An intelligent agricultural production data processing method based on artificial intelligence

Abstract

The application discloses an intelligent agricultural production data processing method based on artificial intelligence and relates to the technical field of intelligent agriculture, and comprises the following steps: standard characteristic vector set construction is carried out based on agricultural production data, and a standard characteristic vector set is obtained; farmland region abnormal state detection is carried out according to the standard characteristic vector set, and an abnormal detection result is obtained; clustering analysis is carried out according to the abnormal detection result, and an abnormal region division result is obtained; abnormal triggering factor analysis is carried out according to the abnormal region division result, and a cause-effect variable set is obtained; abnormal triggering reason reasoning is carried out according to the cause-effect variable set, and a cause-effect reasoning result is obtained; and key factor screening is carried out according to the cause-effect reasoning result, and a farmland key factor set is obtained. The application enhances the physical rationality and dynamic adaptability of path reasoning, improves the adaptability and practicability of the model, and improves the accuracy of abnormal triggering reason reasoning.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of intelligent agriculture, in particular to an intelligent agricultural production data processing method based on artificial intelligence. BACKGROUND

[0002] With the continuous advancement of precision agriculture and intelligent agriculture, a large amount of environmental, crop, physiological, biochemical and other data in the agricultural production process are collected in real time through multi-source sensors, remote sensing satellites and agricultural Internet of Things devices. These heterogeneous, multi-scale and high-dimensional agricultural production data provide a basis for realizing intelligent agricultural management and data-driven decision support. However, the agricultural production system itself has complex nonlinear dynamic characteristics and spatiotemporal coupling characteristics, which makes it difficult to directly infer the causes of abnormal states such as diseases, pests, water stress and soil degradation from the data.

[0003] In the prior art, there are deficiencies in abnormal state cause reasoning: the existing abnormal state cause reasoning methods mostly use graph models based on structure learning, such as PC algorithm, GES, LiNGAM, etc., or use Granger causality models based on time series analysis, which lack the characterization of dynamic indicators such as disturbance propagation intensity and path influence in causal path reasoning, and cannot reflect how variable disturbance is truly transmitted to abnormal states, and cannot adapt to the characteristics of high dynamicity and regional heterogeneity of agricultural systems. SUMMARY

[0004] In view of the deficiencies in the prior art, the present application provides an intelligent agricultural production data processing method based on artificial intelligence to solve the problems raised in the background.

[0005] To achieve the above purpose, the present application provides the following technical solutions:

[0006] In a first aspect, the present application provides an intelligent agricultural production data processing method based on artificial intelligence, comprising the following steps:

[0007] S1, constructing a standard feature vector set based on agricultural production data to obtain a standard feature vector set;

[0008] S2, detecting an abnormal state of a farmland area according to the standard feature vector set to obtain an abnormal detection result;

[0009] S3, performing clustering analysis according to the abnormal detection result to obtain an abnormal area division result;

[0010] S4, performing abnormal cause factor analysis according to the abnormal area division result to obtain a causal variable set;

[0011] S5, performing abnormal cause reasoning according to the causal variable set to obtain a causal reasoning result;

[0012] S6, performing key factor screening according to the result of the causal reasoning to obtain a set of key factors of the farmland.

[0013] Further optimization of the technical solution, the step S2 includes:

[0014] According to the obtained standard feature vector set, the abnormal state of the farmland area is detected, first, spatial representation transformation is performed to describe the abnormal performance mode in the space of the agricultural production scene, then dynamic time gating is performed to adjust the response degree of the model to the current history, then spatiotemporal fusion representation generation is performed to integrate the spatial propagation features and the time dynamic response at the current time, and finally classification output is performed to judge the abnormal prediction probability of each farmland grid unit to obtain the abnormal detection result.

[0015] Further optimization of the technical solution, the spatial representation transformation includes:

[0016] ;

[0017] Wherein:

[0018] : spatial adjacency matrix;

[0019] : standard feature vector matrix at time t, the matrix dimension is MxN, N is the feature dimension number of each farmland grid unit;

[0020] : spatial weight transformation matrix;

[0021] : spatial embedding representation matrix at time t;

[0022] : activation function.

[0023] Further optimization of the technical solution, the dynamic time gating includes:

[0024] ;

[0025] Wherein:

[0026] : global dynamic time gating vector at time t,

[0027] : time dynamic weight matrix;

[0028] : bias vector;

[0029] : average function;

[0030] sigmoid activation function

[0031] Further optimization of the technical solution, the spatio-temporal fusion representation generation includes:

[0032] ;

[0033] Wherein:

[0034] : the fused spatio-temporal feature representation matrix;

[0035] : M-dimensional all-1 vector;

[0036] : fusion weight hyperparameters;

[0037] : replication of the time gating vector.

[0038] Further optimization of the technical solution, the classification output includes:

[0039] ;

[0040] Wherein:

[0041] : the classification result output of the farmland grid unit at time t;

[0042] : classification weight matrix;

[0043] : Softmax conversion is performed on the output vector of each farmland grid unit.

[0044] Further optimization of the technical solution, the abnormal trigger factor analysis in S4 includes:

[0045] According to the obtained abnormal area division result, the artificial perturbation is introduced to the agricultural characteristic variables in the abnormal area, the slight fluctuation of the agricultural characteristic variables is simulated, and the causal impact scoring model is used to analyze the abnormal trigger factor, finally the most causal impact of the abnormal trigger factor is screened out, and the causal variable set of each region is constituted.

[0046] Further optimization of the technical solution, the causal impact scoring model includes:

[0047] ;

[0048] Wherein:

[0049] : the causal score of the nth feature of the jth anomaly region;

[0050] : the anomaly probability of the jth anomaly region before introducing the artificial perturbation;

[0051] : the anomaly probability of the jth anomaly region after introducing the artificial perturbation.

[0052] Further optimization of the technical solution, the anomaly probability comprises:

[0053] ;

[0054] ;

[0055] Wherein:

[0056] : perturbation strength;

[0057] : the nth feature variable of the jth anomaly region;

[0058] : the time step of the time series;

[0059] : the jth anomaly region;

[0060] : the perturbation feature data matrix of the jth anomaly region;

[0061] : the anomaly region original feature data matrix of the jth anomaly region;

[0062] : the anomaly probability function.

[0063] Further optimization of the technical solution, the perturbation feature data matrix comprises:

[0064] ;

[0065] Wherein:

[0066] : the nth dimensional feature variable vector.

[0067] In a second aspect, the embodiments of the present application provide a computer device, comprising a memory and a processor, and the memory stores a computer program, wherein the computer program instructions are executed by the processor to realize the steps of the intelligent agricultural production data processing method based on artificial intelligence according to the first aspect of the present application.

[0068] In a third aspect, an embodiment of the present application provides a computer readable storage medium having stored thereon a computer program, wherein the computer program instructs a processor to implement the steps of the artificial intelligence-based smart agricultural production data processing method according to the first aspect of the present application.

[0069] Compared with the prior art, the present application provides an artificial intelligence-based smart agricultural production data processing method, which has the following beneficial effects:

[0070] The artificial intelligence-based smart agricultural production data processing method uses machine learning and deep learning to simulate the system response after the disturbance of causal variables, forms the response chain between variables in time sequence and space, enhances the physical rationality and dynamic adaptability of path reasoning, models each land separately, matches geographical heterogeneity, is more suitable for field-level management, improves the adaptability and practicality of the model, and models the variable disturbance transmission, improves the accuracy of abnormal cause reasoning. BRIEF DESCRIPTION OF DRAWINGS

[0071] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings needed in the embodiment description will be briefly introduced as follows. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor.

[0072] Fig. 1 A flowchart of the artificial intelligence-based smart agricultural production data processing method according to the present application is shown.

[0073] Fig. 2 A flowchart of the abnormal state detection of the artificial intelligence-based smart agricultural production data processing method according to the present application is shown.

[0074] Fig. 3 A flowchart of the causal influence scoring model of the artificial intelligence-based smart agricultural production data processing method according to the present application is shown.

[0075] Fig. 4 A flowchart of the dynamic disturbance transmission causal chain model of the artificial intelligence-based smart agricultural production data processing method according to the present application is shown. DETAILED DESCRIPTION

[0076] In order to make the above-mentioned objects, features and advantages of the present application more apparent and easy to understand, the specific embodiments of the present application will be described in detail below with reference to the drawings.

[0077] In the following description, numerous specific details are set forth in order to provide a thorough understanding of the present application. However, it will be appreciated that the present application can be practiced in a variety of ways beyond the specifics set forth herein, which can be practiced in any number of manners, depending on the implementation chosen for the present application. The present application can be practiced by employing as many or as few of the specific details described herein as desired or appropriate. Accordingly, the present application is not intended to be limited by the specific embodiments described below, which are given by way of example only.

[0078] Second, the "one embodiment" or "an embodiment" referred to herein means a specific feature, structure, or characteristic under discussion. The phrase "in one embodiment" does not necessarily refer to the same embodiment, although it can. The phrase "in one embodiment" is used intermittently throughout this specification to convey one feature, structure, or characteristic of different embodiments.

[0079] Embodiment One:

[0080] Reference Figs. 1-4 For the first embodiment of the present application, the embodiment provides an intelligent agricultural production data processing method based on artificial intelligence, comprising the following steps:

[0081] S1, constructing a standard feature vector set based on agricultural production data, to obtain a standard feature vector set.

[0082] In this embodiment, the construction of the standard feature vector set includes:

[0083] In the processing of agricultural production data, raw data from different sources usually have different formats, scales, time intervals, spatial distributions, etc. This data heterogeneity makes subsequent data analysis and modeling complex. In order to solve this problem, the core purpose of this step is to convert heterogeneous data sources into a unified standard feature vector set, which is convenient for data analysis and modeling.

[0084] The goal of this step is to convert raw agricultural production data (such as meteorological data, crop growth data, soil data, etc.) into a standardized input format. Raw agricultural production data usually comes from multiple different sources, including multispectral images (image data used to monitor crop growth status, often stored in pixel form, containing red, green, blue, and near-infrared band image data), meteorological data (including temperature, humidity, wind speed, precipitation, etc. Real-time collected environmental information, usually continuous time series data), soil data (including soil conductivity, humidity, pH value, water content, nutrient element (nitrogen, phosphorus, potassium, etc.) concentration, etc.), crop growth data (such as chlorophyll index, crop vigor, aboveground and underground growth conditions, etc.) and other data.

[0085] The different data sources have different frequencies and formats in time collection. By selecting a uniform time scale (such as every hour or every day), the timestamps of all data are unified so that they can be aligned on the same time axis. For those data with different intervals, linear interpolation or spline interpolation (such as Kriging interpolation) is used to fill in the missing data points, so that all data are complete and consistent on the unified time axis. Different agricultural production data usually have different units and dimensions (such as soil moisture is a percentage, and temperature is in Celsius). In order to eliminate these differences, Z-score normalization (subtracting the mean and dividing by the standard deviation) or Min-Max normalization (scaling the data value to 0 to 1) is used for feature standardization, so that the range of all data is compressed to the same scale, avoiding the influence of some features due to the large dimension. Different agricultural production data often have spatial distribution characteristics, such as soil moisture, crop growth, etc. Data are usually collected according to geographical coordinates. In order to perform unified analysis, the data need to be mapped to a unified spatial grid. The farmland is usually divided into grid units (for example, 10m x 10m grid units). The data in each grid unit are combined by interpolation method. For multi-spectral image data, the pixel data are spatially aligned according to the geographical coordinates. Each pixel in the image is mapped to the farmland grid. For other data types (such as weather data, soil data), Kriging interpolation or nearest neighbor interpolation is usually used for spatial interpolation to fill in the grid units without data. In agricultural production, data loss is inevitable, which may be caused by instrument failure, weather interference or human operation error. In order to ensure that the model can handle these incomplete data, an effective missing value processing method is needed. Interpolation method is usually used to fill in the missing values, such as linear interpolation or Kriging interpolation for missing weather data or soil data. These methods can estimate the missing values according to the trend of known data, so as to avoid the influence of missing values on model training.

[0086] After synchronizing and standardizing the original agricultural production data, all data are converted into a unified feature vector. The composition of the feature vector of a farmland grid unit may include weather data, soil data, crop growth data, multi-spectral image data, etc. These multi-dimensional data are converted into a standard feature vector, and finally a standardized feature set of each grid unit is obtained, which includes the standardized data of features at different time steps or from different sources.

[0087] S2, detecting the abnormal state of the farmland area according to the standard feature vector set to obtain an abnormal detection result.

[0088] In this embodiment, the farmland area abnormal state detection includes:

[0089] The standard feature vector set obtained from step S1 has standardized and spatially gridded multi-source heterogeneous data such as meteorological data, crop growth status, soil indicators, and multispectral images. The occurrence of abnormal states, such as crop diseases, soil drought, and waterlogging, is often influenced by spatial neighborhood (e.g., adjacent land spread) and temporal dynamics (e.g., continuous rainfall or high temperature). This step detects abnormal states in farmland areas based on the obtained standard feature vector set. First, a spatial representation transformation is performed to describe the spatial abnormality patterns in agricultural production scenarios, such as disease transmission paths and soil abnormalities. Then, a dynamic time gating is performed to adjust the response degree of the model to historical influences, to some extent, to avoid time redundancy feature interference. Next, a spatio-temporal fusion representation is generated to integrate the spatial propagation features and temporal dynamic responses at the current time. Finally, a classification output is performed to judge the abnormal prediction probability of each farmland grid cell and obtain the abnormal detection result.

[0090] Further, the spatial representation transformation includes:

[0091] ;

[0092] Wherein:

[0093] : Spatial adjacency matrix, representing the topological connection relationship between any two farmland grid cells, with a value of 1 indicating the existence of spatial adjacency, and a value of 0 indicating that the farmland grid cell is not spatially adjacent to itself, obtained from the geographical adjacency relationship between farmland grid cells, with a matrix dimension of MxM, where M is the number of farmland grid cells;

[0094] : Standard feature vector matrix at time t, with a matrix dimension of MxN, where N is the dimension number of the feature vector of each farmland grid cell;

[0095] : Spatial weight transformation matrix, used to map the feature vector of the farmland grid cell to a spatial representation of H dimensions. The size of the dimension number H is set by the user, with the purpose of improving the model's ability to express complex spatio-temporal features, providing sufficient dimensions to support nonlinear learning and controlling the model capacity, avoiding insufficient expression caused by low dimensions, or overfitting caused by high dimensions. The typical setting range is 8-128 dimensions, with a matrix dimension of NxH;

[0096] : the spatial embedding representation matrix of time t, represents the spatial representation of each farmland grid cell after neighborhood propagation and feature compression, the matrix dimension is MxH, the matrix integrates the features of itself and multiple farmland grid cells that exist in spatial adjacency, realizes information fusion in the spatial dimension, so that when the surrounding farmland grid cells appear abnormal and cause the features to change, causing the change of the matrix, even if the current farmland grid cell does not appear abnormal, the abnormality of the surrounding farmland grid cell can also be found through the matrix, the abnormal risk is pre-judged in advance, early abnormal perception is realized, for example, when the features of the surrounding farmland grid cells of a farmland grid cell change, the spatial embedding representation matrix of the farmland grid cell will also change;

[0097] : activation function, represents max(0, x), retains the original input in the positive interval, and outputs 0 in the negative interval, which is used to increase the nonlinear capability.

[0098] Further, the dynamic time gate includes:

[0099] A global time gate mechanism is established to extract the overall trend of all spatial units at the current time, so as to control the response degree of the subsequent model to the current features;

[0100] ;

[0101] Wherein:

[0102] : the global dynamic time gate vector of time t, represents the dynamic state of the entire agricultural area at time t, the vector dimension is H;

[0103] : time dynamic weight matrix, used to map the N-dimensional average feature vector to the H-dimensional gate representation, the matrix dimension is HxN;

[0104] : bias vector, the vector dimension is H;

[0105] : average function, used to average the feature vectors of all farmland grid cells at time t, reflecting the overall state feature, , represents the N-dimensional feature vector of the i-th farmland grid cell at time t, by dimension-by-dimension averaging of the feature vectors of all farmland grid cells, an N-dimensional average feature vector is obtained;

[0106] : Sigmoid activation function, used to map the gate value to [0, 1].

[0107] Further, the spatio-temporal fusion representation generation comprises:

[0108] ;

[0109] wherein:

[0110] : the fused spatio-temporal feature representation matrix, the matrix dimension is M x H;

[0111] : M-dimensional all-1 vector, used for copy operation;

[0112] : fusion weight hyperparameter, which can be set as a fixed value, used to adjust the importance of time and space information;

[0113] : copy of the time gating vector, which means copying the time gating vector with a dimension of 1 x H M times, so as to correspondingly fuse it with each farmland grid cell.

[0114] Further, the classification output comprises:

[0115] ;

[0116] wherein:

[0117] : the classification result output of the farmland grid cell at time t, i.e., the probability of each classification category, the matrix dimension is M x C, C is the number of categories (for example, if the classification categories are normal and abnormal, then C is 2), the i-th row is the classification result vector of the i-th farmland grid cell at time t;

[0118] : classification weight matrix, the matrix dimension is H x C;

[0119] : Softmax conversion is performed on the output vector of each farmland grid cell, so that it satisfies the probability distribution, i.e., the sum of each row element is 1.

[0120] The model describes how to detect the abnormal state of the farmland area according to the standard feature vector set.

[0121] Traditional anomaly detection methods usually ignore the spatial and temporal dependencies, often only considering single-dimensional data features, while agricultural production data has significant spatio-temporal dependencies, and soil moisture, crop growth and other features change over time and have strong local correlation in space. Therefore, traditional anomaly detection methods may result in insufficient detection accuracy, while the present model can simultaneously capture the spatial dependency and temporal variation characteristics of multi-dimensional data in agricultural production, thereby significantly improving the accuracy and real-time performance of anomaly detection.

[0122] The steps of using this model include:

[0123] Data acquisition: obtain the standardized and spatio-temporal gridded data from the standard feature vector set obtained in step S1;

[0124] Spatio-temporal feature fusion: according to the obtained data, calculate the spatial representation of the farmland area and the time gate vector , and fuse them to obtain the spatio-temporal feature representation matrix ;

[0125] Abnormality detection: according to the calculated spatio-temporal features, perform abnormality detection on each farmland grid cell to obtain the classification result of each farmland grid cell, for example, if the classification categories are normal and abnormal, then the probability of each farmland grid cell being normal and abnormal is represented as , which represents that the probability of the 23rd farmland grid cell being determined as an abnormal state at time t is 85%, and the probability of being a normal state is 15%. Set an abnormal threshold, compare the obtained abnormal probability with the abnormal threshold, and if the abnormal probability is greater than the threshold, determine that this farmland grid cell is abnormal.

[0126] S3, according to the abnormality detection result, perform clustering analysis to obtain the abnormal area division result.

[0127] In this embodiment, the clustering analysis includes:

[0128] In step S2, it is determined whether each farmland grid cell is in an abnormal state, but the specific abnormality generated by each farmland grid cell is not clear. Therefore, the goal of this step is to divide the farmland grids with unknown abnormal states into non-overlapping areas on the space according to the abnormal types, and to clearly define the distribution boundaries of each type of problem, providing a spatial semantic basis for subsequent problem cause reasoning.

[0129] This step performs spatial clustering analysis based on the density clustering algorithm DBSCAN (Density-Based Spatial Clustering of Applications with Noise) to obtain the abnormal area division result. This algorithm does not need to preset the number of categories, can handle spatial noise points (i.e. identify "isolated anomalies"), has flexible clustering shape, is suitable for complex field shapes (non-rectangular, irregular boundaries, etc.) in real farmland, and can naturally form spatially continuous abnormal areas.

[0130] According to the abnormality detection result obtained in step S2, a set of abnormal units is extracted from the abnormality detection result, the abnormal units are mapped to their two-dimensional spatial coordinates (the spatial coordinates can be derived from the center point of a plot, a remote sensing pixel coordinate, a latitude and longitude, or a plane coordinate system), a coordinate set is formed, then density clustering parameters are set based on the actual plot size and the remote sensing image resolution, including a neighborhood radius (for controlling the spatial perception range) and a minimum number of neighborhood points (for controlling the clustering threshold), then a clustering operation is performed, the set of abnormal units is input, and a density clustering algorithm is run to obtain a clustering result, the clustering result is a plurality of non-overlapping abnormal regions, and finally the boundary of each abnormal region is calculated, a boundary envelope is extracted, an a-Shape or Convex Hull method (for a small number of points) is used, or an isogram method (for a large number of points) is used, so as to obtain the boundary contour line of each abnormal region, and finally, according to the abnormal region data and the boundary contour line data obtained by the clustering analysis, a complete abnormal region division result is obtained.

[0131] S4. Abnormality causing factor analysis is performed according to the abnormal region division result to obtain a set of causal variables.

[0132] In this embodiment, the abnormality causing factor analysis includes:

[0133] According to the obtained abnormal region division result, a slight disturbance is introduced to the agricultural features in the abnormal region to simulate slight fluctuations of the agricultural features, a causal impact scoring model is used to quantify the impact on the abnormality determination output, abnormality causing factor analysis is performed, and finally the abnormality causing factors with the strongest causal impact are screened out to form a set of causal variables for each region.

[0134] Further, the causal impact scoring model includes:

[0135] ;

[0136] Wherein:

[0137] : the causal score of the nth feature of the jth abnormal region, the greater the value, the stronger the causal relationship between the feature vector and the abnormality causing factor;

[0138] : the abnormal probability of the jth abnormal region before the introduction of the artificial disturbance, which is obtained by calculating the region abnormal probability in step S2;

[0139] : the abnormal probability of the jth abnormal region after the introduction of the artificial disturbance.

[0140] Further, the abnormal probability includes:

[0141] ;

[0142] ;

[0143] wherein:

[0144] : perturbation strength, representing the strength of the introduced artificial perturbation;

[0145] : the nth feature of the jth anomaly region;

[0146] : time step of the time series;

[0147] : the jth anomaly region;

[0148] : perturbation feature matrix of the jth anomaly region, representing the feature vector matrix constructed after the introduction of artificial perturbation, the matrix dimension is T x N, T is the length of the time series data, the time series selected should be able to cover the change process of the anomaly from nothing to something, N is the feature dimension number of the anomaly region;

[0149] : anomaly region original feature matrix of the jth anomaly region, i.e. the feature vector matrix constructed before the introduction of artificial perturbation, which includes the feature vector of T time steps and N dimensions of the anomaly region, the matrix dimension is T x N;

[0150] : anomaly probability function, representing the output of the classification result obtained in step S2, wherein represents the probability of the region appearing abnormal. Since the anomaly region is composed of multiple farmland grid cells, the original feature matrix of the anomaly region and the perturbation feature matrix can be converted into an M x N dimensional standard feature vector matrix at multiple time steps, so as to use the model in step S2 to calculate the anomaly probability.

[0151] Further, the perturbation feature data matrix comprises:

[0152] ;

[0153] wherein:

[0154] : the nth feature vector, obtained by the region standard feature vector set in step S1.

[0155] The model describes how to analyze the influence of the feature variable on the abnormality by perturbing the feature variable.

[0156] The traditional method for analyzing abnormal triggering factors is generally only applicable to large sample scenarios with stable variables and stable structural relationships, has problems of being insensitive to complex nonlinear relationships, being difficult to capture local regional causal characteristics, having low accuracy in judging in a high-noise data missing environment, outputting results being mostly abstract structural diagrams or statistical indicators, being difficult to directly understand, and the like, while the model supports complex nonlinear relationships among multiple variables in agricultural production data, and can be used for agricultural data with less samples and more variable noise, independently evaluates causal variables for each abnormal region, improves adaptability and judgment accuracy, and can obtain a score for each causal variable, facilitating intuitive interpretation and understanding, and improving practicality.

[0157] The use steps of the above model include:

[0158] Disturbance construction: according to the abnormal region obtained in step S3, extracting the feature variable matrix thereof , and injecting a perturbation into the feature variable , constructing a perturbed feature matrix ;

[0159] Abnormal probability calculation: according to the constructed perturbed feature matrix, using the model in step S2, calculating the abnormal probability after injecting the perturbation ;

[0160] Causal impact score: comparing and calculating the abnormal probability after injecting the perturbation and the original abnormal probability , obtaining a causal score . The greater the value of the causal score, the stronger the causal impact, indicating that the causal relationship between the feature variable and the abnormal triggering is stronger, and the feature variable factor is more likely to be a key factor for triggering the abnormality. A causal score threshold is set, and a causal variable set is formed by combining feature variables with a causal score greater than the threshold. For example, there is an abnormality in a certain block, and the causal variables obtained are soil humidity, vegetation index NDVI, and pest index. The abnormality of the block may be related to continuous decline in soil humidity, decline in vegetation index NDVI, and sharp rise in pest index.

[0161] S5, according to the causal variable set, reasoning the abnormal triggering cause to obtain a causal reasoning result.

[0162] In this embodiment, the reasoning of the abnormal triggering cause includes:

[0163] The abnormality-causing variable set obtained in step S4 determines the variable factors causing the abnormality. In this step, the dynamic disturbance transmission causal chain model is used to perform abnormality-causing reason reasoning based on the causal variable set, to analyze and determine how the variables in the causal variable set cause the abnormality, and to obtain the causal reasoning result of each region. The causal reasoning result includes the abnormality-causing causal chain and the causal chain score of each region, which explains the reason why the variables cause the abnormality. The causal chain includes the starting characteristic variable, the intermediate characteristic variable, and the key characteristic variable that lead to the abnormal state. The starting characteristic variable is the source of the causal path and does not directly cause the abnormality, but has an exciting effect on the intermediate characteristic variable. The intermediate characteristic variable does not directly cause the abnormality, but plays a role in amplifying, conducting, or cooperating. The key characteristic variable is close to the abnormal result itself, or its fluctuation directly causes the abnormality. The key characteristic variable that causes the abnormality is obtained through the causal score of the abnormal region characteristic, and the intermediate characteristic variable between the starting variable and the key characteristic variable is reasoned step by step based on the disturbance transmission strength between different characteristic variables, until the starting characteristic variable that leads to the abnormal state is found, thereby forming a complete causal chain.

[0164] Further, the dynamic disturbance transmission causal chain model includes:

[0165] ;

[0166] Wherein:

[0167] : Causal chain score, used to judge the influence degree of the path on the abnormal state. The higher the score, the greater the influence degree.

[0168] : A path chain from a characteristic variable to an abnormal node, for example, soil temperature change affects soil humidity change, which in turn affects the pest index, and finally leads to abnormality.

[0169] : Disturbance transmission strength, reflecting the influence strength of variable u on variable v.

[0170] : Causal score of variable v, obtained by the model in step S4.

[0171] Further, the disturbance transmission strength includes:

[0172] ;

[0173] Wherein:

[0174] : Disturbance transmission strength, reflecting the influence strength of variable p on variable q.

[0175] : value of variable q at time t in region j

[0176] : apply perturbation to variable p in region j : value of variable q at time t+1

[0177] : length of time series data, for example, 24 hours for hourly sampling in agricultural data, T=24.

[0178] The model describes how to infer the cause of the anomaly by analyzing the influence between variables.

[0179] Traditional anomaly cause inference usually uses methods based on time series prediction or structural equation modeling, which are difficult to depict the complex coupling of agricultural multivariate, not suitable for high-dimensional, nonlinear disturbance systems, and not suitable for field-level, spatially heterogeneous agricultural environment, difficult to dynamically adjust and model the transmission process of disturbance, while the model can realize multi-source and multi-modal data integration, suitable for mixed scenarios of continuous variables, discrete variables and remote sensing indexes, and each field is modeled separately, matching geographical heterogeneity, more suitable for field-level management, improving the adaptability and practicality of the model, modeling the transmission of disturbance, improving the accuracy of anomaly cause inference, simulating the system response after the disturbance of causal variables, forming a real response chain between variables in time and space, breaking through the problem that traditional static causal graph cannot reflect the dynamic influence path of variables, enhancing the physical rationality and dynamic adaptability of path reasoning.

[0180] The use of the model includes:

[0181] Data acquisition: obtain the set of causal variables from step S4, get the variables that have an impact on the cause of the anomaly;

[0182] Perturbation transmission calculation: add a perturbation to each variable independently to simulate the response change of other variables, for example, if the air temperature rises by 1℃, whether the vegetation index NDVI fluctuates significantly, calculate the perturbation transmission strength between variables, calculate the causal score according to the causal influence scoring model of step S4, get the causal relationship size of the variable to the cause of the anomaly, according to the perturbation transmission, build a causal path chain, reflect the path from the variable to the cause of the anomaly, for example, vegetation index NDVI-> pest index-> anomaly, or air temperature-> soil moisture-> vegetation index NDVI-> anomaly, etc.

[0183] Abnormal reason inference: According to the calculated disturbance transmission intensity and causal score, the causal chain score of each causal path chain is calculated, and the higher the score, the greater the influence of the path on the abnormal state, and the more it reflects how the variable triggers the anomaly, for example, the causal chain of vegetation index NDVI-> pest index-> anomaly has a lower score than the causal chain of air temperature-> soil humidity-> vegetation index NDVI-> anomaly, indicating that the cause of the abnormality in this area may be temperature changes causing soil humidity changes leading to vegetation index NDVI changes ultimately triggering the anomaly.

[0184] S6, according to the causal inference result, screening the key factors to obtain a set of farmland key factors.

[0185] In this embodiment, the key factor screening includes:

[0186] According to the causal inference result obtained in step S5, a disturbance causal graph with direction and weight is obtained, the node set represents the causal variables, and the edge set represents the disturbance path. The structural importance of each node in the graph is calculated using the PageRank algorithm. The higher the PageRank score of a variable, the more core it is in the entire causal propagation graph and the more likely it is to cause a series of abnormalities downstream. A score threshold is set to screen out nodes with scores greater than the threshold, i.e. the screened key factor variables, which constitute the set of farmland key factors. The variables in this set are the most worthy of attention factors for subsequent agricultural regulation, intervention or risk assessment.

[0187] Embodiment two:

[0188] The embodiment also provides a computer device suitable for the case of the intelligent agricultural production data processing method based on artificial intelligence, including a memory and a processor; the memory is used to store computer executable instructions, and the processor is used to execute the computer executable instructions to realize the intelligent agricultural production data processing method based on artificial intelligence as proposed in the above embodiment.

[0189] The embodiment also provides a storage medium having a computer program stored thereon, which is executed by a processor to realize the intelligent agricultural production data processing method based on artificial intelligence as proposed in the above embodiment.

[0190] The computer device can be a terminal, and the computer device includes a processor, a memory, a communication interface, a display screen and an input device connected by a system bus. The processor of the computer device is configured to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for running the operating system and the computer program in the non-volatile storage medium. The communication interface of the computer device is configured to perform wired or wireless communication with an external terminal. The wireless communication can be achieved by WIFI, an operator network, NFC (Near Field Communication) or other technologies. The display screen of the computer device can be a liquid crystal display screen or an electronic ink display screen. The input device of the computer device can be a touch layer overlaid on the display screen, or can be a key, a trackball or a touchpad arranged on the shell of the computer device, or can be an external keyboard, a touchpad or a mouse, etc.

[0191] If the functions are implemented in the form of software function units and sold or used as independent products, the functions can be stored in a computer readable storage medium. Based on this understanding, the technical solutions of the present application or the parts of the technical solutions that essentially contribute to the prior art or the parts of the technical solutions can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes a plurality of instructions for causing a computer device (which can be a personal computer, a server or a network device, etc.) to execute all or part of the steps of the embodiments of the present application. The storage medium mentioned above includes a U disk, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk and various media that can store program codes.

[0192] The logic and / or steps represented in the flowcharts or otherwise described herein, for example, can be considered as a list of executable instructions for implementing logic functions, which can be specifically embodied in any computer readable medium for use by or in conjunction with an instruction execution system, device or apparatus, such as a computer-based system, a system including a processor or other system that can fetch and execute instructions from the instruction execution system, device or apparatus. For the purpose of the present specification, the "computer readable medium" can be any device that can contain, store, communicate, propagate or transport programs for use by or in conjunction with the instruction execution system, device or apparatus or in conjunction with these instruction execution systems, devices or apparatuses.

[0193] More specific examples (a non-exhaustive list) of the computer-readable medium include the following: an electrical connection (electronic) having one or more wires, a portable computer diskette (magnetic), a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or Flash memory), an optical fiber device, and a portable compact disc read-only memory (CDROM). Additionally, the computer-readable medium can also be paper or another suitable medium upon which the program is printed, as the program can be electronically captured, for example, via an optical scanner, then compiled, interpreted, or otherwise processed, as necessary, and stored in a computer memory.

[0194] It should be understood that aspects of the application can be implemented in hardware, software, firmware or combinations thereof. In the embodiments described above, various steps or methods can be implemented, for example, in software or firmware stored in memory and executed by a suitable instruction execution system. For example, if implemented in hardware, and in another embodiment, any of the following technologies, known in the art, or combinations thereof, can be used: discrete logic circuitry having logic gates for implementing logic functions upon an application of data signals, application specific integrated circuits having appropriate combinational logic gates, programmable gate arrays (PGA), field programmable gate arrays (FPGA), and the like.

[0195] It should be noted that the above examples are merely intended to illustrate the technical solutions of the present application and not to limit the same. Although the present application has been described in detail with reference to the preferred embodiments, those of ordinary skill in the art should understand that the technical solutions of the present application can be modified or replaced by equivalents without departing from the spirit and scope of the present application, and all such modifications or replacements should be included in the scope of the claims of the present application.

Claims

1. A smart agricultural production data processing method based on artificial intelligence, characterized in that, Includes the following steps: S1. Construct a standard feature vector set based on agricultural production data to obtain the standard feature vector set; S2. Detect anomalies in farmland areas based on the standard feature vector set, and obtain the anomaly detection results, specifically including: Based on the obtained standard feature vector set, anomaly detection is performed on farmland areas. First, spatial representation transformation is performed to characterize the spatial anomaly patterns in agricultural production scenarios. Then, dynamic time gating is performed to adjust the current response of the model to historical influences. Next, spatiotemporal fusion representation generation is performed to integrate the spatial propagation features and time dynamic response at the current moment. Finally, classification output is performed to judge the anomaly prediction probability of each farmland grid unit and obtain the anomaly detection result. S3. Perform cluster analysis based on the anomaly detection results to obtain the anomaly region division results; S4. Based on the results of the abnormal area division, conduct anomaly-causing factor analysis to obtain the causal variable set; S5. Based on the set of causal variables, perform reasoning on the causes of the anomalies to obtain the causal reasoning results; S6. Based on the results of causal reasoning, key factors are screened to obtain a set of key factors for farmland. The analysis of anomaly-inducing factors in S4 includes: Based on the obtained results of the abnormal region division, artificial perturbations were introduced into the agricultural characteristic variables in the abnormal regions to simulate slight fluctuations in the agricultural characteristic variables. Then, a causal impact scoring model was used to analyze the abnormal initiating factors. Finally, the abnormal initiating factors with the most causal impact were selected to form the causal variable set for each region. The causal impact scoring model includes: ； in: : The causal score of the nth feature of the jth anomaly region; : The probability of an anomaly in the j-th anomaly region before the introduction of artificial perturbation; : The probability of an anomaly after introducing artificial perturbation into the j-th anomaly region.

2. The method for processing smart agricultural production data based on artificial intelligence according to claim 1, characterized in that, The spatial representation transformation includes: ； in: Spatial adjacency matrix; The standard eigenvector matrix at time t has a dimension of M×N, where N is the number of eigenvector dimensions for each farmland grid cell. Spatial weight transformation matrix; : Spatial embedding representation matrix at time t; Activation function.

3. The method for processing smart agricultural production data based on artificial intelligence according to claim 2, characterized in that, The dynamic time gating includes: ； in: : The global dynamic time-gated vector at time t; Time-based dynamic weight matrix; Bias vector; Find the average function; : Sigmoid activation function.

4. The method for processing smart agricultural production data based on artificial intelligence according to claim 3, characterized in that, The spatiotemporal fusion representation generation includes: ； in: : The fused spatiotemporal feature representation matrix; : An M-dimensional vector of all 1s; : Fusion weight hyperparameter; Copying of time-gated vectors.

5. The method for processing smart agricultural production data based on artificial intelligence according to claim 4, characterized in that, The classification output includes: ； in: : The classification results of the farmland grid cells at time t; Classification weight matrix; : Perform a Softmax transformation on the output vector of each farmland grid cell.

6. The method for processing smart agricultural production data based on artificial intelligence according to claim 5, characterized in that, The anomaly probability includes: ； ； in: : Disturbance intensity; The nth feature of the j-th anomaly region; Time step in a time series; The j-th abnormal region; : The disturbance feature data matrix of the j-th anomalous region; : The original feature data matrix of the j-th abnormal region; : Anomaly probability function.

7. The method for processing smart agricultural production data based on artificial intelligence according to claim 6, characterized in that, The disturbance feature data matrix includes: ； in: : The nth dimension feature variable vector.

Images