Intelligent control system and method based on agricultural big data

Abstract

The present application discloses an intelligent control system and method based on agricultural big data, which obtains soil index data at multiple predetermined time points within a predetermined time period collected by a soil sensor, and growth status parameters of crops at the multiple predetermined time points collected by a crop sensor; performs feature extraction and feature interaction on the soil index data at the multiple predetermined time points and the growth status parameters of crops at the multiple predetermined time points to obtain a soil-crop growth status interaction feature vector; generates irrigation control instructions based on the soil-crop growth status interaction feature vector; the present application can automatically adjust the irrigation operation of farmland to achieve agricultural optimization and efficient production.

Description

Technical Field

[0001] The present application relates to the field of intelligent control technology, and more specifically, to an intelligent control system and method based on agricultural big data. Background Art

[0002] Agricultural production is an important foundation of human society and an important support for economic development. With the advancement of agricultural science and technology, the application of agricultural big data is becoming more and more extensive, which can provide accurate data support for agricultural production and improve agricultural production efficiency and quality. In the process of crop growth, reasonable control of irrigation timing is particularly important for the growth of crops; appropriate irrigation can ensure sufficient soil moisture, promote crop photosynthesis and nutrient absorption, thereby increasing crop growth rate and yield; maintaining appropriate soil moisture can reduce the breeding and spread of pests and diseases, improve crop resistance and immunity, reduce the use of pesticides, reduce production costs, and be more environmentally friendly. Both over-irrigation and under-irrigation can cause harm to crops and farmland. For example, over-irrigation can cause the soil to be too wet, exceeding the amount of water that the crop roots can absorb, which will cause insufficient oxygen supply in the soil, affect the respiration and growth of plant roots, and even cause root suffocation; over-irrigation will increase the moisture content in the soil, making the soil soft and easy to erode. Over-irrigation will cause salt accumulation in the soil. When the salt concentration in the irrigation water is too high, excess salt will accumulate in the soil, forming salinization. Salinization will reduce soil fertility and have an adverse effect on the growth of most crops; furthermore, excessive irrigation will lead to a waste of water resources.

[0003] Traditional irrigation methods mainly rely on personnel to carry out based on experience, but with the development of smart agricultural technology, crop irrigation has gradually realized intelligent control. For example, the Chinese patent application number 201510040333.1 discloses an agricultural irrigation remote control management system and control method, including a cultivation module, an irrigation device, a data processor and a remote control terminal, the cultivation module is connected with a humidity sensor, the humidity sensor is connected with a wireless signal transmitter, the data processor is connected with a wireless signal receiver, the wireless signal transmitter and the wireless signal receiver are connected through a WIFI network to achieve data connection, the setting of the system eliminates the wired device data transmission, reduces the investment cost of wired device data transmission, saves maintenance costs, and also reduces manpower investment, the control method of the present invention is simple to control, the system works stably, saves water resources, monitors the water shortage of vegetation in a timely manner, and improves irrigation efficiency.

[0004] Intelligent control systems similar to the above-mentioned ones, when controlling the timing of irrigation, often only determine whether irrigation is needed based on soil moisture parameters; however, crop growth is staged, and different growth stages have different irrigation requirements, which is not only related to soil moisture, but also closely related to the crop's own transpiration rate, leaf size and other growth characteristics.

[0005] Therefore, in order to provide the most reasonable irrigation control, it is necessary to use various data based on the characteristics of the crops themselves and the soil characteristics as control conditions, so as to make intelligent irrigation decisions based on real-time data to prevent over-irrigation or under-irrigation, thereby affecting the water resource utilization efficiency of the farmland and the health and yield of the crop growth.

[0006] Therefore, an optimized solution based on agricultural big data is needed to realize the above-mentioned intelligent control. Summary of the invention

[0007] In order to solve the above technical problems, the present application is proposed. The embodiment of the present application provides an intelligent control system and method based on agricultural big data, which obtains soil index data at multiple predetermined time points within a predetermined time period collected by a soil sensor, and the growth status parameters of the crops at the multiple predetermined time points collected by a crop sensor; performs feature extraction and feature interaction on the soil index data at the multiple predetermined time points and the growth status parameters of the crops at the multiple predetermined time points to obtain a soil-crop growth status interaction feature vector; based on the soil-crop growth status interaction feature vector, generates irrigation control instructions, which can automatically adjust the irrigation operation of the farmland to achieve agricultural optimization and efficient production.

[0008] In a first aspect, an intelligent control method based on agricultural big data is provided, which comprises:

[0009] Acquire soil index data at a plurality of predetermined time points within a predetermined time period collected by a soil sensor, and growth state parameters of crops at the plurality of predetermined time points collected by a crop sensor; perform feature extraction and feature interaction on the soil index data at the plurality of predetermined time points and the growth state parameters of crops at the plurality of predetermined time points to obtain a soil-crop growth state interaction feature vector;

[0010] generating irrigation control instructions based on the soil-crop growth state interaction feature vector;

[0011] The soil index data at the plurality of predetermined time points and the growth state parameters of the crops at the plurality of predetermined time points are subjected to feature extraction and feature interaction to obtain a soil-crop growth state interaction feature vector, including:

[0012] Arranging the soil index data at the plurality of predetermined time points into a soil index full time series input matrix according to the time dimension and the index sample dimension, and arranging the growth state parameters of the crops at the plurality of predetermined time points into a growth state parameter full time series input matrix according to the time dimension and the state parameter;

[0013] Performing full time series correlation analysis on the soil index full time series input matrix and the growth state parameter full time series input matrix respectively to obtain a soil index time series correlation feature vector and a growth state time series correlation feature vector;

[0014] The soil index time series correlation feature vector and the growth state time series correlation feature vector are subjected to feature interaction to obtain the soil-crop growth state interaction feature vector.

[0015] Preferably, the soil index data include temperature, humidity, pH value and nutrient content.

[0016] Preferably, the growth status parameters include chlorophyll content, leaf area index and transpiration rate.

[0017] Preferably, full time series correlation analysis is performed on the soil indicator full time series input matrix and the growth state parameter full time series input matrix respectively to obtain the soil indicator time series correlation feature vector and the growth state time series correlation feature vector, including: passing the soil indicator full time series input matrix through a soil indicator time series feature extractor based on a first convolutional neural network model to obtain the soil indicator time series correlation feature vector; passing the growth state parameter full time series input matrix through a growth state time series feature extractor based on a second convolutional neural network model to obtain the growth state time series correlation feature vector.

[0018] Preferably, feature interaction is performed on the soil indicator time series association feature vector and the growth state time series association feature vector to obtain the soil-crop growth state interaction feature vector, including: inputting the soil indicator time series association feature vector and the growth state time series association feature vector into a deep feature coupling network to obtain a soil indicator-growth state fine-grained feature interaction matrix; inputting the soil indicator-growth state fine-grained feature interaction matrix into an attention network guided by external knowledge to obtain an external knowledge optimized soil indicator-growth state fine-grained feature interaction matrix; inputting the soil indicator time series association feature vector and the growth state time series association feature vector into a Transformer-guided fine-grained feature enhancement module respectively to obtain an optimized soil indicator time series association feature vector and an optimized growth state time series association feature vector; and feature interaction modulation is performed on the optimized soil indicator time series association feature vector and the optimized growth state time series association feature vector to obtain the soil-crop growth state interaction feature vector.

[0019] Preferably, the soil indicator time series correlation feature vector and the growth state time series correlation feature vector are input into a deep feature coupling network to obtain a soil indicator-growth state fine-grained feature interaction matrix, including: multiplying the transposed vectors of the soil indicator time series correlation feature vector and the growth state time series correlation feature vector to obtain a soil indicator-growth state fine-grained feature interaction matrix.

[0020] Preferably, the soil indicator time series association feature vector and the growth state time series association feature vector are respectively input into the Transformer-guided fine-grained feature enhancement module to obtain the optimized soil indicator time series association feature vector and the optimized growth state time series association feature vector, including: performing a linear transformation on the soil indicator time series association feature vector to obtain a soil indicator time series semantic query feature vector and a soil indicator time series semantic value feature vector; using the external knowledge-optimized soil indicator-growth state fine-grained feature interaction matrix as a key matrix, transforming the soil indicator time series semantic query feature vector and the soil indicator time series semantic value feature vector into a linear transformation module; The key matrix is ​​input into the Transformer-guided fine-grained feature enhancement module to obtain the optimized soil indicator time series association feature vector; the growth state time series association feature vector is linearly transformed to obtain the growth state time series semantic query feature vector and the growth state time series semantic value feature vector; the external knowledge-optimized soil indicator-growth state fine-grained feature interaction matrix is ​​used as the key matrix, and the growth state time series semantic query feature vector, the growth state time series semantic value feature vector and the key matrix are input into the Transformer-guided fine-grained feature enhancement module to obtain the optimized growth state time series association feature vector.

[0021] Preferably, generating irrigation control instructions based on the soil-crop growth state interaction feature vector includes: passing the soil-crop growth state interaction feature vector through a classifier to obtain a classification result, wherein the classification result is used to indicate whether to turn on irrigation; and generating the irrigation control instructions based on the classification result.

[0022] In the second aspect, an intelligent control system based on agricultural big data is provided, which includes:

[0023] A data acquisition module, used to acquire soil index data at a plurality of predetermined time points within a predetermined time period collected by a soil sensor, and growth state parameters of crops at the plurality of predetermined time points collected by a crop sensor;

[0024] A feature extraction and feature interaction module, used for performing feature extraction and feature interaction on the soil index data at the plurality of predetermined time points and the growth state parameters of the crops at the plurality of predetermined time points to obtain a soil-crop growth state interaction feature vector;

[0025] The irrigation control instruction generation module is used to generate irrigation control instructions based on the soil-crop growth state interaction feature vector.

[0026] The soil index data at the plurality of predetermined time points and the growth state parameters of the crops at the plurality of predetermined time points are subjected to feature extraction and feature interaction to obtain a soil-crop growth state interaction feature vector, including:

[0027] Arranging the soil index data at the plurality of predetermined time points into a soil index full time series input matrix according to the time dimension and the index sample dimension, and arranging the growth state parameters of the crops at the plurality of predetermined time points into a growth state parameter full time series input matrix according to the time dimension and the state parameter;

[0028] Performing full time series correlation analysis on the soil index full time series input matrix and the growth state parameter full time series input matrix respectively to obtain a soil index time series correlation feature vector and a growth state time series correlation feature vector;

[0029] The soil index time series correlation feature vector and the growth state time series correlation feature vector are subjected to feature interaction to obtain the soil-crop growth state interaction feature vector.

[0030] This application has at least the following technical effects:

[0031] Compared with the prior art, the present application obtains soil index data at multiple predetermined time points within a predetermined time period collected by a soil sensor, and growth status parameters of crops at the multiple predetermined time points collected by a crop sensor; performs feature extraction and feature interaction on the soil index data at the multiple predetermined time points and the growth status parameters of crops at the multiple predetermined time points to obtain a soil-crop growth status interaction feature vector; generates irrigation control instructions based on the soil-crop growth status interaction feature vector; utilizes soil index data (temperature, humidity, pH value and nutrient content) and growth status parameters of crops (chlorophyll content, leaf area index and transpiration rate), and utilizes an intelligent algorithm for analysis, and generates corresponding irrigation control instructions to automatically adjust the irrigation operation of farmland, thereby achieving agricultural optimization and efficient production. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0033] Figure 1 This is a flow chart of an intelligent control method based on agricultural big data according to an embodiment of the present application.

[0034] Figure 2 This is a schematic diagram of the architecture of an intelligent control method based on agricultural big data according to an embodiment of the present application.

[0035] Figure 3 It is a block diagram of an intelligent control system based on agricultural big data according to an embodiment of the present application. DETAILED DESCRIPTION

[0036] The following will describe the technical solutions in the embodiments of the present application in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0037] Unless otherwise specified, all technical and scientific terms used in the embodiments of the present application have the same meaning as those commonly understood by those skilled in the art of the present application. The terms used in this application are only for the purpose of describing specific embodiments and are not intended to limit the scope of this application.

[0038] In the embodiments of the present application, it should be noted that, unless otherwise specified and limited, the term "connection" should be understood in a broad sense. For example, it can be an electrical connection or a connection between two components. It can be a direct connection or an indirect connection through an intermediate medium. For ordinary technicians in this field, the specific meanings of the above terms can be understood according to the specific circumstances.

[0039] It should be noted that the terms "first\second\third" involved in the embodiments of the present application are only used to distinguish similar objects, and do not represent a specific order for the objects. It is understandable that the specific order or sequence of "first\second\third" can be interchanged where permitted. It should be understood that the objects distinguished by "first\second\third" can be interchanged where appropriate, so that the embodiments of the present application described herein can be implemented in an order other than those illustrated or described herein.

[0040] In modern agricultural production, irrigation is a vital link in agricultural production. In arid areas or areas with seasonal water shortages, irrigation can make up for the lack of precipitation and ensure the normal growth and development of crops. Through reasonable irrigation control, sufficient water can be provided to meet the growth needs of crops and ensure the continuity and stability of agricultural production.

[0041] In order to better control the timing of irrigation, the technical solution of this application is proposed; specifically, Figure 1This is a flow chart of an intelligent control method based on agricultural big data according to an embodiment of the present application. Figure 2 FIG. 1 is a schematic diagram of the architecture of an intelligent control method based on agricultural big data according to an embodiment of the present application. Figure 1 and Figure 2 As shown, the intelligent control method based on agricultural big data includes: 110, acquiring soil index data at multiple predetermined time points within a predetermined time period collected by a soil sensor, and growth status parameters of crops at the multiple predetermined time points collected by a crop sensor; 120, performing feature extraction and feature interaction on the soil index data at the multiple predetermined time points and the growth status parameters of crops at the multiple predetermined time points to obtain a soil-crop growth status interaction feature vector; 130, generating irrigation control instructions based on the soil-crop growth status interaction feature vector.

[0042] In step 110, the accuracy and reliability of the soil sensor and the crop sensor are ensured to obtain accurate data. Appropriate time intervals are arranged for data collection to capture changes in soil and crops. By obtaining soil index data and crop growth status parameters, the soil moisture status, nutrient content and crop growth status can be understood, providing basic data for subsequent irrigation control.

[0043] In step 120, a suitable feature extraction method is selected, such as statistical features, frequency domain features or time domain features, to extract useful features from the raw data. For the interaction of soil and crop features, data mining or machine learning methods, such as feature combination, feature selection or feature fusion, can be used. Through feature extraction and feature interaction, the soil and crop data can be converted into more representative and interpretable feature vectors, thereby improving the information content of the data and providing more accurate input for subsequent irrigation control modeling and decision-making.

[0044] In step 130, according to the specific irrigation needs and goals, a suitable irrigation control algorithm or model is selected, such as a rule-based method, a statistical method, or a machine learning method. Consider factors such as the time, frequency, and amount of irrigation, as well as the water requirement of the crop and the moisture condition of the soil. Intelligent irrigation regulation can be achieved by generating irrigation control instructions based on the soil-crop growth state interaction feature vector. This can reduce manual intervention, improve the accuracy and efficiency of irrigation, avoid the problems of over-irrigation and under-irrigation, and thus improve the water resource utilization efficiency of farmland and the growth quality and yield of crops.

[0045] The intelligent control method based on agricultural big data obtains soil and crop data, extracts features and interacts with features, and finally generates irrigation control instructions to achieve reasonable irrigation. Reasonable irrigation can provide crops with an appropriate amount of water to meet the growth needs of crops. Sufficient water can promote plant photosynthesis and nutrient absorption, and increase crop growth rate and yield. At the same time, reasonable irrigation can also reduce crop stress response, reduce the risk of pests and diseases, and improve crop quality and commerciality.

[0046] In response to the above technical problems, the technical concept of this application is to utilize soil indicator data (temperature, humidity, pH value and nutrient content) and crop growth status parameters (chlorophyll content, leaf area index and transpiration rate), and use intelligent algorithms to analyze and generate corresponding irrigation control instructions to automatically adjust the irrigation operation of the farmland and achieve optimization and efficiency of agricultural production.

[0047] Specifically, soil indicator data can provide information about soil moisture conditions, soil texture, nutrient supply, etc., and analyze the moisture content and available water of the soil to determine whether the soil needs irrigation operations. At the same time, it is not enough to consider only the elements of the soil. For different varieties of crops, the irrigation requirements are also different. The growth status parameters of crops, such as chlorophyll content, leaf area index and transpiration rate, can reflect the growth status and water requirements of crops. Specifically, the chlorophyll content can reflect the photosynthesis intensity of crops, the leaf area index can reflect the leaf coverage of crops, and the transpiration rate can reflect the water evaporation rate of crops. By monitoring and analyzing these parameters, the current water demand of crops can be determined, and corresponding irrigation regulation can be carried out according to the demand.

[0048] Specifically, in step 110, soil index data at a plurality of predetermined time points within a predetermined time period collected by a soil sensor and growth status parameters of crops at the plurality of predetermined time points collected by a crop sensor are acquired.

[0049] Based on this, in the technical solution of the present application, first, soil index data at multiple predetermined time points within a predetermined time period collected by a soil sensor are obtained, wherein the soil index data include temperature, humidity, pH value and nutrient content; at the same time, growth status parameters of crops at the multiple predetermined time points collected by a crop sensor are obtained, wherein the growth status parameters include chlorophyll content, leaf area index and transpiration rate.

[0050] Soil indicator data collected by soil sensors, such as soil moisture, soil temperature, and soil salinity, can be used to understand the moisture status of the soil and the water absorption capacity of crop roots. At the same time, the growth status parameters collected by crop sensors, such as leaf temperature, chlorophyll content, and crop height, can be used to evaluate the water requirement for crop growth. Combining these data, the amount of irrigation required for crops can be accurately calculated to avoid over-irrigation or under-irrigation and ensure that crops receive the right amount of water.

[0051] By monitoring soil index data and crop growth status parameters in real time, we can understand the rate of crop water consumption and soil water replenishment. Based on these data, the frequency of irrigation can be dynamically adjusted to ensure that irrigation matches the water needs of crops. For example, when soil moisture is high, the irrigation time can be delayed or the irrigation amount can be reduced; when soil moisture is low or the crop water demand is high, irrigation can be carried out in advance or the irrigation amount can be increased.

[0052] Data collected by soil sensors and crop sensors can also be used to determine the best time to irrigate. For example, by monitoring the changing trend of soil moisture and the water consumption of crops, it can be determined to irrigate during the period when the water consumption of crops is lower, so as to improve water use efficiency and reduce water loss.

[0053] Through irrigation control instructions generated based on real-time data, over-irrigation can be avoided, soil over-wetting and water accumulation can be prevented, and the risk of soil erosion and nutrient loss can be reduced. At the same time, insufficient irrigation can be avoided to ensure that crops receive sufficient water supply, thereby improving the growth quality and yield of crops. This can effectively save water resources and reduce water waste.

[0054] The irrigation control instructions generated by the data collected by soil sensors and crop sensors can achieve precise regulation of irrigation volume, real-time adjustment of irrigation frequency, optimization of irrigation time, and prevention of over-irrigation and waste of water resources, thereby improving the water resource utilization efficiency of farmland and the growth quality and yield of crops.

[0055] Specifically, in step 120, feature extraction and feature interaction are performed on the soil indicator data at the multiple predetermined time points and the growth status parameters of the crops at the multiple predetermined time points to obtain a soil-crop growth status interaction feature vector, including: arranging the soil indicator data at the multiple predetermined time points into a soil indicator full time series input matrix according to the time dimension and the indicator sample dimension, and arranging the growth status parameters of the crops at the multiple predetermined time points into a growth status parameter full time series input matrix according to the time dimension and the status parameter; performing full time series correlation analysis on the soil indicator full time series input matrix and the growth status parameter full time series input matrix to obtain a soil indicator time series correlation feature vector and a growth status time series correlation feature vector; performing feature interaction on the soil indicator time series correlation feature vector and the growth status time series correlation feature vector to obtain the soil-crop growth status interaction feature vector.

[0056] In this application, firstly, by performing correlation analysis on the soil index full time series input matrix and the growth state parameter full time series input matrix, the time series correlation between them can be found, which helps to understand the mutual influence and correlation between soil index and crop growth state.

[0057] Then, through association analysis, the time series correlation feature vectors of soil indicators and crop growth status can be extracted. These feature vectors capture the important correlation information between soil indicators and crop growth status, which is helpful for further analysis and prediction.

[0058] Next, the feature-by-feature attention layer is used to perform feature interaction based on the attention mechanism on the soil index temporal correlation feature vector and the growth status temporal correlation feature vector, which can further enhance the correlation between them. This helps to better capture the dynamic interaction relationship between soil and crops and provide a more accurate soil-crop growth status interaction feature vector.

[0059] Through the processing of the above steps, a more comprehensive and accurate soil-crop growth status interaction feature vector can be obtained, providing useful information for further farmland management and crop growth prediction.

[0060] Next, the soil index data at the plurality of predetermined time points are arranged into a soil index full time series input matrix according to the time dimension and the index sample dimension, and the growth state parameters of the crops at the plurality of predetermined time points are arranged into a growth state parameter full time series input matrix according to the time dimension and the state parameter. Specifically, the soil index data and the growth state parameters of the crops are respectively subjected to data structuring processing.

[0061] In one embodiment of the present application, full time series correlation analysis is performed on the soil indicator full time series input matrix and the growth state parameter full time series input matrix to obtain a soil indicator time series correlation feature vector and a growth state time series correlation feature vector, including: passing the soil indicator full time series input matrix through a soil indicator time series feature extractor based on a first convolutional neural network model to obtain the soil indicator time series correlation feature vector; passing the growth state parameter full time series input matrix through a growth state time series feature extractor based on a second convolutional neural network model to obtain the growth state time series correlation feature vector.

[0062] Then, the soil indicator full time series input matrix is ​​passed through a soil indicator time series feature extractor based on the first convolutional neural network model to obtain a soil indicator time series associated feature vector; at the same time, the growth state parameter full time series input matrix is ​​passed through a growth state time series feature extractor based on the second convolutional neural network model to obtain a growth state time series associated feature vector.

[0063] Specifically, the convolutional neural network model can effectively extract key features from the input data. By applying the first convolutional neural network model and the second convolutional neural network model to extract features of soil indicators and growth status parameters respectively, their temporal characteristics and correlation information can be captured.

[0064] By extracting the soil index time series correlation feature vector and the growth state time series correlation feature vector, the time series information of soil index and growth state parameters can be converted into more expressive feature vectors. These feature vectors will contain the correlation between soil index and growth state, which will help to better understand the dynamic interaction between soil and crops.

[0065] By merging the soil index time series correlation feature vector and the growth state time series correlation feature vector, a comprehensive soil-crop growth state interaction feature vector can be obtained. This feature vector will comprehensively consider the correlation information between soil index and crop growth state, providing a more comprehensive and accurate feature representation for subsequent irrigation control and farmland management.

[0066] By processing soil indicators and growth state parameters through a feature extractor based on a convolutional neural network model, and arranging and merging their time-series correlation feature vectors, a more expressive and comprehensive soil-crop growth state interaction feature vector can be obtained, thereby improving the effect of farmland management and irrigation control. Then, the soil indicator time-series correlation feature vector and the growth state time-series correlation feature vector are subjected to feature interaction to obtain the soil-crop growth state interaction feature vector. The feature interaction of the present application aims to improve the expressiveness of the model by deeply mining the fine-grained interaction relationship between features and combining external knowledge. This process not only strengthens the interaction between features, but also makes full use of external knowledge to guide model training, thereby significantly improving the performance of the model in processing complex tasks. Specifically, by introducing external knowledge (such as the experience of agricultural experts, historical irrigation data, etc.), feature interaction can more accurately capture the complex relationship between the soil indicator time-series correlation feature vector and the growth state time-series correlation feature vector at the micro level. This includes not only the association between directly observed data, but also those implicit but critical factors for crop growth. For example, the difference in water requirements of specific crop varieties at different development stages can be better understood through external knowledge. External knowledge modulation enables the model to better identify which combinations of soil conditions and crop states require irrigation the most, thereby improving the accuracy of predictions. For example, when soil moisture is near a critical value, the model can decide whether to irrigate based on best practices in similar situations in the past, rather than simply relying on current measurements. This experience-based learning helps avoid misjudgments and ensures that each irrigation decision is optimal. The guidance provided by external knowledge can help the system make more informed irrigation decisions and further optimize the use of water resources and other agricultural inputs. For example, external knowledge can point out that under certain specific conditions, even if the soil looks dry, it does not actually need to be irrigated immediately, or conversely, although the soil moisture is still acceptable, it should be irrigated in advance due to expected weather changes or the upcoming rapid growth period of crops. This can ensure the healthy growth of crops while saving precious water resources.

[0067] In one embodiment of the present application, feature interaction is performed on the soil indicator time series association feature vector and the growth state time series association feature vector to obtain the soil-crop growth state interaction feature vector, including: inputting the soil indicator time series association feature vector and the growth state time series association feature vector into a deep feature coupling network to obtain a soil indicator-growth state fine-grained feature interaction matrix; inputting the soil indicator-growth state fine-grained feature interaction matrix into an attention network guided by external knowledge to obtain an external knowledge optimized soil indicator-growth state fine-grained feature interaction matrix; inputting the soil indicator time series association feature vector and the growth state time series association feature vector into a Transformer-guided fine-grained feature enhancement module respectively to obtain an optimized soil indicator time series association feature vector and an optimized growth state time series association feature vector; and feature interaction modulation is performed on the optimized soil indicator time series association feature vector and the optimized growth state time series association feature vector to obtain the soil-crop growth state interaction feature vector.

[0068] In one embodiment of the present application, the soil index time series correlation feature vector and the growth state time series correlation feature vector are input into a deep feature coupling network to obtain a soil index-growth state fine-grained feature interaction matrix, including: multiplying the transposed vectors of the soil index time series correlation feature vector and the growth state time series correlation feature vector to obtain a soil index-growth state fine-grained feature interaction matrix. Specifically, the process can be expressed as:

[0069]

[0070] in, and are respectively the soil index time series correlation feature vector and the growth state time series correlation feature vector, It is the soil indicator-growth state fine-grained characteristic interaction matrix.

[0071] Specifically, through the deep feature coupling network, the system is able to deeply explore the subtle changes and interactions between soil indicators (such as temperature, moisture, pH value, nutrient content) and crop growth status (such as chlorophyll content, leaf area index, transpiration rate). This fine-grained interaction capture allows the model to more accurately reflect how soil conditions affect crop growth and vice versa. For example, under specific temperature conditions, certain crop varieties may exhibit higher transpiration rates, which can be more accurately modeled through deep feature coupling networks. The fine-grained feature interaction matrix provides richer information and helps improve the accuracy of irrigation control decisions. For example, when soil moisture is low but crops are in a rapid growth period, the model can more accurately identify the increased demand for water from crops and issue irrigation instructions in a timely manner to ensure that crops receive adequate water supply without wasting water resources. This high-precision prediction is critical for optimizing irrigation time and amount.

[0072] Specifically, the soil index-growth state fine-grained feature interaction matrix is ​​input into the attention network guided by external knowledge to obtain the external knowledge optimized soil index-growth state fine-grained feature interaction matrix. Specifically, the process can be expressed as:

[0073]

[0074] in, is the soil index-growth state fine-grained feature interaction matrix, and denotes the first external knowledge attention learnable memory parameter matrix and the second external knowledge attention learnable memory parameter matrix, represents the normalization function, Optimizing the soil indicator-growth state fine-grained feature interaction matrix for external knowledge.

[0075] Specifically, external knowledge (such as the experience of agricultural experts, historical irrigation data, crop growth cycles, etc.) is used to enhance the understanding of specific domain knowledge. By introducing this external information, the attention network guided by external knowledge can more accurately capture which soil and crop state combinations are most critical. This is not limited to the associations between directly observed data, but also includes those implicit but critical factors for crop growth. For example, the differences in water requirements of certain crop varieties at specific development stages can be better captured by external knowledge. This enhanced understanding allows the model to more accurately identify which soil conditions and crop state combinations are most in need of irrigation, thereby improving the accuracy of predictions. When soil moisture is near a critical value, the model can decide whether to irrigate based on best practices in similar situations in the past, rather than simply relying on current measurements. This experience-based learning helps avoid misjudgments and ensures that each irrigation decision is optimal. At the same time, external knowledge can also help the model more accurately judge the best time and amount of irrigation, further improving the quality of decision-making. By incorporating external knowledge, the attention network can not only learn the rules in the training data, but also draw on more widely applicable principles, thereby enhancing the model's adaptability to new environments and unknown situations.

[0076] In one embodiment of the present application, the soil indicator time series association feature vector and the growth state time series association feature vector are respectively input into the Transformer-guided fine-grained feature enhancement module to obtain an optimized soil indicator time series association feature vector and an optimized growth state time series association feature vector, including: performing a linear transformation on the soil indicator time series association feature vector to obtain a soil indicator time series semantic query feature vector and a soil indicator time series semantic value feature vector; using the external knowledge-optimized soil indicator-growth state fine-grained feature interaction matrix as a key matrix, The feature vector and the key matrix are input into the Transformer-guided fine-grained feature enhancement module to obtain the optimized soil indicator time series association feature vector; the growth state time series association feature vector is linearly transformed to obtain the growth state time series semantic query feature vector and the growth state time series semantic value feature vector; the external knowledge-optimized soil indicator-growth state fine-grained feature interaction matrix is ​​used as the key matrix, and the growth state time series semantic query feature vector, the growth state time series semantic value feature vector and the key matrix are input into the Transformer-guided fine-grained feature enhancement module to obtain the optimized growth state time series association feature vector. Specifically, the process can be expressed as follows:

[0077]

[0078]

[0079]

[0080]

[0081]

[0082]

[0083] in, and are respectively the soil index time series correlation feature vector and the growth state time series correlation feature vector, Optimize the soil indicator-growth state fine-grained feature interaction matrix for external knowledge, and They represent the soil index temporal semantic query weight matrix and the soil index temporal semantic query bias vector respectively. and represent the soil index time series semantic value weight matrix and soil index time series semantic value bias vector respectively, is matrix multiplication, and are soil index temporal semantic query feature vector and soil index temporal semantic value feature vector respectively. is the matrix transpose, for The scale is the matrix width multiplied by the matrix height, express function, is the optimized soil index time series correlation feature vector, and They represent the growth state temporal semantic query weight matrix and growth state temporal semantic query bias vector respectively, and They represent the weight matrix of the time series semantic value of the growth state and the bias vector of the time series semantic value of the growth state, and are the growth status temporal semantic query feature vector and the growth status temporal semantic value feature vector, respectively. is the optimized growth state time series correlation feature vector.

[0084] Specifically, firstly, the soil indicator temporal association feature vector and the growth state temporal association feature vector were linearly transformed to generate the corresponding semantic query feature vector and semantic value feature vector. These semantic feature vectors capture the important information in the original feature vector and provide a basis for subsequent feature interaction. Next, the system uses the soil indicator-growth state fine-grained feature interaction matrix obtained by the deep feature coupling network and optimized by the attention network guided by external knowledge as the key matrix. This key matrix contains the micro-level interaction information between soil and crop status and has been further optimized by external knowledge to ensure that it can reflect the most relevant and important association patterns. Then, the soil indicator temporal semantic query feature vector, the soil indicator temporal semantic value feature vector and the key matrix are input into the Transformer-guided fine-grained feature enhancement module. Inside this module, the self-attention mechanism allows the feature vector at each position to interact with the feature vectors at all other positions, thereby enhancing the ability of feature representation. In this way, not only the intrinsic connection between soil indicator features is strengthened, but also it can better benefit from external knowledge, improving the quality of the soil indicator temporal association feature vector. The same processing is also applied to the growth state temporal association feature vector. By performing similar processing on the growth state temporal semantic query feature vector, the growth state temporal semantic value feature vector, and the same key matrix, the system also enhances the intrinsic connection between the growth state features and enables the growth state temporal association feature vector to benefit from external knowledge, thereby improving its quality.

[0085] Specifically, the optimized soil index time series correlation feature vector and the optimized growth state time series correlation feature vector are subjected to feature interaction modulation to obtain the soil-crop growth state interaction feature vector. Specifically, the process can be expressed as follows:

[0086]

[0087] in, is the optimized soil index time series correlation feature vector, is the optimized growth state time series correlation feature vector, is the soil-crop growth state interaction feature vector.

[0088] Specifically, feature interaction modulation is performed by dividing by position, which aims to capture more subtle relationships between the two and generate a comprehensive feature vector to reflect the complex interaction between soil conditions and crop growth status. The operation of dividing by position allows in-depth exploration of the nonlinear relationship between soil indicators and crop growth status. For example, at a specific point in time, if the soil moisture is low but the crop transpiration rate is high, this may indicate that the crop is in a rapid growth period or facing some kind of stress. In this case, dividing by position can amplify these differences, allowing the model to more sensitively capture the subtle changes between soil and crop status. The soil-crop growth status interaction feature vector generated in this way contains richer information and helps to improve the accuracy of irrigation control decisions. For example, when soil moisture is near the critical value, the system can determine whether immediate irrigation is needed based on the information in the interaction feature vector. In addition, for those cases where the soil moisture is acceptable but the crop shows a high transpiration rate, the model can also identify potential needs, thereby making more accurate irrigation decisions.

[0089] Specifically, in the step 130, an irrigation control instruction is generated based on the soil-crop growth state interaction feature vector, including: passing the soil-crop growth state interaction feature vector through a classifier to obtain a classification result, wherein the classification result is used to indicate whether to turn on irrigation; and generating the irrigation control instruction based on the classification result.

[0090] By inputting the soil-crop growth state interaction feature vector into the classifier, a classification result can be obtained. This classification result can indicate whether irrigation should be turned on, that is, whether the current soil and crop state requires irrigation. The classifier can learn the pattern of irrigation demand based on the feature representation of the feature vector, thereby accurately determining whether irrigation is needed.

[0091] Based on the classification results, corresponding irrigation control instructions can be generated. If the classification results indicate that irrigation is required, the generated instructions may include a command to turn on the irrigation system and the corresponding irrigation amount. If the classification results indicate that irrigation is not required, the generated instructions may include a command to turn off the irrigation system. In this way, irrigation control instructions can be dynamically generated according to the real-time soil and crop status to achieve precise irrigation control.

[0092] In the technical solution of the present application, the soil indicator time series association feature vector and the growth state time series association feature vector respectively represent the time series association pattern characteristics between soil indicators and the time series association pattern characteristics of the growth state parameters of crops. In the process of using a feature fine-grained interaction network based on external knowledge modulation to perform feature interaction on the soil indicator time series association feature vector and the growth state time series association feature vector, the parameter sample dimension asymmetry and local encoding pattern difference between the soil indicator time series association feature vector and the growth state time series association feature vector will cause feature fine-grained interaction mismatch and make the soil-crop growth state interaction feature vector have a long-range class regression understanding representation imbalance, thereby deviating from the source time series semantic feature representation class understanding, affecting the accuracy of the classification result.

[0093] Preferably, inputting the soil-crop growth state interaction feature vector into a classifier to obtain a classification result comprises:

[0094] Determine the number of zero eigenvalues ​​in the soil-crop growth state interaction eigenvector, and respectively calculate the reciprocal of the logarithm of the number of zero eigenvalues ​​with base two and the exponential value of the reciprocal of the number of zero eigenvalues ​​with base natural constants to obtain a first soil-crop growth state interaction representation value and a second soil-crop growth state interaction representation value;

[0095] Calculate a power function of each eigenvalue of the soil-crop growth state interaction eigenvector with the inverse of the number of zero eigenvalues ​​as an exponent to obtain a soil-crop growth state interaction structural state vector;

[0096] Performing a dot multiplication of the soil-crop growth state interaction structure state vector and the first soil-crop growth state interaction representation value to obtain a soil-crop growth state interaction information representation vector;

[0097] Performing dot multiplication of the autocorrelation matrix of the soil-crop growth state interaction structure state vector with the second soil-crop growth state interaction representation value and weight hyperparameter respectively to obtain a soil-crop growth state interaction regression understanding matrix;

[0098] Performing matrix multiplication of the soil-crop growth state interaction information representation vector and the soil-crop growth state interaction regression understanding matrix to obtain an optimized soil-crop growth state interaction feature vector;

[0099] The optimized soil-crop growth state interaction feature vector is input into a classifier to obtain a classification result.

[0100] Specifically, the first soil-crop growth state interaction representation value and the second soil-crop growth state interaction representation value are expressed as:

[0101]

[0102]

[0103] in, is the number of zero eigenvalues ​​in the soil-crop growth state interaction eigenvector, is the first soil-crop growth state interaction representation value, is the second soil-crop growth state interaction representation value, is an exponential function with a natural constant as its base

[0104] And the optimized soil-crop growth state interaction feature vector is expressed as:

[0105]

[0106]

[0107]

[0108] in, is the soil-crop growth state interaction feature vector, is the weight hyperparameter, is the soil-crop growth state interactive regression understanding matrix, is the soil-crop growth state interaction information representation vector, is the number of zero eigenvalues ​​in the soil-crop growth state interaction eigenvector, is the first soil-crop growth state interaction representation value, is the second soil-crop growth state interaction representation value, represents a power function in which each eigenvalue of the soil-crop growth state interaction eigenvector is calculated with the inverse of the number of zero eigenvalues ​​as the exponent, represents the transpose of a vector, It is the point multiplication by position. is matrix multiplication, is the optimized soil-crop growth state interaction feature vector.

[0109] Based on this, by modeling the zero dimension of the vector field of the soil-crop growth state interaction feature vector as a global field redundant dependency, effective long-range modeling of the soil-crop growth state interaction feature vector under field linear complexity is achieved. In this way, in the feature regression process of the soil-crop growth state interaction feature vector, structured correlation self-distillation between information representation and class understanding is performed in a selective state space based on feature distribution, thereby maintaining the overall information complexity of the feature distribution of the soil-crop growth state interaction feature vector while promoting the long-range dynamic class understanding balance of its feature convergence, so as to improve the class understanding consistency between the classification regression and feature extraction processes, and improve the accuracy of the classification results obtained by inputting the soil-crop growth state interaction feature vector into the classifier.

[0110] Then, the soil-crop growth state interaction feature vector is passed through a classifier to obtain a classification result, and the classification result is used to indicate whether irrigation is turned on; and then an irrigation control instruction is generated based on the classification result.

[0111] In summary, the intelligent control method 100 based on agricultural big data according to the embodiment of the present application is explained, which utilizes soil index data (temperature, humidity, pH value and nutrient content) and growth status parameters of crops (chlorophyll content, leaf area index and transpiration rate), and uses intelligent algorithms for analysis, and generates corresponding irrigation control instructions to automatically adjust the irrigation operation of the farmland and achieve optimization and efficiency of agricultural production.

[0112] In one embodiment of the present application, Figure 3 FIG. 1 is a block diagram of an intelligent control system based on agricultural big data according to an embodiment of the present application. Figure 3 As shown, according to the embodiment of the present application, the intelligent control system 200 based on agricultural big data includes: a data acquisition module 210, which is used to obtain soil index data at multiple predetermined time points within a predetermined time period collected by a soil sensor, and growth status parameters of crops at the multiple predetermined time points collected by a crop sensor; a feature extraction and feature interaction module 220, which is used to perform feature extraction and feature interaction on the soil index data at the multiple predetermined time points and the growth status parameters of crops at the multiple predetermined time points to obtain a soil-crop growth status interaction feature vector; an irrigation control instruction generation module 230, which is used to generate an irrigation control instruction based on the soil-crop growth status interaction feature vector.

[0113] Specifically, in the intelligent control system based on agricultural big data, the soil index data includes temperature, humidity. The specific functions and operations of each unit and module in the intelligent control system based on agricultural big data have been referred to above. Figure 1 to Figure 2The intelligent control method based on agricultural big data has been introduced in detail, and therefore, its repeated description will be omitted.

[0114] As described above, the intelligent control system 200 based on agricultural big data according to the embodiment of the present application can be implemented in various terminal devices, such as a server for intelligent control based on agricultural big data. In one example, the intelligent control system 200 based on agricultural big data according to the embodiment of the present application can be integrated into the terminal device as a software module and / or a hardware module. For example, the intelligent control system 200 based on agricultural big data can be a software module in the operating system of the terminal device, or it can be an application developed for the terminal device; of course, the intelligent control system 200 based on agricultural big data can also be one of the many hardware modules of the terminal device.

[0115] Alternatively, in another example, the intelligent control system 200 based on agricultural big data and the terminal device may also be separate devices, and the intelligent control system 200 based on agricultural big data may be connected to the terminal device via a wired and / or wireless network, and transmit interactive information in accordance with an agreed data format.

[0116] It should also be noted that in the apparatus, device and method of the present application, each component or each step can be decomposed and / or recombined. Such decomposition and / or recombination should be regarded as equivalent solutions of the present application.

[0117] The above description of the disclosed aspects is provided to enable any person skilled in the art to make or use the present application. Various modifications to these aspects will be readily apparent to those skilled in the art, and the general principles defined herein may be applied to other aspects without departing from the scope of the present application. Therefore, the present application is not intended to be limited to the aspects shown herein, but rather to the widest scope consistent with the principles and novel features disclosed herein.

[0118] The above description has been given for the purpose of illustration and description. In addition, this description is not intended to limit the embodiments of the present application to the forms disclosed herein. Although multiple example aspects and embodiments have been discussed above, those skilled in the art will recognize certain variations, modifications, changes, additions and sub-combinations thereof.

Claims

1. An intelligent control method based on agricultural big data, characterized in that: include: Acquire soil index data collected by a soil sensor at a plurality of predetermined time points within a predetermined time period, and growth state parameters of crops collected by a crop sensor at the plurality of predetermined time points; Performing feature extraction and feature interaction on the soil index data at the plurality of predetermined time points and the growth state parameters of the crops at the plurality of predetermined time points to obtain a soil-crop growth state interaction feature vector; generating irrigation control instructions based on the soil-crop growth state interaction feature vector; The soil index data at the plurality of predetermined time points and the growth state parameters of the crops at the plurality of predetermined time points are subjected to feature extraction and feature interaction to obtain a soil-crop growth state interaction feature vector, including: Arranging the soil index data at the plurality of predetermined time points into a soil index full time series input matrix according to the time dimension and the index sample dimension, and arranging the growth state parameters of the crops at the plurality of predetermined time points into a growth state parameter full time series input matrix according to the time dimension and the state parameter; Performing full time series correlation analysis on the soil index full time series input matrix and the growth state parameter full time series input matrix respectively to obtain a soil index time series correlation feature vector and a growth state time series correlation feature vector; Performing feature interaction on the soil index time series correlation feature vector and the growth state time series correlation feature vector to obtain the soil-crop growth state interaction feature vector; Performing feature interaction on the soil index time series correlation feature vector and the growth state time series correlation feature vector to obtain the soil-crop growth state interaction feature vector, including: Inputting the soil index time series correlation feature vector and the growth state time series correlation feature vector into a deep feature coupling network to obtain a soil index-growth state fine-grained feature interaction matrix; Inputting the soil index-growth state fine-grained feature interaction matrix into an attention network guided by external knowledge to obtain an external knowledge optimized soil index-growth state fine-grained feature interaction matrix; Inputting the soil index time series correlation feature vector and the growth state time series correlation feature vector into the Transformer-guided fine-grained feature enhancement module respectively to obtain an optimized soil index time series correlation feature vector and an optimized growth state time series correlation feature vector; The optimized soil index time series correlation feature vector and the optimized growth state time series correlation feature vector are subjected to feature interaction modulation to obtain the soil-crop growth state interaction feature vector.

2. The intelligent control method based on agricultural big data according to claim 1 is characterized in that: The soil index data include temperature, humidity, pH value and nutrient content.

3. The intelligent control method based on agricultural big data according to claim 2 is characterized in that: The growth status parameters include chlorophyll content, leaf area index and transpiration rate.

4. The intelligent control method based on agricultural big data according to claim 3 is characterized in that: Performing full time series correlation analysis on the soil index full time series input matrix and the growth state parameter full time series input matrix respectively to obtain a soil index time series correlation feature vector and a growth state time series correlation feature vector, including: The soil indicator full time series input matrix is ​​passed through a soil indicator time series feature extractor based on a first convolutional neural network model to obtain the soil indicator time series correlation feature vector; The full time series input matrix of the growth state parameters is passed through a growth state time series feature extractor based on a second convolutional neural network model to obtain the growth state time series associated feature vector.

5. The intelligent control method based on agricultural big data according to claim 4 is characterized in that: The soil indicator time series correlation feature vector and the growth state time series correlation feature vector are input into a deep feature coupling network to obtain a soil indicator-growth state fine-grained feature interaction matrix, including: multiplying the transposed vectors of the soil indicator time series correlation feature vector and the growth state time series correlation feature vector to obtain a soil indicator-growth state fine-grained feature interaction matrix.

6. The intelligent control method based on agricultural big data according to claim 5 is characterized in that: The soil index time series correlation feature vector and the growth state time series correlation feature vector are respectively input into the Transformer-guided fine-grained feature enhancement module to obtain an optimized soil index time series correlation feature vector and an optimized growth state time series correlation feature vector, including: Performing a linear transformation on the soil indicator time series association feature vector to obtain a soil indicator time series semantic query feature vector and a soil indicator time series semantic value feature vector; The external knowledge-optimized soil indicator-growth state fine-grained feature interaction matrix is ​​used as a key matrix, and the soil indicator time series semantic query feature vector, the soil indicator time series semantic value feature vector and the key matrix are input into a Transformer-guided fine-grained feature enhancement module to obtain an optimized soil indicator time series association feature vector; Performing a linear transformation on the growth state time series associated feature vector to obtain a growth state time series semantic query feature vector and a growth state time series semantic value feature vector; The external knowledge-optimized soil indicator-growth state fine-grained feature interaction matrix is ​​used as a key matrix, and the growth state temporal semantic query feature vector, the growth state temporal semantic value feature vector and the key matrix are input into a Transformer-guided fine-grained feature enhancement module to obtain an optimized growth state temporal association feature vector.

7. The intelligent control method based on agricultural big data according to claim 6 is characterized in that: Based on the soil-crop growth state interaction feature vector, an irrigation control instruction is generated, including: Passing the soil-crop growth state interaction feature vector through a classifier to obtain a classification result, wherein the classification result is used to indicate whether irrigation is turned on; Based on the classification result, the irrigation control instruction is generated.

8. An intelligent control system based on agricultural big data, used to execute the intelligent control method based on agricultural big data as claimed in any one of claims 1 to 7, characterized in that: include: A data acquisition module, used to acquire soil index data at a plurality of predetermined time points within a predetermined time period collected by a soil sensor, and growth state parameters of crops at the plurality of predetermined time points collected by a crop sensor; A feature extraction and feature interaction module, used for performing feature extraction and feature interaction on the soil index data at the plurality of predetermined time points and the growth state parameters of the crops at the plurality of predetermined time points to obtain a soil-crop growth state interaction feature vector; The irrigation control instruction generation module is used to generate irrigation control instructions based on the soil-crop growth state interaction feature vector.

9. The intelligent control system based on agricultural big data according to claim 8 is characterized in that: The soil index data at the plurality of predetermined time points and the growth state parameters of the crops at the plurality of predetermined time points are subjected to feature extraction and feature interaction to obtain a soil-crop growth state interaction feature vector, including: Arranging the soil index data at the plurality of predetermined time points into a soil index full time series input matrix according to the time dimension and the index sample dimension, and arranging the growth state parameters of the crops at the plurality of predetermined time points into a growth state parameter full time series input matrix according to the time dimension and the state parameter; Performing full time series correlation analysis on the soil index full time series input matrix and the growth state parameter full time series input matrix respectively to obtain a soil index time series correlation feature vector and a growth state time series correlation feature vector; The soil index time series correlation feature vector and the growth state time series correlation feature vector are subjected to feature interaction to obtain the soil-crop growth state interaction feature vector.

Images