Method and system for correcting strip radiance consistency of multispectral remote sensing image

Abstract

The application provides a multispectral remote sensing image strip radiation consistency correction method and system. The method comprises: acquiring a plurality of remote sensing image strips with lateral or heading overlap relationship; constructing a graph structure based on the coverage range of the plurality of remote sensing image strips; for each connection edge in the graph structure, determining the attribute value of the connection edge based on the invariant pixels identified from the overlapping area of the strips corresponding to the connection edge; for each node, fitting the radiation inconsistency equation between adjacent nodes based on the invariant pixels to obtain the attribute value of the node; based on the attribute values of the nodes and the connection edges in the graph structure, iteratively adjusting the radiation correction parameters of each image strip using a greedy optimization algorithm to determine a set of radiation correction parameters; based on the set of radiation correction parameters, sequentially performing radiation correction on each image strip to generate radiation-consistent multispectral remote sensing image strips. The application can realize the correction of the radiation consistency of the multispectral remote sensing image strips based on the graph optimization mode.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of remote sensing satellites, in particular to a multi-spectral remote sensing image strip radiometric consistency correction method and system. BACKGROUND

[0002] With the rapid development of remote sensing technology, multi-spectral image technology plays an increasingly important role in the fields of environmental monitoring, agricultural management, disaster response, etc. High spatial resolution multi-spectral images can not only accurately capture high-resolution features on the ground, but also effectively reflect the spectral characteristics of ground objects, providing important technical support for accurate analysis and decision-making. However, due to the limitations of sensor performance, changes in atmospheric conditions, and differences in observation geometry, etc., the radiometric response of multi-spectral images often shows significant inconsistency in the time and space dimensions. This radiometric inconsistency not only seriously affects the reliability and accuracy of image data, but also adversely affects the implementation effect of key tasks such as crop growth monitoring, disaster dynamic assessment, and precision decision support. Therefore, obtaining multi-spectral images with temporal and spatial radiometric consistency is of great significance to ensure the scientificity and practicality of remote sensing data in production and application.

[0003] To solve the problem of radiometric inconsistency of multi-spectral images, researchers have proposed a variety of correction methods, including radiometric correction techniques based on statistical models, strip smoothing methods based on image registration, and machine learning driven radiometric consistency optimization techniques, etc. However, these methods generally have the following limitations: First, they often rely heavily on external standard reference data (such as Landsat or Sentinel-2 images), or need to specify an internal reference standard as a benchmark, which not only increases the dependence on external resources (such as only correcting spectral bands compatible with reference data), but also limits the scope of application of the correction method, especially when the reference data cannot cover the target area or time period; second, existing methods mostly fail to fully consider the internal relationship between multi-spectral image strips, and cannot correct multi-temporal and multi-band images as a whole. SUMMARY

[0004] The purpose of the present application is to provide a multi-spectral remote sensing image strip radiometric consistency correction method and system, which can realize the correction of multi-spectral remote sensing image strip radiometric consistency based on graph optimization, eliminate the radiometric inconsistency error between multiple multi-spectral image strips with lateral or heading overlap relationship, without relying on external standard reference data, and also fully considering the internal relationship between multi-spectral image strips, correcting multi-temporal and multi-band images as a whole.

[0005] In a first aspect, the present application provides a multispectral remote sensing image strip radiation consistency correction method, which comprises: acquiring a plurality of remote sensing image strips having a side or heading overlap relationship; constructing a graph structure based on the coverage ranges of the plurality of remote sensing image strips; in the graph structure, each remote sensing image strip is a node, and a connection edge is established between two nodes when there is an overlapping area between the two remote sensing image strips; for each connection edge in the graph structure, determining the attribute value of the connection edge based on the invariant pixels identified from the strip overlapping area corresponding to the connection edge; for each node in the graph structure, fitting the radiation inconsistency equation between adjacent nodes based on the invariant pixels to obtain the attribute value of the node; based on the attribute values of the nodes and the connection edges in the graph structure, iteratively adjusting the radiation correction parameters of each image strip using a greedy optimization algorithm to determine a set of radiation correction parameters; and based on the set of radiation correction parameters, sequentially performing radiation correction on each image strip to generate a set of radiation-consistent multispectral remote sensing image strips.

[0006] Further, the step of determining the attribute value of the connection edge based on the invariant pixels identified from the strip overlapping area corresponding to the connection edge comprises: identifying the invariant pixels in the strip overlapping area corresponding to the connection edge according to a multivariate change detection method; and calculating the radiation inconsistency error between the adjacent nodes corresponding to the connection edge using the identified invariant pixels according to the following first specified formula:

[0007] wherein, represents the radiation inconsistency error between the adjacent node strips and the strip , and respectively represent the ground reflectance values of the th invariant pixel in the overlapping area of the two strips, is the total number of invariant pixels in the overlapping area of the two strips.

[0008] The radiation inconsistency error between the adjacent nodes is taken as the attribute value of the connection edge.

[0009] Further, the step of fitting the radiation inconsistency equation between the adjacent nodes based on the invariant pixels to obtain the attribute value of the node comprises: fitting the radiation inconsistency equation for the node and the connected nodes to obtain a set as the attribute set of the node ; wherein the radiation inconsistency equation is as follows:

[0010] wherein, and respectively represent the adjacent node strips and the strip​ the ground reflectance value of all invariant pixels in the overlapping area, is the slope, is the intercept, is the residual error;

[0011] The attribute set of each node is determined as the attribute value of the node.

[0012] Further, the step of determining the radiation correction parameter set based on the attribute values of the nodes and the connecting edges in the graph structure and iteratively adjusting the radiation correction parameters of each image strip by using a greedy optimization algorithm comprises: for the current round of iteration, performing the following calculation steps: calculating the overall radiation inconsistency error of the graph structure and the radiation inconsistency error of each node in the current round of iteration according to the attribute values of each connecting edge; calculating the correction parameter correction amount of each node according to the attribute values of each node; determining the new correction parameter of each node after correction based on the correction parameter correction amount of each node; calculating the radiation inconsistency error of each node after applying the new correction parameter; calculating the correctable error of each node based on the radiation inconsistency error of each node after applying the new correction parameter and the radiation inconsistency error of each node; determining the target node corresponding to the maximum correctable error, adjusting the radiation correction parameter of the target node to the corresponding new correction parameter, updating the attribute values of the nodes and edges associated with the target node after parameter adjustment, updating the overall radiation inconsistency error of the graph structure based on the updated attribute values; judging whether the difference between the overall radiation inconsistency error of the updated graph structure and the overall radiation inconsistency error of the graph structure in the current round of iteration is greater than a threshold value; if yes, continue to perform the calculation steps in the next round of iteration; if no, stop iteration, and the radiation correction parameter set is composed of the radiation correction parameters of each node.

[0013] Further, the step of calculating the overall radiation inconsistency error of the graph structure and the radiation inconsistency error of each node in the current round of iteration according to the attribute values of each connecting edge comprises: calculating the overall radiation inconsistency error of the graph structure and the radiation inconsistency error of each node in the current round of iteration according to the following two formulas:

[0014] ;

[0015] ;

[0016] wherein, represents the overall radiation inconsistency error of the graph structure; represents the radiation inconsistency error of node ; N and M are the total number of nodes and connecting edges in the graph structure; is the number of nodes connected to node n; is the strip and its first radiometric inconsistency error between adjacent strips.

[0017] Further, the step of calculating the correction parameter correction of each node according to the attribute value of each node comprises: calculating the attribute average value of each node based on the attribute value of each node according to the following formula:

[0018] and ;

[0019] wherein, is the number of nodes connected to the node ; and is the attribute average value of the node ;

[0020] The step of calculating the correction parameter correction of each node according to the attribute value of each node comprises: calculating the correction parameter correction of each node according to the following formula:

[0021] and ;

[0022] wherein, and is the correction parameter correction of the node .

[0023] Further, the step of determining the new correction parameter of each node after correction based on the correction parameter correction of each node comprises: calculating the new correction parameter of each node after correction according to the following formula:

[0024] and ;

[0025] wherein, and is the current radiation correction parameter, and the initial value is 1 and 0 respectively; and is the correction parameter correction of the node ; and is the new correction parameter of the node after correction.

[0026] Further, the step of calculating the correctable error of each node based on the radiation inconsistency error after applying the new correction parameter of each node and the radiation inconsistency error of each node comprises: calculating the correctable error of each node according to the following formula:

[0027] ;

[0028] wherein, indicates the node Correctable error; For nodes The level of radiation inconsistency error after applying the current correction factors. For nodes Radiation inconsistency error after applying the new correction parameters.

[0029] Furthermore, the above-mentioned step of performing radiometric correction on each image strip sequentially based on the radiometric correction parameter set to generate radiometrically consistent multispectral remote sensing image strips includes: using a product reflectance image provided by a third party as a correction reference image, systematically fine-tuning the radiometric correction parameters in the radiometric correction parameter set to obtain optimized radiometric correction parameters; the product reflectance image provided by the third party includes Landsat-8 or Sentinel-2 reflectance images; and performing radiometric correction on each image strip sequentially according to the following formula to generate radiometrically consistent multispectral remote sensing image strips:

[0030] ,in, This is a stripe of surface reflectance image after radiometric consistency correction. This is a stripe of surface reflectance image before correction. and These are the optimized radiation correction parameters.

[0031] Secondly, this application also provides a multispectral remote sensing image strip radiometric consistency correction system, the system comprising: a remote sensing image acquisition module for acquiring multiple remote sensing image strips with lateral or lateral overlap; a graph structure construction module for constructing a graph structure based on the coverage of the multiple remote sensing image strips; in the graph structure, each remote sensing image strip is treated as a node, and a connection edge is established between two nodes when there is an overlapping area between two remote sensing image strips; an attribute value determination module for determining the attribute value of each connection edge in the graph structure based on invariant pixels identified from the overlapping area of ​​the strip corresponding to the connection edge; for each node in the graph structure, a radiometric inconsistency equation between adjacent nodes is fitted based on invariant pixels to obtain the node's attribute value; a correction parameter determination module for iteratively adjusting the radiometric correction parameters of each image strip using a greedy optimization algorithm based on the attribute values ​​of the nodes and connection edges in the graph structure to determine the radiometric correction parameter set; and a radiometric correction module for sequentially performing radiometric correction on each image strip based on the radiometric correction parameter set to generate radiometrically consistent multispectral remote sensing image strips.

[0032] The multispectral remote sensing image strip radiation consistency correction method and system provided by the application first acquires a plurality of remote sensing image strips with a side or heading overlap relationship; then constructs a graph structure based on the coverage ranges of the plurality of remote sensing image strips; in the graph structure, each remote sensing image strip is a node, and a connection edge is established between two nodes when there is an overlap area between the two remote sensing image strips; for each connection edge in the graph structure, the attribute value of the connection edge is determined based on the invariant pixels identified from the strip overlap area corresponding to the connection edge; for each node in the graph structure, the attribute value of the node is obtained by fitting the radiation inconsistency equation between adjacent nodes based on the invariant pixels; finally, based on the attribute values of the nodes and the connection edges in the graph structure, the radiation correction parameters of each image strip are iteratively adjusted using a greedy optimization algorithm to determine a set of radiation correction parameters; and based on the set of radiation correction parameters, each image strip is sequentially radiation corrected to generate a radiation consistent multispectral remote sensing image strip. The application can correct the radiation consistency of multispectral remote sensing image strips based on graph optimization, eliminate the radiation inconsistency error between a plurality of multispectral image strips with a side or heading overlap relationship, without relying on external standard reference data, and can also fully consider the internal relationship between the multispectral image strips to correct the multispectral images as a whole. BRIEF DESCRIPTION OF DRAWINGS

[0033] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the drawings needed in the description of the embodiments or the prior art will be briefly introduced. Obviously, the drawings in the following description are some embodiments of the present application, and other drawings can also be obtained by those skilled in the art without creative labor.

[0034] Figure 1 A flowchart of a multispectral remote sensing image strip radiation consistency correction method provided by an embodiment of the present application;

[0035] Figure 2 A module flowchart provided by an embodiment of the present application;

[0036] Figure 3 A schematic diagram of a comparison result provided by an embodiment of the present application;

[0037] Figure 4 Another schematic diagram of a comparison result provided by an embodiment of the present application;

[0038] Figure 5 A structural block diagram of a multispectral remote sensing image strip radiation consistency correction system provided by an embodiment of the present application. DETAILED DESCRIPTION

[0039] The technical solutions of the present application will be clearly and completely described below in conjunction with the embodiments. Obviously, the described embodiments are only some of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative work fall within the scope of protection of the present application.

[0040] In view of the problems of high dependence on external standard reference data, small applicable range of correction method, and failure to fully consider the internal relationship between multispectral image strips in the correction method of radiation inconsistency in the prior art (such as statistical model-based radiation correction technology, image registration strip smoothing method, and machine learning-driven radiation consistency optimization technology), the embodiments of the present application provide a multispectral remote sensing image strip radiation consistency correction method and system, which can correct the radiation consistency of multispectral remote sensing image strips based on a graph optimization manner, is used for eliminating the radiation inconsistency error between multiple multispectral image strips with side or heading overlapping relationship, does not need to depend on external standard reference data, can also fully consider the internal relationship between multispectral image strips, and corrects multiple time-phase and multi-band images as a whole.

[0041] To facilitate the understanding of the embodiments, first, a multispectral remote sensing image strip radiation consistency correction method disclosed by the embodiments of the present application is introduced in detail.

[0042] A graph is an ideal structure that can effectively represent the internal image mutual radiation relationship. Thanks to the flexibility and extensibility of the graph structure, the graph-based optimization can realize overall and real-time correction of the radiation inconsistency error without depending on any reference data. Therefore, the multispectral remote sensing image strip radiation consistency correction method and system provided by the embodiments of the present application based on graph optimization have important research value and broad application prospect.

[0043] Figure 1 A flowchart of the multispectral remote sensing image strip radiation consistency correction method provided by the embodiments of the present application is shown in the figure, and the method specifically includes the following steps:

[0044] In step S102, multiple remote sensing image strips with side or heading overlapping relationship are acquired; that is, multiple PlanetScope (PS) image strips.

[0045] In step S104, a graph structure is constructed based on the coverage range of the multiple remote sensing image strips; in the graph structure, each remote sensing image strip is taken as a node, and a connection edge between two nodes is established when there is an overlapping area between the two remote sensing image strips.

[0046] In this step, in order to organize the image strips into a whole, when constructing the graph structure, each remote sensing image strip is defined as a node in the graph structure, and if two strips have an overlapping area, a connection edge is established between the nodes corresponding to the two strips. In this embodiment, only the connection edges whose effective observation proportion in the overlapping area is more than 50% and the time interval between the acquisition of the two PS image strips is not more than 2 days are reserved. Since one strip corresponds to one sensor during the acquisition of the remote sensing image, the image strip data whose time interval is not more than 2 days is used to construct the graph structure in this embodiment, so that the reflectivity of the ground object observed by different sensors is ensured to be unchanged, only the data whose effective observation proportion in the overlapping area is more than 50% is reserved, and the reflectivity of the remote sensing image is ensured to be as consistent as possible, thereby improving the accuracy of subsequent data analysis.

[0047] In step S106, for each connection edge in the graph structure, the attribute value of the connection edge is determined based on the invariant pixels identified from the overlapping area of the strips corresponding to the connection edge; and for each node in the graph structure, the attribute value of the node is obtained by fitting the radiation inconsistency equation between adjacent nodes based on the invariant pixels.

[0048] In step S108, based on the attribute values of the nodes and the connection edges in the graph structure, a greedy optimization algorithm is used to iteratively adjust the radiation correction parameters of each image strip to determine a set of radiation correction parameters.

[0049] The specific implementation process of the algorithm can be referred to the specific description in the following embodiments.

[0050] In step S110, based on the set of radiation correction parameters, each image strip is sequentially subjected to radiation correction to generate a set of radiation-consistent multi-spectral remote sensing image strips.

[0051] In a preferred manner, a small amount of data can also be used to fine-tune the radiation correction parameters in the set of radiation correction parameters described above, to ensure that the corrected image strips and the reflectivity product provided by a third party are unified in radiation consistency.

[0052] For example, the product reflectivity image provided by the third party is used as a correction reference image to systematically fine-tune the radiation correction parameters in the set of radiation correction parameters to obtain optimized radiation correction parameters; the product reflectivity image provided by the third party includes the reflectivity image of Landsat-8 or Sentinel-2; and then each image strip is sequentially subjected to radiation correction according to the following formula to generate a set of radiation-consistent multi-spectral remote sensing image strips:

[0053] wherein, is the ground reflectivity image strip after radiation consistency correction, is the ground reflectivity image strip before correction, and These are the optimized radiation correction parameters.

[0054] It should be noted that the above embodiment is a preferred method. If parameter fine-tuning is not required, radiometric correction can also be performed sequentially on each image strip using the radiometric correction parameters from the radiometric correction parameter set. In this case, and These represent the radiation correction parameters in the radiation correction parameter set.

[0055] The multispectral remote sensing image strip radiometric consistency correction method provided in this application embodiment can correct the radiometric consistency of multispectral remote sensing image strips based on graph optimization. It is used to eliminate radiometric inconsistency errors between multiple multispectral image strips with lateral or directional overlap. It does not rely on external standard reference data, and can also fully consider the inherent interrelationships between multispectral image strips to perform overall correction of multi-temporal and multi-band images.

[0056] This application also provides another method for radiometric consistency correction of multispectral remote sensing image stripes, which is implemented based on the above embodiments. This embodiment focuses on describing the process of determining the attribute values ​​of nodes and connecting edges in the graph structure, and the process of determining the set of radiometric correction parameters using a greedy optimization algorithm.

[0057] The step S106 above, "determining the attribute value of the connection edge based on the invariant cells identified from the overlapping strip region corresponding to the connection edge," includes:

[0058] (1) Identify invariant pixels in the overlapping region of the strips corresponding to the connecting edges using the Multivariate Alteration Detection (MAD) method;

[0059] (2) Calculate the radiation inconsistency error between adjacent nodes corresponding to the connecting edge using the identified invariant pixels according to the first specified formula below:

[0060] ,in, Represents adjacent node stripes With strip Radiation inconsistency error, and These represent the first two bands within the overlapping region. The surface reflectance value of an invariant pixel. This represents the total number of invariant pixels within the overlapping regions of the two bands;

[0061] (3) Use the radiation inconsistency error between adjacent nodes as the attribute value of the connecting edge.

[0062] Further, the step of "fitting radiometric inconsistency equations between adjacent nodes based on invariant pixels to obtain attribute values of the nodes" in the above step S106 includes:

[0063] (1) fitting radiometric inconsistency equations (RIE) for the node and the corresponding connected connected nodes to obtain a set as the attribute set of the node ; wherein the radiometric inconsistency equation is as follows:

[0064] wherein and represent the ground reflectance values of all invariant pixels in the overlapping area of the adjacent node strips and , in the radiometric inconsistency equation, is the slope, and is the intercept, and and are taken as the attribute values of the node , which can constitute , i.e., the attribute set of the node , is the residual error;

[0065] (2) determining the attribute set of each node as the attribute value of the node.

[0066] Further, the step of "fitting radiometric inconsistency equations between adjacent nodes based on invariant pixels to obtain attribute values of the nodes" in the above step S106 includes:

[0067] For the current round of iteration, the following calculation steps are performed:

[0068] (1) calculating the overall radiometric inconsistency error of the graph structure and the radiometric inconsistency error of each node in the current round of iteration according to the attribute values of each connected edge;

[0069] Specifically, the overall radiometric inconsistency error of the graph structure and the radiometric inconsistency error of each node in the current round of iteration are calculated according to the following two formulas:

[0070] ;

[0071] ;

[0072] wherein ​the overall radiation inconsistency error of the graph structure; the radiation inconsistency error of the node ; N and M are respectively the total number of nodes and connecting edges in the graph structure; the number of nodes connected to the node n; the radiation inconsistency error between the node and its first adjacent node.

[0073] (2) According to the attribute value of each node, the correction parameter correction amount of each node is calculated;

[0074] In specific implementation, according to the following formula, the attribute average value of each node is calculated based on the attribute value of each node:

[0075] and ;

[0076] wherein, and are respectively the attribute value of the node ; is the number of nodes connected to the node ; and are respectively the attribute average value of the node ;

[0077] According to the following formula, the correction parameter correction amount of each node is calculated:

[0078] and ;

[0079] wherein, and are respectively the correction parameter correction amount of the node .

[0080] (3) Based on the correction parameter correction amount of each node, the new correction parameter of each node after correction is determined;

[0081] In specific implementation, according to the following formula, the new correction parameter of each node after correction is calculated:

[0082] and ;

[0083] wherein, and are respectively the current radiation correction parameter, and the initial value is respectively 1 and 0; and are respectively the correction parameter correction amount of the node ; and For node Corrected new correction parameter.

[0084] (4) Calculate the radiation inconsistency error of each node after applying the new correction parameter;

[0085] (5) Based on the radiation inconsistency error of each node after applying the new correction parameter, and the radiation inconsistency error of each node, calculate the correctable error of each node;

[0086] In specific implementation, the correctable error of each node is calculated according to the following formula:

[0087]

[0088] Among them, represents the correctable error of node ; is the radiation inconsistency error level of node after applying the current correction coefficient, is the radiation inconsistency error of node after applying the new correction parameter.

[0089] (6) Determine the target node corresponding to the maximum correctable error, adjust the radiation correction parameter of the target node to the corresponding new correction parameter, update the attribute values of the nodes and edges associated with the target node after parameter adjustment, and update the overall radiation inconsistency error of the graph structure based on the updated attribute values;

[0090] (7) Determine whether the difference between the overall radiation inconsistency error of the updated graph structure and the overall radiation inconsistency error of the graph structure in the current round of iteration is greater than a threshold value; if yes, continue to perform the calculation step in the next round of iteration; if no, stop iteration, and the radiation correction parameter set is composed of the radiation correction parameters of each node.

[0091] In the embodiments of the present application, the node n with a higher radiation inconsistency error level is preferentially selected for correction, and the correction parameter and of the node are updated, as well as the node attribute , the edge attribute connected to the node , the radiation inconsistency error level of the node , and the overall radiation inconsistency error level of the graph is updated synchronously . Repeat step 5 until the global radiation inconsistency error level does not decrease significantly in the iteration process, and determine the optimal correction parameter set .

[0092] Referring to one embodiment shown in Figure 2 , a multispectral remote sensing image strip radiation consistency correction device includes the following modules:​

[0093] S1. Graph construction module

[0094] S1-1. Graph structure construction module

[0095] Data input: Coverage of PS image strips.

[0096] Data output: All image strips are organized into a complete graph structure according to the side or heading overlap relationship between image strips, taking image strips as nodes in the graph structure.

[0097] S1-2. Invariant pixel identification module

[0098] Data input: Graph structure output by S1-1.

[0099] Data output: Invariant pixels are identified in the overlap area of the two overlapping strips corresponding to each edge in the graph structure.

[0100] S1-3. Graph attribute calculation module

[0101] Data input: Invariant pixels identified by S1-2.

[0102] Data output: A complete graph structure with complete connection edges and node attributes is generated.

[0103] S2. Graph optimization module

[0104] S2-1. Graph optimization module based on greedy strategy

[0105] Data input: Complete graph output by S1.

[0106] Data output: The graph structure is optimized by the greedy strategy to generate corresponding strip image radiometric correction parameters.

[0107] S2-2. Fine-tuning of third-party reflectance product

[0108] Data input: Radiometric correction parameters output by S2-1.

[0109] Data output: Fine-tuned radiometric correction parameters are generated with Landsat-8 and Sentinel-2 reflectance products as reference standards, respectively.

[0110] S2-3. Image strip-by-strip correction

[0111] Data input: Initial correction parameters output by S2-1 unit, or fine-tuned correction parameters output by S2-2 unit.

[0112] Data output: Strip images with completed radiometric correction are generated.

[0113] In another embodiment, as shown in Figure 3 The comparison results of the PS strip images of the research area (a river basin) before and after the radiation consistency correction according to the embodiment of the application are shown in Figure 4 The radiation inconsistency between the image strips has been completely eliminated after the image strips are corrected by the method provided in the embodiment, which further proves the high reliability and significant performance advantage of the method provided in the embodiment. As shown in Table 1, the method provided in the embodiment can not only correct the radiation inconsistency errors of all spectral bands in all original image strips, but also achieve the optimal correction effect, which is manifested as the lowest mean absolute error (MAE) and mean relative error (MRE). Table 1 shows the radiation consistency correction accuracy table of the method of the embodiment, the original strips and the official consistency method of the related company in the research area (the bolded part indicates the optimal score).

[0114] Table 1

[0115]

[0116] In the method for correcting the radiation consistency of the multispectral remote sensing image strips provided in the embodiment of the application, firstly, a plurality of remote sensing image strips with lateral or heading overlap are acquired; a graph structure is constructed according to the coverage range of the strips, the image strips are defined as nodes in the graph structure, and the overlapping regions are defined as edges connecting the nodes. In each edge, the invariant pixels in the overlapping region are identified, and the radiation inconsistency error between the adjacent nodes is calculated based on the invariant pixels as the attribute value of the edge; for each node, the radiation inconsistency equation between the adjacent nodes is fitted by using the invariant pixels, and the fitting coefficients are set as the node attributes. On this basis, based on the attribute values of the nodes and the edges, a greedy optimization algorithm is used to iteratively optimize the radiation parameters of the image strips, so as to gradually reduce the global radiation inconsistency error. Finally, the user can select to fine-tune the correction parameters by a small amount of data, so that the corrected image is consistent with the third-party reflectivity product; or directly use the optimized correction parameters to correct the image strips one by one, to realize the overall radiation consistency correction. The method significantly improves the radiation consistency of the remote sensing image, and provides high-quality data guarantee for image analysis and application.

[0117] Based on the above method embodiment, the embodiment of the application further provides a system for correcting the radiation consistency of multispectral remote sensing image strips, which is described with reference to Figure 5As shown, the system comprises: a remote sensing image acquisition module 502, configured to acquire a plurality of remote sensing image strips having a side or heading overlap relationship; a graph structure construction module 504, configured to construct a graph structure based on the coverage of the plurality of remote sensing image strips; in the graph structure, each remote sensing image strip is taken as a node, and a connection edge between two nodes is established when there is an overlap area between the two remote sensing image strips; an attribute value determination module 506, configured to, for each connection edge in the graph structure, determine an attribute value of the connection edge based on invariant pixels identified from the strip overlap area corresponding to the connection edge; for each node in the graph structure, based on the invariant pixels, fit a radiation inconsistency equation between adjacent nodes to obtain an attribute value of the node; a correction parameter determination module 508, configured to, based on the attribute values of the nodes and the connection edges in the graph structure, iteratively adjust the radiation correction parameters of each image strip using a greedy optimization algorithm to determine a set of radiation correction parameters; and a radiation correction module 510, configured to, based on the set of radiation correction parameters, sequentially perform radiation correction on each image strip to generate a set of radiation-consistent multispectral remote sensing image strips.

[0118] Further, the attribute value determination module 506 is configured to identify the invariant pixels in the strip overlap area corresponding to the connection edge according to a multivariate change detection method; and calculate the radiation inconsistency error between the adjacent nodes corresponding to the connection edge according to the following first specified formula using the identified invariant pixels:

[0119] wherein, denotes the radiation inconsistency error between the adjacent node strips and the strip , and denote the ground reflectance values of the i-th invariant pixel in the overlap area of the two strips, is the total number of invariant pixels in the overlap area of the two strips. The radiation inconsistency error between the adjacent nodes is taken as the attribute value of the connection edge.

[0120] Further, the attribute value determination module 506 is configured to perform radiation inconsistency equation fitting on the node

[0121] and the corresponding connection connection nodes to obtain a set as the attribute set of the node ; wherein the radiation inconsistency equation is as follows:

[0122] wherein, and denote the adjacent node strips and the strip ​​the ground reflectance value of all invariant pixels in the overlapping area, is the slope, is the intercept, is the residual error;

[0123] The attribute set of each node is determined as the attribute value of the node.

[0124] Further, the correction parameter determination module 508 is configured to, for the current round iteration, perform the following calculation steps: calculating the overall radiation inconsistency error of the graph structure in the current round iteration and the radiation inconsistency error of each node according to the attribute values of the respective connection edges; calculating the correction parameter correction amount of each node according to the attribute values of the respective nodes; determining the new correction parameter of each node based on the correction parameter correction amount of each node; calculating the radiation inconsistency error of each node after the new correction parameter is applied to each node; calculating the correctable error of each node based on the radiation inconsistency error of each node after the new correction parameter is applied to each node and the radiation inconsistency error of each node; determining a target node corresponding to the maximum correctable error, adjusting the radiation correction parameter of the target node to the corresponding new correction parameter, updating the attribute values of the nodes and edges associated with the target node after the parameter adjustment, updating the overall radiation inconsistency error of the graph structure based on the updated attribute values; judging whether the difference between the overall radiation inconsistency error of the updated graph structure and the overall radiation inconsistency error of the graph structure in the current round iteration is greater than a threshold value; if yes, continuing to perform the calculation steps in the next round iteration; if no, stopping the iteration, and the radiation correction parameter set is composed of the radiation correction parameters of the nodes.

[0125] Further, the correction parameter determination module 508 is configured to calculate the overall radiation inconsistency error of the graph structure in the current round iteration and the radiation inconsistency error of each node according to the following two formulas:

[0126] ;

[0127] ;

[0128] wherein, represents the overall radiation inconsistency error of the graph structure; represents the radiation inconsistency error of the node ; N and M are respectively the total number of nodes and connection edges in the graph structure; is the number of nodes connected to the node n; is the radiation inconsistency error between the strip and its first adjacent strip.

[0129] Further, the correction parameter determination module 508 is configured to calculate the attribute average value of each node according to the following formula based on the attribute value of each node:

[0130] and ;

[0131] wherein, is the number of nodes connected to the node ; and are the attribute average values of the nodes ;

[0132] The correction parameter correction amount of each node is calculated according to the following formula:

[0133] and ;

[0134] wherein, and are the correction parameter correction amounts of the node .

[0135] Further, the correction parameter determination module 508 is configured to calculate the new correction parameter of each node after correction according to the following formula:

[0136] and ;

[0137] wherein, and are the current radiation correction parameters, and the initial values are 1 and 0 respectively; and are the correction parameter correction amounts of the node ; and are the new correction parameters of the node after correction.

[0138] Further, the correction parameter determination module 508 is configured to calculate the correctable error of each node according to the following formula:

[0139] ;

[0140] wherein, represents the correctable error of the node ; is the radiation inconsistency error level of the node after applying the current correction coefficient, is the radiation inconsistency error of the node after applying the new correction parameter.

[0141] Further, the radiation correction module 510 is further configured to: adopt a product reflectivity image provided by a third party as a correction reference image, and systematically fine tune the radiation correction parameters in the set of radiation correction parameters to obtain optimized radiation correction parameters; the product reflectivity image provided by the third party includes a reflectivity image of Landsat-8 or Sentinel-2; and perform radiation correction on each image strip in sequence according to the following formula to generate a radiation-consistent multi-spectral remote sensing image strip:

[0142] wherein, is the surface reflectivity image strip after radiation consistency correction, is the surface reflectivity image strip before correction, and are the optimized radiation correction parameters, respectively.

[0143] The system provided by the embodiments of the present application has the same implementation principle and technical effects as the foregoing method embodiments. For brevity, the part of the system embodiments not mentioned in the foregoing method embodiments can be referred to the corresponding content in the foregoing method embodiments.

[0144] The embodiments of the present application further provide a computer readable storage medium, which stores computer executable instructions. When the computer executable instructions are called and executed by a processor, the computer executable instructions cause the processor to implement the foregoing method. For details, refer to the foregoing method embodiments, which will not be repeated here.

[0145] The computer program product of the method, device and electronic device provided by the embodiments of the present application includes a computer readable storage medium storing program codes, and the instructions included in the program codes can be used to execute the method described in the foregoing method embodiments. For details, refer to the method embodiments, which will not be repeated here.

[0146] Unless otherwise specifically stated, the relative steps, numerical expressions and values of the components and steps set forth in these embodiments do not limit the scope of the present application.

[0147] If the functions are implemented in the form of software function units and sold or used as independent products, they can be stored in a nonvolatile computer readable storage medium executable by a processor. Based on this understanding, the technical solutions of the present application essentially or the parts that contribute to the prior art or parts of the technical solutions can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes a number of instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present application. The aforementioned storage medium includes: a U disk, a mobile hard disk, a read-only memory (ROM, Read-Only Memory), a random access memory (RAM, Random Access Memory), a magnetic disk or an optical disk, and various media that can store program codes.

[0148] In the description of the present application, it should be noted that the terms "center", "upper", "lower", "left", "right", "vertical", "horizontal", "inner", "outer" and the like indicate the orientation or positional relationship shown in the drawings, and are only for the purpose of facilitating the description of the present application and simplifying the description, and do not indicate or imply that the device or element referred to must have a particular orientation, be constructed and operated in a particular orientation, and therefore cannot be understood as limiting the present application. In addition, the terms "first", "second", "third" are only for the purpose of description, and cannot be understood as indicating or implying relative importance.

[0149] Finally, it should be noted that: the above-described embodiments are only specific embodiments of the present application, used to illustrate the technical solutions of the present application, and are not limited thereto, the protection scope of the present application is not limited thereto, although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that any person skilled in the art within the technical scope disclosed by the present application can modify or easily think of changes to the technical solutions described in the foregoing embodiments, or make equivalent replacements to part of the technical features; and these modifications, changes or replacements do not make the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application, and should be covered within the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

Claims

1. A method for correcting the radiometric consistency of stripes in multispectral remote sensing images, characterized in that, The method includes: Acquire multiple remote sensing image stripes with lateral or directional overlap; A graph structure is constructed based on the coverage of multiple remote sensing image strips; in the graph structure, each remote sensing image strip is a node, and when there is an overlapping area between two remote sensing image strips, a connection edge is established between the two nodes; For each connecting edge in the graph structure, the attribute value of the connecting edge is determined based on the invariant cells identified from the overlapping strip regions corresponding to the connecting edge; for each node in the graph structure, the attribute value of the node is obtained by fitting the radiation inconsistency equation between adjacent nodes based on the invariant cells. Based on the attribute values ​​of nodes and connecting edges in the graph structure, a greedy optimization algorithm is used to iteratively adjust the radiometric correction parameters of each image strip in order to determine the set of radiometric correction parameters. Based on the set of radiometric correction parameters, radiometric correction is performed on each image strip in sequence to generate radiometrically consistent multispectral remote sensing image strips.

2. The method according to claim 1, characterized in that, The step of determining the attribute value of the connection edge based on the invariant cells identified from the overlapping region of the strip corresponding to the connection edge includes: Based on the multivariate change detection method, invariant pixels in the strip overlap region corresponding to the connecting edge are identified; The radiation inconsistency error between adjacent nodes corresponding to the connected edge is calculated using the identified invariant pixels according to the following first specified formula: ,in, Represents adjacent node stripes With strip Radiation inconsistency error, and These represent the first two bands within the overlapping region. The surface reflectance value of each invariant pixel. This represents the total number of invariant pixels within the overlapping regions of the two bands; The radiation inconsistency error between the adjacent nodes is used as the attribute value of the connecting edge.

3. The method according to claim 1, characterized in that, The steps for obtaining node attribute values ​​based on fitting the radiation inconsistency equation between adjacent nodes using invariant pixels include: For nodes and corresponding connections Fitting the radiation inconsistency equation to each connected node yields a set. As a node The set of attributes; wherein, the radiation inconsistency equation is as follows: ; in, and Representing adjacent node stripes respectively With strip Surface reflectance values ​​of all invariant pixels in the overlapping region. The slope The intercept is... For residuals; The set of attributes for each node is determined as the attribute values ​​for that node.

4. The method according to claim 3, characterized in that, Based on the attribute values ​​of nodes and connecting edges in the graph structure, the steps of iteratively adjusting the radiometric correction parameters of each image strip using a greedy optimization algorithm to determine the set of radiometric correction parameters include: For the current iteration, perform the following calculation steps: Based on the attribute values ​​of each connected edge, calculate the overall radiation inconsistency error of the graph structure in the current iteration, as well as the radiation inconsistency error of each node. Calculate the correction amount of the correction parameters for each node based on the attribute values ​​of each node; Based on the correction amount of the correction parameters of each node, determine the new correction parameters of each node after correction; Calculate the radiation inconsistency error at each node after applying the new correction parameters; Based on the radiation inconsistency error of each node after applying the new correction parameters, and the radiation inconsistency error of each node, calculate the correctable error of each node. Identify the target node corresponding to the maximum correctable error, adjust the radiation correction parameters of the target node to the corresponding new correction parameters, update the attribute values ​​of the nodes and edges associated with the target node after parameter adjustment, and update the overall radiation inconsistency error of the graph structure based on the updated attribute values. Determine whether the difference between the overall radiation inconsistency error of the updated graph structure and the overall radiation inconsistency error of the graph structure in the current iteration is greater than a threshold. If so, continue with the calculation steps in the next iteration; If not, stop the iteration and form a set of radiation correction parameters from each node.

5. The method according to claim 4, characterized in that, The steps for calculating the overall radiation inconsistency error of the graph structure in the current iteration, and the radiation inconsistency error of each node, based on the attribute values ​​of each connected edge, include: Calculate the overall radiation inconsistency error and the radiation inconsistency error of each node in the current iteration of the structure shown in the figure below using the following two formulas: ； ； in, This indicates the overall radiation inconsistency error of the diagram structure; Represents a node The radiation inconsistency error; N and M are the total number of nodes and connecting edges in the graph structure, respectively; This represents the number of nodes connected to node n. For strips With strip The radiation inconsistency error between them.

6. The method according to claim 4, characterized in that, The steps for calculating the correction parameter adjustment for each node based on its attribute values ​​include: Calculate the average attribute value of each node using the following formula: and ; in, It is with nodes The number of connected nodes; and For nodes The average value of the attributes; Calculate the correction parameters for each node using the following formula: and ; in, and It is a node The correction parameter adjustment amount.

7. The method according to claim 4, characterized in that, The steps for determining the new correction parameters for each node based on the correction parameter adjustments for each node include: Calculate the new correction parameters for each node using the following formula: and ; in, and These are the current radiation correction parameters, with initial values ​​of 1 and 0 respectively; and For nodes The correction amount of the correction parameter; and For nodes The revised calibration parameters.

8. The method according to claim 4, characterized in that, The steps for calculating the correctable error of each node, based on the radiation inconsistency error after applying the new correction parameters to each node and the radiation inconsistency error of each node, include: Calculate the correctable error of each node using the following formula: ； in, Represents a node Correctable error; For nodes The level of radiation inconsistency error after applying the current correction factors. For nodes Radiation inconsistency error after applying the new correction parameters.

9. The method according to claim 1, characterized in that, The steps of performing radiometric correction on each image strip sequentially based on the radiometric correction parameter set to generate radiometrically consistent multispectral remote sensing image strips include: Using a product reflectance image provided by a third party as a correction reference image, the radiation correction parameters in the radiation correction parameter set are systematically fine-tuned to obtain optimized radiation correction parameters; the product reflectance image provided by the third party includes Landsat-8 or Sentinel-2 reflectance images; The following formula is used to perform radiometric correction on each image strip sequentially to generate radiometrically consistent multispectral remote sensing image strips: ,in, This is a stripe of surface reflectance image after radiometric consistency correction. This is a stripe of surface reflectance image before correction. and These are the optimized radiation correction parameters.

10. A multispectral remote sensing image stripe radiometric consistency correction system, characterized in that, The system includes: The remote sensing image acquisition module is used to acquire multiple remote sensing image strips with lateral or directional overlap. The graph structure construction module is used to construct a graph structure based on the coverage of multiple remote sensing image strips; in the graph structure, each remote sensing image strip is used as a node, and when there is an overlapping area between two remote sensing image strips, a connection edge is established between the two nodes; The attribute value determination module is used to determine the attribute value of each connecting edge in the graph structure based on the invariant pixels identified from the overlapping strip regions corresponding to the connecting edge; and to obtain the attribute value of each node in the graph structure by fitting a radiation inconsistency equation between adjacent nodes based on the invariant pixels. The correction parameter determination module is used to iteratively adjust the radiometric correction parameters of each image strip based on the attribute values ​​of nodes and connecting edges in the graph structure using a greedy optimization algorithm, so as to determine the set of radiometric correction parameters. The radiometric correction module is used to perform radiometric correction on each image strip sequentially based on the set of radiometric correction parameters, thereby generating radiometrically consistent multispectral remote sensing image strips.

Images