Machine learning-based rhizosphere soil health assessment method and system

Abstract

The invention relates to the technical field of data processing, and discloses a rhizosphere soil health assessment method and system based on machine learning. The method comprises the following steps: acquiring multi-dimensional parameter data of a rhizosphere region, processing an abnormal value and calculating a comprehensive index by applying a rhizosphere time sequence median filtering method, establishing an evaluation model by combining a hybrid algorithm with an attention mechanism and fusing static and time sequence characteristics, outputting a health score and judging a state category, and finally, an optimization scheme is generated through a rhizosphere factor weight self-learning algorithm, and complete closed-loop management of dynamic evaluation and intelligent adjustment of rhizosphere soil health is realized. According to the invention, the problem that the rhizosphere microenvironment cannot be dynamically and accurately evaluated in the prior art is solved, and the accuracy and intelligent adjustment capability of rhizosphere soil health evaluation are improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of data processing, in particular to a rhizosphere soil health evaluation method and system based on machine learning. BACKGROUND

[0002] In the prior art, soil health evaluation mainly relies on traditional laboratory analysis methods, which evaluate soil health conditions by collecting soil samples and measuring chemical, physical and biological properties. With the rapid development of Internet of Things, sensor technology and machine learning, some researches have begun to apply machine learning technology to soil pollution evaluation, especially in the prediction of the content of potentially toxic elements (PTE), spatial distribution and source identification. The combination of hyperspectral data and machine learning methods can predict the content of PTE in large-scale areas at low cost, and machine learning algorithms integrated with environmental covariates also provide better performance than traditional statistical methods in spatial prediction.

[0003] However, the prior art has obvious deficiencies: traditional laboratory analysis methods have high time cost, poor spatial representativeness, lack of dynamic monitoring, etc.; existing machine learning methods mainly focus on the static content prediction and large-scale spatial distribution of PTE, and cannot accurately evaluate the dynamic changes of the special microenvironment of the plant rhizosphere; existing algorithms mostly use single machine learning models, lack of effective fusion processing capability for multi-source heterogeneous data of rhizosphere soil, resulting in limited evaluation accuracy.

[0004] Based on the above analysis, the prior art lacks a soil health evaluation method specifically for the characteristics of the rhizosphere microenvironment. Since the rhizosphere region is the most active microenvironment for plant-soil interaction, its soil parameter variation law is significantly different from that of general soil, so it is necessary to develop a hybrid algorithm that can process both static characteristics and dynamic time series information of rhizosphere soil. Further, the prior art cannot automatically generate a targeted adjustment scheme based on the rhizosphere soil health evaluation results, lacking the closed-loop management capability from evaluation to adjustment, which limits the intelligent level and practicality of soil health management. SUMMARY

[0005] The present application provides a rhizosphere soil health evaluation method and system based on machine learning, which solves the problem that the prior art cannot dynamically and accurately evaluate the rhizosphere microenvironment, and improves the accuracy and intelligent adjustment capability of rhizosphere soil health evaluation.

[0006] In a first aspect, the present application provides a rhizosphere soil health evaluation method based on machine learning, which comprises: S1, collecting the pH value, conductivity, temperature and humidity, nutrient concentration and microbial activity data of the target plant rhizosphere region as the rhizosphere soil original data set; S2, the step of correcting the outliers in the rhizosphere soil original data set by the rhizosphere time series median filtering method, calculating the rhizosphere nutrient comprehensive index and the rhizosphere microbial activity index, and obtaining the rhizosphere soil standardized data; S3, the step of training the rhizosphere soil standardized data by a hybrid algorithm, fusing static features and time series information by an attention mechanism, and establishing a rhizosphere soil health evaluation algorithm; S4, the step of outputting a rhizosphere soil health score by the rhizosphere soil health evaluation algorithm, and judging the rhizosphere soil health state category according to the health score change rate; S5, the step of calculating the regulator dosage parameter by a rhizosphere factor weight self-learning algorithm based on the rhizosphere soil health state category, and outputting a rhizosphere soil optimization scheme.

[0007] In a second aspect, the application provides a rhizosphere soil health evaluation system based on machine learning, comprising: A collection module for collecting pH value, conductivity, temperature and humidity, nutrient concentration and microbial activity data in the rhizosphere region of a target plant as a rhizosphere soil original data set; A correction module for correcting outliers in the rhizosphere soil original data set by the rhizosphere time series median filtering method, calculating the rhizosphere nutrient comprehensive index and the rhizosphere microbial activity index, and obtaining the rhizosphere soil standardized data; A training module for training the rhizosphere soil standardized data by a hybrid algorithm, fusing static features and time series information by an attention mechanism, and establishing a rhizosphere soil health evaluation algorithm; An output module for outputting a rhizosphere soil health score by the rhizosphere soil health evaluation algorithm, and judging the rhizosphere soil health state category according to the health score change rate; A calculation module for calculating the regulator dosage parameter by a rhizosphere factor weight self-learning algorithm based on the rhizosphere soil health state category, and outputting a rhizosphere soil optimization scheme.

[0008] In a third aspect, a rhizosphere soil health evaluation device based on machine learning is provided, comprising a memory and at least one processor, the memory having instructions stored therein; the at least one processor invokes the instructions in the memory to enable the rhizosphere soil health evaluation device based on machine learning to perform the rhizosphere soil health evaluation method based on machine learning described above.

[0009] In a fourth aspect, a computer readable storage medium is provided, the computer readable storage medium having instructions stored therein, when executed on a computer, enabling the computer to perform the rhizosphere soil health evaluation method based on machine learning described above.

[0010] In the technical scheme provided in the application, the pH value, conductivity, temperature and humidity, nutrient concentration and microbial activity data of the rhizosphere region of the target plant are collected as the rhizosphere soil original data set. Compared with the traditional soil evaluation method which only focuses on general soil parameters, the application specially collects multi-dimensional data according to the particularity of the rhizosphere microenvironment, which can more accurately reflect the real situation of the interaction between the plant root system and the soil. The rhizosphere time series median filtering method is used to correct abnormal values and calculate the rhizosphere nutrient comprehensive index and the rhizosphere microbial activity index, effectively solving the problem of noise interference of rhizosphere environment data, and a composite evaluation index system specially applicable to rhizosphere soil is constructed, which has higher accuracy and reliability compared with the single parameter evaluation method of the prior art. The rhizosphere soil standardization data is trained by a hybrid algorithm, and the attention mechanism is used to fuse static features and time series information to establish a rhizosphere soil health evaluation algorithm, which overcomes the limitation of the processing capacity of a single model in the prior art, can capture the complex nonlinear relationship between rhizosphere soil parameters, and can identify the time series change pattern, significantly improving the accuracy and stability of the prediction of the rhizosphere soil health state.

[0011] The static feature relationship is processed by the random forest module, and the dynamic change pattern is captured by the time series convolutional neural network module, and the combination of the two can make full use of the space-time features of the rhizosphere environment data, and has stronger generalization ability and adaptability compared with the traditional machine learning method. According to the health score change rate, the rhizosphere soil health state category is determined, which realizes the change from static evaluation to dynamic early warning, and can timely find the early signs of rhizosphere soil health problems. Based on the rhizosphere soil health state category, the rhizosphere factor weight self-learning algorithm is used to calculate the regulator dosage parameter and output the rhizosphere soil optimization scheme, realizing the complete closed-loop management from evaluation to regulation, and compared with the prior art which only provides evaluation results, the application can automatically generate targeted adjustment suggestions, greatly improving the intelligent level of rhizosphere soil management. BRIEF DESCRIPTION OF DRAWINGS

[0012] In order to more clearly illustrate the technical scheme of the embodiments of the application, the drawings needed in the embodiment description will be briefly introduced. Obviously, the drawings in the following description are some embodiments of the application, and other drawings can be obtained by those skilled in the art without creative labor.

[0013] Figure 1 An embodiment schematic diagram of the rhizosphere soil health evaluation method based on machine learning in the embodiments of the application; Figure 2 An embodiment schematic diagram of the rhizosphere soil health evaluation method based on machine learning in the embodiments of the application; Figure 3 A training and verification loss reduction curve of the training effect of the rhizosphere soil health evaluation algorithm based on machine learning in the embodiments of the application; Figure 4 A prediction result correlation comparison chart for the training effect of the rhizosphere soil health assessment algorithm based on machine learning in the embodiment of the application; Figure 5 A back propagation gradient mean display chart for the training effect of the rhizosphere soil health assessment algorithm based on machine learning in the embodiment of the application; Figure 6 An embodiment schematic diagram of the rhizosphere soil health assessment system based on machine learning in the embodiment of the application; Figure 7 A structural schematic block diagram of the rhizosphere soil health assessment device based on machine learning in the embodiment of the application. DETAILED DESCRIPTION

[0014] The embodiment of the application provides a rhizosphere soil health assessment method and system based on machine learning. The terms "first", "second", "third", "fourth" and the like (if any) in the specification and claims of the application and the above-mentioned drawings are used to distinguish similar objects, and do not have to be used to describe a specific order or sequence. It should be understood that the data thus used can be interchanged under appropriate circumstances, so that the embodiments described herein can be implemented in an order other than that illustrated or described herein. In addition, the term "comprising" or "having" and any variations thereof are intended to cover non-exclusive inclusion, for example, a process, method, system, product or device including a series of steps or units does not have to be limited to those steps or units clearly listed, but can include other steps or units not clearly listed or inherent to these processes, methods, products or devices.

[0015] For the sake of understanding, the specific process of the embodiment of the application is described below. Please refer to Figure 1 The embodiment of the application based on machine learning for rhizosphere soil health assessment method includes: S1, collecting the pH value, conductivity, temperature and humidity, nutrient concentration and microbial activity data of the target plant rhizosphere region as the rhizosphere soil original data set; S2, correcting the abnormal values in the rhizosphere soil original data set by the rhizosphere time series median filtering method, calculating the rhizosphere nutrient comprehensive index and the rhizosphere microbial activity index, and obtaining the rhizosphere soil standardized data; S3, training the rhizosphere soil standardized data by a hybrid algorithm, fusing static features and time series information through an attention mechanism, and establishing a rhizosphere soil health assessment algorithm; S4, outputting the rhizosphere soil health score through the rhizosphere soil health assessment algorithm, and judging the rhizosphere soil health state category according to the health score change rate; S5, based on the rhizosphere soil health status category, the regulator dosage parameter is calculated by the rhizosphere factor weight self-learning algorithm, and the rhizosphere soil optimization scheme is output.

[0016] It can be understood that the execution subject of the present application can be a rhizosphere soil health evaluation system based on machine learning, and can also be a terminal or a server, which is not limited here. The server is taken as an example for illustration in the embodiments of the present application.

[0017] Specifically, by collecting multi-dimensional data of the rhizosphere region of the target plant, including pH value, conductivity, temperature and humidity, nutrient concentration and microbial activity, and other key parameters, these data constitute the original data set of the rhizosphere soil. In the data collection process, the sensor network is used to monitor the rhizosphere microenvironment in real time, ensuring high-frequency data collection and accurate recording. The rhizosphere time series median filtering method is used to correct the outliers in the original data, ensuring the accuracy and reliability of the data. For the corrected data, the rhizosphere nutrient comprehensive index and microbial activity index are further calculated, which can more comprehensively reflect the health status of the soil, so as to obtain the standardized rhizosphere soil data. On the basis of the standardized data, the mixed algorithm is used to train the data, and the attention mechanism is combined to fuse the static features and time series information. The mixed algorithm can effectively handle the complex relationship between static features and time series features, and use the deep learning model to capture the nonlinear patterns, establishing a rhizosphere soil health evaluation algorithm. Through the algorithm, the health score of the rhizosphere soil can be output according to the health score of the rhizosphere soil, and the health status category of the rhizosphere soil can be determined by analyzing the health score change rate. According to the score, the system can classify the soil health status into healthy, medium healthy or unhealthy, providing a basis for subsequent soil regulation. Based on the health status category, the system further analyzes the key influencing factors of the soil through the rhizosphere factor weight self-learning algorithm, and automatically calculates the regulator dosage parameter. The algorithm can output the corresponding regulator dosage scheme according to the changes of the factors such as pH value deviation, nutrient concentration and microbial activity of the soil, in order to achieve the purpose of optimizing the health of the rhizosphere soil. This process realizes the closed-loop management from evaluation to regulation, which not only can monitor the health of the rhizosphere soil in real time, but also can provide targeted and accurate soil optimization scheme, greatly improving the intelligence and accuracy of soil management.

[0018] In a specific embodiment, S1 further comprises: deploying a rhizosphere sensor network in the rhizosphere region of the target plant to monitor the rhizosphere microenvironment with multiple parameters, and obtaining rhizosphere environment monitoring data; The rhizosphere environment monitoring data is collected according to a preset time interval, the pH value data is collected every 30 minutes, the conductivity data is collected every 30 minutes, the temperature and humidity data is collected every 60 minutes, the nutrient concentration data is collected every 120 minutes, and the microbial activity data is collected every 120 minutes, to obtain rhizosphere soil time sequence collection data; The rhizosphere soil time sequence collection data is transmitted through a wireless communication module, a time stamp, a sensor identifier and location information are added to each data record, to obtain a rhizosphere soil original data set; Based on the rhizosphere soil original data set, data integrity verification is performed, and missing data is marked for processing, to obtain a rhizosphere soil original data set.

[0019] Specifically, a rhizosphere sensor network is deployed in the target plant rhizosphere area to monitor multiple key parameters in the rhizosphere microenvironment in real time, including pH value, conductivity, temperature and humidity, nutrient concentration and microbial activity. The sensor network synchronously collects these data according to a preset time interval to ensure high-frequency and comprehensive data collection. The pH value and conductivity data are collected every 30 minutes, the temperature and humidity data are collected every 60 minutes, the nutrient concentration data are collected every 120 minutes, and the microbial activity data are also collected every 120 minutes, to form rhizosphere soil time sequence collection data. To ensure smooth and stable data transmission, all collected data are transmitted in real time through a wireless communication module, and a time stamp, a sensor identifier and location information are automatically added to each data record. These information can ensure the integrity and traceability of the data, and generate a rhizosphere soil original data set. After the rhizosphere soil original data set is generated, the system verifies the integrity of the data to ensure that there is no missing data or abnormal records. For missing or incomplete data, the system will mark it for special processing in subsequent data analysis, to obtain a complete and reliable rhizosphere soil original data set. In this way, the entire data collection process ensures the accuracy, timeliness and high integrity of the information, providing a solid data foundation for subsequent rhizosphere soil health assessment.

[0020] For example, during the deployment of the rhizosphere sensor network, sensors are precisely placed within the target plant rhizosphere area to monitor multiple key parameters of the rhizosphere soil in real time. For example, a pH sensor will collect the soil's acid-base value changes in real time, recording data every 30 minutes to ensure adequate monitoring of dynamic changes in the soil environment. A conductivity sensor will also collect data at the same time interval to reflect the dissolved salt content of the soil, helping to analyze the soil's fertility. Temperature and humidity sensors collect data every 60 minutes to capture changes in soil and air temperature and humidity, helping to study the climate adaptability of the rhizosphere environment. Nutrient concentration and microbial activity data are collected every 120 minutes, respectively, to ensure that soil nutrient and microbial activity levels are accurately recorded over a long period of time. These data are transmitted through a wireless communication module and are time-stamped, sensor-identified, and location-informed to ensure that each record can be accurately mapped to a specific time and location, avoiding data confusion or loss.

[0021] For example, during the data integrity verification process, the system checks the collected rhizosphere soil raw data set to ensure that each data item is valid. If some data is missing in some period, the system will automatically mark the missing data and identify it as "missing" or "abnormal", so that appropriate compensation or correction measures can be taken in subsequent data processing. For example, if the pH value data at a certain time fails to be successfully collected, the system will record it through the marking system, and subsequent interpolation methods or other means can be used to fill in the data gap, ensuring the coherence and integrity of the data. In this way, through strict verification and processing of data integrity, the obtained data set can truly reflect the health status of the rhizosphere soil, providing accurate basic data for subsequent evaluation and optimization of the scheme.

[0022] In a specific embodiment, the S2 step further comprises: Range testing is performed on the pH, conductivity, temperature and humidity, nutrient concentration, and microbial activity data in the rhizosphere soil raw data set, and values outside the preset range of the rhizosphere environment are marked as abnormal values to obtain abnormal value marked data; Based on the abnormal value marked data, rhizosphere time series median filtering processing is performed, and the abnormal values are replaced and corrected using the median of the adjacent time point data to obtain rhizosphere soil corrected data; The nitrogen, phosphorus, and potassium concentrations and organic acid concentrations in the rhizosphere soil corrected data are calculated and processed according to the weighted summation formula to obtain a rhizosphere nutrient comprehensive index; The urease activity, phosphatase activity, and invertase activity in the rhizosphere soil corrected data are weighted and calculated according to the enzyme activity weight coefficient to obtain a rhizosphere microbial activity index and rhizosphere soil standardized data.

[0023] Specifically, the pH value, conductivity, temperature and humidity, nutrient concentration and microbial activity data in the rhizosphere soil original data set are subjected to range test, and the values exceeding the preset range of the rhizosphere environment are marked as abnormal values, so as to obtain abnormal value marked data. In the marking process of the abnormal values, the system compares each parameter through the set threshold value, and if a certain data item exceeds the normal range, it will be marked as abnormal, ensuring the data quality and effectiveness of subsequent processing. Based on these abnormal value marked data, rhizosphere time series median filtering processing is performed, and the median value of adjacent time points is used to replace the abnormal value, so as to correct the abnormal items in the data. The calculation formula of the median filtering is: wherein, is the corrected data at the current time, is the original data of the previous and next k time points. This processing method can effectively reduce the influence of noise and retain the overall trend of the data, making the data more smooth and stable, and obtaining the corrected rhizosphere soil data. The nitrogen, phosphorus and potassium concentrations and organic acid concentrations in the rhizosphere soil corrected data are calculated and processed according to the weighted summation formula to obtain the rhizosphere nutrient comprehensive index. Assuming that the weights of each nutrient concentration are , the calculation formula of the rhizosphere nutrient comprehensive index is: wherein, are the nitrogen, phosphorus and potassium concentrations respectively. The urease activity, phosphatase activity and invertase activity in the rhizosphere soil corrected data are also weighted calculated according to the preset enzyme activity weight coefficient to obtain the rhizosphere microbial activity index. The weighted calculation formula of the enzyme activity index is: wherein, are the urease activity, phosphatase activity and invertase activity respectively, is the weight coefficient of each enzyme activity. Through these processing steps, the rhizosphere soil standardized data is obtained, which provides reliable and accurate data support for subsequent health assessment.

[0024] Taking the rhizosphere soil health assessment of a certain farm as an example, a sensor network is deployed in the rhizosphere area of different crops to collect pH value, conductivity, temperature and humidity, nutrient concentration and microbial activity data in real time. Every 30 minutes, the sensor records pH value and conductivity data, every 60 minutes records temperature and humidity data, and every 120 minutes records nutrient concentration and microbial activity data. The collected data is transmitted to the server through the wireless communication module, and the data contains timestamp, sensor identifier and location information. In the data processing stage, it is found that the pH value measurement is abnormal, which exceeds the preset soil pH range (such as 3.0 to 9.0). The abnormal value is marked and corrected by time median filtering method, and the median value of the data at the adjacent time point is used to replace the abnormal value to obtain the corrected data. Then, the system calculates the weighted sum of the concentrations of nitrogen, phosphorus and potassium, assuming that the weights are 0.3, 0.3 and 0.4 respectively, to obtain the rhizosphere nutrient comprehensive index. For microbial activity, the system multiplies the activity data of urease, phosphatase and sucrase by their respective weight coefficients (such as 0.4, 0.3 and 0.3) to obtain the rhizosphere microbial activity index. These processed data are used to generate the rhizosphere soil health assessment report and provide the basis for subsequent soil regulation scheme.

[0025] In a specific embodiment, the S3 step further comprises: inputting the rhizosphere soil standardized data into a random forest module for nonlinear feature relationship learning, modeling and processing the complex correlation between rhizosphere soil parameters to obtain rhizosphere static feature weights; inputting the rhizosphere soil standardized data into a time series convolutional neural network module for time series change pattern recognition, extracting and processing the time series change rule of rhizosphere soil parameters to obtain rhizosphere time series feature vectors; performing attention mechanism fusion calculation based on the rhizosphere static feature weights and the rhizosphere time series feature vectors, weighting and merging the static features and time series information according to dynamic weights to obtain rhizosphere soil comprehensive features; iteratively optimizing and training the rhizosphere soil comprehensive features by gradient descent method to adaptively adjust the parameters of the hybrid algorithm, obtaining the rhizosphere soil health assessment algorithm.

[0026] Specifically, the rhizosphere soil standardized data is input into the random forest module for nonlinear feature relationship learning. The model analyzes the complex correlation between various soil parameters, automatically identifies and establishes the nonlinear relationship between them, which is used to calculate the weights of rhizosphere static features. Through this process, the model can understand the relative importance of different soil parameters in the overall health assessment and provide accurate feature weights for subsequent analysis. For example, assuming that the weights of soil pH value, temperature and humidity, and nutrient concentration are , the comprehensive features of static features can be represented as: wherein, is the rhizosphere static feature, represents the pH value, represents the temperature and humidity, represents the nutrient concentration, are the weights of each feature respectively. These standardized data are input into the time series convolutional neural network module for identification of time series change patterns. The model can identify the trend and pattern of soil state change over time by extracting the time series change rule of soil parameters, thereby generating a rhizosphere time series feature vector. For example, the calculation of the time series feature vector can be represented as: wherein, is the time series feature vector, are the soil parameter data at the previous moment, the current moment and the next moment respectively, denotes the function of extracting time series patterns in the time series convolutional neural network. Based on the rhizosphere static feature weight and the rhizosphere time series feature vector, attention mechanism fusion calculation is performed. The model weights and combines the static features and time series information according to the dynamic weight. In this way, the static characteristics and dynamic change characteristics of the rhizosphere soil are effectively fused to obtain a comprehensive rhizosphere soil feature representation. The rhizosphere soil comprehensive feature is iteratively optimized and trained by gradient descent method. The model continuously adjusts the parameters of the hybrid algorithm for adaptive optimization, so that the health assessment algorithm can accurately predict the soil health status.

[0027] Specifically, in the attention mechanism fusion calculation part, the model combines the rhizosphere static feature and the time series feature, and uses dynamic weights to weight and combine the two parts of information. Specifically, the core of the attention mechanism is to dynamically adjust the weight according to the importance of each feature in the current soil health assessment. Assuming that the rhizosphere static feature weight is and the rhizosphere time series feature weight is , the fused comprehensive feature can be represented in the form of the following weighted sum: wherein, denotes the rhizosphere static feature, denotes the rhizosphere time series feature, and are dynamic weight coefficients in the attention mechanism, representing the importance of static features and time series features in the current health assessment. By calculating these two weight coefficients, the model can adaptively adjust the contribution of features in the assessment according to the characteristics of the actual data. Specifically, and will be continuously updated through the training process, so that the model can better capture the change rule of soil features and obtain a comprehensive rhizosphere soil feature representation, as described in Figure 2The figure shows the rhizosphere soil health assessment process based on machine learning.

[0028] For example, when the rhizosphere soil standardized data is input into the random forest module for nonlinear feature relationship learning, the system analyzes multiple parameters such as pH, conductivity, temperature and humidity, nutrient concentration, and microbial activity. In this process, the random forest algorithm discovers complex nonlinear relationships between different soil parameters through training data, such as hidden correlations between conductivity and nutrient concentration, which help the model assign corresponding feature weights to each parameter, providing accurate static features for subsequent assessment.

[0029] For example, when combining rhizosphere static feature weights and rhizosphere time series feature vectors, the model automatically assigns weights according to the dynamic changes and importance of the data through attention mechanism fusion calculation. For example, when the nutrient concentration of the soil has a greater impact on health assessment at a certain stage, the model will automatically increase the weight of this feature to ensure that the soil health assessment can make more accurate judgments according to the feature changes at different stages.

[0030] For example, when the rhizosphere soil comprehensive features are iteratively optimized and trained through gradient descent, the model adjusts the parameters of the hybrid algorithm through backpropagation to optimize the rhizosphere soil health assessment algorithm. For example, if the evaluation result at a certain stage deviates greatly, the gradient descent method will adjust the model parameters according to the error feedback, so that the system can more accurately predict the soil health status in subsequent training, obtaining an intelligent soil health assessment tool.

[0031] In a specific embodiment, the execution step iteratively optimizes and trains the rhizosphere soil comprehensive features through gradient descent to adaptively adjust the parameters of the hybrid algorithm, obtaining the rhizosphere soil health assessment algorithm process can specifically include the following steps: Match the rhizosphere soil comprehensive features with the rhizosphere soil health annotation data, assign corresponding health score labels to each group of rhizosphere soil samples, and obtain the rhizosphere soil training dataset; Calculate the loss function value based on the rhizosphere soil training dataset, and quantitatively calculate the error between the predicted result and the true label to obtain the rhizosphere soil health prediction error; Calculate the gradient of the rhizosphere soil health prediction error through the backpropagation algorithm, and perform gradient calculation on the gradient values of each layer parameter in the hybrid algorithm to obtain the rhizosphere soil parameter gradient vector; Perform parameter update operation based on the rhizosphere soil parameter gradient vector, and iteratively adjust the hybrid algorithm weight parameters according to the learning rate to obtain the rhizosphere soil health assessment algorithm.

[0032] Specifically, the comprehensive characteristics of the rhizosphere soil are matched with the rhizosphere soil health annotation data, and the system associates each group of rhizosphere soil samples with the corresponding health score label to form a rhizosphere soil training data set. These labels represent the health status of each sample. Through this step, the model obtains a data set for training, which includes multiple features and corresponding health scores. Then, based on the rhizosphere soil training data set, the system calculates the loss function value to measure the error between the predicted result and the true label. Through quantitative calculation, the error is converted into a numerical value to obtain the rhizosphere soil health prediction error. The loss function often uses the mean square error (MSE) or cross entropy loss function. The specific calculation formula is: ,in, is the loss function value, It is The true labels of samples, is the health score predicted by the model, is the number of samples. The rhizosphere soil health prediction error is calculated by gradient through the back propagation algorithm. The model uses the chain rule to derive the parameters of each layer in the hybrid algorithm, calculates the gradient value of each parameter, and obtains the gradient vector of the rhizosphere soil parameters. These gradient values ​​represent the contribution of each parameter to the error, which helps to determine which parameters need to be adjusted and the direction of adjustment. Through back propagation, the model can capture the relationship between features and errors, and optimize the performance of the algorithm based on this information. Based on the rhizosphere soil parameter gradient vector, the system performs a parameter update operation, and iteratively adjusts the weight parameters of the hybrid algorithm through the gradient descent method. The update operation is performed according to the set learning rate to ensure that the model gradually converges to the optimal solution. The learning rate controls the step size of each update. Too large may cause oscillation, and too small may cause slow convergence. Through continuous iterative updates, the optimized rhizosphere soil health assessment algorithm is obtained, which can accurately assess soil health status and has strong generalization ability. Reference Figures 3~5 , showing a comparison chart of the training effect of the rhizosphere soil health assessment algorithm based on machine learning, Figure 3 The training and validation loss reduction curves are shown in Figure 4 The correlation comparison of the prediction results is shown. Figure 5 The back-propagation gradient mean is shown. The method described in the figure, where RF-only: only uses random forests and ignores temporal dynamics, TCN-only: only uses temporal networks and ignores static associations, the method in this paper: is a hybrid machine learning framework that uses random forests to model the nonlinear associations of static parameters, temporal convolutional networks to capture dynamic changes, and combines the attention mechanism to achieve feature adaptive fusion, and jointly optimizes the health score prediction task through the gradient descent method.

[0033] For example, when matching the comprehensive features of rhizosphere soil with the rhizosphere soil health annotation data, the system combines the features of each soil sample (such as pH value, temperature and humidity, nutrient concentration, etc.) with the corresponding health score label to form a training data set.

[0034] For example, a soil sample has a pH value of 6.5, a temperature and humidity of 70%, and a high nutrient concentration. These data will be paired with the health score of the sample (such as 0.85, indicating a healthy state) to form a complete training sample. The system evaluates the prediction accuracy of the model by calculating the loss function value. If the true health score of a soil sample is 0.85, and the model predicts a score of 0.75, the system will calculate the prediction error, reflecting the size of the model error. The smaller the loss function value, the more accurate the model prediction. The model calculates the gradient of each parameter through the backpropagation algorithm based on the loss function value.

[0035] For example, in the hybrid algorithm, if a feature (such as temperature and humidity) has a greater impact on the prediction result, the gradient calculation will result in a larger gradient value for that feature parameter, indicating that this feature contributes more to the error and needs to be adjusted more significantly. The system performs parameter update operations based on the gradient vector. Assuming the learning rate is 0.01, the model will adjust the weight parameters based on this learning rate to gradually reduce the error. After multiple iterations of updating, the model can finally optimize the parameters and accurately predict the soil health status.

[0036] In a specific embodiment, the S4 step further comprises: Inputting the real-time collected rhizosphere sensor data into the rhizosphere soil health assessment algorithm for prediction calculation to quantitatively evaluate the current rhizosphere soil state and obtain the rhizosphere soil health score; Setting a health state judgment threshold based on the rhizosphere soil health score, classifying the health score according to a preset interval, determining the health state when the health score is greater than or equal to 0.8, determining the intermediate health state when the health score is greater than or equal to 0.4 and less than 0.8, and determining the unhealthy state when the health score is less than 0.4, to obtain the rhizosphere soil health state classification result; Calculating the difference of the rhizosphere soil health score within a continuous time window to quantitatively analyze the time variation trend of the health score and obtain the health score change rate; Performing trend judgment operation based on the health score change rate to predict and analyze the development direction of the rhizosphere soil health state and obtain the rhizosphere soil health state category.

[0037] Specifically, the real-time collected rhizosphere sensor data is input into the rhizosphere soil health assessment algorithm for predictive calculation. The model analyzes the real-time data of the soil and obtains the current rhizosphere soil health score, which is used to quantitatively evaluate the health status of the soil. According to the obtained health score, the system sets a health status judgment threshold and classifies the health score according to the preset interval. For example, when the health score is greater than or equal to 0.8, the system determines that the soil state is healthy; when the health score is between 0.4 and 0.8, the system determines that the soil state is in a moderate health state; and when the health score is less than 0.4, the system determines that the soil state is unhealthy. This classification result helps to quickly determine the overall health status of the soil, and thus provides a decision basis for agricultural management. Then, the system calculates the difference of the rhizosphere soil health score in the continuous time window to quantify the time variation trend of the health score. This analysis helps to observe the change of the soil health score over time, reveals the fluctuation of the soil health, and obtains the health score change rate, which further measures the improvement or deterioration speed of the soil health status. Based on the calculated health score change rate, the system performs a trend judgment operation to predict and analyze the development direction of the rhizosphere soil health status. Through this prediction, the system can timely predict the change trend of the soil health status and provide early warning for the future soil health status, so as to provide more accurate regulation and decision support for agricultural management.

[0038] Taking a certain agricultural plantation as an example, the real-time collected rhizosphere sensor data is input into the soil health assessment algorithm for predictive calculation. The sensor monitors the pH value, temperature and humidity, and nutrient concentration of the soil, and the algorithm calculates the soil health score in real time based on these data. Assuming that the health score is 0.75 at this time. According to the set health status judgment threshold, the score 0.75 falls in the interval of the moderate health state, so the soil is determined to be in a moderate health state. Then, the system calculates the difference of the health score in the continuous week, and analyzes that the health score change rate is +0.05, indicating that the soil health status is slowly improving. Based on this change rate, the system predicts that the soil health status may further improve in the next few days, adjusts the early warning system and prepares the corresponding optimization measures, such as increasing fertilizer or adjusting the irrigation plan. Through such analysis, the agricultural management personnel can make timely decisions according to the real-time dynamics of the soil health, effectively manage the soil health, and ensure that the growth conditions of crops remain in the best state.

[0039] In a specific embodiment, the S5 step further comprises: Based on the rhizosphere soil health status category, the key influence factor deviation degree is identified, and the pH value deviation, nutrient content deviation, and microbial activity deviation are quantitatively analyzed and processed to obtain the rhizosphere factor deviation parameter; The rhizosphere factor deviation parameters are input into the rhizosphere factor weight self-learning algorithm for weight calculation, different rhizosphere environmental factors are dynamically weighted according to their influence on soil health, and the rhizosphere factor weight coefficient is obtained; Based on the rhizosphere factor weight coefficient, the regulator dosage calculation is performed, the pH regulator dosage is calculated according to the product of the pH deviation value and the rhizosphere soil volume, the fertilizer dosage is calculated according to the product of the nutrient deviation value and the plant biomass, and the biological agent dosage is calculated according to the product of the microbial activity deviation value and the soil organic matter content, to obtain the regulator dosage parameter; The regulator dosage parameter is associated with the expected regulation effect, and the implementation time and matters needing attention of the regulation measures are integrated to obtain the rhizosphere soil optimization scheme.

[0040] Specifically, based on the rhizosphere soil health state category, the system identifies and analyzes the key factor deviation degree affecting soil health, such as pH value deviation, nutrient content deviation and microbial activity deviation. Through quantitative analysis, the system obtains the deviation parameters of each factor, which reflect the gap between the current rhizosphere soil and the ideal healthy state. The system inputs these rhizosphere factor deviation parameters into the rhizosphere factor weight self-learning algorithm for weight calculation, and the algorithm dynamically allocates weights according to the influence of each factor on soil health. Through this process, the system can automatically adjust the weight of different factors according to the actual state of the soil, maximize the influence on health assessment, and ensure the effectiveness of soil improvement measures. Based on the calculated rhizosphere factor weight coefficient, the system performs regulator dosage calculation. For pH value regulator, the system calculates the required amount of regulator according to the product of the pH deviation value and the soil volume; for fertilizer dosage, the system calculates the required amount of fertilizer according to the product of the nutrient deviation value and the plant biomass; for biological agent, the system calculates the required amount of agent according to the product of the microbial activity deviation value and the soil organic matter content. All calculation results are summarized as regulator dosage parameters to ensure that the amount of each regulator matches the actual needs of the soil. The system then associates the regulator dosage parameter with the expected regulation effect to ensure that the regulation measures can achieve the best effect. According to the analysis results, the system integrates the implementation time and matters needing attention of the regulation measures, and finally generates the rhizosphere soil optimization scheme to provide effective soil improvement strategies for agricultural management. Through this process, soil health management becomes more accurate and intelligent, providing continuous support for the healthy growth of crops.

[0041] For example, in identifying the degree of bias in key influencing factors based on the rhizosphere soil health status category, the system first identifies that the current rhizosphere soil pH value is 5.2, which is lower than the ideal range of 6.0-7.0, the nutrient content of nitrogen, phosphorus and potassium is low, and the microbial activity is also low. These deviations are obtained through quantitative analysis to obtain deviation parameters, for example, the pH value deviation is -0.8, the nutrient deviation is -10%, and the microbial activity deviation is -15%. Then, the system inputs these deviation parameters into the rhizosphere factor weight self-learning algorithm, and according to the influence degree of each factor on soil health, the algorithm assigns dynamic weights to each factor. Assuming that the influence weight of pH value on soil health is 0.5, the influence weight of nutrient content is 0.3, and the influence weight of microbial activity is 0.2. Next, the system calculates the required amount of adjusting agent according to these weight coefficients. For example, for pH adjusting agent, the system determines the required amount by calculating the product of the pH deviation value and the soil volume. If the soil volume is 100 cubic meters, the required amount of adjusting agent is 0.8 (pH deviation) x 100 (volume), resulting in a required adjusting agent amount of 80 units. For fertilizer, assuming that the nutrient deviation is -10% and the plant biomass is 1500 kg, the system will calculate the fertilizer amount as 150 kg. For microbial agents, the system calculates the required amount of microbial agents by multiplying the microbial activity deviation by the soil organic matter content, which is 3%, resulting in a calculation of 45 units of microbial agents. Finally, all these data will generate an adjusting agent amount scheme, which the system compares with the expected adjusting effect to ensure that these measures can achieve the expected goal of soil health.

[0042] For example, in identifying the degree of bias in key influencing factors based on the rhizosphere soil health status category, the system first identifies that the current rhizosphere soil pH value is 5.2, which is lower than the ideal range of 6.0-7.0, the nutrient content of nitrogen, phosphorus and potassium is low, and the microbial activity is also low. These deviations are obtained through quantitative analysis to obtain deviation parameters, for example, the pH value deviation is -0.8, the nutrient deviation is -10%, and the microbial activity deviation is -15%. Then, the system inputs these deviation parameters into the rhizosphere factor weight self-learning algorithm, and according to the influence degree of each factor on soil health, the algorithm assigns dynamic weights to each factor. Assuming that the influence weight of pH value on soil health is 0.5, the influence weight of nutrient content is 0.3, and the influence weight of microbial activity is 0.2. Next, the system calculates the required amount of adjusting agent according to these weight coefficients. For example, for pH adjusting agent, the system determines the required amount by calculating the product of the pH deviation value and the soil volume. If the soil volume is 100 cubic meters, the required amount of adjusting agent is 0.8 (pH deviation) x 100 (volume), resulting in a required adjusting agent amount of 80 units. For fertilizer, assuming that the nutrient deviation is -10% and the plant biomass is 1500 kg, the system will calculate the fertilizer amount as 150 kg. For microbial agents, the system calculates the required amount of microbial agents by multiplying the microbial activity deviation by the soil organic matter content, which is 3%, resulting in a calculation of 45 units of microbial agents. Finally, all these data will generate an adjusting agent amount scheme, which the system compares with the expected adjusting effect to ensure that these measures can achieve the expected goal of soil health.

[0043] The method for evaluating rhizosphere soil health based on machine learning in the embodiments of the present application is described above, and the system for evaluating rhizosphere soil health based on machine learning in the embodiments of the present application is described below. Please refer to Figure 6 The system for evaluating rhizosphere soil health based on machine learning in the embodiments of the present application includes one embodiment as follows: A collection module is configured to collect pH value, conductivity, temperature and humidity, nutrient concentration and microbial activity data of a target plant rhizosphere region as a rhizosphere soil original data set; A correction module is configured to correct outliers in the rhizosphere soil original data set by a rhizosphere time series median filtering method, calculate a rhizosphere nutrient comprehensive index and a rhizosphere microbial activity index, and obtain rhizosphere soil standardized data; A training module is configured to train the rhizosphere soil standardized data by a hybrid algorithm, fuse static features and time series information through an attention mechanism, and establish a rhizosphere soil health evaluation algorithm; An output module is configured to output a rhizosphere soil health score through the rhizosphere soil health evaluation algorithm, and determine a rhizosphere soil health state category according to a health score change rate; A calculation module is configured to calculate an adjusting agent dosage parameter through a rhizosphere factor weight self-learning algorithm based on the rhizosphere soil health state category, and output a rhizosphere soil optimization scheme.

[0044] The system for evaluating rhizosphere soil health based on machine learning in the embodiments of the present application is described above Figure 6 The system for evaluating rhizosphere soil health based on machine learning in the embodiments of the present application is described above

[0045] Referring to Figure 7 The embodiments of the present application also provide a device for evaluating rhizosphere soil health based on machine learning, which can be a server, and the internal structure thereof can be as follows Figure 7The machine learning-based rhizosphere soil health assessment device shown in the figure includes a processor, a memory, a display screen, an input device, a network interface and a database connected through a system bus. Among them, the processor designed by the computer is used to provide computing and control capability. The memory of the machine learning-based rhizosphere soil health assessment device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the machine learning-based rhizosphere soil health assessment device is used to store the corresponding data in this embodiment. The network interface of the machine learning-based rhizosphere soil health assessment device is used to communicate with the external terminal through the network connection. The computer program is executed by the processor to realize the above method.

[0046] Those skilled in the art can understand that, Figure 7 The structure shown in the figure is only a block diagram of part of the structure related to the scheme of the application, and does not constitute a limitation on the machine learning-based rhizosphere soil health assessment device to which the scheme of the application is applied.

[0047] The application also provides a computer readable storage medium, which can be a non-volatile computer readable storage medium, and can also be a volatile computer readable storage medium. The computer readable storage medium stores instructions, and when the instructions run on a computer, the computer executes the steps of the machine learning-based rhizosphere soil health assessment method.

[0048] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working process of the above-mentioned system, system and unit can refer to the corresponding process in the foregoing method embodiments, which will not be repeated here.

[0049] The integrated unit, if realized in the form of a software function unit and sold or used as an independent product, can be stored in a computer readable storage medium. Based on such understanding, the technical scheme of the application or the whole or part of the technical scheme that contributes to the prior art can be embodied in the form of a software product. The computer software product stored in a storage medium includes a plurality of instructions for causing a machine learning-based rhizosphere soil health assessment device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the method described in each embodiment of the application. The foregoing storage medium includes a U disk, a mobile hard disk, a read-only memory (read-only memory, ROM), a random access memory (random access memory, RAM), a magnetic disk or an optical disk, and various media that can store program codes.

[0050] The above examples are only used to illustrate the technical solutions of the present application, but not to limit the present application; although the present application has been described in detail with reference to the foregoing examples, those ordinarily skilled in the art should understand: the technical solutions recorded in the foregoing examples can be modified, or some technical features can be replaced equivalently; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A method for rhizosphere soil health assessment based on machine learning, characterized in that, The method comprises: S1, collecting the pH value, conductivity, temperature and humidity, nutrient concentration and microbial activity data of the rhizosphere region of the target plant as the rhizosphere soil original data set; S2, correcting the abnormal values in the rhizosphere soil original data set by the rhizosphere time series median filtering method, calculating the rhizosphere nutrient comprehensive index and the rhizosphere microbial activity index, and obtaining the rhizosphere soil standardized data; S3, training the rhizosphere soil standardized data by a hybrid algorithm, fusing static features and time series information through an attention mechanism, and establishing a rhizosphere soil health evaluation algorithm; S4, outputting the rhizosphere soil health score by the rhizosphere soil health evaluation algorithm, and judging the rhizosphere soil health state category according to the health score change rate; S5, based on the rhizosphere soil health state category, calculating the regulator dosage parameter by a rhizosphere factor weight self-learning algorithm, and outputting the rhizosphere soil optimization scheme.

2. The machine learning based rhizosphere soil health assessment method of claim 1, wherein, The S1 step further comprises: Deploying a rhizosphere sensor network in the rhizosphere region of the target plant to synchronously monitor the rhizosphere microenvironment and obtain rhizosphere environment monitoring data; Collecting the rhizosphere environment monitoring data at preset time intervals, collecting the pH value data every 30 minutes, collecting the conductivity data every 30 minutes, collecting the temperature and humidity data every 60 minutes, collecting the nutrient concentration data every 120 minutes, and collecting the microbial activity data every 120 minutes to obtain rhizosphere soil time series collection data; Transmitting the rhizosphere soil time series collection data through a wireless communication module, adding a time stamp, a sensor identifier and location information to each data record, and obtaining the rhizosphere soil original data set; Based on the rhizosphere soil original data set, performing data integrity verification and marking missing data to obtain the rhizosphere soil original data set.

3. The machine learning based rhizosphere soil health assessment method as claimed in claim 1, wherein, The S2 step further comprises: Range testing the pH value, conductivity, temperature and humidity, nutrient concentration and microbial activity data in the rhizosphere soil original data set, marking values exceeding the preset range of the rhizosphere environment as abnormal values, and obtaining abnormal value marked data; Based on the abnormal value marked data, performing rhizosphere time series median filtering processing, replacing and correcting abnormal values with the median of adjacent time point data to obtain rhizosphere soil corrected data; Calculating and processing the nitrogen, phosphorus and potassium concentrations and organic acid concentrations in the rhizosphere soil corrected data according to a weighted summation formula to obtain a rhizosphere nutrient comprehensive index; Weighted calculating and processing the urease activity, phosphatase activity and invertase activity in the rhizosphere soil corrected data according to enzyme activity weight coefficients to obtain a rhizosphere microbial activity index and rhizosphere soil standardized data.

4. The machine learning based rhizosphere soil health assessment method as claimed in claim 1, wherein, The S3 step further comprises: Inputting the rhizosphere soil standardized data into a random forest module to learn non-linear feature relationships, modeling the complex correlations between rhizosphere soil parameters, and obtaining rhizosphere static feature weights; Inputting the rhizosphere soil standardized data into a time series convolutional neural network module to identify time series change patterns, extracting the time series change rules of rhizosphere soil parameters, and obtaining a rhizosphere time series feature vector; Perform attention mechanism fusion calculation based on the rhizosphere static feature weight and rhizosphere time sequence feature vector, and perform weighted and combined processing on the static feature and time sequence information according to a dynamic weight to obtain rhizosphere soil comprehensive features; Iterative optimization training is performed on the rhizosphere soil comprehensive features by using the gradient descent method, and adaptive adjustment processing is performed on the mixed algorithm parameters to obtain a rhizosphere soil health evaluation algorithm.

5. The machine learning based rhizosphere soil health assessment method as claimed in claim 4, wherein, The iterative optimization training is performed on the rhizosphere soil comprehensive features by using the gradient descent method, and adaptive adjustment processing is performed on the mixed algorithm parameters to obtain a rhizosphere soil health evaluation algorithm, including: The rhizosphere soil comprehensive features are matched with rhizosphere soil health labeled data, and a corresponding health score label is assigned to each group of rhizosphere soil samples to obtain a rhizosphere soil training data set; Based on the rhizosphere soil training data set, a loss function value is calculated, and the error between the predicted result and the true label is quantitatively calculated to obtain a rhizosphere soil health prediction error; The rhizosphere soil health prediction error is calculated by using the back propagation algorithm, and the gradient value of each layer parameter in the mixed algorithm is derived to obtain a rhizosphere soil parameter gradient vector; Based on the rhizosphere soil parameter gradient vector, a parameter update operation is performed, and the weight parameters of the mixed algorithm are iteratively adjusted according to the learning rate to obtain a rhizosphere soil health evaluation algorithm.

6. The machine learning based rhizosphere soil health assessment method as claimed in claim 1, wherein, The S4 step further includes: The rhizosphere sensor data collected in real time are input into the rhizosphere soil health evaluation algorithm for prediction calculation, and the current rhizosphere soil state is quantitatively evaluated to obtain a rhizosphere soil health score; Based on the rhizosphere soil health score, a health state judgment threshold is set, and the health score is classified according to a preset interval, when the health score is greater than or equal to 0.8, it is determined as a healthy state, when the health score is greater than or equal to 0.4 and less than 0.8, it is determined as a medium health state, and when the health score is less than 0.4, it is determined as an unhealthy state, to obtain a rhizosphere soil health state classification result; The rhizosphere soil health scores in the continuous time window are calculated by using a difference value, and the time variation trend of the health score is quantitatively analyzed to obtain a health score change rate; Based on the health score change rate, a trend judgment operation is performed, and the development direction of the rhizosphere soil health state is predicted and analyzed to obtain a rhizosphere soil health state category.

7. The machine learning based rhizosphere soil health assessment method as claimed in claim 1, wherein, The S5 step further includes: Based on the rhizosphere soil health state category, the key influence factor deviation degree is identified, and the pH value deviation, nutrient content deviation and microbial activity deviation are quantitatively analyzed to obtain rhizosphere factor deviation parameters; The rhizosphere factor deviation parameters are input into a rhizosphere factor weight self-learning algorithm for weight calculation, and different rhizosphere environmental factors are dynamically weighted and distributed according to their influence on soil health to obtain rhizosphere factor weight coefficients; Based on the rhizosphere factor weight coefficient, the regulator dosage calculation is performed, the pH regulator dosage is calculated according to the product of the pH deviation value and the rhizosphere soil volume, the fertilizer dosage is calculated according to the product of the nutrient deviation value and the plant biomass, and the biological agent dosage is calculated according to the product of the microbial activity deviation value and the soil organic matter content, to obtain the regulator dosage parameter; The regulator dosage parameter is associated with the expected regulation effect, and the implementation time and matters needing attention of the regulation measures are integrated to obtain the rhizosphere soil optimization scheme.

8. A machine learning based rhizosphere soil health assessment system, characterized in that, The machine learning-based rhizosphere soil health evaluation system for implementing the machine learning-based rhizosphere soil health evaluation method according to any one of claims 1-7 comprises: A collection module for collecting pH, conductivity, temperature and humidity, nutrient concentration and microbial activity data in the rhizosphere region of the target plant as the rhizosphere soil original data set; A correction module for correcting outliers in the rhizosphere soil original data set by a rhizosphere time series median filtering method, calculating a rhizosphere nutrient comprehensive index and a rhizosphere microbial activity index, and obtaining a rhizosphere soil standardized data; A training module for training the rhizosphere soil standardized data by a hybrid algorithm, fusing static features and time series information through an attention mechanism, and establishing a rhizosphere soil health evaluation algorithm; An output module for outputting a rhizosphere soil health score through the rhizosphere soil health evaluation algorithm, and determining the rhizosphere soil health state category according to the health score change rate; A calculation module for calculating regulator dosage parameters based on the rhizosphere soil health state category through a rhizosphere factor weight self-learning algorithm, and outputting a rhizosphere soil optimization scheme. 9.A rhizosphere soil health assessment device based on machine learning, characterized by, A computer program stored in a memory and executable on a processor, wherein the processor implements the machine learning-based rhizosphere soil health evaluation method according to any one of claims 1-7 when executing the computer program.

10. A computer-readable storage medium having stored thereon a computer program, characterized in that, The computer program, when executed by the processor, causes the processor to perform the machine learning-based rhizosphere soil health evaluation method according to any one of claims 1-7.

Images