Soil moisture status assessment and quality classification method for cultivated land based on machine learning

Through a machine learning-based method, combined with remote sensing and Internet of Things technology, the problem that traditional soil moisture assessment technology is difficult to fully reflect the spatial variability and time-consuming and labor-consuming of soil moisture is solved, and the rapid and accurate assessment of soil moisture content is achieved, which improves the Evaluation of accuracy and reliability provides support for precision agricultural management.

CN118761684BActive Publication Date: 2024-11-15YUNNAN HANZHE TECHN CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202411245657.4
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-09-06
Publication Date
2024-11-15
Estimated Expiration
2044-09-06

AI Technical Summary

Technical Problem

Traditional soil moisture assessment technology is difficult to fully reflect the spatial variability of soil moisture, which is time-consuming and labor-intensive, and the evaluation results are greatly affected by human factors, lack objectivity and consistency, making it difficult to achieve large-scale and high-frequency monitoring.

Method used

Using a machine learning-based method, combining remote sensing technology and Internet of Things technology, multi-source historical data is obtained for geometric correction and radiation calibration, a training data set is constructed, and soil moisture content is evaluated using the R-CNN model, and the model is optimized through the objective function. Comprehensive evaluation values ​​of soil moisture status were generated and graded.

Benefits of technology

It realizes rapid and accurate assessment of soil moisture content, improves assessment accuracy and reliability, reduces costs, and provides support for precise agricultural management and optimized water resource allocation.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118761684B_ABST
    Figure CN118761684B_ABST
Patent Text Reader

Abstract

The invention relates to the technical field of cultivated land quality detection, and in particular to a method for evaluating and grading the soil moisture status of cultivated land based on machine learning. The method comprises the following steps: acquiring multi-source historical data of cultivated land, and performing geometric correction, radiation calibration and atmospheric correction to form a training data set and a test data set; training a soil moisture content evaluation index model based on the training data set, and calculating the loss value between the test data set and the predicted soil moisture content evaluation index, and optimizing the soil moisture content evaluation index model by setting an objective function; calculating a predicted value of a soil moisture content index of cultivated land by completing the trained soil moisture content evaluation index model, performing quantitative calculation on the predicted value of the soil moisture content index of cultivated land, and performing comprehensive evaluation and grading on a comprehensive evaluation value of the soil moisture status based on a set judgment logic to generate a quality distribution map of the cultivated land.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of cultivated land quality detection, and in particular to a method for evaluating the soil moisture status and quality grading of cultivated land based on machine learning. Background Art

[0002] Soil moisture is one of the key factors affecting the quality of cultivated land and crop growth. Accurately assessing and monitoring the soil moisture status of cultivated land is of great significance for agricultural production management and decision-making. At present, traditional soil moisture assessment technology usually uses field sampling and laboratory analysis. Due to the limited sampling points of field sampling, it is not only difficult to fully reflect the spatial variability of soil moisture in the cultivated land area, but also the sampling and analysis process is time-consuming and labor-intensive, making it difficult to achieve large-scale and high-frequency monitoring. In addition, the assessment results are greatly affected by human factors and lack objectivity and consistency. Based on this, in order to overcome the above problems, we designed a method for assessing the soil moisture status and quality grading of cultivated land based on machine learning. Summary of the invention

[0003] The purpose of the present invention is to provide a method for evaluating the moisture status and quality grading of cultivated land soil based on machine learning. It utilizes remote sensing technology and Internet of Things technology, and combines machine learning methods. It can not only realize the rapid and accurate evaluation of the moisture content of cultivated land soil, improve the evaluation accuracy and reliability, but also greatly improve the evaluation efficiency and reduce the cost. It provides strong support for precision agricultural management and optimal allocation of water resources, and has broad application prospects.

[0004] The embodiments of the present invention are implemented by the following technical solutions:

[0005] A method for evaluating soil moisture status and grading quality of cultivated land based on machine learning, the method includes the following steps:

[0006] Obtain multi-source historical data of cultivated land, perform geometric correction, radiation calibration and atmospheric correction on the multi-source historical data of cultivated land to construct training data sets and test data sets;

[0007] The preset soil moisture content evaluation index model is trained based on the training data set, and the loss value between the test data set and the predicted soil moisture content evaluation index is calculated. At the same time, the soil moisture content evaluation index model is optimized by setting the objective function, and the trained soil moisture content evaluation index model is output;

[0008] The predicted value of the soil moisture content index of cultivated land is calculated by completing the trained soil moisture content assessment index model, and the predicted value of the soil moisture content index of cultivated land is quantified to obtain a comprehensive assessment value of the soil moisture condition. Based on the set judgment logic, the comprehensive assessment value of the soil moisture condition is comprehensively evaluated and graded to generate a quality distribution map of cultivated land.

[0009] Optionally, the multi-source historical data of the cultivated land specifically includes: set resolution remote sensing image data, soil physical and chemical property data and meteorological data.

[0010] Optionally, the soil moisture content assessment index model is specifically an R-CNN model, which is composed of an input layer, a convolutional layer, a region proposal layer, a RoI pooling layer, a fully connected layer and an output layer, wherein the RoI pooling layer introduces channel attention and spatial attention.

[0011] Optionally, the calculation formula of the input layer is:

[0012]

[0013] in, is the remote sensing image data, H, W, and C are the height, width, and number of bands of the remote sensing image, respectively. is the soil physical and chemical property data, N is the number of samples, For meteorological data, , They are the characteristic dimensions of soil physical and chemical property data and meteorological data respectively;

[0014] The calculation formula of the convolutional layer is:

[0015]

[0016] in, , are the weights and biases of the convolutional layer, is the activation function;

[0017] The calculation formula of the region proposal layer is:

[0018]

[0019] in, is the weight of the region proposal layer, and R is the generated candidate region set;

[0020] The calculation formula of the RoI pooling layer is:

[0021]

[0022]

[0023]

[0024]

[0025]

[0026] in, is the feature map after pooling, , are the weights of channel attention, is the sigmoid activation function, for The convolution operation, is element-wise multiplication, For the flattening operation;

[0027] The calculation formula of the fully connected layer is:

[0028]

[0029] in, , are the weights and biases of the fully connected layer;

[0030] The calculation formula of the output layer is:

[0031]

[0032] in, , are the weights and biases of the output layer.

[0033] Optionally, the soil moisture content assessment index model is optimized by setting an objective function, wherein the optimization target is the optimal comprehensive performance of the soil moisture content assessment index model, and the calculation formula is:

[0034]

[0035] in, is the objective function, t is the time step, its value range is 1 to T, T is the total number of time steps, n is the sample number, its value range is 1 to N, N is the total number of samples, is the predicted value of soil moisture content of the nth sample by the soil moisture content evaluation index model at time step t, that is, , is the prediction utility function, which is used to measure the accuracy of the predicted value of the soil moisture content evaluation index model. is a resource cost function, which is used to measure the resource cost required to obtain and process the nth sample data. is the absolute value of the change in the predicted value of the soil moisture content assessment index model for the nth sample between time steps t and t-1, which is used to measure the stability of the prediction results. is the regularization term of the output layer parameters, which is used to control the complexity of the soil moisture content evaluation index model, that is, .

[0036] Optionally, the predicted value of the soil moisture content index of the cultivated land is quantitatively calculated to obtain a comprehensive assessment value of the soil moisture status, which is specifically:

[0037] The predicted value of soil moisture content of cultivated land is calculated by completing the trained soil moisture content evaluation index model ;

[0038] The predicted value of soil moisture content of cultivated land is normalized Map to the [0,1] interval to get the quantized value ;

[0039] Quantize the value The exponential weighted algorithm is used for calculation to obtain a comprehensive assessment value of soil moisture conditions.

[0040] Optionally, the quantized value Bring in the exponential weighted algorithm for calculation, and the specific calculation formula is:

[0041]

[0042] in, is the optimal value of soil moisture content, , To adjust the parameters.

[0043] Optionally, based on the set judgment logic, the comprehensive evaluation value of the soil moisture condition is comprehensively evaluated and graded, which is specifically:

[0044] Comprehensive assessment value based on soil moisture conditions , the soil moisture conditions are divided into the first level, the second level, the third level and the fourth level;

[0045] Construct a piecewise function:

[0046]

[0047] in, They are preset thresholds for classifying different soil moisture levels.

[0048] Threshold intervals according to soil moisture status levels Execution judgment:

[0049] Comprehensive assessment value for determining soil moisture status Is it in the interval If yes, the quality of the cultivated land is determined to be the first level; if no, proceed to the next step;

[0050] Comprehensive assessment value for determining soil moisture status Is it in the interval If yes, the quality of cultivated land is judged as the second level; if no, proceed to the next step;

[0051] Comprehensive assessment value for determining soil moisture status Is it in the interval If yes, the quality of cultivated land is judged as the third level; if not, the comprehensive assessment value of soil moisture status is In the interval The quality of the cultivated land is judged to be the fourth level, among which the quality of the first level of cultivated land is greater than the quality of the second level of cultivated land, which is greater than the quality of the third level of cultivated land, which is greater than the quality of the fourth level of cultivated land.

[0052] The technical solution of the embodiment of the present invention has at least the following advantages and beneficial effects:

[0053] The embodiments of the present invention utilize remote sensing technology and Internet of Things technology, combined with machine learning methods, which can not only realize the rapid and accurate assessment of the moisture content of cultivated land soil, improve the assessment accuracy and reliability, but also greatly improve the assessment efficiency and reduce the cost, providing strong support for precision agricultural management and optimal allocation of water resources, and has broad application prospects. BRIEF DESCRIPTION OF THE DRAWINGS

[0054] Figure 1 A schematic flow chart of a method for evaluating soil moisture status and grading quality of cultivated land based on machine learning provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0055] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Generally, the components of the embodiments of the present invention described and shown in the drawings here can be arranged and designed in various different configurations.

[0056] like Figure 1 As shown, Figure 1 A schematic flow chart of a method for evaluating soil moisture status and grading quality of cultivated land based on machine learning provided in an embodiment of the present invention.

[0057] In one embodiment, the present invention provides a method for evaluating soil moisture status and grading quality of cultivated land based on machine learning, the method comprising the following steps:

[0058] Obtain multi-source historical data of cultivated land, perform geometric correction, radiation calibration and atmospheric correction on the multi-source historical data of cultivated land to construct training data sets and test data sets;

[0059] The preset soil moisture content evaluation index model is trained based on the training data set, and the loss value between the test data set and the predicted soil moisture content evaluation index is calculated. At the same time, the soil moisture content evaluation index model is optimized by setting the objective function, and the trained soil moisture content evaluation index model is output;

[0060] The predicted value of the soil moisture content index of cultivated land is calculated by completing the trained soil moisture content assessment index model, and the predicted value of the soil moisture content index of cultivated land is quantified to obtain a comprehensive assessment value of the soil moisture condition. Based on the set judgment logic, the comprehensive assessment value of the soil moisture condition is comprehensively evaluated and graded to generate a quality distribution map of cultivated land.

[0061] Specifically, the multi-source historical data of cultivated land includes: remote sensing image data of set resolution, soil physical and chemical property data and meteorological data.

[0062] In this embodiment, the soil moisture content assessment index model is specifically an R-CNN model, which is composed of an input layer, a convolutional layer, a region proposal layer, a RoI pooling layer, a fully connected layer and an output layer, wherein the RoI pooling layer introduces channel attention and spatial attention.

[0063] The input layer receives multi-source data such as high-resolution remote sensing images, soil physical and chemical properties data, and meteorological data, and its calculation formula is:

[0064]

[0065] in, is the remote sensing image data, H, W, and C are the height, width, and number of bands of the remote sensing image, respectively. is the soil physical and chemical property data, N is the number of samples, For meteorological data, , They are the characteristic dimensions of soil physical and chemical properties data and meteorological data respectively.

[0066] The convolution layer extracts the local features of the remote sensing image through convolution operation, and its calculation formula is:

[0067]

[0068] in, , are the weights and biases of the convolutional layer, is the activation function.

[0069] The region proposal layer generates candidate regions, and its calculation formula is:

[0070]

[0071] in, is the weight of the region proposal layer, and R is the set of generated candidate regions.

[0072] The RoI pooling layer performs feature extraction and pooling operations on the candidate region, and its calculation formula is:

[0073]

[0074] This embodiment introduces channel attention and spatial attention in the RoI pooling layer to enhance the soil moisture content assessment index model's attention to key areas and features. The calculation formula of the channel attention is:

[0075]

[0076] The calculation formula of the spatial attention is:

[0077]

[0078] The calculation formula of the attention feature map is:

[0079]

[0080] Then the pooled remote sensing image features, soil physical and chemical properties and meteorological characteristics are integrated:

[0081]

[0082] in, is the feature map after pooling, , are the weights of channel attention, is the sigmoid activation function, for The convolution operation, is element-wise multiplication, For the flattening operation.

[0083] The fully connected layer performs feature transformation through a fully connected operation, and its calculation formula is:

[0084]

[0085] in, , are the weights and biases of the fully connected layer.

[0086] The output layer maps the output of the fully connected layer to the predicted value of soil moisture content through linear regression, and its calculation formula is:

[0087]

[0088] in, , are the weights and biases of the output layer.

[0089] In the specific application of this embodiment, the soil moisture content evaluation index model is optimized by setting the objective function, and the optimization goal is to optimize the comprehensive performance of the soil moisture content evaluation index model, and the calculation formula is:

[0090]

[0091] in, is the objective function, t is the time step, its value range is 1 to T, T is the total number of time steps, n is the sample number, its value range is 1 to N, N is the total number of samples, is the predicted value of soil moisture content of the nth sample by the soil moisture content evaluation index model at time step t, that is, , is the prediction utility function, which is used to measure the accuracy of the predicted value of the soil moisture content evaluation index model. is a resource cost function, which is used to measure the resource cost required to obtain and process the nth sample data. is the absolute value of the change in the predicted value of the soil moisture content assessment index model for the nth sample between time steps t and t-1, which is used to measure the stability of the prediction results. is the regularization term of the output layer parameters, which is used to control the complexity of the soil moisture content evaluation index model, that is, .

[0092] In practice, 0.6, 0.3, and 0.1 are all weight coefficients. The significance of setting the weight coefficients to these values ​​in this embodiment is that: the weight coefficient of the prediction utility is set to 0.6, indicating that in this embodiment, the prediction utility of the soil moisture content assessment index model is emphasized, so that the soil moisture content assessment index model can accurately predict the soil moisture content and provide valuable information for agricultural production. The weight coefficient of resource cost is set to 0.3, indicating that in this embodiment, the resource cost of acquiring and processing data is taken into account, while ensuring the prediction utility, the resource consumption is reduced as much as possible. The weight coefficient of prediction stability is set to 0.1, indicating that in this embodiment, the prediction results of the soil moisture content assessment index model are further considered to maintain a certain stability over time to avoid drastic fluctuations. The reason why this embodiment is set to 0.1 is that the importance of stability is relatively low compared to the prediction utility and resource cost, so the weight coefficient set in this embodiment is relatively small. In addition, 0.01 is specifically the weight coefficient of the regularization term of the output layer parameter. The value of 0.01 is set to 0.01 to indicate that in this embodiment, the size of the output layer parameters is controlled by regularization to prevent the soil moisture content assessment index model from overfitting. This embodiment takes the optimal comprehensive performance of the soil moisture content assessment index model as the optimization goal, which can enable the soil moisture content assessment index model to achieve a good balance in prediction utility, resource cost, prediction stability and complexity, thereby realizing accurate prediction and evaluation of soil moisture content.

[0093] In this embodiment, the predicted value of the soil moisture content index of the cultivated land is quantitatively calculated to obtain a comprehensive evaluation value of the soil moisture status, which is specifically:

[0094] The predicted value of soil moisture content of cultivated land is calculated by completing the trained soil moisture content evaluation index model ;

[0095] The predicted value of soil moisture content of cultivated land is normalized Map to the [0,1] interval to get the quantized value ;

[0096] Quantize the value The exponential weighted algorithm is used for calculation to obtain a comprehensive assessment value of soil moisture conditions.

[0097] Specifically, the quantized value Bring in the exponential weighted algorithm for calculation, and the specific calculation formula is:

[0098]

[0099] in, is the optimal value of soil moisture content, , To adjust the parameters.

[0100] During implementation, the comprehensive evaluation value of soil moisture conditions is comprehensively evaluated and graded based on the set judgment logic, which is specifically:

[0101] Comprehensive assessment value based on soil moisture conditions , the soil moisture conditions are divided into the first level, the second level, the third level and the fourth level;

[0102] Construct a piecewise function:

[0103]

[0104] in, They are preset thresholds for classifying different soil moisture levels.

[0105] Threshold intervals according to soil moisture status levels Execution judgment:

[0106] Comprehensive assessment value for determining soil moisture status Is it in the interval If yes, the quality of the cultivated land is determined to be the first level; if no, proceed to the next step;

[0107] Comprehensive assessment value for determining soil moisture status Is it in the interval If yes, the quality of cultivated land is judged as the second level; if no, proceed to the next step;

[0108] Comprehensive assessment value for determining soil moisture status Is it in the interval If yes, the quality of cultivated land is judged as the third level; if not, the comprehensive assessment value of soil moisture status is In the interval The quality of the cultivated land is judged to be the fourth level, among which the quality of the first level of cultivated land is greater than the quality of the second level of cultivated land, which is greater than the quality of the third level of cultivated land, which is greater than the quality of the fourth level of cultivated land.

[0109] The above are only preferred embodiments of the present invention and are not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A method for evaluating soil moisture status and grading quality of cultivated land based on machine learning, characterized in that: The steps of the method include: Obtain multi-source historical data of cultivated land, perform geometric correction, radiation calibration and atmospheric correction on the multi-source historical data of cultivated land to construct training data sets and test data sets; The preset soil moisture content evaluation index model is trained based on the training data set, and the loss value between the test data set and the predicted soil moisture content evaluation index is calculated. At the same time, the soil moisture content evaluation index model is optimized by setting the objective function, and the trained soil moisture content evaluation index model is output; The predicted value of the soil moisture content index of the cultivated land is calculated by completing the trained soil moisture content evaluation index model, and the predicted value of the soil moisture content index of the cultivated land is quantified to obtain the comprehensive evaluation value of the soil moisture status. Based on the set judgment logic, the comprehensive evaluation value of the soil moisture status is comprehensively evaluated and graded to generate a quality distribution map of the cultivated land. The multi-source historical data of the cultivated land specifically includes: remote sensing image data of set resolution, soil physical and chemical property data and meteorological data; The soil moisture content assessment index model is specifically an R-CNN model, which is composed of an input layer, a convolutional layer, a region proposal layer, a RoI pooling layer, a fully connected layer and an output layer, wherein the RoI pooling layer introduces channel attention and spatial attention; The calculation formula of the input layer is: in, is the remote sensing image data, H, W, and C are the height, width, and number of bands of the remote sensing image, respectively. is the soil physical and chemical property data, N is the number of samples, For meteorological data, , They are the characteristic dimensions of soil physical and chemical property data and meteorological data respectively; The calculation formula of the convolutional layer is: in, , are the weights and biases of the convolutional layer, is the activation function; The calculation formula of the region proposal layer is: in, is the weight of the region proposal layer, and R is the generated candidate region set; The calculation formula of the RoI pooling layer is: in, is the feature map after pooling, , are the weights of channel attention, is the sigmoid activation function, for The convolution operation, is element-wise multiplication, For the flattening operation; The calculation formula of the fully connected layer is: in, , are the weights and biases of the fully connected layer; The calculation formula of the output layer is: in, , are the weights and biases of the output layer; The soil moisture content evaluation index model is optimized by setting an objective function, and the optimization goal is to optimize the comprehensive performance of the soil moisture content evaluation index model. The calculation formula is: in, is the objective function, t is the time step, its value range is 1 to T, T is the total number of time steps, n is the sample number, its value range is 1 to N, N is the total number of samples, is the predicted value of soil moisture content of the nth sample by the soil moisture content evaluation index model at time step t, that is, , is the prediction utility function, which is used to measure the accuracy of the predicted value of the soil moisture content evaluation index model. is a resource cost function, which is used to measure the resource cost required to obtain and process the nth sample data. is the absolute value of the change in the predicted value of the soil moisture content assessment index model for the nth sample between time steps t and t-1, which is used to measure the stability of the prediction results. is the regularization term of the output layer parameters, which is used to control the complexity of the soil moisture content evaluation index model, that is, ; The predicted value of the soil moisture content index of the cultivated land is quantitatively calculated to obtain a comprehensive assessment value of the soil moisture status, which is specifically: The predicted value of soil moisture content of cultivated land is calculated by completing the trained soil moisture content evaluation index model ; The predicted value of soil moisture content of cultivated land is normalized Map to the [0,1] interval to get the quantized value ; Quantize the value The exponential weighted algorithm is used for calculation to obtain a comprehensive assessment value of soil moisture conditions; The quantized value Bring in the exponential weighted algorithm for calculation, and the specific calculation formula is: in, is the optimal value of soil moisture content, , To adjust the parameters.

2. The method for evaluating soil moisture status and grading quality of cultivated land based on machine learning according to claim 1, characterized in that: Based on the set judgment logic, the comprehensive evaluation value of the soil moisture condition is comprehensively evaluated and classified, which is specifically: Comprehensive assessment value based on soil moisture conditions , the soil moisture conditions are divided into the first level, the second level, the third level and the fourth level; Construct a piecewise function: in, They are preset thresholds for classifying different soil moisture levels. Threshold intervals according to soil moisture status levels Execution judgment: Comprehensive assessment value for determining soil moisture status Is it in the interval If yes, the quality of the cultivated land is determined to be the first level; if no, proceed to the next step; Comprehensive assessment value for determining soil moisture status Is it in the interval If yes, the quality of cultivated land is judged as the second level; if no, proceed to the next step; Comprehensive assessment value for determining soil moisture status Is it in the interval If yes, the quality of cultivated land is judged as the third level; if not, the comprehensive assessment value of soil moisture status is In the interval The quality of the cultivated land is judged to be the fourth level, among which the quality of the first level of cultivated land is greater than the quality of the second level of cultivated land, which is greater than the quality of the third level of cultivated land, which is greater than the quality of the fourth level of cultivated land.