Agricultural environment intelligent monitoring method and system based on multi-source information fusion

Abstract

The invention provides an agricultural environment intelligent monitoring method and system based on multi-source information fusion, and relates to the technical field of agricultural environment intelligent monitoring, and the method comprises the following steps: periodically collecting K kinds of monitoring data through a plurality of sensors at a period T, and forming K kinds of multi-source initial time sequence data sequences, each initial time sequence comprises N collected parameters ending the current; performing the following operations on each initial time sequence data sequence: identifying abnormal data points; extracting cultivation operation and time points of crops in the time range of the initial time sequence data sequence, and calculating an abnormal time sequence to be corrected based on the abnormal data points; performing internal influence filtering on the to-be-corrected abnormal time sequence to obtain a filtered time sequence data sequence; and after a multi-source filtering time sequence is obtained, carrying out environmental anomaly judgment through an environmental anomaly recognition model. According to the invention, interference caused by agricultural cultivation operation can be identified and corrected in agricultural environment monitoring based on various sensors.

Description

Technical Field

[0001] This invention relates to the field of intelligent monitoring technology for agricultural environment, and more specifically, to an intelligent monitoring method and system for agricultural environment based on multi-source information fusion. Background Technology

[0002] As modern agriculture develops towards precision and intelligence, agricultural environmental monitoring has become an important technical means to ensure stable crop growth and increase yield.

[0003] Generally, intelligent monitoring of the agricultural environment relies on multiple sensors for data collection. However, during agricultural production, manual operations such as watering, fertilizing, and spraying pesticides can significantly disrupt sensor data. For example, watering can cause drastic fluctuations in soil moisture within a short period, while fertilization can alter the electrical conductivity of the soil solution. If this operational interference is not effectively filtered, it can lead to misjudgments by environmental anomaly identification models, reducing monitoring accuracy. Conversely, directly discarding data from the operational phase can result in incomplete monitoring and missed detections.

[0004] Therefore, how to identify and correct interference caused by agricultural operations in agricultural environmental monitoring based on multiple sensors, thereby improving the accuracy of environmental anomaly identification, has become a technical problem that urgently needs to be solved in the field of intelligent agricultural monitoring. Summary of the Invention

[0005] The purpose of this invention is to provide an intelligent monitoring method and system for agricultural environment based on multi-source information fusion, which can identify and correct interference caused by agricultural and aquaculture operations in agricultural environment monitoring based on multiple sensors.

[0006] This invention is achieved through the following technical solution: A method for intelligent monitoring of the agricultural environment based on multi-source information fusion includes the following steps: K types of monitoring data are collected periodically by multiple sensors with period T, forming K initial time series data sequences from multiple sources. Each initial time series sequence includes N parameters collected up to the current time. Perform the following operations on each initial time series data sequence: Identify outlier data points; Extracting the initial time series data sequence for crops within a time range The planting and breeding operations and the corresponding time points of the breeding operations are used to calculate the abnormal time series sequence to be corrected based on the abnormal data points. Internal influence filtering is performed on the time series sequence to be corrected to obtain the filtered time series data sequence; After obtaining the multi-source filtered time series, environmental anomaly judgment is performed through the environmental anomaly identification model.

[0007] Preferably, the various monitoring data include ambient temperature, ambient humidity, soil moisture, soil conductivity, carbon dioxide concentration, and oxygen concentration.

[0008] Preferably, the method for obtaining the identified abnormal data points is as follows: Obtain the difference between adjacent data: ; in, For the k-th initial time series data sequence The nth parameter and the (n-1)th parameter The difference, , ; If the absolute value of the difference between adjacent data points is greater than a preset first difference threshold, then the data at the next time point is considered the abnormal data point.

[0009] Preferably, the method for calculating the abnormal time series sequence to be corrected based on abnormal data points is as follows: Get the The duration of the impact of the aquaculture operation described in type k on the monitoring data described in type k ; Determine the difference between the abnormal data point and the time point of the r-th type of aquaculture operation; if the difference is not... The internal method excludes the breeding operation described in the rth type. ; Get the The time points of the aforementioned breeding operations , The value range is for aquaculture operations that were not excluded, and the collection time of the abnormal data points is obtained. ; Get the The duration of the effects of the aforementioned aquaculture operations ; Get The maximum value is taken as the termination time of the comprehensive and lasting impact; In the corresponding initial time series data sequence, all parameters collected from the abnormal data point within the time of termination of the comprehensive continuous impact are formed to form the abnormal time series sequence to be corrected.

[0010] Preferably, the step of obtaining the first The duration of the impact of the aquaculture operation described in type k on the monitoring data described in type k The method involves extraction through the following experiments: At the point of time For crops, the r-th type of aquaculture operation is performed, and K types of monitoring data are collected periodically according to cycle T, starting from time point. A reference time series of data is set to record the monitoring data after the r-th type of aquaculture operation is performed. Each reference time series of data includes G parameters. Obtain the adjacent differences of the k-th reference time series data sequence: ; in, For the k-th reference time series data sequence The Parameters and the Parameters The difference, ; Obtain the adjacent differences of all reference time-series data sequences whose absolute values ​​are less than the preset second difference threshold, and sort them in time to form a time-series difference sequence; Perform the following iterations: Initialize q=1; If the difference between the corresponding acquisition time points of the q-th parameter and the (q+b-th parameter) of the time-series difference sequence is less than a preset time threshold... If the iteration stops, then stop; otherwise, update the value of q to q = q + 1. Let the value of q be the value when the iteration stops. , Let the first of the time series difference sequences be... The parameters are , Obtain the k-th reference time series data sequence. The acquisition time point of each parameter, and the time point relative to the time point The difference is the first Duration of the effects of the aforementioned aquaculture operations .

[0011] Preferably, the method for filtering the internal influence of the abnormal time series to be corrected is to perform the following operations for all k=1,2,…,K: Assignment The time points mentioned in the experiment The value of the kth type of monitoring data collected in the previous instance is used to obtain the first... The k-th reference time series data sequence under farming and aquaculture operations and the Number of sprays per unit area for planting and breeding operations , The value range is for aquaculture operations that were not eliminated; The time series sequence of the anomalies to be corrected for the kth type of monitoring data Make corrections: ; ; in, The abnormal time series sequence to be corrected The y-th parameter The corrected value, for The x-th parameter, where x is an intermediate parameter. For the actual time point Time The amount of spray applied per unit area in planting and breeding operations. The abnormal time series sequence to be corrected The time point at which the y-th parameter is collected, and the range of values ​​for y is... All parameters, The function representing rounding. Not less than And the difference does not exceed the preset threshold for the number of sprays per unit area.

[0012] Preferably, the method for judging environmental anomalies using an environmental anomaly identification model is as follows: A multi-source data matrix H is constructed based on the filtered time series data from the multiple sources. The k-th row of the multi-source data matrix is ​​the k-th filtered time series data sequence, where k = 1, 2, ..., K. Feature matrix extraction based on multi-source data matrix; Environmental anomaly detection is performed based on the feature matrix.

[0013] Preferably, the method for extracting the feature matrix based on feature fusion from a multi-source data matrix is ​​as follows: ; in, Let represent the k-th element of the feature matrix E, where the dimension of the feature matrix E is . , Represents the k-th row of the multi-source data matrix H. and These represent the preset upper and lower thresholds for the monitoring data corresponding to the k-th filtered time-series data sequence, respectively, and sum(.) represents summing all values ​​within the parentheses. This represents truth value calculation.

[0014] Preferably, the method for judging environmental anomalies based on the feature matrix is ​​as follows: Based on the feature matrix E, the judgment matrix P is obtained: ; Where e is the natural constant, and Let P be the weights and biases obtained from the fitting experiment, and let P be the dimension of the judgment matrix. ; If the value of the kth element of the judgment matrix P is greater than the preset percentage threshold, then the monitoring data corresponding to the kth filtered time series data sequence is judged to be abnormal.

[0015] This invention also provides an intelligent agricultural environment monitoring system based on multi-source information fusion, applied to the aforementioned intelligent agricultural environment monitoring method based on multi-source information fusion, comprising: The data acquisition module is used to periodically collect K types of monitoring data through multiple sensors at a period of T, forming K initial time series data sequences from multiple sources. Each initial time series sequence includes N parameters collected up to the current time. The data correction module performs the following operations on each initial time-series data sequence: Identify outlier data points; Extracting the initial time series data sequence for crops within a time range The planting and breeding operations and the corresponding time points of the breeding operations are used to calculate the abnormal time series sequence to be corrected based on the abnormal data points. Internal influence filtering is performed on the time series sequence to be corrected to obtain the filtered time series data sequence; The anomaly detection module is used to detect environmental anomalies by obtaining the filtered time series data from multiple sources and then using the environmental anomaly identification model.

[0016] The technical solution of the present invention has at least the following advantages and beneficial effects: This invention periodically collects K types of monitoring data using multiple sensors, avoiding the information limitations of a single data source and providing a more comprehensive reflection of the true state of the agricultural environment. This invention identifies abnormal data points and, in conjunction with R types of events and their time points in crop farming operations, makes targeted corrections to the data, effectively eliminating false anomalies caused by operations such as watering and fertilizing, and improving data reliability. This invention inputs the filtered time-series data into an environmental anomaly identification model for anomaly identification, without the need to directly delete monitoring data during the aquaculture operation stage, thus ensuring data integrity and reducing the possibility of missed detection. This invention enables automated correction and filtering of monitoring data, further enhancing the intelligence level of agricultural environmental monitoring. It is applicable to environmental monitoring of different types of crops and is easy to promote and implement. Attached Figure Description

[0017] Figure 1 This is a flowchart illustrating the intelligent agricultural environment monitoring method based on multi-source information fusion provided in Embodiment 1 of the present invention. Figure 2 This is a schematic diagram of the structure of the intelligent agricultural environment monitoring system based on multi-source information fusion provided in Embodiment 1 of the present invention. Detailed Implementation

[0018] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. The components of the embodiments of the present invention described and shown in the accompanying drawings can generally be arranged and designed in various different configurations.

[0019] Example 1 This embodiment provides an intelligent monitoring method for the agricultural environment based on multi-source information fusion. (See attached document.) Figure 1 This includes the following steps: Step S1: Data acquisition is performed by periodically collecting K types of monitoring data using multiple sensors at a period of T, forming K initial time-series data sequences from multiple sources. Each initial time-series sequence includes N parameters up to the current acquisition point. The preferred monitoring data include ambient temperature, ambient humidity, soil moisture, soil conductivity, carbon dioxide concentration, and oxygen concentration, all of which can be obtained using existing sensors.

[0020] Step S2: Eliminate the data impact caused by aquaculture operations, which means performing the following operations on each initial time series data sequence: Step S21 identifies anomalous data points. Specifically, the method for obtaining anomalous data points is as follows: Obtain the difference between adjacent data: ; in, For the k-th initial time series data sequence The nth parameter and the (n-1)th parameter The difference, , ; If the absolute value of the difference between adjacent data points is greater than a preset first difference threshold, then the data at the next time point is considered the abnormal data point.

[0021] In other words, abnormal jumps in numerical values ​​are initially considered as anomalous data points because certain aquaculture operations can cause significant changes in monitoring data within a short period. For example, soil moisture may increase dramatically after watering, or soil conductivity may increase dramatically in a short time after fertilization. These data then gradually return to normal over time. Because such sudden and significant changes can lead to misjudgments, data correction is necessary. By analyzing the differences between adjacent data points, anomalous data points that may be caused by aquaculture operations can be quickly and effectively located.

[0022] Step S22: Extract the initial time series data sequence for crops within the time range. The planting and breeding operations and the corresponding time points of the breeding operations are used to calculate the abnormal time series sequence to be corrected based on the abnormal data points.

[0023] In this embodiment, the method for calculating the abnormal time series sequence to be corrected based on abnormal data points is as follows: Get the The duration of the impact of the aquaculture operation described in type k on the monitoring data described in type k ; Determine the difference between the abnormal data point and the time point of the r-th type of aquaculture operation; if the difference is not... The internal method excludes the breeding operation described in the rth type. ; Get the The time points of the aforementioned breeding operations , The value range is for aquaculture operations that were not excluded, and the collection time of the abnormal data points is obtained. ; Get the The duration of the effects of the aforementioned aquaculture operations ; Get The maximum value is taken as the termination time of the comprehensive and lasting impact; In the corresponding initial time series data sequence, all parameters collected from the abnormal data point within the time of termination of the comprehensive continuous impact are formed to form the abnormal time series sequence to be corrected.

[0024] Based on this, the acquisition of the first The duration of the impact of the aquaculture operation described in type k on the monitoring data described in type k The method involves extraction through the following experiments: At the point of time Perform the r-th type of aquaculture operation on crops, that is... For the r-th aquaculture operation in the experiment, K monitoring data were collected periodically according to period T, starting from time point [r]. A reference time series of data is set to record the monitoring data after the r-th type of aquaculture operation is performed. Each reference time series of data includes G parameters. Obtain the adjacent differences of the k-th reference time series data sequence: ; in, For the k-th reference time series data sequence The Parameters and the Parameters The difference, ; Obtain the adjacent differences of all reference time-series data sequences whose absolute values ​​are less than the preset second difference threshold, and sort them in time to form a time-series difference sequence; Perform the following iterations: Initialize q=1; If the difference between the corresponding acquisition time points of the q-th parameter and the (q+b-th parameter) of the time-series difference sequence is less than a preset time threshold... σ If the iteration stops, then stop; otherwise, update the value of q to q = q + 1. Let the value of q be the value when the iteration stops. , Let the first of the time series difference sequences be... The parameters are , Obtain the k-th reference time series data sequence. The acquisition time point of each parameter, and the time point relative to the time point The difference is the first Duration of the effects of the aforementioned aquaculture operations .

[0025] After aquaculture operations cause a significant change in certain monitoring data, the data will recover to the level before the operation within a certain period of time. During this recovery phase, the changes in monitoring data will be faster than normal. Therefore, this embodiment utilizes this characteristic to find the duration of the impact. Thus, this embodiment first extracts all cases where the values ​​at adjacent time points are less than a second difference threshold, forming a time-series difference sequence. Each value in the time-series difference sequence is the difference between the data at the next time point and the data at the previous time point. Therefore, each value in the time-series difference sequence corresponds to a collection time, that is, the collection time of the data at the next time point. Since it is arranged in chronological order, the difference between the corresponding collection time points of the q-th parameter and the (q+b)-th parameter is less than the preset time threshold. σ This indicates that the q-th parameter and the (q+b)-th parameter are also the result of several consecutive cycles in actual data collection, which means that the overall trend of data change has decreased. Therefore, it is considered that the influence of the aquaculture operation has ended. b can be 1-3. It is a very small fault tolerance parameter.

[0026] It should be noted that the required amount of spraying in the aquaculture operation during the experiment is consistent with that in step S23, in order to eliminate errors in the calculation of the duration of the effect caused by different spraying amounts. Furthermore, the experiment was conducted under conditions free of abnormalities.

[0027] Step S23: Perform internal influence filtering on the abnormal time series sequence to be corrected to obtain the filtered time series data sequence.

[0028] As a preferred embodiment, the method for filtering the internal influence of the abnormal time series to be corrected is to perform the following operations for all k=1,2,…,K: Assignment The time points mentioned in the experiment The value of the kth type of monitoring data collected in the previous instance is used to obtain the first... The k-th reference time series data sequence under farming and aquaculture operations and the Number of sprays per unit area for planting and breeding operations , The value range is for aquaculture operations that were not eliminated; The time series sequence of the anomalies to be corrected for the kth type of monitoring data Make corrections: ; ; in, The abnormal time series sequence to be corrected The y-th parameter The corrected value, for The x-th parameter, where x is an intermediate parameter. For the actual time point Time The amount of spray applied per unit area in planting and breeding operations. The abnormal time series sequence to be corrected The time point at which the y-th parameter is collected, and the range of values ​​for y is... All parameters, The function representing rounding. Not less than And the difference does not exceed the preset threshold for the number of sprays per unit area. Not less than The reason for ensuring that the difference does not exceed the preset threshold for the number of applications per unit area is to avoid obtaining a decisive duration of the experimental impact, and also to avoid excessive errors in the calculation. Generally, there are several relatively standard farming methods for the same crop. Data such as standard watering and fertilization amounts per unit area can be obtained based on these standard farming methods, and experiments can be conducted based on this standard data.

[0029] The above scheme uses data outliers as starting points. Based on the duration of the impact and the actual time of the aquaculture operations, it obtains a time-series sequence of anomalies to be corrected, thus determining the amount of data to be corrected, making the correction more accurate and targeted. Furthermore, the correction is based on experimental data. Because the experiment was conducted under stable and normal conditions, data without aquaculture operations are approximated as constant, and data changes are considered to be caused by aquaculture operations. Correct errors caused by the amount of spray.

[0030] Unlike traditional anomaly handling methods that rely solely on numerical statistics, this embodiment directly correlates the timing of aquaculture operations with changes in monitoring data. This allows for more targeted correction of abnormal data and elimination of the impact of aquaculture operations. By obtaining the duration of the continuous impact of aquaculture operations on different monitoring data through experiments, the subjectivity of empirically set parameters is avoided, making the correction results more scientific and universal.

[0031] Step S3: Finally, environmental anomaly judgment can be performed. That is, after obtaining the multi-source filtered time series, environmental anomaly judgment is performed through the environmental anomaly identification model.

[0032] In this embodiment, the method for judging environmental anomalies using an environmental anomaly identification model is as follows: A multi-source data matrix H is constructed based on the filtered time series data from the multiple sources. The k-th row of the multi-source data matrix is ​​the k-th filtered time series data sequence, where k = 1, 2, ..., K. Feature matrix extraction based on multi-source data matrix; Environmental anomaly detection is performed based on the feature matrix.

[0033] Based on the above scheme, the method for extracting the feature matrix by feature fusion based on multi-source data matrix is ​​as follows: ; in, Let represent the k-th element of the feature matrix E, where the dimension of the feature matrix E is . , Represents the k-th row of the multi-source data matrix H. and These represent the preset upper and lower thresholds for the monitoring data corresponding to the k-th filtered time-series data sequence, respectively, and sum(.) represents summing all values ​​within the parentheses. This represents truth value calculation.

[0034] Finally, the method for judging environmental anomalies based on the feature matrix is ​​as follows: Based on the feature matrix E, the judgment matrix P is obtained: ; Where e is the natural constant, and Let P be the weights and biases obtained from the fitting experiment, and let P be the dimension of the judgment matrix. ; If the value of the kth element of the judgment matrix P is greater than the preset percentage threshold, then the monitoring data corresponding to the kth filtered time series data sequence is judged to be abnormal.

[0035] The above scheme extracts the proportion of data that exceeds the normal value range for each type of parameter, and uses the sigmoid function to implement a non-linear mapping of the degree of abnormality to determine whether an anomaly has occurred. For each parameter in the sequence within parentheses, a truth value judgment is performed, outputting either true 1 or false 0, thus transforming the numerical sequence formed by the k-th row of the original multi-source data matrix. The mapping is a sequence of true values, all of which are 0 or 1. A true value (1) indicates that the data exceeds the normal range. Therefore, summing each parameter of the true value sequence gives the number of data points exceeding the normal range, and dividing by N gives the anomaly percentage. Performing this operation on each row means performing the above operation separately for each type of monitoring data. The k-th element of the feature matrix E represents the proportion of anomalies in the monitored data corresponding to the k-th filtered time series data sequence. The weight parameters, biases, and percentage thresholds in the model can all be obtained through experimental fitting, allowing for adaptive adjustment based on different agricultural scenarios and improving the applicability of the method.

[0036] Example 2 This embodiment provides an intelligent agricultural environment monitoring system based on multi-source information fusion, applied to the aforementioned intelligent agricultural environment monitoring method based on multi-source information fusion. (See reference...) Figure 2 ,include: The data acquisition module is used to periodically collect K types of monitoring data through multiple sensors at a period of T, forming K initial time series data sequences from multiple sources. Each initial time series sequence includes N parameters collected up to the current time. The data correction module performs the following operations on each initial time-series data sequence: Identify outlier data points; Extract the R-type aquaculture operations and corresponding time points of the aquaculture operations for crops within the time range of the initial time series data sequence, and calculate the abnormal time series sequence to be corrected based on the abnormal data points; Internal influence filtering is performed on the time series sequence to be corrected to obtain the filtered time series data sequence; The anomaly detection module is used to detect environmental anomalies by obtaining the filtered time series data from multiple sources and then using the environmental anomaly identification model.

[0037] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for intelligent monitoring of the agricultural environment based on multi-source information fusion, characterized in that, Includes the following steps: K types of monitoring data are collected periodically by multiple sensors with period T, forming K initial time series data sequences from multiple sources. Each initial time series sequence includes N parameters collected up to the current time. Perform the following operations on each initial time series data sequence: Identify outlier data points; Extracting the initial time series data sequence for crops within a time range The planting and breeding operations and the corresponding time points of the breeding operations are used to calculate the abnormal time series sequence to be corrected based on the abnormal data points. Internal influence filtering is performed on the time series sequence to be corrected to obtain the filtered time series data sequence; After obtaining the multi-source filtered time series, environmental anomaly judgment is performed through the environmental anomaly identification model.

2. The intelligent monitoring method for agricultural environment based on multi-source information fusion according to claim 1, characterized in that, The various monitoring data include ambient temperature, ambient humidity, soil moisture, soil conductivity, carbon dioxide concentration, and oxygen concentration.

3. The intelligent monitoring method for agricultural environment based on multi-source information fusion according to claim 1, characterized in that, The method for obtaining the identified abnormal data points is as follows: Obtain the difference between adjacent data: ； in, For the k-th initial time series data sequence The nth parameter and the (n-1)th parameter The difference, , K is the total number of the initial time series data sequences, and N is the number of parameters in each initial time series sequence; If the absolute value of the difference between adjacent data points is greater than a preset first difference threshold, then the data at the next time point is considered the abnormal data point.

4. The intelligent monitoring method for agricultural environment based on multi-source information fusion according to claim 3, characterized in that, The method for calculating the abnormal time series sequence to be corrected based on abnormal data points is as follows: Get the The duration of the impact of the aquaculture operation described in type k on the monitoring data described in type k ; Determine the difference between the abnormal data point and the time point of the r-th type of aquaculture operation; if the difference is not... The internal method excludes the breeding operation described in the rth type. ; Get the The time points of the aforementioned breeding operations , The value range is for aquaculture operations that were not excluded, and the collection time of the abnormal data points is obtained. ; Get the The duration of the effects of the aforementioned aquaculture operations ; Get The maximum value is taken as the termination time of the comprehensive and lasting impact; In the corresponding initial time series data sequence, all parameters collected from the abnormal data point within the time of termination of the comprehensive continuous impact are formed to form the abnormal time series sequence to be corrected.

5. The intelligent monitoring method for agricultural environment based on multi-source information fusion according to claim 4, characterized in that, The acquisition of the first The duration of the impact of the aquaculture operation described in type k on the monitoring data described in type k The method involves extraction through the following experiments: At the point of time For crops, the r-th type of aquaculture operation is performed, and K types of monitoring data are collected periodically according to cycle T, starting from time point. A reference time series of data is set to record the monitoring data after the r-th type of aquaculture operation is performed. Each reference time series of data includes G parameters. Obtain the adjacent differences of the k-th reference time series data sequence: ； in, For the k-th reference time series data sequence The Parameters and the Parameters The difference, ; Obtain the adjacent differences of all reference time-series data sequences whose absolute values ​​are less than the preset second difference threshold, and sort them in time to form a time-series difference sequence; Perform the following iterations: Initialize q=1; If the difference between the corresponding acquisition time points of the q-th parameter and the (q+b-th parameter) of the time-series difference sequence is less than a preset time threshold... If the iteration stops, then stop; otherwise, update the value of q to q = q + 1. Let the value of q be the value when the iteration stops. , Let the first of the time series difference sequences be... The parameters are , Obtain the k-th reference time series data sequence. The acquisition time point of each parameter, and the time point relative to the time point The difference is the first Duration of the effects of the aforementioned aquaculture operations .

6. The intelligent monitoring method for agricultural environment based on multi-source information fusion according to claim 5, characterized in that, The method for filtering the internal influence of the abnormal time series to be corrected is to perform the following operations for all k=1,2,…,K: Assignment The time points mentioned in the experiment The value of the kth type of monitoring data collected in the previous instance is used to obtain the first... The k-th reference time series data sequence under farming and aquaculture operations and the Number of sprays per unit area for planting and breeding operations , The value range is for aquaculture operations that were not eliminated; The time series sequence of the anomalies to be corrected for the kth type of monitoring data Make corrections: ; ； in, The abnormal time series sequence to be corrected The y-th parameter The corrected value, for The x-th parameter, where x is an intermediate parameter. For the actual time point Time The amount of spray applied per unit area in planting and breeding operations. The abnormal time series sequence to be corrected The time point at which the y-th parameter is collected, and the range of values ​​for y is... All parameters, The function representing rounding. Not less than And the difference does not exceed the preset threshold for the number of sprays per unit area.

7. The intelligent monitoring method for agricultural environment based on multi-source information fusion according to claim 1, characterized in that, The method for judging environmental anomalies using an environmental anomaly identification model is as follows: A multi-source data matrix H is constructed based on the filtered time series data from the multiple sources. The k-th row of the multi-source data matrix is ​​the k-th filtered time series data sequence, where k = 1, 2, ..., K. Feature matrix extraction based on multi-source data matrix; Environmental anomaly detection is performed based on the feature matrix.

8. The intelligent monitoring method for agricultural environment based on multi-source information fusion according to claim 7, characterized in that, The method for extracting feature matrices based on feature fusion using multi-source data matrices is as follows: ； in, Let represent the k-th element of the feature matrix E, where the dimension of the feature matrix E is . , Represents the k-th row of the multi-source data matrix H. and These represent the preset upper and lower thresholds for the monitoring data corresponding to the k-th filtered time-series data sequence, respectively, and sum(.) represents summing all values ​​within the parentheses. This represents truth value calculation.

9. The intelligent monitoring method for agricultural environment based on multi-source information fusion according to claim 8, characterized in that, The method for judging environmental anomalies based on the feature matrix is ​​as follows: Based on the feature matrix E, the judgment matrix P is obtained: ； Where e is the natural constant, and Let P be the weights and biases obtained from the fitting experiment, and let P be the dimension of the judgment matrix. ; If the value of the kth element of the judgment matrix P is greater than the preset percentage threshold, then the monitoring data corresponding to the kth filtered time series data sequence is judged to be abnormal.

10. An intelligent agricultural environment monitoring system based on multi-source information fusion, applied to the intelligent agricultural environment monitoring method based on multi-source information fusion as described in any one of claims 1-9, characterized in that, include: The data acquisition module is used to periodically collect K types of monitoring data through multiple sensors at a period of T, forming K initial time series data sequences from multiple sources. Each initial time series sequence includes N parameters collected up to the current time. The data correction module performs the following operations on each initial time-series data sequence: Identify outlier data points; Extracting the initial time series data sequence for crops within a time range The planting and breeding operations and the corresponding time points of the breeding operations are used to calculate the abnormal time series sequence to be corrected based on the abnormal data points. Internal influence filtering is performed on the time series sequence to be corrected to obtain the filtered time series data sequence; The anomaly detection module is used to detect environmental anomalies by obtaining the filtered time series data from multiple sources and then using the environmental anomaly identification model.

Images