Intelligent control method and device for water and fertilizer of outdoor cabbage

Abstract

The application provides an open-field cabbage water and fertilizer intelligent regulation method and device, and applies to the crop water and fertilizer precision regulation technical field.The method comprises the following steps: synchronously collecting short-impending meteorological data, cabbage body growth image data and hierarchical soil moisture data of an open-field cabbage planting area; taking a natural day as a time reference, integrating and processing the collected multi-source heterogeneous data, and generating daily characteristic data of a unified scale and representing growth correlation characteristics of the cabbage; inputting the daily characteristic data into a multi-source fusion intelligent decision model to obtain irrigation regulation parameters and fertilization regulation parameters of the open-field cabbage; and performing water and fertilizer supply operations on the open-field cabbage according to the irrigation regulation parameters and the fertilization regulation parameters.

Description

Technical Field

[0001] This invention relates to the field of precision water and fertilizer regulation technology for crops, and in particular to a method and device for intelligent water and fertilizer regulation of open-field cabbage. Background Technology

[0002] Open-field cabbage is widely cultivated in the open field, with its growth environment completely exposed to the natural climate. However, global climate change has led to frequent extreme weather events such as high temperatures, droughts, rainstorms, and abnormally low temperatures. Fluctuations in temperature, light, and precipitation significantly affect the physiological processes of cabbage, such as transpiration and photosynthesis, as well as the formation of morphological processes such as rosette expansion and head compaction. This makes its water and fertilizer requirements highly dynamic and unpredictable.

[0003] Current water and fertilizer management for open-field cabbage mostly relies on experience-based operations or static model regulation. Some intelligent technologies make decisions based on only a single factor, such as carrying out irrigation based on soil moisture, formulating plans based on weather forecasts, or inferring needs through remote sensing monitoring of crop growth.

[0004] However, experience and static models are slow to respond, which can easily lead to an imbalance between water and fertilizer supply and demand. Furthermore, extensive operations can result in resource waste and environmental problems such as eutrophication of water bodies. Smart technologies driven by a single factor lack multi-dimensional collaborative perception and integrated decision-making on "climate change - crop growth - water and fertilizer demand". Soil moisture monitoring ignores the relationship between crop transpiration and climate, weather forecasts are divorced from actual crop response, and growth monitoring is constrained by cost and real-time performance and it is difficult to build accurate demand models. All of these cannot adapt to the dynamic water and fertilizer demand of open-field cabbage. Summary of the Invention

[0005] This invention provides a method and device for intelligent water and fertilizer regulation of open-field cabbage, which can solve the problems of water and fertilizer imbalance, resource waste and high environmental risks caused by the dynamic and varied water and fertilizer requirements of open-field cabbage due to climate change. It can achieve precise dynamic regulation of water and fertilizer, improve resource efficiency, ensure stable and high-quality production, and enhance climate adaptability.

[0006] This invention provides a method for intelligent water and fertilizer regulation of open-field cabbage, comprising: simultaneously collecting short-term meteorological data, cabbage growth image data, and stratified soil moisture data of the open-field cabbage planting area; integrating and processing the collected multi-source heterogeneous data using a natural day as the time base to generate daily-level feature data of uniform scale and characterizing the growth-related features of cabbage; inputting the daily-level feature data into a multi-source fusion intelligent decision model to obtain irrigation and fertilization regulation parameters for the open-field cabbage; and performing water and fertilizer supply operations on the open-field cabbage according to the irrigation and fertilization regulation parameters; wherein, the multi-source fusion intelligent decision model includes a meteorological feature extraction branch, a soil feature extraction branch, an image feature extraction branch, and a cross-modal feature fusion unit, the cross-modal feature fusion unit being used to adaptively adjust the fusion ratio of image features and meteorological features by using a gating weight generated by splicing meteorological features and soil features.

[0007] This invention also provides an intelligent water and fertilizer control device for open-field cabbage, comprising the following modules: a data acquisition module, a processing module, and a supply module; the data acquisition module is used to simultaneously acquire short-term meteorological data, cabbage growth image data, and stratified soil moisture data of the open-field cabbage planting area; the processing module is used to integrate and process the acquired multi-source heterogeneous data with natural days as the time base to generate daily feature data of a unified scale that characterizes the growth correlation features of cabbage; the daily feature data is input into a multi-source fusion intelligent decision model to obtain irrigation control parameters and fertilization control parameters for open-field cabbage; the supply module is used to perform water and fertilizer supply operations on the open-field cabbage according to the irrigation control parameters and the fertilization control parameters; wherein, the multi-source fusion intelligent decision model includes a meteorological feature extraction branch, a soil feature extraction branch, an image feature extraction branch, and a cross-modal feature fusion unit, the cross-modal feature fusion unit is used to adaptively adjust the fusion ratio of image features and meteorological features by using a gating weight generated by splicing meteorological features and soil features.

[0008] The present invention also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the intelligent water and fertilizer control method for open-field cabbage as described above.

[0009] The present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the intelligent water and fertilizer control method for open-field cabbage as described above.

[0010] The present invention also provides a computer program product, including a computer program, which, when executed by a processor, implements the intelligent water and fertilizer control method for open-field cabbage as described above.

[0011] The intelligent water and fertilizer regulation method and device for open-field cabbage provided by this invention achieves multi-dimensional collaborative perception of "climate change-crop growth-soil supply" by simultaneously collecting short-term meteorological data, cabbage growth image data, and stratified soil moisture data. This overcomes the limitations of existing technologies that rely on single-factor decision-making and avoids problems such as soil moisture monitoring ignoring the relationship between climate and crops and weather forecasts being out of sync with actual crop responses. By integrating multi-source heterogeneous data based on natural days to generate daily feature data, the time scale differences between different data sources are eliminated, providing a unified input basis that fits the growth pattern of cabbage for accurate decision-making and solving the defects of lag in response of empirical or static models. With the help of a multi-source fusion intelligent decision-making model containing three feature extraction branches and cross-modal feature fusion units, especially by adaptively adjusting the fusion ratio of image and meteorological features through the gating weight generated by splicing meteorological and soil features, the method can dynamically analyze the impact mechanism of climate on the physiological and water and fertilizer requirements of cabbage and accurately match the dynamically changing water and fertilizer requirements of open-field cabbage. In this way, water and fertilizer can be supplied on demand, effectively solving the problem of imbalance between water and fertilizer supply and demand, reducing environmental risks such as resource waste and eutrophication of water bodies, while enhancing adaptability to climate change and ensuring high yield and quality of open-field cabbage. Attached Figure Description

[0012] To more clearly illustrate the technical solutions in this 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 some embodiments of this invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0013] Figure 1 This is a flowchart illustrating the intelligent water and fertilizer control method for open-field cabbage provided by the present invention.

[0014] Figure 2 This is a flowchart illustrating the multi-source fusion intelligent decision-making model provided by the present invention;

[0015] Figure 3 This is a schematic diagram of the intelligent water and fertilizer regulation device for open-field cabbage provided by the present invention.

[0016] Figure 4 This is a schematic diagram of the structure of the electronic device provided by the present invention. Detailed Implementation

[0017] To make the objectives, technical solutions, and advantages of this application clearer, the technical solutions of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0018] It should be noted that in the embodiments of this application, the words "exemplary" or "for example" are used to indicate examples, illustrations, or explanations. Any embodiment or design scheme described as "exemplary" or "for example" in the embodiments of this application should not be construed as being more preferred or advantageous than other embodiments or design schemes. Specifically, the use of the words "exemplary" or "for example" is intended to present the relevant concepts in a specific manner.

[0019] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element. Furthermore, it should be noted that the scope of the methods and apparatuses in the embodiments of this application is not limited to performing functions in the order shown or discussed, but may also include performing functions substantially simultaneously or in the reverse order, depending on the functions involved. For example, the described methods may be performed in a different order than described, and various steps may be added, omitted, or combined. Additionally, features described with reference to certain examples may be combined in other examples.

[0020] To facilitate a clear description of the technical solutions of the embodiments of this application, the terms "first" and "second" are used in the embodiments of this application to distinguish the same or similar items with essentially the same function and effect. Those skilled in the art can understand that the terms "first" and "second" are not intended to limit the quantity or execution order.

[0021] This application describes some exemplary embodiments for illustrative purposes. It should be understood that this application may be implemented in other ways not specifically shown in the accompanying drawings.

[0022] like Figure 1As shown, this application provides a method for intelligent water and fertilizer regulation of open-field cabbage, which can be applied to an intelligent water and fertilizer regulation device for open-field cabbage. The method may include steps S101-S104:

[0023] S101, the open-field cabbage water and fertilizer intelligent control device simultaneously collects short-term meteorological data, cabbage growth image data, and stratified soil moisture data of the open-field cabbage planting area.

[0024] Optionally, short-term meteorological data may include air temperature (°C), relative humidity (%), photosynthetically active radiation (PAR) (μmol / m² / s), precipitation (mm / h), and wind speed (m / s); cabbage growth image data may include red-green-blue images (RGB images) of cabbage collected by devices such as cameras and mobile phones; stratified soil moisture data may include volumetric water content (VWC), electrical conductivity (EC), and soil temperature at the root zone and permeation zone.

[0025] Specifically, for short-term meteorological data, the intelligent water and fertilizer control device for open-field cabbage can capture indicators such as air temperature, humidity, photosynthetically active radiation, precipitation, and wind speed in real time through field meteorological sensors deployed in the open-field cabbage planting area. When extreme weather events such as high temperature (such as exceeding 35°C) are detected, it can automatically trigger real-time data upload to ensure that no key climate information is missed.

[0026] Based on the image data of the cabbage plant's growth, the open-field cabbage water and fertilizer intelligent control device can use a near-ground camera fixed in the field to take standard overhead images three times a day. At the same time, it can combine a mobile APP to take 45° oblique photos twice a week during the key growth stages of cabbage (such as the rosette stage and the heading stage). In case of extreme weather, additional photos will be taken to record the growth status of cabbage plants, such as plant height, leaf area, and leaf color (such as whether it is yellowing due to nitrogen deficiency, or whether the leaf tips are withered and potassium deficient).

[0027] For stratified soil moisture data, wireless sensors can be buried in the root layer (the main water and fertilizer absorption area of ​​cabbage) at a depth of 20cm and the permeation layer (the key layer to avoid water and fertilizer leakage and waste) at a depth of 40cm. The open-field cabbage water and fertilizer intelligent control device can monitor soil volumetric water content, soil electrical conductivity and soil temperature through wireless sensors. Under normal circumstances, data is collected once every 10 minutes. Before and after irrigation and fertilization, in order to accurately grasp the changes in soil condition, the frequency is increased to once every 5 minutes. In addition, the wireless sensors will be deployed in zones according to the differences in soil texture in the field to ensure comprehensive data coverage.

[0028] Meanwhile, all collected data will be appended with a unified UTC timestamp. When an extreme weather event is triggered, the open-field cabbage water and fertilizer intelligent control device can forcibly and synchronously retrieve the latest cabbage growth image data and soil moisture data within the first preset time period, achieving time matching of the three types of data.

[0029] S102, the open-field cabbage water and fertilizer intelligent control device uses the natural day as the time base to integrate and process the collected multi-source heterogeneous data, generating daily feature data of a unified scale that characterizes the growth-related features of cabbage.

[0030] The intelligent water and fertilizer control device for open-field cabbage uses the natural day as the time base to integrate and process multi-source heterogeneous data to generate daily-level feature data with a unified scale that characterizes the growth-related features of cabbage. This includes: statistical analysis of short-term meteorological data collected on the day to obtain a daily meteorological feature set, which includes daily average temperature, diurnal temperature range, and daily total photosynthetically active radiation integral; dynamic analysis of stratified soil moisture data collected on the day to generate a daily soil feature set, which includes the rate of change of soil moisture content and the trend of change of electrical conductivity; screening and fusion of cabbage growth image data collected on the day, merging images from different time periods using a weighted method based on the imaging time period to generate a daily image feature set; and unifying the scale and format of the daily meteorological feature set, the daily soil feature set, and the daily image feature set, removing redundant information to form the daily-level feature data.

[0031] Specifically, in order to eliminate the temporal scale differences of multi-source heterogeneous data and construct input features that are in sync with agricultural management decisions, spatiotemporal alignment can be performed on the different source data collected by S101. Through spatiotemporal alignment, the instantaneous changes of meteorological elements, the diurnal evolution of image information, and the continuous dynamics of soil conditions are unified to the same scale, forming a feature representation consistent with the logic of agricultural operations, and providing an interpretable spatiotemporal context for the model.

[0032] Optionally, based on the physiological rhythms of cabbage and agricultural management practices, the basic time unit can be the natural day to perform spatiotemporal alignment of multi-source heterogeneous data:

[0033] For short-term meteorological data, the intelligent water and fertilizer control device for open-field cabbage can statistically analyze the temperature, humidity, and photosynthetically active radiation collected on the same day, and calculate the daily average temperature, diurnal temperature range, and daily total photosynthetically active radiation integral.

[0034] It should be noted that the average daily temperature is used to reflect the average intensity of daytime heat stress in order to model the basic impact of temperature on transpiration. For example, when the average daily temperature is above 30°C, transpiration of cabbage is enhanced. The diurnal temperature range is the difference between the highest and lowest air temperatures of the day, used to reflect the stimulating effect on the physiological activities of cabbage. For example, a temperature difference of more than 8°C is required during the heading stage to promote nutrient accumulation. The total daily photosynthetically active radiation integral is used to quantify the total amount of light energy input, which is directly related to the photosynthetic capacity of cabbage. The higher the radiation, the greater the fertilizer requirement.

[0035] Furthermore, the intelligent water and fertilizer control device for open-field cabbage can also use preset high temperature intensity to transform extreme weather events into continuous characteristics, thereby obtaining cumulative stress intensity. The cumulative stress intensity is used to quantify the cumulative stress effect of preset high temperature on water demand.

[0036] Specifically, the intelligent water and fertilizer control device for open-field cabbage can use the intensity of high temperature to transform extreme events into continuous characteristics, quantifying the cumulative stress effect of high temperature on water demand.

[0037] For example, at temperatures above 35°C, accumulating 1 unit of stress intensity per hour:

[0038] ;

[0039] in, Indicates the cumulative stress intensity. This indicates the instantaneous air temperature.

[0040] For stratified soil moisture data, the intelligent water and fertilizer control device for open-field cabbage can calculate the rate of change of soil moisture content and the trend of change of electrical conductivity in the root layer and the permeation layer. The rate of change of soil moisture content is used to quantify the dynamics of water potential in the root layer to infer changes in crop transpiration intensity and soil water retention. The trend of change of electrical conductivity is used to reflect fluctuations in soil fertility. For example, a decrease in electrical conductivity after irrigation indicates that fertilizer has been diluted or leached.

[0041] For example, the formula for calculating the rate of change of soil moisture content can be:

[0042] ;

[0043] in, This indicates the volumetric water content of the soil layer at the beginning of the day. This indicates the volumetric water content of the soil layer at the end of the day. Indicates a time interval.

[0044] For the image data of the growth of cabbage, the open-field cabbage water and fertilizer intelligent control device can filter the images collected at different times of the day, give higher weight to the characteristics of sufficient midday light and clear imaging, generate a single daily image through weighted fusion, and then extract features directly related to the growth status such as plant height, leaf area ratio, and leaf color from it.

[0045] For example, the formula for feature extraction and time-weighted fusion of three standard overhead views per day is:

[0046] ;

[0047] in, This represents the feature vector of a single cabbage growth image obtained after feature extraction. This represents the weight value of the k-th acquired image during fusion. This represents an image feature extraction network. This represents the image acquired in the kth acquisition. Indicates the image acquisition time. The midday time is represented by τ, which controls the time decay weight to ensure that the noon image has the highest weight.

[0048] Finally, the open-field cabbage water and fertilizer intelligent control device can unify the scale and format of the above-processed meteorological, soil, and image feature data, remove redundant information that is irrelevant to cabbage growth, and form a complete set of daily feature data.

[0049] It should be noted that the above steps not only solve the problem of inconsistent time scales of multi-source heterogeneous data, making it difficult to use directly for decision-making; at the same time, the generated daily feature data focuses on indicators related to cabbage growth, avoiding the disorder of the original data, and providing accurate and unified input basis for the subsequent multi-source fusion intelligent decision-making model. Compared with relying on scattered data, it can significantly improve the accuracy of subsequent water and fertilizer regulation decisions and reduce regulation deviations caused by data disorder.

[0050] S103, the open-field cabbage water and fertilizer intelligent control device inputs the daily feature data into the multi-source fusion intelligent decision-making model to obtain the irrigation control parameters and fertilization control parameters for open-field cabbage.

[0051] The multi-source fusion intelligent decision-making model includes a meteorological feature extraction branch, a soil feature extraction branch, an image feature extraction branch, and a cross-modal feature fusion unit. The cross-modal feature fusion unit is used to adaptively adjust the fusion ratio of image features and meteorological features by using a gating weight generated by splicing meteorological features and soil features.

[0052] Optionally, the multi-source fusion intelligent decision-making model includes three parallel feature extraction branches, which realize cross-modal feature interaction through dynamic gating fusion, and finally output the regulation decision by the water and fertilizer co-production prediction head.

[0053] Optionally, the input layer of the meteorological feature extraction branch receives a daily meteorological feature set, which is activated by the first fully connected layer and then embedded with an optical-thermal coupling unit to obtain fused meteorological features. These features are then input into the second fully connected layer. The optical-thermal coupling unit is used to analyze the synergistic mechanism between photosynthetically active radiation and temperature. A high-temperature feature enhancement channel is provided before the second fully connected layer. An additional subnet is activated through a gating mechanism to calculate thermal stress features and fuse them with the meteorological features output by the second fully connected layer to obtain enhanced features. The enhanced features are then merged with the main path output and linearly transformed to obtain a meteorological feature vector.

[0054] Specifically, such as Figure 2 As shown, this meteorological feature extraction branch processes the daily meteorological feature set. The input layer receives the daily meteorological feature set, and the first fully connected layer (64 nodes) uses ReLU activation. An optical-temperature coupling unit is embedded to realize nonlinear feature combination. The optical-temperature coupling unit is constructed as follows:

[0055] ;

[0056] in, This is a Sigmoid function used to map the coupling strength between photosynthetically active radiation and temperature to the (0, 1) interval; The input features include photosynthetically active radiation (RAEP) and a learnable weight matrix. Daily average temperature and its squared terms (Capturing nonlinear response), phototemperature product term , This is a bias term used to adjust the baseline value for the linear transformation.

[0057] It should be noted that the photothermal coupling term It is a coefficient between 0 and 1. If the photothermal coupling term... A higher light-temperature synergy indicates a good photosynthetic effect in cabbage, resulting in high photosynthetic efficiency and potentially a greater demand for water and fertilizer. Conversely, a lower light-temperature synergy indicates a poor photosynthetic effect (e.g., high temperature and strong light inhibit photosynthesis), affecting the physiological activities of cabbage and requiring corresponding changes in water and fertilizer needs.

[0058] Next, the optical-thermal coupling term will be... The fused meteorological features are obtained by splicing them with the original meteorological features. Input the second fully connected layer:

[0059] ;

[0060] in, This refers to the original meteorological characteristics, including basic meteorological parameters such as temperature, humidity, wind speed, and rainfall. Used for Each dimension is assigned a different "importance weight" to achieve a "weighted combination" of the original meteorological characteristics. The bias term is used to adjust the "baseline value" of the linear transformation, compensating for systematic deviations between the original and target features. Through linear transformation, the "dispersed original meteorological features" can be initially integrated into more representative intermediate meteorological features. The ReLU activation function introduces nonlinearity into the intermediate meteorological features after linear transformation and filters out invalid features. This represents channel-wise multiplication, which is the element-wise multiplication of two vectors of the same dimension.

[0061] It should be noted that after processing the original meteorological features through linear transformation and ReLU activation, and then weighting them with the photothermal coupling intensity, a fused meteorological feature that better reflects the physiological response of cabbage can be obtained. .

[0062] A high-temperature feature enhancement channel is added before the second fully connected layer (32 nodes). When the high-temperature index > 5, an additional 16-node subnet is activated through a gating mechanism.

[0063] ;

[0064] in, This represents the number of hours during which the temperature is higher than the preset temperature value (e.g., the cumulative number of hours above 35℃). This means that the daily high temperature index is truncated and normalized to make it more consistent with the physiological law of cabbage's "linear response followed by saturation" to high temperatures, thus avoiding excessive interference from extreme values. This represents the base temperature feature vector, which includes the daily average temperature. and highest temperature , Represents the weight matrix. This represents mapping temperature features to an 8-dimensional latent space. This represents the Sigmoid activation function. The values ​​∈[0,1] represent continuous gating values, which can more comprehensively reflect the growth stress of cabbage under hot conditions. The maximum activation intensity is reached when the value is >5.

[0065] when At that time, activate an additional 16-node subnet:

[0066] ;

[0067] in, This is the weight matrix. For bias terms, By multiplying elements one by one, the fusion of meteorological characteristics and thermal stress characteristics can be achieved.

[0068] Finally, the enhanced features will be... By merging with the main path output, a more comprehensive and accurate meteorological feature vector is obtained. :

[0069] ;

[0070] in, This is the weight matrix. This is a bias term.

[0071] On-demand scaling is achieved through a gating mechanism, keeping the model lightweight in non-high-temperature scenarios and activating a dedicated subnet only when encountering continuous thermal stress to capture complex responses such as accelerated evaporation and photosynthetic inhibition caused by high temperatures.

[0072] The output layer uses Batch Normalization to ensure uniform feature scale.

[0073] Optionally, the soil feature extraction branch takes a daily-level soil feature set as input, achieves nonlinear mapping through a fully connected layer combined with the GELU activation function, and uses hard-coded weights to force the negative influence of the rate of change of soil moisture content on the rate of change of electrical conductivity; the original rate of change signal is fused with the activation features through residual connections to obtain the soil feature vector.

[0074] Specifically, such as Figure 2 As shown, the soil feature extraction branch takes the rate of change of soil moisture content (ΔSWC) and the rate of change of electrical conductivity (ΔEC) as inputs. A nonlinear mapping is achieved through a 32-node fully connected layer combined with the GELU activation function. Hard-coded weights (-0.8) are used to enforce the negative influence of ΔSWC on ΔEC, simulating the fundamental physical laws of "drainage leading to increased salinity" and "irrigation leading to decreased salinity." Finally, the original rate of change signals are directly fused with the activation features through residual connections, enhancing gradient propagation efficiency while preserving the physical meaning of the observed data. The final output is a 32-dimensional soil feature vector to characterize the dynamic patterns of water and salt transport.

[0075] This design models the core mechanism of the water-salt diffusion-convection process through a minimalist architecture, making it suitable for irrigation decision support in edge computing scenarios. Hard-coded water-salt constraints enforce a negative correlation between ΔEC and ΔSWC (e.g., EC should increase when SWC decreases), and fixed negative weights are applied to the ΔSWC feature channels.

[0076] ;

[0077] in, This indicates a characteristic related to the rate of change in soil electrical conductivity.

[0078] The residual connection is:

[0079] ;

[0080] in, This indicates soil characteristics after applying the "water and fertilizer negative correlation constraint". This represents the original soil monitoring data. This represents the soil feature vector.

[0081] Optionally, the above image feature extraction branch includes an image segmentation and location coding layer, a Transformer coding layer, and a multi-scale feature enhancement layer; the image segmentation and location coding layer is used to segment the image into pixel blocks and assign different weights to the canopy spatial structure through location coding; the Transformer coding layer is used to establish a spatial association network between leaves based on a self-attention mechanism; the multi-scale feature enhancement layer is used to integrate global canopy coverage and local new leaf proportion to obtain image feature vectors.

[0082] Specifically, such as Figure 2 As shown, the image segmentation and location coding layer includes an image segmentation processing module and a spatial location coding module. The image segmentation processing module can uniformly divide the acquired cabbage growth image into independent 16×16 pixel blocks. By dividing the data into small blocks, the local details of the cabbage canopy are completely preserved, such as the unfolding shape of new leaves and whether the leaf edges show signs of nutrient deficiency or stress, avoiding the loss of detail information caused by large convolutional kernels in traditional CNNs. The spatial location coding module can assign coding information corresponding to the spatial location of each 16×16 pixel block in the image. In particular, it assigns higher location weights to the new leaf area in the center of the cabbage canopy and automatically reduces the weights to the soil background area near the image edge. This coding method accurately records the spatial structure of the cabbage canopy, effectively reducing the interference of the soil background on the extraction of crop growth features and ensuring that subsequent modules focus on the analysis of the growth status of the cabbage itself.

[0083] The Transformer encoding layer is a coding network layer with a self-attention mechanism at its core, containing a feature association calculation module and a feature weight allocation module. The feature association calculation module uses the self-attention mechanism to perform global spatial association analysis on the image features after block and positional encoding, establishing an association network between leaves in different regions of the cabbage canopy. The feature weight allocation module assigns weights to features in different regions based on the association strength, forming typical association patterns. For example, high association weights are assigned to the features of the central new leaf region and the stem base, reflecting the activity of the cabbage's growth points through the interaction of their features (the close association and vibrant features between the new leaves and the stem base indicate vigorous growth); strong association weights are established for the yellowing areas of the leaves and the vein features, accurately indicating whether the cabbage has symptoms of nitrogen or potassium deficiency by utilizing the distribution pattern of the yellowing areas along the veins, overcoming the limitation of traditional CNNs that can only capture local features and cannot establish global spatial associations.

[0084] The multi-scale feature enhancement layer comprises a global feature extraction submodule and a local feature extraction submodule. The global feature extraction submodule uses HSV color space analysis to globally quantify the color distribution and brightness of the cabbage canopy, calculating the canopy coverage area ratio in the image to reflect the density of the cabbage population (e.g., low canopy coverage during the rosette stage may indicate slow growth). The local feature extraction submodule employs a feature point matching algorithm to identify and match feature points such as the outline and texture of new leaves in the canopy, and statistically analyzes the proportion of new leaves in the canopy to characterize the growth vitality of individual cabbage plants (e.g., a high proportion of new leaves indicates vigorous metabolism). The two submodules are integrated to obtain an image feature vector, enabling multi-dimensional feature enhancement of cabbage growth and providing comprehensive image feature support for subsequent cross-modal feature fusion and water and fertilizer demand decisions.

[0085] Optionally, the cross-modal feature fusion unit achieves adaptive fusion of cross-modal features through a gating mechanism. The specific network structure is as follows: meteorological feature vectors and soil feature vectors are concatenated into a 64-dimensional vector, which is then processed by a sigmoid-activated fully connected layer (32 nodes) to generate gating weights. The fusion computation adopts a dual-path architecture: the main path multiplies the gating weights by the image feature vector, while the auxiliary path multiplies the difference between the gating weights and the meteorological feature vectors. Finally, the fused features are obtained by element-wise addition.

[0086] Optionally, the prediction head of the multi-source fusion intelligent decision-making model includes a shared hidden layer and a multi-task output structure. The specific network structure is as follows: a 128-node fully connected layer (ReLU activation) with fused feature inputs, divided into three control channels: the irrigation channel is directly connected to the linear output layer; the nitrogen fertilizer channel adds a 64-node subnet (ReLU activation) and a PAR mask multiplier; the potash fertilizer channel uses the same structure but connects to an EC constraint module. All output layers are post-loaded with a ReLU function to ensure non-negative outputs, and L2 regularization with a weight of 0.1 is added at the end.

[0087] It should be noted that the training of the multi-source fusion intelligent decision-making model adopts a two-stage strategy: first, fix the image feature extraction branch and pre-train the meteorological-soil fusion module, and then fine-tune the overall architecture end-to-end, combining the mean square error (irrigation amount) and quantile loss (fertilizer amount) loss functions to ensure the prediction robustness under extreme climate scenarios.

[0088] S104. The intelligent water and fertilizer control device for open-field cabbage performs water and fertilizer supply operations on open-field cabbage according to the irrigation control parameters and the fertilizer control parameters.

[0089] In this embodiment, by simultaneously collecting short-term meteorological data, cabbage growth image data, and stratified soil moisture data, a multi-dimensional collaborative perception of "climate change - crop growth - soil supply" is achieved. This overcomes the limitations of existing technologies that rely solely on single-factor decision-making and avoids problems such as soil moisture monitoring ignoring the relationship between climate and crops and weather forecasts being divorced from actual crop responses. By integrating multi-source heterogeneous data based on natural days to generate daily-level feature data, the time scale differences between different data sources are eliminated, providing a unified input basis that fits the growth pattern of cabbage for accurate decision-making and solving the defects of lag in response of empirical or static models. With the help of a multi-source fusion intelligent decision-making model containing three feature extraction branches and cross-modal feature fusion units, especially by adaptively adjusting the fusion ratio of image and meteorological features through the gating weight generated by splicing meteorological and soil features, the influence mechanism of climate on the physiological and water and fertilizer requirements of cabbage can be dynamically analyzed, and the dynamic water and fertilizer requirements of open-field cabbage can be accurately matched. In this way, water and fertilizer can be supplied on demand, effectively solving the problem of imbalance between water and fertilizer supply and demand, reducing environmental risks such as resource waste and eutrophication of water bodies, while enhancing adaptability to climate change and ensuring high yield and quality of open-field cabbage.

[0090] The foregoing mainly describes the solutions provided by the embodiments of this application from a methodological perspective. To achieve the above functions, it includes corresponding hardware structures and / or software modules for executing each function. Those skilled in the art should readily recognize that, in conjunction with the units and algorithm steps of the various examples described in the embodiments disclosed herein, the embodiments of this application can be implemented in hardware or a combination of hardware and computer software. Whether a function is executed in hardware or by computer software driving hardware depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0091] It should be noted that the device in the embodiments of this application includes a virtual device and a physical device. The virtual device can be an intelligent water and fertilizer control device for open-field cabbage, and the physical device can include electronic devices, computer storage media, and computer program products.

[0092] The intelligent water and fertilizer regulation method for open-field cabbage provided in this application embodiment can be executed by an intelligent water and fertilizer regulation device for open-field cabbage, or a control module for intelligent water and fertilizer regulation of open-field cabbage within the intelligent water and fertilizer regulation device. This application embodiment uses the intelligent water and fertilizer regulation device for open-field cabbage to execute the intelligent water and fertilizer regulation method for open-field cabbage as an example to illustrate the intelligent water and fertilizer regulation device for open-field cabbage provided in this application embodiment.

[0093] It should be noted that the embodiments of this application can divide the intelligent water and fertilizer control device for open-field cabbage into functional modules based on the above method examples. For example, each function can be divided into its own functional modules, or two or more functions can be integrated into one processing module. The integrated modules can be implemented in hardware or as software functional modules. Optionally, the module division in the embodiments of this application is illustrative and only represents one logical functional division; other division methods may be used in actual implementation.

[0094] like Figure 3 As shown in the figure, this application provides an intelligent water and fertilizer control device 300 for open-field cabbage. The intelligent water and fertilizer control device 300 for open-field cabbage includes: a data acquisition module 301, a processing module 302, and a supply module 303;

[0095] The acquisition module 301 is used to simultaneously acquire short-term meteorological data, cabbage growth image data, and stratified soil moisture data of the open-field cabbage planting area.

[0096] The processing module 302 is used to integrate and process the collected multi-source heterogeneous data with natural days as the time base to generate daily feature data of uniform scale that characterizes the growth correlation of cabbage; and input the daily feature data into the multi-source fusion intelligent decision model to obtain the irrigation control parameters and fertilization control parameters of open field cabbage.

[0097] The supply module 303 is used to perform water and fertilizer supply operations on open-field cabbage according to the irrigation control parameters and the fertilizer control parameters.

[0098] The multi-source fusion intelligent decision-making model includes a meteorological feature extraction branch, a soil feature extraction branch, an image feature extraction branch, and a cross-modal feature fusion unit. The cross-modal feature fusion unit is used to adaptively adjust the fusion ratio of image features and meteorological features by using a gating weight generated by splicing meteorological features and soil features.

[0099] Optionally, the processing module 302 is used to perform statistical analysis on the short-term meteorological data collected on the same day to obtain a daily meteorological feature set, which includes the daily average temperature, diurnal temperature range, and daily total photosynthetically active radiation integral; to perform dynamic analysis on the stratified soil moisture data collected on the same day to generate a daily soil feature set, which includes the rate of change of soil moisture content, the trend of change of electrical conductivity, the soil's ability to supply water and fertilizer, and the characteristics of leakage risk; to screen and fuse the cabbage plant growth image data collected on the same day, and to merge images from different time periods using a weighted method according to the imaging time period to generate a daily image feature set; and to unify the scale and format of the daily meteorological feature set, the daily soil feature set, and the daily image feature set, and to remove redundant information to form the daily feature data.

[0100] Optionally, the processing module 302 is used to convert extreme weather events into continuous features using a preset high temperature intensity to obtain cumulative stress intensity, wherein the cumulative stress intensity is used to quantify the cumulative stress effect of the preset high temperature on water demand.

[0101] Optionally, the input layer of the meteorological feature extraction branch receives a daily meteorological feature set, which is activated by the first fully connected layer and then embedded with an optical-thermal coupling unit to obtain fused meteorological features. These features are then input into the second fully connected layer. The optical-thermal coupling unit is used to analyze the synergistic mechanism between photosynthetically active radiation and temperature. A high-temperature feature enhancement channel is provided before the second fully connected layer. An additional subnet is activated through a gating mechanism to calculate thermal stress features and fuse them with the meteorological features output by the second fully connected layer to obtain enhanced features. The enhanced features are then merged with the main path output and linearly transformed to obtain a meteorological feature vector.

[0102] Optionally, the soil feature extraction branch takes a daily-level soil feature set as input, achieves nonlinear mapping through a fully connected layer combined with the GELU activation function, and uses hard-coded weights to force the negative influence of the rate of change of soil moisture content on the rate of change of electrical conductivity; the original rate of change signal is fused with the activation features through residual connections to obtain the soil feature vector.

[0103] Optionally, the image feature extraction branch includes an image segmentation and location coding layer, a Transformer coding layer, and a multi-scale feature enhancement layer; the image segmentation and location coding layer is used to segment the image into pixel blocks and assign different weights to the canopy spatial structure through location coding; the Transformer coding layer is used to establish a spatial association network between leaves based on a self-attention mechanism; the multi-scale feature enhancement layer is used to integrate global canopy coverage and local new leaf proportion to obtain an image feature vector.

[0104] In this embodiment, by simultaneously collecting short-term meteorological data, cabbage growth image data, and stratified soil moisture data, a multi-dimensional collaborative perception of "climate change - crop growth - soil supply" is achieved. This overcomes the limitations of existing technologies that rely solely on single-factor decision-making and avoids problems such as soil moisture monitoring ignoring the relationship between climate and crops and weather forecasts being divorced from actual crop responses. By integrating multi-source heterogeneous data based on natural days to generate daily-level feature data, the time scale differences between different data sources are eliminated, providing a unified input basis that fits the growth pattern of cabbage for accurate decision-making and solving the defects of lag in response of empirical or static models. With the help of a multi-source fusion intelligent decision-making model containing three feature extraction branches and cross-modal feature fusion units, especially by adaptively adjusting the fusion ratio of image and meteorological features through the gating weight generated by splicing meteorological and soil features, the influence mechanism of climate on the physiological and water and fertilizer requirements of cabbage can be dynamically analyzed, and the dynamic water and fertilizer requirements of open-field cabbage can be accurately matched. In this way, water and fertilizer can be supplied on demand, effectively solving the problem of imbalance between water and fertilizer supply and demand, reducing environmental risks such as resource waste and eutrophication of water bodies, while enhancing adaptability to climate change and ensuring high yield and quality of open-field cabbage.

[0105] Figure 4 An example is a schematic diagram of the physical structure of an electronic device, such as... Figure 4 As shown, the electronic device may include: a processor 410, a communications interface 420, a memory 430, and a communications bus 440, wherein the processor 410, the communications interface 420, and the memory 430 communicate with each other through the communications bus 440. The processor 410 can call logic instructions in the memory 430 to execute a smart water and fertilizer regulation method for open-field cabbage. This method includes: simultaneously collecting short-term meteorological data, cabbage growth image data, and stratified soil moisture data of the open-field cabbage planting area; integrating and processing the collected multi-source heterogeneous data using a natural day as the time base to generate daily-level feature data of a unified scale that characterizes the growth-related features of cabbage; inputting the daily-level feature data into a multi-source fusion intelligent decision model to obtain irrigation and fertilization regulation parameters for the open-field cabbage; and performing water and fertilizer supply operations on the open-field cabbage based on the irrigation and fertilization regulation parameters. The multi-source fusion intelligent decision model includes a meteorological feature extraction branch, a soil feature extraction branch, an image feature extraction branch, and a cross-modal feature fusion unit. The cross-modal feature fusion unit is used to adaptively adjust the fusion ratio of image features and meteorological features by using a gating weight generated by splicing meteorological and soil features.

[0106] Furthermore, the logical instructions in the aforementioned memory 430 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0107] On the other hand, the present invention also provides a computer program product, which includes a computer program that can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer can execute the intelligent water and fertilizer regulation method for open-field cabbage provided by the above methods. The method includes: synchronously collecting short-term meteorological data, cabbage growth image data, and stratified soil moisture data of the open-field cabbage planting area; integrating and processing the collected multi-source heterogeneous data with natural days as the time base to generate daily feature data of uniform scale and characterizing the growth correlation features of cabbage; inputting the daily feature data into a multi-source fusion intelligent decision model to obtain irrigation regulation parameters and fertilization regulation parameters for open-field cabbage; and performing water and fertilizer supply operations on open-field cabbage according to the irrigation regulation parameters and the fertilization regulation parameters. The multi-source fusion intelligent decision model includes a meteorological feature extraction branch, a soil feature extraction branch, an image feature extraction branch, and a cross-modal feature fusion unit. The cross-modal feature fusion unit is used to adaptively adjust the fusion ratio of image features and meteorological features by using a gating weight generated by splicing meteorological features and soil features.

[0108] In another aspect, the present invention also provides a non-transitory computer-readable storage medium storing a computer program thereon. When executed by a processor, the computer program implements the intelligent water and fertilizer regulation method for open-field cabbage provided by the above methods. The method includes: synchronously collecting short-term meteorological data, cabbage growth image data, and stratified soil moisture data of the open-field cabbage planting area; integrating and processing the collected multi-source heterogeneous data with a natural day as the time base to generate daily-level feature data of a unified scale that characterizes the growth-related features of cabbage; inputting the daily-level feature data into a multi-source fusion intelligent decision model to obtain irrigation regulation parameters and fertilization regulation parameters for open-field cabbage; and performing water and fertilizer supply operations on the open-field cabbage according to the irrigation regulation parameters and the fertilization regulation parameters. The multi-source fusion intelligent decision model includes a meteorological feature extraction branch, a soil feature extraction branch, an image feature extraction branch, and a cross-modal feature fusion unit. The cross-modal feature fusion unit is used to adaptively adjust the fusion ratio of image features and meteorological features by using a gating weight generated by splicing meteorological features and soil features.

[0109] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0110] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0111] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for intelligent regulation of water and fertilizer in open-field cabbage, characterized in that, include: Simultaneously collect short-term meteorological data, cabbage growth image data, and stratified soil moisture data from the open-field cabbage planting area; Using the natural day as the time base, the collected multi-source heterogeneous data are integrated and processed to generate daily feature data with a unified scale that characterizes the growth correlation features of cabbage. The daily-level feature data is input into the multi-source fusion intelligent decision-making model to obtain the irrigation and fertilization control parameters for open-field cabbage. Based on the irrigation control parameters and the fertilization control parameters, water and fertilizer supply operations are performed on open-field cabbage; The multi-source fusion intelligent decision model includes a meteorological feature extraction branch, a soil feature extraction branch, an image feature extraction branch, and a cross-modal feature fusion unit. The cross-modal feature fusion unit is used to adaptively adjust the fusion ratio of image features and meteorological features by using a gating weight generated by splicing meteorological features and soil features. The process involves integrating and processing multi-source heterogeneous data collected using the natural day as a time base to generate daily-level feature data of a unified scale that characterizes the correlation features of cabbage growth, including: Statistical analysis is performed on the short-term meteorological data collected on the same day to obtain a daily meteorological feature set, which includes the daily average temperature, diurnal temperature range, and daily total photosynthetically active radiation integral. Dynamic analysis is performed on the stratified soil moisture data collected on the same day to generate a daily soil feature set, which includes the rate of change of soil moisture content and the trend of change of electrical conductivity. The growth image data of cabbage collected on the same day were screened and fused. Images from different time periods were merged using a weighted method based on the imaging time period to generate a daily image feature set. The daily meteorological feature set, the daily soil feature set, and the daily image feature set are scaled and formatted, and redundant information is removed to form the daily feature data. The method further includes: using a preset high temperature intensity to transform extreme weather events into continuous characteristics to obtain cumulative stress intensity, wherein the cumulative stress intensity is used to quantify the cumulative stress effect of the preset high temperature on water demand.

2. The intelligent water and fertilizer regulation method for open-field cabbage according to claim 1, characterized in that, The input layer of the meteorological feature extraction branch receives a daily meteorological feature set. After being activated by the first fully connected layer, it is embedded with an optical-thermal coupling unit to obtain fused meteorological features, which are then input into the second fully connected layer. The optical-thermal coupling unit is used to analyze the synergistic mechanism between photosynthetically active radiation and temperature. A high-temperature feature enhancement channel is set in front of the second fully connected layer. An additional subnet is activated through a gating mechanism to calculate thermal stress features and fuse them with the meteorological features output by the second fully connected layer to obtain enhanced features. The enhanced features are merged with the main path output and then linearly transformed to obtain the meteorological feature vector.

3. The intelligent water and fertilizer regulation method for open-field cabbage according to claim 1, characterized in that, The soil feature extraction branch takes the daily soil feature set as input, achieves nonlinear mapping through a fully connected layer combined with the GELU activation function, and uses hard-coded weights to force the negative influence of the rate of change of soil moisture content on the rate of change of electrical conductivity; the original rate of change signal is fused with the activation features through residual connection to obtain the soil feature vector.

4. The intelligent water and fertilizer regulation method for open-field cabbage according to claim 1, characterized in that, The image feature extraction branch includes an image segmentation and positional encoding layer, a Transformer encoding layer, and a multi-scale feature enhancement layer. The image segmentation and positional encoding layer is used to segment the image into pixel blocks and assign different weights to the canopy spatial structure through positional encoding. The Transformer encoding layer is used to establish a spatial association network between leaves based on a self-attention mechanism. The multi-scale feature enhancement layer is used to integrate global canopy coverage and local new leaf proportion to obtain image feature vectors.

5. An intelligent water and fertilizer control device for open-field cabbage using the intelligent water and fertilizer control method as described in any one of claims 1 to 4, characterized in that, include: The module consists of a data acquisition module, a processing module, and a supply module. The acquisition module is used to simultaneously acquire short-term meteorological data, cabbage growth image data, and stratified soil moisture data of the open-field cabbage planting area. The processing module is used to integrate and process the collected multi-source heterogeneous data with natural days as the time base to generate daily feature data of uniform scale that characterizes the growth correlation of cabbage; the daily feature data is input into the multi-source fusion intelligent decision model to obtain the irrigation control parameters and fertilization control parameters of open field cabbage. The supply module is used to perform water and fertilizer supply operations on open-field cabbage according to the irrigation control parameters and the fertilizer control parameters. The multi-source fusion intelligent decision-making model includes a meteorological feature extraction branch, a soil feature extraction branch, an image feature extraction branch, and a cross-modal feature fusion unit. The cross-modal feature fusion unit is used to adaptively adjust the fusion ratio of image features and meteorological features by using a gating weight generated by splicing meteorological features and soil features.

6. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the intelligent water and fertilizer control method for open-field cabbage as described in any one of claims 1 to 4.

7. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the intelligent water and fertilizer control method for open-field cabbage as described in any one of claims 1 to 4.

8. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by the processor, it implements the intelligent water and fertilizer control method for open-field cabbage as described in any one of claims 1 to 4.

Images