phase by combining the inverse function of the Gompertz function model. Step 106: For each candidate value of maximum concentration, the candidate value of maximum concentration and the corresponding maximum specific growth rate and lag phase duration are divided into a candidate parameter group. Combined with the Gompertz function model, the prediction error of each candidate parameter group is calculated, and the final growth curve prediction parameters are determined according to the candidate parameter group with the smallest prediction error. Step 107: Calculate the key time points of the growth period of the fermentation broth experimental sample based on the growth curve prediction parameters, so as to generate a growth prediction curve of the fermentation process based on the key growth points.

[0025] For ease of understanding, the following explains some key terms in this embodiment: The method for predicting microbial growth curves during fermentation refers to a technical solution that predicts microbial growth trends and key growth stages by modeling and analyzing the changes in microbial numbers over time during fermentation. This method aims to provide a means to stably and accurately predict microbial growth curves even under data sparsity conditions.

[0026] The sampling time point refers to the time node that is pre-set during the fermentation experiment for collecting experimental samples of the fermentation broth.

[0027] Microscopic images are images obtained by observing and photographing experimental samples of fermentation broth using a microscope. These images contain visual information such as the morphology, quantity, and distribution of microorganisms, and are the basic data for microbial counting and concentration conversion.

[0028] Prior knowledge of microbial growth refers to known theories and experiences regarding the laws governing microbial growth in the field of microbiology. For example, microbial growth typically follows an S-shaped curve, including phases such as lag phase, logarithmic phase, and stationary phase, and there exists a maximum growth concentration. This knowledge is used to guide the prediction and range limitation of model parameters.

[0029] The Gompertz function model is a nonlinear mathematical model commonly used to describe biological growth processes. This model can fit the S-shaped growth curve of microorganisms well, and its parameters can characterize key growth characteristics such as initial concentration, maximum concentration, maximum specific growth rate, and duration of lag phase.

[0030] Maximum specific growth rate refers to the maximum relative rate at which the number or concentration of microorganisms increases per unit time during microbial growth. This parameter reflects the growth potential of microorganisms under optimal growth conditions.

[0031] The duration of the lag phase refers to the period after microorganisms are inoculated into a new culture medium, during which they adapt to the new environment and begin to divide and proliferate. During this phase, the number of microorganisms typically does not increase significantly.

[0032] Critical growth points refer to specific time points in the microbial growth curve that have significant biological importance, such as the start of the logarithmic growth phase, the moment of maximum growth rate, and the end of the logarithmic growth phase. These time points are instructive for understanding and controlling the fermentation process.

[0033] More specifically, this embodiment provides a method for predicting the microbial growth curve of a fermentation process, the main technical features of which are described in detail below: First, microscopic images of the fermentation broth samples are collected at preset sampling time points. In one implementation, researchers can manually extract small amounts of fermentation broth from the fermenter at preset time points, such as every few hours, prepare them as microscope slides, and then photograph them using an optical microscope equipped with a digital camera to obtain microscopic images of the microorganisms. Another implementation method involves placing the fermentation broth samples in a simple observation device that triggers a camera at a fixed position to capture images at preset time points.

[0034] Secondly, image recognition and target object statistics are performed based on the microscopic image to obtain the number of microorganisms at different sampling time points, and the concentration data corresponding to different sampling time points is calculated based on the number of microorganisms. Specifically, manual counting can be used, where the observer directly identifies and marks individual microorganisms on the microscopic image and then counts them. Alternatively, basic image processing software can be used to preliminarily identify and count pixel clusters in the image by setting a fixed brightness threshold and size range, thereby obtaining the number of microorganisms. Subsequently, the obtained number of microorganisms is divided by the known volume of the microscope's field of view to estimate the concentration data of the microorganisms.

[0035] Furthermore, based on the concentration data corresponding to non-initial time points and combined with prior knowledge of microbial growth, the estimated range of the maximum concentration can be determined. For example, based on known culture medium types and microbial species, and referring to historical experimental data or biological manuals, a reasonable range for the maximum microbial concentration can be artificially set, such as from a certain minimum concentration value to a certain maximum concentration value. Alternatively, a possible range for the maximum concentration can be initially defined based on observed concentration data from non-initial time points.

[0036] Based on this, and according to the estimated range and a preset step size parameter, multiple candidate values ​​for the maximum concentration are determined. Specifically, once the estimated range for the maximum concentration is determined, for example, [C_min, C_max], the system can start from C_min and increment by a fixed amount (i.e., the step size parameter), for example, every 10 increments. 6 1 cell / mL, generating a series of discrete concentration values ​​until C_max is reached or exceeded. These discrete values ​​constitute a candidate set for the maximum concentration.

[0037] Subsequently, based on each candidate maximum concentration, the deviation between that candidate value and the concentration data at the initial time point is calculated. Then, based on this deviation, the maximum specific growth rate and the duration of the lag phase are obtained by solving the inverse Gompertz function model. For each candidate maximum concentration, and given the known initial time point concentration and at least one non-initial time point concentration, the maximum specific growth rate and the duration of the lag phase at that specific maximum concentration can be derived backwards using the mathematical properties of the Gompertz function model through algebraic transformations or numerical iteration methods. Essentially, this process involves finding growth kinetic parameters that match the current candidate maximum concentration.

[0038] For each candidate maximum concentration value, the candidate maximum concentration value, along with the corresponding maximum specific growth rate and lag phase duration, are grouped into a candidate parameter group. Using the Gompertz function model, the prediction error of each candidate parameter group is calculated, and the final growth curve prediction parameters are determined based on the candidate parameter group with the smallest prediction error. Specifically, each set consisting of the candidate maximum concentration value, maximum specific growth rate, and lag phase duration is considered a potential growth curve parameter combination. Using these parameters, the Gompertz function model can predict the microbial concentration at all sampling time points. By comparing the differences between these predicted concentrations and the actual observed concentrations, such as by calculating the mean squared error or sum of squared errors, the prediction accuracy of each candidate parameter group can be quantified. Finally, the parameter group with the smallest error is selected as the growth curve prediction parameter that best represents the actual growth of the microorganisms, thus achieving the fitting operation of the growth curve prediction parameters under extremely sparse conditions with only 2-3 data points.

[0039] More specifically, the implementation process of the above-mentioned few-sample-specific fitting method can be seen in the following example: First, input the measured data; generally, only data from three time points are needed: ), , in =0, Obtained directly from image counts at the initial time point (0h); Next, determine the maximum concentration. The search space: based on prior knowledge of microbial growth, It must be greater than the measured maximum value, and cannot be infinitely large. The search range is:

[0040] Preferred, The reference value is generally 1.1~1.2. The reference value is generally 2.9~3.0.

[0041] Then, based on this search range, all candidate values ​​are enumerated with a fixed step size. Then, for each candidate... calculate Using the first non-zero data point and the second non-zero data point The solution is obtained analytically using the inverse function of the Gompertz model. and No iterative optimization is required; Next, by calculating the prediction error for the known data points, the group with the smallest error is selected as the final fitting result. In addition, before calculating the prediction error, selective constraint filtering can be performed to eliminate candidate parameter groups that violate conventional biological constraints. Then, for the remaining candidate groups, the prediction error for the known data points is calculated. Alternatively, if none of the candidate groups meet the constraints, the user can be prompted that "the current data points are insufficient or there are anomalies. It is recommended to increase the sampling points."

[0042] Finally, based on the growth curve prediction parameters, the key time points of the growth period for the fermentation broth experimental sample are calculated to generate a growth prediction curve for the fermentation process based on these key growth points. Once the optimal growth curve prediction parameters are determined, these parameters can be substituted into preset mathematical formulas or used through numerical analysis methods to calculate important inflection points in the microbial growth curve, such as the start time of the logarithmic phase, the time of maximum growth rate, and the end time of the logarithmic phase. Based on these key time points and the complete Gompertz model parameters, a smooth and predictive microbial growth curve can be plotted.

[0043] This embodiment determines the estimated range of the maximum concentration by introducing prior knowledge of microbial growth, and generates multiple candidate values ​​for the maximum concentration within this range. Then, by combining the inverse function of the Gompertz function model, the corresponding maximum specific growth rate and lag phase duration are determined for each candidate value. This method effectively transforms the multi-parameter fitting problem into an iterative search and verification of a single parameter (maximum concentration). Unlike traditional methods that rely on continuous high-density data acquisition from numerous sensors, this method is specifically designed for sparse data conditions with only 2-3 effective data points. By introducing prior knowledge of microbial growth to constrain the range of the maximum concentration, the multi-parameter fitting problem is reduced to a single-parameter search problem, thus enabling stable solution of the Gompertz model parameters even under extremely sparse data conditions. By calculating the prediction error of each candidate parameter group and selecting the group with the smallest error, reliable growth curve prediction parameters can be stably extracted from limited data.

[0044] Furthermore, the method in this embodiment can generate growth prediction curves and key time points of the growth period in real time during the middle of the experiment (e.g., after 12 hours of data collection), such as predicting the start of the logarithmic phase at 3 hours and the moment of maximum growth rate at 8 hours. This real-time prediction capability allows experimenters to adjust subsequent experimental plans based on the prediction results, such as increasing the sampling frequency before the arrival of the logarithmic phase, thereby greatly improving the guidance and interactivity of the experiment and achieving teaching and research goals that are difficult to achieve with existing technologies.

[0045] In summary, the method of this embodiment can stably and accurately predict microbial growth curves under data sparsity conditions and provide real-time information on key growth stages, effectively overcoming the challenges faced by existing technologies in teaching scenarios and having better practical value.

[0046] In some of the above embodiments, this application proposes a method for predicting the microbial growth curve of a fermentation process. This method involves periodically acquiring microscopic images of fermentation broth samples, performing image recognition and target object statistics to obtain the number of microorganisms, then calculating concentration data, and finally using a Gompertz function model to predict the growth curve. Furthermore, this application proposes that after periodically acquiring microscopic images of the fermentation broth samples, if the sample carrier is a hemocytometer, the grid lines in the microscopic images are identified using image recognition and Hough transform processing; based on these grid lines, several counting regions are divided within the microscopic images.

[0047] Among them, a hemocytometer is a glass slide with a precisely graduated grid, commonly used for counting microorganisms or cells. The grid lines on its surface are stable geometric features that can serve as references in image processing. Image recognition refers to methods that use computer vision technology to analyze and understand image content to extract specific information. This can include, but is not limited to, various techniques based on feature extraction, pattern matching, machine learning, or deep learning. For example, edge detection algorithms (such as the Canny operator) can be used to highlight line features in an image. Hough transform is a widely used feature extraction technique in image processing, particularly adept at detecting geometric shapes such as straight lines and circles in images. For identifying grid lines in microscopic images, Hough transform can effectively map points in image space to parameter space, determining the parameters of the line through the peak values ​​in the parameter space, thereby accurately locating the grid lines. This can be achieved through a standard Hough transform, mapping each edge point in the image to a series of curves in parameter space, and determining the line through the intersection of the curves; or by using a probabilistic Hough transform, randomly sampling edge points to improve computational efficiency and robustness.

[0048] Dividing the counting region based on the grid lines refers to defining specific sub-regions for microbial counting in a microscopic image after accurately identifying the grid lines, based on the geometric information of the grid lines (such as intersections, line segment lengths, and directions). These regions are typically standard square or rectangular units enclosed by the grid lines. The division can be achieved by calculating the coordinates of all intersections of the grid lines and then constructing rectangular regions using these intersections as vertices; or by performing morphological operations (such as closure operations) on the identified grid lines to generate closed counting units, facilitating automatic alignment of the same region during multi-time-point sampling and ensuring spatial traceability of counting data from different time points.

[0049] Automatic alignment of the same region refers to the system's ability to automatically match and calibrate the counting region in the current image with the corresponding physical region in images from previous time points, based on microscopic images acquired at different time points. This can be achieved by recording the relative position of each counting region in the image or through image registration techniques. For example, coordinate transformations between images can be performed based on the geometric features of grid lines (such as relative position, angle, and spacing), thereby aligning images from different time points to a unified coordinate system and ensuring that subsequent counting is performed within the same physical region. Spatial traceability means that the specific physical region on the hemocytometer from which the microbial data collected at different time points originated can be clearly traced. This is crucial for verifying the accuracy of experimental results, analyzing the growth dynamics of microorganisms within a specific region, and eliminating errors caused by inconsistent sampling areas.

[0050] This application addresses the issue of inconsistent counting regions in multi-timepoint sampling using the aforementioned methods for predicting microbial growth curves during fermentation processes by introducing the identification and utilization of grid lines on a hemocytometer. Specifically, after acquiring the microscopic image, image recognition and Hough transform processing are used to accurately identify the grid lines in the image. These grid lines are inherent and stable structural features of the hemocytometer. Once the grid lines are identified, the system can accurately delineate several standard, physically meaningful counting regions in the microscopic image based on these grid lines. In subsequent multi-timepoint sampling, even if the microscope's field of view shifts slightly, the system can automatically align the counting regions in the current image with the corresponding regions in previous timepoint images by re-identifying the grid lines and using them as a reference. This automatic alignment mechanism based on grid lines ensures that the number of microorganisms counted at different time points always originates from the same physical region on the hemocytometer, thus making the counting data at different time points spatially traceable. In this way, the accuracy and consistency of microbial quantity statistics are significantly improved, providing more reliable raw data for subsequent concentration data conversion, Gompertz function model parameter solving, and growth curve prediction, thereby improving the accuracy of the entire prediction method.

[0051] In some of the above embodiments, this application proposes that if the carrier of the fermentation broth experimental sample is a hemocytometer, then the grid lines in the microscopic image are identified through image recognition and Hough transform processing; based on the grid lines, several counting regions are divided in the microscopic image to facilitate automatic alignment of the same region during multi-time-point sampling. Furthermore, this application proposes that after identifying the grid lines in the microscopic image, the process includes: subtracting the region mask corresponding to the grid lines from the original microscopic image, and then performing microbial image segmentation on the remaining region.

[0052] One technique involves subtracting the region corresponding to the grid lines from the original microscopic image. This technique aims to isolate or remove the identified grid line regions in the microscopic image through image processing. Its purpose is to eliminate the interference of grid lines on subsequent microbial identification. Specifically, after identifying the grid line positions, a binary mask image corresponding to the grid line regions can be generated. Then, the pixel values ​​in the original microscopic image corresponding to the grid line regions in the mask image can be modified, for example, by setting them to the background color, zero, or performing interpolation, thereby achieving the "subtraction" of the grid line regions. Another implementation method is to directly mark the grid line regions in the image data structure and skip these regions in subsequent processing. Microbial image segmentation is then performed on the remaining regions. This technique refers to identifying and separating microorganisms only in areas of the image that do not contain grid lines after the grid line regions have been subtracted. Its purpose is to accurately distinguish individual microorganisms in the image from the background, providing a basis for subsequent counting. Specifically, threshold-based segmentation methods, such as global thresholding or local adaptive thresholding, can be used to separate microbial pixels from background pixels. In addition, edge detection-based segmentation methods, such as the Canny or Sobel operators, can be used to identify the contours of microorganisms. Furthermore, machine learning or deep learning models, such as U-Net or Mask R-CNN, can be used for pixel-level precise segmentation of microorganisms. To avoid interference from grid lines in the counting targets, this technique clarifies the purpose and effect of the above operations. Grid lines, as prominent features in an image, may have pixel values ​​and shapes similar to microorganisms. If not processed, they can easily be misidentified as microorganisms during image segmentation, or lead to inaccurate boundary identification of microorganisms, thus affecting the accuracy of microorganism counts. By removing the grid line region, it can be ensured that subsequent microorganism image segmentation is performed only in clean background areas, effectively avoiding this interference.

[0053] The proposed solution involves first masking the regions corresponding to grid lines in the original microscopic image after identifying them. This step ensures that only microorganisms and background information are retained in the image, eliminating grid lines as a potential source of interference. Subsequently, microbial image segmentation is performed on the remaining image regions after grid line removal. In this way, the microbial segmentation algorithm can focus on identifying and separating real microbial individuals without being misled by grid line edges or textures. This processing flow makes microbial identification and counting more accurate and reliable, thus providing high-quality input data for subsequent conversion of concentration data based on microbial quantity, determination of the maximum concentration prediction range, solving the Gompertz function model parameters, and finally generating fermentation process growth prediction curves. This solution effectively improves the accuracy and robustness of the entire prediction method through refined image preprocessing.

[0054] In some embodiments described above in this application, microscopic images of fermentation broth experimental samples are acquired at regular intervals, and grid lines in the microscopic images are identified to divide the samples into several counting regions, thereby enabling microbial quantity statistics. Building upon this, this application further proposes that after acquiring the microscopic images of the fermentation broth experimental samples at regular intervals, and before identifying the grid lines in the microscopic images using image recognition and Hough transform processing, preprocessing of the microscopic images is included. This preprocessing includes: median filtering for noise removal and adaptive histogram equalization for enhancing the contrast between grid lines and microorganisms.

[0055] Preprocessing refers to a series of operations performed on microscopic images before subsequent analysis (such as grid line identification and microbial segmentation) to improve image quality, remove interference, or enhance specific features. Its purpose is to improve the accuracy and robustness of subsequent processing. Besides median filtering and adaptive histogram equalization, preprocessing can also include gamma correction, Gaussian blurring, and bilateral filtering. Median filtering is a non-linear digital filtering technique commonly used to remove salt-and-pepper noise or speckle noise from images while preserving edge information as much as possible. Its principle is to replace the value of a pixel with the median value of its neighborhood. In addition to median filtering, Gaussian filtering, mean filtering, and bilateral filtering can also be used to remove image noise. Adaptive histogram equalization is an image enhancement technique that enhances local contrast by equalizing the histogram of local image regions, especially suitable for images with uneven overall brightness distribution or low contrast in local areas. Its purpose is to make grid lines and microbial details clearer, facilitating subsequent identification and segmentation. In addition to adaptive histogram equalization, other methods such as global histogram equalization and contrast-limited adaptive histogram equalization (CLAHE) can be used to enhance image contrast.

[0056] This application's solution introduces a preprocessing step for the microscopic images after periodically acquiring them, and before identifying grid lines in the images using image recognition and Hough transform. This preprocessing first involves median filtering to remove noise, effectively filtering out random noise caused by limitations of the acquisition environment or equipment, resulting in cleaner images. Subsequently, adaptive histogram equalization is performed to enhance the contrast between grid lines and microorganisms. This process optimizes the brightness distribution in local image regions, significantly improving the contrast between grid lines and background, and between microorganisms and background, making previously blurred grid lines and microbial outlines clearer. Through the synergistic effect of these two preprocessing steps, the quality of the original microscopic images is significantly improved, laying a solid foundation for accurate grid line identification. Clear grid line identification ensures accurate division of counting regions, guaranteeing the accuracy of microbial quantity statistics at different sampling time points, thereby improving the reliability of concentration data and ultimately making the growth curves predicted based on the Gompertz function model more accurate.

[0057] The above technical solution introduces median filtering and adaptive histogram equalization before performing grid line recognition on the microscopic images of fermentation broth experimental samples. Median filtering effectively removes noise from the image, preventing it from interfering with subsequent image analysis. Adaptive histogram equalization significantly enhances the contrast between grid lines and microorganisms in the microscopic image, making their features more prominent. The combination of these two preprocessing methods greatly improves the quality of the microscopic image, thereby significantly improving the accuracy of grid line recognition through image recognition and Hough transform. The accuracy of grid line recognition directly affects the precision of counting region division, thus ensuring the reliability of microbial quantity statistics and concentration data conversion. Ultimately, it provides high-quality input data for accurate prediction of microbial growth curves during fermentation, effectively solving the problem of decreased prediction accuracy due to poor original image quality.

[0058] In some embodiments described above in this application, a method for predicting microbial growth curves during fermentation is proposed. This method involves acquiring microscopic images, counting microbial numbers, converting concentration data, and combining this with a Gompertz function model to solve for parameters and calculate prediction errors, ultimately generating a growth prediction curve. Based on this, this application further proposes the following formula for the Gompertz function model:

[0059] In the formula, Let be the microbial concentration at time t. The initial microbial concentration. For maximum concentration, For the maximum specific growth rate, The duration of the lag period is given by ε, and e is a natural constant.

[0060] It should be noted that the Gompertz function model is a widely used mathematical model for describing biological growth processes, especially suitable for fitting the growth curves of microorganisms under limited resource conditions. It can effectively reflect the lag phase, logarithmic growth phase, and stationary phase in the microbial growth process, characterized by the asymmetric change in growth rate over time. This model is typically used to predict the trend of microbial quantity or biomass changes over time. The mathematical formula of the Gompertz function model is explicitly given to provide a precise and unambiguous mathematical foundation for subsequent calculations. This formula defines the quantitative relationship between the various parameters in the model, ensuring the model's consistency and operability across different application scenarios. The microbial concentration at time t refers to the actual or predicted concentration of microorganisms in the fermentation broth experimental sample at any given time point t. This concentration is the core output variable described and predicted by the Gompertz function model, reflecting the growth state of microorganisms at a specific time point. The initial microbial concentration at the initial time point refers to the concentration of microorganisms in the fermentation broth experimental sample at the start of the fermentation process, i.e., at time point t=0. This parameter is the starting point of the microbial growth curve, providing the initial conditions for the model. Maximum concentration refers to the highest concentration of microorganisms that can be achieved during the entire fermentation process. This parameter reflects the upper limit of the number of microorganisms that the fermentation system can support, and is usually limited by factors such as nutrient depletion or accumulation of metabolites. Maximum specific growth rate refers to the instantaneous rate at which the concentration of microorganisms increases most rapidly during the logarithmic growth phase. This parameter is an important indicator of microbial growth viability, reflecting the reproductive capacity of microorganisms under optimal growth conditions. Lag phase duration refers to the time required for microorganisms to enter the logarithmic growth phase from inoculation. During this period, microorganisms typically adapt to the new environment, preparing for subsequent rapid growth. This parameter reflects the adaptability of microorganisms to environmental changes. e is an irrational number, approximately equal to 2.71828, and is an important constant in mathematics, widely used in exponential and logarithmic functions. In the Gompertz function model, e is used as the base to construct the exponential term to describe the nonlinear change in microbial concentration.

[0061] This application provides a solid mathematical foundation for predicting microbial growth curves in fermentation processes by clarifying the specific formula of the Gompertz function model. The method requires solving the inverse function of the Gompertz function model, calculating prediction errors, and ultimately predicting the growth curve based on parameters such as the candidate maximum concentration, initial time point concentration data, maximum specific growth rate, and lag phase duration. By clarifying the mathematical expression of the model, it ensures that all these calculation steps are based on a unified and accurate mathematical framework. Specifically, once the key parameters such as the microbial concentration at the initial time point, maximum concentration, maximum specific growth rate, and lag phase duration are determined, the microbial concentration at any time t can be calculated using this explicit formula. This allows for the accurate calculation of the prediction error for each candidate parameter group after dividing the candidate maximum concentration, along with its corresponding maximum specific growth rate and lag phase duration, into a candidate parameter group using the Gompertz function model. This clarity eliminates ambiguity in model application, making the parameter solution process more standardized and the prediction results more reliable.

[0062] Accordingly, based on the above Gompertz function model expression, the formula for solving can be derived. and The expression for the inverse function is:

[0063] In some embodiments described above in this application, a method for predicting microbial growth curves during fermentation is proposed, which generates a predicted growth curve for the fermentation process by determining growth curve prediction parameters. Based on this, this application further proposes a calculation formula for the key time points of the growth period, which includes specific mathematical expressions for the start of the logarithmic phase, the moment of maximum growth rate, and the end of the logarithmic phase:

[0064] In the formula, Represents the starting point of the logarithmic period. Represents the moment of maximum growth rate. Represents the end of the logarithmic period. The initial microbial concentration. For maximum concentration, For the maximum specific growth rate, The duration of the lag period is given by ε, and e is a natural constant.

[0065] It should be noted that the calculation formulas for the key time points in the growth period are mathematical models used to transform the predicted parameters of the microbial growth curve into time points with clear biological significance. These formulas define the critical time point (logarithmic phase start) for the microorganism to transition from the slow adaptation phase to the rapid proliferation logarithmic phase, the moment when the microbial population growth rate reaches its peak (maximum growth rate moment), and the critical time point (logarithmic phase end) for the microorganism to transition from the rapid proliferation logarithmic phase to the slowed growth stationary phase. These formulas ensure that the determination of key time points is mathematically rigorous and repeatable. One implementation method is to directly substitute the determined Gompertz model parameters into the preset mathematical expressions for numerical calculation to obtain the specific values ​​of each key time point. Another implementation method is to construct a calculation module that can receive the Gompertz function and its derivative, and based on these functional relationships, determine the time points that meet specific biological conditions (e.g., the growth rate reaches its maximum, the inflection point of the growth curve, etc.) through numerical or analytical solutions. These time points are consistent with the meaning expressed by the provided formulas. The initial microbial concentration, maximum concentration, maximum specific growth rate, and duration of the lag phase are the core parameters for predicting microbial growth curves. These parameters are obtained by fitting a Gompertz function model and are substituted as input variables into the calculation formula for key time points in the growth period to accurately determine the start and end times of each growth stage.

[0066] By providing explicit formulas for calculating key time points in the growth cycle, this application ensures that the determination of key time points such as the start of the logarithmic phase, the moment of maximum growth rate, and the end of the logarithmic phase is standardized, objective, and highly accurate. This avoids errors caused by subjective judgment or lack of experience, enabling the growth prediction curve of the fermentation process to more accurately reflect the actual growth dynamics of microorganisms, thereby providing more reliable and refined data support for the monitoring, control, and optimization of the fermentation process.

[0067] In some embodiments, this application further proposes that the sampling time points include: an initial time point and two non-initial time points, wherein the non-initial time points are 6 hours and 12 hours from the initial time point, respectively.

[0068] The sampling time point refers to the specific moment during fermentation when microscopic images of the fermentation broth are collected. The selection of these time points directly affects the accuracy of subsequent microbial quantity statistics, concentration data conversion, and Gompertz function model parameter solving. Appropriate sampling time points can capture key stages of microbial growth, providing effective data support for the model. For example, sampling time points can be preset based on the expected growth cycle of the microorganisms or adjusted based on historical data and experience. The initial time point refers to the first sampling moment at the start of the fermentation process, usually marked as time zero (0h). The sample data collected at this time point represents the initial state of microbial growth and is an important benchmark for subsequent calculations of microbial concentration changes, determination of candidate maximum concentration values, and Gompertz function model parameters. Data from the initial time point is typically used to determine the initial microbial concentration. Non-initial time points refer to other sampling moments besides the initial time point. Data from these time points is used to capture the dynamic changes of microorganisms during growth, especially information on the logarithmic growth phase and stationary phase, thus providing crucial data for predicting the maximum concentration, maximum specific growth rate, and duration of the lag phase. Non-initial time points can be flexibly set according to experimental needs, microbial species, or fermentation process characteristics. For example, in addition to 6h and 12h mentioned in this application, they can also be 4h, 8h, 10h, 16h, etc. The 6h time point refers to the sampling time 6 hours after the initial time point. Selecting this time point for sampling aims to capture the state of microorganisms transitioning from the lag phase to the logarithmic growth phase or having entered the early stage of the logarithmic growth phase. At this stage, the number of microorganisms usually begins to increase significantly, providing important early growth data for solving subsequent model parameters. The 12h time point refers to the sampling time 12 hours after the initial time point. Selecting this time point for sampling aims to capture the state of microorganisms in the mid-to-late logarithmic growth phase or about to enter the stationary phase. At this stage, the number of microorganisms usually reaches a high level, providing more sufficient data support for accurately estimating the maximum concentration and maximum specific growth rate.

[0069] This application's approach limits sampling to three time points: the initial time point, the 6-hour time point, and the 12-hour time point. This allows for the acquisition of representative microbial growth data in the early stages of fermentation. Specifically, the data at the initial time point provides baseline information on microbial growth, namely the initial microbial concentration. Sampling at the 6-hour time point effectively captures crucial information about the transition from the lag phase to the logarithmic growth phase, at which point the microbial population begins to grow rapidly. This provides early dynamic data for estimating the duration of the lag phase and the maximum specific growth rate in the subsequent Gompertz function model. Next, sampling is performed again at the 12-hour time point. At this point, the microorganisms have typically entered the rapid growth phase or are nearing the stationary phase. The concentration data at this time point is crucial for accurately predicting the maximum concentration range and further refining the calculation of the maximum specific growth rate. By using data from these three specific time points, the main stages of microbial growth can be effectively covered, avoiding the waste of resources from excessive sampling while ensuring sufficient data for solving the model parameters. This strategy enables the Gompertz function model to more accurately fit the microbial growth curve based on a limited but crucial set of data points, thereby improving the reliability of predictions.

[0070] Following the step logic of the fermentation process microbial growth curve prediction method mentioned in the above embodiments, this embodiment further illustrates the above technical solution in more detail through a more specific example: Suppose that in a microbial fermentation experiment, user A needs to predict the growth curve of a certain yeast. Due to experimental limitations, user A can only collect fermentation broth samples at three preset sampling time points: the initial time point (0 hours), 6 hours, and 12 hours.

[0071] First, at 0, 6, and 12 hours, User A observed the fermentation broth sample using a microscope and took microscopic images. These images were then input into the processing system.

[0072] Subsequently, the system performs image recognition and target object statistics on these microscopic images. For example, using a simplified image processing algorithm, yeast cells in the images are identified and counted, thus obtaining the number of microorganisms at 0 hours, 6 hours, and 12 hours. Assuming these numbers are converted to concentration data, with an initial concentration of 1.2 × 10⁻⁶... 6 Cells / mL, concentration at 6 hours: 2.8 x 10⁻⁶ 6 Cells / mL, concentration at 12 hours: 8.6 x 10⁻⁶ 6 Cells / mL

[0073] Next, based on the concentration data from non-initial time points at 6 and 12 hours, and combined with prior knowledge of yeast growth under the current culture conditions, the system determines the estimated range of the maximum concentration. For example, based on experience, the maximum concentration of this yeast is typically around 10.3 x 10⁻⁶. 6 Up to 25.0 x 10 6 Between cells per milliliter.

[0074] Based on this estimated range, the system generates a series of candidate maximum concentration values ​​according to preset step size parameters. These candidate values ​​may include 10.3 x 10 6 10.4 x 10 6 10.5 x 10 6 Until 25.0 x 10 6 Cells / mL

[0075] For each candidate maximum concentration value, the system calculates the deviation between the candidate value and the concentration data at the initial time point (0 hours). Then, using the inverse function of the Gompertz function model and the concentration data at 0 hours, 6 hours, and 12 hours, it solves for the maximum specific growth rate and the duration of the lag phase corresponding to that candidate maximum concentration value. For example, when the candidate maximum concentration value is 13.6 x 10⁻⁶... 6 When the cell / mL ratio is 1, the maximum specific growth rate is calculated to be 0.5 h. -1 The duration of the lag period is 6.0 hours.

[0076] Furthermore, the system combines each candidate maximum concentration value and its corresponding maximum specific growth rate and lag phase duration into a candidate parameter set. For example, a candidate parameter set might be {maximum concentration = 13.6 x 10⁻⁶}. 6 Maximum specific growth rate = 0.5 h -1 The system uses a Gompertz function model to predict microbial concentrations at 0, 6, and 12 hours based on each candidate parameter set, and calculates the prediction error (e.g., mean squared error) between these predictions and actual observations. By comparing the prediction errors of all candidate parameter sets, the system ultimately determines the candidate parameter set with the smallest prediction error as the final growth curve prediction parameters. Assume the final determined parameter set is {maximum concentration = 13.6 x 10⁻⁶}. 6 Maximum specific growth rate = 0.5 h -1 The duration of the lag period is 6 hours.

[0077] Finally, based on these finalized growth curve prediction parameters, the system calculates the key time points of the yeast fermentation broth experimental sample's growth period, such as the start of the logarithmic phase, the moment of maximum growth rate, and the end of the logarithmic phase. For example, the calculated start of the logarithmic phase is 6 hours, the moment of maximum growth rate is 15 hours, and the end of the logarithmic phase is 24 hours. Based on these key nodes and the complete Gompertz model, the system generates a complete growth prediction curve for the fermentation process, intuitively displaying the growth trend of the yeast throughout the fermentation process. Using this method, user A can obtain reliable growth curve predictions with only a few data points and understand in advance the arrival of key growth stages, thus guiding subsequent experimental operations. Furthermore, the system can automatically output the above prediction results at the 12th hour of the experiment and prompt the student: "The logarithmic phase is expected to begin in 6 hours; it is recommended to increase sampling and observation frequency around 15 hours."

[0078] The above is a detailed description of an embodiment of a method for predicting the microbial growth curve of a fermentation process provided by this application. The following is a detailed description of an embodiment of a device for predicting the microbial growth curve of a fermentation process provided by this application.

[0079] Please see Figure 2 This application provides an embodiment of a microbial growth curve prediction device for a fermentation process, the device comprising: The image acquisition unit 201 is used to acquire microscopic images of the fermentation broth experimental sample at preset sampling time points. The microbial concentration conversion unit 202 is used to perform image recognition and target object statistics based on the microscopic image to obtain the number of microorganisms corresponding to different sampling time points, and to convert the concentration data corresponding to different sampling time points based on the number of microorganisms. The maximum concentration range prediction unit 203 is used to determine the predicted range of the maximum concentration based on the concentration data corresponding to two non-initial time points and in combination with prior knowledge of microbial growth. The maximum concentration candidate value determination unit 204 is used to determine multiple maximum concentration candidate values ​​according to the estimated range and a preset step size parameter. The inverse function solving unit 205 is used to calculate the deviation between the candidate maximum concentration value and the concentration data at the initial time point based on each candidate maximum concentration value, and then obtain the maximum specific growth rate and the duration of the lag phase by solving the inverse function of the Gompertz function model based on the deviation value. The growth curve prediction parameter determination unit 206 is used to divide the maximum concentration candidate value and the corresponding maximum specific growth rate and lag phase duration into a candidate parameter group for each maximum concentration candidate value, calculate the prediction error of each candidate parameter group in combination with the Gompertz function model, and determine the final growth curve prediction parameters based on the candidate parameter group with the smallest prediction error. The growth prediction curve generation unit 207 is used to calculate the key time points of the growth period of the fermentation broth experimental sample based on the growth curve prediction parameters, so as to generate a growth prediction curve of the fermentation process based on the key nodes of the growth period.

[0080] This device combines the maximum concentration prediction range with the inverse function solution of the Gompertz function model using a parameter iterative search method. This allows for stable identification of growth curve parameters even with only 2-3 data points, achieving real-time prediction of key growth phase points and supporting inquiry-based learning. Specifically, in junior high school biological fermentation experiments, due to time constraints and operational limitations, only the initial time point and two non-initial time points (e.g., 6h and 12h) concentration data are typically available. Traditional nonlinear least squares methods often fail to converge or easily get trapped in local optima due to poor parameter identification. This embodiment introduces prior knowledge of microbial growth through a maximum concentration range prediction unit, limiting the maximum concentration to a reasonable range. A maximum concentration candidate value determination unit generates discrete candidate values, and an inverse function solution unit derives the maximum specific growth rate and lag phase duration for each candidate value. The growth curve prediction parameter determination unit filters the optimal parameter set by minimizing the prediction error, effectively avoiding the instability of multi-parameter fitting under sparse data. Based on this, the growth prediction curve generation unit calculates key time points such as the start of the logarithmic phase and the moment of maximum growth rate in real time based on the final parameters, enabling students to know the changes in the growth stage in advance during the experiment and achieve the inquiry-based learning goal of "experimenting, predicting and verifying at the same time".

[0081] In some embodiments, this application further proposes an intelligent microscope device, including a memory and a processor; the memory is used to store program code corresponding to the above-described method for predicting the microbial growth curve of the fermentation process; the processor is used to read and execute the program code to implement the method for predicting the microbial growth curve of the fermentation process.

[0082] A smart microscope device is a unit that integrates optical microscopy with computational processing capabilities. It not only allows for microscopic observation of fermentation broth samples but also enables automated analysis and processing of acquired images, thereby predicting microbial growth curves. This device can be a standalone, embedded system or an integrated system connecting a microscope to an external computing unit (e.g., an industrial computer or dedicated server) via a data interface. Memory is a hardware component used to store data and instructions. It can include various types, such as random access memory (RAM) for temporary storage and program execution, and non-volatile memory (e.g., flash memory, solid-state drives, or hard disk drives) for persistent storage of the operating system, application programs (such as program code for microbial growth curve prediction methods), and experimental data. The processor is the core computing unit of the smart microscope device, responsible for executing the program code in memory and performing various data processing and calculations. The processor can be a general-purpose central processing unit (CPU), such as a processor based on ARM or x86 architecture, or a dedicated processor optimized for specific tasks, such as a digital signal processor (DSP) or a graphics processing unit (GPU), to efficiently handle image recognition and complex mathematical model calculations. Program code refers to a set of instructions written in a specific programming language. After being read and executed by a processor, these instructions enable the various functions of the microbial growth curve prediction method for the fermentation process described above. This code can exist in the form of compiled machine code, interpreted scripts, or firmware; they are the concrete carriers of the method's logic and algorithms. Reading and executing the program code refers to the process by which the processor retrieves program instructions from memory and decodes and operates them according to their logical order. This process is the foundation for automated and intelligent analysis, enabling abstract prediction methods to be transformed into practical, runnable computational flows.

[0083] By integrating the microbial growth curve prediction method for the fermentation process into an intelligent microscope, the above technical solution significantly improves the automation level and efficiency of the prediction process. This equipment can autonomously complete the entire process from microscopic image acquisition to growth curve prediction, significantly reducing manual intervention, effectively avoiding human error, and lowering the skill requirements for operators. This integrated solution enables continuous and real-time monitoring and analysis of microbial growth curves, providing timely and accurate data support for the optimization and control of the fermentation process. Furthermore, the combination of dedicated hardware and optimized software ensures the stable operation of complex algorithms and the reliability and repeatability of prediction results, transforming abstract prediction methods into an efficient, convenient, and easily deployable practical tool, thereby improving the intelligence level and production efficiency of the fermentation industry.

[0084] In some embodiments, this application further proposes that the computer-readable storage medium stores program code for being read and executed by a processor to implement the method for predicting the microbial growth curve of the fermentation process.

[0085] Computer-readable storage media refers to any physical medium capable of storing digital data and readable by a computer system. Its function is to serve as a carrier of program code, enabling the program code to be persistently stored and accessed and loaded by the processor when needed. Possible implementations include, but are not limited to: non-volatile memory, such as hard disk drives, solid-state drives, read-only memory, flash memory, or optical discs; and volatile memory, such as random access memory, used as temporary storage space during program execution. Program code refers to a set of instructions written in a computer-understandable language, used to instruct the processor to perform specific operations and calculations. Its function is to encapsulate the specific logic and algorithms of the microbial growth curve prediction method for the fermentation process, enabling the method to be executed automatically by a computer. Program code can exist in various forms, such as compiled executable files, interpreted scripts, bytecode, or source code. The processor is the core computing unit of the computer system, responsible for interpreting and executing the instructions in the program code, performing data processing and control operations. Its function is to execute various calculations and logical judgments in the microbial growth curve prediction method for the fermentation process according to the instructions of the program code, thereby driving the completion of the entire prediction process. The processor can be a central processing unit, graphics processing unit, microcontroller, or application-specific integrated circuit, etc. "Read and execute" refers to the process by which a processor retrieves program code from a computer-readable storage medium and processes these instructions line by line according to the order and logic defined in the code. Specifically, the processor first loads the program code from the storage medium into memory, then decodes and executes the instructions and manages the data flow through its internal control unit and arithmetic logic unit. This process is fundamental to implementing any software function, ensuring that the microbial growth curve prediction method for the fermentation process can be correctly run by the computer system. Implementing the aforementioned microbial growth curve prediction method for the fermentation process means that the processor reads and executes the program code stored on the computer-readable storage medium, thereby completely and automatically completing all steps and functions of the method. This means that from image acquisition, microbial quantity statistics, concentration conversion, maximum concentration range estimation, parameter solving, prediction error calculation to the generation of the final growth curve, all these complex calculations and logical operations will be automatically completed by the computer system without human intervention.

[0086] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the terminals, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0087] In the several embodiments provided in this application, it should be understood that the disclosed terminals, devices, and methods can be implemented in other ways. For example, the device embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be an indirect coupling or communication connection between devices or units through some interfaces, and may be electrical, mechanical, or other forms.

[0088] The terms “first,” “second,” “third,” “fourth,” etc. (if present) in the specification and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a particular order or sequence. It should be understood that such data can be interchanged where appropriate so that embodiments of the application described herein can be implemented, for example, in orders other than those illustrated or described herein. Furthermore, the terms “comprising” and “having,” and any variations thereof, are intended to cover a non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0089] It should be understood that in this application, "at least one (item)" means one or more, and "more than" means two or more. "And / or" is used to describe the relationship between related objects, indicating that three relationships can exist. For example, "A and / or B" can represent three cases: only A exists, only B exists, and both A and B exist simultaneously, where A and B can be singular or plural. The character " / " generally indicates that the preceding and following related objects are in an "or" relationship. "At least one (item) of the following" or similar expressions refer to any combination of these items, including any combination of single or plural items. For example, at least one (item) of a, b, or c can represent: a, b, c, "a and b", "a and c", "b and c", or "a and b and c", where a, b, and c can be single or multiple.

[0090] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0091] Furthermore, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0092] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention.

[0093] The above-described embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application.

Claims

1. A method for predicting microbial growth curves during fermentation, characterized in that, include: Microscopic images of the fermentation broth experimental samples were collected at preset sampling time points. Based on the microscopic images, image recognition and target object statistics are performed to obtain the number of microorganisms corresponding to different sampling time points, and the concentration data corresponding to different sampling time points is calculated based on the number of microorganisms. Based on the concentration data corresponding to two non-initial time points, and combined with prior knowledge of microbial growth, the estimated range of the maximum concentration is determined. Based on the estimated range, and according to the preset step size parameters, multiple candidate values ​​for maximum concentration are determined; Based on each candidate maximum concentration value, the deviation between the candidate maximum concentration value and the concentration data at the initial time point is calculated. Then, based on the deviation value, the maximum specific growth rate and the duration of the lag phase are obtained by solving the inverse function of the Gompertz function model. For each candidate value of maximum concentration, the candidate value of maximum concentration and the corresponding maximum specific growth rate and lag phase duration are divided into a candidate parameter group. Combined with the Gompertz function model, the prediction error of each candidate parameter group is calculated, and the final growth curve prediction parameters are determined according to the candidate parameter group with the smallest prediction error. Based on the growth curve prediction parameters, the key time points of the growth period of the fermentation broth experimental sample are calculated, so as to generate a growth prediction curve of the fermentation process based on the key growth period nodes.

2. The method for predicting microbial growth curves in a fermentation process according to claim 1, characterized in that, After periodically acquiring microscopic images of the fermentation broth experimental samples, the process also includes: If the carrier of the fermentation broth experimental sample is a hemocytometer, then the grid lines in the microscopic image are identified by image recognition and Hough transform processing. Based on the grid lines, several counting regions are divided in the microscopic image.

3. The method for predicting microbial growth curves in a fermentation process according to claim 2, characterized in that, After identifying the grid lines in a microscopic image, the following steps are also included: The mask corresponding to the grid lines is removed from the original microscopic image, and then the remaining area is segmented into microbial images.

4. The method for predicting microbial growth curves in a fermentation process according to claim 2, characterized in that, After periodically acquiring microscopic images of the fermentation broth experimental samples, and before identifying the grid lines in the microscopic images using image recognition and Hough transform processing, the process further includes: The microscopic image is preprocessed, including median filtering for noise removal and adaptive histogram equalization for enhancing the contrast between grid lines and microorganisms.

5. The method for predicting microbial growth curves in a fermentation process according to claim 1, characterized in that, The formula for the Gompertz function model is: In the formula, Let be the microbial concentration at time t. The initial microbial concentration. For maximum concentration, For the maximum specific growth rate, The duration of the lag period is given by ε, and e is a natural constant.

6. The method for predicting microbial growth curves in a fermentation process according to claim 1, characterized in that, The formula for calculating the key time points during the growth period is: In the formula, Represents the starting point of the logarithmic period. Represents the moment of maximum growth rate. Represents the end of the logarithmic period. The initial microbial concentration. For maximum concentration, For the maximum specific growth rate, The duration of the lag period is given by ε, and e is a natural constant.

7. A method for predicting microbial growth curves in a fermentation process according to any one of claims 1 to 6, characterized in that, The sampling time points include: an initial time point and two non-initial time points, wherein the non-initial time points are the 6-hour and 12-hour time points starting from the initial time point, respectively.

8. A device for predicting the microbial growth curve of a fermentation process, characterized in that, include: The image acquisition unit is used to acquire microscopic images of the fermentation broth experimental samples at preset sampling time points. The microbial concentration conversion unit is used to perform image recognition and target object statistics based on the microscopic image to obtain the number of microorganisms corresponding to different sampling time points, and to convert the concentration data corresponding to different sampling time points based on the number of microorganisms. The maximum concentration range prediction unit is used to determine the predicted range of the maximum concentration based on the concentration data corresponding to two non-initial time points and combined with prior knowledge of microbial growth. The maximum concentration candidate value determination unit is used to determine multiple maximum concentration candidate values ​​according to the estimated range and a preset step size parameter. The inverse function solving unit is used to calculate the deviation between the candidate maximum concentration and the concentration data at the initial time point based on each candidate maximum concentration, and then use the deviation to solve the inverse function of the Gompertz function model to obtain the maximum specific growth rate and the duration of the lag phase. The growth curve prediction parameter determination unit is used to divide the maximum concentration candidate value, the corresponding maximum specific growth rate and lag phase duration into a candidate parameter group for each maximum concentration candidate value, calculate the prediction error of each candidate parameter group in combination with the Gompertz function model, and determine the final growth curve prediction parameters based on the candidate parameter group with the smallest prediction error. The growth prediction curve generation unit is used to calculate the key time points of the growth period of the fermentation broth experimental sample based on the growth curve prediction parameters, so as to generate a growth prediction curve of the fermentation process based on the key time points of the growth period.

9. An intelligent microscope device, characterized in that, include: Memory and processor; The memory is used to store program code, which corresponds to the method for predicting microbial growth curves in fermentation processes as described in any one of claims 1 to 7. The processor is used to read and execute the program code to implement the method for predicting the microbial growth curve of the fermentation process.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium contains program code that is read and executed by a processor to implement the method for predicting microbial growth curves in a fermentation process as described in any one of claims 1 to 7.

Images