A method and system for evaluating the improvement effect of moderately and severely salinized farmland

Abstract

The present invention relates to the technical field of arable land improvement effect evaluation, and specifically discloses a method and system for evaluating the improvement effect of moderately to severely salinized arable land. The method comprises receiving an arable land range input by an administrator, randomly selecting evaluation points within the arable land range, and determining the evaluation level of the evaluation points; obtaining evaluation data containing point tags and time tags based on evaluation equipment installed at the evaluation points, and constructing an evaluation data matrix containing the point tags; receiving an evaluation instruction input by the administrator, extracting row data from the evaluation data matrix based on the input time of the evaluation instruction, analyzing the extracted row data, and outputting an evaluation result; and synchronously training an evaluation model based on the evaluation result. The present invention installs information collection equipment within the arable land, remotely obtains soil data, and performs real-time evaluation of the improvement effect. At the same time, a neural network model based on remote sensing images is trained based on the evaluation results, further improving recognition speed and enhancing timeliness.

Description

Technical Field

[0001] The present invention relates to the technical field of farmland improvement effect evaluation, and in particular to a method and system for evaluating the improvement effect of moderately and severely salinized farmland. Background Art

[0002] The improvement of moderately to severely salinized farmland is a complex and long-term process that requires comprehensive monitoring of multiple indicators to evaluate the improvement effect and timely adjust the improvement measures.

[0003] In the process of improving moderately and severely salinized farmland, it is necessary to evaluate the improvement effect and change the corresponding improvement measures to obtain better improvement results; however, the existing improvement effect evaluation process is a segmented process, and only after an improvement cycle is completed will an overall evaluation be carried out, which is very timeless. Summary of the Invention

[0004] The purpose of the present invention is to provide a method and system for evaluating the improvement effect of moderately and severely salinized farmland, so as to solve the problems raised in the above background technology.

[0005] To achieve the above object, the present invention provides the following technical solutions:

[0006] A method for evaluating the improvement effect of moderately and severely salinized farmland, the method comprising:

[0007] Receive the cultivated land range input by the administrator, randomly select evaluation points within the cultivated land range, and determine the evaluation level of the evaluation point; wherein the evaluation level is used to determine the data type of the data to be obtained at the evaluation point;

[0008] Based on the evaluation equipment installed at the evaluation point, the evaluation data containing point tags and time tags are obtained, and an evaluation data matrix containing the point tags is constructed; the time tag adopts relative time, and time zero is the improvement start time; the number of columns in the evaluation data matrix is ​​the same as the total number of data types, and the number of rows corresponds to time;

[0009] Receive the evaluation instruction input by the administrator, extract row data from the evaluation data matrix based on the input time of the evaluation instruction, analyze the extracted row data, and output the evaluation results;

[0010] When receiving the evaluation instruction input by the administrator, remote sensing images are obtained synchronously, and the remote sensing image at the time of improvement startup is obtained as the baseline image. The baseline image and the remote sensing image are used as features, and the evaluation results are used as labels to train the evaluation model; the application probability of the evaluation model and the application probability of the analysis process of the evaluation data matrix are adjusted according to the error rate of the evaluation model.

[0011] As a further solution of the present invention, the steps of receiving the cultivated land range input by the administrator, randomly selecting evaluation points within the cultivated land range, and determining the evaluation levels of the evaluation points include:

[0012] Receive the cultivated land range input by the administrator;

[0013] The cultivated land area is rasterized, and then a grid is inserted into the cultivated land area according to the preset unit size; the left and top lines of the grid are tangent to the cultivated land area;

[0014] Randomly select grid nodes as evaluation points and simultaneously determine the evaluation level of the evaluation points;

[0015] Update the selection probability of each grid node and execute the selection process cyclically;

[0016] The total number of selected evaluation points is recorded, and when the total number reaches a preset threshold, the selected evaluation points and their evaluation levels are output.

[0017] As a further solution of the present invention, the process of updating the selection probability of each grid node and cyclically executing the selection process includes:

[0018] Where, Represents a grid node The probability of selecting Represents a grid node The eigenvalue at is the total number of grid nodes in the horizontal direction, is the total number of grid nodes in the longitudinal direction; and The values ​​of are all positive integers; ; is the preset correction factor, For the selected The evaluation level of each evaluation point; For the selected Evaluation points and grid nodes distance;

[0019] When the grid node When selected as an evaluation point, Where, For grid nodes The evaluation level of The independent variable is The classification function is The increasing function of .

[0020] As a further solution of the present invention, the steps of receiving an evaluation instruction input by an administrator, extracting row data from the evaluation data matrix based on the input time of the evaluation instruction, analyzing the extracted row data, and outputting the evaluation results include:

[0021] Receive the evaluation instruction input by the administrator and record the input time of the evaluation instruction;

[0022] Extract the row data corresponding to the input time in the evaluation data matrix;

[0023] Traverse the preset standard data set based on row data and calculate the matching degree;

[0024] Taking the matching degree as the evaluation result at the input moment;

[0025] The standard data set is a theoretical data range of all data types at different preset moments.

[0026] As a further solution of the present invention: the step of traversing a preset standard data set based on row data and calculating the matching degree includes:

[0027] Read the standard data set at the input time;

[0028] Compare the row data extracted from the evaluation data matrix with each row data in the standard row data group and calculate the matching degree;

[0029] In the comparison process, the row data extracted from the evaluation data matrix are padded with columns, and the padded elements are preset default values.

[0030] As a further solution of the present invention, when receiving the evaluation instruction input by the administrator, a remote sensing image is synchronously acquired, a remote sensing image at the time of improvement startup is acquired as a reference image, the reference image and the remote sensing image are used as features, and the evaluation results are used as labels to train the evaluation model; and the step of adjusting the application probability of the evaluation model and the application probability of the analysis process of the evaluation data matrix according to the error rate of the evaluation model includes:

[0031] Acquire a remote sensing image at the time of improvement startup as a reference image;

[0032] When receiving the evaluation instruction input by the administrator, the remote sensing image is synchronously acquired as the remote sensing image at the input moment;

[0033] The reference image and the remote sensing image at the input moment are taken as an image group, and the evaluation results at the input moment are read to construct a sample set; the image group is the feature corresponding to the input, and the evaluation results are the label corresponding to the output;

[0034] An evaluation model is trained based on the sample set, and the application probability of the evaluation model and the application probability of the analysis process of the evaluation data matrix are adjusted according to the error rate of the evaluation model;

[0035] Among them, the application probability of the evaluation model is inversely proportional to the error rate, and the application probability of the analysis process of the evaluation data matrix is ​​directly proportional.

[0036] The technical solution of the present invention also provides a system for evaluating the improvement effect of moderately and severely salinized farmland, the system comprising:

[0037] The evaluation point planning module is used to receive the cultivated land range input by the administrator, randomly select evaluation points within the cultivated land range, and determine the evaluation level of the evaluation points; wherein the evaluation level is used to determine the data type of the data to be obtained at the evaluation point;

[0038] A data matrix construction module is used to obtain evaluation data containing point tags and time tags based on evaluation equipment installed at evaluation points, and to construct an evaluation data matrix containing point tags; the time tags use relative time, with time zero being the improvement start time; the number of columns in the evaluation data matrix is ​​the same as the total number of data types, and the number of rows corresponds to time;

[0039] The row data analysis module is used to receive the evaluation instruction input by the administrator, extract the row data from the evaluation data matrix based on the input time of the evaluation instruction, analyze the extracted row data, and output the evaluation results;

[0040] The model training application module is used to synchronously obtain remote sensing images when receiving evaluation instructions input by the administrator, obtain the remote sensing image at the time of improvement startup as the reference image, use the reference image and the remote sensing image as features, and use the evaluation results as labels to train the evaluation model; adjust the application probability of the evaluation model and the application probability of the analysis process of the evaluation data matrix according to the error rate of the evaluation model.

[0041] As a further solution of the present invention: the evaluation point planning module includes:

[0042] A range receiving unit, used for receiving the cultivated land range input by the administrator;

[0043] The grid insertion unit is used to rasterize the cultivated land range and then insert a grid within the cultivated land range according to the preset unit size; the left and top lines of the grid are tangent to the cultivated land range;

[0044] Random selection unit, used to randomly select grid nodes as evaluation points and simultaneously determine the evaluation level of the evaluation points;

[0045] A loop execution unit is used to update the selection probability of each grid node and loop through the selection process;

[0046] The first output unit is used to record the total number of selected evaluation points, and when the total number reaches a preset threshold, output the selected evaluation points and their evaluation levels.

[0047] As a further solution of the present invention: the row data analysis module includes:

[0048] An instruction receiving unit, configured to receive an evaluation instruction input by an administrator and record the time at which the evaluation instruction is input;

[0049] A row data extraction unit is used to extract row data corresponding to an input moment in the evaluation data matrix;

[0050] A traversal matching unit is used to traverse a preset standard data set based on row data and calculate the matching degree;

[0051] a first output unit, configured to use the matching degree as an evaluation result at the input moment;

[0052] The standard data set is a theoretical data range of all data types at different preset moments.

[0053] As a further solution of the present invention: the model training application module includes:

[0054] A reference image acquisition unit, used to acquire a remote sensing image at the time of improvement start-up as a reference image;

[0055] A real-time image acquisition unit is used to synchronously acquire a remote sensing image when receiving an evaluation instruction input by an administrator, as the remote sensing image at the input moment;

[0056] A sample set construction unit is configured to take the reference image and the remote sensing image at the input moment as an image group, read the evaluation result at the input moment, and construct a sample set; wherein the image group is a feature corresponding to the input, and the evaluation result is a label corresponding to the output;

[0057] An application probability updating unit, used for training an evaluation model based on a sample set, and adjusting the application probability of the evaluation model and the application probability of the analysis process of the evaluation data matrix according to the error rate of the evaluation model;

[0058] Among them, the application probability of the evaluation model is inversely proportional to the error rate, and the application probability of the analysis process of the evaluation data matrix is ​​directly proportional.

[0059] Compared with the existing technology, the beneficial effects of the present invention are: the present invention installs information collection equipment in the cultivated land, remotely obtains soil data, and conducts real-time evaluation of the improvement effect. At the same time, the neural network model based on remote sensing images is trained according to the evaluation results, which further improves the recognition speed and is extremely timely. When the administrator needs it, the evaluation effect at the current moment can be obtained at any time. BRIEF DESCRIPTION OF THE DRAWINGS

[0060] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments of the present invention.

[0061] Figure 1 This is a flowchart of the evaluation method for improving moderately and severely salinized farmland.

[0062] Figure 2 This is a structural diagram of the system for evaluating the improvement effect of moderately and severely salinized farmland. DETAILED DESCRIPTION

[0063] In order to make the technical problems, technical solutions and beneficial effects to be solved by the present invention more clearly understood, the present invention is further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0064] Figure 1 This is a flowchart of a method for evaluating the improvement effect of moderately and severely salinized farmland. In an embodiment of the present invention, a method for evaluating the improvement effect of moderately and severely salinized farmland includes:

[0065] Step S100: receiving the cultivated land range input by the administrator, randomly selecting evaluation points within the cultivated land range, and determining the evaluation level of the evaluation points; wherein the evaluation level is used to determine the data type of the data to be obtained at the evaluation point;

[0066] The technical solution of the present invention is applied to the scenario of farmland improvement, specifically the improvement scenario of moderately to severely salinized farmland. Any improvement process has a range, that is, the range of the farmland that needs to be improved. The range is input by the administrator and is generally a set of boundary coordinates. The input method can be to input a coordinate set and fit the coordinates in the coordinate set to obtain the farmland range. In addition, another method is to directly input a touch screen signal to directly delineate a farmland range. After obtaining the farmland range, some points within the farmland range are selected as evaluation points, and information collection equipment is installed at the evaluation points to collect soil parameters. The type of soil data obtained at different evaluation points is different, which is determined by the evaluation level parameter. The higher the evaluation level, the more types of information collection equipment are installed and the more data types are obtained. In the technical solution of the present invention, soil parameters generally include soil salt content, soil acidity and alkalinity, soil organic matter content, soil texture and soil microbial indicators. In addition, they can also include crop growth parameters, including vegetation coverage and crop physiological parameters. These parameters have corresponding sensors in the existing technology. Some specific examples are as follows:

[0067] Soil salt content:

[0068] Conductivity sensors indirectly measure the salt content of soil solutions by measuring their electrical conductivity. Because salt ions in soil solutions conduct electricity, conductivity is positively correlated with salt content. Commonly used conductivity sensors include four-electrode sensors, which can be buried in the soil for real-time measurements.

[0069] Near-infrared spectral sensors use soil's absorption of near-infrared light to infer salinity. Different salts have specific absorption peaks in the near-infrared band, and salt concentration can be estimated by analyzing the reflected spectrum. This sensor can be mounted on drones or ground vehicles for rapid, large-scale monitoring.

[0070] Soil pH:

[0071] Glass electrode pH sensor: Based on the principle of ion exchange, when a glass electrode comes into contact with a soil solution, hydrogen ions are exchanged on the surface of the glass membrane, generating a potential difference. This potential difference can be measured to calculate the soil pH value. The electrode must be inserted into the soil extract for measurement.

[0072] Fiber optic pH sensor: Utilizing fluorescent materials that are sensitive to pH, when the fluorescent material comes into contact with the soil solution, its fluorescence intensity or wavelength will change with the pH value. By transmitting the light signal through the optical fiber and performing detection and analysis, real-time, in-situ monitoring of soil pH can be achieved.

[0073] Soil organic matter content:

[0074] Sensor based on near-infrared spectroscopy: The organic matter in the soil has a characteristic absorption peak in the near-infrared spectral region. By collecting the near-infrared spectrum of the soil and establishing a corresponding prediction model, the soil organic matter content can be quickly estimated.

[0075] Thermogravimetric analyzers (ex situ, real-time monitoring): In a laboratory, soil samples are heated at a specific temperature program, and the change in sample mass is measured to determine the organic matter content. While not a real-time on-site sensor, they provide highly accurate results and can be used to calibrate other rapid detection methods.

[0076] Soil texture:

[0077] Laser particle size analyzer: This instrument measures the angle and intensity of light scattered by a laser in a suspension of soil particles, calculating the particle size distribution based on Mie scattering theory to determine soil texture. This instrument requires soil samples to be processed and measured in a laboratory.

[0078] Time domain reflectometry (TDR) combined with model: TDR is mainly used to measure soil moisture content, but by establishing an empirical model related to soil texture and combining it with information such as the soil dielectric constant measured by TDR, soil texture can be indirectly estimated.

[0079] Soil microorganism information:

[0080] Microbial sensors: Based on the biochemical reaction between microorganisms and specific substrates, these reactions are converted into detectable electrical or optical signals. For example, specific microorganisms are immobilized on the surface of an electrode. When the target substrate in the soil solution reacts with the microorganisms, changes in the electrode potential or current are caused, thereby enabling the detection of microbial activity or related substrate concentrations.

[0081] It is worth mentioning that the costs of different sensors are different. If you want to improve the fit between the selection process of evaluation points and the actual situation, then when determining the evaluation level of the evaluation point, you can use cost as a parameter, that is, the evaluation level corresponds to the cost. The higher the evaluation level, the higher the cost. Based on the determined cost combination, the information collection equipment is installed at the evaluation point; when combining information collection equipment based on the determined cost, it is necessary to select the combination scheme with the largest number of information collection equipment.

[0082] In addition, regarding the cost of information collection equipment, under normal circumstances, the cost of information collection equipment is fixed, but in this application, the cost of information collection equipment can be adjusted. The cost can be adjusted according to the importance of the information collection equipment. The higher the importance, the lower the cost. For necessary information collection equipment, it can be set to zero cost or negative cost. After adjustment, necessary information collection equipment will be installed at each evaluation point, and more important information collection equipment will be installed.

[0083] Step S200: Based on the evaluation equipment installed at the evaluation point, the evaluation data containing point tags and time tags are obtained, and an evaluation data matrix containing the point tags is constructed; the time tags use relative time, and time zero is the improvement start time; the number of columns in the evaluation data matrix is ​​the same as the total number of data types, and the number of rows corresponds to time;

[0084] An evaluation device is installed at the evaluation point, and soil data is obtained based on the evaluation device. The acquisition frequency is a preset value. In order to simplify the process, the acquisition frequency of all evaluation devices can be set to the same value; a transmission module is set in the evaluation device to upload the obtained data (called evaluation data) to the main terminal. After receiving the evaluation data, the main terminal stores it in the form of a matrix, which is called an evaluation data matrix; the evaluation data includes a point label and a time label. The point label is used to indicate at which evaluation point the data is obtained, and the time label is used to indicate when the data is obtained. The reason for using these two labels is that due to different transmission speeds, the data received by the main terminal is disordered, and the data needs to be integrated according to the two labels. The integration method is to construct an evaluation data matrix; specifically, the time label uses relative time, and time zero is the start time of improvement, that is, the time when the cultivated land improvement starts is time zero; the number of columns of the evaluation data matrix is ​​the same as the total number of data types, each column corresponds to an information collection device, the number of rows corresponds to time, and each row corresponds to a time; the obtained evaluation data matrix is ​​the matrix of each evaluation.

[0085] Step S300: receiving an evaluation instruction input by an administrator, extracting row data from the evaluation data matrix based on the input time of the evaluation instruction, analyzing the extracted row data, and outputting an evaluation result;

[0086] When the administrator needs to evaluate the improvement effect, the administrator inputs the evaluation instruction. The execution subject of this method locates the corresponding row in the evaluation data matrix based on the input time of the evaluation instruction and extracts the row data; analyzes the row data and outputs the evaluation result; the process of analyzing the row data can use the existing data analysis model, or pre-determine some standard data, and compare the extracted row data with the standard data to obtain the evaluation result.

[0087] Step S400: When receiving the evaluation instruction input by the administrator, remote sensing images are synchronously acquired, and the remote sensing image at the time of improvement startup is acquired as the reference image. The reference image and the remote sensing image are used as features, and the evaluation results are used as labels to train the evaluation model; the application probability of the evaluation model and the application probability of the analysis process of the evaluation data matrix are adjusted according to the error rate of the evaluation model.

[0088] In an example of the technical solution of the present invention, a remote sensing-based effect evaluation process is also introduced. A remote sensing image at the moment of improvement startup is obtained as a reference image. When receiving the evaluation instruction input by the administrator, the remote sensing image is obtained synchronously. The remote sensing image and the reference image jointly determine an image group. The difference in the images is the difference generated in the improvement process. It is used as an independent variable, and the evaluation result based on the evaluation data matrix at the same time is read as the dependent variable. A sample set from the independent variable to the dependent variable is constructed, and a neural network model can be trained, which is called an evaluation model. The function of the evaluation model is to obtain the evaluation result based on the remote sensing image.

[0089] The process of determining the evaluation results based on the evaluation data matrix is ​​real-time. The evaluation model simplifies this process through deep learning. The evaluation results determined based on the evaluation data matrix are used as the true value to determine the error rate of the evaluation model. The use frequency of the two evaluation methods can be adjusted according to the error rate. Under the extreme state, the error rate of the evaluation model is zero. At this time, the evaluation model is fully used, the evaluation speed is very fast, and the working pressure of the information collection equipment is very small. There is almost no need to collect and upload data, which increases the service life (most of the information collection equipment in this application is consumable equipment).

[0090] Regarding the specific working conditions of the technical solution of the present invention, for a piece of soil, the administrator will regularly perform some improvement measures. This application can remotely and frequently determine the improvement effect at each moment and provide timely feedback to the administrator.

[0091] Regarding step S100, the steps of receiving the cultivated land range input by the administrator, randomly selecting evaluation points within the cultivated land range, and determining the evaluation levels of the evaluation points include:

[0092] Receive the cultivated land range input by the administrator;

[0093] The cultivated land area is rasterized, and then a grid is inserted into the cultivated land area according to the preset unit size; the left and top lines of the grid are tangent to the cultivated land area;

[0094] Randomly select grid nodes as evaluation points and simultaneously determine the evaluation level of the evaluation points;

[0095] Update the selection probability of each grid node and execute the selection process cyclically;

[0096] The total number of selected evaluation points is recorded, and when the total number reaches a preset threshold, the selected evaluation points and their evaluation levels are output.

[0097] The scope of cultivated land is input by the administrator. After receiving the scope of cultivated land, the scope of cultivated land is rasterized to obtain a bitmap, and a grid is inserted into the rasterized scope of cultivated land. The left and upper lines of the grid are tangent to the scope of cultivated land, and the size of the grid unit is a preset value. The actual meaning of the grid is very simple, which is the grid function in some existing map software; the purpose of limiting the tangency is to predetermine the origin position of the grid.

[0098] After the grid is inserted, a grid node is randomly selected in the grid as an evaluation point, the evaluation level of the evaluation point is determined synchronously, the selection probability of each grid node is updated, and the selection process is executed cyclically; among them, the initial selection probability of each grid node is the same.

[0099] Finally, the total number of selected evaluation points is recorded. When the total number of selected evaluation points is large enough, the selected evaluation points and their evaluation levels are output.

[0100] Specifically, the process of updating the selection probability of each grid node and cyclically executing the selection process includes:

[0101] Where, Represents a grid node The probability of selecting Represents a grid node The eigenvalue at is the total number of grid nodes in the horizontal direction, is the total number of grid nodes in the longitudinal direction; and The values ​​of are all positive integers; ; is the preset correction factor, For the selected The evaluation level of each evaluation point; For the selected Evaluation points and grid nodes distance;

[0102] When the grid node When selected as an evaluation point, Where, For grid nodes The evaluation level of The independent variable is The classification function is The increasing function of .

[0103] The core of the above content is the eigenvalue. For any position, the influence of the selected evaluation point on the position is calculated. The greater the distance from the evaluation point, the smaller the influence. The greater the evaluation level of the evaluation point, the greater the influence. That is, the above content Item; for any position, the influence of all evaluation points on it is superimposed, and a decreasing function is compounded on the influence, that is, the greater the influence on a certain position, the smaller the probability of being selected, thereby obtaining the selection probability of the position; further, the greater the influence on a certain position, the smaller the probability of being selected, but when it is really selected, the corresponding evaluation level will also be smaller, thereby making the evaluation points as evenly distributed as possible, and the distribution of evaluation levels more in line with reality (the larger the evaluation level, the higher the cost).

[0104] Regarding step S300, the steps of receiving the evaluation instruction input by the administrator, extracting row data from the evaluation data matrix based on the input time of the evaluation instruction, analyzing the extracted row data, and outputting the evaluation results include:

[0105] Receive the evaluation instruction input by the administrator and record the input time of the evaluation instruction;

[0106] Extract the row data corresponding to the input time in the evaluation data matrix;

[0107] Traverse the preset standard data set based on row data and calculate the matching degree;

[0108] Taking the matching degree as the evaluation result at the input moment;

[0109] The standard data set is a theoretical data range of all data types at different preset moments.

[0110] The above content defines the specific evaluation process, receiving the evaluation instructions input by the administrator, recording the input time of the evaluation instructions, extracting the row data corresponding to the input time in the evaluation data matrix, traversing the preset standard data set based on the row data, calculating the matching degree, and using the matching degree as the evaluation result at the input time.

[0111] Among them, the standard data set is a collection of standard data ranges for each information collection device. It needs to include all information collection devices. The standard data of each information collection device is not a fixed value, but a range. If the measured data is within the range, it is considered to be standard.

[0112] Each row of data is a data set obtained by the installed information collection device. It is compared with the standard data set at the same moment to determine whether the data of each information collection device is standard, and then a matching degree is output.

[0113] Specifically, the step of traversing a preset standard data set based on row data and calculating the matching degree includes:

[0114] Read the standard data set at the input time;

[0115] Compare the row data extracted from the evaluation data matrix with each row data in the standard row data group and calculate the matching degree;

[0116] In the comparison process, the row data extracted from the evaluation data matrix are padded with columns, and the padded elements are preset default values.

[0117] The above content provides a detailed description of the calculation process of the maximum matching degree. The standard data set at the input moment is read, the row data extracted from the evaluation data matrix is ​​compared with each row data in the standard row data group, the matching degree is calculated, the calculated matching degrees are compared, and the maximum matching degree is output.

[0118] It should be noted that the number of data in the row data is different from the number of data in the standard data set (the row data is the information collection equipment actually installed, and the standard data set is all information collection equipment). Therefore, data filling is required, that is, inserting a data at the missing position in the row data. The inserted data is the default value, such as NULL. When compared, it does not match all ranges. Therefore, it must be non-standard. This also means that the fewer information collection devices installed, the lower the corresponding matching degree. This indirectly raises the evaluation standard for the improvement effect, making it more difficult to achieve excellent results (actually very high, but it will also be evaluated as a medium value), prompting administrators to perform more calibrations. On the contrary, if the missing data is evaluated as standard data, then the evaluation process will have no reference significance (not collecting data means full marks).

[0119] The matching degree calculation process is as follows: for any data in the row data, query the standard data range in the standard data set to determine whether the data belongs to the standard data range, which is one if it does, and zero if it does not; calculate the number of elements that are one and divide it by the total number of elements in the standard data set to obtain the matching degree.

[0120] Regarding step S400, when receiving the evaluation instruction input by the administrator, synchronously acquiring a remote sensing image, acquiring a remote sensing image at the time of improvement startup as a reference image, using the reference image and the remote sensing image as features, and using the evaluation results as labels to train the evaluation model; and adjusting the application probability of the evaluation model and the application probability of the analysis process of the evaluation data matrix according to the error rate of the evaluation model include the following steps:

[0121] Acquire a remote sensing image at the time of improvement startup as a reference image;

[0122] When receiving the evaluation instruction input by the administrator, the remote sensing image is synchronously acquired as the remote sensing image at the input moment;

[0123] The reference image and the remote sensing image at the input moment are taken as an image group, and the evaluation results at the input moment are read to construct a sample set; the image group is the feature corresponding to the input, and the evaluation results are the label corresponding to the output;

[0124] An evaluation model is trained based on the sample set, and the application probability of the evaluation model and the application probability of the analysis process of the evaluation data matrix are adjusted according to the error rate of the evaluation model;

[0125] Among them, the application probability of the evaluation model is inversely proportional to the error rate, and the application probability of the analysis process of the evaluation data matrix is ​​directly proportional.

[0126] In one example of the technical solution of the present invention, a remote sensing image at the moment of improvement startup is obtained as a reference image. When receiving the evaluation instruction input by the administrator, a remote sensing image is synchronously obtained as the remote sensing image at the input moment. The reference image and the remote sensing image at the input moment are taken as an image group, the evaluation result at the input moment is read, a sample set is constructed, and an evaluation model is trained based on the sample set. The application probability of the evaluation model and the application probability of the analysis process of the evaluation data matrix are adjusted according to the error rate of the evaluation model. The application probability of the evaluation model is inversely proportional to the error rate, and the application probability of the analysis process of the evaluation data matrix is ​​directly proportional.

[0127] The above content actually provides an adversarial architecture. When the error rate is small enough, an evaluation model is used to replace the evaluation process based on information collection equipment. The advantage of this method is that data acquisition is more convenient and intuitive. In addition, the evaluation speed is also faster.

[0128] Figure 2 The structure diagram of the system for evaluating the improvement effect of moderately and severely salinized farmland is shown in FIG. In an embodiment of the present invention, a system for evaluating the improvement effect of moderately and severely salinized farmland is provided. The system 10 includes:

[0129] The evaluation point planning module 11 is used to receive the cultivated land range input by the administrator, randomly select evaluation points within the cultivated land range, and determine the evaluation level of the evaluation points; wherein the evaluation level is used to determine the data type of the data to be obtained at the evaluation point;

[0130] The data matrix construction module 12 is used to obtain evaluation data containing point tags and time tags based on the evaluation equipment installed at the evaluation point, and construct an evaluation data matrix containing the point tags; the time tags use relative time, and time zero is the improvement start time; the number of columns in the evaluation data matrix is ​​the same as the total number of data types, and the number of rows corresponds to time;

[0131] The row data analysis module 13 is used to receive the evaluation instruction input by the administrator, extract the row data from the evaluation data matrix based on the input time of the evaluation instruction, analyze the extracted row data, and output the evaluation result;

[0132] The model training application module 14 is used to synchronously obtain remote sensing images when receiving evaluation instructions input by the administrator, obtain the remote sensing image at the time of improvement startup as the reference image, use the reference image and the remote sensing image as features, and use the evaluation results as labels to train the evaluation model; adjust the application probability of the evaluation model and the application probability of the analysis process of the evaluation data matrix according to the error rate of the evaluation model.

[0133] Furthermore, the evaluation point planning module 11 includes:

[0134] A range receiving unit, used for receiving the cultivated land range input by the administrator;

[0135] The grid insertion unit is used to rasterize the cultivated land range and then insert a grid within the cultivated land range according to the preset unit size; the left and top lines of the grid are tangent to the cultivated land range;

[0136] Random selection unit, used to randomly select grid nodes as evaluation points and simultaneously determine the evaluation level of the evaluation points;

[0137] A loop execution unit is used to update the selection probability of each grid node and loop through the selection process;

[0138] The first output unit is used to record the total number of selected evaluation points, and when the total number reaches a preset threshold, output the selected evaluation points and their evaluation levels.

[0139] Specifically, the row data analysis module 13 includes:

[0140] An instruction receiving unit, configured to receive an evaluation instruction input by an administrator and record the time at which the evaluation instruction is input;

[0141] A row data extraction unit is used to extract row data corresponding to an input moment in the evaluation data matrix;

[0142] A traversal matching unit is used to traverse a preset standard data set based on row data and calculate the matching degree;

[0143] a first output unit, configured to use the matching degree as an evaluation result at the input moment;

[0144] The standard data set is a theoretical data range of all data types at different preset moments.

[0145] Furthermore, the model training application module 14 includes:

[0146] A reference image acquisition unit, used to acquire a remote sensing image at the time of improvement start-up as a reference image;

[0147] A real-time image acquisition unit is used to synchronously acquire a remote sensing image when receiving an evaluation instruction input by an administrator, as the remote sensing image at the input moment;

[0148] A sample set construction unit is configured to take the reference image and the remote sensing image at the input moment as an image group, read the evaluation result at the input moment, and construct a sample set; wherein the image group is a feature corresponding to the input, and the evaluation result is a label corresponding to the output;

[0149] An application probability updating unit, used for training an evaluation model based on a sample set, and adjusting the application probability of the evaluation model and the application probability of the analysis process of the evaluation data matrix according to the error rate of the evaluation model;

[0150] Among them, the application probability of the evaluation model is inversely proportional to the error rate, and the application probability of the analysis process of the evaluation data matrix is ​​directly proportional.

[0151] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A method for evaluating the improvement effect of moderately and severely salinized farmland, characterized in that: The method comprises: Receive the cultivated land range input by the administrator, randomly select evaluation points within the cultivated land range, and determine the evaluation level of the evaluation point; wherein the evaluation level is used to determine the data type of the data to be obtained at the evaluation point; Based on the evaluation equipment installed at the evaluation point, the evaluation data containing point tags and time tags are obtained, and an evaluation data matrix containing the point tags is constructed; the time tag adopts relative time, and time zero is the improvement start time; the number of columns in the evaluation data matrix is ​​the same as the total number of data types, and the number of rows corresponds to time; Receive the evaluation instruction input by the administrator, extract row data from the evaluation data matrix based on the input time of the evaluation instruction, analyze the extracted row data, and output the evaluation results; When receiving the evaluation instruction input by the administrator, a remote sensing image is synchronously acquired, and a remote sensing image at the time of improvement startup is acquired as a reference image. The reference image and the remote sensing image are used as features, and the evaluation results are used as labels to train the evaluation model; the application probability of the evaluation model and the application probability of the analysis process of the evaluation data matrix are adjusted according to the error rate of the evaluation model; The steps of receiving the evaluation instruction input by the administrator, extracting row data from the evaluation data matrix based on the input time of the evaluation instruction, analyzing the extracted row data, and outputting the evaluation result include: Receive the evaluation instruction input by the administrator and record the input time of the evaluation instruction; Extract the row data corresponding to the input time in the evaluation data matrix; Read the standard data set at the input time; Compare the row data extracted from the evaluation data matrix with each row data in the standard row data group and calculate the matching degree; In the comparison process, the row data extracted from the evaluation data matrix are padded with columns, and the padded elements are preset default values; Taking the matching degree as the evaluation result at the input moment; The standard data set is a theoretical data range of all data types at different preset moments.

2. The method for evaluating improvement effects of moderately and severely salinized farmland according to claim 1, wherein: The steps of receiving the cultivated land range input by the administrator, randomly selecting evaluation points within the cultivated land range, and determining the evaluation levels of the evaluation points include: Receive the cultivated land range input by the administrator; The cultivated land area is rasterized, and then a grid is inserted into the cultivated land area according to the preset unit size; the left and top lines of the grid are tangent to the cultivated land area; Randomly select grid nodes as evaluation points and simultaneously determine the evaluation level of the evaluation points; Update the selection probability of each grid node and execute the selection process cyclically; The total number of selected evaluation points is recorded, and when the total number reaches a preset threshold, the selected evaluation points and their evaluation levels are output.

3. The method for evaluating improvement effects of moderately and severely salinized farmland according to claim 2, wherein: The process of updating the selection probability of each grid node and cyclically executing the selection process includes: Where, Represents a grid node The probability of selecting Represents a grid node The eigenvalue at is the total number of grid nodes in the horizontal direction, is the total number of grid nodes in the longitudinal direction; and The values ​​of are all positive integers; ; is the preset correction factor, For the selected The evaluation level of each evaluation point; For the selected Evaluation points and grid nodes distance; When the grid node When selected as an evaluation point, Where, For grid nodes The evaluation level of The independent variable is The classification function is The increasing function of .

4. The method for evaluating improvement effects of moderately and severely salinized farmland according to claim 1, wherein: When receiving the evaluation instruction input by the administrator, the remote sensing image is synchronously acquired, the remote sensing image at the time of improvement startup is acquired as the reference image, the reference image and the remote sensing image are used as features, and the evaluation result is used as a label to train the evaluation model; The steps of adjusting the application probability of the evaluation model and the application probability of the analysis process of the evaluation data matrix according to the error rate of the evaluation model include: Acquire a remote sensing image at the time of improvement startup as a reference image; When receiving the evaluation instruction input by the administrator, the remote sensing image is synchronously acquired as the remote sensing image at the input moment; The reference image and the remote sensing image at the input moment are taken as an image group, and the evaluation results at the input moment are read to construct a sample set; the image group is the feature corresponding to the input, and the evaluation results are the label corresponding to the output; An evaluation model is trained based on the sample set, and the application probability of the evaluation model and the application probability of the analysis process of the evaluation data matrix are adjusted according to the error rate of the evaluation model; Among them, the application probability of the evaluation model is inversely proportional to the error rate, and the application probability of the analysis process of the evaluation data matrix is ​​directly proportional.

5. A system for evaluating the improvement effect of moderately and severely salinized farmland, characterized in that: The system comprises: The evaluation point planning module is used to receive the cultivated land range input by the administrator, randomly select evaluation points within the cultivated land range, and determine the evaluation level of the evaluation points; wherein the evaluation level is used to determine the data type of the data to be obtained at the evaluation point; A data matrix construction module is used to obtain evaluation data containing point tags and time tags based on evaluation equipment installed at evaluation points, and to construct an evaluation data matrix containing point tags; the time tags use relative time, with time zero being the improvement start time; the number of columns in the evaluation data matrix is ​​the same as the total number of data types, and the number of rows corresponds to time; The row data analysis module is used to receive the evaluation instruction input by the administrator, extract the row data from the evaluation data matrix based on the input time of the evaluation instruction, analyze the extracted row data, and output the evaluation results; The model training application module is used to synchronously acquire remote sensing images when receiving evaluation instructions input by the administrator, acquire the remote sensing image at the time of improvement startup as the reference image, use the reference image and the remote sensing image as features, and use the evaluation results as labels to train the evaluation model; adjust the application probability of the evaluation model and the application probability of the analysis process of the evaluation data matrix according to the error rate of the evaluation model; The row data analysis module includes: An instruction receiving unit, configured to receive an evaluation instruction input by an administrator and record the time at which the evaluation instruction is input; A row data extraction unit is used to extract row data corresponding to an input moment in the evaluation data matrix; A traversal matching unit is used to traverse a preset standard data set based on row data and calculate the matching degree; a first output unit, configured to use the matching degree as an evaluation result at the input moment; The standard data set is a theoretical data range of all data types at different preset times; The content of calculating the matching degree by traversing the preset standard data set based on row data includes: Read the standard data set at the input time; Compare the row data extracted from the evaluation data matrix with each row data in the standard row data group and calculate the matching degree; In the comparison process, the row data extracted from the evaluation data matrix are padded with columns, and the padded elements are preset default values.

6. The system for evaluating improvement effects of moderately and severely salinized farmland according to claim 5, characterized in that: The evaluation point planning module includes: A range receiving unit, used for receiving the cultivated land range input by the administrator; The grid insertion unit is used to rasterize the cultivated land range and then insert a grid within the cultivated land range according to the preset unit size; the left and top lines of the grid are tangent to the cultivated land range; Random selection unit, used to randomly select grid nodes as evaluation points and simultaneously determine the evaluation level of the evaluation points; A loop execution unit is used to update the selection probability of each grid node and loop through the selection process; The first output unit is used to record the total number of selected evaluation points, and when the total number reaches a preset threshold, output the selected evaluation points and their evaluation levels.

7. The system for evaluating improvement effects of moderately and severely salinized farmland according to claim 5, characterized in that: The model training application module includes: A reference image acquisition unit, used to acquire a remote sensing image at the time of improvement start-up as a reference image; A real-time image acquisition unit is used to synchronously acquire a remote sensing image when receiving an evaluation instruction input by an administrator, as the remote sensing image at the input moment; A sample set construction unit is configured to take the reference image and the remote sensing image at the input moment as an image group, read the evaluation result at the input moment, and construct a sample set; wherein the image group is a feature corresponding to the input, and the evaluation result is a label corresponding to the output; An application probability updating unit, used for training an evaluation model based on a sample set, and adjusting the application probability of the evaluation model and the application probability of the analysis process of the evaluation data matrix according to the error rate of the evaluation model; Among them, the application probability of the evaluation model is inversely proportional to the error rate, and the application probability of the analysis process of the evaluation data matrix is ​​directly proportional.

Images