An internet of things crop ion generation and delivery system based on ai data detection

Abstract

This invention discloses an IoT-based crop ion generation and delivery system using AI data detection, belonging to the field of smart agriculture IoT technology. The system includes a data acquisition module that acquires in-situ detection sequences from multiple ion concentration sensors in the crop root region and root environment parameter sequences; a state construction module that constructs an instantaneous absorption state vector based on the ion concentration fluctuation characteristics of all micro-regions and the root environment parameter sequences; a mode discrimination module that obtains absorption mode transition factors based on the geometric shape of the evolution trajectory of historical instantaneous absorption state vectors in vector space; a distribution prediction module that uses the absorption mode transition factors and the current instantaneous absorption state vector to predict the offset of crop rhizosphere ion concentration distribution at the next moment; and an execution control module that generates adjustment control quantities for the ion generator and delivery pipeline valves based on the offset. This scheme can reduce the impact of sensor noise on regulation and improve the responsiveness and spatial targeting of ion supply to dynamic changes in rhizosphere absorption.

Description

Technical Field

[0001] This invention relates to the field of smart agricultural Internet of Things (IoT) technology, specifically to an IoT-based crop ion generation and delivery system based on AI data detection. Background Technology

[0002] In the regulation of ion generation and transport in crop rhizosphere, existing technologies typically rely on single or average ion concentration detection values ​​in the root region, timed control, or fixed threshold comparisons to drive ion generators and transport pipeline valves. Some systems directly average or weight data from multiple sensors as the basis for ion supply regulation. This approach uses absolute concentration values ​​or simple statistical results as control inputs, failing to decompose the fluctuation components of the in-situ detection sequence and thus unable to extract the ion concentration fluctuation characteristics of each sensor's micro-region. Sensor drift, transient disturbances, and local concentration spikes can easily directly influence control decisions, causing frequent fluctuations or lags in ion supply. Spatial differences across multiple micro-regions are averaged, making it difficult to reflect the non-uniform changes in root absorption dynamics across the rhizosphere. Existing control methods lack measurement of the geometric shape of historical state vector evolution trajectories, failing to effectively identify absorption mode transitions. When root absorption demand undergoes a mode shift, ion generation and transport continue to operate according to the previous mode, easily leading to oversupply or undersupply.

[0003] To address the aforementioned issues, it is necessary to separate the fluctuation components from the in-situ detection sequences of multiple ion concentration sensors in the root region, obtain the ion concentration fluctuation characteristics of each micro-region, and integrate root environmental parameters to construct an instantaneous absorption state vector describing the dynamics of ion absorption in crop roots. Furthermore, it is also necessary to measure the geometric shape of the evolution trajectory of the state vector in the vector space based on the instantaneous absorption state vectors of multiple historical moments prior to the current moment, obtain the absorption mode transition factor, and use this factor and the current state vector to predict the offset of crop rhizosphere ion concentration distribution at the next moment, so as to guide the adjustment of ion generators and delivery pipeline valves.

[0004] Furthermore, existing systems typically rely on manual observation and experience to determine nutrient deficiencies in crops, lacking automatic deficiency identification and targeted nutrient supplementation mechanisms based on in-situ detection data. This makes it difficult to respond promptly in the early stages of deficiency, impacting normal crop growth and yield. Therefore, in addition to ion concentration regulation, it is necessary to enhance the system's capabilities by automatically determining deficiency states and generating nutrient supplementation strategies, enabling it to proactively replenish the specific nutrient ions required by crops. Summary of the Invention

[0005] To address the problems of existing crop ion generation and transport control relying on absolute concentration values ​​or simple statistics, failing to extract micro-region fluctuation characteristics, and having difficulty identifying the evolution of absorption patterns, this invention provides an IoT-based crop ion generation and transport system based on AI data detection. The aim is to predict the offset of rhizosphere ion concentration distribution by in-situ detection sequence fluctuation component decomposition and state vector evolution trajectory geometric measurement, and to generate adjustment and control quantities for ion generators and transport pipeline valves, so that the ion supply adapts to the dynamic changes in crop root absorption.

[0006] To achieve the above objectives, the present invention provides the following technical solution: The present invention provides an IoT crop ion generation and delivery system based on AI data detection, including a data acquisition module, a fluctuation decomposition module, a state construction module, a pattern discrimination module, a distribution prediction module, and an execution control module. The data acquisition module obtains in-situ detection sequences from multiple ion concentration sensors in the crop root region, along with root environmental parameter sequences at the corresponding acquisition times. The fluctuation decomposition module decomposes each in-situ detection sequence into fluctuation components, extracting ion concentration fluctuation characteristics of the micro-regions where each sensor is located. The state construction module constructs an instantaneous absorption state vector describing the dynamics of ion absorption in the crop roots based on the ion concentration fluctuation characteristics of all micro-regions and the root environmental parameter sequences. The mode discrimination module obtains the absorption mode transition factor at the current moment by measuring the geometric shape of the evolution trajectory of the state vector in the vector space, based on the instantaneous absorption state vectors at multiple historical moments prior to the current moment. The distribution prediction module uses the absorption mode transition factor and the instantaneous absorption state vector at the current moment to predict the distribution offset of crop rhizosphere ion concentration at the next moment. The execution control module generates adjustment control quantities for the ion generation and delivery system based on the ion concentration distribution offset, and applies them to the ion generator and delivery pipeline valves. This system integrates multi-micro-region ion concentration fluctuations with root environmental parameters, enabling it to identify transition trends in root absorption patterns, predict rhizosphere ion distribution shifts, and adjust ion generation and transport accordingly. This improves the temporal and spatial matching between ion supply and crop absorption needs, reducing regulatory lag and oversupply.

[0007] The fluctuation decomposition module arranges each in-situ detection sequence in chronological order to form the original concentration time series of the corresponding sensor. It then performs multi-scale sliding window segmentation on the original concentration time series, calculating the relative arrangement pattern of the sequence values ​​within each window to obtain a window pattern descriptor. All window pattern descriptors are sorted according to the center time of each window to form a pattern evolution sequence. Adjacent windows with the same descriptor in the pattern evolution sequence are merged, and the number of consecutive segments after merging is used as the ion concentration fluctuation characteristic of the micro-region where the sensor is located. This characteristic reflects the frequency of changes in the ion concentration fluctuation pattern, effectively expressing the ion perturbation characteristics of the root micro-region.

[0008] Preferably, the window pattern descriptor is generated as follows: all concentration values ​​within the window are sorted in ascending order to obtain the ranking of each value within the window; the ranking of each value within the window is arranged according to the original chronological order of the values ​​within the window to form a ranking vector; the ranking vector is converted into a binary encoded string of a preset length as the window pattern descriptor. This method can compactly encode the concentration fluctuation pattern and reduce the impact of noise on subsequent pattern recognition.

[0009] The state construction module concatenates the ion concentration fluctuation characteristics of all micro-regions at the same acquisition time according to the spatial arrangement of the sensors to form the concentration fluctuation distribution vector at that time; it also arranges various environmental parameters in the root environmental parameter sequence at the same acquisition time according to a preset order to form the environmental parameter vector at that time; finally, it concatenates the concentration fluctuation distribution vector and the environmental parameter vector end to end to obtain the instantaneous absorption state vector at that time. This state vector integrates the spatial fluctuation information of ion concentration and the root environment information, and can describe the dynamics of ion absorption in crop roots more completely.

[0010] The mode discrimination module arranges the instantaneous absorption state vectors from multiple consecutive historical moments prior to the current moment in chronological order, forming a state vector sequence. It calculates the transition difference metric between any two adjacent state vectors in the state vector sequence, obtaining a transition difference sequence. Each transition difference metric in the transition difference sequence is then accumulated chronologically to obtain the cumulative transition path length. Finally, the cumulative transition path length is divided by the direct difference metric between the first and last state vectors in the state vector sequence to obtain the absorption mode transition factor at the current moment. This transition factor reflects the geometric curvature and transition intensity of the state vector's evolution trajectory in the vector space, which is beneficial for capturing the trend of absorption modes transitioning from stable to drastic changes.

[0011] Preferably, the transition difference metric between two adjacent state vectors is obtained as follows: a vector difference operation is performed on two adjacent state vectors in the state vector sequence to obtain a difference vector; the sign of each element in the difference vector is calculated to form a sign sequence; adjacent positions in the sign sequence where the sign changes are marked, and an interval length sequence is constructed using the interval lengths between the marked positions; the mean of all interval lengths in the interval length sequence is used as the transition difference metric between the two adjacent state vectors. This difference metric reflects the consistency of the direction of change in each dimension of the state vector, making the trajectory geometry measurement more sensitive to multi-dimensional state changes.

[0012] The distribution prediction module acquires the absorption mode transition factors corresponding to a preset number of historical moments prior to the current moment, forming a historical sequence of transition factors. It calculates the differences between adjacent elements in this historical sequence to obtain a sequence of transition factor changes. Based on the numerical distribution range of the transition factor change sequence, it determines the fluctuation range boundary of the estimated transition factor value for the next moment. The boundary value with the smallest difference from the current moment's absorption mode transition factor within this fluctuation range boundary is selected as the predicted value for the next moment's transition factor. The predicted value for the next moment's transition factor is multiplied by the instantaneous absorption state vector of the current moment to obtain the distribution shift of crop rhizosphere ion concentration for the next moment. This prediction method determines a reasonable fluctuation range for the next moment by using the historical transition factor change range and imposes boundary constraints on the predicted value, thereby reducing the impact of abnormal fluctuations on the prediction results and improving the stability of the rhizosphere ion concentration distribution shift prediction.

[0013] Preferably, the boundary of the fluctuation range is determined as follows: all values ​​in the transition factor change sequence are arranged in ascending order to obtain an ordered change sequence; the first change value and the second change value located at preset quantiles are extracted from the ordered change sequence, wherein the first change value is less than the second change value; the sum of the current absorption mode transition factor and the first change value is used as the lower boundary of the fluctuation range; the sum of the current absorption mode transition factor and the second change value is used as the upper boundary of the fluctuation range. Using quantiles to determine the boundary can adapt to the distribution characteristics of historical change amplitudes and avoid excessively large prediction intervals caused by individual extreme changes.

[0014] The execution control module decomposes the ion concentration distribution offset according to the spatial location of each sensor, obtaining the desired concentration change direction for each micro-region. For each micro-region, the opening adjustment direction of the corresponding branch valve in the delivery pipeline is determined based on the sign of the desired concentration change direction. The sum of the absolute values ​​of the desired concentration change directions of all micro-regions is used as the total output adjustment range of the ion generator. The valve opening adjustment amount of each branch is allocated according to the proportion of the absolute value of the desired concentration change direction of each micro-region to the total output adjustment range of the ion generator. This control method can perform differentiated adjustment according to the ion offset direction and amplitude of different micro-regions in the rhizosphere, so that the total output of the ion generator and the valve opening of each branch move in a coordinated manner, improving the spatial targeting of ion delivery.

[0015] Preferably, the desired concentration change direction is obtained as follows: The offset component values ​​corresponding to the positions of each sensor are extracted from the ion concentration distribution offset; the sign of each offset component value is determined; if the offset component value is greater than zero, it is marked as a positive direction; if it is less than zero, it is marked as a negative direction; the spatial coordinates of the micro-region where each sensor is located are associated with the corresponding positive or negative direction mark for that micro-region to obtain the desired concentration change direction for that micro-region. By associating spatial coordinates with direction marks, the predicted offset can be mapped to specific delivery branches, which is beneficial for adjusting the supply according to the needs of each micro-region.

[0016] The technical effects and advantages provided by the present invention in the above technical solution are as follows: After constructing the original concentration time series for each in-situ detection sequence in chronological order, multi-scale sliding window segmentation is performed. Within each window, a window pattern descriptor is generated based on the relative arrangement of concentration values. Then, adjacent windows with the same descriptor are sorted and merged according to the center time of the window. The number of consecutive segments after merging is used as the ion concentration fluctuation characteristic of the micro-region. This processing transforms the absolute ion concentration values ​​into the ranking relationship within the window, reducing the interference of sensor baseline drift, transient noise, and local outliers on control decisions. The number of consecutive segments obtained after merging adjacent windows with the same pattern can reflect the complexity and frequency of changes in the ion concentration fluctuation structure of the micro-region, and better reflects the local dynamic changes caused by root absorption than directly using the concentration mean or variance. Based on this characteristic, a concentration fluctuation distribution vector is spliced ​​according to the spatial arrangement of the sensors, and then concatenated end-to-end with the root environmental parameter vector to form an instantaneous absorption state vector. This allows the state expression to simultaneously carry rhizosphere spatial distribution information and environmental condition information, making pattern discrimination and distribution prediction spatially targeted. After obtaining the instantaneous absorption state vectors from multiple consecutive historical moments prior to the current moment, the transition difference measure between adjacent state vectors is calculated and accumulated sequentially to obtain the cumulative transition path length. The cumulative transition path length is then divided by the direct difference measure between the first and last state vectors to obtain the absorption mode transition factor at the current moment. This ratio describes the curvature of the trajectory of the state vector in vector space; the more pronounced the curvature, the closer the absorption mode is to a non-stationary transition. Compared to fixed thresholds or single concentration change rate judgments, this factor can capture the dynamic transformation trend of absorption from the geometric shape of the historical state sequence, reducing sensitivity to anomalies at individual sampling points. When predicting the offset of the rhizosphere ion concentration distribution at the next moment, the adjacent differences in the historical sequence of transition factors are used to determine the boundary of the fluctuation interval, and the boundary with the smallest difference from the current transition factor is selected as the predicted value for the next moment. This value is then multiplied by the current instantaneous absorption state vector to obtain the offset. The predicted offset includes the components of each micro-division, which can be decomposed according to the spatial location of the sensor to obtain the desired concentration change direction. This further generates the corresponding branch valve opening adjustment direction and the total output adjustment amplitude of the ion generator, so that ion generation and transport spatially match the rhizosphere concentration change trend, reducing supply lag and local oversupply or undersupply when the absorption mode changes. Attached Figure Description

[0017] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this invention. For those skilled in the art, other drawings can be obtained based on these drawings.

[0018] Figure 1 This is a schematic diagram of an IoT-based crop ion generation and delivery system based on AI data detection. Figure 2This is a flowchart for extracting characteristic quantities of ion concentration fluctuations; Figure 3 It is a curve showing the change of characteristic quantities of ion concentration fluctuation in a multi-sensor micro-region; Figure 4 It is a measure of the difference in the evolution of crop root absorption state vector and the curve of the change of absorption mode transition factor; Figure 5 This is a schematic diagram showing the direction and distribution of the opening adjustment of the micro-zone valve. Detailed Implementation

[0019] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0020] See Figure 1 This invention provides an IoT crop ion generation and delivery system based on AI data detection, including a data acquisition module, a fluctuation decomposition module, a state construction module, a pattern discrimination module, a distribution prediction module, and an execution control module.

[0021] The data acquisition module acquires in-situ detection sequences from multiple ion concentration sensors in the crop root region, as well as root environmental parameter sequences at the corresponding acquisition times. The in-situ detection sequence is a data sequence formed by arranging the ion concentration detection values ​​obtained by a single sensor at continuous acquisition times in chronological order, while the root environmental parameter sequence is a data sequence formed by arranging the root environmental parameters in chronological order within the same acquisition time period.

[0022] The fluctuation decomposition module decomposes the fluctuation components of each in-situ detection sequence and extracts the ion concentration fluctuation characteristics of the micro-region where each sensor is located.

[0023] The state construction module constructs an instantaneous absorption state vector describing the dynamics of ion absorption in crop roots based on the ion concentration fluctuation characteristics of all microregions and the root environment parameter sequence.

[0024] The mode discrimination module obtains the absorption mode transition factor at the current moment by measuring the geometric shape of the evolution trajectory of the state vector in the vector space based on the instantaneous absorption state vectors of multiple historical moments before the current moment.

[0025] The distribution prediction module uses the current absorption mode transition factor and the current instantaneous absorption state vector to predict the distribution shift of crop rhizosphere ion concentration at the next moment.

[0026] The execution control module generates adjustment control quantities for the ion generation and delivery system based on the ion concentration distribution offset, and applies these quantities to the ion generator and delivery pipeline valves.

[0027] During system operation, the data acquisition module continuously acquires sensor data and root environmental parameter data, the fluctuation decomposition module decomposes the sensor data into fluctuation components, the state construction module forms an instantaneous absorption state vector based on fluctuation characteristics and environmental parameters, the mode discrimination module calculates the absorption mode transition factor using historical instantaneous absorption state vectors, the distribution prediction module obtains the ion concentration distribution offset based on the absorption mode transition factor and the current instantaneous absorption state vector, and the execution control module adjusts the ion generator and delivery pipeline valves based on the ion concentration distribution offset.

[0028] In specific implementation, please refer to Figure 2 The fluctuation decomposition module acquires the in-situ detection sequence of each ion concentration sensor and arranges the in-situ detection sequences in chronological order of acquisition time to form the original concentration time series corresponding to the ion concentration sensor.

[0029] The original concentration time series is segmented using a multi-scale sliding window method. This multi-scale sliding window segmentation employs a preset set of window lengths. Each window length value in the preset set is no greater than the length of the original concentration time series. Optionally, the preset window length set is determined based on the length of the original concentration time series and the sampling interval of the ion concentration sensor, ensuring that the time span corresponding to each window length value covers at least two sampling periods. For each window length value in the preset set, starting from the beginning of the original concentration time series, the window position is moved according to a preset sliding step until the end of the window reaches the end of the original concentration time series, resulting in multiple windows. Each window covers multiple consecutive concentration values ​​in the original concentration time series.

[0030] For any given window, sort all concentration values ​​within the window in ascending order to obtain the ranking of each concentration value within the window. The ranking starts from 1, with the smallest concentration value corresponding to ranking 1, and the largest concentration value corresponding to ranking the total number of concentration values ​​in the window.

[0031] The ranking of each concentration value within the window is arranged according to the original chronological order of the concentration values ​​within the window, forming a ranking vector. The rank vector is... The element is the first element in the window. The sorting position corresponding to the concentration value at each sampling time.

[0032] The positional arrangement vector is converted into a binary encoded string of a preset length, which serves as the window mode descriptor. During the conversion, each position in the positional arrangement vector is converted into a binary value of a fixed number of bits. All binary values ​​are then concatenated according to the positional arrangement vector to form the binary encoded string. The fixed number of bits satisfies... The smallest integer of , where The number of concentration values ​​within the window. The number of binary bits used for each sorting position. The length of the binary encoded string for the window mode descriptor. for: in, The number of concentration values ​​within the window. The number of binary bits used for each sorting position The length of the binary encoded string of the window mode descriptor. The value satisfies The smallest integer.

[0033] All window pattern descriptors are sorted according to the window center time to form a pattern evolution sequence. The window center time is the midpoint between the acquisition time of the first concentration value and the acquisition time of the last concentration value within the window.

[0034] Adjacent windows with the same descriptors in the pattern evolution sequence are merged. During the merging process, starting with the first window pattern descriptor in the pattern evolution sequence, adjacent window pattern descriptors are compared sequentially. If two adjacent window pattern descriptors are the same, they are grouped into the same continuous segment; if they are different, a new continuous segment is created starting with the next window pattern descriptor. After completing all adjacent comparisons, the number of continuous segments is counted, and this count is used as the ion concentration fluctuation characteristic of the micro-region where the sensor is located, corresponding to the original concentration time series.

[0035] See Figure 3 In the figure, the horizontal axis represents the window length of the multi-scale sliding window, in units of "windows," ranging from 20 to 500. The vertical axis represents the ion concentration fluctuation characteristics of the micro-region where each sensor is located under the corresponding window length. In the legend, the four curves correspond to the ion concentration fluctuation characteristics of the micro-regions where sensor 1, sensor 2, sensor 3, and sensor 4 are located, respectively, and the curve types are solid line, dashed line, dotted line, and dotted-dashed line.

[0036] As can be seen from the figure, as the window length increases, the ion concentration fluctuation characteristics of the micro-regions where the four sensors are located generally show an upward trend. This indicates that the larger the multi-scale sliding window segmentation window length, the more continuous segments are obtained in the pattern evolution sequence, showing an increase in the complexity of ion concentration fluctuation.

[0037] Specifically, the micro-region where sensor 3 is located has the highest fluctuation characteristic quantity, with the curve at the top overall. It tends to stabilize when the window length exceeds about 300, with a maximum value of about 22. The micro-region where sensor 1 is located is the second highest, with the fluctuation characteristic quantity curve following closely behind, and a maximum value of about 21. The micro-regions where sensors 2 and 4 are located have lower fluctuation characteristic quantities, and their curves are relatively close, with a maximum value of about 18 and 16 respectively. They tend to stabilize after the window length exceeds about 400.

[0038] When the window length is small, the fluctuation characteristic increases rapidly, especially between 20 and 150 window lengths, where the fluctuation characteristic of each sensor micro-region increases significantly, indicating that the short-timescale fluctuation component is relatively abundant within this range. As the window length continues to increase, the growth rate of the fluctuation characteristic slows down and tends to stabilize, reflecting that the evolution of ion concentration fluctuation patterns tends to stabilize within long-timescale windows.

[0039] In practical implementation, the state construction module obtains the ion concentration fluctuation characteristics of all micro-regions at each acquisition time from the fluctuation decomposition module. For any acquisition time, the state construction module concatenates the ion concentration fluctuation characteristics of all micro-regions at the same acquisition time according to the spatial arrangement order of the sensors, forming a concentration fluctuation distribution vector. The spatial arrangement order of the sensors is pre-set based on the deployment positions of multiple ion concentration sensors in the crop root region, and this fixed spatial arrangement order is used for all acquisition times. The dimension of the concentration fluctuation distribution vector is equal to the total number of ion concentration sensors. The element is the first element arranged in the spatial order of the sensors. The characteristic quantity of ion concentration fluctuation in the micro-region where each sensor is located at the corresponding acquisition time, among which It is a positive integer. The value ranges from 1 to the total number of ion concentration sensors.

[0040] The state construction module arranges various environmental parameters from the root environmental parameter sequence at the same acquisition time in a preset order, forming an environmental parameter vector. The various environmental parameters in the root environmental parameter sequence include soil temperature, soil moisture, soil pH, soil electrical conductivity, light intensity, air temperature, and air humidity. The preset order is soil temperature, soil moisture, soil pH, soil electrical conductivity, light intensity, air temperature, and air humidity. The dimension of the environmental parameter vector equals the total number of environmental parameter types. The element is the first element in a preset order. The values ​​of each environmental parameter type at the corresponding acquisition time, among which It is a positive integer. The value range is from 1 to the total number of environmental parameter types.

[0041] The concentration fluctuation distribution vector and the environmental parameter vector are concatenated end-to-end to obtain the instantaneous absorption state vector at the corresponding acquisition time. In the instantaneous absorption state vector, all elements of the concentration fluctuation distribution vector are arranged at the beginning, and all elements of the environmental parameter vector are arranged at the end. Represented as: in, This represents the instantaneous absorption state vector at the current moment; This represents the concentration fluctuation distribution vector at the current moment; This represents the environmental parameter vector at the current moment; the square brackets indicate that the concentration fluctuation distribution vector and the environmental parameter vector at the current moment are concatenated end to end. The dimension of the instantaneous absorption state vector is the sum of the dimensions of the concentration fluctuation distribution vector and the environmental parameter vector.

[0042] In practice, the pattern discrimination module obtains the instantaneous absorption state vectors from multiple consecutive historical moments preceding the current moment from the state construction module. The number of consecutive historical moments is a positive integer, preset according to the system sampling frequency and the dynamic change rate of crop root ion absorption, and is no less than 3.

[0043] The instantaneous absorption state vectors from multiple consecutive historical moments are arranged in chronological order to form a state vector sequence. Each element in the state vector sequence is the instantaneous absorption state vector for the corresponding historical moment.

[0044] For any two adjacent state vectors in the state vector sequence, perform a vector difference operation on the two adjacent state vectors to obtain a difference vector. Each element in the difference vector is the difference between the corresponding element of the latter state vector and the corresponding element of the former state vector. The dimension of the difference vector is the same as the dimension of the instantaneous absorption state vector.

[0045] For each element in the difference vector, calculate its sign to form a sign sequence. Signs include positive, negative, and zero. When an element in the difference vector is greater than zero, the corresponding element in the sign sequence is positive; when an element in the difference vector is less than zero, the corresponding element in the sign sequence is negative; when an element in the difference vector is equal to zero, the corresponding element in the sign sequence is zero.

[0046] Mark the adjacent positions in the symbol sequence where the symbols change. A change in symbols means that two adjacent symbols in the symbol sequence are no longer the same, including changes between positive and negative signs, positive and zero signs, and negative and zero signs. The marked position is the index position of two adjacent symbols in the symbol sequence.

[0047] Construct an interval length sequence based on the distances between marker positions. Each interval length in the interval length sequence is the index difference between two adjacent marker positions in the symbol sequence. The interval length sequence also includes the interval length from the start position of the symbol sequence to the first marker position and the interval length from the last marker position to the end position of the symbol sequence. Each interval length in the interval length sequence takes a positive integer value.

[0048] The mean of all interval lengths in the interval length sequence is used as a measure of the transition difference between two adjacent state vectors. The mean interval length is obtained by summing all interval lengths in the interval length sequence and dividing by the number of interval lengths.

[0049] The process of repeatedly performing vector difference operation, symbol sequence generation, symbol change labeling, interval length sequence construction, and interval length mean calculation on each pair of adjacent state vectors in the state vector sequence yields the transition difference sequence. Each element in the transition difference sequence corresponds to the transition difference measure between a pair of adjacent state vectors in the state vector sequence.

[0050] The cumulative transfer path length is obtained by summing up each transfer difference measure in the transfer difference sequence in chronological order. The cumulative transfer path length is the sum of all transfer difference measures in the transfer difference sequence.

[0051] Take the first and last state vectors in the state vector sequence, and perform a vector difference operation on them to obtain the first-to-last difference vector. Following the same calculation method as the transition difference metric between adjacent state vectors, calculate the positive / negative sign, sign sequence, sign change marker, interval length sequence, and mean interval length for the first-to-last difference vector to obtain the direct difference metric. The direct difference metric represents the direct difference between the first and last state vectors in the state vector sequence.

[0052] Dividing the cumulative transfer path length by the direct difference measure yields the absorption mode transition factor at the current time. The absorption mode transition factor is calculated using the following formula: in, This represents the absorption mode transition factor at the current moment. Indicates the cumulative transfer path length. This represents a direct measure of difference. Cumulative transfer path length. This is the sum of all transfer difference measures in the transfer difference sequence, arranged chronologically. Direct difference measure. It is a numerical measure of the direct difference between the first and last state vectors in a state vector sequence.

[0053] See Figure 4 In the figure, the horizontal axis represents historical moments in "times," and the vertical axis represents measurement values. Three curves are plotted: cumulative transfer path length, direct difference measure, and absorption mode transition factor. The curve colors and line types correspond to the blue solid line, orange dashed line, and green dotted line in the legend, respectively.

[0054] As shown in the figure, the cumulative transfer path length curve exhibits periodic fluctuations, with a value ranging from approximately 1.5 to 5.5. The alternating peaks and troughs indicate significant fluctuations in the transfer differences between state vectors at historical moments. The direct difference measurement curve shows relatively small overall fluctuations, remaining stable within the range of 1.5 to 2.5, suggesting relatively smooth changes in the direct difference between the first and last state vectors. The absorption mode transition factor curve is relatively flat, with a value generally maintained between 1.0 and 2.0. Its overall trend is synchronized with the cumulative transfer path length curve, but with smaller fluctuations.

[0055] In practical implementation, the distribution prediction module obtains the absorption mode transition factors corresponding to a preset number of historical moments prior to the current moment from the mode discrimination module. The preset number is a positive integer, pre-set based on the system sampling frequency and the dynamic change rate of crop root ion absorption, and is no less than 2. The preset number of historical moments are consecutive historical moments prior to the current moment. The absorption mode transition factors corresponding to the preset number of historical moments are arranged in chronological order to form a historical sequence of transition factors.

[0056] Calculate the differences between adjacent elements in the transition factor history sequence to obtain the transition factor change sequence. Each element in the transition factor change sequence is the difference between the transition factor of the next absorption mode and the transition factor of the previous absorption mode in the transition factor history sequence. The length of the transition factor change sequence is the length of the transition factor history sequence minus 1.

[0057] Arrange all values ​​in the transition factor change sequence in ascending order to obtain an ordered change sequence. The values ​​in the ordered change sequence are arranged from smallest to largest.

[0058] The first and second change values ​​located at preset quantiles are extracted from the ordered change sequence, where the first change value is less than the second change value. The preset quantiles include the first and second quantiles. The first quantile is set at 25%, and the second quantile at 75%. The first quantile is set at 25% because it represents the value at the one-quarter position in the ordered change sequence, and the second quantile is set at 75% because it represents the value at the three-quarters position. These 25% and 75% quantiles cover the middle 50% of the ordered change sequence, removing the influence of extreme changes at both ends on the fluctuation range. Using the position corresponding to the first element in the ordered change sequence as 0% and the position corresponding to the last element as 100%, the values ​​corresponding to the 25% and 75% quantiles are obtained through linear position mapping.

[0059] The lower boundary of the fluctuation range is determined by using the sum of the current absorption mode transition factor and the first change value, and the upper boundary is determined by using the sum of the current absorption mode transition factor and the second change value. The lower and upper boundaries of the fluctuation range together constitute the fluctuation range of the estimated transition factor value at the next time step.

[0060] The boundary value with the smallest difference between the absorption mode transition factor at the current moment and the boundary of the fluctuation range is selected as the predicted value of the transition factor at the next moment. The absolute value of the difference between the lower boundary of the fluctuation range and the absorption mode transition factor at the current moment is calculated to obtain the lower boundary difference value; the absolute value of the difference between the upper boundary of the fluctuation range and the absorption mode transition factor at the current moment is calculated to obtain the upper boundary difference value. The lower boundary difference value and the upper boundary difference value are compared. If the lower boundary difference value is less than the upper boundary difference value, the lower boundary of the fluctuation range is selected as the predicted value of the transition factor at the next moment; if the lower boundary difference value is greater than the upper boundary difference value, the upper boundary of the fluctuation range is selected as the predicted value of the transition factor at the next moment; if the lower boundary difference value is equal to the upper boundary difference value, the lower boundary of the fluctuation range is selected as the predicted value of the transition factor at the next moment.

[0061] Multiplying the predicted transition factor value at the next time step by the instantaneous absorption state vector at the current time step yields the shift in crop rhizosphere ion concentration distribution at the next time step. The shift in crop rhizosphere ion concentration distribution at the next time step is calculated using the following formula: in, The offset of the crop rhizosphere ion concentration distribution at the next moment is represented by a vector. The predicted value of the transition factor at the next moment is a scalar. The instantaneous absorption state vector at the current moment is a vector; the multiplication sign indicates multiplying the scalar. with vector Multiply each element in the vector. The dimension of the crop rhizosphere ion concentration distribution offset at the next time step is the same as the dimension of the instantaneous absorption state vector at the current time step.

[0062] In practice, the execution control module obtains the crop rhizosphere ion concentration distribution offset for the next time step from the distribution prediction module. The ion concentration distribution offset is a vector, with the same dimension as the instantaneous absorption state vector at the current moment. The first element of the ion concentration distribution offset, matching the total number of ion concentration sensors, corresponds to the position of each sensor, while the latter part corresponds to the environmental parameter offset components.

[0063] The execution control module extracts the offset component values ​​corresponding to the positions of each sensor from the ion concentration distribution offset. The first [element / component] in the ion concentration distribution offset... The element is the first The offset component values ​​of each sensor corresponding to a micro-region It is a positive integer. The value ranges from 1 to the total number of ion concentration sensors.

[0064] For each extracted offset component value, a sign determination is performed. When the offset component value is greater than zero, the micro-region corresponding to the offset component value is marked as the positive direction; when the offset component value is less than zero, the micro-region corresponding to the offset component value is marked as the negative direction; when the offset component value is equal to zero, the micro-region corresponding to the offset component value is marked as the zero direction.

[0065] The spatial coordinates of each sensor's micro-region are associated with the corresponding positive, negative, or zero direction markers to obtain the desired concentration change direction for each micro-region. The spatial coordinates of the micro-regions are pre-stored based on the deployment location of the ion concentration sensors in the crop root region.

[0066] For each micro-region, the direction of valve opening adjustment in the corresponding branch of the delivery pipeline is determined based on the sign of the expected concentration change direction in the micro-region. When the expected concentration change direction of the micro-region is positive, the valve opening of the corresponding branch is increased; when the expected concentration change direction of the micro-region is negative, the valve opening of the corresponding branch is decreased; when the expected concentration change direction of the micro-region is zero, the valve opening of the corresponding branch remains unchanged.

[0067] The sum of the absolute values ​​of the expected concentration change directions in all micro-regions is used as the total output adjustment amplitude of the ion generator. The sum of the absolute values ​​of the expected concentration change directions in all micro-regions is obtained by adding the absolute values ​​of the offset components of each micro-region.

[0068] The valve opening adjustment amount for each branch is allocated based on the proportion of the absolute value of the desired concentration change direction in each micro-region to the total output adjustment amplitude of the ion generator. The valve opening adjustment amount is calculated using the following formula: in, Indicates the first The valve opening adjustment amount of each branch; This indicates the total output adjustment range of the ion generator; Indicates the first The offset component values ​​of each micro-region; Indicates the first The absolute value of the expected concentration change direction in each micro-region; For the summation index, The value range is 1 to ; Indicates the total number of micro-regions; This represents the sum of the absolute values ​​indicating the direction of desired concentration changes in all micro-regions. The total output adjustment range of the ion generator. equal .

[0069] The total output adjustment range of the ion generator and the valve opening adjustment range of each branch together constitute the regulation and control quantities of the ion generation and delivery system. The total output adjustment range of the ion generator is used to adjust the total output ion volume of the ion generator, and the valve opening adjustment range of each branch is used to adjust the opening of the corresponding delivery pipeline valve. The regulation and control quantities act on the ion generator and the delivery pipeline valves.

[0070] In its implementation, the system also includes a nutrient enhancement decision module. This module obtains the ion concentration fluctuation characteristics of each micro-region from the fluctuation decomposition module, acquires the root environment parameter sequence from the data acquisition module, and obtains the instantaneous absorption state vector at the current moment. The module has a built-in crop nutrient deficiency judgment rule library. This library stores the normal concentration range thresholds and critical deficiency values ​​for different nutrients at various growth stages for different crop varieties. For each nutrient element, the library contains the lower limit of its normal concentration and the critical deficiency concentration in the root environment. Nutrient elements include macroelements (nitrogen, phosphorus, potassium) and microelements (iron, manganese, zinc, copper, boron, molybdenum, etc.). The nutrient enhancement decision module performs nutrient deficiency judgment for each micro-region: it extracts the concentration components corresponding to each nutrient element from the ion concentration fluctuation characteristics of that micro-region and compares these components with the critical deficiency concentration of the corresponding nutrient element in the nutrient deficiency judgment rule library. When the concentration component of a certain nutrient element is lower than its critical deficiency concentration, the micro-region is determined to be deficient in that nutrient element.

[0071] When a nutrient deficiency is determined in a specific micro-region, the nutrient supplementation decision module further identifies the type and severity of the deficiency. The deficiency type refers to the nutrient element whose concentration is below the critical deficiency concentration. The severity of the deficiency is determined by the difference between the concentration and the critical deficiency concentration; the larger the difference, the more severe the deficiency. Deficiency severity is categorized into three levels: mild, moderate, and severe. Mild deficiency corresponds to a concentration below the critical deficiency concentration but above the lower limit of normal concentration; moderate deficiency corresponds to a concentration below the lower limit of normal concentration but above 50% of the lower limit of normal concentration; and severe deficiency corresponds to a concentration below 50% of the lower limit of normal concentration.

[0072] The nutrient enhancement decision module generates a nutrient enhancement strategy based on the type and severity of nutrient deficiency. The strategy includes the types of nutrient ions to be supplemented and their corresponding amounts. The types of nutrient ions to be supplemented correspond to the type of deficiency. The supplementation amount is determined based on the severity of the deficiency: for mild deficiency, the supplementation amount is 1.2 times the normal supply; for moderate deficiency, it is 1.5 times the normal supply; and for severe deficiency, it is 2.0 times the normal supply. The normal supply amount is the baseline supply amount of this nutrient element for the micro-region under current environmental conditions. This baseline supply amount is determined by the nutrient enhancement decision module using a pre-defined supply lookup table based on soil temperature, soil moisture, and soil pH from the root environmental parameter sequence.

[0073] The nutrient enhancement decision module sends the generated nutrient enhancement strategy to the execution control module. Based on the strategy, the execution control module controls the ion generator to start generating the corresponding nutrient ions and opens the branch valves in the delivery pipeline corresponding to the nutrient-deficient micro-region, delivering the generated nutrient ions to the corresponding micro-region. When multiple micro-regions simultaneously exhibit nutrient deficiency, the execution control module executes the nutrient enhancement operation sequentially according to the priority of the deficiency severity, with higher priority for micro-regions with more severe deficiencies. After executing the nutrient enhancement operation, the nutrient enhancement decision module continuously monitors the changes in the ion concentration fluctuation characteristics of the corresponding micro-region. When it determines that the concentration of the corresponding nutrient element in the micro-region has returned to the normal range, the nutrient enhancement decision module sends a stop nutrient enhancement command to the execution control module, which then stops the generation of the nutrient ions and closes the corresponding branch valve. Through the above-mentioned nutrient enhancement decision module settings, the system can proactively identify the type and severity of nutrient deficiency when crop roots show signs of nutrient deficiency, generate targeted nutrient enhancement strategies, and execute them precisely. This achieves a functional upgrade from passively adjusting ion concentration to actively supplementing crop nutrients, improving the intelligence level of crop nutrient supply.

[0074] See Figure 5In the graph, the vertical axis represents the valve opening adjustment amount, ranging from 0 to approximately 3.0, and the horizontal axis represents the total number of micro-regions, approximately 500 in total. The legend indicates three desired directions of concentration change: positive (red dot), negative (blue triangle), and zero (green square).

[0075] As can be seen from the figure, the valve opening adjustment amounts in the positive and negative directions are relatively evenly distributed, and the values ​​mainly fluctuate between 0 and 2.0, with a few points exceeding 2.0 and the highest approaching 2.8. This reflects that when the execution control module adjusts the valve opening of the delivery pipeline according to the ion concentration distribution offset for different micro-regions, the adjustment amount has a large dynamic change. The valve opening adjustment amount in the zero direction is always zero, which meets the control requirement of keeping the valve opening of the micro-region in the zero direction unchanged in the above embodiment.

[0076] The positive and negative directions of the valve opening adjustment are interspersed throughout the entire micro-region, with no obvious clustering areas. This indicates that the system's execution control module can make fine adjustments to the desired concentration change direction in each micro-region, ensuring the dynamic balance and precise delivery of ion concentration in the crop roots.

[0077] Overall, the execution control module determines the sign of the offset component values ​​of corresponding micro-regions in the ion concentration distribution offset and combines this with the spatial coordinates of the micro-regions to adjust the valve opening for different micro-regions. The valve opening adjustment is allocated according to the proportion of the absolute value of the micro-region offset component to the total output adjustment range, ensuring that the total output of the ion generator is linked to the valve opening of each branch, thereby achieving intelligent control of the ion generation and delivery system.

[0078] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application.

Claims

1. An IoT-based crop ion generation and delivery system based on AI data detection, characterized in that, include: The data acquisition module acquires in-situ detection sequences from multiple ion concentration sensors in the crop root region, as well as root environmental parameter sequences at the corresponding acquisition times. The fluctuation decomposition module decomposes the fluctuation components of each in-situ detection sequence and extracts the ion concentration fluctuation characteristics of the micro-region where each sensor is located. The state construction module constructs an instantaneous absorption state vector describing the dynamics of ion absorption in crop roots, based on the ion concentration fluctuation characteristics of all microregions and the root environment parameter sequence. The mode discrimination module obtains the absorption mode transition factor at the current moment by measuring the geometric shape of the evolution trajectory of the state vector in the vector space based on the instantaneous absorption state vectors of multiple historical moments before the current moment. The distribution prediction module uses the current absorption mode transition factor and the current instantaneous absorption state vector to predict the crop rhizosphere ion concentration distribution shift at the next moment. The execution control module generates adjustment control quantities for the ion generation and delivery system based on the ion concentration distribution offset, and applies these quantities to the ion generator and delivery pipeline valves.

2. The IoT crop ion generation and delivery system based on AI data detection according to claim 1, characterized in that, The step of performing fluctuation component decomposition on each in-situ detection sequence and extracting the ion concentration fluctuation characteristics of the micro-region where each sensor is located includes: Each in-situ detection sequence is arranged in chronological order to form the original concentration time series of the corresponding sensor; The original concentration time series is segmented into multi-scale sliding windows, and the relative arrangement pattern of the sequence values ​​is calculated within each window to obtain the window pattern descriptor; All window pattern descriptors are sorted according to the window center time to form a pattern evolution sequence; Adjacent windows with the same descriptor in the pattern evolution sequence are merged, and the number of consecutive segments after merging is used as the characteristic quantity of ion concentration fluctuation in the micro-region where the sensor is located.

3. The IoT crop ion generation and delivery system based on AI data detection according to claim 2, characterized in that, The step of constructing an instantaneous absorption state vector describing the dynamics of ion absorption in crop roots based on the ion concentration fluctuation characteristics of all microregions and the root environment parameter sequence includes: The ion concentration fluctuation characteristics of all micro-regions at the same acquisition time are spliced ​​together according to the spatial arrangement of the sensor to form the concentration fluctuation distribution vector at that time. The environmental parameters in the root environmental parameter sequence at the same acquisition time are arranged in a preset order to form the environmental parameter vector at that time. By concatenating the concentration fluctuation distribution vector and the environmental parameter vector end to end, the instantaneous absorption state vector at that moment is obtained.

4. The IoT crop ion generation and delivery system based on AI data detection according to claim 3, characterized in that, The step of obtaining the absorption mode transition factor at the current moment by measuring the geometric shape of the evolution trajectory of the state vector in the vector space based on the instantaneous absorption state vectors of multiple historical moments prior to the current moment includes: Arrange the instantaneous absorption state vectors of multiple consecutive historical moments before the current moment in chronological order to form a state vector sequence; Calculate the transition difference measure between two adjacent state vectors in the state vector sequence to obtain the transition difference sequence; The cumulative transfer path length is obtained by sequentially summing each transfer difference metric in the transfer difference sequence in chronological order. The absorption mode transition factor at the current moment is obtained by dividing the cumulative transition path length by the direct difference measure between the first and last state vectors in the state vector sequence.

5. The IoT crop ion generation and delivery system based on AI data detection according to claim 4, characterized in that, The step of predicting the shift in crop rhizosphere ion concentration distribution at the next moment using the absorption mode transition factor and the instantaneous absorption state vector at the current moment includes: Obtain the absorption mode transition factors corresponding to a preset number of historical moments before the current moment, and form a historical sequence of transition factors; Calculate the difference between adjacent elements in the historical sequence of the transition factor to obtain the sequence of changes in the transition factor. Based on the numerical distribution range of the transition factor change sequence, determine the fluctuation range boundary of the estimated transition factor value at the next moment. The boundary value with the smallest difference between the absorption mode transition factor and the current moment is selected as the predicted value of the transition factor for the next moment within the boundary of the fluctuation range. Multiply the predicted value of the transition factor at the next moment by the instantaneous absorption state vector at the current moment to obtain the offset of the crop rhizosphere ion concentration distribution at the next moment.

6. The IoT crop ion generation and delivery system based on AI data detection according to claim 5, characterized in that, The step of generating a control quantity for the ion generation and delivery system based on the ion concentration distribution offset, and applying it to the ion generator and the valves in the delivery pipeline, includes: The ion concentration distribution offset is decomposed according to the spatial position corresponding to each sensor to obtain the expected concentration change direction of each micro-region; For each micro-region, the opening adjustment direction of the corresponding branch in the delivery pipeline valve is determined based on the positive or negative sign of the expected concentration change direction in that micro-region. The sum of the absolute values ​​of the expected concentration change directions in all micro-regions is used as the total output adjustment range of the ion generator. The valve opening adjustment amount of each branch is allocated according to the proportion of the absolute value of the expected concentration change direction of each micro-region to the total output adjustment range of the ion generator.

7. The IoT crop ion generation and delivery system based on AI data detection according to claim 2, characterized in that, The step of calculating the relative arrangement pattern of sequence values ​​within each window to obtain a window pattern descriptor includes: Sort all concentration values ​​in the window in ascending order to obtain the ranking of each value within the window; Arrange the sorting order of each value in the window according to the original time order of the values ​​in the window to form a sorting vector; The positional arrangement vector is converted into a binary encoded string of a preset length, which is used as the window mode descriptor of the window.

8. The IoT crop ion generation and delivery system based on AI data detection according to claim 4, characterized in that, The step of calculating the transition difference metric between two adjacent state vectors in the state vector sequence to obtain the transition difference sequence includes: Perform a vector difference operation on two adjacent state vectors in the state vector sequence to obtain the difference vector; Calculate the sign of each element in the difference vector to form a sign sequence; Mark the adjacent positions in the symbol sequence where the symbol changes, and construct an interval length sequence based on the interval length between the marked positions; The mean of all interval lengths in the interval length sequence is used as a measure of the transition difference between two adjacent state vectors.

9. The IoT crop ion generation and delivery system based on AI data detection according to claim 5, characterized in that, The step of determining the fluctuation range boundary of the estimated value of the transition factor at the next moment based on the numerical distribution range of the transition factor change sequence includes: Arrange all the values ​​in the transition factor change sequence in order of magnitude to obtain an ordered change sequence; Extract the first change value and the second change value located at a preset quantile from the ordered change value sequence, wherein the first change value is less than the second change value; The sum of the absorption mode transition factor at the current moment and the first change value is used as the lower boundary of the fluctuation range. The sum of the absorption mode transition factor at the current moment and the second change value is used as the upper boundary of the fluctuation range.

10. The IoT crop ion generation and delivery system based on AI data detection according to claim 6, characterized in that, The step of decomposing the ion concentration distribution offset according to the spatial position corresponding to each sensor to obtain the desired concentration change direction of each micro-region includes: Extract the offset component values ​​corresponding to the positions of each sensor from the ion concentration distribution offset; For each offset component value, determine the sign: if the offset component value is greater than zero, mark it as a positive direction; if it is less than zero, mark it as a negative direction. By associating the spatial coordinates of each sensor's micro-region with the corresponding positive or negative direction marker, the desired concentration change direction of that micro-region can be obtained.

Images