Irrigation area land utilization classification method and system

Abstract

The invention relates to an irrigation area land utilization classification method and system. A dynamic sample migration and logic conversion mechanism is provided through the classification method and system, cross-year reuse of high-precision samples is achieved through spectrum similarity measurement and Euclidean distance, and the problem of long-time sequence sample labeling is solved; and introducing a space-time logic constraint mechanism, defining a land utilization type conversion rule, and automatically correcting an unreasonable classification result. The precision and rationality of long-time sequence land utilization classification can be remarkably improved, and a new thought and a reliable reference method are provided for land utilization change monitoring.

Description

Technical Field

[0001] The present invention relates to the technical field of remote sensing image analysis and geographic information systems, and in particular to a method and system for classifying land use in irrigated areas. The method and system are particularly suitable for long-term dynamic monitoring of land use in complex irrigated area ecosystems. Background Art

[0002] Land use classification is a key research topic in geographic information systems (GIS) and remote sensing technology, and is crucial for land resource management, ecological and environmental monitoring, and urban planning. In recent years, the Google Earth Engine (GEE) platform, with its powerful data storage and computing capabilities, has become a crucial tool for remote sensing data processing and analysis. The GEE platform facilitates the acquisition and processing of long-term remote sensing imagery data and, combined with machine learning algorithms, enables land use classification.

[0003] However, traditional land-use classification methods have two major limitations: First, sample labeling relies on field surveys conducted in the current year, and classification based on remote sensing imagery from a single year is difficult to support long-term analysis. Second, they ignore the spatiotemporal continuity of land-use types, leading to illogical classification results such as "water bodies suddenly changing to construction land." Therefore, how to introduce a logical transformation mechanism into the classification process to ensure the rationality of classification results is a key issue in current land-use classification research. Summary of the Invention

[0004] In view of the defects in the prior art, the purpose of the present invention is to provide an irrigation area land use classification method, which can solve the problems of low accuracy of long-term land use classification and lack of logical conversion mechanism in the prior art.

[0005] In order to achieve the above purpose, the technical solution adopted by the present invention is:

[0006] A method for classifying land use in irrigation areas, characterized by comprising the following steps:

[0007] S1, based on a specific year, obtains Landsat images of that year and the previous n years, performs radiation correction, cloud mask and minimum cloud synthesis, extracts B1-B7 bands and calculates NDVI, NDVI1, NDVI2, NDWI and EVI indices of spring, summer, autumn and winter every year through the following formulas (1)-(5), and constructs a multi-feature time series dataset:

[0008]

[0009] In the above formulas, NDVI is the normalized difference vegetation index; NDVI1 is the custom vegetation index 1; NDVI2 is the custom vegetation index 2; NDWI is the water body index; EVI is the enhanced vegetation index; NIR is the reflectance of the near-infrared band; RED is the reflectance of the red band; GREEN is the reflectance of the green band; SWIR1 is the reflectance of the shortwave infrared band 1; SWIR3 is the reflectance of the shortwave infrared band 3;

[0010] The Landsat images of the above historical years are: multispectral data and training sample data of the Landsat series satellites acquired and processed online by the GEE platform;

[0011] S2, based on the Landsat image of a specific year described in step S1, labeling the land use types therein and generating sample points;

[0012] The land use types include: construction land, farmland, natural land, forest and grassland, and water bodies;

[0013] S3, for the Landsat images n years before a specific year, use a similar feature matching method to migrate the sample points of the specific year obtained in step S2 to the Landsat images of each year in the n years before the specific year to generate a cross-year training set;

[0014] S4, using the migration samples of each of the n years before the specific year obtained in step S3 as input, outputting the land use classification results of each of the n years before the specific year year by year based on the random forest algorithm;

[0015] S5, construct a land use type conversion matrix and define prohibited conversion rules. If the land use type classification result of a certain year in the n years before the specific year violates the above prohibited conversion rules, the model will be retroactively corrected to the result of the previous year.

[0016] Based on the above solution, the construction of the multi-feature time series dataset in step S1 specifically includes:

[0017] (1) Using the Level-2 surface reflectance product of Landsat 5 / 7 / 8, the image quality of the study area was screened and images with cloud cover less than 20% were selected;

[0018] (2) Using the MASK algorithm for cloud detection and masking;

[0019] (3) The images are synthesized using the median synthesis method, and the missing areas are filled using other sensors in the same period. Four images of spring, summer, autumn and winter are synthesized every year.

[0020] Based on the above solution, the migration of sample points in a specific year to Landsat images of each of the n years before the specific year in step S3 specifically includes:

[0021] (1) Calculate the spectral similarity between the migration sample and the target year image through spectral angular distance and Euclidean distance;

[0022] (2) Set the spectral angle distance threshold to >0.9 and the Euclidean distance threshold to <0.15 to eliminate abnormal samples caused by changes in surface cover and screen samples that meet the requirements.

[0023] Based on the above solution, the conversion rules described in step S5 include:

[0024] (1) Allowable conversion types: construction land → natural land, forest and grassland → natural land / farmland, natural land → construction land / forest and grassland / farmland / water body, farmland → forest and grassland / natural land, water body → natural land;

[0025] (2) Prohibited conversion types: construction land → forest and grassland / farmland / water body; forest and grassland → construction land / water body; farmland → construction land / water body; water body → forest and grassland / farmland / construction land.

[0026] Another object of the present invention is to provide an irrigation area land use classification system.

[0027] In order to achieve the above purpose, the technical solution adopted by the present invention is:

[0028] An irrigation area land use classification system using the above land use classification method is characterized by comprising the following interconnected modules:

[0029] Data preprocessing module: used for image correction, feature extraction and time series data set construction;

[0030] Sample management module: realizes sample labeling, migration and quality assessment;

[0031] Classification engine module: integrates random forest algorithm and GEE parallel computing interface;

[0032] Logic verification module: loads the conversion rule library and performs logic conversion correction.

[0033] Based on the above solution, the built-in rule library of the logic verification module supports user-defined transformation matrices and provides a visual interface to display the spatial differences before and after correction.

[0034] Yet another object of the present invention is to provide an electronic device and a computer-readable storage medium.

[0035] In order to achieve the above purpose, the technical solution adopted by the present invention is:

[0036] An electronic device, characterized in that: the electronic device includes a memory, a processor and a computer program stored in the memory, and when the processor executes the computer program, the steps of the above-mentioned irrigation area land use classification method are implemented.

[0037] A computer-readable storage medium stores a computer program, wherein the computer program implements the steps of the above method when executed by a processor.

[0038] The irrigation area land use classification method and system described in the present invention have the following beneficial effects:

[0039] The present invention provides a dynamic sample migration and logical conversion mechanism: this mechanism uses spectral similarity measurement and Euclidean distance to achieve cross-year reuse of high-precision samples, solving the problem of long-term sample labeling; this mechanism can significantly improve the accuracy and rationality of long-term land use classification, providing new ideas and reliable reference methods for land use change monitoring.

[0040] The present invention also provides a spatiotemporal logical constraint model, which can be used to construct a land use conversion probability matrix and automatically identify and correct classification results that violate the laws of land type evolution. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] The present invention has the following accompanying drawings:

[0042] Figure 1 A flow chart of a land use classification method for irrigation areas provided by the present invention;

[0043] Figure 2 A schematic diagram of the hardware structure of a possible electronic device provided by the present invention;

[0044] Figure 3 A schematic diagram of the hardware structure of a possible computer-readable storage medium provided by the present invention;

[0045] Figure 4 This is a comparison chart of land use classification accuracy before and after logical correction finally obtained in an embodiment of the present invention;

[0046] Figure 5 This is a structural diagram of the irrigation area land use classification system provided by the present invention. DETAILED DESCRIPTION

[0047] The specific embodiments of the present invention are described in detail below with reference to the accompanying drawings, but it should be understood that the protection scope of the present invention is not limited by the specific embodiments.

[0048] Unless expressly stated otherwise, throughout the specification and claims, the term "comprise" or variations such as "include" or "comprising", etc., will be understood to include the stated elements or components but not to exclude other elements or other components.

[0049] Example:

[0050] S1, such as Figure 1 As shown in the figure, based on the GEE platform, Landsat series images from 2000 to 2024 were obtained, and radiometric correction, cloud mask and minimum cloud cover synthesis were performed. The B1-B7 bands were extracted and the NDVI, NDVI1, NDVI2, NDWI and EVI indices of spring, summer, autumn and winter were calculated respectively by formulas (1)-(5) to construct a multi-feature time series dataset.

[0051]

[0052] The specific process of constructing a multi-feature time series dataset is as follows:

[0053] Based on the Google Earth Engine (GEE) platform, we acquired Landsat images from 2000 to 2024, including Landsat 5 (2000-2006, 2008-2010), Landsat 7 (2007, 2011-2013), and Landsat 8 (2014-2023). We then performed the following data preprocessing:

[0054] Landsat imagery was converted to TOA reflectivity to eliminate atmospheric effects and improve data consistency across years. Cloud masking was performed using the Landsat imagery's built-in quality mask. Imagery from 2000 to 2024 was quality-screened based on the Landsat cloud mask algorithm. Images with less than 20% cloud cover were used directly. Images were synthesized using the median composite method, with missing areas filled in using images from other sensors from the same period. Due to damage to Landsat 7 satellite bands, the Landsat 7 imagery was strip-repaired, and missing areas were filled in using images from other sensors from the same period. If no satellite images were available from other sensors, images from a similar time in the previous year were used first. If any missing areas remained, images from a similar time in the next year were used.

[0055] S2, generates sample points based on high-resolution images in 2024.

[0056] Using the high-resolution remote sensing images provided by Google Earth and combined with manual visual interpretation, the main land use types in 2024 were marked, including: construction land (buildings, urban areas), farmland (arable land), natural land (bare land, desert), forest and grassland (forests, grasslands) and water bodies (lakes, rivers, reservoirs), etc., to generate sample points.

[0057] S3, sample migration to images from 2000 to 2023.

[0058] The feature similarity matching algorithm (spectral angular distance, Euclidean distance) is used to project the 2024 sample points onto the historical imagery. The eigenvectors (B1-B7) of the 2024 sample points are calculated, and the spectral angular distance threshold is set to be greater than 0.9 and the Euclidean distance threshold is set to be less than 0.15. The pixels with the most similar spectra in the historical imagery are found and migrated if they meet the threshold requirements as the corresponding samples for that year. A cross-year training sample set from 2000 to 2023 is constructed for subsequent land use classification.

[0059] S4, based on random forest classification and logistic correction.

[0060] A random forest classifier (RF) was used for classification with the following parameters: number of decision trees: 100, sampling ratio: 70%, and input bag fraction: 0.8. The classification results for the current year were compared with those for the previous year to identify areas where land use type changes had occurred. Masks were added to these areas and the classification results were compared against Table 1. Land use type conversion results that met the requirements were retained. For those that did not meet the requirements, the results from the previous year were filled in to ensure the legitimacy of the land use change.

[0061] The accuracy of the crop classification results after correction was verified using the confusion matrix and visual points collected from Google Earth high-resolution images of the same period. The overall accuracy and Kappa coefficient were calculated: the overall accuracy is equal to the ratio of correctly classified sampling points to the total number of sampling points; the Kappa coefficient represents the error reduction ratio of this classification compared to completely random classification. When the Kappa coefficient is 1, it means that the two judgment results are completely consistent, when it is -1, it means that there is no consistency at all. When it is greater than 0.75, it means that a fairly satisfactory degree of consistency has been achieved. The evaluation results are as follows: Figure 4 shown.

[0062] Table 1 Land use type conversion rules

[0063]

[0064] like Figure 2As shown, an embodiment of the present invention provides an electronic device, including a memory 1310, a processor 1320, and a computer program 1311 stored in the memory 1310 and executable on the processor 1320. When the processor 1320 executes the computer program 1311, the following steps are implemented:

[0065] S1, based on the GEE platform, remote sensing image data from 2000 to 2024 were screened and processed, and image synthesis was performed using multiple bands (B1-B7) and vegetation indices (such as NDVI, NDWI, etc.);

[0066] S2. Generate sample points by manually identifying land use types such as construction land, farmland, natural land, forest and grassland, and water bodies in 2024;

[0067] S3. Using sample migration technology, migrate the sample points from 2024 to other years and perform land use classification for the corresponding years;

[0068] S4. Introduce a logical conversion mechanism based on the classification results, consider the impact of the previous year's land use on the next year's classification results, and ensure the rationality of land use changes.

[0069] like Figure 3 As shown, this embodiment provides a computer-readable storage medium 1400 on which a computer program 1411 is stored. When the computer program 1411 is executed by a processor, the following steps are implemented:

[0070] S1, based on the GEE platform, remote sensing image data from 2000 to 2024 were screened and processed, and image synthesis was performed using multiple bands (B1-B7) and vegetation indices (such as NDVI, NDWI, etc.);

[0071] S2. Generate sample points by manually identifying land use types such as construction land, farmland, natural land, forest and grassland, and water bodies in 2024;

[0072] S3. Using sample migration technology, migrate the sample points from 2024 to other years and perform land use classification for the corresponding years;

[0073] S4. Introduce a logical conversion mechanism based on the classification results, consider the impact of the previous year's land use on the classification results of the next year, and ensure the rationality of land use changes.

[0074] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Furthermore, the present application may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to magnetic disk storage, CD-ROM, optical storage, etc.) containing computer application code.

[0075] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 a process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0076] These computer program instructions may also be stored in a computer readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 a process or multiple processes and / or boxes Figure 1 The function specified in one or more boxes.

[0077] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operational steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing the instructions executed on the computer or other programmable device for implementing the process. Figure 1 a process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0078] The foregoing descriptions of specific exemplary embodiments of the present invention are for purposes of illustration and description. These descriptions are not intended to limit the invention to the precise forms disclosed, and it is apparent that many modifications and variations are possible in light of the foregoing teachings. The exemplary embodiments have been selected and described for the purpose of explaining the specific principles of the invention and their practical application, thereby enabling those skilled in the art to make and utilize various exemplary embodiments of the invention and various options and variations. It is intended that the scope of the invention be defined by the claims and their equivalents.

[0079] The contents not described in detail in this specification belong to the prior art known to those skilled in the art.

Claims

1. A method for classifying land use in irrigation areas, characterized in that: The steps include: S1, based on a specific year, obtains Landsat images of that year and the previous n years, performs radiation correction, cloud mask and minimum cloud synthesis, extracts B1-B7 bands and calculates NDVI, NDVI1, NDVI2, NDWI and EVI indices of spring, summer, autumn and winter every year through the following formulas (1)-(5), and constructs a multi-feature time series dataset: In the above formulas, NDVI is the normalized difference vegetation index; NDVI1 is the custom vegetation index 1; NDVI2 is the custom vegetation index 2; NDWI is the water body index; EVI is the enhanced vegetation index; NIR is the reflectance of the near-infrared band; RED is the reflectance of the red band; GREEN is the reflectance of the green band; SWIR1 is the reflectance of the shortwave infrared band 1; SWIR3 is the reflectance of the shortwave infrared band 3; S2, based on the Landsat image of a specific year described in step S1, labeling the land use types therein and generating sample points; The land use types include: construction land, farmland, natural land, forest and grassland, and water bodies; S3, for the Landsat images n years before a specific year, use a similar feature matching method to migrate the sample points of the specific year obtained in step S2 to the Landsat images of each year in the n years before the specific year to generate a cross-year training set; S4, using the migration samples of each of the n years before the specific year obtained in step S3 as input, outputting the land use classification results of each of the n years before the specific year year by year based on the random forest algorithm; S5, construct a land use type conversion matrix and define prohibited conversion rules. If the land use type classification result of a certain year in the n years before the specific year violates the above prohibited conversion rules, the model will be retroactively corrected to the result of the previous year.

2. The method for classifying land use in an irrigation area according to claim 1, wherein: The construction of the multi-feature time series dataset in step S1 specifically includes: (1) Quality screening of images of the study area was conducted, and images with cloud cover less than 20% were selected; (2) Using the MASK algorithm for cloud detection and masking; (3) The images are synthesized using the median synthesis method, and the missing areas are filled using other sensors in the same period. Four images of spring, summer, autumn and winter are synthesized every year.

3. The method for classifying land use in an irrigation area according to claim 1, wherein: The migration of the sample points of a specific year to the Landsat images of each of the n years before the specific year in step S3 specifically includes: (1) Calculate the spectral similarity between the migration sample and the target year image through spectral angular distance and Euclidean distance; (2) Set the spectral angle distance threshold and the Euclidean distance threshold to eliminate abnormal samples caused by changes in surface cover and screen samples that meet the requirements.

4. The method for classifying land use in an irrigation area according to claim 1, wherein: The conversion rules described in step S5 include: (1) Allowable conversion types: construction land → natural land, forest and grassland → natural land / farmland, natural land → construction land / forest and grassland / farmland / water body, farmland → forest and grassland / natural land, water body → natural land; (2) Prohibited conversion types: construction land → forest and grassland / farmland / water body; forest and grassland → construction land / water body; farmland → construction land / water body; water body → forest and grassland / farmland / construction land.

5. An irrigation area land use classification system using the irrigation area land use classification method according to any one of claims 1 to 4, characterized in that: It includes the following interconnected modules: Data preprocessing module: used for image correction, feature extraction and time series data set construction; Sample management module: realizes sample labeling, migration and quality assessment; Classification engine module: integrates random forest algorithm and GEE parallel computing interface; Logic verification module: loads the conversion rule library and performs logic conversion correction.

6. The irrigation area land use classification system according to claim 5, wherein: The logic verification module has a built-in rule library that supports user-defined transformation matrices and provides a visual interface to display the spatial differences before and after correction.

7. An electronic device, characterized in that: The electronic device includes a memory, a processor, and a computer program stored in the memory. When the processor executes the computer program, the steps of the irrigation area land use classification method according to any one of claims 1 to 4 are implemented.

8. A computer scale storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 4 are implemented.

Images