Agricultural irrigation intelligent monitoring system based on multi-modal data

Abstract

The invention relates to the technical field of agricultural irrigation, in particular to an agricultural irrigation intelligent monitoring system based on multi-modal data. The system comprises a data acquisition module for acquiring humidity data in a period by using a humidity sensor on each area of a farmland to form a humidity sequence; acquiring different types of irrigation data of each time of irrigation in different regional periods; the region similarity analysis module is used for acquiring the soil state similarity of the two regions according to the humidity difference and the irrigation data difference in the periods of the two regions; the clustering analysis module is used for clustering all the regions to obtain different class clusters; the areas in one class cluster form a to-be-analyzed area; and the water demand analysis module obtains the final water demand coefficient of the to-be-analyzed region, constructs a water demand model, and then brings the area of the to-be-analyzed region and the final water demand coefficient into the water demand model to obtain the water demand of the to-be-analyzed region. The method can accurately determine the amount of water needing to be irrigated in each to-be-analyzed area.

Description

Technical Field

[0001] This invention relates to the field of agricultural irrigation technology, and specifically to an intelligent monitoring system for agricultural irrigation based on multimodal data. Background Technology

[0002] In recent years, the development of technologies such as the Internet of Things (IoT), sensor networks, and artificial intelligence has provided new ideas for intelligent monitoring of agricultural irrigation. These technologies are expected to effectively solve problems associated with traditional irrigation methods, such as reliance on manual experience for water usage judgment and control, high subjectivity, and untimely response. This can help avoid water waste, soil compaction, and poor crop growth caused by over- or under-irrigation. However, agricultural irrigation demand is influenced by multiple factors, including soil moisture. Monitoring systems based on a single data source cannot fully reflect the true state of farmland, lacking adaptability and robustness. Multimodal data fusion technology can simultaneously collect and analyze multiple types of data, significantly improving monitoring accuracy and decision reliability by leveraging the complementarity between data. Therefore, it has broad application prospects in the field of intelligent agricultural irrigation.

[0003] In existing agricultural irrigation scenarios, due to variations in terrain, soil texture, nutrient retention capacity, drainage conditions, and sunlight exposure across different locations within farmland, the actual water requirements of crops often exhibit significant spatial unevenness. However, current irrigation methods often employ a uniform or overall average water allocation strategy, lacking the ability to effectively perceive and model spatial differences within the field. This makes it difficult to implement differentiated regulation based on soil properties and water dynamics in different areas. This not only leads to insufficient irrigation accuracy but also easily results in over-irrigation in some areas, causing root hypoxia and disease proliferation, or under-irrigation in others, causing crop wilting and stunted growth. Consequently, it affects water resource utilization efficiency, crop yield, and quality, hindering the development of precision and intelligent agricultural irrigation. Summary of the Invention

[0004] To address the aforementioned technical problems, the present invention aims to provide an intelligent monitoring system for agricultural irrigation based on multimodal data. The specific technical solution adopted is as follows:

[0005] One embodiment of the present invention provides an intelligent monitoring system for agricultural irrigation based on multimodal data, the system comprising:

[0006] The data acquisition module is used to divide the farmland into different areas evenly, and to deploy humidity sensors on each area to collect humidity data within a cycle, forming a humidity sequence; it also collects different types of irrigation data for each irrigation in different areas within a cycle.

[0007] The regional similarity analysis module is used to obtain the soil state similarity between two regions based on the differences in humidity sequences within two periods and the differences in irrigation data of different types of irrigation during each irrigation.

[0008] The clustering analysis module is used to cluster all regions based on the distance between humidity sensors in each pair of regions and the similarity of soil conditions to obtain different clusters; regions within a cluster constitute a region to be analyzed.

[0009] The water demand analysis module is used to analyze the humidity sequence corresponding to each humidity sensor in the area to be analyzed and obtain the final water demand coefficient; construct a water demand model and solve the parameters; and input the area of ​​the area to be analyzed and the final water demand coefficient into the water demand model to obtain the water demand of the area to be analyzed.

[0010] Preferably, data on different types of irrigation are collected for each irrigation cycle in different regions, including:

[0011] The irrigation data for a single irrigation in a region within a cycle includes the start time of irrigation, the duration of irrigation, and the water output per unit time at each moment during the irrigation process. The water output per unit time at each moment during an irrigation process forms a sequence, denoted as the irrigation rate sequence.

[0012] Preferably, the soil condition similarity between the two regions is obtained based on the differences in humidity sequences within the two regions over a period and the differences in irrigation data for different types of irrigation during each irrigation, including:

[0013] An irrigation combination is formed by combining one irrigation session in one region with one irrigation session in another region, and all irrigation combinations between the two regions are obtained. The absolute value of the difference between the start times of the two irrigation sessions in an irrigation combination is recorded as the irrigation start time difference of the irrigation combination. The absolute value of the difference between the durations of the two irrigation sessions in an irrigation combination is recorded as the irrigation duration difference of the irrigation combination. The DTW matching distance of the irrigation rate sequences of the two irrigation sessions in an irrigation combination is calculated using the DTW algorithm. The absolute value of the difference between the means of the two irrigation rate sequences in the irrigation combination is recorded as the first difference of the irrigation combination. The product of the DTW matching distance and the first difference is the irrigation rate difference of the irrigation combination. The DTW matching distance of the humidity sequences of the two regions is calculated. The absolute value of the difference between the means of the humidity sequences of the two regions is recorded as the second difference. The product of the DTW matching distance and the second difference of the humidity sequences of the two regions is the humidity difference between the two regions.

[0014] Based on the differences in irrigation start time, irrigation duration, irrigation rate, and humidity between the two regions for each irrigation combination, the similarity of irrigation start time, irrigation duration, irrigation rate, and humidity between the two regions were obtained respectively.

[0015] Soil condition similarity between the two regions was obtained based on the similarity of irrigation start time, irrigation duration, irrigation rate, and humidity.

[0016] Preferably, the similarity of irrigation start time, irrigation duration, irrigation rate, and humidity between the two regions is obtained based on the differences in irrigation start time, irrigation duration, irrigation rate, and humidity between the two regions for each irrigation combination, including:

[0017] The similarity of irrigation start time between the two regions is obtained by normalizing the value of the difference in irrigation start time between all irrigation combinations in the two regions by subtracting the mean of the difference in irrigation duration between all irrigation combinations in the two regions from the first preset value; the similarity of irrigation duration between the two regions is obtained by normalizing the value of the difference in irrigation rate between all irrigation combinations in the two regions by subtracting the mean of the difference in irrigation rate between all irrigation combinations in the two regions from the first preset value; and the similarity of humidity between the two regions is obtained by normalizing the value of the difference in humidity between the two regions by subtracting the mean of the difference in humidity between the two regions from the first preset value.

[0018] Preferably, the soil state similarity between the two regions is obtained based on the similarity of irrigation start time, irrigation duration, irrigation rate, and humidity, including:

[0019] The soil condition similarity between the two regions is obtained by multiplying and square-taking the similarity of irrigation start time, irrigation duration, irrigation rate, and humidity between the two regions.

[0020] Preferably, clustering is performed on all regions based on the distance between humidity sensors in each pair of regions and the similarity of soil conditions to obtain different clusters, including:

[0021] The distance between humidity sensors in two regions is negatively correlated using an exponential function with a base of the natural constant, and multiplied by the soil condition similarity between the two regions to obtain the confidence coefficient of the same irrigation type in the two regions.

[0022] Clusters were obtained by using the confidence coefficient of the same irrigation type in every two regions and the HDBSCAN clustering algorithm.

[0023] Preferably, the final water demand coefficient is analyzed by analyzing the humidity sequences corresponding to each humidity sensor within the analysis area, including:

[0024] The slope of the humidity sequence corresponding to a humidity sensor in the region to be analyzed is obtained at each moment, forming a slope sequence. A negative correlation mapping is performed on the time distance from a moment in the humidity sequence to the current time using an exponential function with a base of the natural constant, serving as the weight for that moment. The slope value at each moment is then mapped using the exponential function with a base of the natural constant to obtain the mapped value for each moment. A weighted average of the mapped values ​​at each moment in the slope sequence is then performed using the weights at each moment to obtain the humidity trend change factor of the humidity sensor. Finally, a weighted average of the humidity at each moment in the humidity sequence is performed using the weights at each moment to obtain the humidity evaluation value of the humidity sensor.

[0025] The water demand coefficient of the region corresponding to the humidity sensor is obtained by negatively mapping the product of the humidity trend change factor and the humidity evaluation value of a humidity sensor using an exponential function with the natural constant as the base; the final water demand coefficient of the region to be analyzed is obtained based on the water demand coefficients of all regions corresponding to humidity sensors in the region to be analyzed.

[0026] Preferably, the final water demand coefficient of the area to be analyzed is obtained based on the water demand coefficients of the areas corresponding to all humidity sensors within the area to be analyzed, including:

[0027] The crop coefficient of a region to be analyzed is determined based on the growth stage of the crop in that region. The average water demand coefficient of all humidity sensors in that region is calculated and multiplied by the crop coefficient of that region to obtain the final water demand coefficient of that region.

[0028] Preferably, the water demand model is as follows:

[0029] L j =θ×W j ×A j +β,

[0030] Among them, L j Let represent the water demand of the j-th region to be analyzed; θ and β represent the parameters of the water demand model; W j A represents the final water demand coefficient for j regions to be analyzed; j This represents the area of ​​the j-th region to be analyzed.

[0031] The embodiments of this invention have at least the following beneficial effects: First, farmland is evenly divided into different areas. Then, humidity data within a cycle is collected by humidity sensors deployed in each area to form a humidity sequence. At the same time, different types of irrigation data for each irrigation within the cycle in different areas are also acquired. The humidity data and irrigation data are combined for analysis to obtain the state similarity between each two areas. Further, based on the distance between humidity sensors and soil state similarity between each two areas, all areas are clustered to obtain different clusters. Areas within a cluster form an area to be analyzed. Finally, the humidity sequence corresponding to each humidity sensor in the area to be analyzed is analyzed to obtain the final water demand coefficient. A water demand model is constructed and the parameters are solved. The area of ​​the area to be analyzed and the final water demand coefficient are substituted into the water demand model to obtain the water demand of the area to be analyzed, which can realize precise irrigation for different areas. Attached Figure Description

[0032] To more clearly illustrate the technical solutions and advantages in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0033] Figure 1 This is a system block diagram of an intelligent monitoring system for agricultural irrigation based on multimodal data, provided as an embodiment of the present invention. Detailed Implementation

[0034] To further illustrate the technical means and effects adopted by the present invention to achieve its intended purpose, the following, in conjunction with the accompanying drawings and preferred embodiments, details the specific implementation, structure, features, and effects of an intelligent agricultural irrigation monitoring system based on multimodal data proposed according to the present invention. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. Furthermore, specific features, structures, or characteristics in one or more embodiments can be combined in any suitable form.

[0035] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0036] The following description, in conjunction with the accompanying drawings, details a specific scheme for an intelligent agricultural irrigation monitoring system based on multimodal data provided by the present invention.

[0037] Example:

[0038] The main application scenarios of this invention are:

[0039] By sensing and modeling the spatial differences within farmland, differentiated irrigation regulation can be implemented based on soil properties and water dynamics in different areas, thereby improving the precision and intelligence of irrigation, increasing water resource utilization efficiency, and enhancing crop yield and quality.

[0040] Please see Figure 1 The diagram illustrates a system block diagram of an intelligent agricultural irrigation monitoring system based on multimodal data, provided by an embodiment of the present invention. The system includes the following modules:

[0041] The data acquisition module is used to divide the farmland into different areas evenly, and to deploy humidity sensors in each area to collect humidity data within a cycle, forming a humidity sequence; it also collects different types of irrigation data for each irrigation in different areas within a cycle.

[0042] In the process of intelligent monitoring of agricultural irrigation, the geographical environment of different locations within farmland may vary, such as elevation, soil type, degree of shading, and drainage conditions. These factors can lead to differences in soil moisture and nutrient retention capacity at different locations. Even after the same irrigation, the growth status and water requirements of crops in different locations may show significant differences. For example, lower-lying areas may be prone to waterlogging, while higher-lying areas with loose soil may lose water too quickly. If a uniform irrigation strategy is adopted, it can easily lead to over-irrigation in some areas, causing root hypoxia or disease, while other areas may experience water shortage and crop wilting.

[0043] Therefore, to address the aforementioned phenomena, personalized monitoring and data collection are needed for different locations within farmland. This requires refined analysis combining data such as soil moisture and irrigation information to develop targeted irrigation strategies for each location, achieving truly precise and intelligent irrigation management. This example uses rapeseed fields for analysis; the same principle applies to other farmland.

[0044] Specifically, the farmland is divided into different areas, and a humidity sensor is placed at the center of each area.

[0045] The implementation cycle is seven days, which can be adjusted by the operator based on actual conditions. Image data of the farmland is collected during the most recent cycle; the images are taken from above and include the entire crop. Humidity data for that cycle is also collected using sensors deployed throughout the area, creating a humidity sequence; one humidity sequence corresponds to one region.

[0046] Because soil conditions vary in different locations, the required water volume may also differ. Therefore, it is necessary to collect different types of irrigation data for each irrigation in each region. The irrigation data includes the start time of irrigation, the duration of irrigation, and the water output per unit time at each moment during an irrigation in a region. The water output per unit time at each moment during an irrigation forms a sequence, which is denoted as the irrigation rate sequence.

[0047] This allows us to obtain the humidity sequence for each region within a cycle, as well as the irrigation start time, irrigation duration, and irrigation rate sequence for each irrigation.

[0048] The regional similarity analysis module is used to obtain the soil state similarity between two regions based on the differences in humidity sequences within two periods and the differences in irrigation data of different types of irrigation during each irrigation.

[0049] For individual feature data, such as the irrigation start time and duration for each irrigation session in two regions, the similarity can be directly calculated by comparing the differences in these values. However, for humidity sequences and irrigation rate sequences for each irrigation session in two regions, which are time-series data, the similarity can be calculated using the Dynamic Time Warping (DTW) algorithm. Combining the similarity of the individual feature data and the time-series feature data yields the overall similarity between the two regions. Based on this similarity, cluster analysis is performed on the regions corresponding to the humidity sensors, facilitating the division of the regions into multiple clusters and enabling precise zoning irrigation for different land clusters.

[0050] Therefore, when calculating the similarity of relevant data for any two regions, since the number of irrigations in one cycle may differ between the two regions, the amount of data collected may also differ. Therefore, when calculating the similarity of individual feature data for any two regions, a combination method is used. For example, if region a was irrigated three times and region b was irrigated only once, then when calculating the similarity of the irrigation duration between regions a and b, the calculation is performed by combining each irrigation in region a with each irrigation in region b, and then obtaining the average of all combinations to obtain the difference in irrigation duration between the two regions, and thus the similarity of the data.

[0051] Soil condition similarity between the two regions was obtained by comparing the differences in humidity sequences within the two regions over a period of time and the differences in irrigation data for different types of irrigation during each irrigation.

[0052] Specifically, an irrigation combination is formed by combining one irrigation session in one region with one irrigation session in another region, and all irrigation combinations between the two regions are obtained. The absolute value of the difference between the start times of two irrigation sessions in an irrigation combination is recorded as the irrigation start time difference of the irrigation combination. The absolute value of the difference between the durations of two irrigation sessions in an irrigation combination is recorded as the irrigation duration difference of the irrigation combination. The DTW matching distance of the irrigation rate sequences of two irrigation sessions in an irrigation combination is calculated using the DTW algorithm, and the absolute value of the difference between the means of the two irrigation rate sequences in the irrigation combination is recorded as the first difference of the irrigation combination. The product of the DTW matching distance and the first difference is the irrigation rate difference of the irrigation combination. The DTW matching distance of the humidity sequences of two regions is calculated, and the absolute value of the difference between the means of the humidity sequences of two regions is recorded as the second difference. The product of the DTW matching distance and the second difference of the humidity sequences of two regions is the humidity difference between the two regions.

[0053] Furthermore, based on the differences in irrigation start time, irrigation duration, irrigation rate, and humidity between the two regions, the similarity of irrigation start time, irrigation duration, irrigation rate, and humidity between the two regions were obtained.

[0054] Specifically, the similarity of irrigation start time between two regions is obtained by subtracting the mean of the differences in irrigation start time of all irrigation combinations from the first preset value and normalizing the value; the similarity of irrigation duration between two regions is obtained by subtracting the mean of the differences in irrigation duration of all irrigation combinations from the first preset value and normalizing the value; the similarity of irrigation rate between two regions is obtained by subtracting the mean of the differences in irrigation rate of all irrigation combinations from the first preset value and normalizing the value; and the similarity of humidity between two regions is obtained by subtracting the mean of the differences in humidity between two regions from the first preset value.

[0055] Next, the soil state similarity between the two regions is obtained based on the similarity of irrigation start time, irrigation duration, irrigation rate, and humidity. The first preset value is set to 1.

[0056] Specifically, the soil condition similarity between the two regions is obtained by multiplying and taking the square root of the similarity of irrigation start time, irrigation duration, irrigation rate, and humidity between the two regions.

[0057] The specific calculation model for soil state similarity is as follows:

[0058]

[0059] Where Z(a,b) represents the soil state similarity between region a and region b, m represents 4, because the two regions have four similarities: irrigation start time similarity, irrigation duration similarity, irrigation rate similarity, and humidity similarity, and C(a,b) represents the soil state similarity between region a and region b. i This represents the i-th similarity among irrigation start time similarity, irrigation duration similarity, irrigation rate similarity, and humidity similarity; This means multiplying the similarities in irrigation start time, irrigation duration, irrigation rate, and humidity between regions a and b, and then taking the m=4th power. The summation and raising the power of m ensures that all four features are similar. This allows us to obtain the soil condition similarity between any two regions.

[0060] The clustering analysis module is used to cluster all regions based on the distance between humidity sensors in each pair of regions and the similarity of soil conditions to obtain different clusters; regions within a cluster constitute a region to be analyzed.

[0061] The above analysis can yield the similarity of soil conditions between any two regions. However, in actual irrigation, due to the permeability and diffusion of water in the soil, the humidity levels of adjacent regions often influence each other, thus interfering with the irrigation analysis results. Therefore, it is necessary to further consider the deployment locations of humidity sensors in different regions, and, based on similarity calculations, comprehensively consider spatial proximity to conduct a reliability assessment of the soil conditions between any two regions. By introducing this reliability index, soil region categories can be more rationally classified, improving the accuracy and reliability of zonal irrigation analysis.

[0062] Therefore, based on the distance between humidity sensors in each pair of regions and the similarity of soil conditions, the probability of classifying each pair of regions as belonging to the same irrigation type is calculated, and the probability of being classified as belonging to the same irrigation type is denoted as the confidence coefficient of the same irrigation type. Specifically, an exponential function with a base of the natural constant is used to perform a negative correlation mapping on the distance between humidity sensors in the two regions, and this is multiplied by the similarity of soil conditions in the two regions to obtain the confidence coefficient of the same irrigation type for the two regions.

[0063] The specific calculation model for the reliability coefficient of the same irrigation type is as follows:

[0064] B(a,b)=Z(a,b)×e -d(a,b) ,

[0065] Where B(a,b) represents the probability that regions a and b are classified into the same irrigation type, i.e., the confidence coefficient of regions a and b belonging to the same irrigation type. Z(a,b) represents the soil condition similarity between regions a and b. The larger this value, the greater the similarity of the soil conditions between the two regions, and therefore the higher the confidence of classifying them into the same irrigation type. e represents the natural constant, and d(a,b) represents the distance between the humidity sensors in region a and region b. The smaller this distance, the greater the likelihood of e being classified into the same irrigation type. -d(a,b) The higher the value, the greater the likelihood that the two regions can be classified as belonging to the same irrigation type. This allows us to obtain the confidence coefficient for the same irrigation type for each pair of regions.

[0066] Then, the HDBSCAN (Hierarchical Density-Based Spatial Clustering of Applications with Noise) clustering algorithm is used. HDBSCAN is a density-based hierarchical clustering algorithm that combines the advantages of DBSCAN and hierarchical clustering. This technique is existing and will not be elaborated further. Different regions are divided into multiple clusters based on the confidence coefficient of the same irrigation type between every two regions using the HDBSCAN clustering algorithm. Then, based on the location of regions within each cluster, their corresponding coverage areas are mapped onto the farmland space, thereby dividing the entire land into several regions with similar soil moisture characteristics. In other words, regions within a cluster are merged to obtain a single region to be analyzed. This region division method based on region clustering results can effectively reflect the spatial differences in soil moisture within farmland, providing a reliable basis for subsequent zoning irrigation decisions.

[0067] The water demand analysis module is used to analyze the humidity sequence corresponding to each humidity sensor in the area to be analyzed and obtain the final water demand coefficient; construct a water demand model and solve the parameters; and input the area of ​​the area to be analyzed and the final water demand coefficient into the water demand model to obtain the water demand of the area to be analyzed.

[0068] The above-mentioned clustering results of all regions divide the farmland into multiple regions to be analyzed. Further analysis is needed based on the humidity sequence corresponding to each humidity sensor in each region to obtain the water shortage status of the region to be analyzed, and conversely, the water demand status.

[0069] The final water demand coefficient is analyzed based on the humidity sequences corresponding to each humidity sensor within the analysis area. Specifically, the slope of the humidity sequence corresponding to a humidity sensor in the analysis area is obtained at each moment, forming a slope sequence. A negative correlation mapping is performed on the time distance from a moment in the humidity sequence corresponding to the humidity sensor to the current time using an exponential function with a base of the natural constant, serving as the weight for that moment. The slope value at each moment is then mapped using the exponential function with a base of the natural constant to obtain the mapped value for each moment. A weighted average of the mapped values ​​at each moment in the slope sequence is then performed using the weights at each moment to obtain the humidity trend change factor of the humidity sensor. Finally, a weighted average of the humidity at each moment in the humidity sequence is performed using the weights at each moment to obtain the humidity evaluation value of the humidity sensor.

[0070] The specific calculation model for the humidity trend change factor is as follows:

[0071]

[0072] Where Q represents the humidity trend change factor corresponding to a humidity sensor in the region to be analyzed; g represents the number of humidity data points in the humidity sequence corresponding to that humidity sensor in the region to be analyzed, that is, the number of moments in the humidity sequence; e represents the natural constant; k s Let represent the slope of the slope sequence formed by the slopes at each time step in the humidity sequence, at the s-th time step. Let t(n,s) represent the first mapping value at time s, where the slope at the first time is represented by a first preset value of 1; t(n,s) represents the time distance between the s-th time and the current time in the humidity sequence. The smaller this value, the higher the weight e at time s. -t(n,s) The larger the value, the greater the impact of the slope at time s on the evaluation of the change in the current humidity data; This represents the weighted average of all the slope mapping values. The smaller this value, the faster the humidity data in the area corresponding to the humidity sensor decreases.

[0073] The calculation of humidity assessment values ​​aims to avoid the interference of fluctuations in humidity values ​​at a single moment on the overall soil condition assessment, thereby obtaining a more stable and comprehensive humidity assessment result. The specific calculation model for humidity assessment values ​​is as follows:

[0074]

[0075] Where X represents the humidity evaluation value of a humidity sensor in the area to be analyzed, and F s This represents the s-th humidity level in the humidity sequence corresponding to the humidity sensor in the area to be analyzed.

[0076] Therefore, the water demand coefficient of the region corresponding to the humidity sensor is obtained by combining the humidity trend change factor and humidity evaluation value of a humidity sensor within the region to be analyzed. Specifically, the water demand coefficient of the region corresponding to the humidity sensor is obtained by negatively mapping the product of the humidity trend change factor and humidity evaluation value of a humidity sensor using an exponential function with a base of the natural constant. The smaller the humidity trend change factor and humidity evaluation value, the greater the water demand of the region corresponding to the humidity sensor, and vice versa.

[0077] Furthermore, the final water requirement coefficient for the analyzed area is obtained based on the water requirement coefficients corresponding to all humidity sensors within the analyzed area. This requires considering the crop growth stage within the analyzed area to determine the final water requirement coefficient. Specifically, a neural network model is used to extract and analyze features from crop images in each analyzed area to identify the current growth stage of the crop (e.g., seedling stage, vegetative growth stage, flowering stage, grain-filling stage, etc.). Based on the identified growth stage, and using the FAO-56 method (a standard for calculating crop water requirements published by the Food and Agriculture Organization of the United Nations, providing calculation methods based on crop coefficient K and other relevant data related to crop growth stages), the crop coefficient K for each analyzed area is obtained (this coefficient reflects the water requirement of the crop at the current growth stage, providing a basis for subsequent irrigation calculations).

[0078] The average water demand coefficient of all humidity sensors within a given area is calculated, and this average is multiplied by the crop coefficient of that area to obtain the final water demand coefficient. A higher crop coefficient indicates a greater water demand from the crops in the analyzed area.

[0079] This yields the final water demand coefficient for each region to be analyzed, and a water demand model is then constructed based on the final water demand coefficient and area of ​​each region. The specific water demand model is as follows:

[0080] L j =θ×W j ×A j +β,

[0081] Among them, L j Let represent the water demand of the j-th region to be analyzed; θ and β represent the parameters of the water demand model; W j A represents the final water demand coefficient for j regions to be analyzed; j This represents the area of ​​the j-th region to be analyzed.

[0082] Furthermore, it is necessary to solve for the parameters in the water demand model. These parameters can be obtained using linear regression, least squares, or other statistical fitting methods. Specifically, the final water demand coefficient of each region within the analyzed area before each irrigation is obtained within a given period; the mean of the final water demand coefficients of all regions within the analyzed area before each irrigation is calculated, denoted as the average final water demand coefficient of the analyzed area within that period; the mean of the irrigation water volume during each irrigation of all regions within the analyzed area is calculated, denoted as the average irrigation water volume of the analyzed area within that period; using the average irrigation water volume of each analyzed area within each period as the dependent variable, and the average final water demand coefficient and area of ​​each analyzed area within each period as independent variables, a regression algorithm is applied to fit and obtain the parameters θ and β of the water demand model.

[0083] This yields a water demand model with calculated parameters θ and β. The area of ​​the region to be analyzed and the currently obtained final water demand coefficient are then substituted into the model to calculate the water demand for the next irrigation. By utilizing historical irrigation experience, the model parameters reflect the actual water demand of crops, improving the accuracy and reliability of irrigation prediction. Furthermore, the parameters can be dynamically adjusted based on new irrigation data, enabling model adaptation and optimization.

[0084] The water demand of each area to be analyzed can be obtained through the above methods. The water volume can be regulated by an irrigation control system (such as a pipeline network controlled by a solenoid valve, a sprinkler or drip irrigation device), and then the corresponding amount of water can be irrigated, thereby achieving differentiated irrigation for each area to be analyzed.

[0085] It should be noted that the order of the above embodiments of the present invention is merely for descriptive purposes and does not represent the superiority or inferiority of the embodiments. Furthermore, the above description focuses on specific embodiments of this specification. Additionally, the processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired results. In some embodiments, multitasking and parallel processing are possible or may be advantageous.

[0086] The various embodiments in this specification are described in a progressive manner. The same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on describing the differences from other embodiments.

[0087] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. An intelligent monitoring system for agricultural irrigation based on multimodal data, characterized in that, The system includes: The data acquisition module is used to divide the farmland into different areas evenly, and to deploy humidity sensors on each area to collect humidity data within a cycle, forming a humidity sequence; it also collects different types of irrigation data for each irrigation in different areas within a cycle. The regional similarity analysis module is used to obtain the soil state similarity between two regions based on the differences in humidity sequences within two periods and the differences in irrigation data of different types of irrigation during each irrigation. The clustering analysis module is used to cluster all regions based on the distance between humidity sensors in each pair of regions and the similarity of soil conditions to obtain different clusters; regions within a cluster constitute a region to be analyzed. The water demand analysis module is used to analyze the humidity sequence corresponding to each humidity sensor in the area to be analyzed and obtain the final water demand coefficient; construct a water demand model and solve the parameters; and input the area of ​​the area to be analyzed and the final water demand coefficient into the water demand model to obtain the water demand of the area to be analyzed.

2. The intelligent monitoring system for agricultural irrigation based on multimodal data according to claim 1, characterized in that, The collection of different types of irrigation data for each irrigation cycle in different regions includes: The irrigation data for a single irrigation in a region within a cycle includes the start time of irrigation, the duration of irrigation, and the water output per unit time at each moment during the irrigation process. The water output per unit time at each moment during an irrigation process forms a sequence, denoted as the irrigation rate sequence.

3. The intelligent monitoring system for agricultural irrigation based on multimodal data according to claim 1, characterized in that, The method of obtaining soil state similarity between two regions based on differences in humidity sequences within two periods and differences in irrigation data of different types during each irrigation includes: An irrigation combination is formed by combining one irrigation session in one region with one irrigation session in another region, and all irrigation combinations between the two regions are obtained. The absolute value of the difference between the start times of the two irrigation sessions in an irrigation combination is recorded as the irrigation start time difference of the irrigation combination. The absolute value of the difference between the durations of the two irrigation sessions in an irrigation combination is recorded as the irrigation duration difference of the irrigation combination. The DTW matching distance of the irrigation rate sequences of the two irrigation sessions in an irrigation combination is calculated using the DTW algorithm. The absolute value of the difference between the means of the two irrigation rate sequences in the irrigation combination is recorded as the first difference of the irrigation combination. The product of the DTW matching distance and the first difference is the irrigation rate difference of the irrigation combination. The DTW matching distance of the humidity sequences of the two regions is calculated. The absolute value of the difference between the means of the humidity sequences of the two regions is recorded as the second difference. The product of the DTW matching distance and the second difference of the humidity sequences of the two regions is the humidity difference between the two regions. Based on the differences in irrigation start time, irrigation duration, irrigation rate, and humidity between the two regions for each irrigation combination, the similarity of irrigation start time, irrigation duration, irrigation rate, and humidity between the two regions were obtained respectively. Soil condition similarity between the two regions was obtained based on the similarity of irrigation start time, irrigation duration, irrigation rate, and humidity.

4. The intelligent monitoring system for agricultural irrigation based on multimodal data according to claim 3, characterized in that, The method of obtaining the similarity of irrigation start time, irrigation duration, irrigation rate, and humidity between the two regions based on the differences in irrigation start time, irrigation duration, irrigation rate, and humidity between the two regions for each irrigation combination includes: The similarity of irrigation start time between the two regions is obtained by normalizing the value of the difference in irrigation start time between all irrigation combinations in the two regions by subtracting the mean of the difference in irrigation duration between all irrigation combinations in the two regions from the first preset value; the similarity of irrigation duration between the two regions is obtained by normalizing the value of the difference in irrigation rate between all irrigation combinations in the two regions by subtracting the mean of the difference in irrigation rate between all irrigation combinations in the two regions from the first preset value; and the similarity of humidity between the two regions is obtained by normalizing the value of the difference in humidity between the two regions by subtracting the mean of the difference in humidity between the two regions from the first preset value.

5. The intelligent monitoring system for agricultural irrigation based on multimodal data according to claim 3, characterized in that, The method of obtaining soil state similarity between two regions based on similarity in irrigation start time, irrigation duration, irrigation rate, and humidity includes: The soil condition similarity between the two regions is obtained by multiplying and square-taking the similarity of irrigation start time, irrigation duration, irrigation rate, and humidity between the two regions.

6. The intelligent monitoring system for agricultural irrigation based on multimodal data according to claim 1, characterized in that, The method of clustering all regions based on the distance between humidity sensors in each pair of regions and the similarity of soil conditions to obtain different clusters includes: The distance between humidity sensors in two regions is negatively correlated using an exponential function with a base of the natural constant, and multiplied by the soil condition similarity between the two regions to obtain the confidence coefficient of the same irrigation type in the two regions. Clusters were obtained by using the confidence coefficient of the same irrigation type in every two regions and the HDBSCAN clustering algorithm.

7. The intelligent monitoring system for agricultural irrigation based on multimodal data according to claim 1, characterized in that, The analysis of the humidity sequences corresponding to each humidity sensor within the area to be analyzed to obtain the final water demand coefficient includes: The slope of the humidity sequence corresponding to a humidity sensor in the region to be analyzed is obtained at each moment, forming a slope sequence. A negative correlation mapping is performed on the time distance from a moment in the humidity sequence to the current time using an exponential function with a base of the natural constant, serving as the weight for that moment. The slope value at each moment is then mapped using the exponential function with a base of the natural constant to obtain the mapped value for each moment. A weighted average of the mapped values ​​at each moment in the slope sequence is then performed using the weights at each moment to obtain the humidity trend change factor of the humidity sensor. Finally, a weighted average of the humidity at each moment in the humidity sequence is performed using the weights at each moment to obtain the humidity evaluation value of the humidity sensor. The water demand coefficient of the region corresponding to the humidity sensor is obtained by negatively mapping the product of the humidity trend change factor and the humidity evaluation value of a humidity sensor using an exponential function with the natural constant as the base; the final water demand coefficient of the region to be analyzed is obtained based on the water demand coefficients of all regions corresponding to humidity sensors in the region to be analyzed.

8. The intelligent monitoring system for agricultural irrigation based on multimodal data according to claim 7, characterized in that, The step of obtaining the final water demand coefficient of the area to be analyzed based on the water demand coefficients of the areas corresponding to all humidity sensors within the area to be analyzed includes: The crop coefficient of a region to be analyzed is determined based on the growth stage of the crop in that region. The average water demand coefficient of all humidity sensors in that region is calculated and multiplied by the crop coefficient of that region to obtain the final water demand coefficient of that region.

9. The intelligent monitoring system for agricultural irrigation based on multimodal data according to claim 1, characterized in that, The water demand model is as follows: L j =θ×W j ×A j +a, Among them, L j Let represent the water demand of the j-th region to be analyzed; θ and β represent the parameters of the water demand model; W j A represents the final water demand coefficient for j regions to be analyzed; j This represents the area of ​​the j-th region to be analyzed.

Images