Incremental PID (Proportion Integration Differentiation) control-based commercial crop remote sensing data complementation method and device

Abstract

The invention discloses an incremental PID control-based commercial crop remote sensing data completion method and device. The method specifically comprises the following steps of S1, acquiring commercial crop remote sensing data from a server and storing the commercial crop remote sensing data; s2, constructing an economic crop remote sensing data matrix according to the economic crop remote sensing data; s3, initializing an incremental PID controller according to the constructed commercial crop remote sensing data matrix; s4, constructing a target loss function according to the commercial crop remote sensing data matrix and an incremental PID controller, performing iterative training, and extracting a hidden factor matrix; and S5, according to the extracted implicit factor matrix, calculating a commercial crop remote sensing data complementation value and storing the commercial crop remote sensing data complementation value.

Description

Technical Field

[0001] The present invention relates to the technical field of economic crops, and in particular to an economic crop remote sensing data completion method and device based on incremental PID control. Background Art

[0002] With the advancement of remote sensing technology and agricultural digitalization, remote sensing monitoring of cash crops has become a key tool for agricultural informatization. Analysis of remote sensing data can support various agricultural decision-making processes, including growth assessment, yield verification, and pest and disease monitoring. However, due to limitations in remote sensing data acquisition, such as sensor coverage, cloud cover, and imaging frequency, significant amounts of missing or incomplete data are often present, significantly impacting the spatiotemporal continuity of the data and the accuracy of subsequent analysis. Therefore, effectively completing cash crop remote sensing data and improving its integrity and usability have become key issues in current agricultural remote sensing research.

[0003] Existing remote sensing data completion methods often rely on techniques such as interpolation, spatiotemporal modeling, or deep learning. However, these methods often suffer from high computational costs, weak generalization, and sensitivity to missing patterns when working with high-dimensional remote sensing observations. Furthermore, with the expansion of monitoring areas and the increasing diversity of remote sensing dimensions, cash crop remote sensing data exhibits high-dimensional and sparse characteristics. This poses significant challenges to existing methods in terms of computational resources and time costs, limiting their application in large-scale remote sensing completion scenarios. Therefore, how to efficiently and accurately complete cash crop remote sensing data has become a critical issue that needs to be addressed. Summary of the Invention

[0004] Aiming at the problem of low accuracy in the existing technology for completing remote sensing data of economic crops, the present invention proposes a method for completing remote sensing data of economic crops based on incremental PID control.

[0005] In order to achieve the above object, the present invention provides the following technical solutions:

[0006] The economic crop remote sensing data completion method based on incremental PID control specifically includes the following steps:

[0007] S1: collects and stores economic crop remote sensing data from the server;

[0008] S2: construct cash crop remote sensing data matrix based on cash crop remote sensing data;

[0009] S3: Initialize the incremental PID controller based on the constructed economic crop remote sensing data matrix;

[0010] S4: Construct the target loss function based on the economic crop remote sensing data matrix and the incremental PID controller and perform iterative training to extract the latent factor matrix;

[0011] S5: Based on the extracted latent factor matrix, the completion value of economic crop remote sensing data is calculated and stored.

[0012] Preferably, in S1, the received economic crop remote sensing data is stored in the form of a triple; the triple is expressed as Z = (x, y, c), where x represents the column index of the pixel in the economic crop remote sensing data, corresponding to the horizontal coordinate in space; y represents the row index of the pixel in the economic crop remote sensing data, corresponding to the vertical coordinate in space; c represents the observation value of the pixel in a specific band.

[0013] Preferably, in said S2, all triples Z = (x, y, c) are constructed into a second-order economic crop remote sensing data matrix R of size I × J I×J , where I is the total number of horizontal indices of pixels in the remote sensing data, J is the total number of vertical indices of pixels in the remote sensing data, and the element r in the matrix ij It represents the observation value of the pixel located in the i-th column and the j-th row in a specific band, 1≤i≤I, 1≤j≤J, and Λ is the set of known observation values in the economic crop remote sensing data matrix R.

[0014] Preferably, in S3, the total number of incremental PID controllers is |Λ|, where |Λ| represents the number of known observation values in the economic crop remote sensing data matrix R, that is, each incremental PID controller corresponds to each known observation value in the economic crop remote sensing data matrix R.

[0015] Preferably, the S4 includes:

[0016] S4-1: Initialize the process parameters involved in the completion of economic crop remote sensing data;

[0017] S4-2: Construct target loss function based on cash crop remote sensing data and cash crop remote sensing data matrix;

[0018] S4-3: Iteratively optimize the target loss function ε using the standard stochastic gradient descent algorithm and the incremental PID controller;

[0019] S4-4: Determine whether the loss function ε has reached the convergence condition in the economic crop remote sensing data matrix R. If so, terminate the training and output the first latent factor matrix M and the second latent factor matrix N; otherwise, continue the iteration.

[0020] Preferably, in S4-1, the process parameters include latent feature matrices M and N; latent feature dimension F; convergence termination threshold δ; learning step length γ; regularization coefficient λ; wherein,

[0021] The latent feature dimension F determines the latent feature dimension of the latent factor matrices M and N and is initialized to a positive integer;

[0022] The sizes of the latent feature matrices M and N are determined by the dimension values of each order of the corresponding economic crop remote sensing data matrix R, that is, M is a latent factor matrix with I rows and F columns, and N is a latent factor matrix with J rows and F columns. The two latent factor matrices are initialized with small random positive numbers respectively.

[0023] The convergence termination threshold δ is a parameter used to determine whether the iterative process has converged, and is initialized with a very small positive number;

[0024] The regularization coefficient λ is a constant that controls the regularization effect of the related elements of the latent factor matrices M and N during training and is initialized to a small positive number.

[0025] Preferably, in S4-2, the target loss function is expressed by the following formula:

[0026]

[0027] In formula (1), ε represents the target loss function; M and N represent the first latent factor matrix and the second latent factor matrix after decomposition of the economic crop remote sensing data matrix R; r ij represents the observed value of the pixel located in the i-th column and the j-th row in the economic crop remote sensing data matrix R in a specific band; Λ is the set of known observed values in the economic crop remote sensing data matrix R; m i is the i-th row vector in the first latent factor matrix M, n j is the j-th row vector in the second latent factor matrix N; <·,·> represents the inner product of two vectors; ||·||2 represents the L2 norm of a vector; λ represents the regularization coefficient of the latent factor matrix; Indicates the completion value of economic crop remote sensing data.

[0028] Preferably, in S4-3, the iterative optimization method is:

[0029] First, the stochastic gradient descent algorithm is used for optimization. The optimization formula is as follows:

[0030]

[0031] In formula (2), m i is the i-th row vector in the first latent factor matrix M, n j is the j-th row vector in the second latent factor matrix N; Indicates that ε is in (m i ,n j ) on m i The gradient, Indicates that ε is in (m i ,nj ) on n j The gradient of ;γ is the learning step size;

[0032] Then, the incremental PID controller is used to calculate the value of each observation r ij The instantaneous error e ij To control the learning step size γ in the above optimization formula, where the instantaneous error e ij =r ij – <m i ,n j >, the incremental PID controller calculates the learning step increment as follows:

[0033]

[0034] In formula (3), is the incremental PID controller for the observed value r ij The learning step increment calculated at the tth iteration; K P is the proportional gain in the incremental PID controller, K I is the integral gain in the incremental PID controller, K D is the derivative gain in the incremental PID controller; is the observation value r at the tth iteration ij The instantaneous error, is the observation value r at the t-1th iteration ij The instantaneous error, is the observation value r at the t-2th iteration ij The instantaneous error of

[0035] Then the observed value r ij The learning step size at the tth iteration is:

[0036]

[0037] In formula (4), is the observed value r ij The learning step size at iteration t; is the observed value r ij The learning step size at the t-1th iteration; is the incremental PID controller for the observed value r ij The learning step increment calculated at iteration t.

[0038] Preferably, in said S5, the economic crop remote sensing data complement value The calculation formula is:

[0039]

[0040] In formula (5), represents the economic crop remote sensing data completion value, that is, the economic crop remote sensing data completion value of the pixel in the i-th column and the j-th row; m i is the i-th row vector in the first latent factor matrix M, n j is the j-th row vector in the second latent factor matrix N.

[0041] The present invention also provides an economic crop remote sensing data completion device based on incremental PID control, comprising a data acquisition module, a storage module, a matrix construction module, an incremental PID control module, a latent factor extraction module and an output module;

[0042] The data acquisition module is used to obtain economic crop remote sensing data from the server and store it;

[0043] A storage module, used for storing received economic crop remote sensing data and economic crop remote sensing data complement values;

[0044] The matrix construction module is used to construct the economic crop remote sensing data matrix R based on the economic crop remote sensing data;

[0045] The incremental PID control module is used to initialize the number of incremental PID controllers according to the economic crop remote sensing data matrix R;

[0046] The latent factor extraction module is used to construct the target loss function based on the economic crop remote sensing data matrix and the initialized incremental PID controller, and extract the latent factor matrix;

[0047] The output module is used to output the economic crop remote sensing data completion value according to the extracted latent factor matrix.

[0048] In summary, due to the adoption of the above technical solution, compared with the prior art, the present invention has at least the following beneficial effects:

[0049] The present invention provides a method and device for completing remote sensing data of economic crops based on incremental PID control. The method and device are specifically used for remote sensing data of economic crops, and can complete missing or abnormal remote sensing observation values with high precision and good temporal and spatial consistency, thereby improving the integrity and availability of remote sensing data, thereby providing reliable data support for economic crop growth monitoring, pest and disease control, yield estimation, etc., and can be widely used in economic crop growth monitoring, smart agriculture and other fields. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] Figure 1 Schematic diagram of a method for completing economic crop data based on incremental PID control according to an exemplary embodiment of the present invention.

[0051] Figure 2 Schematic diagram of an economic crop data completion device based on incremental PID control according to an exemplary embodiment of the present invention.

[0052] Figure 3 FIG. 4 is a schematic diagram of a storage module according to an exemplary embodiment of the present invention.

[0053] Figure 4 FIG. 4 is a schematic diagram of a latent factor extraction module according to an exemplary embodiment of the present invention. DETAILED DESCRIPTION

[0054] The present invention will be further described in detail below with reference to the examples and specific implementation methods. However, this should not be understood as limiting the scope of the present invention to the following examples, as all technologies implemented based on the present invention fall within the scope of the present invention.

[0055] In the description of the present invention, it should be understood that the terms "longitudinal", "transverse", "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside", "outside", etc., indicating the orientation or position relationship, are based on the orientation or position relationship shown in the accompanying drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as limiting the present invention.

[0056] like Figure 1 As shown, the present invention provides a method and device for completing economic crop data based on incremental PID control, which specifically includes the following steps:

[0057] S1: Receive the economic crop remote sensing data completion instruction sent by the server, collect the economic crop remote sensing data from the server and store it.

[0058] The server may send instructions in the following manner: regularly, according to the notification of the device, or according to the notification of a server, to send instructions requiring the completion of economic crop remote sensing data to the data receiving module.

[0059] In this embodiment, the received economic crop remote sensing data is stored in the form of a triple; the triple is expressed as Z = (x, y, c), where x represents the column index of the pixel in the economic crop remote sensing data, corresponding to the horizontal coordinate in space; y represents the row index of the pixel in the economic crop remote sensing data, corresponding to the vertical coordinate in space; c represents the observation value of the pixel in a specific band.

[0060] S2: Construct the economic crop remote sensing data matrix R based on the economic crop remote sensing data.

[0061] In this embodiment, all triples Z = (x, y, c) are constructed into a second-order economic crop remote sensing data matrix R of size I × J I×J, where I is the total number of horizontal indices of pixels in the remote sensing data, J is the total number of vertical indices of pixels in the remote sensing data, and the element r in the matrix ij represents the observed value of the pixel in the i-th column and j-th row in a specific band, 1≤i≤I, 1≤j≤J, and Λ is the set of known observations in the economic crop remote sensing data matrix R. Due to factors such as weather, sensor occlusion, and data loss, remote sensing data often contain a large number of missing pixels, and the observed values are non-negative physical quantities. Therefore, the constructed economic crop remote sensing data matrix R is a non-negative sparse matrix.

[0062] S3: Initialize the incremental PID controller based on the constructed economic crop remote sensing data matrix R.

[0063] In this embodiment, according to the economic crop remote sensing data matrix constructed by the economic crop remote sensing data matrix construction module, the total number of initialized incremental PID controllers is |Λ|, where |Λ| represents the number of known observation values in the economic crop remote sensing data matrix R, that is, each incremental PID controller corresponds to each known observation value in the economic crop remote sensing data matrix R.

[0064] S4: Construct the target loss function based on the economic crop remote sensing data and the economic crop remote sensing data matrix and perform iterative training to extract the latent factor matrix.

[0065] S4-1: Initialize the process parameters involved in the completion of economic crop remote sensing data.

[0066] The process parameters include latent feature matrices M and N; latent feature dimension F; convergence termination threshold δ; learning step size γ; regularization coefficient λ; where,

[0067] The latent feature dimension F determines the latent feature dimension of the latent factor matrices M and N and is initialized to a positive integer;

[0068] The sizes of the latent feature matrices M and N are determined by the dimension values of each order of the corresponding economic crop remote sensing data matrix R, that is, M is a latent factor matrix with I rows and F columns, and N is a latent factor matrix with J rows and F columns. The two latent factor matrices are initialized with small random positive numbers respectively.

[0069] The convergence termination threshold δ is a parameter used to determine whether the iterative process has converged, and is initialized with a very small positive number;

[0070] The regularization coefficient λ is a constant that controls the regularization effect of the related elements of the latent factor matrices M and N during training and is initialized to a small positive number.

[0071] S4-2: Construct the target loss function ε based on the economic crop remote sensing data and the economic crop remote sensing data matrix, which is expressed by the following formula:

[0072]

[0073] In formula (1), ε represents the target loss function; M and N represent the first latent factor matrix and the second latent factor matrix after decomposition of the economic crop remote sensing data matrix R; r ij represents the observed value of the pixel located in the i-th column and the j-th row in the economic crop remote sensing data matrix R in a specific band; Λ is the set of known observed values in the economic crop remote sensing data matrix R; m i is the i-th row vector in the first latent factor matrix M, n j is the j-th row vector in the second latent factor matrix N; <·,·> represents the inner product of two vectors; ||·||2 represents the L2 norm of a vector; λ represents the regularization coefficient of the latent factor matrix; Indicates the completion value of economic crop remote sensing data.

[0074] The objective loss function uses Euclidean distance as the optimization target; and uses Tikhonov regularization to constrain the latent factor matrices M and N to prevent overfitting problems;

[0075] S4-3: Use the standard stochastic gradient descent algorithm (SGD) to iteratively optimize the target loss function ε so that the value of the target loss function ε quickly approaches the minimum. The optimization formula is as follows:

[0076]

[0077] In formula (2), m i is the i-th row vector in the first latent factor matrix M, n j is the j-th row vector in the second latent factor matrix N; Indicates that ε is in (m i ,n j ) on m i The gradient, Indicates that ε is in (m i ,n j ) on n j The gradient of ;γ is the learning step size;

[0078] Then, the incremental PID controller is used to calculate the value of each observation r ij The instantaneous error e ij To control the learning step length γ in the above optimization formula. Among them, the instantaneous error e ij =r ij – <m i ,n j >. The incremental PID controller calculates the learning step increment as follows:

[0079]

[0080] In formula (3), is the incremental PID controller for the observed value r ij The learning step increment calculated at the tth iteration; K P is the proportional gain in the incremental PID controller, K I is the integral gain in the incremental PID controller, K D is the derivative gain in the incremental PID controller; is the observation value r at the tth iteration ij The instantaneous error, is the observation value r at the t-1th iteration ij The instantaneous error, is the observation value r at the t-2th iteration ij The instantaneous error of

[0081] Then the observed value r ij The learning step size at the tth iteration is:

[0082]

[0083] In formula (4), is the observed value r ij The learning step size at iteration t; is the observed value r ij The learning step size at the t-1th iteration; is the incremental PID controller for the observed value r ij The learning step increment calculated at iteration t.

[0084] S4-4: Determine whether the training iteration process of the loss function ε on the known data set Λ in the economic crop remote sensing data matrix R reaches the convergence condition. After the convergence condition is reached, terminate the training and output the first latent factor matrix M and the second latent factor matrix N.

[0085] In this embodiment, the convergence judgment condition is that the absolute value of the difference between the calculated value of the target loss function ε and the value of the target loss function ε of the previous iteration is less than the threshold δ.

[0086] S5: Based on the extracted latent factor matrix, the completion value of economic crop remote sensing data is calculated and stored.

[0087] In this embodiment, when the target loss function ε converges on the known data set Λ, the first latent factor matrix M and the second latent factor matrix N obtained by training when the target loss function ε reaches the minimum value are used to calculate the highest accuracy of the economic crop remote sensing data complement value of the pixel in the i-th column and the j-th row in a specific band.

[0088]

[0089] In formula (5), represents the economic crop remote sensing data completion value, that is, the economic crop remote sensing data completion value of the pixel in the i-th column and the j-th row; m i is the i-th row vector in the first latent factor matrix M, n j is the j-th row vector in the second latent factor matrix N.

[0090] The present invention provides a method and device for completing remote sensing data of economic crops based on incremental PID control. The method and device are specifically used for remote sensing data of economic crops, and can complete missing or abnormal remote sensing observation values with high precision and good temporal and spatial consistency, thereby improving the integrity and availability of remote sensing data, thereby providing reliable data support for economic crop growth monitoring, pest and disease control, yield estimation, etc., and can be widely used in economic crop growth monitoring, smart agriculture and other fields.

[0091] Based on the above-mentioned economic crop remote sensing data completion method based on incremental PID control, such as Figure 2 As shown, the present invention also provides an economic crop remote sensing data completion device based on incremental PID control, including a data acquisition module 1, a storage module 2, a matrix construction module 3, an incremental PID control module 4, a latent factor extraction module 5 and an output module 6.

[0092] The output end of the data acquisition module 1 is connected to the first input end of the storage module 2, the output end of the storage module 2 is respectively connected to the input end of the matrix construction module 3, the output end of the matrix construction module 3 is connected to the input end of the incremental PID control module 4, the output end of the incremental PID control module 4 is connected to the input end of the latent factor extraction module 5, the output end of the latent factor extraction module 5 is connected to the input end of the output module 6, and the output end of the output module 6 is connected to the second input end of the storage module 2.

[0093] The data acquisition module 1 is used to acquire remote sensing data of economic crops from the server and store the data.

[0094] The storage module 2 is used to store the received economic crop remote sensing data and economic crop remote sensing data completion values.

[0095] In this embodiment, Figure 3 As shown, the storage module 2 includes a first storage unit 21 and a second storage unit 22;

[0096] The first storage unit 21 is used to store the received economic crop remote sensing data in a structured manner in the form of a triple. The triple is represented by Z = (x, y, c), where x represents the column index of the pixel in the economic crop remote sensing data, corresponding to the horizontal coordinate in space; y represents the row index of the pixel in the economic crop remote sensing data, corresponding to the vertical coordinate in space; and c represents the observation value of the pixel in a specific band.

[0097] The second storage unit 22 is used to store the economic crop remote sensing data supplement value output by the output module 6 .

[0098] In this embodiment, the matrix construction module 3 is used to construct the economic crop remote sensing data matrix R according to the economic crop remote sensing data.

[0099] Construct all triples Z = (x, y, c) into a second-order economic crop remote sensing data matrix R of size I × J I ×J , where I is the total number of horizontal indices of pixels in the remote sensing data, J is the total number of vertical indices of pixels in the remote sensing data, and the element r in the matrix ij It represents the observation value of the pixel located in the i-th column and the j-th row in a specific band, 1≤i≤I, 1≤j≤J, and Λ is the set of known observation values in the economic crop remote sensing data matrix R.

[0100] In this embodiment, the incremental PID control module 4 is used to initialize the number of incremental PID controllers according to the economic crop remote sensing data matrix R.

[0101] The number of incremental PID controllers is Λ|, corresponding to the number of known observation values in the economic crop remote sensing data matrix R, and each incremental PID controller corresponds to each known observation value in the economic crop remote sensing data matrix R.

[0102] The latent factor extraction module 5 is used to construct a target loss function based on the economic crop remote sensing data matrix and the initialized incremental PID controller, and extract the latent factor matrix.

[0103] like Figure 4 As shown, the latent factor extraction module 5 includes an initialization unit 51, a target loss function construction unit 52 and a training unit 53; the output end of the initialization unit 51 is connected to the input end of the target loss function construction unit 52, the output end of the target loss function construction unit 52 is connected to the input end of the training unit 53, and the output end of the training unit 53 is connected to the output module 6.

[0104] Initialization unit 51, used for initializing process parameters involved in completing economic crop remote sensing data;

[0105] The target loss function construction unit 52 is used to construct the target loss function according to the economic crop remote sensing data, the economic crop remote sensing data matrix, and the process parameters;

[0106] The training unit 53 is used to train and optimize the target loss function and extract the latent factor matrix.

[0107] The output module 6 is used to output the economic crop remote sensing data completion value according to the extracted latent factor matrix.

[0108] The present invention also provides an electronic device, which includes a processor, and the processor is used to run a computer program stored in a memory, so that the electronic device implements the steps of the economic crop remote sensing data completion method based on incremental PID control in the above embodiment.

[0109] The present invention also provides a computer-readable storage medium storing a computer program. When the computer program is executed on a processor, the computer program implements the steps of the economic crop remote sensing data completion method based on incremental PID control in the above embodiment.

[0110] A computer program includes computer program code, which may be in source code form, object code form, executable files, or some intermediate form. Computer-readable media may include at least any entity or device capable of carrying computer program code to an electronic device, recording media, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signals, telecommunications signals, and software distribution media. Examples include USB flash drives, removable hard drives, magnetic disks, or optical disks. In some jurisdictions, due to legislation and patent practice, computer-readable media cannot be electrical carrier signals or telecommunications signals.

[0111] Those skilled in the art will appreciate that the above-mentioned embodiments are specific examples for implementing the present invention, and that in actual applications, various changes may be made thereto in form and detail without departing from the spirit and scope of the present invention.

Claims

1. A method for completing economic crop remote sensing data based on incremental PID control, characterized in that: The specific steps include: S1: collects and stores economic crop remote sensing data from the server; S2: construct cash crop remote sensing data matrix based on cash crop remote sensing data; S3: Initialize the incremental PID controller based on the constructed economic crop remote sensing data matrix; S4: Construct the target loss function based on the economic crop remote sensing data matrix and the incremental PID controller and perform iterative training to extract the latent factor matrix; S5: Based on the extracted latent factor matrix, the completion value of economic crop remote sensing data is calculated and stored.

2. The economic crop remote sensing data completion method based on incremental PID control according to claim 1, characterized in that: In S1, the received economic crop remote sensing data is stored in the form of a triple; the triple is expressed as Z = (x, y, c), where x represents the column index of the pixel in the economic crop remote sensing data, corresponding to the horizontal coordinate in space; y represents the row index of the pixel in the economic crop remote sensing data, corresponding to the vertical coordinate in space; c represents the observation value of the pixel in a specific band.

3. The economic crop remote sensing data completion method based on incremental PID control according to claim 1, characterized in that: In S2, all triples Z = (x, y, c) are constructed into a second-order economic crop remote sensing data matrix R of size I × J I ×J , where I is the total number of horizontal indices of pixels in the remote sensing data, J is the total number of vertical indices of pixels in the remote sensing data, and the element r in the matrix ij It represents the observation value of the pixel located in the i-th column and the j-th row in a specific band, 1≤i≤I, 1≤j≤J, and Λ is the set of known observation values in the economic crop remote sensing data matrix R.

4. The economic crop remote sensing data completion method based on incremental PID control according to claim 1, characterized in that: In S3, the total number of incremental PID controllers is |Λ|, where |Λ| represents the number of known observation values in the economic crop remote sensing data matrix R, that is, each incremental PID controller corresponds to each known observation value in the economic crop remote sensing data matrix R.

5. The economic crop remote sensing data completion method based on incremental PID control according to claim 1, characterized in that: The S4 includes: S4-1: Initialize the process parameters involved in the completion of economic crop remote sensing data; S4-2: Construct target loss function based on cash crop remote sensing data and cash crop remote sensing data matrix; S4-3: Iteratively optimize the target loss function ε using the standard stochastic gradient descent algorithm and the incremental PID controller; S4-4: Determine whether the loss function ε has reached the convergence condition in the economic crop remote sensing data matrix R. If so, terminate the training and output the first latent factor matrix M and the second latent factor matrix N; otherwise, continue the iteration.

6. The economic crop remote sensing data completion method based on incremental PID control according to claim 5, characterized in that: In said S4-1, the process parameters include latent feature matrices M and N; latent feature dimension F; Convergence termination threshold δ; learning step size γ; regularization coefficient λ; where, The latent feature dimension F determines the latent feature dimension of the latent factor matrices M and N and is initialized to a positive integer; The sizes of the latent feature matrices M and N are determined by the dimension values of each order of the corresponding economic crop remote sensing data matrix R, that is, M is a latent factor matrix with I rows and F columns, and N is a latent factor matrix with J rows and F columns. The two latent factor matrices are initialized with small random positive numbers respectively. The convergence termination threshold δ is a parameter used to determine whether the iterative process has converged, and is initialized with a very small positive number; The regularization coefficient λ is a constant that controls the regularization effect of the related elements of the latent factor matrices M and N during training and is initialized to a small positive number.

7. The economic crop remote sensing data completion method based on incremental PID control according to claim 5, characterized in that: In S4-2, the objective loss function is expressed by the following formula: In formula (1), ε represents the target loss function; M and N represent the first latent factor matrix and the second latent factor matrix after decomposition of the economic crop remote sensing data matrix R; r ij represents the observed value of the pixel located in the i-th column and the j-th row in the economic crop remote sensing data matrix R in a specific band; Λ is the set of known observed values in the economic crop remote sensing data matrix R; m i is the i-th row vector in the first latent factor matrix M, n j is the j-th row vector in the second latent factor matrix N; <·,·> represents the inner product of two vectors; ||·||2 represents the L2 norm of a vector; λ represents the regularization coefficient of the latent factor matrix; Indicates the completion value of economic crop remote sensing data.

8. The economic crop remote sensing data completion method based on incremental PID control according to claim 5, characterized in that: In S4-3, the iterative optimization method is: First, the stochastic gradient descent algorithm is used for optimization. The optimization formula is as follows: In formula (2), m i is the i-th row vector in the first latent factor matrix M, n j is the j-th row vector in the second latent factor matrix N; Indicates that ε is in (m i ,n j ) with respect to the gradient of mi, Indicates that ε is in (m i ,n j ) on n j The gradient of ;γ is the learning step size; Then, the incremental PID controller is used to calculate the value of each observation r. ij The instantaneous error e ij To control the learning step size γ in the above optimization formula ， Among them, the instantaneous error e ij =r ij – <m i ,n j >, the incremental PID controller calculates the learning step increment as follows: In formula (3), is the incremental PID controller for the observed value r ij The learning step increment calculated at the tth iteration; K P is the proportional gain in the incremental PID controller, K I is the integral gain in the incremental PID controller, K D is the derivative gain in the incremental PID controller; is the observation value r at the tth iteration ij The instantaneous error, is the observation value r at the t-1th iteration ij The instantaneous error, is the observation value r at the t-2th iteration ij The instantaneous error of Then the observed value r ij The learning step size at the tth iteration is: In formula (4), is the observed value r ij The learning step size at iteration t; is the observed value r ij The learning step size at the t-1th iteration; is the incremental PID controller for the observed value r ij The learning step increment calculated at iteration t.

9. The economic crop remote sensing data completion method based on incremental PID control according to claim 1, characterized in that: In S5, the economic crop remote sensing data completion value The calculation formula is: In formula (5), represents the economic crop remote sensing data completion value, that is, the economic crop remote sensing data completion value of the pixel in the i-th column and the j-th row; m i is the i-th row vector in the first latent factor matrix M, n j is the j-th row vector in the second latent factor matrix N.

10. An economic crop remote sensing data completion device based on incremental PID control according to any one of claims 1 to 9, characterized in that: It includes data acquisition module, storage module, matrix construction module, incremental PID control module, latent factor extraction module and output module; The data acquisition module is used to obtain economic crop remote sensing data from the server and store it; A storage module, used for storing received economic crop remote sensing data and economic crop remote sensing data complement values; The matrix construction module is used to construct the economic crop remote sensing data matrix R based on the economic crop remote sensing data; The incremental PID control module is used to initialize the number of incremental PID controllers according to the economic crop remote sensing data matrix R; The latent factor extraction module is used to construct the target loss function based on the economic crop remote sensing data matrix and the initialized incremental PID controller, and extract the latent factor matrix; The output module is used to output the economic crop remote sensing data completion value according to the extracted latent factor matrix.

Images