Land cover classification method and device, equipment and medium

Abstract

The embodiment of the invention discloses a land cover classification method and device, equipment and a medium. The method comprises the steps of collecting initial land cover types of each grid of a target area in three periods; performing reclassification according to the initial land cover types of each grid in the three periods to obtain standard land cover types of each grid in the target area in the three periods; acquiring remote sensing images of the target area in three periods, and performing spectral analysis based on the remote sensing images to obtain remote sensing feature data and spectral angular distances of each grid of the target area in different periods; and correcting the land cover type of the inverse continuity grid according to the remote sensing feature data and the spectral angular distance of each grid of the target area in different periods, and determining the target land cover type of the inverse continuity grid in the second period. And land cover type correction is carried out on the inverse continuity grids with inconsistent land cover types in three periods, so that the influence of false change on classification is eliminated, and the classification precision is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of hydrological forecasting, and in particular to a land cover classification method and device, equipment and medium. Background Art

[0002] Land cover classification is an important research topic in environmental monitoring, ecological protection and resource management. Faced with the dual challenges of global climate change and intensified human activities, high-precision land cover classification is particularly important in supporting sustainable development decision-making.

[0003] The introduction of multi-temporal datasets allows researchers to more comprehensively capture information on land surface changes, but this also introduces the problem of pseudo-variation. Multi-temporal data refers to data from multiple time periods. Pseudo-variation refers to differences in land cover classification results caused by non-real surface changes, such as sensor errors, changing atmospheric conditions, and seasonal factors. This pseudo-variation can reduce classification accuracy and affect the assessment of true surface changes.

[0004] Therefore, existing methods are usually unable to effectively distinguish between real changes and false changes, resulting in reduced classification accuracy, which in turn affects the scientific nature of environmental management and policy making. Summary of the Invention

[0005] Based on this, it is necessary to propose land cover classification methods, devices, equipment and media to address the above problems, so as to eliminate the influence of pseudo-changes on the classification results of grid land cover as much as possible.

[0006] To achieve the above objectives, the present application provides a first aspect of a land cover classification method, the method comprising:

[0007] Collecting the initial land cover type of each grid in the target area in three preset periods, wherein the three periods include a first period, a second period, and a third period, the first period is before the second period, and the second period is before the third period;

[0008] Based on a preset classification system, reclassifying the initial land cover type of each grid in the target area in the three periods to obtain the standard land cover type of each grid in the target area in the three periods;

[0009] Acquiring remote sensing images of the target area during the three periods, performing spectral analysis based on the remote sensing images, and obtaining remote sensing feature data and spectral angular distances of each grid in the target area during different periods;

[0010] The land cover type of the inverse continuity grid is corrected according to the remote sensing characteristic data and spectral angular distance of each grid in the target area at different periods, and the target land cover type of the inverse continuity grid in the second period is determined, wherein the inverse continuity grid is a grid whose standard land cover type is the same as that of the first period and the third period, and whose standard land cover type is different from that of the first period.

[0011] Furthermore, the land cover type of the inverse continuity grid is corrected according to the remote sensing feature data and spectral angle distance of each grid in the target area at different periods to determine the target land cover type of the inverse continuity grid in the second period, specifically including:

[0012] Determining a first classification result of the land cover type of the inverse continuity grid by comparing the spectral angular distances of the inverse continuity grid at different time periods with a spectral angular distance threshold;

[0013] Constructing a sample data set based on remote sensing feature data and standard land cover types of high continuity grids in different periods, wherein the high continuity grids are grids having the same standard land cover types in the first period, the second period, and the third period;

[0014] Performing model training based on the sample data set to obtain a land cover classification model;

[0015] Inputting the remote sensing characteristic data of the inverse continuity grid at different periods into the land cover classification model for reclassification to obtain a second classification result of the land cover type of the inverse continuity grid;

[0016] The target land cover type of the inverse continuity grid in the second period is determined according to the first classification result and the second classification result by using a voting method.

[0017] Furthermore, the step of comparing the spectral angular distances of the inverse continuity grid at different time periods with a spectral angular distance threshold to determine a first classification result of the land cover type of the inverse continuity grid specifically includes:

[0018] Determining the spectral angular distance threshold according to the spectral angular distance of the high continuity grid;

[0019] Determining a first classification result of the land cover type of the inverse continuity grid by comparing the spectral angular distance of the inverse continuity grid at different periods with the spectral angular distance threshold;

[0020] The first classification result is determined by the following formula:

[0021]

[0022] Where, LC1 is the first classification result; LC T1 , LC T2 and LC T3 are the standard land cover classifications of the target grid in the first period, the second period, and the third period, respectively, wherein the target grid is any one of all the grids in the target area; SAD1, SAD2, and SAD3 are the spectral angular distances of the target grid between the first period and the second period, the spectral angular distances between the second period and the third period, and the spectral angular distances between the third period and the first period, respectively; T SAD1 and T SAD2 They are respectively the spectral angular distance thresholds during the first period and the second period, and the spectral angular distance thresholds during the second period and the third period.

[0023] Furthermore, the spectral angle distance threshold is determined by the following formula:

[0024]

[0025] Where, is the spectral angle distance threshold between the tth period and the t+1th period, and is the mean and standard deviation of the spectral angular distance between period t and period t+1.

[0026] Furthermore, the sample dataset is constructed based on the remote sensing feature data of the high-continuity grid at different periods and the standard land cover type, specifically including:

[0027] Sort the spectral angular distances of the high continuity grid in each period by size, and select the top m target spectral angular distances with the smallest spectral angular distances in each period;

[0028] Obtaining remote sensing characteristic data and standard land cover types of target high continuity grids corresponding to the target spectral angle distance in three periods;

[0029] A sample dataset is constructed based on the remote sensing characteristic data of the target high-continuity grid in three periods and standard land cover types.

[0030] Furthermore, the land cover classification model includes a support vector machine and a decision tree;

[0031] Then, the remote sensing feature data of the inverse continuity grid at different periods is input into the land cover classification model for reclassification to obtain a second classification result of the land cover type of the inverse continuity grid, specifically including:

[0032] Inputting the remote sensing feature data of the inverse continuity grid at different periods into the support vector machine to obtain a support vector machine classification result;

[0033] Inputting the remote sensing feature data of the inverse continuity grid at different periods into the decision tree to obtain a decision tree classification result;

[0034] The second classification result includes a support vector machine classification result and a decision tree classification result;

[0035] The support vector machine classification result is obtained by the following formula:

[0036]

[0037] Where LC2 is the classification result of the support vector machine, N is the number of support vectors, and a i is the Lagrange multiplier, K(·) is the preset kernel function, (x i ,y i ) is the training sample, x is the input remote sensing feature data, and b is the bias term;

[0038] The decision tree classification result is obtained by the following formula:

[0039]

[0040] Where LC3 is the classification result of the decision tree, is the probability of the category in the decision tree leaf node, x is the input remote sensing feature data,

[0041] LC∈D, where D represents all land cover types.

[0042] Furthermore, the determining of the target land cover type of the inverse continuity grid in the second period according to the first classification result and the second classification result by using a voting method specifically includes:

[0043] Determining the target land cover type after the inverse continuity grid is corrected according to the first classification result, the support vector machine classification result, and the decision tree classification result using a voting method;

[0044] The target land cover type of the inverse continuity grid in the second period is determined by the following formula:

[0045]

[0046] Where, LC Final is the target land cover type of the inverse continuity grid in the second period; LC RFis any land cover type; LC1, LC2 and LC3 are respectively the first classification result, the support vector machine classification result and the decision tree classification result of the inverse continuity grid.

[0047] To achieve the above-mentioned purpose, the second aspect of the present application provides a land cover classification device, the device comprising: a data acquisition module, a reclassification module and a data calculation module;

[0048] The data collection module is used to collect the initial land cover type of each grid in the target area in three preset periods, wherein the three periods include a first period, a second period and a third period, the first period is before the second period, and the second period is before the third period;

[0049] The reclassification module is configured to reclassify the initial land cover type of each grid in the target area during the three periods based on a preset classification system to obtain a standard land cover type of each grid in the target area during the three periods;

[0050] The data calculation module is used to obtain remote sensing images of the target area in the three periods, perform spectral analysis based on the remote sensing images, and obtain remote sensing feature data and spectral angular distance of each grid in the target area at different periods;

[0051] The land cover type of the inverse continuity grid is corrected according to the remote sensing characteristic data and spectral angular distance of each grid in the target area at different periods, and the target land cover type of the inverse continuity grid in the second period is determined, wherein the inverse continuity grid is a grid whose standard land cover type is the same as that of the first period and the third period, and whose standard land cover type is different from that of the first period.

[0052] To achieve the above-mentioned objectives, the third aspect of the present application provides a computer-readable storage medium storing a computer program. When the computer program is executed by a processor, the processor executes the steps of the method described in the first aspect.

[0053] To achieve the above-mentioned objectives, the fourth aspect of the present application provides a computer device, comprising a memory and a processor, wherein the memory stores a computer program, and when the computer program is executed by the processor, the processor executes the steps of the method described in the first aspect.

[0054] The embodiments of the present invention have the following beneficial effects:

[0055] An embodiment of the present invention provides a land cover classification method, comprising: collecting initial land cover types for each grid in a target area during three preset periods, wherein the three periods include a first period, a second period, and a third period, wherein the first period precedes the second period, and the second period precedes the third period; reclassifying the initial land cover types for each grid in the target area during the three periods based on a preset classification system to obtain standard land cover types for each grid in the target area during the three periods; acquiring remote sensing images of the target area during the three periods, performing spectral analysis based on the remote sensing images to obtain remote sensing feature data and spectral angular distances for each grid in the target area during the different periods; and correcting the land cover types of inverse continuity grids based on the remote sensing feature data and spectral angular distances for each grid in the target area during the different periods to determine the target land cover type for the inverse continuity grid during the second period, wherein the inverse continuity grid is a grid having the same standard land cover type for the first and third periods, and having a different standard land cover type for the second period than for the first period. By correcting the land cover types of inverse continuity grids whose land cover types are inconsistent during the three periods, the present invention eliminates the impact of pseudo-changes on classification, thereby improving classification accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0056] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0057] in:

[0058] Figure 1 Schematic diagram of the process of the land cover classification method according to an embodiment of the present invention;

[0059] Figure 2 This is a structural block diagram of the land cover classification device according to an embodiment of the present invention.

[0060] Figure 3 2 is a diagram showing the internal structure of a computer device in an embodiment of the present invention. DETAILED DESCRIPTION

[0061] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.

[0062] In order to eliminate the influence of pseudo-change on land cover classification results, the embodiment of the present invention proposes a land cover classification method, which corrects the land cover type of grids that may have pseudo-changes, making the classification results more accurate. Figure 1 , Figure 1 Schematic diagram of a land cover classification method according to an embodiment of the present invention. The method includes:

[0063] Step 110 : Collect the initial land cover types of each grid in the target area in three preset periods, wherein the three periods include a first period, a second period, and a third period, the first period is before the second period, and the second period is before the third period.

[0064] First, the target area that needs to be corrected for land cover type is selected and the target area is divided into grids so that the land cover type can be corrected for each grid separately.

[0065] Next, a multi-period land use remote sensing monitoring dataset (CNLUCC) is downloaded and a multi-temporal dataset for the target area is clipped from the CNLUCC dataset. The multi-temporal dataset contains land cover data from multiple periods, and the land cover data includes the land cover type for each grid. In this embodiment of the present invention, land cover data for the target area from three different periods are clipped to determine which grids need to be corrected based on the land cover data from the three different periods.

[0066] The three periods include a first period, a second period, and a third period, where the first period is before the second period, and the second period is before the third period. For example, the three periods may be 1980, 1990, and 2000.

[0067] Step 120 : Based on a preset classification system, reclassify the initial land cover types of each grid in the target area in the three periods to obtain the standard land cover types of each grid in the target area in the three periods.

[0068] Based on the actual classification needs and application scenarios, a classification system suitable for the target area is determined. For example, the classification of the target area can include six categories: cultivated land, forest land, grassland, water area, urban and rural industrial and mining residential land, and unused land. The reclassification work is to merge the grids with similar land cover types in the target area obtained in step 110 into a preset classification system. For example, in the multi-temporal dataset, the land cover type of grid A in the target area is river, and the land cover type of grid B is lake. The land cover types of grids A and B can be reclassified as water area type 1; the land cover type of grid C in the target area is evergreen coniferous forest, and the land cover type of grid D is broad-leaved forest. The land cover types of grids C and D can be reclassified as forest area type 1.

[0069] By reclassifying the land cover type of each grid in the target area, the consistency analysis of subsequent grids and the correction of land cover types can be facilitated.

[0070] Step 130: Acquire remote sensing images of the target area at three time periods, perform spectral analysis based on the remote sensing images, and obtain remote sensing feature data and spectral angular distance of each grid in the target area at different time periods.

[0071] In an embodiment of the present invention, remote sensing images of the target area at three periods are acquired based on Landsat images, and spectral angular distance and remote sensing feature data of each grid at different periods are calculated based on the remote sensing images of the target area at three periods.

[0072] The smaller the spectral angle distance, the smaller the spectral change of the grid, indicating that the land cover type will basically not change. Therefore, the size of the spectral angle distance of the target area at different times can be used to determine whether there is a false change in the grid.

[0073] In the embodiment of the present invention, the calculation formula of the spectral angular distance is:

[0074]

[0075] Where SAD is the spectral angular distance, n is the number of bands, x i and y i are the reflectance of the i-th band of the target grid in the remote sensing images of any two periods, and the target grid is any one of all the grids in the target area.

[0076] Remote sensing feature data can include spectral and index features. Spectral features are direct or indirect information extracted from the spectral data of remote sensing images, reflecting the spectral response characteristics of land features in different bands. Index features are calculated by mathematically combining multiple bands and are typically used to enhance or highlight the characteristics of specific land features. Therefore, changes in the spectral and index features of each grid in the target area over time can be used to correct grid land cover types that exhibit spurious changes.

[0077] Step 140: Correct the land cover type of the inverse continuity grid based on the remote sensing characteristic data and spectral angular distance of each grid in the target area at different periods, and determine the target land cover type of the inverse continuity grid in the second period. The inverse continuity grid is a grid whose standard land cover type is the same as that of the first period and the third period, and whose standard land cover type is different from that of the first period.

[0078] In this embodiment of the present invention, all grid types are divided into three categories: high continuity grids, inverse continuity grids, and non-inverse continuity grids. A high continuity grid is a grid with the same standard land cover type for all three periods; an inverse continuity grid is a grid with the same standard land cover type for the first and third periods, but a different standard land cover type for the second period; and a non-inverse continuity grid is any grid other than a high continuity grid and an inverse continuity grid.

[0079] The inverse continuity grid is considered a grid with potential spurious changes. Remote sensing feature data and spectral angular distances from different periods are used to correct the land cover type of the inverse continuity grid in the second period. By correcting the land cover type of the inverse continuity grid with inconsistent land cover types in the three periods, the impact of spurious changes on classification is eliminated, thereby improving classification accuracy.

[0080] In one embodiment of the present invention, two land cover type reclassification methods are proposed: one is based on spectral angular distance judgment, and the other is model judgment. The final land cover type is determined by the judgment results of the two methods to improve the accuracy of the classification results. Based on this, step 140, based on the remote sensing feature data and spectral angular distance of each grid in the target area at different time periods, the land cover type of the inverse continuity grid is corrected, and the specific implementation method for determining the target land cover type of the inverse continuity grid in the second time period includes:

[0081] Step 410: Determine the first classification result of the land cover type of the inverse continuity grid by comparing the spectral angular distance of the inverse continuity grid at different time periods with the spectral angular distance threshold.

[0082] In the embodiment of the present invention, the first classification result of the land cover type of the inverse continuity grid is determined by comparing the spectral angular distances of the inverse continuity grid at different time periods with a spectral angular distance threshold.

[0083] Specifically, Step 410, based on the comparison of the spectral angular distances of the inverse continuity grid at different time periods with the spectral angular distance threshold, determines the first classification result of the land cover type of the inverse continuity grid, specifically including:

[0084] Step 411: Determine the spectral angular distance threshold according to the spectral angular distance of the high-continuity grid.

[0085] Since the high-continuity grids are grids with the same standard land cover types in the three periods, it can be considered that there is no pseudo-change in the high-continuity grids or the pseudo-change is small and negligible. Therefore, the spectral angular distance threshold can be determined based on the spectral angular distance of the high-continuity grids to ensure the accuracy of the classification results.

[0086] In the embodiment of the present invention, the spectral angular distance threshold is determined by the following formula:

[0087]

[0088] Where, is the spectral angle distance threshold between the tth period and the t+1th period, and is the mean and standard deviation of the spectral angular distance between period t and period t+1.

[0089] Step 412: Determine the first classification result of the land cover type of the inverse continuity grid by comparing the spectral angular distance of the inverse continuity grid at different periods with the spectral angular distance threshold. The first classification result is determined by the following formula:

[0090]

[0091] Where, LC1 is the first classification result; LC T1 , LC T2 and LC T3 are the standard land cover classifications of the target grid in the first, second and third periods, respectively, where the target grid is any grid in all the grids in the target area; SAD1, SAD2 and SAD3 are the spectral angular distances of the target grid between the first and second periods, the spectral angular distances between the second period and the third period, and the spectral angular distances between the third period and the first period, respectively; T SAD1 and T SAD2 are the spectral angular distance thresholds between the first and second periods, and the spectral angular distance thresholds between the second and third periods, respectively.

[0092] For ease of understanding, the following example illustrates the above formula, but is not limited to this example: First, based on the Landsat 4-5TM images of 1980, 1990 and 2000, the spectral angular distances of all grids in the target area between 1980-1990, 1990-2000, and 1980-2000 are calculated, which are expressed as SAD1, SAD2, and SAD3 respectively. The spectral angular distance thresholds of high continuity grids in the two periods of 1980-1990 and 1990-2000 are determined, which are T SAD1 and T SAD2 The obtained spectral angular distance and spectral angular distance threshold are then substituted into the first classification result formula to obtain the first classification result, so as to modify the land cover type of the inverse continuity grid in 1990 based on the first classification result.

[0093] Step 420, construct a sample data set based on the remote sensing characteristic data and standard land cover types of the high continuity grid in different periods, wherein the high continuity grid is a grid with the same standard land cover types in the first period, the second period and the third period.

[0094] Considering that the standard land cover types of high-continuity grids are the same in the three periods, there is no pseudo-change or the pseudo-change is small and negligible. The relevant data of high-continuity grids are selected as training samples for model training, and the final trained model has higher accuracy.

[0095] Specifically, the standard land cover types of high-continuity grids in three periods and the remote sensing feature data corresponding to the standard land cover types in each period, such as spectral features and index features, are obtained. A sample data set is constructed based on the standard land cover types and remote sensing feature data of high-continuity grids in three periods to train the model.

[0096] In one embodiment of the present invention, Step 420, constructing a sample dataset based on remote sensing feature data of high-continuity grids at different periods and standard land cover types, specifically includes:

[0097] Step 421. Sort the spectral angular distances of the high continuity grids in each period by size, and select the top m target spectral angular distances with the smallest spectral angular distances in each period; obtain the remote sensing feature data and standard land cover types of the target high continuity grids corresponding to the target spectral angular distances in three periods.

[0098] Step 422: Construct a sample data set based on the remote sensing characteristic data and standard land cover types of the target high-continuity grid in three periods.

[0099] Specifically, the spectral angular distances of high continuity grids in each period are sorted from small to large, and the high continuity grids whose spectral angular distances are in the top m or top m% in each period are selected. The sample dataset S is constructed based on the remote sensing feature data of the high continuity grids and the standard land cover types, which can be expressed by the following formula:

[0100] S={(x i ,y i )|x i ∈X,SAD(x i ,TT)≤P m}

[0101] Where X is the set of all high continuity grids, y i is the grid x i The land cover type is determined by multi-temporal datasets; SAD(x i ,t) represents the grid x iThe spectral angular distance in time period T; P 20 is the top m% quantile of the spectral angular distance of all high continuity grids in time period TT.

[0102] Step 430: Perform model training based on the sample data set to obtain a land cover classification model.

[0103] In an embodiment of the present invention, the land cover classification model includes a support vector machine (SVM) and a decision tree. Training samples and validation samples are randomly extracted from a sample dataset at a preset ratio to obtain trained SVM and decision trees. The SVM and decision trees are used to reclassify the land cover of the target area. The SVM classification result is obtained by calculating the weighted sum of the support vectors and finding the maximum classification decision value. The decision tree classification result is obtained by recursively partitioning the input features within the decision tree until they fall into a leaf node and selecting the category with the highest proportion within that node.

[0104] Step 440: Input the remote sensing characteristic data of the inverse continuity grid at different periods into the land cover classification model for reclassification to obtain the second classification result of the land cover type of the inverse continuity grid.

[0105] Specifically, the second classification result is determined by using a support vector machine and a decision tree, and the second classification result includes a support vector machine classification result and a decision tree classification result.

[0106] Firstly, the remote sensing feature data of the inverse continuity grid at different periods are input into the support vector machine to obtain the support vector machine classification results.

[0107] The support vector machine classification result is obtained by the following formula:

[0108]

[0109] Where LC2 is the classification result of support vector machine, N is the number of support vectors, and a i is the Lagrange multiplier, K(·) is the preset kernel function, (x i ,y i ) is the training sample, x is the input remote sensing feature data, and b is the bias term.

[0110] Secondly, the remote sensing feature data of the inverse continuity grid at different periods are input into the decision tree to obtain the decision tree classification results.

[0111] The decision tree classification result is obtained by the following formula:

[0112]

[0113] Where LC3 is the decision tree classification result, is the probability of the category in the decision tree leaf node, x is the input remote sensing feature data,

[0114] LC∈D, where D represents all land cover types.

[0115] Step 450: Determine the target land cover type of the inverse continuity grid in the second period based on the first classification result and the second classification result using a voting method.

[0116] Specifically: The voting method is used to determine the target land cover type of the inverse continuity grid in the second period according to the first classification results, the support vector machine classification results and the decision tree classification results.

[0117] The target land cover type of the second period of the inverse continuity grid is determined by the following formula:

[0118]

[0119] Where, LC Final is the target land cover type after the inverse continuity grid is corrected; LC RF is any land cover type; LC1, LC2 and LC3 are the first classification result of the inverse continuity grid, the support vector machine classification result and the decision tree classification result respectively.

[0120] In one embodiment of the present invention, a land cover classification device is also provided, which can be referred to Figure 2 , Figure 2 2 is a structural block diagram of a land cover classification device according to an embodiment of the present invention. The device includes: a data acquisition module 201 , a reclassification module 202 and a data calculation module 203 .

[0121] The data collection module 201 is used to collect the initial land cover type of each grid in the target area in three preset periods, wherein the three periods include a first period, a second period, and a third period, the first period is before the second period, and the second period is before the third period;

[0122] The reclassification module 202 is configured to reclassify the initial land cover types of each grid in the target area in the three periods based on a preset classification system to obtain the standard land cover types of each grid in the target area in the three periods;

[0123] The data calculation module 203 is used to obtain remote sensing images of the target area at three periods, perform spectral analysis based on the remote sensing images, and obtain remote sensing feature data and spectral angular distance of each grid in the target area at different periods;

[0124] The land cover type of the inverse continuity grid is corrected according to the remote sensing characteristic data and spectral angle distance of each grid in the target area at different periods, and the target land cover type of the second period of the inverse continuity grid is determined. Among them, the inverse continuity grid is a grid with the same standard land cover type in the first and third periods, and a different standard land cover type in the second period from that in the first period.

[0125] The land cover classification device proposed in this paper treats the inverse continuity grid as a grid that may contain spurious changes. It uses remote sensing feature data and spectral angular distances from different time periods to correct the land cover type of the inverse continuity grid in the second period. By correcting the land cover type of the inverse continuity grid with inconsistent land cover types across the three time periods, the impact of spurious changes on classification is eliminated, thereby improving classification accuracy.

[0126] Figure 3 FIG1 shows the internal structure of a computer device in one embodiment of the present invention. The computer device can be a terminal or a system. Figure 3 As shown, the computer device includes a processor, a memory, and a network interface connected via a system bus. The memory includes a non-volatile storage medium and an internal memory. The non-volatile storage medium of the computer device stores an operating system and may also store a computer program. When the computer program is executed by the processor, the processor can implement the various steps in the above method embodiment. The internal memory may also store a computer program. When the computer program is executed by the processor, the processor can implement the various steps in the above method embodiment. It will be understood by those skilled in the art that Figure 3 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than shown in the figure, or combine certain components, or have a different component arrangement.

[0127] In one embodiment, a computer device is provided, including a memory and a processor. The memory stores a computer program. When the computer program is executed by the processor, the processor executes each step in the above method embodiment.

[0128] In one embodiment, a computer-readable storage medium is provided, which stores a computer program. When the computer program is executed by a processor, the processor executes the steps in the above method embodiment.

[0129] Those skilled in the art will appreciate that all or part of the processes in the above-mentioned embodiments can be implemented by instructing the relevant hardware through a computer program. The program can be stored in a non-volatile computer-readable storage medium. When the program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM).

[0130] The technical features of the above embodiments can be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0131] The above-described embodiments merely represent several implementation methods of the present application. While the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the present invention. It should be noted that a person of ordinary skill in the art may make various modifications and improvements without departing from the spirit of the present invention, all of which fall within the scope of protection of the present invention. Therefore, the scope of protection of the present invention shall be determined by the appended claims.

Claims

1. A land cover classification method, characterized in that: The method comprises: Collecting the initial land cover type of each grid in the target area in three preset periods, wherein the three periods include a first period, a second period, and a third period, the first period is before the second period, and the second period is before the third period; Based on a preset classification system, reclassify the initial land cover type of each grid in the target area in the three periods to obtain the standard land cover type of each grid in the target area in the three periods; Acquiring remote sensing images of the target area during the three periods, performing spectral analysis based on the remote sensing images, and obtaining remote sensing feature data and spectral angular distances of each grid in the target area during different periods; The land cover type of the inverse continuity grid is corrected according to the remote sensing characteristic data and spectral angular distance of each grid in the target area at different periods, and the target land cover type of the inverse continuity grid in the second period is determined, wherein the inverse continuity grid is a grid whose standard land cover type is the same as that of the first period and the third period, and whose standard land cover type is different from that of the first period.

2. The method according to claim 1, wherein The step of correcting the land cover type of the inverse continuity grid according to the remote sensing feature data and spectral angle distance of each grid in the target area at different periods, and determining the target land cover type of the inverse continuity grid in the second period, specifically includes: Determining a first classification result of the land cover type of the inverse continuity grid by comparing the spectral angular distances of the inverse continuity grid at different time periods with a spectral angular distance threshold; Constructing a sample data set based on remote sensing feature data and standard land cover types of high continuity grids in different periods, wherein the high continuity grids are grids with the same standard land cover types in the first period, the second period, and the third period; Performing model training based on the sample data set to obtain a land cover classification model; Inputting the remote sensing characteristic data of the inverse continuity grid at different periods into the land cover classification model for reclassification to obtain a second classification result of the land cover type of the inverse continuity grid; The target land cover type of the inverse continuity grid in the second period is determined according to the first classification result and the second classification result by using a voting method.

3. The method according to claim 2, wherein The determining of the first classification result of the land cover type of the inverse continuity grid by comparing the spectral angular distances of the inverse continuity grid at different time periods with the spectral angular distance threshold value specifically includes: Determining the spectral angular distance threshold according to the spectral angular distance of the high continuity grid; Determining a first classification result of the land cover type of the inverse continuity grid by comparing the spectral angular distance of the inverse continuity grid at different periods with the spectral angular distance threshold; The first classification result is determined by the following formula: Where, LC1 is the first classification result; LC T1 , LC T2 and LC T3 are the standard land cover classifications of the target grid in the first period, the second period, and the third period, respectively, wherein the target grid is any one of all the grids in the target area; SAD1, SAD2, and SAD3 are the spectral angular distances of the target grid between the first period and the second period, the spectral angular distances between the second period and the third period, and the spectral angular distances between the third period and the first period, respectively; T SAD1 and T SAD2 They are respectively the spectral angular distance thresholds during the first period and the second period, and the spectral angular distance thresholds during the second period and the third period.

4. The method according to claim 3, wherein The spectral angle distance threshold is determined by the following formula: Where, is the spectral angle distance threshold between the tth period and the t+1th period, and is the mean and standard deviation of the spectral angular distance between period t and period t+1.

5. The method according to claim 2, wherein The sample dataset is constructed based on the remote sensing characteristic data of high-continuity grids at different periods and standard land cover types, specifically including: Sort the spectral angular distances of the high continuity grid in each period by size, and select the top m target spectral angular distances with the smallest spectral angular distances in each period; Obtaining remote sensing characteristic data and standard land cover types of target high continuity grids corresponding to the target spectral angle distance in three periods; A sample dataset is constructed based on the remote sensing characteristic data of the target high-continuity grid in three periods and standard land cover types.

6. The method according to claim 2, wherein The land cover classification model includes a support vector machine and a decision tree; Then, the remote sensing feature data of the inverse continuity grid at different periods is input into the land cover classification model for reclassification to obtain a second classification result of the land cover type of the inverse continuity grid, specifically including: Inputting the remote sensing feature data of the inverse continuity grid at different periods into the support vector machine to obtain a support vector machine classification result; Inputting the remote sensing feature data of the inverse continuity grid at different periods into the decision tree to obtain a decision tree classification result; The second classification result includes a support vector machine classification result and a decision tree classification result; The support vector machine classification result is obtained by the following formula: Where LC2 is the classification result of the support vector machine, N is the number of support vectors, and a i is the Lagrange multiplier, K(·) is the preset kernel function, (x i ,y i ) is the training sample, x is the input remote sensing feature data, and b is the bias term; The decision tree classification result is obtained by the following formula: Where LC3 is the classification result of the decision tree, is the probability of the category in the decision tree leaf node, x is the input remote sensing feature data, LC∈D, where D represents all land cover types.

7. The method according to claim 6, wherein Determining the target land cover type of the inverse continuity grid in the second period according to the first classification result and the second classification result by using a voting method specifically includes: Determining the target land cover type after the inverse continuity grid is corrected according to the first classification result, the support vector machine classification result, and the decision tree classification result using a voting method; The target land cover type of the inverse continuity grid in the second period is determined by the following formula: Where, LC Final is the target land cover type of the inverse continuity grid in the second period; LC RF is any land cover type; LC1, LC2 and LC3 are respectively the first classification result, the support vector machine classification result and the decision tree classification result of the inverse continuity grid.

8. A land cover classification device, characterized in that: The device includes: a data acquisition module, a reclassification module and a data calculation module; The data collection module is used to collect the initial land cover type of each grid in the target area in three preset periods, wherein the three periods include a first period, a second period and a third period, the first period is before the second period, and the second period is before the third period; The reclassification module is configured to reclassify the initial land cover type of each grid in the target area during the three periods based on a preset classification system to obtain a standard land cover type of each grid in the target area during the three periods; The data calculation module is used to obtain remote sensing images of the target area in the three periods, perform spectral analysis based on the remote sensing images, and obtain remote sensing feature data and spectral angular distance of each grid in the target area at different periods; The land cover type of the inverse continuity grid is corrected according to the remote sensing characteristic data and spectral angular distance of each grid in the target area at different periods, and the target land cover type of the inverse continuity grid in the second period is determined, wherein the inverse continuity grid is a grid whose standard land cover type is the same as that of the first period and the third period, and whose standard land cover type is different from that of the first period.

9. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the processor is caused to perform the steps of the method according to any one of claims 1 to 7.

10. A computer device comprising a memory and a processor, characterized in that: The memory stores a computer program, and when the computer program is executed by the processor, the processor is caused to perform the steps of the method according to any one of claims 1 to 7.

Images