A greenhouse access management method and system

Abstract

The present application provides a method and system for managing the entry and exit of a greenhouse. The method includes: obtaining the basic information, health status, historical entry and exit pattern data, and environmental parameters of the greenhouse of the object to be entered or left in the target time period; using a pre-trained prediction model to perform multi-level processing on the historical entry and exit pattern data, predicting the entry and exit demand information of the object in the target time period based on the multi-level processing results, and determining the initial disinfection process of the object according to the entry and exit demand information; the multi-levels include the potential distribution learning level and the graph structure modeling level; dynamically adjusting the initial disinfection process according to the basic information, health status, and environmental parameters of the greenhouse of the object to obtain the target disinfection process; executing the target disinfection process to obtain the disinfection result of the object, and determining whether to allow the object to enter or leave the greenhouse according to the disinfection result of the object. The technical solution provided by the present application can improve management efficiency.

Description

Technical Field

[0001] The embodiments of the present application relate to the technical field of greenhouses, and in particular to a method and system for managing entry and exit of greenhouses. Background Art

[0002] A greenhouse is a facility used to protect plant growth. It provides a suitable growth environment for plants by controlling the internal temperature, humidity, light and other environmental conditions. With the development of modern agriculture, greenhouses are being used more and more widely, not only for the cultivation of crops such as vegetables and fruits, but also for flowers, medicinal materials and other fields. However, the management and operation of greenhouses face a series of challenges, especially in the management of personnel and material access.

[0003] Existing greenhouse entry and exit management methods usually rely on manual management and simple automated equipment. For example, entry and exit management is completed through manual registration or card swiping systems. Manual registration and inspection are time-consuming and labor-intensive, prone to errors and omissions, especially during peak hours, when managers are under great pressure and prone to management errors. In addition, the disinfection process is often fixed, which may lead to incomplete disinfection or excessive disinfection, affecting entry and exit efficiency. Summary of the invention

[0004] The embodiments of the present application provide a greenhouse entry and exit management method and system to solve the problem of low greenhouse entry and exit management efficiency in the prior art.

[0005] In a first aspect, an embodiment of the present application provides a method for managing entry and exit of a greenhouse, comprising: The basic information, health status, historical entry and exit pattern data and environmental parameters of the objects to be entered or left in the greenhouse within the target time period are obtained through the intelligent corridor system connected between the greenhouses; The historical entry and exit pattern data are processed at multiple levels using a pre-trained prediction model, and the entry and exit demand information of the object within the target time period is predicted based on the multi-level processing results, and the initial disinfection process of the object is determined according to the entry and exit demand information; the multi-levels include a potential distribution learning level and a graph structure modeling level; According to the basic information and health status of the object and the environmental parameters of the greenhouse, the initial disinfection process is dynamically adjusted to obtain a target disinfection process; Execute the target disinfection process to obtain the object disinfection result, and determine whether to allow the object to enter or leave the greenhouse according to the object disinfection result.

[0006] Optionally, the pre-trained prediction model is used to perform multi-level processing on the historical entry and exit pattern data, and the entry and exit demand information of the object in the target time period is predicted based on the multi-level processing results, including: The historical entry-exit pattern data is processed at the potential distribution learning level using a pre-trained variational autoencoder to obtain object potential features; the historical entry-exit pattern data is processed at the graph structure modeling level using a pre-trained graph neural network to obtain object graph features; A target attention mechanism is used to perform weighted fusion processing on the object potential features and the object graph features to obtain weighted comprehensive features; the target attention mechanism is a multi-head attention mechanism that introduces residual connections; According to the weighted comprehensive features, a target prediction algorithm is adopted to predict the entry and exit demand information of the object within the target time period. The target prediction algorithm includes an adaptive reinforcement learning algorithm, a deep reinforcement learning algorithm and a multi-layer perceptron.

[0007] Optionally, a target attention mechanism is used to perform weighted fusion processing on the object potential features and the object graph features to obtain weighted comprehensive features, including: Using a normalization method to perform normalization processing on the object potential feature and the object graph feature respectively, to obtain a standardized potential feature and a standardized graph feature; The standardized latent features and the standardized graph features are respectively linearly transformed by using a multi-head linear transformation method in a multi-head attention mechanism to obtain multiple groups of first query vectors, first key vectors, and first value vectors of the standardized latent features, and multiple groups of second query vectors, second key vectors, and second value vectors of the standardized graph features; A multi-head attention score calculation method is used to perform dot product calculations on multiple groups of first query vectors and first key vectors of the standardized latent features, and multiple groups of second query vectors and second key vectors of the standardized graph features, to obtain multiple first attention score matrices of the standardized latent features and multiple second attention score matrices of the standardized graph features; Using a multi-head scaling method to scale the multiple first attention score matrices and the multiple second attention score matrices respectively, to obtain multiple scaled first attention score matrices and multiple scaled second attention score matrices; Using a multi-head softmax function to normalize the multiple scaled first attention score matrices and the multiple scaled second attention score matrices, respectively, to obtain multiple first attention weight matrices and multiple second attention weight matrices; According to the multiple first attention weight matrices, a multi-head weighted processing method is used to perform weighted processing on the first value vector to obtain multiple weighted potential features; according to the multiple second attention weight matrices, a multi-head weighted processing method is used to perform weighted processing on the second value vector to obtain multiple weighted graph features; Generating multi-head weighted latent features and multi-head weighted graph features respectively according to the plurality of weighted latent features and the plurality of weighted graph features; The multi-head weighted latent feature is residually connected with the standardized latent feature by using an adaptive residual connection method to obtain a target weighted latent feature, and the multi-head weighted graph feature is residually connected with the standardized graph feature to obtain a target weighted graph feature; The target weighted latent features and the target weighted graph features are fused by using a feature concatenation method to obtain weighted comprehensive features.

[0008] Optionally, the first value vector is weighted according to the multiple first attention weight matrices using a multi-head weighted processing method to obtain multiple weighted potential features, including: Using the multiple first attention weight matrices to perform dot multiplication operations on the first value vector respectively, and introducing position coding to correct each dot multiplication result to obtain multiple corrected first potential features; The modified first latent feature is nonlinearly transformed through a ReLU activation function to obtain a plurality of weighted latent features.

[0009] Optionally, the target prediction algorithm is used to predict the entry and exit demand information of the object within the target time period according to the weighted comprehensive features, and the target prediction algorithm includes an adaptive reinforcement learning algorithm, a deep reinforcement learning algorithm and a multi-layer perceptron, including: Adopting an adaptive reinforcement learning algorithm to perform feature selection and enhancement processing on the weighted comprehensive features to obtain optimized features; Using a deep reinforcement learning algorithm to perform strategy optimization processing on the optimization features to obtain an optimization strategy; A multi-layer perceptron is used to perform prediction processing on the optimization strategy to obtain the entry and exit demand information of the object within the target time period.

[0010] Optionally, dynamically adjusting the initial disinfection process according to the basic information and health status of the object and the environmental parameters of the greenhouse to obtain a target disinfection process includes: A decision tree algorithm is used to preliminarily classify the basic information, health status and environmental parameters of the greenhouse of the object to obtain a preliminary classification result; Using a support vector machine to refine the preliminary classification result to obtain a refined processing result; A random forest algorithm is used to comprehensively evaluate the basic information, health status and environmental parameters of the greenhouse of the object to obtain a comprehensive evaluation result; Using a convolutional neural network to perform deep feature extraction on the historical entry and exit pattern data of the object and the environmental parameters of the greenhouse to obtain a deep feature vector; The preliminary classification result, the refinement processing result, the comprehensive evaluation result and the deep feature vector are comprehensively processed by an integrated learning method to obtain a target disinfection process.

[0011] Optionally, the use of a convolutional neural network to perform deep feature extraction on the historical entry and exit pattern data of the object and the environmental parameters of the greenhouse to obtain a deep feature vector includes: Using the first convolutional layer to extract local features from the historical entry and exit pattern data of the object and the environmental parameters of the greenhouse to obtain a local feature map; Using a first pooling layer to downsample the local feature map to obtain a downsampled feature map; Using a combination of multiple second convolutional layers and second pooling layers to perform multi-layer feature extraction on the downsampled feature map to obtain a multi-level feature map; Using a fully connected layer to integrate the multi-level feature map to obtain a feature vector; An activation function is used to perform nonlinear transformation on the feature vector, and a normalization method is used to normalize the feature vector after the nonlinear transformation to obtain a deep feature vector.

[0012] In a second aspect, an embodiment of the present application provides a greenhouse entry and exit management system, including: An acquisition module is used to acquire basic information, health status, historical entry and exit mode data, and environmental parameters of the objects to be entered or left the greenhouse within a target time period through an intelligent corridor system connected between the greenhouses; A processing prediction module, used to use a pre-trained prediction model to perform multi-level processing on the historical entry and exit pattern data, predict the entry and exit demand information of the object within the target time period based on the multi-level processing results, and determine the initial disinfection process of the object according to the entry and exit demand information; the multi-levels include a potential distribution learning level and a graph structure modeling level; A dynamic adjustment module, used to dynamically adjust the initial disinfection process according to the basic information and health status of the object and the environmental parameters of the greenhouse to obtain a target disinfection process; The execution determination module is used to execute the target disinfection process to obtain the object disinfection result, and determine whether to allow the object to enter or leave the greenhouse according to the object disinfection result.

[0013] In a third aspect, an embodiment of the present application provides a computing device, including a processing component and a storage component; the storage component stores one or more computer instructions; the one or more computer instructions are used to be called and executed by the processing component to implement a greenhouse entry and exit management method as described in any one of the first aspects.

[0014] In a fourth aspect, an embodiment of the present application provides a computer storage medium storing a computer program, which, when executed by a computer, implements a method for managing entry and exit of a greenhouse as described in any one of the first aspects.

[0015] In an embodiment of the present application, a greenhouse entry and exit management method is provided, which obtains the basic information, health status, historical entry and exit mode data and environmental parameters of the object to be entered or left in the greenhouse within the target time period through an intelligent corridor system connected between each greenhouse; uses a pre-trained prediction model to process the historical entry and exit mode data at multiple levels, predicts the entry and exit demand information of the object within the target time period based on the multi-level processing results, and determines the initial disinfection process of the object according to the entry and exit demand information; the multiple levels include a potential distribution learning level and a graph structure modeling level; dynamically adjusts the initial disinfection process according to the basic information, health status and environmental parameters of the object to obtain a target disinfection process; executes the target disinfection process to obtain the disinfection result of the object, and determines whether to allow the object to enter or leave the greenhouse according to the disinfection result of the object. The prediction model in the embodiment of the present application can more comprehensively capture the complex relationship between the historical entry and exit mode and environmental parameters of the object through multi-level processing such as the potential distribution learning level and the graph structure modeling level, thereby improving the accuracy of the prediction; based on the accurate prediction results, the present embodiment can dynamically adjust the disinfection process to ensure that each object can be properly processed, thereby improving the flexibility and efficiency of management. In the process of making predictions based on multi-level processing results, the latent features and graph features are weightedly fused through the target attention mechanism to generate more comprehensive weighted comprehensive features, which can improve the accuracy of the prediction.

[0016] These and other aspects of the present application will become more clearly understood in the description of the following embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, a brief introduction will be given below to the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0018] Figure 1 A flow chart of a greenhouse entry and exit management method provided in an embodiment of the present application; Figure 2 A schematic diagram of the structure of a greenhouse entry and exit management system provided in an embodiment of the present application; Figure 3 A schematic diagram of the structure of a computing device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0019] In order to enable those skilled in the art to better understand the solution of the present application, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application.

[0020] In some of the processes described in the specification and claims of this application and the above-mentioned figures, multiple operations that appear in a specific order are included, but it should be clearly understood that these operations may not be executed in the order in which they appear in this article or executed in parallel. The sequence numbers of the operations, such as 11, 12, etc., are only used to distinguish between different operations, and the sequence numbers themselves do not represent any execution order. In addition, these processes may include more or fewer operations, and these operations may be executed in sequence or in parallel. It should be noted that the descriptions of "first", "second", etc. in this article are used to distinguish different messages, devices, modules, etc., do not represent the order of precedence, and do not limit the "first" and "second" to different types.

[0021] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of this application.

[0022] Figure 1 A flow chart of a greenhouse entry and exit management method provided in an embodiment of the present application, such as Figure 1 As shown, the method includes: S11. Obtain basic information, health status, historical entry and exit pattern data, and environmental parameters of the greenhouses of objects to enter or leave the greenhouses within a target time period through an intelligent corridor system connected between the greenhouses.

[0023] It should be understood that the smart corridor system refers to an automated channel connecting various greenhouses, equipped with various types of sensors (such as temperature sensors, humidity sensors, etc.), cameras, Radio Frequency Identification (RFID) readers and writers and other equipment, which are used to collect and transmit the basic information, health status, historical entry and exit mode data and environmental parameters of the greenhouse.

[0024] It should also be understood that an object can refer to a person or a material, wherein the basic information of a person includes name, identity, purpose of entry or exit, etc. The basic information of a material includes material identification, physical properties, supplier information, production information, use information, etc. Among them, the material identification can be the material name (such as seeds, fertilizers, pesticides, etc.), material number, batch number, etc. Physical properties include quantity (such as weight, volume or number of pieces), specifications (such as size, model, etc.), and packaging methods (such as bagging, boxing, etc.). Supplier information includes supplier name, supplier address and contact information. Production information includes production date, expiration date, production location, etc. The health status of a person includes physical health status, whether there is a history of disease, whether it carries pathogens, etc. The health status of a material includes quality information, which includes quality certificates, inspection reports, ingredient information, etc. Historical entry and exit pattern data refers to the entry and exit records of an object over a period of time in the past, including entry and exit time, frequency, path, etc. The environmental parameters of a greenhouse can refer to environmental conditions such as temperature, humidity, light intensity, and carbon dioxide concentration in the greenhouse, which are used to evaluate the impact of the environment on the object and ensure the safety and effectiveness of entry and exit management.

[0025] S12. Use a pre-trained prediction model to perform multi-level processing on historical entry and exit pattern data, predict the entry and exit demand information of the object within the target time period based on the multi-level processing results, and determine the initial disinfection process of the object based on the entry and exit demand information; the multi-levels include the potential distribution learning level and the graph structure modeling level.

[0026] It should be understood that the pre-trained prediction model provided in the embodiment of the present application, on the one hand, processes the historical entry and exit pattern data at the potential distribution learning level, learns the potential distribution of the data, and captures the intrinsic structure and pattern of the data; on the other hand, it processes the historical entry and exit pattern data at the graph structure modeling level to capture the relationship and mutual influence between objects. Through the above multi-level processing, this embodiment can enhance the prediction model's ability to handle complex relationships and improve the accuracy of the prediction.

[0027] It should also be understood that the initial disinfection process is a set of disinfection procedures determined after preliminary processing of the object's entry and exit demand information within the target time period based on the prediction model. This set of disinfection procedures is formulated based on the object's historical entry and exit pattern data and the multi-level processing results of the prediction model to provide a basic disinfection plan as the basis for subsequent dynamic adjustments.

[0028] Among them, the initial disinfection process includes determining the disinfection method, determining the disinfection time, selecting the disinfectant, determining the disinfection area, designing the operation steps, etc. Among them, the disinfection method may include chemical disinfection, physical disinfection (such as ultraviolet irradiation), biological disinfection and other methods. The disinfection time is the specific time for disinfection to ensure the continuity and effectiveness of the disinfection process. Disinfectant selection refers to selecting a suitable disinfectant according to the type and characteristics of the object. The disinfection area can be a specific area designated for disinfection, such as an entrance, an exit, a specific disinfection room, etc. The operation step design is used to provide detailed operation steps, including preparations before disinfection, precautions during the disinfection process, and post-disinfection treatment. The embodiment of the present application determines the initial disinfection process in advance through a predictive model, which can reduce the time for temporary decision-making on site and improve management efficiency.

[0029] S13. According to the basic information and health status of the object and the environmental parameters of the greenhouse, the initial disinfection process is dynamically adjusted to obtain the target disinfection process.

[0030] Among them, the target disinfection process is the final disinfection process that is dynamically adjusted according to the basic information, health status and environmental parameters of the greenhouse of the object. It can ensure that each object can receive appropriate disinfection treatment and improve the flexibility and safety of management.

[0031] S14, executing the target disinfection process to obtain the object disinfection result, and determining whether to allow the object to enter or leave the greenhouse according to the object disinfection result.

[0032] Among them, the object disinfection result is the disinfection status of the object after executing the target disinfection process, including whether the disinfection standard is met, whether there are residual pathogens, etc. The object disinfection result is used to evaluate the disinfection effect and decide whether to allow the object to enter or leave the greenhouse.

[0033] For example, there is a subject who needs to enter a greenhouse. According to the processing results of the prediction model, the initial disinfection process preliminarily determined may be as follows: Disinfection method: chemical disinfection; Disinfection time: 5 minutes; Disinfectant selection: 0.1% sodium hypochlorite solution; Disinfection area: entrance disinfection room. Operation steps: Preparation before disinfection: Ensure that the disinfection room is well ventilated, and prepare disinfectants and disinfection equipment. Disinfection process: The subject enters the disinfection room. Spray 0.1% sodium hypochlorite solution to ensure that the subject is covered all over. After the spraying is completed, the subject stays in the disinfection room for 5 minutes. Post-disinfection processing: The subject leaves the disinfection room and enters the greenhouse.

[0034] Although the initial disinfection process is preliminarily determined, it is not fixed. According to the basic information of the object, the health status and the environmental parameters of the greenhouse, this embodiment can dynamically adjust the initial disinfection process to ensure that each object can receive the most appropriate disinfection treatment. For example: when the health status of the object is not good, increase the disinfection time or use a stronger disinfectant. Alternatively, adjust the concentration of the disinfectant or the disinfection time according to changes in temperature and humidity. Alternatively, when new pathogens are discovered, immediately adjust the disinfection process and take more stringent disinfection measures. In this way, the initial disinfection process not only provides basic disinfection guarantees, but also has the ability to flexibly adjust to ensure the safety and effectiveness of greenhouse entry and exit management.

[0035] In the embodiment of the present application, the pre-trained prediction model can more comprehensively capture the complex relationship between the historical entry and exit patterns of the object and the environmental parameters through multi-level processing (latent distribution learning level and graph structure modeling level), thereby improving the accuracy of the prediction. Based on accurate prediction results, the present embodiment can dynamically adjust the disinfection process and other management measures to ensure that each object can be properly treated and improve the flexibility and efficiency of management. In addition, accurate predictions and flexible management measures help to timely discover and deal with potential risk points, reduce the possibility of pest transmission and cross-infection, and improve the overall safety and sanitation level of the greenhouse. Efficient entry and exit management can reduce unnecessary waiting time and operation steps and improve production efficiency. For example, a fast and accurate disinfection process can speed up the circulation of personnel and materials and reduce production interruptions. Through precise entry and exit management, the stability of environmental parameters such as temperature and humidity in the greenhouse can be maintained to provide plants with optimal growth conditions.

[0036] In some optional embodiments, S12, using a pre-trained prediction model to perform multi-level processing on the historical entry and exit pattern data, and predicting the entry and exit demand information of the object in the target time period based on the multi-level processing results, including: Step 121: Use a pre-trained variational autoencoder to process the historical input / output pattern data at the potential distribution learning level to obtain the object potential features; use a pre-trained graph neural network to process the historical input / output pattern data at the graph structure modeling level to obtain the object graph features.

[0037] Among them, the variational autoencoder is a generative model that combines the principles of autoencoders and Bayesian inference. By learning the latent distribution of data, it can maintain the diversity and authenticity of generated samples while generating new samples. The specific steps are as follows: Collect historical in-out pattern data and organize them into a format suitable for input encoder. Input historical in-out pattern data to the encoder to obtain the mean and variance of the latent variables. Use reparameterization techniques to generate latent variables. Input the latent variables into the decoder to generate reconstructed data. By optimizing the loss function, the reconstructed data is made as close to the original data as possible, while ensuring that the distribution of the latent variables is close to the standard normal distribution. Use the encoder to map the historical in-out pattern data to the latent space to obtain the object's latent features. The object's latent features are features obtained after processing at the latent distribution learning level, reflecting the potential laws of the object's in-out pattern.

[0038] Graph neural network is a deep learning model specially designed for processing graph structure data. In greenhouse entry and exit management, graph neural network can be used to process the historical entry and exit pattern data of objects, capture the relationship and mutual influence between objects, and thus obtain object graph features. The specific steps are as follows: (1) Construct graph structure: Each object is a node, and the node features include the basic information and health status of the object. Edges are the relationship between objects, such as the frequency of simultaneous entry and exit, path similarity, etc. (2) Node representation initialization: The initial representation of each node can be the basic information, health status, historical entry and exit pattern data of the object. (3) Message passing: Collect information from adjacent nodes, such as information about objects that enter and exit at the same time. Update the representation of the target node based on the collected information to capture the relationship and mutual influence between objects. (4) Multi-layer stacking: Through multi-layer message passing and update operations, the graph neural network gradually deepens its understanding of the relationship between objects. (5) Output layer: Through the node representation of the last layer, generate object graph features to reflect the complex relationship and pattern between objects.

[0039] For example, node A: basic information (staff member A), health status (good), historical entry and exit pattern data (enter at 9 am and leave at 5 pm every day). Node B: basic information (material B), health status (good), historical entry and exit pattern data (enter at 9 am every Monday and leave at 5 pm every Friday). Collect information about node B, the neighbor of node A, and update the representation of node A. Collect information about node A, the neighbor of node B, and update the representation of node B. Through multiple layers of message passing and update operations, the model gradually deepens its understanding of the relationship between nodes A and B. The output layer generates graph features of nodes A and B, reflecting the complex relationship and pattern between them.

[0040] Step 122: Use the target attention mechanism to perform weighted fusion processing on the object potential features and the object graph features to obtain weighted comprehensive features; the target attention mechanism is a multi-head attention mechanism that introduces residual connections. Through the multi-head attention mechanism, the prediction model can focus on different parts of different features, improve the effect of feature fusion, and thus improve the prediction accuracy of the prediction model. Residual connections help alleviate the gradient vanishing problem in the prediction model.

[0041] Step 123: According to the weighted comprehensive features, a target prediction algorithm is used to predict the entry and exit demand information of the object within the target time period. The target prediction algorithm includes an adaptive reinforcement learning algorithm, a deep reinforcement learning algorithm and a multi-layer perceptron.

[0042] The adaptive reinforcement learning algorithm is a machine learning algorithm that improves the prediction ability of the prediction model through continuous iteration and optimization. The adaptive reinforcement learning algorithm gradually improves the accuracy of the prediction by gradually increasing the complexity of the prediction model. The deep reinforcement learning algorithm improves the prediction and decision-making ability of the prediction model in complex environments by learning the optimal strategy. The multi-layer perceptron learns the complex mapping relationship between input features and outputs through multiple layers of nonlinear transformations, thereby improving the expressive power of the prediction model.

[0043] The embodiment of the present application is based on multi-level processing results, combined with a target attention mechanism and a target prediction algorithm, to improve the prediction accuracy of input and output demand information, thereby minimizing the deviation between the initial disinfection process and the target disinfection process as much as possible, and improving the accuracy of the initial disinfection process.

[0044] In the above embodiment, the multi-head attention mechanism reduces the risk of overfitting that may exist in a single attention head by focusing on the input data from multiple perspectives. Even if some heads perform poorly, other heads can provide effective information compensation, thereby enhancing the overall robustness of the model. Accordingly, step 122 uses the target attention mechanism to perform weighted fusion processing on the object potential features and the object graph features to obtain weighted comprehensive features, including: Step a1: normalize the object potential features and the object graph features respectively using a normalization method to obtain a standardized potential feature and a standardized graph feature. The normalization method can eliminate the dimensional differences between features, make the feature data have the same scale, and improve the performance of the prediction model.

[0045] Step a2, using the multi-head linear transformation method in the multi-head attention mechanism to perform linear transformation on the standardized latent features and the standardized graph features, respectively, to obtain multiple groups of first query vectors, first key vectors, and first value vectors of the standardized latent features, as well as multiple groups of second query vectors, second key vectors, and second value vectors of the standardized graph features. In this embodiment, the multi-head linear transformation method maps features to different subspaces (or heads) through linear transformation, prepares for subsequent attention score calculation, and improves the expressiveness of the model.

[0046] Step a3: Use a multi-head attention score calculation method to perform dot product calculations on multiple groups of first query vectors and first key vectors of standardized latent features, and multiple groups of second query vectors and second key vectors of standardized graph features, to obtain multiple first attention score matrices of standardized latent features and multiple second attention score matrices of standardized graph features.

[0047] For example, The first query vector corresponding to the attention head is ;No. The first key vector corresponding to the attention head .in, To standardize the latent features The original representation of the first query vector, is the weight matrix of the first query vector. To standardize the latent features The original representation of the first key vector, is the weight matrix of the first key vector.

[0048] Similarly, The second query vector corresponding to the attention head is ;No. The second key vector corresponding to the attention head .in, The first The original representation of the second query vector, is the weight matrix of the second query vector. The first The original representation of the second key vector, is the weight matrix of the second key vector.

[0049] First attention score in the first attention score matrix for: ; in, yes The transpose of is the dimension of the first bond vector, is a positional encoding. is a learnable bias term. is a learnable scalar. is the weight matrix. is the concatenation of the first query vector and the first key vector.

[0050] Second attention score in the second attention score matrix for: ; in, yes The transpose of is the dimension of the second key vector. is a positional encoding. is a learnable bias term. is a learnable scalar. is the weight matrix. is the concatenation of the second query vector and the second key vector.

[0051] This embodiment can quantify the similarity between different features through attention score calculation, provide a basis for subsequent weight allocation, and improve the accuracy of feature fusion.

[0052] Step a4: Use a multi-head scaling method to scale multiple first attention score matrices and multiple second attention score matrices respectively to obtain multiple scaled first attention score matrices and multiple scaled second attention score matrices.

[0053] In step a4, the multi-head scaling method uses the following formula: ; ; in, is the scaled first attention score in the scaled first attention score matrix, is the second attention score after scaling in the scaled second attention score matrix; scaling factor and They are respectively used to prevent the corresponding first attention score and second attention score from being too large, thereby improving the numerical stability of the model.

[0054] In this embodiment, multi-head scaling can prevent the dot product result from being too large and improve the numerical stability of the model.

[0055] Step a5, use a multi-head softmax function to normalize multiple scaled first attention score matrices and multiple scaled second attention score matrices respectively, and obtain multiple first attention weight matrices and multiple second attention weight matrices. The multi-head softmax function converts the attention score into a probability distribution, indicating the importance of different features, and improving the effect of feature fusion. In addition, both the first attention weight matrix and the second attention weight matrix can be dynamically generated to adjust the importance of each feature according to the input data. This dynamic adjustment mechanism enables the model to flexibly adjust the importance of each feature under different tasks and different input conditions, thereby improving the adaptability and flexibility of the model.

[0056] It should be understood that the multi-head softmax function can be formulated as follows: ; ; in, is the first attention score after scaling, is the second attention score after scaling, and are the first attention weight in the first attention weight matrix and the second attention weight in the second attention weight matrix, respectively.

[0057] Step a6: weight the first value vector using a multi-head weighted processing method according to multiple first attention weight matrices to obtain multiple weighted potential features; weight the second value vector using a multi-head weighted processing method according to multiple second attention weight matrices to obtain multiple weighted graph features. The embodiment of the present application dynamically adjusts the importance of different features through weighted summation, improves the effect of feature fusion, and enhances the expressiveness of the model.

[0058] Step a7: Generate multi-head weighted latent features and multi-head weighted graph features respectively according to the multiple weighted latent features and the multiple weighted graph features. The embodiment of the present application integrates multiple weighted features together to form a richer feature representation and improve the expression ability of the model.

[0059] Step a8, using an adaptive residual connection method to perform residual connection on the multi-head weighted latent features and the standardized latent features to obtain target weighted latent features, and performing residual connection on the multi-head weighted graph features and the standardized graph features to obtain target weighted graph features. In an embodiment of the present application, the adaptive residual connection method retains the information of the original features such as the standardized latent features and the standardized graph features, prevents the gradient from disappearing or exploding, and improves the stability and performance of the model.

[0060] Step a9: Use a feature concatenation method to fuse the target weighted latent features and the target weighted graph features to obtain a weighted comprehensive feature.

[0061] It can be seen from the above description that the embodiment of the present application can generate richer and more accurate feature representations through normalization, multi-head linear transformation, attention score calculation, multi-head scaling and multi-head softmax function, and improve the expressiveness of the model. Through the multi-head attention mechanism and residual connection, the model can capture the complex relationship between different features, dynamically adjust the importance of features, improve the effect of feature fusion, and thus improve the accuracy of prediction. In the multi-head weighted processing process, each head will dynamically generate attention weights based on the input data, and these weights are used to weighted sum the features. This dynamic adjustment mechanism enables the model to flexibly adjust the importance of each feature under different tasks and different input conditions, improving the adaptability and flexibility of the model.

[0062] In the above embodiment, step a6, weighting the first value vector using a multi-head weighted processing method according to multiple first attention weight matrices to obtain multiple weighted potential features, includes: step a61, using multiple first attention weight matrices to perform dot multiplication operations on the first value vector respectively, and introducing position encoding to correct each dot multiplication result to obtain multiple corrected first potential features. Step a62, nonlinearly transforming the corrected first potential features through the ReLU activation function to obtain multiple weighted potential features.

[0063] Among them, the weighted processing can be adopted as follows: ; in, is the dot product result, is the first attention weight in the first attention weight matrix, For the The original representation of the first value vector.

[0064] Introducing positional encoding Correct each point product result: ; Among them, the left side of the equal sign is the first latent feature after correction; Nonlinear transformation is performed through the ReLU activation function: ; in, is the weighted latent feature.

[0065] Similarly, in step s6, the second value vector is weighted according to the multiple second attention weight matrices using a multi-head weighted processing method to obtain multiple weighted graph features, including: step a63, using multiple second attention weight matrices to perform dot multiplication operations on the second value vector respectively, and introducing position encoding to correct each dot multiplication result to obtain multiple corrected second potential features. Step a64, the corrected second potential features are nonlinearly transformed through the ReLU activation function to obtain multiple weighted graph features.

[0066] Among them, the weighted processing can be adopted as follows: ; in, is the dot product result, is the first attention weight in the first attention weight matrix For the The original representation of the second value vector.

[0067] Introducing positional encoding Correct each point product result: ; Among them, the left side of the equal sign is the modified second latent characteristic; Nonlinear transformation is performed through the ReLU activation function: ; in, is a weighted graph feature.

[0068] In the above process, the dot product operation can weight the first value vector according to the first attention weight in the first attention weight matrix, and weight the second value vector according to the second attention weight in the second attention weight matrix, highlighting important features and suppressing unimportant features, so that the model can flexibly focus on the key parts of the input data; position encoding is a method for encoding each position information in a sequence, which helps the model understand the position information of the feature and improves the model's ability to process data. Through the dot product operation and the ReLU activation function, the model can learn more complex feature representations. Each head can learn a specific type of feature, and the final output is a combination of these features. This combination enables the model to better represent the complex structure of the input data and improve the model's expressiveness.

[0069] In the above embodiment, step 123, according to the weighted comprehensive features, a target prediction algorithm is used to predict the entry and exit demand information of the object within the target time period, and the target prediction algorithm includes an adaptive reinforcement learning algorithm, a deep reinforcement learning algorithm and a multi-layer perceptron, including: step b1, using an adaptive reinforcement learning algorithm to perform feature selection and enhancement processing on the weighted comprehensive features to obtain optimized features. Step b2, using a deep reinforcement learning algorithm to perform strategy optimization processing on the optimized features to obtain an optimized strategy. Step b3, using a multi-layer perceptron to predict the optimization strategy and obtain the entry and exit demand information of the object within the target time period.

[0070] In an embodiment of the present application, the adaptive reinforcement learning algorithm gradually adjusts the importance of features through an iterative process, selects the most relevant features and enhances their weights, and removes irrelevant or redundant features. This dynamic adjustment enables the model to flexibly select and enhance features according to changes in input data. The adaptive reinforcement learning algorithm continuously adjusts the weights and parameters of the model so that it can better cope with changes in input data. This adaptive adjustment improves the robustness of the model. The deep reinforcement learning algorithm learns the optimal decision-making strategy through interaction with the environment to maximize cumulative rewards. This strategy optimization enables the model to select the optimal action based on the current state and environment. The deep reinforcement learning algorithm can handle complex input-output relationships and capture hidden patterns and laws. Through the learning ability of deep neural networks, the model can understand and process high-dimensional, nonlinear data. The multi-layer perceptron learns the complex mapping relationship between input features and outputs through multi-layer nonlinear transformations. This nonlinear transformation enables the model to capture the nonlinear relationship in the input data and improve the accuracy of the prediction. The multi-layer perceptron can quickly predict on large-scale data sets. Through parallel computing and optimized algorithms, the model can achieve real-time prediction. Therefore, by selecting and enhancing features through the adaptive reinforcement learning algorithm, optimizing strategies through the deep reinforcement learning algorithm, and performing prediction processing through the multi-layer perceptron, the prediction model can more accurately predict the entry and exit demand information of the object within the target time period.

[0071] In some optional embodiments, S13, dynamically adjust the initial disinfection process according to the basic information, health status and environmental parameters of the greenhouse of the object to obtain the target disinfection process, including: step 131, use the decision tree algorithm to preliminarily classify the basic information, health status and environmental parameters of the greenhouse of the object to obtain a preliminary classification result. Step 132, use the support vector machine to refine the preliminary classification result to obtain a refined processing result. Step 133, use the random forest algorithm to comprehensively evaluate the basic information, health status and environmental parameters of the greenhouse of the object to obtain a comprehensive evaluation result. Step 134, use a convolutional neural network to extract deep features of the historical entry and exit pattern data of the object and the environmental parameters of the greenhouse to obtain a deep feature vector. Step 135, use an integrated learning method to comprehensively process the preliminary classification results, the refined processing results, the comprehensive evaluation results and the deep feature vector to obtain the target disinfection process.

[0072] In an embodiment of the present application, the decision tree algorithm can quickly perform preliminary classification of the basic information, health status and environmental parameters of the greenhouse of the object, and provide a quick response. The preliminary classification results are refined by a support vector machine to improve the accuracy and reliability of the classification. Random forests can effectively process high-dimensional data such as the basic information, health status and environmental parameters of the greenhouse of the object, and improve the accuracy of the evaluation. The historical entry and exit pattern data of the object and the environmental parameters of the greenhouse are deeply extracted through a convolutional neural network to capture the deep information in the data. The preliminary classification results, refined processing results, comprehensive evaluation results and deep feature vectors are comprehensively processed by an integrated learning method to improve the accuracy of the target disinfection process prediction.

[0073] In the above embodiment, step 134, using a convolutional neural network to perform deep feature extraction on the historical entry and exit pattern data of the object and the environmental parameters of the greenhouse to obtain a deep feature vector, includes: step c1, using the first convolutional layer to perform local feature extraction on the historical entry and exit pattern data of the object and the environmental parameters of the greenhouse to obtain a local feature map. Step c2, using the first pooling layer to downsample the local feature map to obtain a downsampled feature map. Step c3, using a combination of multiple second convolutional layers and second pooling layers to perform multi-layer feature extraction on the downsampled feature map to obtain a multi-level feature map. Step c4, using a fully connected layer to integrate the features of the multi-level feature map to obtain a feature vector. Step c5, using an activation function to perform nonlinear transformation on the feature vector, and using a normalization method to standardize the feature vector after nonlinear transformation to obtain a deep feature vector.

[0074] In this embodiment, the first convolutional layer can extract local features of input data such as the historical entry and exit pattern data of the object and the environmental parameters of the greenhouse by means of a sliding window, and can capture local patterns and structures in the input data. The first pooling layer reduces the spatial size of the local feature map through a downsampling operation, reduces the computational complexity, and retains important features. Through the combination of multiple second convolutional layers and second pooling layers, the model can extract more advanced features layer by layer, capturing the deep patterns in the historical entry and exit pattern data of the object and the environmental parameters of the greenhouse. The fully connected layer can capture global information in multi-level feature maps and improve the global understanding ability of the model. The activation function introduces nonlinear factors, and the normalization method standardizes the feature vectors to the same scale to improve the stability of the model.

[0075] Figure 2 A schematic diagram of a greenhouse entry and exit management system provided in an embodiment of the present application is shown in FIG. Figure 2 As shown, the system includes: The acquisition module 21 is used to obtain the basic information, health status, historical entry and exit mode data and environmental parameters of the objects to enter or leave the greenhouse within the target time period through the intelligent corridor system connected between the greenhouses.

[0076] The processing prediction module 22 is used to use a pre-trained prediction model to perform multi-level processing on the historical entry and exit pattern data, predict the entry and exit demand information of the object within the target time period based on the multi-level processing results, and determine the initial disinfection process of the object based on the entry and exit demand information; the multiple levels include the potential distribution learning level and the graph structure modeling level.

[0077] The dynamic adjustment module 23 is used to dynamically adjust the initial disinfection process according to the basic information and health status of the object and the environmental parameters of the greenhouse to obtain the target disinfection process.

[0078] The execution determination module 24 is used to execute the target disinfection process to obtain the object disinfection result, and determine whether to allow the object to enter or leave the greenhouse according to the object disinfection result.

[0079] Figure 2 The greenhouse access management system can be implemented Figure 1 The implementation principle and technical effect of the greenhouse entry and exit management method described in the illustrated embodiment will not be repeated. The specific manner in which each module and unit performs operations in the greenhouse entry and exit management system in the above embodiment has been described in detail in the embodiment of the method, and will not be elaborated here.

[0080] In one possible design, Figure 2 The greenhouse entry and exit management system of the illustrated embodiment can be implemented as a computing device, such as Figure 3 As shown, the computing device may include a storage component 31 and a processing component 32 .

[0081] The storage component 31 stores one or more computer instructions, wherein the one or more computer instructions are called and executed by the processing component 32 .

[0082] The processing component 32 is used to: obtain the basic information, health status, historical entry and exit pattern data and environmental parameters of the objects to be entered or left the greenhouse within the target time period through the intelligent corridor system connected between the greenhouses; use a pre-trained prediction model to perform multi-level processing on the historical entry and exit pattern data, predict the entry and exit demand information of the object within the target time period based on the multi-level processing results, and determine the initial disinfection process of the object according to the entry and exit demand information; the multi-levels include the potential distribution learning level and the graph structure modeling level; dynamically adjust the initial disinfection process according to the basic information, health status and environmental parameters of the object to obtain the target disinfection process; execute the target disinfection process to obtain the object disinfection result, and determine whether to allow the object to enter or leave the greenhouse according to the object disinfection result.

[0083] The processing component 32 may include one or more processors to execute computer instructions to complete all or part of the steps in the above method. Of course, the processing component may also be implemented by one or more application-specific integrated circuits (ASIC), digital signal processors (DSP), digital signal processing devices (DSPD), programmable logic devices (PLD), field programmable gate arrays (FPGA), controllers, microcontrollers, microprocessors or other electronic components to perform the above method.

[0084] The storage component 31 is configured to store various types of data to support operations at the terminal. The storage component can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as random access memory (RAM), static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, magnetic disk or optical disk.

[0085] Of course, the computing device may also include other components, such as input / output interfaces, display components, communication components, etc.

[0086] The input / output interface provides an interface between the processing component and the peripheral interface module, which may be an output device, an input device, etc.

[0087] The communication component is configured to facilitate, among other things, wired or wireless communications between the computing device and other devices.

[0088] Among them, the computing device can be a physical device or an elastic computing host provided by a cloud computing platform, etc. In this case, the computing device can refer to a cloud server, and the above-mentioned processing components, storage components, etc. can be basic server resources rented or purchased from the cloud computing platform.

[0089] The present application also provides a computer storage medium storing a computer program, wherein the computer program can achieve the above-mentioned Figure 1 The greenhouse entry and exit management method of the illustrated embodiment.

[0090] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the systems and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0091] The system embodiment described above is merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of this embodiment. Those of ordinary skill in the art may understand and implement it without creative work.

[0092] Through the description of the above implementation methods, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus a necessary general hardware platform, and of course, can also be implemented by hardware. Based on this understanding, the above technical solution is essentially 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, a disk, an optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0093] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit it. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A method for managing entry and exit of a greenhouse, characterized in that: include: The basic information, health status, historical entry and exit pattern data and environmental parameters of the objects to be entered or left in the greenhouse within the target time period are obtained through the intelligent corridor system connected between the greenhouses; The historical entry and exit pattern data are processed at multiple levels using a pre-trained prediction model, and the entry and exit demand information of the object within the target time period is predicted based on the multi-level processing results, and the initial disinfection process of the object is determined according to the entry and exit demand information; the multi-levels include a potential distribution learning level and a graph structure modeling level; According to the basic information and health status of the object and the environmental parameters of the greenhouse, the initial disinfection process is dynamically adjusted to obtain a target disinfection process; Execute the target disinfection process to obtain the object disinfection result, and determine whether to allow the object to enter or leave the greenhouse according to the object disinfection result.

2. The method according to claim 1, characterized in that The pre-trained prediction model is used to perform multi-level processing on the historical entry and exit pattern data, and the entry and exit demand information of the object in the target time period is predicted based on the multi-level processing results, including: The historical entry-exit pattern data is processed at the potential distribution learning level using a pre-trained variational autoencoder to obtain object potential features; the historical entry-exit pattern data is processed at the graph structure modeling level using a pre-trained graph neural network to obtain object graph features; A target attention mechanism is used to perform weighted fusion processing on the object potential features and the object graph features to obtain weighted comprehensive features; the target attention mechanism is a multi-head attention mechanism that introduces residual connections; According to the weighted comprehensive features, a target prediction algorithm is adopted to predict the entry and exit demand information of the object within the target time period. The target prediction algorithm includes an adaptive reinforcement learning algorithm, a deep reinforcement learning algorithm and a multi-layer perceptron.

3. The method according to claim 2, characterized in that The target attention mechanism is used to perform weighted fusion processing on the object potential features and the object graph features to obtain weighted comprehensive features, including: Using a normalization method to perform normalization processing on the object potential feature and the object graph feature respectively, to obtain a standardized potential feature and a standardized graph feature; The standardized latent features and the standardized graph features are respectively linearly transformed by using a multi-head linear transformation method in a multi-head attention mechanism to obtain multiple groups of first query vectors, first key vectors, and first value vectors of the standardized latent features, and multiple groups of second query vectors, second key vectors, and second value vectors of the standardized graph features; A multi-head attention score calculation method is used to perform dot product calculations on multiple groups of first query vectors and first key vectors of the standardized latent features, and multiple groups of second query vectors and second key vectors of the standardized graph features, to obtain multiple first attention score matrices of the standardized latent features and multiple second attention score matrices of the standardized graph features; Using a multi-head scaling method to scale the multiple first attention score matrices and the multiple second attention score matrices respectively, to obtain multiple scaled first attention score matrices and multiple scaled second attention score matrices; Using a multi-head softmax function to normalize the multiple scaled first attention score matrices and the multiple scaled second attention score matrices, respectively, to obtain multiple first attention weight matrices and multiple second attention weight matrices; According to the multiple first attention weight matrices, a multi-head weighted processing method is used to perform weighted processing on the first value vector to obtain multiple weighted potential features; according to the multiple second attention weight matrices, a multi-head weighted processing method is used to perform weighted processing on the second value vector to obtain multiple weighted graph features; Generating multi-head weighted latent features and multi-head weighted graph features respectively according to the plurality of weighted latent features and the plurality of weighted graph features; The multi-head weighted latent feature is residually connected with the standardized latent feature by using an adaptive residual connection method to obtain a target weighted latent feature, and the multi-head weighted graph feature is residually connected with the standardized graph feature to obtain a target weighted graph feature; The target weighted latent features and the target weighted graph features are fused by using a feature concatenation method to obtain weighted comprehensive features.

4. The method according to claim 3, characterized in that The first value vector is weighted according to the multiple first attention weight matrices using a multi-head weighted processing method to obtain multiple weighted potential features, including: Using the multiple first attention weight matrices to perform dot multiplication operations on the first value vector respectively, and introducing position coding to correct each dot multiplication result to obtain multiple corrected first potential features; The modified first latent feature is nonlinearly transformed through a ReLU activation function to obtain a plurality of weighted latent features.

5. The method according to claim 2, characterized in that: The target prediction algorithm is used to predict the entry and exit demand information of the object within the target time period according to the weighted comprehensive features. The target prediction algorithm includes an adaptive reinforcement learning algorithm, a deep reinforcement learning algorithm and a multi-layer perceptron, including: Adopting an adaptive reinforcement learning algorithm to perform feature selection and enhancement processing on the weighted comprehensive features to obtain optimized features; Using a deep reinforcement learning algorithm to perform strategy optimization processing on the optimization features to obtain an optimization strategy; A multi-layer perceptron is used to perform prediction processing on the optimization strategy to obtain the entry and exit demand information of the object within the target time period.

6. The method according to claim 1, characterized in that The method of dynamically adjusting the initial disinfection process according to the basic information and health status of the object and the environmental parameters of the greenhouse to obtain the target disinfection process includes: A decision tree algorithm is used to preliminarily classify the basic information, health status and environmental parameters of the greenhouse of the object to obtain a preliminary classification result; Using a support vector machine to refine the preliminary classification result to obtain a refined processing result; A random forest algorithm is used to comprehensively evaluate the basic information, health status and environmental parameters of the greenhouse of the object to obtain a comprehensive evaluation result; Using a convolutional neural network to perform deep feature extraction on the historical entry and exit pattern data of the object and the environmental parameters of the greenhouse to obtain a deep feature vector; The preliminary classification result, the refinement processing result, the comprehensive evaluation result and the deep feature vector are comprehensively processed by an integrated learning method to obtain a target disinfection process.

7. The method according to claim 6, characterized in that The convolutional neural network is used to extract deep features from the historical entry and exit pattern data of the object and the environmental parameters of the greenhouse to obtain a deep feature vector, including: Using the first convolutional layer to extract local features from the historical entry and exit pattern data of the object and the environmental parameters of the greenhouse to obtain a local feature map; Using a first pooling layer to downsample the local feature map to obtain a downsampled feature map; Using a combination of multiple second convolutional layers and second pooling layers to perform multi-layer feature extraction on the downsampled feature map to obtain a multi-level feature map; Using a fully connected layer to integrate the multi-level feature map to obtain a feature vector; An activation function is used to perform nonlinear transformation on the feature vector, and a normalization method is used to normalize the feature vector after the nonlinear transformation to obtain a deep feature vector.

8. A greenhouse entry and exit management system, characterized in that: include: An acquisition module is used to acquire basic information, health status, historical entry and exit mode data, and environmental parameters of the objects to be entered or left the greenhouse within a target time period through an intelligent corridor system connected between the greenhouses; A processing prediction module, used to use a pre-trained prediction model to perform multi-level processing on the historical entry and exit pattern data, predict the entry and exit demand information of the object within the target time period based on the multi-level processing results, and determine the initial disinfection process of the object according to the entry and exit demand information; the multi-levels include a potential distribution learning level and a graph structure modeling level; A dynamic adjustment module, used to dynamically adjust the initial disinfection process according to the basic information and health status of the object and the environmental parameters of the greenhouse to obtain a target disinfection process; The execution determination module is used to execute the target disinfection process to obtain the object disinfection result, and determine whether to allow the object to enter or leave the greenhouse according to the object disinfection result.

9. A computing device, characterized in that It comprises a processing component and a storage component; the storage component stores one or more computer instructions; the one or more computer instructions are used to be called and executed by the processing component to implement a greenhouse entry and exit management method as described in any one of claims 1 to 7.

10. A computer storage medium, characterized in that: A computer program is stored, and when the computer program is executed by a computer, a greenhouse entry and exit management method as described in any one of claims 1 to 7 is implemented.

Images