Saline-alkali soil fertilization method and system based on intelligent recommendation

Abstract

The invention discloses a saline-alkali soil fertilization method and system based on intelligent recommendation. A map model based on saline-alkali soil is constructed, monitoring areas are divided for preset saline-alkali soil, similarity analysis is carried out on environment characteristic data of every two monitoring areas, and a feasible prediction area group is judged; in a real-time monitoring period, carrying out soil salinity and alkalinity detection on the sampling region, and importing detection data into an LSTM prediction model to carry out salinity and alkalinity content value prediction so as to obtain predicted salinity and alkalinity data of a feasible prediction region group; by predicting saline-alkali data and historical saline-alkali content data, fertilization and irrigation schemes are set for a plurality of monitoring areas, and intelligent scheme recommendation is performed based on different monitoring areas. Through the method, the saline-alkali content of the saline-alkali soil can be effectively predicted, the consumption of manpower and material resources for detecting soil is reduced, and the regulation and control capability of the saline-alkali soil is improved.

Description

Technical Field

[0001] The present invention relates to the field of intelligent analysis of saline-alkali land data, and more specifically, to a saline-alkali land fertilization method and system based on intelligent recommendation. Background Art

[0002] Saline-alkali land is a general term for saline soil and alkaline soil. It has the characteristics of low organic matter content, low soil fertility, poor physical and chemical properties, etc., which has a serious impact on crop growth. Traditional fertilization methods are often based on experience or simple soil tests, which are difficult to accurately judge the soil fertility status and crop needs. In addition, the detection and analysis efficiency of saline-alkali land is low, and the analysis accuracy of different environmental areas in saline-alkali land is low. Soil detection and analysis heavily rely on manual detection data. There is no efficient information technology to effectively analyze and evaluate different regions. There is also a lack of corresponding saline-alkali land data prediction methods, which seriously affects the effective analysis of fertilization plans, resulting in low fertilization efficiency and even aggravation of soil salinization. Therefore, the development of a saline-alkali land fertilization method and system based on intelligent recommendation is of great significance to improving the agricultural production efficiency of saline-alkali land. Summary of the invention

[0003] The present invention overcomes the defects of the prior art and proposes a saline-alkali land fertilization method and system based on intelligent recommendation.

[0004] A first aspect of the present invention provides a saline-alkali land fertilization method based on intelligent recommendation, comprising:

[0005] Obtain preset saline-alkali land area information, and construct a map model based on the saline-alkali land through the area information;

[0006] In the map model, the preset saline-alkali land is divided into multiple monitoring areas by presetting monitoring points, and the soil environmental data of the monitoring areas are obtained;

[0007] In the soil environmental data, feature vectors of multiple dimensions are generated corresponding to various environmental parameters, and environmental characteristic data of each monitoring area are formed;

[0008] Conduct similarity analysis on the environmental characteristic data of the monitoring area, mark the areas with expected similarity and geographically continuous, and use each group of marked continuous areas as a feasible prediction area group;

[0009] In a preset historical time period, the salinity and alkali content data of the feasible prediction area group are extracted and serialized, and the serialized data are imported into the LSTM prediction model for training. A monitoring area is selected in each feasible prediction area group as the sampling area;

[0010] In a real-time monitoring cycle, soil salinity and alkalinity are tested in the sampling area, and the test data are imported into the LSTM prediction model to perform numerical prediction of salinity and alkalinity content, and the predicted salinity and alkalinity data of the feasible prediction area group are obtained;

[0011] By predicting salinity data and historical salinity data, fertilization and irrigation plans are set for multiple monitoring areas, and intelligent plans are recommended based on different monitoring areas.

[0012] In this solution, the preset saline-alkali land area information is obtained, and a map model based on the saline-alkali land is constructed through the area information, specifically:

[0013] Obtaining preset saline-alkali land area information, the area information including area, area outline, planting range and monitoring point location information;

[0014] A two-dimensionally visualized map model is constructed using the regional information.

[0015] In this solution, in the map model, the preset saline-alkali land is divided into multiple monitoring areas by preset monitoring points, and the soil environmental data of the monitoring areas are obtained, specifically:

[0016] In the map model, obtain the location of the preset monitoring point;

[0017] Based on the locations of preset monitoring points, the preset saline-alkali land is divided into regions to form multiple monitoring areas, ensuring that each monitoring area corresponds to a preset monitoring point;

[0018] Soil environmental data of each monitoring area are collected within a preset historical time period.

[0019] In this solution, in the soil environmental data, feature vectors of multiple dimensions are generated corresponding to multiple environmental parameters, and environmental feature data of each monitoring area is formed, specifically:

[0020] In the soil environmental data, each environmental parameter is used as a dimension, and the corresponding parameter value is used as the value of the corresponding dimension to generate a multi-dimensional feature vector;

[0021] The multi-dimensional feature vector is used as the environmental feature data, and features are extracted for each monitoring area to form the environmental feature data of each monitoring area.

[0022] In this scheme, similarity analysis is performed on the environmental characteristic data of the monitoring area, and the areas with expected similarity and geographically continuous are marked, and each group of marked continuous areas is used as a feasible prediction area group, specifically:

[0023] The similarity calculation is performed on the environmental characteristic data of the monitoring area. The calculation process is to calculate the environmental characteristic data of the two selected monitoring areas based on the standard Euclidean distance method, and obtain the environmental similarity between the two monitoring areas through the distance value;

[0024] By presetting a similarity threshold, it is determined whether every two monitoring areas are similar, and all monitoring areas are determined, and the monitoring areas with similarity are marked to form a similar monitoring group. Based on all monitoring areas, multiple similar monitoring groups are generated;

[0025] A similar monitoring group includes at least two monitoring areas;

[0026] In similar monitoring groups, ensure that each monitoring area has similarity with at least one monitoring area in the group;

[0027] In a similar monitoring group, geographically continuous monitoring areas are screened, and a new group is formed based on the screening results, which is marked as a feasible prediction area group;

[0028] All similar monitoring groups are screened and marked to form multiple feasible prediction area groups.

[0029] In this scheme, within a preset historical time period, the salinity and alkali content data of the feasible prediction area group are extracted and the data is serialized, the serialized data is imported into the LSTM prediction model for training, and a monitoring area is selected as the sampling area in each feasible prediction area group, specifically:

[0030] A feasible forecast area group was used as the unit of analysis;

[0031] Acquiring salinity and alkali content data of all monitoring areas in the feasible prediction area group within a preset historical time period;

[0032] Sorting the salinity and alkali content data based on the geographical continuity order between monitoring areas in a feasible prediction area group, and serializing the sorted data to form serialized data;

[0033] Build an LSTM prediction model and initialize parameters and set loss function;

[0034] The serialized data is divided into a test set and a training set based on a preset ratio and imported into the LSTM prediction model for cyclic prediction training. During the training process, the model parameters are optimized through the loss function until the prediction accuracy reaches the preset standard.

[0035] Record the specific model parameters of the LSTM prediction model at this time, and associate the model parameters with the one feasible prediction area group;

[0036] In each feasible prediction area group, a monitoring area at a starting point location is selected as the sampling area based on the continuity of the geographical location.

[0037] In this scheme, within a real-time monitoring cycle, soil salinity detection is performed on the sampling area, and the detection data is imported into the LSTM prediction model for numerical prediction of salinity content to obtain the predicted salinity data of the feasible prediction area group, specifically:

[0038] A feasible forecast area group was used as the unit of analysis;

[0039] In a real-time monitoring cycle, soil salinity and alkalinity detection is performed on a sampling area in a feasible prediction area group to obtain detection data;

[0040] Import the detection data into the LSTM prediction model to perform numerical prediction of salinity and alkali content and obtain prediction sequence data;

[0041] The predicted sequence data are analyzed, and the data are matched to each monitoring area in a feasible prediction area group to obtain the predicted salinity data of each monitoring area in the group.

[0042] In this scheme, by predicting salinity data and historical salinity data, fertilization and irrigation schemes are set for multiple monitoring areas, and intelligent schemes are recommended based on different monitoring areas, specifically:

[0043] In a feasible prediction area group, the salinity and alkali changes in the area are evaluated through the detection data and predicted salinity and alkali data of the sampling area, and the salt content control index and alkalinity control index of the soil are generated;

[0044] Analyzing fertilization and irrigation plans based on the soil salt content control index and alkalinity control index, and generating a predicted regional fertilization plan;

[0045] According to multiple feasible forecast area groups, fertilization plans corresponding to multiple forecast areas are generated;

[0046] In the map model, all monitoring areas outside the feasible prediction area group are marked as non-prediction areas. Real-time salinity and alkali data detection is carried out on the non-prediction areas. Combined with historical salinity and alkali content data, matching selection analysis is carried out in the pre-fertilization schemes to screen out a variety of real-time fertilization schemes.

[0047] Based on multiple predicted regional fertilization plans and multiple real-time fertilization plans, intelligent plans are recommended for preset saline-alkali areas.

[0048] The second aspect of the present invention further provides a saline-alkali land fertilization system based on intelligent recommendation, the system comprising: a memory, a processor, the memory comprising a saline-alkali land fertilization program based on intelligent recommendation, the saline-alkali land fertilization program based on intelligent recommendation being executed by the processor to implement the following steps:

[0049] Obtain preset saline-alkali land area information, and construct a map model based on the saline-alkali land through the area information;

[0050] In the map model, the preset saline-alkali land is divided into multiple monitoring areas by presetting monitoring points, and the soil environmental data of the monitoring areas are obtained;

[0051] In the soil environmental data, feature vectors of multiple dimensions are generated corresponding to various environmental parameters, and environmental characteristic data of each monitoring area are formed;

[0052] Conduct similarity analysis on the environmental characteristic data of the monitoring area, mark the areas with expected similarity and geographically continuous, and use each group of marked continuous areas as a feasible prediction area group;

[0053] In a preset historical time period, the salinity and alkali content data of the feasible prediction area group are extracted and serialized, and the serialized data are imported into the LSTM prediction model for training. A monitoring area is selected in each feasible prediction area group as the sampling area;

[0054] In a real-time monitoring cycle, soil salinity and alkalinity are tested in the sampling area, and the test data are imported into the LSTM prediction model to perform numerical prediction of salinity and alkalinity content, and the predicted salinity and alkalinity data of the feasible prediction area group are obtained;

[0055] By predicting salinity data and historical salinity data, fertilization and irrigation plans are set for multiple monitoring areas, and intelligent plans are recommended based on different monitoring areas.

[0056] The second aspect of the present invention further provides a saline-alkali land fertilization system based on intelligent recommendation, the system comprising: a memory, a processor, the memory comprising a saline-alkali land fertilization program based on intelligent recommendation, the saline-alkali land fertilization program based on intelligent recommendation being executed by the processor to implement the following steps:

[0057] The third aspect of the present invention also provides a computer-readable storage medium, which includes a saline-alkali land fertilization program based on intelligent recommendation. When the saline-alkali land fertilization program based on intelligent recommendation is executed by a processor, the steps of the saline-alkali land fertilization method based on intelligent recommendation as described in any one of the above items are implemented.

[0058] The present invention discloses a saline-alkali land fertilization method and system based on intelligent recommendation. By constructing a map model based on saline-alkali land, the preset saline-alkali land is divided into monitoring areas, and the environmental characteristic data of every two monitoring areas are analyzed for similarity, and a feasible prediction area group is determined; within a real-time monitoring cycle, the soil salinity and alkalinity of the sampling area are detected, and the detection data is imported into the LSTM prediction model for numerical prediction of the salinity and alkali content, and the predicted salinity and alkali data of the feasible prediction area group are obtained; by predicting the salinity and alkali data and the historical salinity and alkali content data, fertilization and irrigation schemes are set for multiple monitoring areas, and intelligent schemes are recommended based on different monitoring areas. Through the present invention, it is possible to effectively predict the salinity and alkali content of saline-alkali land, reduce the human and material consumption of soil detection, and improve the regulation and control capacity of saline-alkali land. BRIEF DESCRIPTION OF THE DRAWINGS

[0059] Figure 1 A flow chart of a saline-alkali land fertilization method based on intelligent recommendation according to the present invention is shown;

[0060] Figure 2 A block diagram of a saline-alkali land fertilization system based on intelligent recommendation according to the present invention is shown. DETAILED DESCRIPTION

[0061] In order to more clearly understand the above-mentioned purpose, features and advantages of the present invention, the present invention is further described in detail below in conjunction with the accompanying drawings and specific embodiments. It should be noted that the embodiments of the present application and the features in the embodiments can be combined with each other without conflict.

[0062] In the following description, many specific details are set forth to facilitate a full understanding of the present invention. However, the present invention may also be implemented in other ways different from those described herein. Therefore, the protection scope of the present invention is not limited to the specific embodiments disclosed below.

[0063] Figure 1 A flow chart of a saline-alkali land fertilization method based on intelligent recommendation of the present invention is shown.

[0064] like Figure 1 As shown, the first aspect of the present invention provides a saline-alkali land fertilization method based on intelligent recommendation, comprising:

[0065] S102, obtaining preset saline-alkali land area information, and constructing a map model based on the saline-alkali land through the area information;

[0066] S104, in the map model, the preset saline-alkali land is divided into a plurality of monitoring areas by preset monitoring points, and soil environmental data of the monitoring areas are obtained;

[0067] S106, generating feature vectors of multiple dimensions corresponding to multiple environmental parameters in the soil environmental data, and forming environmental feature data for each monitoring area;

[0068] S108, performing similarity analysis on the environmental characteristic data of the monitoring area, marking the areas with expected similarity and continuous geographical locations, and taking each group of marked continuous areas as a feasible prediction area group;

[0069] S110, extracting salinity data of feasible prediction area groups within a preset historical time period and serializing the data, importing the serialized data into an LSTM prediction model for training, and selecting a monitoring area in each feasible prediction area group as a sampling area;

[0070] S112, within a real-time monitoring cycle, soil salinity detection is performed on the sampling area, and the detection data is imported into the LSTM prediction model to perform numerical prediction of salinity content, so as to obtain predicted salinity data of the feasible prediction area group;

[0071] S114, by predicting salinity data and historical salinity data, fertilization and irrigation plans are set for multiple monitoring areas, and intelligent plans are recommended based on different monitoring areas.

[0072] According to an embodiment of the present invention, the step of obtaining preset saline-alkali land regional information and constructing a map model based on the saline-alkali land through the regional information is specifically as follows:

[0073] Obtaining preset saline-alkali land area information, the area information including area, area outline, planting range and monitoring point location information;

[0074] A two-dimensionally visualized map model is constructed using the regional information.

[0075] It should be noted that the map model is a visualization model used for visual display of current saline-alkali data and corresponding area display, which can conveniently and intuitively understand the impact distribution of saline-alkali land.

[0076] According to an embodiment of the present invention, in the map model, the preset saline-alkali land is divided into multiple monitoring areas by preset monitoring points, and the soil environmental data of the monitoring areas is obtained, specifically:

[0077] In the map model, obtain the location of the preset monitoring point;

[0078] Based on the locations of preset monitoring points, the preset saline-alkali land is divided into regions to form multiple monitoring areas, ensuring that each monitoring area corresponds to a preset monitoring point;

[0079] Soil environmental data of each monitoring area are collected within a preset historical time period.

[0080] It should be noted that the preset monitoring points include multiple ones. The preset historical time period is generally a monitoring period set by a user, and it is necessary to collect enough environmental characteristic data to perform effective characteristic analysis on the environmental data of the area. Soil environmental data includes a variety of environmental parameters, such as soil temperature, humidity, evaporation, conductivity, etc.

[0081] According to an embodiment of the present invention, in the soil environmental data, feature vectors of multiple dimensions are generated corresponding to multiple environmental parameters, and environmental feature data of each monitoring area is formed, specifically:

[0082] In the soil environmental data, each environmental parameter is used as a dimension, and the corresponding parameter value is used as the value of the corresponding dimension to generate a multi-dimensional feature vector;

[0083] The multi-dimensional feature vector is used as the environmental feature data, and features are extracted for each monitoring area to form the environmental feature data of each monitoring area.

[0084] It should be noted that, in the feature vector, one environmental parameter corresponds to one dimension.

[0085] According to an embodiment of the present invention, similarity analysis is performed on the environmental feature data of the monitoring area, and regions with expected similarity and continuous geographical locations are marked, and each group of marked continuous regions is used as a feasible prediction region group, specifically:

[0086] The similarity calculation is performed on the environmental characteristic data of the monitoring area. The calculation process is to calculate the environmental characteristic data of the two selected monitoring areas based on the standard Euclidean distance method, and obtain the environmental similarity between the two monitoring areas through the distance value;

[0087] By presetting a similarity threshold, it is determined whether every two monitoring areas are similar, and all monitoring areas are determined, and the monitoring areas with similarity are marked to form a similar monitoring group. Based on all monitoring areas, multiple similar monitoring groups are generated;

[0088] A similar monitoring group includes at least two monitoring areas;

[0089] In similar monitoring groups, ensure that each monitoring area has similarity with at least one monitoring area in the group;

[0090] In a similar monitoring group, geographically continuous monitoring areas are screened, and a new group is formed based on the screening results, which is marked as a feasible prediction area group;

[0091] All similar monitoring groups are screened and marked to form multiple feasible prediction area groups.

[0092] It should be noted that in the similarity obtained by the distance value, the greater the distance, the smaller the similarity. The feasible prediction area group includes at least two monitoring areas. In the process of forming multiple feasible prediction area groups, the areas with related and continuous monitoring data can be classified and grouped to form multiple groups, so as to realize the subsequent accurate and efficient fertilization plan analysis and saline-alkali land regulation.

[0093] In the present invention, the monitoring area is first divided according to the saline-alkali land, and the similarity and geographical location continuity are judged based on the environmental characteristics to select different area groups, namely, feasible prediction area groups. In the feasible prediction area group, the areas within the group have a high degree of environmental characteristic similarity and correlation with the detection data. Based on this, the corresponding changes in salinity and alkali content have a certain correlation with the detection values, and therefore have a certain prediction feasibility. Furthermore, the present invention trains a prediction model for the historical data of each feasible prediction area group. The prediction model adopts a deep prediction model based on sequence data to simulate the changes in salinity and alkali values ​​in continuous areas in the region. Different feasible prediction area groups are associated with different model parameters, so that each feasible prediction area group corresponds to a trained prediction model. In the real-time monitoring cycle, by combining the sampling area data with the prediction model, accurate data prediction can be achieved for the remaining areas in the feasible prediction area group, which greatly reduces the number of manual sampling and detection frequency, and provides an efficient saline-alkali land analysis method based on informatization. It is worth mentioning that the current existing land sampling and detection process is relatively time-consuming and labor-intensive. For large-scale saline-alkali land, full-area sampling and analysis is generally adopted, which is a time-consuming and labor-intensive process. Therefore, through the present invention, it is possible to achieve accurate data prediction for multiple regions based on a single sampling and analysis, greatly improve the monitoring and analysis efficiency of saline-alkali land, reduce repetitive sampling and analysis processes, and achieve cost reduction and efficiency improvement.

[0094] According to an embodiment of the present invention, within a preset historical time period, the salinity and alkali content data of the feasible prediction area group are extracted and the data is serialized, the serialized data is imported into the LSTM prediction model for training, and a monitoring area is selected in each feasible prediction area group as a sampling area, specifically:

[0095] A feasible forecast area group was used as the unit of analysis;

[0096] Acquiring salinity and alkali content data of all monitoring areas in the feasible prediction area group within a preset historical time period;

[0097] Sorting the salinity and alkali content data based on the geographical continuity order between monitoring areas in a feasible prediction area group, and serializing the sorted data to form serialized data;

[0098] Build an LSTM prediction model and initialize parameters and set loss function;

[0099] The serialized data is divided into a test set and a training set based on a preset ratio and imported into the LSTM prediction model for cyclic prediction training. During the training process, the model parameters are optimized through the loss function until the prediction accuracy reaches the preset standard.

[0100] Record the specific model parameters of the LSTM prediction model at this time, and associate the model parameters with the one feasible prediction area group;

[0101] In each feasible prediction area group, a monitoring area at a starting point location is selected as the sampling area based on the continuity of the geographical location.

[0102] It should be noted that the salt-alkali content data is sorted based on the geographical continuity order between the monitoring areas in a feasible prediction area group. Specifically, a regional route is formed based on the continuous areas, and the salt-alkali content data corresponding to each monitoring area is serialized based on the order in the regional route. For example, a preset saline-alkali area is divided into multiple monitoring areas by a grid. In a feasible prediction area group, multiple horizontal continuous monitoring areas are included. The salt-alkali content data can be serialized based on the order from left to right (or from right to left) of multiple continuous monitoring areas. Each unit data after serialization corresponds to the data of each monitoring area. It is worth mentioning that in each feasible prediction area group, based on the continuity of the geographical location, a monitoring area at a starting point position is selected as the sampling area. Applied to the above example, the leftmost monitoring area can be used as the sampling area in the order from left to right of multiple continuous monitoring areas, because it can be used as a starting point to connect the remaining monitoring areas in terms of geographical location. Subsequently, the salt-alkali data of the sampling area and the prediction model can be used to predict the salt-alkali data of the remaining continuous monitoring areas.

[0103] The specific model parameters of the LSTM prediction model are recorded at this time, and the model parameters are associated with the one feasible prediction area group. Since the environmental characteristics of each feasible prediction area group are different, the corresponding salinity and alkali content are also different, and the change rules are different. Therefore, the present invention trains prediction models for different groups respectively, generates model parameters for corresponding groups, and there are differences in the prediction model parameters of different groups.

[0104] LSTM, or Long Short-Term Memory, is a special recurrent neural network (RNN) used to predict data. In the embodiment of the present invention, since the environmental characteristics in similar and adjacent continuous monitoring areas are correlated and continuous, the corresponding salinity content has a certain correlation. The present invention uses salinity content as serialized data to perform corresponding prediction training, which can effectively mine the salinity change law in continuous areas, realize accurate prediction of saline-alkali land, improve the information integration ability of saline-alkali land, reduce unnecessary sampling monitoring, reduce the manpower and material resources consumed by saline-alkali land monitoring, and achieve cost reduction and efficiency improvement.

[0105] According to an embodiment of the present invention, within a real-time monitoring cycle, soil salinity detection is performed on the sampling area, and the detection data is imported into the LSTM prediction model to perform numerical prediction of salinity content, and the predicted salinity data of the feasible prediction area group is obtained, specifically:

[0106] A feasible forecast area group was used as the unit of analysis;

[0107] In a real-time monitoring cycle, soil salinity and alkalinity detection is performed on a sampling area in a feasible prediction area group to obtain detection data;

[0108] Import the detection data into the LSTM prediction model to perform numerical prediction of salinity and alkali content and obtain prediction sequence data;

[0109] The predicted sequence data are analyzed, and the data are matched to each monitoring area in a feasible prediction area group to obtain the predicted salinity data of each monitoring area in the group.

[0110] It should be noted that the detection data is specifically salt-alkali content data. The salt-alkali content data includes values ​​such as the salt content and alkalinity of the soil. In a feasible prediction area group, except for the sampling area for detection data, the remaining monitoring areas correspond to predicted salt-alkali data.

[0111] According to an embodiment of the present invention, by predicting salinity data and historical salinity data, fertilization and irrigation schemes are set for multiple monitoring areas, and intelligent schemes are recommended based on different monitoring areas, specifically:

[0112] In a feasible prediction area group, the salinity and alkali changes in the area are evaluated through the detection data and predicted salinity and alkali data of the sampling area, and the salt content control index and alkalinity control index of the soil are generated;

[0113] Analyzing fertilization and irrigation plans based on the soil salt content control index and alkalinity control index, and generating a predicted regional fertilization plan;

[0114] According to multiple feasible forecast area groups, fertilization plans corresponding to multiple forecast areas are generated;

[0115] In the map model, all monitoring areas outside the feasible prediction area group are marked as non-prediction areas. Real-time salinity and alkali data detection is carried out on the non-prediction areas. Combined with historical salinity and alkali content data, matching selection analysis is carried out in the pre-fertilization schemes to screen out a variety of real-time fertilization schemes.

[0116] Based on multiple predicted regional fertilization plans and multiple real-time fertilization plans, intelligent plans are recommended for preset saline-alkali areas.

[0117] It should be noted that in the intelligent scheme recommendation, the predicted area fertilization scheme can be applied to the fertilization scheme recommendation of the feasible predicted area group, and its specific scheme is a scheme dynamically analyzed based on the predicted data; multiple real-time fertilization schemes are fixed preferred schemes, which are applicable to non-predicted areas. Pre-set fertilization schemes are fixed preset recommended schemes, including multiple ones.

[0118] When analyzing fertilization and irrigation plans, the quality requirements of irrigation water sources, including conductivity and pH value, can be determined based on corresponding indicators, and the use of low-salt water for irrigation can be dynamically controlled to help reduce salt accumulation in the soil. The fertilization plan can set the amount of fertilizer and the salt ratio of the fertilizer based on indicators to control the degree of soil salinization.

[0119] According to an embodiment of the present invention, it also includes:

[0120] Mark a feasible prediction area group as the current group;

[0121] In the second monitoring cycle, the second soil environment data of the monitoring area in the current group is obtained;

[0122] Extracting features from the second soil environment data to obtain second environment feature data;

[0123] Based on multiple monitoring areas in the current group, the second environmental feature data is subjected to variance calculation. During the calculation process, the data difference is calculated using the standard Euclidean distance to calculate the difference between two types of second environmental feature data, and a variance result is obtained to determine whether the variance result is within a preset variance range;

[0124] If not, a secondary similarity analysis and judgment is performed based on multiple monitoring areas in the current group to determine areas with expected similarity and geographical continuity. Based on the judgment results, the monitoring areas in the current group are updated to form a new feasible prediction area group.

[0125] It should be noted that with the changes in environmental factors, the environmental characteristics between the monitoring areas within the feasible prediction area group will also change accordingly. When the changes exceed expectations, the monitoring areas within the feasible prediction area group will no longer have the corresponding correlation. For example, if a certain area experiences severe sudden pollution or sudden changes in climate and environmental factors, it is necessary to update the feasible prediction area group in real time to dynamically adapt to different environmental changes and ensure the applicability and accuracy of the prediction model. At the same time, the fertilization plan can be updated synchronously to improve the dynamic planning capability of the plan in information analysis.

[0126] Figure 2 A block diagram of a saline-alkali land fertilization system based on intelligent recommendation according to the present invention is shown.

[0127] The second aspect of the present invention further provides a saline-alkali land fertilization system 2 based on intelligent recommendation, the system comprising: a memory 21, a processor 22, the memory comprising a saline-alkali land fertilization program based on intelligent recommendation, the saline-alkali land fertilization program based on intelligent recommendation being executed by the processor to implement the following steps:

[0128] Obtain preset saline-alkali land area information, and construct a map model based on the saline-alkali land through the area information;

[0129] In the map model, the preset saline-alkali land is divided into multiple monitoring areas by presetting monitoring points, and the soil environmental data of the monitoring areas are obtained;

[0130] In the soil environmental data, feature vectors of multiple dimensions are generated corresponding to various environmental parameters, and environmental characteristic data of each monitoring area are formed;

[0131] Conduct similarity analysis on the environmental characteristic data of the monitoring area, mark the areas with expected similarity and geographically continuous, and use each group of marked continuous areas as a feasible prediction area group;

[0132] In a preset historical time period, the salinity and alkali content data of the feasible prediction area group are extracted and serialized, and the serialized data are imported into the LSTM prediction model for training. A monitoring area is selected in each feasible prediction area group as the sampling area;

[0133] In a real-time monitoring cycle, soil salinity and alkalinity are tested in the sampling area, and the test data are imported into the LSTM prediction model to perform numerical prediction of salinity and alkalinity content, and the predicted salinity and alkalinity data of the feasible prediction area group are obtained;

[0134] By predicting salinity data and historical salinity data, fertilization and irrigation plans are set for multiple monitoring areas, and intelligent plans are recommended based on different monitoring areas.

[0135] According to an embodiment of the present invention, the step of obtaining preset saline-alkali land regional information and constructing a map model based on the saline-alkali land through the regional information is specifically as follows:

[0136] Obtaining preset saline-alkali land area information, the area information including area, area outline, planting range and monitoring point location information;

[0137] A two-dimensionally visualized map model is constructed using the regional information.

[0138] It should be noted that the map model is a visualization model used for visual display of current saline-alkali data and corresponding area display, which can conveniently and intuitively understand the impact distribution of saline-alkali land.

[0139] According to an embodiment of the present invention, in the map model, the preset saline-alkali land is divided into multiple monitoring areas by preset monitoring points, and the soil environmental data of the monitoring areas is obtained, specifically:

[0140] In the map model, obtain the location of the preset monitoring point;

[0141] Based on the locations of preset monitoring points, the preset saline-alkali land is divided into regions to form multiple monitoring areas, ensuring that each monitoring area corresponds to a preset monitoring point;

[0142] Soil environmental data of each monitoring area are collected within a preset historical time period.

[0143] It should be noted that the preset monitoring points include multiple ones. The preset historical time period is generally a monitoring period set by a user, and it is necessary to collect enough environmental characteristic data to perform effective characteristic analysis on the environmental data of the area. Soil environmental data includes a variety of environmental parameters, such as soil temperature, humidity, evaporation, conductivity, etc.

[0144] According to an embodiment of the present invention, in the soil environmental data, feature vectors of multiple dimensions are generated corresponding to multiple environmental parameters, and environmental feature data of each monitoring area is formed, specifically:

[0145] In the soil environmental data, each environmental parameter is used as a dimension, and the corresponding parameter value is used as the value of the corresponding dimension to generate a multi-dimensional feature vector;

[0146] The multi-dimensional feature vector is used as the environmental feature data, and features are extracted for each monitoring area to form the environmental feature data of each monitoring area.

[0147] It should be noted that, in the feature vector, one environmental parameter corresponds to one dimension.

[0148] According to an embodiment of the present invention, similarity analysis is performed on the environmental feature data of the monitoring area, and regions with expected similarity and continuous geographical locations are marked, and each group of marked continuous regions is used as a feasible prediction region group, specifically:

[0149] The similarity calculation is performed on the environmental characteristic data of the monitoring area. The calculation process is to calculate the environmental characteristic data of the two selected monitoring areas based on the standard Euclidean distance method, and obtain the environmental similarity between the two monitoring areas through the distance value;

[0150] By presetting a similarity threshold, it is determined whether every two monitoring areas are similar, and all monitoring areas are determined, and the monitoring areas with similarity are marked to form a similar monitoring group. Based on all monitoring areas, multiple similar monitoring groups are generated;

[0151] A similar monitoring group includes at least two monitoring areas;

[0152] In similar monitoring groups, ensure that each monitoring area has similarity with at least one monitoring area in the group;

[0153] In a similar monitoring group, geographically continuous monitoring areas are screened, and a new group is formed based on the screening results, which is marked as a feasible prediction area group;

[0154] All similar monitoring groups are screened and marked to form multiple feasible prediction area groups.

[0155] It should be noted that in the similarity obtained by the distance value, the greater the distance, the smaller the similarity. The feasible prediction area group includes at least two monitoring areas. In the process of forming multiple feasible prediction area groups, the areas with related and continuous monitoring data can be classified and grouped to form multiple groups, so as to realize the subsequent accurate and efficient fertilization plan analysis and saline-alkali land regulation.

[0156] In the present invention, the monitoring area is first divided according to the saline-alkali land, and the similarity and geographical location continuity are judged based on the environmental characteristics to select different area groups, namely, feasible prediction area groups. In the feasible prediction area group, the areas within the group have a high degree of environmental characteristic similarity and correlation with the detection data. Based on this, the corresponding changes in salinity and alkali content have a certain correlation with the detection values, and therefore have a certain prediction feasibility. Furthermore, the present invention trains a prediction model for the historical data of each feasible prediction area group. The prediction model adopts a deep prediction model based on sequence data to simulate the changes in salinity and alkali values ​​in continuous areas in the region. Different feasible prediction area groups are associated with different model parameters, so that each feasible prediction area group corresponds to a trained prediction model. In the real-time monitoring cycle, by combining the sampling area data with the prediction model, accurate data prediction can be achieved for the remaining areas in the feasible prediction area group, which greatly reduces the number of manual sampling and detection frequency, and provides an efficient saline-alkali land analysis method based on informatization. It is worth mentioning that the current existing land sampling and detection process is relatively time-consuming and labor-intensive. For large-scale saline-alkali land, full-area sampling and analysis is generally adopted, which is a time-consuming and labor-intensive process. Therefore, through the present invention, it is possible to achieve accurate data prediction for multiple regions based on a single sampling and analysis, greatly improve the monitoring and analysis efficiency of saline-alkali land, reduce repetitive sampling and analysis processes, and achieve cost reduction and efficiency improvement.

[0157] According to an embodiment of the present invention, within a preset historical time period, the salinity and alkali content data of the feasible prediction area group are extracted and the data is serialized, the serialized data is imported into the LSTM prediction model for training, and a monitoring area is selected in each feasible prediction area group as a sampling area, specifically:

[0158] A feasible forecast area group was used as the unit of analysis;

[0159] Acquiring salinity and alkali content data of all monitoring areas in the feasible prediction area group within a preset historical time period;

[0160] Sorting the salinity and alkali content data based on the geographical continuity order between monitoring areas in a feasible prediction area group, and serializing the sorted data to form serialized data;

[0161] Build an LSTM prediction model and initialize parameters and set loss function;

[0162] The serialized data is divided into a test set and a training set based on a preset ratio and imported into the LSTM prediction model for cyclic prediction training. During the training process, the model parameters are optimized through the loss function until the prediction accuracy reaches the preset standard.

[0163] Record the specific model parameters of the LSTM prediction model at this time, and associate the model parameters with the one feasible prediction area group;

[0164] In each feasible prediction area group, a monitoring area at a starting point location is selected as the sampling area based on the continuity of the geographical location.

[0165] It should be noted that the salt-alkali content data is sorted based on the geographical continuity order between the monitoring areas in a feasible prediction area group. Specifically, a regional route is formed based on the continuous areas, and the salt-alkali content data corresponding to each monitoring area is serialized based on the order in the regional route. For example, a preset saline-alkali area is divided into multiple monitoring areas by a grid. In a feasible prediction area group, multiple horizontal continuous monitoring areas are included. The salt-alkali content data can be serialized based on the order from left to right (or from right to left) of multiple continuous monitoring areas. Each unit data after serialization corresponds to the data of each monitoring area. It is worth mentioning that in each feasible prediction area group, based on the continuity of the geographical location, a monitoring area at a starting point position is selected as the sampling area. Applied to the above example, the leftmost monitoring area can be used as the sampling area in the order from left to right of multiple continuous monitoring areas, because it can be used as a starting point to connect the remaining monitoring areas in terms of geographical location. Subsequently, the salt-alkali data of the sampling area and the prediction model can be used to predict the salt-alkali data of the remaining continuous monitoring areas.

[0166] The specific model parameters of the LSTM prediction model are recorded at this time, and the model parameters are associated with the one feasible prediction area group. Since the environmental characteristics of each feasible prediction area group are different, the corresponding salinity and alkali content are also different, and the change rules are different. Therefore, the present invention trains prediction models for different groups respectively, generates model parameters for corresponding groups, and there are differences in the prediction model parameters of different groups.

[0167] LSTM, or Long Short-Term Memory, is a special recurrent neural network (RNN) used to predict data. In the embodiment of the present invention, since the environmental characteristics in similar and adjacent continuous monitoring areas are correlated and continuous, the corresponding salinity content has a certain correlation. The present invention uses salinity content as serialized data to perform corresponding prediction training, which can effectively mine the salinity change law in continuous areas, realize accurate prediction of saline-alkali land, improve the information integration ability of saline-alkali land, reduce unnecessary sampling monitoring, reduce the manpower and material resources consumed by saline-alkali land monitoring, and achieve cost reduction and efficiency improvement.

[0168] According to an embodiment of the present invention, within a real-time monitoring cycle, soil salinity detection is performed on the sampling area, and the detection data is imported into the LSTM prediction model to perform numerical prediction of salinity content, and the predicted salinity data of the feasible prediction area group is obtained, specifically:

[0169] A feasible forecast area group was used as the unit of analysis;

[0170] In a real-time monitoring cycle, soil salinity and alkalinity detection is performed on a sampling area in a feasible prediction area group to obtain detection data;

[0171] Import the detection data into the LSTM prediction model to perform numerical prediction of salinity and alkali content and obtain prediction sequence data;

[0172] The predicted sequence data are analyzed, and the data are matched to each monitoring area in a feasible prediction area group to obtain the predicted salinity data of each monitoring area in the group.

[0173] It should be noted that the detection data is specifically salt-alkali content data. The salt-alkali content data includes values ​​such as the salt content and alkalinity of the soil. In a feasible prediction area group, except for the sampling area for detection data, the remaining monitoring areas correspond to predicted salt-alkali data.

[0174] According to an embodiment of the present invention, by predicting salinity data and historical salinity data, fertilization and irrigation schemes are set for multiple monitoring areas, and intelligent schemes are recommended based on different monitoring areas, specifically:

[0175] In a feasible prediction area group, the salinity and alkali changes in the area are evaluated through the detection data and predicted salinity and alkali data of the sampling area, and the salt content control index and alkalinity control index of the soil are generated;

[0176] Analyzing fertilization and irrigation plans based on the soil salt content control index and alkalinity control index, and generating a predicted regional fertilization plan;

[0177] According to multiple feasible forecast area groups, fertilization plans corresponding to multiple forecast areas are generated;

[0178] In the map model, all monitoring areas outside the feasible prediction area group are marked as non-prediction areas. Real-time salinity and alkali data detection is carried out on the non-prediction areas. Combined with historical salinity and alkali content data, matching selection analysis is carried out in the pre-fertilization schemes to screen out a variety of real-time fertilization schemes.

[0179] Based on multiple predicted regional fertilization plans and multiple real-time fertilization plans, intelligent plans are recommended for preset saline-alkali areas.

[0180] It should be noted that in the intelligent scheme recommendation, the predicted area fertilization scheme can be applied to the fertilization scheme recommendation of the feasible predicted area group, and its specific scheme is a scheme dynamically analyzed based on the predicted data; multiple real-time fertilization schemes are fixed preferred schemes, which are applicable to non-predicted areas. Pre-set fertilization schemes are fixed preset recommended schemes, including multiple ones.

[0181] When analyzing fertilization and irrigation plans, the quality requirements of irrigation water sources, including conductivity and pH value, can be determined based on corresponding indicators, and the use of low-salt water for irrigation can be dynamically controlled to help reduce salt accumulation in the soil. The fertilization plan can set the amount of fertilizer and the salt ratio of the fertilizer based on indicators to control the degree of soil salinization.

[0182] The third aspect of the present invention also provides a computer-readable storage medium, which includes a saline-alkali land fertilization program based on intelligent recommendation. When the saline-alkali land fertilization program based on intelligent recommendation is executed by a processor, the steps of the saline-alkali land fertilization method based on intelligent recommendation as described in any one of the above items are implemented.

[0183] The present invention discloses a saline-alkali land fertilization method and system based on intelligent recommendation. By constructing a map model based on saline-alkali land, the preset saline-alkali land is divided into monitoring areas, and the environmental characteristic data of every two monitoring areas are analyzed for similarity, and a feasible prediction area group is determined; within a real-time monitoring cycle, the soil salinity and alkalinity of the sampling area are detected, and the detection data is imported into the LSTM prediction model for numerical prediction of the salinity and alkali content, and the predicted salinity and alkali data of the feasible prediction area group are obtained; by predicting the salinity and alkali data and the historical salinity and alkali content data, fertilization and irrigation schemes are set for multiple monitoring areas, and intelligent schemes are recommended based on different monitoring areas. Through the present invention, it is possible to effectively predict the salinity and alkali content of saline-alkali land, reduce the human and material consumption of soil detection, and improve the regulation and control capacity of saline-alkali land.

[0184] In the several embodiments provided in the present application, it should be understood that the disclosed devices and methods can be implemented in other ways. The device embodiments described above are only schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation, such as: multiple units or components can be combined, or can be integrated into another system, or some features can be ignored or not executed. In addition, the coupling, direct coupling, or communication connection between the components shown or discussed can be through some interfaces, and the indirect coupling or communication connection of the devices or units can be electrical, mechanical or other forms.

[0185] The units described above as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units; they may be located in one place or distributed on multiple network units; some or all of the units may be selected according to actual needs to achieve the purpose of the present embodiment.

[0186] In addition, all functional units in the embodiments of the present invention may be integrated into one processing unit, or each unit may be separately used as a unit, or two or more units may be integrated into one unit; the above-mentioned integrated units may be implemented in the form of hardware or in the form of hardware plus software functional units.

[0187] Those skilled in the art can understand that: all or part of the steps of implementing the above method embodiments can be completed by hardware related to program instructions, and the aforementioned program can be stored in a computer-readable storage medium. When the program is executed, it executes the steps of the above method embodiments; and the aforementioned storage medium includes: mobile storage devices, read-only memories (ROM, Read-Only Memory), random access memories (RAM, Random Access Memory), disks or optical disks, and other media that can store program codes.

[0188] Alternatively, if the above-mentioned integrated unit of the present invention is implemented in the form of a software function module and sold or used as an independent product, it can also be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the embodiment of the present invention can be essentially or partly reflected in the form of a software product that contributes to the prior art. The computer software product is stored in a storage medium and includes several instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the methods described in each embodiment of the present invention. The aforementioned storage medium includes: various media that can store program codes, such as mobile storage devices, ROM, RAM, magnetic disks or optical disks.

[0189] The above description is only a specific implementation mode of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can easily think of changes or substitutions within the technical scope disclosed by the present invention, which should be covered by the protection scope of the present invention.

Claims

1. A saline-alkali land fertilization method based on intelligent recommendation, characterized in that: include: Obtain preset saline-alkali land area information, and construct a map model based on the saline-alkali land through the area information; In the map model, the preset saline-alkali land is divided into multiple monitoring areas by presetting monitoring points, and the soil environmental data of the monitoring areas are obtained; In the soil environmental data, feature vectors of multiple dimensions are generated corresponding to various environmental parameters, and environmental characteristic data of each monitoring area are formed; Conduct similarity analysis on the environmental characteristic data of the monitoring area, mark the areas with expected similarity and geographically continuous, and use each group of marked continuous areas as a feasible prediction area group; In a preset historical time period, the salinity and alkali content data of the feasible prediction area group are extracted and serialized, and the serialized data are imported into the LSTM prediction model for training. A monitoring area is selected in each feasible prediction area group as the sampling area; In a real-time monitoring cycle, soil salinity and alkalinity are tested in the sampling area, and the test data are imported into the LSTM prediction model to perform numerical prediction of salinity and alkalinity content, and the predicted salinity and alkalinity data of the feasible prediction area group are obtained; By predicting salinity data and historical salinity data, fertilization and irrigation plans are set for multiple monitoring areas, and intelligent plans are recommended based on different monitoring areas.

2. A saline-alkali land fertilization method based on intelligent recommendation according to claim 1, characterized in that: The step of obtaining preset saline-alkali land area information and constructing a map model based on the saline-alkali land through the area information is as follows: Obtaining preset saline-alkali land area information, the area information including area, area outline, planting range and monitoring point location information; A two-dimensionally visualized map model is constructed using the regional information.

3. The saline-alkali land fertilization method based on intelligent recommendation according to claim 1, characterized in that: In the map model, the preset saline-alkali land is divided into multiple monitoring areas by preset monitoring points, and the soil environmental data of the monitoring areas are obtained, specifically: In the map model, obtain the location of the preset monitoring point; Based on the locations of preset monitoring points, the preset saline-alkali land is divided into regions to form multiple monitoring areas, ensuring that each monitoring area corresponds to a preset monitoring point; Soil environmental data of each monitoring area are collected within a preset historical time period.

4. A saline-alkali land fertilization method based on intelligent recommendation according to claim 3, characterized in that: In the soil environmental data, feature vectors of multiple dimensions are generated corresponding to multiple environmental parameters, and environmental feature data of each monitoring area is formed, specifically: In the soil environmental data, each environmental parameter is used as a dimension, and the corresponding parameter value is used as the value of the corresponding dimension to generate a multi-dimensional feature vector; The multi-dimensional feature vector is used as the environmental feature data, and features are extracted for each monitoring area to form the environmental feature data of each monitoring area.

5. A saline-alkali land fertilization method based on intelligent recommendation according to claim 4, characterized in that: The similarity analysis is performed on the environmental characteristic data of the monitoring area, and the areas with expected similarity and continuous geographical location are marked, and each group of marked continuous areas is used as a feasible prediction area group, specifically: The similarity calculation is performed on the environmental characteristic data of the monitoring area. The calculation process is to calculate the environmental characteristic data of the two selected monitoring areas based on the standard Euclidean distance method, and obtain the environmental similarity between the two monitoring areas through the distance value; By presetting a similarity threshold, it is determined whether every two monitoring areas are similar, and all monitoring areas are determined, and the monitoring areas with similarity are marked to form a similar monitoring group. Based on all monitoring areas, multiple similar monitoring groups are generated; A similar monitoring group includes at least two monitoring areas; In similar monitoring groups, ensure that each monitoring area has similarity with at least one monitoring area in the group; In a similar monitoring group, geographically continuous monitoring areas are screened, and a new group is formed based on the screening results, which is marked as a feasible prediction area group; All similar monitoring groups are screened and marked to form multiple feasible prediction area groups.

6. A saline-alkali land fertilization method based on intelligent recommendation according to claim 5, characterized in that: In a preset historical time period, the salinity and alkali content data of the feasible prediction area group are extracted and serialized, and the serialized data are imported into the LSTM prediction model for training, and a monitoring area is selected as the sampling area in each feasible prediction area group, specifically: A feasible forecast area group was used as the unit of analysis; Acquiring salinity and alkali content data of all monitoring areas in the feasible prediction area group within a preset historical time period; Sorting the salinity and alkali content data based on the geographical continuity order between monitoring areas in a feasible prediction area group, and serializing the sorted data to form serialized data; Build an LSTM prediction model and initialize parameters and set loss function; The serialized data is divided into a test set and a training set based on a preset ratio and imported into the LSTM prediction model for cyclic prediction training. During the training process, the model parameters are optimized through the loss function until the prediction accuracy reaches the preset standard. Record the specific model parameters of the LSTM prediction model at this time, and associate the model parameters with the one feasible prediction area group; In each feasible prediction area group, a monitoring area at a starting point location is selected as the sampling area based on the continuity of the geographical location.

7. A saline-alkali land fertilization method based on intelligent recommendation according to claim 6, characterized in that: In a real-time monitoring cycle, soil salinity and alkalinity detection is performed on the sampling area, and the detection data is imported into the LSTM prediction model to perform numerical prediction of salinity and alkali content, and the predicted salinity and alkali data of the feasible prediction area group are obtained, specifically: A feasible forecast area group was used as the unit of analysis; In a real-time monitoring cycle, soil salinity and alkalinity detection is performed on a sampling area in a feasible prediction area group to obtain detection data; Import the detection data into the LSTM prediction model to perform numerical prediction of salinity and alkali content and obtain prediction sequence data; The predicted sequence data are analyzed, and the data are matched to each monitoring area in a feasible prediction area group to obtain the predicted salinity data of each monitoring area in the group.

8. The saline-alkali land fertilization method based on intelligent recommendation according to claim 7, characterized in that: By predicting salinity data and historical salinity data, fertilization and irrigation plans are set for multiple monitoring areas, and intelligent plans are recommended based on different monitoring areas, specifically: In a feasible prediction area group, the salinity and alkali changes in the area are evaluated through the detection data and predicted salinity and alkali data of the sampling area, and the salt content control index and alkalinity control index of the soil are generated; Analyzing fertilization and irrigation plans based on the soil salt content control index and alkalinity control index, and generating a predicted regional fertilization plan; According to multiple feasible forecast area groups, fertilization plans corresponding to multiple forecast areas are generated; In the map model, all monitoring areas outside the feasible prediction area group are marked as non-prediction areas. Real-time salinity and alkali data detection is carried out on the non-prediction areas. Combined with historical salinity and alkali content data, matching selection analysis is carried out in the pre-fertilization schemes to screen out a variety of real-time fertilization schemes. Based on multiple predicted regional fertilization plans and multiple real-time fertilization plans, intelligent plans are recommended for preset saline-alkali areas.

9. A saline-alkali land fertilization system based on intelligent recommendation, characterized in that: The system includes: a memory and a processor, wherein the memory includes a saline-alkali land fertilization program based on intelligent recommendation, and the saline-alkali land fertilization program based on intelligent recommendation implements the following steps when executed by the processor: Obtain preset saline-alkali land area information, and construct a map model based on the saline-alkali land through the area information; In the map model, the preset saline-alkali land is divided into multiple monitoring areas by presetting monitoring points, and the soil environmental data of the monitoring areas are obtained; In the soil environmental data, feature vectors of multiple dimensions are generated corresponding to various environmental parameters, and environmental characteristic data of each monitoring area are formed; Conduct similarity analysis on the environmental characteristic data of the monitoring area, mark the areas with expected similarity and geographically continuous, and use each group of marked continuous areas as a feasible prediction area group; In a preset historical time period, the salinity and alkali content data of the feasible prediction area group are extracted and serialized, and the serialized data are imported into the LSTM prediction model for training. A monitoring area is selected in each feasible prediction area group as the sampling area; In a real-time monitoring cycle, soil salinity and alkalinity are tested in the sampling area, and the test data are imported into the LSTM prediction model to perform numerical prediction of salinity and alkalinity content, and the predicted salinity and alkalinity data of the feasible prediction area group are obtained; By predicting salinity data and historical salinity data, fertilization and irrigation plans are set for multiple monitoring areas, and intelligent plans are recommended based on different monitoring areas.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium includes a saline-alkali land fertilization program based on intelligent recommendation. When the saline-alkali land fertilization program based on intelligent recommendation is executed by a processor, the steps of the saline-alkali land fertilization method based on intelligent recommendation as described in any one of claims 1 to 8 are implemented.

Images