Intelligent fresh-keeping parameter control method and system for fruit fresh-keeping refrigeration house

Abstract

The invention relates to the technical field of fruit fresh-keeping, and particularly discloses an intelligent fresh-keeping parameter control method and system for a fruit fresh-keeping refrigeration house. Firstly, variety and maturity state information of crisp persimmons is collected, and meanwhile, various fresh-keeping parameters in a storage environment are monitored in real time; the method comprises the following steps of: performing global time sequence correlation feature mining on various fresh-keeping parameters by adopting a deep learning-based data processing technology to obtain global storage environment state features, and meanwhile, performing vectorization processing on variety and maturity state information of the crisp persimmons by adopting a one-hot coding technology, and taking the vectorized information as prior information of fresh-keeping storage of the crisp persimmons, so as to obtain a fresh-keeping storage result of the crisp persimmons. And then, through association and interaction fusion of the two, comprehensive understanding of the fresh-keeping requirements of the crisp persimmons is realized, so that an appropriate fresh-keeping temperature is intelligently recommended. Therefore, more accurate preservation control can be realized by comprehensively considering internal attributes and external preservation environment factors of the crisp persimmons, so that the preservation period of the crisp persimmons is effectively prolonged, and the commodity value of the crisp persimmons is improved.

Description

Technical Field

[0001] The present application relates to the technical field of fruit preservation, and more specifically, to an intelligent preservation parameter control method and system of a fruit preservation cold storage. Background Art

[0002] As consumers' requirements for food quality continue to increase, fruit and vegetable preservation technology has become one of the key areas of concern in modern agriculture and the food industry. As a fruit rich in nutrients and with a unique taste, the market demand for persimmons is growing. However, persimmons are prone to softening and rotting after picking, resulting in a short shelf life, which seriously affects market supply and economic benefits. Therefore, how to effectively extend the shelf life of persimmons and improve their commodity value has become an important topic of current research.

[0003] Traditional methods of persimmon preservation mainly rely on empirical temperature control and simple environmental parameter monitoring, such as using cold storage to maintain a certain low temperature environment to delay the ripening and aging process of the fruit. However, this method often ignores the differences between persimmon varieties and the differences in the sensitivity of fruits at different maturity states to preservation conditions, resulting in uneven preservation effects and unable to meet the market's continued demand for high-quality persimmons.

[0004] Therefore, an optimized intelligent preservation parameter control method for fruit preservation cold storage is expected. Summary of the invention

[0005] In order to solve the above technical problems, the present application is proposed. The embodiment of the present application provides an intelligent preservation parameter control method and system for a fruit preservation cold storage, which first collects the variety and maturity status information of crisp persimmons, and monitors the various preservation parameters in the storage environment in real time, and uses data processing technology based on deep learning to perform global time series correlation feature mining on various preservation parameters to obtain global storage environment status characteristics. At the same time, the variety and maturity status information of crisp persimmons are vectorized by using unique hot encoding technology, and used as prior information for crisp persimmon preservation storage, and then the two are correlated and interactively fused to achieve a comprehensive understanding of the crisp persimmon preservation needs, so as to intelligently recommend suitable preservation temperatures. In this way, by comprehensively considering the intrinsic properties of crisp persimmons and external preservation environment factors, more accurate preservation control can be achieved, thereby effectively extending the shelf life of crisp persimmons and improving their commodity value.

[0006] Accordingly, according to one aspect of the present application, a method for controlling intelligent preservation parameters of a fruit preservation cold storage is provided, comprising: Receiving the variety information and maturity status label of the persimmon input by the user; Acquire a data set of real-time preservation parameters in the warehouse collected by the sensor matrix, wherein the real-time preservation parameters in the warehouse include temperature value, humidity value, oxygen concentration value and carbon dioxide concentration value; Mining the time series correlation feature of the fresh-keeping parameters on the data set of the real-time fresh-keeping parameters in the warehouse to obtain a time series correlation feature vector of the fresh-keeping parameters in the warehouse; Vectorizing the variety information of the persimmon and the maturity status label to obtain a cascade vector of persimmon fresh-keeping storage prior information; Interactively fuse the persimmon fresh-keeping storage prior information cascade vector and the in-store fresh-keeping parameter time series correlation feature vector to obtain a prior information-in-store fresh-keeping parameter significant interactive fusion representation vector; Based on the prior information-storage fresh-keeping parameter significant interaction fusion representation vector, the adjustment strategy of the storage fresh-keeping temperature is determined.

[0007] The "inside the warehouse" refers to the interior of the AI ​​fresh-keeping warehouse.

[0008] In the above-mentioned intelligent preservation parameter control method of the fruit preservation cold storage, the cascade vector of the prior information on the preservation and storage of persimmons and the temporal correlation feature vector of the preservation parameters in the warehouse are interactively fused, including: extracting the implicit correlation features between the cascade vector of the prior information on the preservation and storage of persimmons and the temporal correlation feature vector of the preservation parameters in the warehouse as conditional feature vectors; based on the conditional feature vectors, guiding the cascade vector of the prior information on the preservation and storage of persimmons and the temporal correlation feature vector of the preservation parameters in the warehouse to perform cross-domain interactive fusion based on the attention mechanism to obtain the prior information-in-warehouse preservation parameter significant interactive fusion representation vector.

[0009] In the above-mentioned intelligent preservation parameter control method of the fruit preservation cold storage, the data set of the real-time preservation parameters in the storage is subjected to the preservation parameter time series correlation feature mining to obtain the storage parameter time series correlation feature vector, including: the data set of the real-time preservation parameters in the storage is subjected to data structuring processing according to the time dimension and the parameter sample dimension to obtain the storage parameter time series joint matrix; the storage parameter time series joint matrix is ​​input into the storage parameter time series correlation pattern feature miner based on the converter structure to obtain the storage parameter time series correlation feature vector.

[0010] In the above-mentioned intelligent preservation parameter control method of the fruit preservation cold storage, the variety information of the crisp persimmon and the maturity status label are vectorized to obtain a cascade vector of the crisp persimmon preservation and storage prior information, including: the variety information of the crisp persimmon and the maturity status label are one-hot encoded to obtain a crisp persimmon variety one-hot encoded vector and a maturity status label one-hot encoded vector, and the crisp persimmon variety one-hot encoded vector and the maturity status label one-hot encoded vector are cascaded to obtain the crisp persimmon preservation and storage prior information cascade vector.

[0011] In the above-mentioned intelligent preservation parameter control method of the fruit preservation cold storage, the implicit correlation feature between the persimmon preservation storage prior information cascade vector and the in-warehouse preservation parameter time series correlation feature vector is extracted as a conditional feature vector, including: inputting the persimmon preservation storage prior information cascade vector and the in-warehouse preservation parameter time series correlation feature vector into an implicit correlation feature capture network to obtain a prior information-in-warehouse preservation parameter implicit correlation feature vector; performing feature activation based on the Sigmoid function on the prior information-in-warehouse preservation parameter implicit correlation feature vector to obtain the conditional feature vector.

[0012] In the above-mentioned intelligent preservation parameter control method of the fruit preservation cold storage, based on the conditional feature vector, the persimmon preservation storage prior information cascade vector and the in-warehouse preservation parameter time series association feature vector are guided to perform cross-domain interactive fusion based on the attention mechanism to obtain the prior information-in-warehouse preservation parameter significant interactive fusion representation vector, including: based on the semantic relevance of the persimmon preservation storage prior information cascade vector and the in-warehouse preservation parameter time series association feature vector relative to the conditional feature vector, the persimmon preservation storage prior information cascade vector and the in-warehouse preservation parameter time series association feature vector are feature modulated to obtain the modulated persimmon preservation storage prior information cascade vector and the modulated in-warehouse preservation parameter time series association feature vector; the modulated persimmon preservation storage prior information cascade vector, the modulated in-warehouse preservation parameter time series association feature vector and the conditional feature vector are cross-domain interactively encoded to obtain the prior information-in-warehouse preservation parameter significant interactive fusion representation vector.

[0013] In the above-mentioned intelligent preservation parameter control method for the fruit preservation cold storage, based on the semantic relevance of the crisp persimmon preservation storage prior information cascade vector and the in-warehouse preservation parameter time series association feature vector relative to the conditional feature vector, the crisp persimmon preservation storage prior information cascade vector and the in-warehouse preservation parameter time series association feature vector are feature modulated to obtain the modulated crisp persimmon preservation storage prior information cascade vector and the modulated in-warehouse preservation parameter time series association feature vector, including: calculating the first semantic relevance factor of the crisp persimmon preservation storage prior information cascade vector relative to the conditional feature vector; calculating the second semantic relevance factor of the in-warehouse preservation parameter time series association feature vector for the conditional feature vector; normalizing the first semantic relevance factor and the second semantic relevance factor, and using the normalized first semantic relevance factor and the second semantic relevance factor to weighted modulate the crisp persimmon preservation storage prior information cascade vector and the in-warehouse preservation parameter time series association feature vector to obtain the modulated crisp persimmon preservation storage prior information cascade vector and the modulated in-warehouse preservation parameter time series association feature vector.

[0014] In the above-mentioned intelligent preservation parameter control method of the fruit preservation cold storage, the cascade vector of the prior information on the storage of crisp persimmons after modulation, the temporal correlation feature vector of the storage parameters in the storage after modulation and the conditional feature vector are cross-domain interactively encoded to obtain the prior information-storage storage preservation parameter significant interaction fusion representation vector, including: using the cascade vector of the prior information on the storage of crisp persimmons after modulation as the query vector, the temporal correlation feature vector of the storage parameters in the storage after modulation as the key vector and the conditional feature vector as the value vector, and performing significant guided interaction between features based on the converter structure on the cascade vector of the prior information on the storage of crisp persimmons after modulation, the temporal correlation feature vector of the storage parameters in the storage after modulation and the conditional feature vector to obtain the prior information-storage storage preservation parameter significant interaction fusion representation vector.

[0015] In the above-mentioned intelligent preservation parameter control method for the fruit preservation cold storage, the adjustment strategy of the preservation temperature in the storage is determined based on the prior information-the significant interaction fusion representation vector of the preservation parameters in the storage, including: inputting the prior information-the significant interaction fusion representation vector of the preservation parameters in the storage into the decoder-based preservation parameter optimization module to obtain an optimization result, and the optimization result is the recommended preservation temperature value; based on the optimization result, determining the adjustment strategy of the preservation temperature in the storage.

[0016] According to another aspect of the present application, an intelligent fresh-keeping parameter control system for a fruit fresh-keeping cold storage is provided, which includes: A persimmon fresh-keeping storage prior information acquisition module is used to receive the persimmon variety information and maturity status label input by the user; A fresh-keeping environment data acquisition module is used to acquire a data set of real-time fresh-keeping parameters in the warehouse collected by the sensor matrix, wherein the real-time fresh-keeping parameters in the warehouse include temperature value, humidity value, oxygen concentration value and carbon dioxide concentration value; A fresh-keeping parameter time series association coding module is used to perform fresh-keeping parameter time series association feature mining on the data set of real-time fresh-keeping parameters in the library to obtain a fresh-keeping parameter time series association feature vector in the library; A priori information vectorization processing module, used for performing vectorization processing on the variety information of the crisp persimmon and the maturity status label to obtain a cascade vector of the crisp persimmon fresh-keeping storage priori information; A feature interactive fusion module is used to interactively fuse the persimmon preservation storage prior information cascade vector and the in-store preservation parameter time series correlation feature vector to obtain a prior information-in-store preservation parameter significant interactive fusion representation vector; The fresh-keeping temperature adjustment module is used to determine the adjustment strategy of the fresh-keeping temperature in the warehouse based on the prior information-the significant interaction fusion representation vector of the fresh-keeping parameters in the warehouse.

[0017] Compared with the prior art, the intelligent preservation parameter control method and system of the fruit preservation cold storage provided by the present application first collects the variety and maturity status information of the crisp persimmon, and monitors the various preservation parameters in the storage environment in real time, and uses the data processing technology based on deep learning to perform global time series correlation feature mining on various preservation parameters to obtain the global storage environment status characteristics. At the same time, the unique hot encoding technology is used to vectorize the variety and maturity status information of the crisp persimmon, and use it as the prior information of the crisp persimmon preservation storage, and then through the correlation and interactive fusion of the two, a comprehensive understanding of the crisp persimmon preservation needs is achieved, so as to intelligently recommend the appropriate preservation temperature. In this way, by comprehensively considering the intrinsic properties of the crisp persimmon and the external preservation environment factors, more accurate preservation control can be achieved, thereby effectively extending the shelf life of the crisp persimmon and improving its commodity value. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] By describing the embodiments of the present application in more detail in conjunction with the accompanying drawings, the above and other purposes, features and advantages of the present application will become more apparent. The accompanying drawings are used to provide a further understanding of the embodiments of the present application and constitute a part of the specification. Together with the embodiments of the present application, they are used to explain the present application and do not constitute a limitation of the present application. In the accompanying drawings, the same reference numerals generally represent the same components or steps.

[0019] Figure 1 This is a flow chart of an intelligent preservation parameter control method for a fruit preservation cold storage according to an embodiment of the present application.

[0020] Figure 2This is a data flow diagram of an intelligent preservation parameter control method for a fruit preservation cold storage according to an embodiment of the present application.

[0021] Figure 3 This is a flowchart of step S130 in the intelligent preservation parameter control method of the fruit preservation cold storage according to an embodiment of the present application.

[0022] Figure 4 This is a flowchart of step S150 in the intelligent preservation parameter control method of the fruit preservation cold storage according to an embodiment of the present application.

[0023] Figure 5 This is a block diagram of an intelligent preservation parameter control system for a fruit preservation cold storage according to an embodiment of the present application.

[0024] Figure 6 This is a schematic diagram comparing the ethylene release rate of persimmons in AI fresh-keeping storage and ordinary cold storage.

[0025] Figure 7 This is a schematic diagram comparing the respiration rates of persimmons in AI fresh-keeping storage and ordinary cold storage.

[0026] Figure 8 This is a schematic diagram comparing the hardness of crisp persimmons in AI fresh-keeping storage and ordinary cold storage.

[0027] Fig. 9 This is a schematic diagram comparing the hard fruit rate of crisp persimmons in AI fresh-keeping storage and ordinary cold storage.

[0028] Fig.10 This is a schematic diagram comparing the L value of the peel of crisp persimmons in AI fresh-keeping storage and ordinary cold storage.

[0029] Fig.11 This is a schematic diagram comparing the h-value of the peel of crisp persimmons in the AI ​​fresh-keeping storage and ordinary cold storage.

[0030] Fig.12 This is a schematic diagram comparing the cross-section of persimmon fruits in the AI ​​fresh-keeping storage and ordinary cold storage.

[0031] Fig.13 This is a schematic diagram comparing the chilling injury rate of persimmons in AI fresh-keeping storage and ordinary cold storage.

[0032] Fig.14 This is a schematic diagram comparing the chilling injury index of persimmons in AI fresh-keeping storage and ordinary cold storage.

[0033] Fig.15 This is a schematic diagram comparing the chilling injury symptoms of persimmons in AI fresh-keeping storage and ordinary cold storage.

[0034] Fig.16 This is a schematic diagram comparing the weight loss rate of crisp persimmons in AI fresh-keeping storage and ordinary cold storage.

[0035] Fig.17This is a schematic diagram comparing the decay rates of persimmons in AI fresh-keeping storage and ordinary cold storage.

[0036] The AI ​​fresh-keeping warehouse is an AI fresh-keeping warehouse constructed based on the intelligent fresh-keeping parameter control method of the fruit fresh-keeping cold storage of the present application. DETAILED DESCRIPTION

[0037] Below, the exemplary embodiments according to the present application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application, and it should be understood that the present application is not limited to the exemplary embodiments described here.

[0038] Figure 1 This is a flow chart of an intelligent preservation parameter control method for a fruit preservation cold storage according to an embodiment of the present application. Figure 2 Schematic diagram of data flow of the intelligent fresh-keeping parameter control method of the fruit fresh-keeping cold storage according to the embodiment of the present application. Figure 1 and Figure 2 As shown, according to the embodiment of the present application, the intelligent preservation parameter control method of the fruit preservation cold storage includes the following steps: S110, receiving the variety information and maturity status label of the crisp persimmon input by the user; S120, obtaining a data set of real-time preservation parameters in the warehouse collected by the sensor matrix, wherein the real-time preservation parameters in the warehouse include temperature value, humidity value, oxygen concentration value and carbon dioxide concentration value; S130, mining the preservation parameter time series correlation feature of the real-time preservation parameter data set in the warehouse to obtain the time series correlation feature vector of the preservation parameter in the warehouse; S140, vectorizing the variety information of the crisp persimmon and the maturity status label to obtain a cascade vector of prior information of crisp persimmon preservation storage; S150, interactively fusing the cascade vector of prior information of crisp persimmon preservation storage and the time series correlation feature vector of the preservation parameter in the warehouse to obtain a significant interactive fusion representation vector of prior information-preservation parameters in the warehouse; S160, determining the adjustment strategy of the preservation temperature in the warehouse based on the significant interactive fusion representation vector of prior information-preservation parameters in the warehouse.

[0039] In the above-mentioned intelligent preservation parameter control method of the fruit preservation cold storage, the step S110 receives the variety information and maturity status label of the crisp persimmon input by the user. It should be understood that different varieties of crisp persimmons have different requirements for environmental factors such as temperature and humidity. For example, some varieties may be more cold-resistant, while others require higher temperatures to maintain their freshness. Similarly, crisp persimmons at different stages of maturity may have different requirements for preservation conditions. By understanding the specific varieties and maturity of crisp persimmons, we can better understand the changes that may occur in crisp persimmons under specific storage conditions (such as water loss, sugar conversion, etc.), thereby providing more personalized preservation suggestions to extend the shelf life and reduce losses.

[0040] In order to effectively manage and optimize the preservation process of persimmons, it is first necessary to receive the persimmon variety information and maturity status labels provided by users. This process involves multiple links, including user interface design, data input, verification, transmission and storage. The following is a detailed description of this process: In order to ensure that users can easily and quickly input relevant information about persimmons, it is necessary to design an intuitive and easy-to-operate user interface. The user interface can be a web form, a mobile application, or a touch screen interface on a device. For example, a user can access a dedicated website through a browser and select or fill in the variety name and maturity status of the persimmon in the form. The web form should have the functions of variety selection and maturity selection. The variety selection can be a drop-down menu that lists all known persimmon varieties, and the user selects the corresponding variety from the list; the maturity selection can be a group of radio buttons that list different maturity levels, such as immature, semi-mature, mature, and over-mature, and the user selects the option that best suits the actual situation. In addition, the form should also have a submit button. After the user completes the information input, click the submit button to send the data to the backend system.

[0041] Another way is to develop a mobile app where users can select the variety and maturity of persimmons through a drop-down menu or input box. The mobile app should have similar functions. The variety selection can be a drop-down menu or search box. Users can enter keywords of the variety name, and the system will automatically display matching varieties for users to choose; the maturity selection can be a set of radio buttons or a slider bar, and users can select or adjust the maturity status of persimmons. Similarly, there should be a submit button in the application. After the user completes the information input, click the submit button to send the data to the backend system.

[0042] If the fresh-keeping warehouse is equipped with a touch screen device, a simple operation interface can also be designed directly on the device for users to input information. The touch screen interface should have the functions of variety selection and maturity selection. Variety selection can be a drop-down menu or search box. Users can enter keywords of the variety name, and the system will automatically display matching varieties for users to choose; maturity selection can be a group of radio buttons or a sliding bar. Users select or adjust the maturity status of persimmons. Similarly, there should be a submit button on the touch screen interface. After the user completes the information input, click the submit button to send the data to the backend system.

[0043] The variety information of persimmons is usually a limited list of choices. To ensure accuracy, common persimmon varieties can be collected and organized in advance, and then provided to users for selection. For example, a drop-down menu can be provided to list all known persimmon varieties, and users can select the corresponding variety from the drop-down menu. Common persimmon varieties include European sweet persimmon, Japanese sweet persimmon, four-season persimmon, and Korean sweet persimmon. In addition, a search box can be provided, and users can enter keywords for the variety name, and the system will automatically display matching varieties for users to choose. For example, the user enters "Japan" and the system displays "Japanese sweet persimmon". For some uncommon varieties, users can be allowed to manually enter the variety name, but users need to be prompted to enter the full variety name to ensure accurate identification. For example, users can enter "local sweet persimmon".

[0044] The maturity status of persimmons can be classified according to appearance characteristics or other indicators. It can usually be divided into several levels, such as immature, semi-mature, mature and over-mature. In order to facilitate user selection, a radio button can be provided for each maturity level, and the user selects the option that best suits the actual situation. For example, four radio buttons can be provided, corresponding to immature, semi-mature, mature and over-mature. In addition, a slider can be used to allow users to intuitively select a continuous value between immature and over-mature. For example, the slider ranges from 0 to 100, 0 represents immature and 100 represents over-mature. In addition, pictures of persimmons in different maturity states can be displayed, and users can select the closest state based on the comparison between the real object and the picture. For example, four pictures can be displayed, corresponding to immature, semi-mature, mature and over-mature persimmons.

[0045] In order to ensure that the information entered is accurate, the data entered by the user needs to be verified. First, make sure that the user must enter the variety information and maturity status, otherwise it cannot be submitted. For example, if the user does not select the variety information, the system prompts "Please select the crisp persimmon variety"; if the user does not select the maturity status, the system prompts "Please select the crisp persimmon maturity." Secondly, for the manually entered variety name, check whether it meets the preset format requirements. For example, the variety name must not be empty, and the length should be within a certain range (such as 2-50 characters). Finally, perform a logical check. For example, some varieties may not have a specific maturity status. The system should be able to identify these unreasonable situations and give prompts. For example, if the user selects "Local Sweet Persimmon" but selects "Overripe", the system prompts "Local sweet persimmons usually do not reach an overripe state, please reselect."

[0046] The information input by the user needs to be safely and efficiently transmitted to the backend system for further processing. To this end, a RESTful API interface can be designed to receive user input data from the frontend. For example, the POST request URL can be / api / crisp_persimmon / input, and the request body example is as follows: {"variety":"Japanese persimmon","maturity":"mature"} To ensure the security of data transmission, use the HTTPS protocol to prevent information from being stolen or tampered with. For example, configure the Web server to use an SSL certificate and establish a secure connection between the client and the server. During the data transmission process, if there is a network interruption or other abnormal situation, there should be a corresponding error handling mechanism to ensure that the user can resubmit or get an error prompt. For example, when the network is interrupted, the client prompts "Network connection failed, please check your network settings"; the server returns an error code and error message, and the client displays the corresponding prompt information according to the error code.

[0047] The received user input data needs to be stored in the database for subsequent processing and analysis. To this end, design a suitable database table structure to store the crisp persimmon variety information and maturity status labels entered by the user. For example, you can create a table called crisp_persimmon_inputs with the following fields: id: primary key, auto-increment ID variety: variety name, string type maturity: maturity status, string type input_time: input time, date and time type user_id: user ID, integer (optional) To improve query efficiency, create indexes for key fields. For example, create an index for the variety field to speed up queries by variety; create an index for the input_time field to speed up queries by time range. To prevent data loss, back up the database regularly. For example, automatically back up the database every morning and store the backup files in a safe location, such as a cloud storage service.

[0048] Through the above steps, it can be ensured that the persimmon variety information and maturity status labels entered by the user are accurately and completely transmitted to the system, providing reliable basic data for the subsequent optimization of preservation parameters. Specifically, the user interface design should be simple and intuitive, data input should be accurate, data verification should be strict and reasonable, data storage should be safe and efficient, and data storage should be standardized and reliable. This process not only improves the efficiency of data management, but also provides a solid data foundation for the intelligent preservation of persimmons.

[0049] In the above-mentioned intelligent preservation parameter control method of the fruit preservation cold storage, the step S120 obtains a data set of real-time preservation parameters in the storage collected by the sensor matrix, wherein the real-time preservation parameters in the storage include temperature value, humidity value, oxygen concentration value and carbon dioxide concentration value. It should be understood that the preservation effect of crisp persimmon is directly affected by the storage environment conditions. For example, temperature directly affects the metabolic rate of the fruit, which is the most basic and most critical factor affecting the preservation of crisp persimmons. Too high temperature will accelerate the aging of the fruit; humidity affects the moisture retention of the fruit, and too high or too low humidity may cause the fruit to lose water or mold; oxygen and carbon dioxide concentrations will affect the respiration of the fruit, and appropriate gas concentrations can slow down the respiration rate of the fruit and delay the aging process. Therefore, by comprehensively acquiring various preservation parameter data, the preservation environment of crisp persimmons can be fully understood, which helps to avoid fruit damage or deterioration due to poor environmental conditions.

[0050] In order to effectively manage and optimize the preservation process of persimmons, it is necessary to monitor and collect the environmental parameters in the fresh-keeping warehouse in real time. These parameters include temperature, humidity, oxygen concentration and carbon dioxide concentration. The following are the detailed implementation steps of this process: First, a sensor matrix needs to be deployed in the fresh-keeping warehouse. These sensors are responsible for real-time monitoring of various environmental parameters in the warehouse. The deployment of the sensor matrix should consider location selection, quantity configuration, and type selection. Location selection is crucial, and sensors should be evenly distributed in different locations in the warehouse to ensure the comprehensiveness and representativeness of the data. Especially in large fresh-keeping warehouses, environmental conditions in different areas may vary, so sensors need to be installed at multiple points. Quantity configuration is also very important. According to the size and shape of the warehouse, the number of sensors should be reasonably configured to ensure coverage of a wide area. Common sensor types include temperature sensors, humidity sensors, oxygen concentration sensors, and carbon dioxide concentration sensors. Select high-precision, stable and reliable sensors to ensure the accuracy and reliability of the data.

[0051] Next, in order to efficiently collect and manage the data collected by the sensors, a data acquisition system needs to be designed. The system should have a data acquisition module, a data transmission module, a data storage module, and a data processing module. The data acquisition module is responsible for connecting each sensor by wired or wireless means. Wired connections usually use RS-485 or Modbus protocols, while wireless connections can use wireless communication technologies such as Wi-Fi, Bluetooth, or Zigbee. The data transmission module transmits the collected data to the central server through the network. Common network transmission methods include Ethernet, Wi-Fi, cellular networks (such as 4G / 5G), etc. The data storage module is responsible for receiving and storing data from each sensor. It can use a relational database (such as MySQL, PostgreSQL) or a time series database (such as InfluxDB). The data processing module performs preliminary processing on the collected data, such as filtering, calibration, and anomaly detection, to ensure the accuracy and consistency of the data.

[0052] In order to timely discover and deal with environmental changes in the fresh-keeping warehouse, it is necessary to establish a real-time data monitoring system. The system should have functions such as real-time data display, alarm function, historical data query and data analysis. Through the graphical interface or dashboard, the temperature, humidity, oxygen concentration and carbon dioxide concentration in the warehouse are displayed in real time, so that users can understand the current environmental conditions at a glance. When a parameter exceeds the preset threshold, the system should immediately issue an alarm to notify relevant personnel to take measures. For example, when the temperature exceeds the set upper limit, the system can send an alarm through SMS, email or APP push. The historical data query function allows users to view the trend of environmental parameter changes in any time period, which is convenient for analyzing and optimizing the preservation conditions. The data analysis function performs statistical analysis on the collected data and generates reports and charts to help users better understand the laws of environmental changes in the warehouse.

[0053] For subsequent advanced analysis and modeling, the collected data needs to be structured. The specific steps include data cleaning, time alignment, data formatting and data storage. Data cleaning is to remove invalid or abnormal data points to ensure the quality of data. For example, outliers caused by sensor failure are eliminated. Time alignment is to align the data collected by different sensors according to the timestamp to ensure the consistency of the data of each parameter at the same time point. Data formatting is to convert the raw data into a unified format for subsequent processing. For example, convert the temperature value to Celsius and the humidity value to a percentage. Data storage is to store the processed data in the database and generate a time series joint matrix of the preservation parameters in the library. Each row of the time series joint matrix represents a time point, and each column represents an environmental parameter.

[0054] In order to ensure the security and integrity of data, a series of data security and backup measures need to be taken. Data encryption is the use of encryption technology (such as TLS / SSL) during data transmission to prevent data from being stolen or tampered with. Permission management is to set access rights so that only authorized users can view and operate data. Data backup is to regularly back up the database to prevent data loss. Backup files should be stored in a secure location, such as a cloud storage service or an external hard drive.

[0055] Through the above steps, the real-time environmental parameters in the fresh-keeping warehouse can be effectively acquired and managed, providing reliable data support for the intelligent preservation of crisp persimmons. Specifically, the deployment of the sensor matrix should consider location selection, quantity configuration and type selection to ensure the comprehensiveness and representativeness of the data. The data acquisition system should have data acquisition, transmission, storage and processing functions to ensure efficient data management. The real-time data monitoring system should have data display, alarm, query and analysis functions to timely detect and handle environmental changes. Data structured processing should ensure the quality and consistency of the data and provide a basis for subsequent analysis and modeling. Data security and backup measures should ensure the security and integrity of the data to prevent data loss and leakage.

[0056] This process not only improves the efficiency of data management, but also provides a solid foundation for optimizing the preservation of persimmons. By real-time monitoring and analyzing the environmental parameters in the warehouse, the preservation conditions can be adjusted in time to extend the shelf life of persimmons and improve their market competitiveness and economic benefits. At the same time, this technical solution also provides a reference and reference for the preservation of other perishable agricultural products, promoting the healthy development of the entire preservation industry.

[0057] In the above-mentioned intelligent preservation parameter control method for the fruit preservation cold storage, the step S130 performs preservation parameter time series correlation feature mining on the data set of real-time preservation parameters in the cold storage to obtain a preservation parameter time series correlation feature vector in the cold storage. Figure 3 FIG. 1 is a flow chart of step S130 in the intelligent fresh-keeping parameter control method of the fruit fresh-keeping cold storage according to an embodiment of the present application. Figure 3 As shown, the step S130 includes: S131, structuring the data set of the real-time preservation parameters in the warehouse according to the time dimension and the parameter sample dimension to obtain the in-warehouse preservation parameter time series joint matrix; S132, inputting the in-warehouse preservation parameter time series joint matrix into the in-warehouse preservation parameter time series association pattern feature miner based on the converter structure to obtain the in-warehouse preservation parameter time series association feature vector.

[0058] Specifically, the step S131 performs data structuring processing on the data set of the real-time preservation parameters in the warehouse according to the time dimension and the parameter sample dimension to obtain the time series joint matrix of the preservation parameters in the warehouse. It should be understood that the preservation effect of crisp persimmons is the result of the synergistic effect of multiple preservation parameters, and there is a certain dynamic correlation between the various preservation parameters. For example, the change in oxygen concentration may have a certain correlation with the change in carbon dioxide concentration. Therefore, in order to fully explore the synergistic correlation of various preservation parameters in the time dimension, in the technical solution of the present application, the data set of the real-time preservation parameters in the warehouse is firstly processed by data structuring processing according to the time dimension and the parameter sample dimension to maintain the time series and correlation of the data, and form a time series joint matrix of the preservation parameters in the warehouse, which helps to better capture and utilize the correlation and dependency of various preservation parameters in the time dimension.

[0059] Specifically, in step S132, the in-store fresh-keeping parameter time series joint matrix is ​​input into the in-store fresh-keeping parameter time series association pattern feature miner based on the converter structure to obtain the in-store fresh-keeping parameter time series association feature vector. That is, in order to reveal the interaction and influence between different fresh-keeping parameters, the present application adopts a converter structure model to construct an in-store fresh-keeping parameter time series association pattern feature miner, which is based on the self-attention mechanism and can model and analyze the correlation between different fresh-keeping parameters, capture the long-range dependency relationship in the in-store fresh-keeping parameter time series joint matrix and the interaction effect between multiple parameters, thereby obtaining the in-store fresh-keeping parameter time series association feature vector, which provides a scientific basis for the subsequent formulation of fresh-keeping strategies.

[0060] In the above-mentioned intelligent preservation parameter control method of the fruit preservation cold storage, the step S140 vectorizes the variety information of the crisp persimmon and the maturity status label to obtain a cascade vector of the crisp persimmon preservation storage prior information. In a specific example of the present application, the step S140 includes: one-hot encoding the variety information of the crisp persimmon and the maturity status label to obtain a crisp persimmon variety one-hot encoding vector and a maturity status label one-hot encoding vector, and cascading the crisp persimmon variety one-hot encoding vector and the maturity status label one-hot encoding vector to obtain the crisp persimmon preservation storage prior information cascade vector. That is, considering that the variety information of the crisp persimmon and the maturity status label are categorical variables, both provide important prior information for the preservation control of the crisp persimmon. Therefore, in order to realize the quantitative representation of the crisp persimmon preservation storage prior information, so as to facilitate the multimodal data joint analysis with the preservation parameter characteristics, the present application uses the one-hot encoding technology to vectorize the variety information of the crisp persimmon and the maturity status label. Those skilled in the art should know that one-hot encoding is a method of converting categorical variables into binary vectors, where each category corresponds to a unique binary vector, so that the variety and maturity status information of persimmons can be converted into a machine-recognizable format and accurately reflect the intrinsic properties of persimmons. Furthermore, by cascading the binary one-hot encoding vectors of the variety information of persimmons and the maturity status label, accurate and comprehensive prior information is provided for the preservation control of persimmons.

[0061] In the above-mentioned intelligent preservation parameter control method of the fruit preservation cold storage, the step S150 interactively fuses the persimmon preservation storage prior information cascade vector and the in-store preservation parameter time series correlation feature vector to obtain a prior information-in-store preservation parameter significant interactive fusion representation vector. It should be understood that the persimmon preservation storage prior information cascade vector describes the intrinsic characteristics of the persimmon itself, and the in-store preservation parameter time series correlation feature vector reveals the time series changes of the persimmon storage environment. By fusing the information of the two, the preservation requirements of persimmons in a specific storage environment can be more comprehensively understood, so as to make more accurate preservation decisions. It is worth mentioning that in order to fully capture and utilize the interactive effect between the persimmon preservation storage prior information and the preservation environment parameters, the present application proposes a feature-guided interaction method based on implicit correlation features, which guides the information fusion process between the two by mining the implicit correlation features between the persimmon preservation storage prior information and the preservation environment parameters, and can effectively identify and strengthen the important correlation patterns between the two, thereby enhancing the pertinence and effectiveness of the preservation strategy.

[0062] Figure 4 FIG. 1 is a flow chart of step S150 in the intelligent fresh-keeping parameter control method of the fruit fresh-keeping cold storage according to an embodiment of the present application. Figure 4As shown, the step S150 includes: S151, extracting the implicit correlation features between the cascade vector of the prior information on the preservation and storage of persimmons and the temporal correlation feature vector of the in-warehouse preservation parameters as a conditional feature vector; S152, based on the conditional feature vector, guiding the cascade vector of the prior information on the preservation and storage of persimmons and the temporal correlation feature vector of the in-warehouse preservation parameters to perform cross-domain interactive fusion based on the attention mechanism to obtain the prior information-in-warehouse preservation parameters significant interactive fusion representation vector.

[0063] Specifically, the step S151 includes: inputting the cascade vector of the crisp persimmon preservation storage prior information and the time series correlation feature vector of the in-warehouse preservation parameters into an implicit correlation feature capture network to obtain a priori information-in-warehouse preservation parameter implicit correlation feature vector; and performing feature activation based on the Sigmoid function on the prior information-in-warehouse preservation parameter implicit correlation feature vector to obtain the conditional feature vector.

[0064] In a specific example of the present application, the cascade vector of the prior information on the storage of fresh persimmons and the time-series correlation feature vector of the in-warehouse preservation parameters are input into an implicit correlation feature capture network to obtain a priori information-in-warehouse preservation parameter implicit correlation feature vector, including: fusing the cascade vector of the prior information on the storage of fresh persimmons and the time-series correlation feature vector of the in-warehouse preservation parameters according to the position points and then passing them through a neural network layer based on the tanh function to obtain the priori information-in-warehouse preservation parameter implicit correlation feature vector.

[0065] The process can be expressed as follows:

[0066]

[0067] in, represents the persimmon fresh-keeping storage prior information cascade vector, represents the time series correlation feature vector of the storage parameter, Indicates adding by position point. Represents the prior information-the implicit correlation feature vector of the storage parameters in the library, and Represent the weight matrix and bias parameters of the neural network layer, respectively. represents the hyperbolic tangent function, represents the sigmoid activation function, represents the conditional feature vector.

[0068] That is, it first uses an implicit correlation feature capture network to mine the implicit correlation shared between the persimmon preservation storage prior information cascade vector and the in-library preservation parameter time series correlation feature vector, and uses the Sigmoid function to perform nonlinear transformation on the extracted implicit correlation features, and uses the smoothness characteristics of the Sigmoid function to adjust the feature weights to emphasize key information and suppress irrelevant noise, thereby constructing a conditional feature vector to guide the subsequent feature interaction fusion process.

[0069] Specifically, the step S152 includes: based on the semantic relevance of the cascade vector of the prior information on fresh-keeping storage of crisp persimmons and the time-series association feature vector of the in-warehouse fresh-keeping parameters relative to the conditional feature vector, feature modulation is performed on the cascade vector of the prior information on fresh-keeping storage of crisp persimmons and the time-series association feature vector of the in-warehouse fresh-keeping parameters to obtain the modulated cascade vector of the prior information on fresh-keeping storage of crisp persimmons and the time-series association feature vector of the in-warehouse fresh-keeping parameters after modulation; cross-domain interactive encoding is performed on the cascade vector of the prior information on fresh-keeping storage of crisp persimmons after modulation, the time-series association feature vector of the in-warehouse fresh-keeping parameters after modulation, and the conditional feature vector to obtain the prior information-in-warehouse fresh-keeping parameter significant interactive fusion representation vector.

[0070] In a specific example of the present application, feature modulation is performed on the persimmon fresh-keeping storage prior information cascade vector and the in-store fresh-keeping parameter time series associated feature vector, including: firstly calculating the first semantic correlation factor of the persimmon fresh-keeping storage prior information cascade vector relative to the conditional feature vector, and calculating the second semantic correlation factor of the in-store fresh-keeping parameter time series associated feature vector for the conditional feature vector; The above process can be expressed as follows:

[0071]

[0072] in, , and They represent the first characteristic value, the first characteristic value of the time series correlation characteristic vector of the storage parameter eigenvalue and the conditional eigenvector eigenvalues, is the absolute value symbol, represents the logarithmic function with base 2, represents the exponential function with base e, It represents the length of the cascade vector of the prior information on the fresh-keeping storage of persimmons, and the lengths of the cascade vector of the prior information on the fresh-keeping storage of persimmons, the time series association feature vector of the in-warehouse fresh-keeping parameters and the conditional feature vector are the same.

[0073] That is, calculating the first semantic relevance factor of the cascade vector of the prior information on the preservation and storage of persimmons relative to the conditional feature vector, including: calculating the dot division vector between the cascade vector of the prior information on the preservation and storage of persimmons and the conditional feature vector, and calculating the base-two logarithmic value of the absolute value of each eigenvalue in the dot division vector to obtain a semantic relevance representation vector; calculating the dot product vector between the semantic relevance representation vector and the cascade vector of the prior information on the preservation and storage of persimmons, and calculating an exponential function value with base e and the sum of the eigenvalues ​​of the dot product vector as an exponent to obtain the first semantic relevance factor.

[0074] Then, the first semantic relevance factor and the second semantic relevance factor are normalized, and the normalized first semantic relevance factor and the second semantic relevance factor are used to perform weighted modulation on the crisp persimmon preservation storage prior information cascade vector and the in-warehouse preservation parameter time series association feature vector to obtain the modulated crisp persimmon preservation storage prior information cascade vector and the modulated in-warehouse preservation parameter time series association feature vector.

[0075] The process can be expressed as follows:

[0076]

[0077]

[0078]

[0079] in, and denote the first semantic relevance factor and the second semantic relevance factor respectively, and denote the normalized first semantic relevance factor and the second semantic relevance factor, respectively. and They respectively represent the cascade vector of the prior information on the storage of crisp persimmon after modulation and the temporal correlation feature vector of the storage parameters after modulation.

[0080] That is, by performing semantic relevance evaluation on the cascade vector of the fresh-keeping storage prior information of the crisp persimmons and the time-series correlation feature vector of the in-warehouse fresh-keeping parameters respectively with the conditional feature vector, the importance of the two to the fresh-keeping effect of the crisp persimmons is determined, and the original cascade vector of the fresh-keeping storage prior information of the crisp persimmons and the time-series correlation feature vector of the in-warehouse fresh-keeping parameters are weighted and modulated accordingly, thereby effectively emphasizing the key information in the two and suppressing irrelevant information.

[0081] In a specific example of the present application, the cascade vector of prior information on the storage of fresh-keeping persimmons after modulation, the temporal association feature vector of the storage parameters in the warehouse after modulation, and the conditional feature vector are cross-domain interactively encoded, including: using the cascade vector of prior information on the storage of fresh-keeping persimmons after modulation as a query vector, the temporal association feature vector of the storage parameters in the warehouse after modulation as a key vector and the conditional feature vector as a value vector, and performing significant guided interaction between features based on a converter structure on the cascade vector of prior information on the storage of fresh-keeping persimmons after modulation, the temporal association feature vector of the storage parameters in the warehouse after modulation, and the conditional feature vector to obtain a significant interaction fusion representation vector of the prior information-in-warehouse storage parameters. More specifically, the cascade vector of the prior information on the storage of fresh-keeping persimmons after modulation is multiplied by the transposed vector of the temporal correlation feature vector of the in-warehouse fresh-keeping parameters after modulation and then divided by the square root of the length of the temporal correlation feature vector of the in-warehouse fresh-keeping parameters to obtain a priori information-in-warehouse fresh-keeping parameter query attention score matrix; the prior information-in-warehouse fresh-keeping parameter query attention score matrix is ​​multiplied by the conditional feature vector after passing through a softmax function to obtain the prior information-in-warehouse fresh-keeping parameter significant interaction fusion representation vector.

[0082] The above process is expressed as follows:

[0083] in, represents the transpose of a vector, represents the matrix multiplication operation, is the normalized exponential function, represents the significant interaction fusion representation vector of the prior information and the fresh-keeping parameters in the library, and They represent the cascade vector of prior information on fresh-keeping storage of crisp persimmon after modulation and the temporal correlation feature vector of fresh-keeping parameters in the storage after modulation, Indicates the length of the time series correlation feature vector of the storage parameter, represents the conditional eigenvector.

[0084] That is, in the end, the modulated cascade vector of the fresh-keeping storage prior information of persimmons is used as the query vector, the modulated time-series correlation feature vector of the in-warehouse fresh-keeping parameters is used as the key vector, and the conditional feature vector is used as the value vector. The Transformer architecture is used to guide the interaction between features, and the complex relationship between the fresh-keeping storage prior information of persimmons and the fresh-keeping environment parameters is dynamically captured and utilized based on the self-attention mechanism, so as to achieve deep interactive fusion of cross-modal features and generate a significant interactive fusion representation vector of prior information-in-warehouse fresh-keeping parameters, thereby achieving a comprehensive consideration of the intrinsic properties of persimmons and the storage environment, so as to ensure the provision of customized preservation strategies at different preservation stages.

[0085] In the above-mentioned intelligent preservation parameter control method for the fruit preservation cold storage, the step S160 determines the adjustment strategy of the preservation temperature in the storage based on the prior information-storage preservation parameter significant interaction fusion representation vector. In a specific example of the present application, the step S160 includes: inputting the prior information-storage preservation parameter significant interaction fusion representation vector into the decoder-based preservation parameter optimization module to obtain an optimization result, the optimization result is a recommended preservation temperature value; based on the optimization result, determining the adjustment strategy of the storage preservation temperature.

[0086] That is, the decoder model is used to perform feature learning and decoding prediction on the prior information-storage preservation parameter significant interaction fusion representation vector. The decoder model abstracts and transforms the prior information-storage preservation parameter significant interaction fusion representation vector layer by layer through a multi-layer neural network, and can fully understand the preservation requirements and storage environment status of crisp persimmons, so as to provide a suitable preservation temperature value for crisp persimmons based on the comprehensive consideration of the variety information, maturity status and current preservation environment status of crisp persimmons, so as to achieve the purpose of extending the shelf life of crisp persimmons. Furthermore, the temperature in the warehouse can be adjusted in real time according to the recommendation results of the decoder model to ensure that the crisp persimmons are stored under the best preservation conditions. For example, when the temperature in the warehouse is higher than the recommended value, the power of the refrigeration equipment is increased to slowly reduce the temperature in the warehouse to avoid the risk of fruit rot. On the contrary, if the temperature in the warehouse is detected to be lower than the recommended value, the power of the refrigeration equipment is appropriately reduced to ensure that the crisp persimmons are stored within a suitable temperature range to prevent fruit frostbite.

[0087] In summary, the intelligent preservation parameter control method of the fruit preservation cold storage according to the embodiment of the present application is explained, which first collects the variety and maturity status information of the crisp persimmon, and monitors the various preservation parameters in the storage environment in real time, and uses data processing technology based on deep learning to perform global time series correlation feature mining on various preservation parameters to obtain global storage environment status characteristics. At the same time, the unique hot encoding technology is used to vectorize the variety and maturity status information of the crisp persimmon, and use it as prior information for the preservation storage of the crisp persimmon, and then through the correlation and interactive fusion of the two, a comprehensive understanding of the preservation needs of the crisp persimmon is achieved, so as to intelligently recommend the appropriate preservation temperature. In this way, by comprehensively considering the intrinsic properties of the crisp persimmon and the external preservation environment factors, more accurate preservation control can be achieved, thereby effectively extending the shelf life of the crisp persimmon and improving its commodity value.

[0088] Figure 5 FIG. 1 is a block diagram of an intelligent fresh-keeping parameter control system for a fruit fresh-keeping cold storage according to an embodiment of the present application. Figure 5 As shown, according to the embodiment of the present application, the intelligent preservation parameter control system 100 of the fruit preservation cold storage includes: a persimmon preservation storage prior information acquisition module 110, which is used to receive the variety information and maturity status label of the persimmon input by the user; a preservation environment data acquisition module 120, which is used to obtain a data set of real-time preservation parameters in the storage collected by the sensor matrix, wherein the real-time preservation parameters in the storage include temperature value, humidity value, oxygen concentration value and carbon dioxide concentration value; a preservation parameter time series association coding module 130, which is used to mine the preservation parameter time series association features of the data set of real-time preservation parameters in the storage to obtain the storage environment data set. The storage temperature is adjusted according to the storage temperature, and the storage temperature is adjusted according to the storage temperature. The storage temperature is adjusted according to the storage temperature, and the storage temperature is adjusted according to the storage temperature. The storage temperature is adjusted according to the storage temperature, and the storage temperature is adjusted according to the storage temperature. The storage temperature is adjusted according to the storage temperature, and the storage temperature is adjusted according to the storage temperature.

[0089] Here, those skilled in the art can understand that the specific operations of each module in the intelligent fresh-keeping parameter control system of the above-mentioned fruit fresh-keeping cold storage have been referred to above. Figures 1 to 4 The description of the intelligent preservation parameter control method of the fruit preservation cold storage has been introduced in detail, and therefore, its repeated description will be omitted.

[0090] Example 1 This application adopts the intelligent preservation parameter control method of the above-mentioned fruit preservation cold storage to build a crisp persimmon AI preservation storage, selects the same batch of Yangfeng persimmons with the same size, without mechanical damage and pests and diseases as research materials, takes two groups of Yangfeng persimmons of the same number of fruits and stores them in the AI ​​preservation cold storage and the ordinary cold storage based on fixed preservation parameter control for comparative experiments, observes the Yangfeng persimmons at a fixed time and place, and records the fruit corruption in detail, including whether there are mold spots, softening degree, whether there is peculiar smell, etc. Among them, the fruits stored in the AI ​​preservation storage are the experimental group, and the fruits stored in the ordinary cold storage are the control group.

[0091] 1. Fruit respiration rate and ethylene release rate Six fruits were taken from each group and placed in a sealed 3L glass jar under storage conditions for 24 h. The fruit respiration rate and ethylene release rate were measured using an F-900 Portable Ethylene Analyzer (Felix, USA). Figure 6 As shown, during the storage period, the Yangfeng persimmons in the experimental group showed a significantly lower ethylene release rate, which was significantly different from that in the control group. No ethylene release was detected in either group during the first 14 days of storage. Starting from the 28th day, the ethylene release rate rose rapidly until it reached a peak on the 56th day, then fell slightly and then rose again. Throughout the observation period, the ethylene release rate of the experimental group continued to remain below that of the control group. What is particularly noteworthy is that on the 56th day of storage, the ethylene release rate of the experimental group was significantly lower than that of the control group, which fully verified that the application of the AI ​​preservation technology described in this application can effectively inhibit the ethylene release of Yangfeng persimmons during storage.

[0092] like Figure 7 As shown, during the entire storage period, the respiration rate of the experimental group was always lower than that of the control group. In the first 28 days of storage, although the respiration rates of the Yangfeng persimmons in the experimental and control groups were similar, the respiration rate of the experimental group was still slightly lower than that of the control group. From the 28th day, the respiration rates of both groups increased, but the rate and amplitude of increase in the control group significantly exceeded that of the experimental group. On the 56th day, the respiration rate of the experimental group began to show a downward trend, while that of the control group continued to increase until it began to slow down on the 70th day. Finally, on the 84th day of storage, the difference in the respiration rate of Yangfeng persimmons treated in the two groups became significant. Based on this, it is shown that the above-mentioned AI preservation technology can effectively regulate the respiration rate of Yangfeng persimmons during storage.

[0093] 2. Fruit hardness and hard fruit rate Six fruits were taken for each group, and one point was taken from the equator of the yin and yang sides of each fruit for measurement, and the average value was taken as the measured value. The texture analyzer Texture Analyser XTplus (TA.XTplus) (Stable Micro Systems, UK) was used for measurement, puncture test, P / 5 probe, speed 1mm / s, displacement 10cm, trigger force 0.049N. Fruits with hardness greater than 8kg / cm2 at both points on the yin and yang sides were considered hard fruits. Hard fruit rate = number of hard fruits / total number of fruits measured × 100%. Figure 8 and Fig. 9 As shown in the figure, the fruit firmness of the control group decreased rapidly in the early stage of storage, especially from the 42nd to the 56th day, when the firmness dropped sharply from 10.76kg / cm² to 7.31kg / cm², a decrease of about 32%. After the 56th day, the rate of decline of the firmness slowed down and basically stabilized at about 5.60kg / cm². In contrast, the firmness of the fruit in the experimental group decreased more slowly, and its firmness remained at about 13.07kg / cm² on the 84th day of storage.

[0094] 3. Peel color Six fruits were taken for each group, and one point was taken at the equator of the yin and yang sides of each fruit for measurement, and the average value was taken as the measured value. Minolta Chroma Meter CR-400 (Minolta Konica, Japan) was used to measure L* and h° values ​​(C light source, 2° observation angle, measurement diameter 8 mm). L* represents brightness, L*=100 represents white, and L*=0 represents black. h° represents hue angle, which represents the actual perceived color. As the fruit matures, the color of the peel changes from green (180°) to yellow (90°) and approaches orange, away from yellow (90°) and toward red (0°). Fig.10 As shown in the figure, the L* value of the Yangfeng sweet persimmon peel showed a significant increase in the early stage of storage. The L* value of the control group climbed to the highest point on the 28th day, then dropped sharply, and rebounded slightly after hitting the bottom on the 56th day; while the L* value of the experimental group reached its peak on the 42nd day, and then steadily declined. Throughout the storage period, the L* value of the experimental group was always higher than that of the control group. Through variance analysis, it was found that in the first 28 days of storage, the difference in the L* value of the peel between the experimental group and the control group was not obvious, but from the 28th day, the difference became significant. As shown in the figure, the L* value of the fruit peel of the experimental group and the control group was not significantly different, but from the 28th day, the difference became significant. Fig.11 As shown in the figure, with the passage of storage time, the hue angle of Yangfeng sweet persimmon gradually decreases, especially in the early stage of storage, the hue angle decreases rapidly, and in the later stage, the change tends to be stable. During the storage process, the hue angle of Yangfeng sweet persimmon stored in AI fresh-keeping warehouse continues to maintain the highest level, while that in ordinary cold storage is second, and the difference between the two is particularly significant only in a specific period of time.

[0095] like Fig.12As shown, the fruit skin of the control group showed obvious browning, the color of the flesh became darker, and it showed a gel-like texture. In contrast, the fruit skin of the experimental group did not change significantly during storage, and no gelation occurred. Only in the later stages of storage was some slight browning found in some of the flesh. By comparing the surfaces, cross sections, and longitudinal sections of the two groups of fruits, it can be concluded that the AI ​​preservation technology of this application has a significant effect in maintaining the appearance quality of the fruit.

[0096] 4. Fruit chilling injury index The main manifestations of chilling injury to Yangfeng sweet persimmon are browning of the flesh and peel, gelation of the flesh, and stickiness of the juice. The chilling injury degree of the fruit is divided into 0 to 4 levels according to the size of the chilling injury area of ​​the fruit: Level 0, no chilling injury; Level 1, chilling injury area <25%; Level 2, 25%≤ chilling injury area <50%; Level 3, 50%≤ chilling injury area <75%; Level 4, chilling injury area ≥75%. 15 fruits were observed repeatedly in each group. Chilling injury rate (%) = number of fruits with chilling injury / total number of fruits observed * 100%. The chilling injury index of persimmon fruit is calculated according to the following formula:

[0097] like Fig.13 , Fig.14 and Fig.15 As shown in the figure, on the 28th day of storage, the control group began to show symptoms of chilling injury. In contrast, the experimental group did not begin to show similar signs of chilling injury until the 56th day of storage. Throughout the storage period, the chilling injury index of the experimental group was always significantly lower than that of the control group. Specifically, the chilling injury index of the control group began to gradually increase on the 28th day, and increased rapidly after the 56th day, and this growth trend continued until the 84th day. By the 84th day, the chilling injury index of the control group had climbed to 45.34. During the entire storage process of the experimental group, the chilling injury index only began to show a slight upward trend after the 56th day. Even so, the growth trend of the chilling injury index of the experimental group remained gentle and stable throughout the storage period, and until the 84th day, the growth of the index was still controlled at 10.05.

[0098] 5. Weight loss rate and decay rate like Fig.16 and Fig.17 As shown, during the storage period, the weight loss rate and spoilage rate of the experimental group were significantly lower than those of the control group, which proved that the storage technology and treatment methods adopted by the experimental group had significant advantages in maintaining fruit moisture and effectively inhibiting spoilage.

[0099] In summary, during the storage of Yangfeng persimmons, the AI ​​fresh-keeping warehouse demonstrated its excellent performance and significant advantages, not only maintaining the long-term freshness of the fruit, but also significantly reducing the weight loss rate and mold rate. In addition, the AI ​​fresh-keeping warehouse also performed well in maintaining the hardness and soluble solids content of the fruit, fully demonstrating the extraordinary ability of the AI ​​fresh-keeping warehouse in maintaining the freshness and taste of the fruit. Therefore, it can be clearly known that compared with traditional cold storage, the AI ​​fresh-keeping warehouse is more suitable for the long-term storage of Yangfeng persimmons.

[0100] Finally, it should be noted that the above embodiments are only used to illustrate the technical solution of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solution of the present invention can be modified or replaced by equivalents without departing from the spirit and scope of the technical solution of the present invention.

Claims

1. An intelligent preservation parameter control method for a fruit preservation cold storage, characterized in that: include: Receiving the variety information and maturity status label of the persimmon input by the user; Acquire a data set of real-time preservation parameters in the warehouse collected by the sensor matrix, wherein the real-time preservation parameters in the warehouse include temperature value, humidity value, oxygen concentration value and carbon dioxide concentration value; Mining the time series correlation feature of the fresh-keeping parameters on the data set of the real-time fresh-keeping parameters in the warehouse to obtain a time series correlation feature vector of the fresh-keeping parameters in the warehouse; Vectorizing the variety information of the persimmon and the maturity status label to obtain a cascade vector of persimmon fresh-keeping storage prior information; Interactively fuse the persimmon fresh-keeping storage prior information cascade vector and the in-store fresh-keeping parameter time series correlation feature vector to obtain a prior information-in-store fresh-keeping parameter significant interactive fusion representation vector; Based on the prior information-storage fresh-keeping parameter significant interaction fusion representation vector, determine the adjustment strategy of the storage fresh-keeping temperature; Among them, the cascade vector of the prior information on the fresh-keeping storage of persimmons and the temporal correlation feature vector of the in-warehouse fresh-keeping parameters are interactively fused, including: extracting the implicit correlation features between the cascade vector of the prior information on the fresh-keeping storage of persimmons and the temporal correlation feature vector of the in-warehouse fresh-keeping parameters as conditional feature vectors; based on the conditional feature vectors, guiding the cascade vector of the prior information on the fresh-keeping storage of persimmons and the temporal correlation feature vector of the in-warehouse fresh-keeping parameters to perform cross-domain interactive fusion based on the attention mechanism to obtain the prior information-in-warehouse fresh-keeping parameters significant interactive fusion representation vector.

2. The intelligent fresh-keeping parameter control method for a fruit fresh-keeping cold storage according to claim 1 is characterized in that: Mining the time series correlation feature of the fresh-keeping parameters on the data set of the real-time fresh-keeping parameters in the library to obtain the time series correlation feature vector of the fresh-keeping parameters in the library includes: The data set of the real-time fresh-keeping parameters in the warehouse is subjected to data structuring processing according to the time dimension and the parameter sample dimension to obtain a time series joint matrix of the fresh-keeping parameters in the warehouse; The in-store fresh-keeping parameter time series joint matrix is ​​input into the in-store fresh-keeping parameter time series association pattern feature miner based on the converter structure to obtain the in-store fresh-keeping parameter time series association feature vector.

3. The intelligent fresh-keeping parameter control method for a fruit fresh-keeping cold storage according to claim 2 is characterized in that: Vectorization is performed on the variety information of the persimmon and the maturity status label to obtain a cascade vector of persimmon fresh-keeping storage prior information, including: The variety information of the crisp persimmon and the maturity status label are one-hot encoded to obtain a crisp persimmon variety one-hot encoding vector and a maturity status label one-hot encoding vector, and the crisp persimmon variety one-hot encoding vector and the maturity status label one-hot encoding vector are cascaded to obtain a cascade vector of the crisp persimmon preservation storage prior information.

4. The intelligent fresh-keeping parameter control method for a fruit fresh-keeping cold storage according to claim 3 is characterized in that: Extracting the implicit correlation feature between the persimmon preservation storage prior information cascade vector and the in-store preservation parameter time series correlation feature vector as a conditional feature vector includes: Inputting the persimmon preservation storage prior information cascade vector and the in-store preservation parameter time series correlation feature vector into an implicit correlation feature capture network to obtain a prior information-in-store preservation parameter implicit correlation feature vector; The prior information-library preservation parameter implicit correlation feature vector is activated based on the Sigmoid function to obtain the conditional feature vector.

5. The intelligent fresh-keeping parameter control method of the fruit fresh-keeping cold storage according to claim 4 is characterized in that: Based on the conditional feature vector, the persimmon preservation storage prior information cascade vector and the in-store preservation parameter time series correlation feature vector are guided to perform cross-domain interactive fusion based on the attention mechanism to obtain the prior information-in-store preservation parameter significant interactive fusion representation vector, including: Based on the semantic relevance of the crisp persimmon fresh-keeping storage prior information cascade vector and the in-warehouse fresh-keeping parameter time series association feature vector relative to the condition feature vector, feature modulation is performed on the crisp persimmon fresh-keeping storage prior information cascade vector and the in-warehouse fresh-keeping parameter time series association feature vector to obtain a modulated crisp persimmon fresh-keeping storage prior information cascade vector and a modulated in-warehouse fresh-keeping parameter time series association feature vector; The prior information cascade vector of the post-modulation crisp persimmon preservation storage, the temporal correlation feature vector of the post-modulation storage preservation parameters and the conditional feature vector are cross-domain interactively encoded to obtain a significant interactive fusion representation vector of the prior information-storage preservation parameters.

6. The intelligent fresh-keeping parameter control method of the fruit fresh-keeping cold storage according to claim 5 is characterized in that: Based on the semantic relevance of the crisp persimmon fresh-keeping storage prior information cascade vector and the in-warehouse fresh-keeping parameter time series association feature vector relative to the condition feature vector, feature modulation is performed on the crisp persimmon fresh-keeping storage prior information cascade vector and the in-warehouse fresh-keeping parameter time series association feature vector to obtain a modulated crisp persimmon fresh-keeping storage prior information cascade vector and a modulated in-warehouse fresh-keeping parameter time series association feature vector, including: Calculate the first semantic relevance factor of the persimmon fresh-keeping storage prior information cascade vector relative to the conditional feature vector; Calculate a second semantic correlation factor of the in-store fresh-keeping parameter time series association feature vector to the condition feature vector; The first semantic relevance factor and the second semantic relevance factor are normalized, and the normalized first semantic relevance factor and the second semantic relevance factor are used to perform weighted modulation on the crisp persimmon preservation storage prior information cascade vector and the in-warehouse preservation parameter time series association feature vector to obtain the modulated crisp persimmon preservation storage prior information cascade vector and the modulated in-warehouse preservation parameter time series association feature vector.

7. The intelligent fresh-keeping parameter control method of the fruit fresh-keeping cold storage according to claim 6 is characterized in that: The prior information cascade vector of the post-modulation fresh-keeping storage of crisp persimmon, the temporal correlation feature vector of the post-modulation storage fresh-keeping parameter, and the conditional feature vector are cross-domain interactively encoded to obtain the prior information-storage fresh-keeping parameter significant interactive fusion representation vector, including: Taking the cascade vector of the prior information on the storage of fresh-keeping persimmons after modulation as the query vector, the temporal association feature vector of the storage parameters in the warehouse after modulation as the key vector and the conditional feature vector as the value vector, the cascade vector of the prior information on the storage of fresh-keeping persimmons after modulation, the temporal association feature vector of the storage parameters in the warehouse after modulation and the conditional feature vector are subjected to significant guided interaction among features based on the converter structure to obtain the significant interaction fusion representation vector of the prior information-the storage parameters in the warehouse.

8. The intelligent fresh-keeping parameter control method for a fruit fresh-keeping cold storage according to claim 7, characterized in that: Based on the prior information-the storage parameter significant interaction fusion representation vector, the adjustment strategy of the storage temperature is determined, including: Inputting the prior information-library preservation parameter significant interaction fusion representation vector into the decoder-based preservation parameter optimization module to obtain an optimization result, wherein the optimization result is a recommended preservation temperature value; Based on the optimization result, an adjustment strategy for the fresh-keeping temperature in the warehouse is determined.

9. An intelligent fresh-keeping parameter control system for a fruit fresh-keeping cold storage, characterized in that: include: A persimmon fresh-keeping storage prior information acquisition module is used to receive the persimmon variety information and maturity status label input by the user; A fresh-keeping environment data acquisition module is used to acquire a data set of real-time fresh-keeping parameters in the warehouse collected by the sensor matrix, wherein the real-time fresh-keeping parameters in the warehouse include temperature value, humidity value, oxygen concentration value and carbon dioxide concentration value; A fresh-keeping parameter time series association coding module is used to perform fresh-keeping parameter time series association feature mining on the data set of real-time fresh-keeping parameters in the library to obtain a fresh-keeping parameter time series association feature vector in the library; A priori information vector processing module, used for vectorizing the variety information of the crisp persimmon and the maturity status label to obtain a cascade vector of the crisp persimmon fresh-keeping storage priori information; A feature interactive fusion module is used to interactively fuse the persimmon preservation storage prior information cascade vector and the in-store preservation parameter time series correlation feature vector to obtain a prior information-in-store preservation parameter significant interactive fusion representation vector; The fresh-keeping temperature adjustment module is used to determine the adjustment strategy of the fresh-keeping temperature in the warehouse based on the prior information-the fresh-keeping parameter significant interaction fusion representation vector in the warehouse.

Images