Markov chain-based planting structure adjustment and prediction method and system

Abstract

The invention belongs to the technical field of agricultural remote sensing, and relates to a Markov chain-based planting structure adjustment and prediction method and system. The method comprises the following steps: enabling each planting structure adjustment mode to correspond to one state of a Markov chain model at a pixel scale to form a complete state space; the time step length of the Markov chain is set, and the time interval of state transition is defined; constructing a Markov chain transfer matrix model; calculating a matrix transition probability; determining a target planting structure adjustment mode; and performing iterative deduction on the initial planting system state by using a matrix multiplication and power operation method, simulating the evolution process of the target planting structure adjustment mode, and predicting the future evolution trend of the target planting structure adjustment mode. According to the method, future planting structure evolution is accurately predicted, the evolution process of crop planting types along with time is reflected, accurate quantitative deduction of future planting structure evolution is realized, multi-step planting structure dynamic prediction can be carried out, and the stability and reliability of a prediction result are improved.

Description

Technical Field

[0001] The present invention belongs to the field of agricultural remote sensing technology, and in particular relates to a planting structure adjustment and prediction method and system based on Markov chains. Background Art

[0002] Accurately identifying the changing trends of planting structures at the regional scale, understanding their spatial evolution paths and making scientific predictions have become core technical issues in the field of agricultural remote sensing and land resource monitoring.

[0003] Most existing methods use crop type difference analysis, which involves comparing the classification results of remote sensing identification from two years before and after to identify the type and scope of changes in cultivated land use.

[0004] The two-period difference method, which fragments continuous processes, cannot accurately reflect the actual patterns of crop structure adjustment. Actual crop structure change typically exhibits multi-year, multi-stage evolutionary characteristics, with significant historical inertia and cumulative effects. This is a long-term, gradual process, which the two-period difference method struggles to capture. Methods that omit phased or gradual adjustment trends and only detect changes in adjacent years are prone to missing phased or gradual adjustment trends within the crop structure adjustment process, failing to capture more representative adjustment paths and resulting in a less comprehensive and in-depth understanding of crop structure change. When multi-period remote sensing data is used to compensate for the shortcomings of a single period, directly combining crop type change paths over multiple years leads to a sharp increase in the number of state combinations, resulting in a large number of low-frequency, atypical change sequences. These sequences are difficult to classify as typical adjustment patterns associated with policy or climate context, which in turn weakens the stability and practical value of pattern recognition, making existing methods inefficient and ineffective when processing multi-period data. Summary of the Invention

[0005] In order to solve the above technical problems, the present invention provides a planting structure adjustment and prediction method and system based on Markov chain.

[0006] In a first aspect, the present invention provides a method for adjusting and predicting planting structure based on a Markov chain, comprising: Each planting structure adjustment mode corresponds to a state of the Markov chain model at the pixel scale, forming a complete state space; the planting structure adjustment mode includes at least a single-cropping planting mode, a double-cropping planting mode and a two-year three-cropping planting mode; Set the time step of the Markov chain to define the time interval between state transitions; Construct a Markov chain transfer matrix model between different planting structure adjustment modes in time and space; the Markov chain transfer matrix model is used to describe the state transition probability of each planting structure adjustment mode within a set time step; Based on the constructed Markov chain transfer matrix model, the matrix transfer probability is calculated; each element in the Markov chain transfer matrix model reflects the probability of state transition between different planting structure adjustment modes; Determine the target planting structure adjustment mode according to the preset relationship between the matrix transfer probability and the planting structure adjustment mode; Based on the Markov chain transfer matrix model, the initial planting system state is iteratively deduced using matrix multiplication and power operation methods to simulate the evolution process of the target planting structure adjustment pattern of the future planting system. Based on the dynamic model evaluation and optimization mechanism, the evolution trend of the future target planting structure adjustment pattern of each planting system is predicted.

[0007] In a second aspect, the present invention provides a planting structure adjustment and prediction system based on a Markov chain, comprising a state space composition unit, a setting unit, a model construction unit, a model processing unit, a first output unit, and a second output unit; The state space forming unit is used to correspond each planting structure adjustment mode to a state of the Markov chain model at the pixel scale, thereby forming a complete state space; the planting structure adjustment mode includes at least a single-cropping planting mode, a double-cropping planting mode, and a two-year three-cropping planting mode; The setting unit is used to set the time step of the Markov chain and define the time interval between state transitions; A model building unit is used to build a Markov chain transfer matrix model between different planting structure adjustment modes in time and space; the Markov chain transfer matrix model is used to describe the state transition probability of each planting structure adjustment mode within a set time step; The model processing unit is used to calculate the matrix transfer probability based on the constructed Markov chain transfer matrix model; each element in the Markov chain transfer matrix model reflects the probability of state transfer between different planting structure adjustment modes; A first output unit is configured to determine a target planting structure adjustment mode according to a preset relationship between the matrix transition probability and the planting structure adjustment mode; The second output unit is used to iteratively deduce the initial planting system state based on the Markov chain transfer matrix model using matrix multiplication and power operation methods, simulate the evolution process of the target planting structure adjustment pattern of the future planting system, and predict the evolution trend of the future target planting structure adjustment pattern of each planting system based on the dynamic model evaluation and optimization mechanism.

[0008] On the basis of the above technical solution, the present invention can also be improved as follows.

[0009] Furthermore, the state space of the Markov chain can be represented as follows: Indicates the total number of planting structure adjustment patterns, Indicates the There are three planting structure adjustment modes, and the state space is , No. The state space corresponding to the planting structure adjustment mode is , No. The state space corresponding to the planting structure adjustment mode is , then the state space of the Markov chain is Expressed as: .

[0010] Furthermore, let the Markov chain transfer matrix model be , the Markov chain transfer matrix is The matrix of The state space corresponding to the planting structure adjustment mode is , No. The state space corresponding to the planting structure adjustment mode is , the matrix transition probability is , matrix transition probability Represents the state space Transfer to state space The probability of time step is , Indicates in Year to From the state space Transfer to state space The number of times, Indicates in Year to State space between years The total number of times, then: .

[0011] Furthermore, based on the preset relationship between the matrix transition probability and the planting structure adjustment pattern, the target planting structure adjustment pattern is determined, including: analyzing the transition probability between each state in the Markov chain transfer matrix model, and screening the planting system types greater than the set threshold based on the transition probability to obtain the target planting structure adjustment pattern.

[0012] Furthermore, the evolution trend of the future target planting structure adjustment pattern of each planting system is predicted, including: Based on the current observation data, the proportion of various planting structure adjustment modes is counted to form the initial state vector and determine the initial state distribution; Perform short-term evolution deduction based on the transfer matrix, and predict the state of the next time step based on the current state through matrix multiplication to obtain the prediction result; The prediction results are compared with the actual observation data to verify the accuracy of the model in short-term prediction.

[0013] Furthermore, the prediction results are compared with the actual observation data, including: using the root mean square error or mean absolute error to evaluate the error level; if the prediction error is greater than the set threshold, the transition probability is updated or the classification of the state space is adjusted.

[0014] Furthermore, the initial planting system state is iteratively deduced using the matrix multiplication and power operation method, including: using the power of the transfer matrix to calculate the distribution of the planting structure adjustment pattern after several time steps in the future to obtain the evolution trend; constructing the distribution state of the planting structure adjustment pattern of the current planting system as the initial state vector, and multiplying the initial state vector with the transfer matrix to obtain the state vector of the next time step. Let the initial state vector be , the transfer matrix is , the state vector for the next time step is , ; Assume the number of time steps is , No. The state vector for each time step is , continue to perform matrix power operations, after The state vector after time steps is .

[0015] Furthermore, based on the dynamic model evaluation and optimization mechanism, including: training and testing the implant system data in different time periods through the cross-validation method, and calculating the error index to verify the Markov chain transfer matrix model.

[0016] Furthermore, different weights are assigned to the matrix transfer probabilities in different time periods, and the Markov chain transfer matrix model is extended by introducing state information of multiple time steps; different weights are assigned to the matrix transfer probabilities according to external environmental characteristics, including climatic conditions and human intervention characteristics; climatic conditions include rainfall and temperature; and human intervention characteristics include irrigation water volume.

[0017] The beneficial effects of the present invention are: (1) This paper uses Markov chain analysis to capture the gradual evolution of cropping systems by relying on multi-time series data. By integrating years of cropping system change data, an accurate transition probability matrix model is constructed, which can reflect the long-term trend of crop type changes, accurately predict the future evolution of cropping structures, and reflect the evolution of crop planting types over time. (2) The present invention achieves accurate quantitative deduction of the future evolution of planting structure by constructing a Markov chain transfer matrix model of the changes in planting system over many years; at the same time, by combining matrix multiplication and power operation methods, it can perform multi-step dynamic prediction of planting structure, which can not only accurately characterize short-term changes but also effectively infer long-term trends; in addition, the model evaluation and dynamic optimization mechanism are introduced in the prediction process to further improve the stability and reliability of the prediction results. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Figure 1 Schematic diagram of the planting structure adjustment and prediction method based on Markov chain provided in Example 1 of the present invention; Figure 2 This is a principle block diagram of the Markov chain-based planting structure adjustment and prediction system provided in Example 2 of the present invention. DETAILED DESCRIPTION

[0019] To make the objectives, technical solutions, and advantages of the embodiments of the present invention more clear, the technical solutions of the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only 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 herein can be arranged and designed in various different configurations.

[0020] Example 1 As an example, as shown in the attached Figure 1 As shown, to solve the above technical problems, this embodiment provides a planting structure adjustment and prediction method based on Markov chain, including: Each planting structure adjustment mode corresponds to a state of the Markov chain model at the pixel scale, forming a complete state space; the planting structure adjustment mode includes at least a single-cropping planting mode, a double-cropping planting mode and a two-year three-cropping planting mode; Set the time step of the Markov chain to define the time interval between state transitions; Construct a Markov chain transfer matrix model between different planting structure adjustment modes in time and space; the Markov chain transfer matrix model is used to describe the state transition probability of each planting structure adjustment mode within a set time step; Based on the constructed Markov chain transfer matrix model, the matrix transfer probability is calculated; each element in the Markov chain transfer matrix model reflects the probability of state transition between different planting structure adjustment modes; Determine the target planting structure adjustment mode according to the preset relationship between the matrix transfer probability and the planting structure adjustment mode; Based on the Markov chain transfer matrix model, the initial planting system state is iteratively deduced using matrix multiplication and power operation methods to simulate the evolution process of the target planting structure adjustment pattern of the future planting system. Based on the dynamic model evaluation and optimization mechanism, the evolution trend of the future target planting structure adjustment pattern of each planting system is predicted.

[0021] The Markov chain transfer matrix model is obtained based on the statistics of multi-period remote sensing classification data. Each element in the matrix model reflects the probability relationship of the transition from one planting system to another planting system. The model can characterize the dynamic evolution process of planting structure in the time dimension, providing basic support for the identification of subsequent planting structure adjustment patterns and the prediction of future evolution trends.

[0022] Traditional agricultural planting structure adjustment methods usually only rely on data from a single time point or relatively simple models, which makes it difficult to fully consider the long-term evolution process of the planting structure adjustment pattern of the planting system. The present invention adopts the core advantages of Markov chain analysis and relies on multi-time series data to capture the gradual evolution of the planting structure adjustment pattern. By integrating years of planting system change data, an accurate transition probability matrix model is constructed, which can reflect the long-term trend of crop type changes, rather than simply relying on local differences in adjacent years. This can more accurately predict the future evolution of planting structure and reflect the evolution of crop planting types over time.

[0023] The present invention achieves quantitative deduction of the future evolution of planting structures by constructing a Markov chain transfer matrix model of the changes in planting systems over many years. Traditional methods often find it difficult to effectively capture the complex and dynamic change relationships between different planting systems, resulting in unstable or large deviations in the prediction results. The present invention utilizes the Markov chain's characteristic of no aftereffect, that is, the transfer of the planting structure adjustment mode is only related to the current state, and the historical state will not have a direct impact on the future. It only relies on the current state and its transfer rules to accurately simulate the future evolution process; at the same time, combined with matrix multiplication and power operation methods, it can perform multi-step dynamic prediction of planting structures, which can not only accurately characterize short-term changes, but also effectively infer long-term trends; in addition, the introduction of model evaluation and dynamic optimization mechanisms in the prediction process further improves the stability and reliability of the prediction results.

[0024] Specifically, a cropping structure adjustment model refers to an agricultural production model that rotates or intercrops crops over a specific period of time, based on climate, soil conditions, and crop growth cycles. These include, but are not limited to, single-cropping (growing only one crop per year), double-cropping (growing two different crops per year, such as corn in the spring and winter wheat in the fall), and triple-cropping (growing three different crops over a two-year period).

[0025] Alternatively, the state space of the Markov chain can be represented as follows: Indicates the total number of planting structure adjustment patterns, Indicates the There are three planting structure adjustment modes, and the state space is , No. The state space corresponding to the planting structure adjustment mode is , No. The state space corresponding to the planting structure adjustment mode is , then the state space of the Markov chain is Expressed as: .

[0026] The time step is the time interval between the state transitions of the Markov chain during the study. Assume that the time step is set to Year, indicating that from Year to The state transition process between years. The Markov chain transfer matrix model refers to the transfer matrix from one state space (i.e., the planting structure adjustment pattern) to another state space.

[0027] Optionally, let the Markov chain transfer matrix model be , the Markov chain transfer matrix is The matrix of The state space corresponding to the planting structure adjustment mode is , No. The state space corresponding to the planting structure adjustment mode is , the matrix transition probability is , matrix transition probability Represents the state space Transfer to state space The probability of time step is , Indicates in Year to From the state space Transfer to state space The number of times, Indicates in Year to State space between years The total number of times, then: .

[0028] Optionally, predict the evolution trend of the future target planting structure adjustment pattern of each planting system, including: Based on the current observation data, the proportion of various planting structure adjustment modes is counted to form the initial state vector and determine the initial state distribution; Perform short-term evolution deduction based on the transfer matrix, and predict the state of the next time step based on the current state through matrix multiplication to obtain the prediction result; The prediction results are compared with the actual observation data to verify the accuracy of the model in short-term prediction.

[0029] Crop structure adjustment patterns can characterize the transition patterns and trends between different crop structure adjustment patterns over a specific time scale. They describe the directionality, frequency, and stability of transitions between cropping patterns over different time periods, revealing the evolutionary trends of crop structure. For example, in a region that was originally dominated by a single-cropping pattern, some areas gradually shifted to a double-cropping pattern due to advances in agricultural technology and changes in market demand. By constructing a Markov chain transition matrix, we can quantify the probability of this transition from a single-cropping pattern to a double-cropping pattern, thereby extracting the evolutionary patterns and trends of crop structure adjustment in that region.

[0030] Optionally, by analyzing the transition probabilities between states in the Markov chain transfer matrix model, the target planting structure adjustment pattern is obtained by screening the planting structure adjustment patterns that are greater than a set threshold based on the transition probabilities.

[0031] Taking a study area as an example, an analysis of changes in cropping structure adjustment patterns between 2005 and 2010 revealed three types of cropping structure adjustment patterns: single-season corn, winter wheat and corn rotation, and fallow land. The constructed Markov transition matrix showed that the transition probability from single-season corn to winter wheat and corn rotation was 0.40, and the transition probability from fallow land to winter wheat and corn rotation was 0.30, both exceeding the set threshold of 0.30. This indicates significant trends in cropping structure adjustment patterns in the region, from single-season corn to rotation, and from fallow land to rotation.

[0032] Optionally, predict the evolution trend of the future target planting structure adjustment pattern of each planting system, including: Based on the current observation data, the proportion of various planting structure adjustment modes is counted to form the initial state vector and determine the initial state distribution; Perform short-term evolution deduction based on the transfer matrix, and predict the state of the next time step based on the current state through matrix multiplication to obtain the prediction result; The prediction results are compared with the actual observation data to verify the accuracy of the model in short-term prediction.

[0033] Specifically, matrix multiplication is a method of multiplying and summing the corresponding elements of the rows and columns of the matrix.

[0034] Optionally, the prediction results are compared with the actual observation data, including: using the root mean square error or mean absolute error to evaluate the error level; if the prediction error is greater than a set threshold, updating the transition probability or adjusting the classification of the state space.

[0035] Optionally, the initial planting system state is iteratively deduced using a matrix multiplication and power operation method, including: using the power of the transfer matrix to calculate the distribution of the planting structure adjustment pattern after several time steps in the future to obtain the evolution trend; constructing the distribution state of the planting structure adjustment pattern of the current planting system as the initial state vector, multiplying the initial state vector with the transfer matrix to obtain the state vector of the next time step, and assuming that the initial state vector is , the transfer matrix is , the state vector for the next time step is , ; Assume the number of time steps is , No. The state vector for each time step is , continue to perform matrix power operations, after The state vector after time steps is After completing the short-term verification and optimization of a single step, a multi-step long-term deduction is carried out. The power of the transfer matrix is ​​used to calculate the distribution of the planting structure adjustment pattern after several future time steps, and finally the long-term evolution trend is obtained. At the same time, the steady-state distribution of the Markov chain is derived to analyze the stable planting structure adjustment pattern that the planting system may achieve after long-term evolution.

[0036] Specifically, the current crop structure adjustment pattern is first constructed as an initial state vector. This is then multiplied by the Markov chain transition matrix to obtain the predicted state for the next time step. Multi-step predictions can be achieved by applying exponentiation to the transition matrix. Each self-multiplication of the matrix represents the state transition process within a time step. Through continuous matrix multiplication and exponentiation, the evolution trend of the crop structure over time can be iteratively deduced.

[0037] Optionally, based on a dynamic model evaluation and optimization mechanism, the method includes: training and testing the implant system data of different time periods through a cross-validation method, and calculating error indicators to verify the Markov chain transfer matrix model.

[0038] Optionally, different weights are assigned to the matrix transition probabilities according to the features in different time periods, and the Markov chain transition matrix model is extended by introducing state information of multiple time steps.

[0039] Sensitivity analysis verifies the stability of the Markov chain transition matrix model to changes in initial conditions and transition probabilities. Based on the results of error assessment and sensitivity analysis, model parameters are dynamically optimized, including adjusting transition probabilities, introducing time-varying weights, or employing multi-order Markov chain augmented modeling. Optionally, external auxiliary variables (such as climate factors and policy changes) can be introduced to further improve model adaptability and prediction accuracy. Sensitivity analysis involves systematically adjusting the initial conditions and transition probabilities of the Markov chain transition matrix model to observe their impact on prediction results and assess which factors are most sensitive to these changes, thereby guiding parameter optimization. Adjusting transition probabilities dynamically modifies the transition probabilities in the transition matrix based on observed changes or external environmental factors (such as climate change and policy adjustments). Specifically, the transition frequencies between different crop structure adjustment patterns are adjusted to better reflect actual conditions. Introducing time-varying weights assigns different weights to the transition probabilities of each time step based on the impact of different time periods, thereby emphasizing the influence of key time periods (such as policy releases or seasonal changes) and ensuring that these periods have a more prominent impact on the Markov chain transition matrix model's predictions. Enhanced modeling using multi-order Markov chains involves converting the Markov chain transfer matrix model into a multi-order Markov chain by introducing historical information over a longer time step, expanding the state space and enhancing the model's ability to capture long-term trends. Introducing external auxiliary variables involves adjusting model parameters, when necessary, based on changes in external variables such as climate factors (such as temperature and precipitation) or policy changes (such as agricultural subsidies and land use policies), further improving the model's adaptability and forecasting accuracy.

[0040] Example 2 Based on the same principle as the method shown in Example 1 of the present invention, as shown in the attached Figure 2 As shown, an embodiment of the present invention further provides a planting structure adjustment and prediction system based on a Markov chain, comprising a state space composition unit, a setting unit, a model construction unit, a model processing unit, a first output unit, and a second output unit; The state space forming unit is used to correspond each planting structure adjustment mode to a state of the Markov chain model at the pixel scale, thereby forming a complete state space; the planting structure adjustment mode includes at least a single-cropping planting mode, a double-cropping planting mode, and a two-year three-cropping planting mode; The setting unit is used to set the time step of the Markov chain and define the time interval between state transitions; A model building unit is used to build a Markov chain transfer matrix model between different planting structure adjustment modes in time and space; the Markov chain transfer matrix model is used to describe the state transition probability of each planting structure adjustment mode within a set time step; The model processing unit is used to calculate the matrix transfer probability based on the constructed Markov chain transfer matrix model; each element in the Markov chain transfer matrix model reflects the probability of state transfer between different planting structure adjustment modes; A first output unit is configured to determine a target planting structure adjustment mode according to a preset relationship between the matrix transition probability and the planting structure adjustment mode; The second output unit is used to iteratively deduce the initial planting system state based on the Markov chain transfer matrix model using matrix multiplication and power operation methods, simulate the evolution process of the target planting structure adjustment pattern of the future planting system, and predict the evolution trend of the future target planting structure adjustment pattern of each planting system based on the dynamic model evaluation and optimization mechanism.

[0041] Alternatively, the state space of the Markov chain can be represented as follows: Indicates the total number of planting structure adjustment patterns, Indicates the There are three planting structure adjustment modes, and the state space is , No. The state space corresponding to the planting structure adjustment mode is , No. The state space corresponding to the planting structure adjustment mode is , then the state space of the Markov chain is Expressed as: .

[0042] Optionally, let the Markov chain transfer matrix model be , the Markov chain transfer matrix is The matrix of The state space corresponding to the planting structure adjustment mode is , No. The state space corresponding to the planting structure adjustment mode is , the matrix transition probability is , matrix transition probability Represents the state space Transfer to state space The probability of time step is , Indicates in Year to From the state space Transfer to state space The number of times, Indicates in Year to State space between years The total number of times, then: .

[0043] Optionally, the target planting structure adjustment pattern is determined based on a preset relationship between the matrix transition probability and the planting structure adjustment pattern, including: analyzing the transition probability between each state in the Markov chain transfer matrix model, and screening the planting system types greater than a set threshold based on the transition probability to obtain the target planting structure adjustment pattern.

[0044] Optionally, predict the evolution trend of the future target planting structure adjustment pattern of each planting system, including: Based on the current observation data, the proportion of various planting structure adjustment modes is counted to form the initial state vector and determine the initial state distribution; Perform short-term evolution deduction based on the transfer matrix, and predict the state of the next time step based on the current state through matrix multiplication to obtain the prediction result; The prediction results are compared with the actual observation data to verify the accuracy of the model in short-term prediction.

[0045] Optionally, the prediction results are compared with the actual observation data, including: using the root mean square error or mean absolute error to evaluate the error level; if the prediction error is greater than a set threshold, updating the transition probability or adjusting the classification of the state space.

[0046] Optionally, the initial planting system state is iteratively deduced using a matrix multiplication and power operation method, including: using the power of the transfer matrix to calculate the distribution of the planting structure adjustment pattern after several time steps in the future to obtain the evolution trend; constructing the distribution state of the planting structure adjustment pattern of the current planting system as the initial state vector, multiplying the initial state vector with the transfer matrix to obtain the state vector of the next time step, and assuming that the initial state vector is , the transfer matrix is , the state vector for the next time step is , ; Assume the number of time steps is , No. The state vector for each time step is , continue to perform matrix power operations, after The state vector after time steps is .

[0047] Optionally, based on a dynamic model evaluation and optimization mechanism, the method includes: training and testing the implant system data of different time periods through a cross-validation method, and calculating error indicators to verify the Markov chain transfer matrix model.

[0048] Optionally, different weights are assigned to the matrix transfer probabilities in different time periods, and the Markov chain transfer matrix model is expanded by introducing state information of multiple time steps; different weights are assigned to the matrix transfer probabilities according to external environmental characteristics including climate condition characteristics and human intervention characteristics; climate condition characteristics include rainfall and temperature; and human intervention characteristics include irrigation water volume.

[0049] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention. Those skilled in the art will readily appreciate that various modifications and variations of the present invention are possible. Any modifications, equivalent substitutions, or improvements made within the spirit and principles of the present invention shall be included within the scope of protection of the present invention.

Claims

1. The planting structure adjustment and prediction method based on Markov chain is characterized by: include: Each planting structure adjustment mode corresponds to a state of the Markov chain model at the pixel scale, forming a complete state space; The cropping structure adjustment model includes at least a single cropping model, a double cropping model, and a two-year three-cropping model; Set the time step of the Markov chain to define the time interval between state transitions; Construct a Markov chain transfer matrix model between different planting structure adjustment modes in time and space; the Markov chain transfer matrix model is used to describe the state transition probability of each planting structure adjustment mode within a set time step; Based on the constructed Markov chain transfer matrix model, the matrix transfer probability is calculated; Each element in the Markov chain transition matrix model reflects the probability of state transition between different planting structure adjustment modes; Determine the target planting structure adjustment mode according to the preset relationship between the matrix transfer probability and the planting structure adjustment mode; Based on the Markov chain transfer matrix model, the initial planting system state is iteratively deduced using matrix multiplication and power operation methods to simulate the evolution process of the target planting structure adjustment pattern of the future planting system. Based on the dynamic model evaluation and optimization mechanism, the evolution trend of the future target planting structure adjustment pattern of each planting system is predicted.

2. The Markov chain-based planting structure adjustment and prediction method according to claim 1, characterized in that: The state space representation of the Markov chain is: Indicates the total number of planting structure adjustment patterns, Indicates the There are three planting structure adjustment modes, and the state space is , No. The state space corresponding to the planting structure adjustment mode is , No. The state space corresponding to the planting structure adjustment mode is , then the state space of the Markov chain is Expressed as: .

3. The Markov chain-based planting structure adjustment and prediction method according to claim 1, characterized in that: Assume that the Markov chain transfer matrix model is , the Markov chain transfer matrix is The matrix of The state space corresponding to the planting structure adjustment mode is , No. The state space corresponding to the planting structure adjustment mode is , the matrix transition probability is , matrix transition probability Represents the state space Transfer to state space The probability of time step is , Indicates in Year to From the state space Transfer to state space The number of times, Indicates in Year to State space between years The total number of times, then: 。 4. The Markov chain-based planting structure adjustment and prediction method according to claim 1, characterized in that: According to the preset relationship between the matrix transition probability and the planting structure adjustment pattern, the target planting structure adjustment pattern is determined, including: analyzing the transition probability between each state in the Markov chain transfer matrix model, and screening the planting structure adjustment pattern greater than the set threshold based on the transition probability to obtain the target planting structure adjustment pattern.

5. The Markov chain-based planting structure adjustment and prediction method according to claim 1, characterized in that: Predict the evolution trend of future target planting structure adjustment patterns for each planting system, including: Based on the current observation data, the proportion of various planting structure adjustment modes is counted to form the initial state vector and determine the initial state distribution; Perform short-term evolution deduction based on the transfer matrix, and predict the state of the next time step based on the current state through matrix multiplication to obtain the prediction result; The prediction results are compared with the actual observation data to verify the accuracy of the model in short-term prediction.

6. The Markov chain-based planting structure adjustment and prediction method according to claim 5, characterized in that: Compare the prediction results with the actual observation data, including: using the root mean square error or mean absolute error to evaluate the error level; if the prediction error is greater than the set threshold, update the transition probability or adjust the classification of the state space.

7. The Markov chain-based planting structure adjustment and prediction method according to claim 1, characterized in that: The initial planting system state is iteratively deduced using the matrix multiplication and power operation method, including: using the power of the transfer matrix to calculate the distribution of the planting structure adjustment pattern after several time steps in the future to obtain the evolution trend; constructing the distribution state of the planting structure adjustment pattern of the current planting system as the initial state vector, multiplying the initial state vector with the transfer matrix to obtain the state vector of the next time step, and setting the initial state vector as , the transfer matrix is , the state vector for the next time step is , ; Assume the number of time steps is , No. The state vector for each time step is , continue to perform matrix power operations, after The state vector after time steps is .

8. The Markov chain-based planting structure adjustment and prediction method according to claim 1, characterized in that: Based on the dynamic model evaluation and optimization mechanism, it includes: training and testing the implant system data of different time periods through the cross-validation method, and calculating the error index to verify the Markov chain transfer matrix model.

9. The Markov chain-based planting structure adjustment and prediction method according to claim 1, characterized in that: Different weights are assigned to the matrix transfer probabilities in different time periods, and the Markov chain transfer matrix model is extended by introducing state information of multiple time steps. Different weights are assigned to the matrix transfer probabilities according to external environmental characteristics, including climatic conditions and human intervention characteristics. Climate conditions include rainfall and temperature, and human intervention characteristics include irrigation water volume.

10. The Markov chain-based planting structure adjustment and prediction system is characterized by: It includes a state space forming unit, a setting unit, a model building unit, a model processing unit, a first output unit and a second output unit; The state space forming unit is used to correspond each planting structure adjustment mode to a state of the Markov chain model at the pixel scale, thereby forming a complete state space; the planting structure adjustment mode includes at least a single-cropping planting mode, a double-cropping planting mode, and a two-year three-cropping planting mode; The setting unit is used to set the time step of the Markov chain and define the time interval between state transitions; A model building unit is used to build a Markov chain transfer matrix model between different planting structure adjustment modes in time and space; the Markov chain transfer matrix model is used to describe the state transition probability of each planting structure adjustment mode within a set time step; A model processing unit, for calculating matrix transfer probability based on the constructed Markov chain transfer matrix model; Each element in the Markov chain transition matrix model reflects the probability of state transition between different planting structure adjustment modes; A first output unit is configured to determine a target planting structure adjustment mode according to a preset relationship between the matrix transition probability and the planting structure adjustment mode; The second output unit is used to iteratively deduce the initial planting system state based on the Markov chain transfer matrix model using matrix multiplication and power operation methods, simulate the evolution process of the target planting structure adjustment pattern of the future planting system, and predict the evolution trend of the future target planting structure adjustment pattern of each planting system based on the dynamic model evaluation and optimization mechanism.

Images