Soybean intelligent soaking method and system based on soybean size and mass transformation model, and computer

Abstract

The invention provides an intelligent soybean soaking method and system based on a soybean size and mass transformation model and a computer, and belongs to the technical field of soybean product processing. The method comprises the following steps: monitoring a soybean soaking process in real time through an image sensor, obtaining an image, and obtaining a soybean size index in the soaking process through an image recognition model; according to soybean food processing requirements on soybean soaking degree (quality), soybean character indexes and soaking condition indexes, obtaining prediction time of soybean soaking ending through a soybean quality prediction model; and obtaining real-time mass increase data of the soaked soybeans through a soybean size and mass transformation model, and ending soaking when the real-time mass increase data meets requirements. According to the invention, the soybean soaking work is intelligentized, so that fine management of soybean soaking is realized, invalid soaking time and operation cost are reduced, stable high-quality soybean products are ensured to be obtained, and digital intelligent production transformation of the soybean product processing industry is promoted.

Description

Technical Field

[0001] The present invention relates to the technical field of soy product processing, and particularly relates to an intelligent soy soaking method, system and computer based on a soy size and quality transformation model. Background Art

[0002] Soy products have a long history of production and consumption in China and are an important source of high-quality protein in the dietary structure of Chinese residents. At present, the market scale of the soy product industry is continuously expanding, and the output value has exceeded 100 billion yuan. The soy product processing industry is transforming towards large-scale digital and intelligent production.

[0003] At present, soy product processing enterprises still rely on manual judgment of the soaking state of soybeans during the soy soaking process. On the one hand, if the soybeans are not soaked enough, the prepared soy milk is yellowish in color and the protein extraction rate is low; if soaked excessively, the obtained soy milk is dark in color, gray in color, the coagulability decreases, and the prepared tofu is soft and brittle. On the other hand, different soy products require different soaking states of soybeans. Manual judgment is time-consuming and laborious and cannot accurately set the soaking time, making it difficult to meet the needs of stable high-quality soy product processing. Facing the huge raw material use and increasingly strict quality control requirements of modern soy product processing enterprises, an intelligent soy soaking method is needed to realize the digital and intelligent production transformation in the soy soaking stage.

[0004] At present, a research has proposed an automatic soybean feeding system and method. In CN118505610A, through the extraction and recognition of key frames of the soaking video, the maximum projections in the vertical and horizontal directions are obtained as the true lengths of the soybean grain length and grain width, and all soybean grain length and grain width data greater than their thresholds are set as the conditions for automatic soybean feeding. However, it is found in practice that the grain orientation of soybeans cannot be controlled during the soy soaking process, the maximum projections in the vertical and horizontal directions are not the true data of the soybean grain length and grain width, and after the soybean grain length and grain width no longer increase, its quality still increases significantly. It can be seen that this automatic system and method still have deficiencies such as manual setting and poor accuracy, and need to be improved in aspects such as intelligent prediction and automatic setting. Summary of the Invention

[0005] The technical problem to be solved by the present invention is to propose an intelligent soy soaking method based on a soy size and quality transformation model for realizing intelligent soaking of soybeans in different soy product processing processes in view of the above-mentioned deficiencies of the prior art.

[0006] To achieve the object of the present invention, the technical solution adopted by the present invention is as follows: An intelligent soy soaking method based on a soy size and quality transformation model, comprising the following steps: S1. Real-time monitor the soy soaking process through an image sensor and obtain images, and obtain the size indexes of soybeans during the soaking process through an image recognition model; S2. According to the requirements (quality) of the degree of soaked soybeans for soybean food processing, the soybean trait indicators, and the soaking condition indicators, obtain the predicted time when the soaking of soybeans ends through the soybean quality prediction model; S3. According to the soybean trait indicators, the soaking condition indicators, and the size indicators during the soaking process, obtain the real-time quality growth data of the soaked soybeans through the soybean size and quality transformation model. When the real-time quality growth data does not meet the requirements, correct the predicted time when the soaking of soybeans ends. When it meets the requirements, end the soaking.

[0007] Preferably, the image recognition model in step S1 is obtained by performing semantic segmentation annotation, preprocessing, and YOLO model training on the real-time soaking image data of soybeans; the soybean quality prediction model in step S2 is obtained by numerically optimizing the nonlinear model based on the soybean quality growth data and soaking time data of soybeans with different traits under different soaking conditions; the soybean size and quality transformation model in step S3 is obtained by training with the multivariate adaptive regression spline machine learning algorithm based on the real-time size indicators and real-time quality growth data of soybeans with different traits under different soaking conditions.

[0008] Preferably, in step S3 of the above technical solution, the soybean trait indicators include: 100-seed weight; the soaking condition indicators include: soaking water temperature, pH, soaking time; the soaking size indicators include: major axis, minor axis, and thickness of soybeans; the real-time quality growth data of soybeans includes one or more of the soybean mass growth rate and water absorption rate.

[0009] Preferably, step S1 is specifically: obtain the real-time soaking image of soybeans through the image sensor at set time intervals, and input the image into the image recognition model for processing; the model performs object detection on each soybean in the image to generate corresponding bounding boxes; subsequently, extract the longest distance within each bounding box as the number of pixels of the major axis, and output the longest distance perpendicular to the major axis as the minor axis and thickness data; use the constant proportionality coefficient of known pixels to the actual distance to convert the number of pixels into the real length of the three-dimensional size.

[0010] Preferably, step S2 is specifically: input the soybean trait indicators and the soaking temperature conditions into the soybean quality prediction model to obtain the predicted soaking end time under the requirements (quality) of the degree of soybean food soaking.

[0011] Preferably, the soybean quality prediction model in step S2 is a non - linear mathematical model, including: visco - elastic model, Boltzmann model, Logistic model. The soybean quality prediction model is based on soybean trait indicators, quality growth data, soaking condition indicators, and soybean size indicators, and obtains non - linear mathematical model parameters through an optimization algorithm. Under fixed soybean trait indicators and soaking temperature conditions, the soybean quality prediction model is a specific and unique non - linear mathematical model, which can predict the quality growth data and size indicators during the entire soaking process.

[0012] Preferably, step S3 is specifically: inputting the soybean trait indicators, soaking condition indicators, and real - time soybean size indicators during the soaking process into the soybean size - quality transformation model to obtain real - time soybean quality growth data.

[0013] Preferably, the algorithm of the soybean size - quality transformation model in step S3 includes multivariate adaptive regression splines. The soybean size - quality transformation model is trained through the multivariate adaptive regression splines algorithm based on soybean trait indicators, size indicators, soaking condition indicators, and quality growth data, and learns the relationship between soybean trait indicators, size indicators, soaking condition indicators, and soybean quality growth data. When the soaking temperature changes, the model can output real - time quality growth data according to the real - time temperature, and input it into the corresponding soybean quality prediction model to correct the predicted end time of soaking.

[0014] Further preferably, the image recognition model uses object detection based on prior boxes as the learning strategy, and stochastic gradient descent and non - maximum suppression as the optimization algorithms; the learning strategy of the soybean size - quality transformation model is piece - wise regression, adaptive basis function selection, and interaction effect modeling, and the optimization algorithm is backward pruning; the optimization algorithm of the soybean quality prediction model is the Levenberg - Marquardt algorithm.

[0015] The present invention also provides a smart soybean soaking system based on the soybean size - quality transformation model, including: A soaking module, including a soaking tank and its configured various automatic valves, which is used to soak soybeans and stop soaking according to the processor's instructions; A data acquisition module, including an image sensor, a pH sensor, a temperature sensor, a liquid level sensor, a processor, and a display, which is used to acquire real - time soaking images of soybeans, soaking condition indicators, and soybean trait indicators; A data processing module, including an image recognition model, a soybean quality prediction model, and a soybean size - quality transformation model, which is used to process the acquired various data respectively, including the following parts: Processing the acquired real - time soaking images of soybeans through the image recognition model to obtain the true length of the three - dimensional size; According to the requirements for the degree (quality) of soaked soybeans in soy food processing and the soaking condition indicators, the predicted time for the end of soybean soaking is obtained through the soybean quality prediction model; According to the trait indicators of soybeans, the soaking condition indicators, and the size indicators during the soaking process, the real-time quality growth data of soaked soybeans is obtained through the soybean size and quality transformation model; The execution module includes a necessary execution program to control the soaking module. According to the data output by the data processing module, when the real-time quality growth data does not meet the requirements, the predicted time for the end of soybean soaking is corrected, and when it meets the requirements, the soaking ends.

[0016] The present invention also provides a computer including a soybean intelligent soaking system based on the soybean size and quality transformation model. The computer includes a computer program that can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer can execute any one of the above methods.

[0017] Compared with the prior art, the present invention achieves the following technical effects: The present invention trains the image recognition model, the soybean quality prediction model, and the soybean size and quality transformation model through historical data. Based on the trained models, new image data is collected at fixed time intervals and input into the image recognition model to obtain the soybean size indicators; according to the requirements for the degree (quality) of soaked soybeans in soy food processing, the soybean trait indicators, and the soaking condition indicators, the predicted time for the end of soybean soaking is obtained through the soybean quality prediction model; the soybean trait indicators, the soaking condition indicators, and the soybean size indicators during the soaking process are input into the soybean size and quality transformation model to obtain the real-time quality growth data of soaked soybeans. When it does not meet the requirements, the predicted time for the end of soybean soaking is corrected, and when it meets the requirements, the soaking ends. Through the present invention, the soybean soaking work is made intelligent, thereby realizing the refined management of soybean soaking, reducing the ineffective soaking time and operation cost, ensuring the acquisition of stable high-quality soy products, and promoting the digital transformation of the soy product processing industry. Description of the Drawings

[0018] For ease of explanation, the present invention is described in detail by the following specific embodiments and drawings Figure 1 It is a schematic flow chart of a soybean intelligent soaking method based on the soybean size and quality transformation model provided by the present invention.

[0019] Figure 2 It is a schematic diagram of a soybean intelligent soaking system based on the soybean size and quality transformation model provided by the present invention. Detailed Description of the Invention

[0020] The following are specific embodiments of the present invention and, in combination with the accompanying drawings, further describe the technical solutions of the present invention. However, the present invention is not limited to these embodiments. In the following description, specific details such as specific configurations are provided only to help comprehensively understand the embodiments of the present invention. Therefore, those skilled in the art should clearly understand that various changes and modifications can be made to the embodiments described herein without departing from the scope and spirit of the present invention.

[0021] It should be noted that, without conflict, the embodiments in the present invention and the features in the embodiments can be combined with each other.

[0022] The following combines Figure 1 - Figure 2 to describe an embodiment of the intelligent soybean soaking method and system based on the soybean size and quality transformation model provided by the present invention.

[0023] Embodiment 1

[0024] Figure 1 is a schematic flowchart of the intelligent soybean soaking method based on the soybean size and quality transformation model provided by the present invention. Referring to Figure 1 , an intelligent soybean soaking method based on the soybean size and quality transformation model provided by the present invention may include: Step S1, monitoring the soybean soaking process in real time through an image sensor to obtain images, and obtaining the size index of the soybeans during the soaking process through an image recognition model; Step S2, according to the requirements for the degree (quality) of the soaked soybeans in soybean food processing, the soybean trait index, and the soaking condition index, obtaining the predicted time for the end of soybean soaking through the soybean quality prediction model; Step S3, according to the soybean trait index, the soaking condition index, and the size index during the soaking process, obtaining the real-time quality growth data of the soaked soybeans through the soybean size and quality transformation model. When the real-time quality growth data does not meet the requirements, correct the predicted time for the end of soybean soaking. When it meets the requirements, end the soaking.

[0025] It should be noted that the execution subject of the intelligent soybean soaking method based on the soybean size and quality transformation model provided by the present invention can be any terminal-side device that meets the technical requirements.

[0026] Before monitoring and predicting the end time of soybean soaking for different soy products through this method, the image data, trait index, soaking condition index, size index during the soaking process, and quality growth data of the soybeans can be obtained from the historical soy product raw material processing data as a training set, and model training can be carried out respectively in combination with different learning strategies and optimization algorithms to obtain an image recognition model, a soybean quality prediction model, and a soybean size and quality transformation model.

[0027] In this embodiment, the soybean trait indicators include: 100-seed weight; the soaking condition indicators include: soaking water temperature, pH, soaking time; the soaking size indicators include: the major axis, minor axis, and thickness of the soybean; the soybean mass growth data includes any one of the following or any combination thereof: soybean mass growth rate, water absorption rate.

[0028] In this embodiment, the obtained real-time soaking images of soybeans are input into an image recognition model. After preprocessing, object detection is performed on each soybean in the image to generate corresponding bounding boxes. Subsequently, the longest distance within each bounding box is extracted as the number of pixels of the major axis, and the longest distance perpendicular to the major axis is output as the data of the minor axis and thickness. Using the constant proportionality coefficient between known pixels and actual distances, the number of pixels is converted into the real length of the three-dimensional size; It should be noted that the image acquisition device includes an industrial image sensor and a fixed light source, and a calibration plate is used for size calibration to improve the accuracy and robustness of the soybean size indicators.

[0029] In this embodiment, the image recognition model includes the YOLO model; the soybean mass prediction model mainly includes non-linear mathematical models, including: viscoelastic model, Boltzmann model, Logistic model; the soybean size and mass transformation model algorithm includes: multivariate adaptive regression splines.

[0030] During the intelligent soaking process of soybean products raw materials, the obtained soybean trait indicators, soaking condition indicators, size indicators during soaking, and soybean real-time mass growth data are input into the image recognition model, soybean mass prediction model, and soybean size and mass transformation model to obtain the predicted end time of soybean soaking and monitor the soaking process. If the soaking temperature changes, the soybean size and mass transformation model outputs real-time mass growth data according to the real-time temperature, and inputs it into the corresponding soybean mass prediction model to correct the predicted end time of soaking, continuously improving the accuracy of the predicted end time. Through the soybean intelligent soaking method based on the soybean size and mass transformation model, enterprises can achieve refined management of soybean soaking, reduce ineffective soaking time, better control the quality of soybean products, and improve product competitiveness.

[0031] Embodiment 2

[0032] For the soybean intelligent soaking system based on the soybean size and mass transformation model, the soybean intelligent soaking system based on the soybean size and mass transformation model described in this embodiment can be correspondingly referred to the soybean intelligent soaking method based on the soybean size and mass transformation model in Embodiment 1.

[0033] Referring to Figure 2 , a soybean intelligent soaking system based on the soybean size and mass transformation model provided by the present invention may include: Soaking module, including a soaking tank and its configured various automatic valves, for soaking soybeans and stopping soaking according to processor instructions; Data acquisition module, including an image sensor, a pH sensor, a temperature sensor, a liquid level sensor, a processor and a display, for acquiring real-time soaking images of soybeans, soaking condition indicators, and soybean trait indicators; Data processing module, including an image recognition model, a soybean quality prediction model, and a soybean size and quality transformation model, for respectively processing the acquired data, including the following parts: Processing the acquired real-time soaking images of soybeans through the image recognition model to obtain the true length of the three-dimensional size; According to the requirement of the degree (quality) of soaked soybeans for soybean food processing and the soaking condition indicators, obtaining the predicted time for the end of soybean soaking through the soybean quality prediction model; According to the soybean trait indicators, soaking condition indicators, and size indicators during soaking, obtaining the real-time quality growth data of soaked soybeans through the soybean size and quality transformation model.

[0034] Execution module, including necessary execution programs to control the soaking module. According to the data output by the data processing module, when the real-time quality growth data does not meet the requirements, correcting the predicted time for the end of soybean soaking, and when it meets the requirements, ending the soaking.

[0035] A soybean intelligent soaking system based on the soybean size and quality transformation model provided by this embodiment further includes: Prediction module, for: predicting the size indicators and quality growth data of soybeans under different traits and different soaking conditions according to the soybean quality prediction model.

[0036] Embodiment 3

[0037] A computer, the computer includes a computer program of the soybean intelligent soaking system based on the soybean size and quality transformation model in Embodiment 2. The computer program can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer can execute the method in Embodiment 1 above. Specifically, the computer program includes several instructions to enable a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in various embodiments of the present invention. And the foregoing storage medium includes: USB flash drive, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), magnetic disk or optical disk, etc., which can store program codes.

[0038] Through the description of the above embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus a necessary general hardware platform, and of course, it can also be implemented by hardware. Based on such an understanding, the essence of the above technical solution or the part that contributes to the prior art can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., including several instructions to enable a computer device (which can be a personal computer, server, or network device, etc.) to execute the method described in the embodiment or some parts of the embodiment.

[0039] Those skilled in the art to which this application pertains can make various modifications, supplements, or use similar ways to substitute the described specific embodiments, but will not deviate from the inventive concept of this application or exceed the scope defined by the appended claims.

Claims

1. An intelligent soybean soaking method based on a soybean size and quality transformation model, characterized in that, It includes the following steps: S1. Monitor the soybean soaking process in real time through an image sensor and obtain images, and obtain the size index of soybeans during the soaking process through an image recognition model; S2. According to the degree requirement of soaked soybeans for soybean food processing, soybean trait indexes and soaking condition indexes, obtain the predicted time for the end of soybean soaking through a soybean quality prediction model; S3. According to the soybean trait indexes, soaking condition indexes, and size indexes during the soaking process, obtain the real-time quality growth data of soaked soybeans through a soybean size and quality transformation model. When the real-time quality growth data does not meet the requirements, correct the predicted time for the end of soybean soaking. When it meets the requirements, end the soaking.

2. The intelligent soybean soaking method based on the soybean size and quality transformation model according to claim 1, characterized in that, The image recognition model in step S1 is obtained by performing semantic segmentation annotation, preprocessing, and YOLO model training on the real-time soybean soaking image data; the soybean quality prediction model in step S2 is obtained by numerical optimization of a non-linear model based on the soybean quality growth data and soaking time data of different trait soybeans under different soaking conditions; the soybean size and quality transformation model in step S3 is obtained by training through a multivariate adaptive regression spline machine learning algorithm based on the real-time size indexes and real-time quality growth data of different trait soybeans under different soaking conditions.

3. The intelligent soybean soaking method based on the soybean size and quality transformation model according to claim 1, wherein In step S3, the soybean trait indexes include: 100-seed weight; the soaking condition indexes include: soaking water temperature, pH, soaking time; the soaking size indexes include: soybean major axis, minor axis, thickness; the real-time quality growth data of soybeans includes one or more of the soybean mass growth rate and water absorption rate.

4. A soybean intelligent soaking method based on a soybean size and quality transformation model according to claim 1, characterized in that, The specific step S1 is: obtain the real-time soybean soaking image through the image sensor at a set time interval, and input the image into the image recognition model for processing; the model performs object detection on each soybean in the image to generate a corresponding bounding box; subsequently, extract the longest distance within each bounding box as the major axis pixel number, and output the longest distance perpendicular to the major axis as the minor axis and thickness data; use the known constant proportional coefficient between pixels and actual distance to convert the pixel number into the real length of the three-dimensional size.

5. The intelligent soybean soaking method based on the soybean size and quality transformation model according to claim 1, characterized in that, The specific step S2 is: input the soybean trait indexes and soaking temperature conditions into the soybean quality prediction model to obtain the predicted soaking end time under the degree requirement of soybean food soaking.

6. The intelligent soybean soaking method based on the soybean size and quality transformation model according to claim 5, wherein The soybean quality prediction model in step S2 is a non-linear mathematical model, including: viscoelastic model, Boltzmann model, Logistic model. The soybean quality prediction model obtains the parameters of the non-linear mathematical model through an optimization algorithm based on the soybean trait indexes, quality growth data, soaking condition indexes, and soybean size indexes.

7. A soybean intelligent soaking method based on a soybean size and quality transformation model according to claim 1, characterized in that, The specific steps of step S3 are as follows: input the trait indicators of soybeans, soaking condition indicators, and real-time soybean size indicators during the soaking process into the soybean size and quality transformation model to obtain real-time soybean quality growth data; the algorithm of the soybean size and quality transformation model in step S3 includes multivariate adaptive regression splines, and the soybean size and quality transformation model is trained based on soybean trait indicators, size indicators, soaking condition indicators, and quality growth data through the multivariate adaptive regression spline algorithm to learn the relationship between soybean trait indicators, size indicators, soaking condition indicators, and soybean quality growth data.

8. A soybean intelligent soaking method based on a soybean size and quality transformation model according to any one of claims 1-7, characterized in that, The image recognition model uses object detection based on prior boxes as the learning strategy, and stochastic gradient descent and non-maximum suppression as the optimization algorithms; the learning strategy of the soybean size and quality transformation model is piecewise regression, adaptive basis function selection, and interaction effect modeling, and the optimization algorithm is backward pruning; the optimization algorithm of the soybean quality prediction model is the Levenberg-Marquardt algorithm.

9. A soybean intelligent soaking system based on a soybean size and quality transformation model, characterized in that, It includes: A soaking module, including a soaking tank and its configured automatic valves, for soaking soybeans and stopping soaking according to the instructions of the processor; A data acquisition module, including an image sensor, a pH sensor, a temperature sensor, a liquid level sensor, a processor, and a display, for acquiring real-time soaking images of soybeans, soaking condition indicators, and soybean trait indicators; A data processing module, including an image recognition model, a soybean quality prediction model, and a soybean size and quality transformation model, for respectively processing the acquired data, including the following parts: Processing the acquired real-time soaking images of soybeans through the image recognition model to obtain the true length of the three-dimensional size; According to the degree requirement of soaking soybeans for soybean food processing and the soaking condition indicators, obtaining the predicted time for the end of soybean soaking through the soybean quality prediction model; According to the trait indicators of soybeans, soaking condition indicators, and size indicators during the soaking process, obtaining real-time soybean quality growth data through the soybean size and quality transformation model; An execution module, including necessary execution programs to control the soaking module. According to the data output by the data processing module, when the real-time quality growth data does not meet the requirements, the predicted time for the end of soybean soaking is corrected, and when it meets the requirements, the soaking is ended.

10. A computer including the soybean intelligent soaking system based on the soybean size and quality transformation model according to claim 9, characterized in that, The computer includes a computer program of the intelligent soybean soaking system based on the soybean size and quality transformation model. The computer program can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer can execute the method according to any one of claims 1-8.

Images