Grain storage mildew detection method and system based on extreme learning machine, computer equipment and storage medium

Abstract

The grain storage mildew detection method based on the extreme learning machine comprises the following steps: periodically collecting grain storage environment parameters in a grain depot in a grain storage process; the grain storage environment parameters are input into a pre-trained mildew detection model for detection to obtain a mildew condition, the mildew detection model is obtained based on training of a regularization extreme learning machine model, and the method for training the mildew detection model comprises the steps of setting a plurality of initial regularization coefficients within a preset value limit; optimizing the initial regularization coefficient by using a subtraction average optimization algorithm to obtain an optimal regularization coefficient; performing optimization adjustment on the basic extreme learning machine model by using the optimal regularization coefficient to obtain a regularization extreme learning machine model; and constructing a sample set, and training the regularization extreme learning machine model by using the sample set. According to the invention, the mildew condition of the grain can be detected more efficiently, and the safety of the stored grain is facilitated.

Description

Technical Field

[0001] The present invention relates to the technical field of grain storage, and in particular to a method, system, computer equipment and storage medium for detecting stored grain mildew based on an extreme learning machine. Background Art

[0002] Food security means ensuring that everyone has access to and can afford sufficient food to sustain life and health at all times. Grain storage is a crucial component of ensuring food security. Given the vast quantities of stored grain, scientific storage management and optimized storage levels are crucial measures to improve grain quality and reduce losses. Generally speaking, the main factors contributing to grain loss are insect pests, temperature, humidity, and mold. Circulation fumigation, widely used in grain storage, effectively suppresses pests and reduces losses caused by them. However, abnormal temperature and humidity can directly lead to active microbial growth in grain, leading to mold.

[0003] To keep abreast of grain mold conditions, existing technologies primarily predict mold based on the temperature and humidity environments in granaries. However, because mold is a complex process, and mold is difficult to detect with the naked eye during its initial growth stages, it's difficult to detect in the early stages of infection. Once discovered, mold activity is often at its most active, directly leading to grain spoilage and inedibility. Grain condition monitoring systems that only monitor temperature, or both temperature and humidity, often only detect overheating or overhumidity when mold is severe. This delays early warning of the grain storage environment, posing significant risks to grain storage safety and quality.

[0004] To more efficiently determine whether grain in a granary is moldy or likely to become moldy, numerous methods have emerged in the prior art. Among these, methods based on deep learning models have gradually become mainstream. Most existing methods use models based on BP neural networks for mold detection. BP neural networks are feedforward neural network models. The weights between the input layer, output layer, and hidden layer of this algorithm are iteratively obtained using a gradient descent method. The network architecture consists of two parts: forward propagation and backpropagation. While this approach offers strong learning capabilities and performance, it suffers from significant issues during the iterative process, such as long training times and a tendency to fall into local optimality and be unable to find a solution. Summary of the Invention

[0005] In order to address the deficiencies in the prior art, the present invention provides a stored grain mold detection method, system, computer equipment, and storage medium based on an extreme learning machine, which can more efficiently detect the moldiness of grain and is beneficial to the safety of stored grain.

[0006] In order to achieve the above object, the specific scheme adopted by the present invention is: a method for detecting stored grain mold based on an extreme learning machine, comprising the following steps: During the grain storage process, the grain storage environment parameters in the grain depot are collected periodically; The grain storage environment parameters are input into a pre-trained mold detection model to detect the mold situation. The mold detection model is trained based on a regularized extreme learning machine model. The method for training the mold detection model includes: Set multiple initial regularization coefficients within the preset value limits; The initial regularization coefficient is optimized using the subtraction average optimization algorithm to obtain the optimal regularization coefficient; The basic extreme learning machine model is optimized and adjusted using the optimal regularization coefficient to obtain a regularized extreme learning machine model; a sample set is constructed, and the regularized extreme learning machine model is trained using the sample set.

[0007] As a further optimization of the above-mentioned stored grain mold detection method based on extreme learning machine: a method of setting multiple initial regularization coefficients within a preset value limit is: C i =lo+r i *(hi-lo) i=1,2...N; Where r is a random value in the numerical interval [0,1], hi and lo are constants used to regulate the maximum and minimum limits of the regularization coefficient C, respectively, and N is the number of initial regularization coefficients.

[0008] As a further optimization of the above-mentioned stored grain mold detection method based on the extreme learning machine: the method of optimizing the initial regularization coefficient using the subtraction mean optimization algorithm includes: Perform a subtraction search on the initial regularization coefficient as follows: C i,j =C i,1 +(C j,1 -I*C i,1 )*sign(fit i -fit j ) i=1,2...N,j=1,2...N; Where I is a random number in the numerical interval [1,2], sign is the sign function signum, and fit is the objective function value of the individual particle; Update the optimal particle. The specific method is:

[0009] As a further optimization of the above-mentioned stored grain mold detection method based on extreme learning machine: the loss function of the regularized extreme learning machine model is: Where n is the number of hidden layer units in the regularized extreme learning machine model, x is the input grain storage environment parameter, I is the identity matrix, B is the regularized weight matrix, H is the hidden layer output matrix, T = [T_train1, T_train2, ..., T_train M ] T is the true value matrix, The output of the regularized extreme learning machine model is: The objective function of the regularized extreme learning machine model is: in, is the L2 norm regularization term or penalty term.

[0010] As a further optimization of the above-mentioned stored grain mold detection method based on extreme learning machine: the output weight solution condition of the regularized extreme learning machine model is: If H T If H is a non-singular matrix, then If H T If H is a singular matrix, then

[0011] As a further optimization of the above-mentioned stored grain mold detection method based on extreme learning machine: the grain storage environment parameters include temperature parameters, humidity parameters and carbon dioxide concentration parameters, wherein the temperature parameter includes multiple temperature point values ​​corresponding to different positions in the grain warehouse, the humidity parameter includes multiple humidity point values ​​corresponding to different positions in the grain warehouse, and the carbon dioxide concentration parameter includes multiple carbon dioxide concentration point values ​​corresponding to different positions in the grain warehouse.

[0012] A stored grain mold detection system based on an extreme learning machine is used to implement the above-mentioned stored grain mold detection method based on an extreme learning machine. The system includes: A grain condition detection subsystem, used to obtain the grain storage environment parameters; A model training subsystem is used to adjust and optimize the basic extreme learning machine model to obtain a regularized extreme learning machine model, and to train the regularized extreme learning machine model to obtain a mold detection model; The extreme learning machine algorithm prediction subsystem is used to run the mold detection model to detect the grain storage environment parameters to obtain the mold situation.

[0013] Computer equipment, including memory and processors; The memory is used to store computer programs; The processor is used to read and execute the computer program to implement the above-mentioned stored grain mold detection method based on extreme learning machine.

[0014] A storage medium is used to store a computer program, which, when executed, implements the above-mentioned method for detecting stored grain mold based on an extreme learning machine.

[0015] Beneficial effects: The present invention detects grain moldiness based on a regularized extreme learning machine. Compared with the existing method based on BP neural network, the overall efficiency is higher. In addition, the present invention further uses the subtraction average optimization algorithm to perform real-time optimal estimation of the regularization coefficient. Compared with the conventional algorithm that sets the coefficient to a fixed value, the proposed method can estimate the most appropriate regularization coefficient without setting an initial value, which can further improve the training efficiency of the extreme learning machine. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 It is a structural block diagram of the detection system of the present invention; Figure 2 It is a schematic diagram of the regularized extreme learning machine. DETAILED DESCRIPTION

[0017] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0018] A stored grain mold detection method based on extreme learning machine, including S1 to S2.

[0019] S1. During grain storage, periodically collect environmental parameters within the grain depot. These environmental parameters include temperature, humidity, and carbon dioxide concentration. The temperature parameter includes multiple temperature values ​​corresponding to different locations within the depot, the humidity parameter includes multiple humidity values ​​corresponding to different locations within the depot, and the carbon dioxide concentration parameter includes multiple carbon dioxide concentration values ​​corresponding to different locations within the depot. Temperature and humidity are common parameters that can easily lead to grain mold and will not be discussed in detail here. Furthermore, because grain respires under the influence of oxygen and temperature, releasing carbon dioxide, grain depots employ measures such as cooling and air isolation to slow or even eliminate this respiration. If local environmental conditions are not properly controlled during storage, and if proper ventilation is lacking, respiration can become more active, and carbon dioxide can gradually accumulate in temperature-abnormal layers or moldy areas. The more intense the respiration, the higher the carbon dioxide content. Therefore, in addition to temperature and humidity, the carbon dioxide content within the grain pile is also an important indicator of whether grain is moldy.

[0020] Furthermore, a sensor array can be deployed in the grain warehouse. This array consists of multiple sensor nodes, each equipped with specific sensors such as temperature, humidity, and carbon dioxide concentration. Based on this sensor array, each sensor node can simultaneously acquire the values ​​of multiple parameter points, which is more efficient and easier to network, enabling the efficient transmission of grain storage environment parameters.

[0021] S2. Input the stored grain environmental parameters into a pre-trained mold detection model to detect mold conditions. The mold detection model is trained based on a regularized extreme learning machine model. Compared to traditional models such as BP neural networks, the different parameters of the extreme learning machine (ELM) have no dependencies. Since the weights and thresholds between the input layer and the hidden layer are randomly generated by a set function, there is no need for gradient iteration optimization, which meets the goals of rapid convergence and efficient learning. Therefore, the present invention constructs a mold detection model based on the extreme learning machine.

[0022] In order to further improve the performance of the mold detection model, the method for training the mold detection model includes S100 to S103.

[0023] While extreme learning machines (ELMs) offer greater efficiency than traditional BP neural network models, they still struggle when processing large amounts of data. Grain storage lasts for long periods of time, resulting in a significant amount of data on storage environment parameters. Therefore, the present invention optimizes the basic ELM to create a regularized ELM, which is more capable of processing large amounts of data.

[0024] like Figure 2 As shown, the regularization coefficient is a very important parameter in a regularized extreme learning machine. It is necessary to select an appropriate regularization coefficient in advance and set it to a fixed value. However, this process is very time-consuming and requires repeated testing to achieve. To improve the training efficiency of the model, especially to reduce the time required to determine the regularization coefficient, the present invention uses a subtractive mean optimization algorithm to optimize the regularization coefficient. The specific method is as follows.

[0025] S100: setting a plurality of initial regularization coefficients within a preset value limit. Specifically, the method of setting a plurality of initial regularization coefficients within a preset value limit is: C i =lo+r i *(hi-lo)i=1,2...N; Where r is a random value in the numerical interval [0,1], hi and lo are constants, which are used to regulate the maximum and minimum limit values ​​of the regularization coefficient C, respectively. N is the number of initial regularization coefficients. The larger the value of N, the easier it is to find the optimal value of the regularization coefficient. The smaller the value of V, the faster the convergence efficiency.

[0026] S101, using a subtraction mean optimization algorithm to optimize the initial regularization coefficient to obtain an optimal regularization coefficient. Specifically, the method of optimizing the initial regularization coefficient using the subtraction mean optimization algorithm includes: Perform a subtraction search on the initial regularization coefficient as follows: C i,j =C i,1 +(C j,1 -I*C i,1 )*sign(fit i -fit j )i=1,2...N,j=1,2...N; Where I is a random number in the numerical interval [1,2], sign is the sign function signum, and fit is the objective function value of the individual particle; Update the optimal particle. The specific method is:

[0027] In the above formula The new coefficient value obtained by subtracting the average is determined by determining whether the new C value exceeds the maximum or minimum limit. If so, the maximum boundary value is assigned to the new C value. Finally, based on the value of the objective function, the updated coefficient value is selected or the initial coefficient value is retained to complete the update of the optimal particle. This achieves the optimal estimate of the regularization coefficient C.

[0028] Compared with the traditional extreme learning machine, the present invention uses the subtraction balance optimization algorithm to optimize the regularization coefficient C without setting a fixed coefficient C, and estimates the most appropriate regularization coefficient based on the minimum condition of the objective function. It can significantly improve the efficiency of the regularization coefficient determination process, thereby improving the training efficiency of the mold detection model.

[0029] S102. Optimize and adjust the basic extreme learning machine model using the optimal regularization coefficient to obtain a regularized extreme learning machine model. The loss function of the regularized extreme learning machine model is: Where n is the number of hidden layer units in the regularized extreme learning machine model, x is the input grain storage environment parameter, I is the identity matrix, B is the regularized weight matrix, H is the hidden layer output matrix, T = [T_train1, T_train2, ..., T_train M ] T is the true value matrix.

[0030] The output of the regularized extreme learning machine model is:

[0031] The objective function of the regularized extreme learning machine model is: in, is the L2 norm regularization term or penalty term.

[0032] The output weight solution condition of the regularized extreme learning machine model is: If H T If H is a non-singular matrix, then If H T If H is a singular matrix, then

[0033] S103: Construct a sample set, and use the sample set to train a regularized extreme learning machine model.

[0034] More specifically, in one embodiment of the present invention, the optimization process of the regularized extreme learning machine can be expressed as follows.

[0035] First, construct the regularized extreme learning machine model. The output f of the extreme learning machine model is n And the loss function L can be described by the following formula: Where n is the number of hidden layer units, x is the input, and the model objective function after adding the regularization term is: It is the L2 norm regularization term or penalty term, which is balanced by the regularization coefficient C to improve the model's ability to resist interference from outliers. This is achieved by maximizing the distance between edge separation classes, so that the model maintains sufficient intervals in the classification task and improves the correct classification rate of new samples.

[0036] Then, the solution obtained by combining the optimal conditions of the Lagrangian function is: where I is the identity matrix, is the regularized weight matrix, H is the hidden layer output matrix, T=[T_train1,T_train2,...,T_train M ] T For the true value matrix, a regularized extreme learning prediction model with stronger anti-interference ability is constructed.

[0037] Then, the regularization coefficient is optimized using the subtraction mean optimization algorithm. Specifically, a series of C values ​​are initialized first, that is, the regularization parameters are initialized: C i =lo+r i *(hi-lo) i=1,2...N; (5) Where r is a random value between [0, 1]; hi and lo are constants that define the maximum and minimum limits of the regularization coefficient C, respectively; and N is the number of particles set by the algorithm. A larger value makes it easier to find the optimal value, while a smaller value results in higher convergence efficiency. Each initial regularization coefficient can be considered a particle in the subtractive mean optimization algorithm.

[0038] Next, perform subtraction for individual particle searches: C i,j =C i,1 +(C j,1 -I*C i,1 )*sign(fit i -fit j ) i=1,2...N,j=1,2...N; (6) Among them, I is a random number generated between [1,2]; sign is the sign function signum; fit is the objective function value of the individual particle.

[0039] Then, the optimal particle is updated: In the above formula The new coefficient value obtained by subtracting the average is determined by formula (8) to determine whether the new C value exceeds the maximum or minimum limit. If it exceeds the limit, the maximum boundary value is assigned to the new C value. Finally, based on the value of the objective function, the updated coefficient value is selected or the initial coefficient value is retained to complete the update of the optimal particle. This achieves the optimal estimate of the regularization coefficient C.

[0040] Finally, in the optimized extreme learning machine, it can be seen from formula (2) that in the process of training the extreme learning machine model, the optimal condition for learning and training is to minimize the loss function L. Transforming formula (2) yields: Hβ=Τ;(10) Since the number of hidden layer neurons and activation function have been determined, here the number of hidden layer neurons is 250 and the activation function is sigmoid function. Therefore, the training process of the extreme learning machine only needs to find the least squares solution of the above formula. That's it.

[0041] After the regularization is introduced, the regularization parameter C is optimized online. Since the inverse of the hidden layer output matrix H is required, the H matrix needs to be judged. Combining formula (4) and formula (9), we can get: If H T If H is a non-singular matrix, then H -1 =(H T H) -1 H T ,have: If H T If H is a singular matrix, then H -1 =H T (H T H) -1 ,have:

[0042] The specific setting method of the parameters of the optimized regularized extreme learning machine is shown in Table 1.

[0043] Table 1 Model parameters of the extreme learning machine after optimization

[0044] like Figure 1 As shown, the present invention further provides a stored grain moldy detection system based on an extreme learning machine, which is used to implement the above-mentioned stored grain moldy detection method based on an extreme learning machine. The system includes a grain condition detection subsystem, a model training subsystem, an extreme learning machine algorithm prediction subsystem and a grain moldy situation early warning subsystem.

[0045] The grain condition detection subsystem is used to obtain the grain storage environment parameters.

[0046] The model training subsystem is used to adjust and optimize the basic extreme learning machine model to obtain a regularized extreme learning machine model, and to train the regularized extreme learning machine model to obtain a mold detection model.

[0047] The extreme learning machine algorithm prediction subsystem is used to run the mold detection model to detect the grain storage environment parameters to obtain the mold situation.

[0048] The grain mold early warning subsystem is used to send early warning signals to staff and may include at least one grain storage status display terminal.

[0049] Those skilled in the art can clearly understand that, for the convenience and simplicity of description, the specific working process of the system and device described above can refer to the corresponding process in the aforementioned method embodiment, and will not be repeated here. In the several embodiments provided in the present disclosure, it should be understood that the disclosed system, device and method can be implemented in other ways. The device embodiments described above are merely schematic. For example, the division of the modules is only a logical function division. There may be other division methods in actual implementation. For example, multiple modules or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some communication interfaces, indirect coupling or communication connection of devices or modules, which can be electrical, mechanical or other forms.

[0050] The modules described as separate components may or may not be physically separate, and the components shown as modules may or may not be physical modules, that is, they may be located in one place or distributed across multiple network modules. Some or all of the modules may be selected to achieve the purpose of this embodiment according to actual needs.

[0051] The present invention also provides a computer device including a memory and a processor.

[0052] The memory is used to store computer programs.

[0053] The processor is used to read and execute the computer program to implement the above-mentioned stored grain mold detection method based on extreme learning machine.

[0054] Among them, the memory, as a carrier for resource storage, can be a read-only memory, random access memory, disk or CD, etc. The resources stored thereon can include an operating system, computer programs, etc., and the storage method can be temporary storage or permanent storage. The operating system is used to manage and control the various hardware devices and computer programs on the computer device, which can be Windows Server, Netware, Unix, Linux, etc. In addition to including a computer program that can be used to complete the adaptive emotion regulation method based on personalized reconfigurable music executed by the computer device disclosed in any of the aforementioned embodiments, the computer program can further include a computer program that can be used to complete other specific tasks. The processor can adopt a general-purpose processor product based on X86, IA64, RISC, MIPS and ARM architectures.

[0055] In some embodiments, the processor may be a central processing unit (CPU), a microprocessor or other data processing chip for running program codes stored in the memory or processing data.

[0056] The embodiment of the present invention does not specifically limit the type of computer device mentioned. The computer device may be a portable computer device such as a mobile phone, a tablet computer, a personal digital assistant (PDA), a wearable device, a laptop computer, or the like. Exemplary embodiments of portable computer devices include but are not limited to portable computer devices equipped with iOS, Android, Microsoft, or other operating systems. The above-mentioned portable computer device may also be other portable computer devices, such as a laptop computer with a touch-sensitive surface (e.g., a touch panel). It should also be understood that in some other embodiments of the present invention, the computer device may not be a portable computer device, but a desktop computer with a touch-sensitive surface (e.g., a touch panel).

[0057] Finally, the present invention provides a storage medium for storing a computer program, which, when executed, implements the above-mentioned method for detecting stored grain mold based on an extreme learning machine.

[0058] The above description of the disclosed embodiments is intended to enable one skilled in the art to implement or use the present invention. Various modifications to these embodiments will be readily apparent to one skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention is not limited to the embodiments shown herein but is intended to conform to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for detecting stored grain mold based on an extreme learning machine, characterized in that: The steps include: During the grain storage process, the grain storage environment parameters in the grain depot are collected periodically; The grain storage environment parameters are input into a pre-trained mold detection model to detect the mold situation. The mold detection model is trained based on a regularized extreme learning machine model. The method for training the mold detection model includes: Set multiple initial regularization coefficients within the preset value limits; The initial regularization coefficient is optimized using the subtraction average optimization algorithm to obtain the optimal regularization coefficient; The basic extreme learning machine model is optimized and adjusted using the optimal regularization coefficient to obtain a regularized extreme learning machine model; Construct a sample set and use it to train the regularized extreme learning machine model.

2. The method for detecting stored grain mold based on an extreme learning machine according to claim 1, wherein: The method to set multiple initial regularization coefficients within the preset value limit is: C i =lo+r i *(hi-lo)i=1,2...N; Where r is a random value in the numerical interval [0,1], hi and lo are constants used to regulate the maximum and minimum limits of the regularization coefficient C, respectively, and N is the number of initial regularization coefficients.

3. The method for detecting stored grain mold based on an extreme learning machine according to claim 2, wherein: Methods for optimizing the initial regularization coefficient using the subtraction mean optimization algorithm include: Perform a subtraction search on the initial regularization coefficient as follows: C i,j =C i,1 +(C j,1 -I*C i,1 )*sign(fit i -fit j )i=1,2...N,j=1,2...N; Where I is a random number in the numerical interval [1,2], sign is the sign function signum, and fit is the objective function value of the individual particle; Update the optimal particle. The specific method is:

4. The method for detecting stored grain mold based on an extreme learning machine according to claim 3, wherein: The loss function of the regularized extreme learning machine model is: Where n is the number of hidden layer units in the regularized extreme learning machine model, x is the input grain storage environment parameter, I is the identity matrix, B is the regularized weight matrix, H is the hidden layer output matrix, T = [T_train1, T_train2, ..., T_train M ] T is the true value matrix, The output of the regularized extreme learning machine model is: The objective function of the regularized extreme learning machine model is: in, is the L2 norm regularization term or penalty term.

5. The method for detecting stored grain mold based on an extreme learning machine according to claim 4, wherein: The output weight solution condition of the regularized extreme learning machine model is: If H T If H is a non-singular matrix, then If H T If H is a singular matrix, then 6. The method for detecting stored grain mold based on an extreme learning machine according to claim 1, wherein: The grain storage environment parameters include temperature parameters, humidity parameters and carbon dioxide concentration parameters, wherein the temperature parameters include multiple temperature point values ​​corresponding to different positions in the grain warehouse, the humidity parameters include multiple humidity point values ​​corresponding to different positions in the grain warehouse, and the carbon dioxide concentration parameters include multiple carbon dioxide concentration point values ​​corresponding to different positions in the grain warehouse.

7. A stored grain mold detection system based on an extreme learning machine is characterized by: For implementing the method for detecting stored grain mold based on an extreme learning machine as described in any one of claims 1 to 6, the system comprises: A grain condition detection subsystem, used to obtain the grain storage environment parameters; A model training subsystem is used to adjust and optimize the basic extreme learning machine model to obtain a regularized extreme learning machine model, and to train the regularized extreme learning machine model to obtain a mold detection model; The extreme learning machine algorithm prediction subsystem is used to run the mold detection model to detect the grain storage environment parameters to obtain the mold situation.

8. Computer equipment, characterized in that including memory and processor; The memory is used to store computer programs; The processor is used to read and execute the computer program to implement the stored grain mold detection method based on the extreme learning machine as described in any one of claims 1-6.

9. A storage medium, characterized in that Used to store a computer program, which, when executed, implements the method for detecting stored grain mold based on an extreme learning machine as described in any one of claims 1 to 6.

Images