Regional land utilization data change detection method based on remote sensing image

Abstract

The invention relates to the field of remote sensing technology and land resource management, and discloses a remote sensing image-based regional land utilization data change detection method, which comprises the following steps of: obtaining multi-temporal remote sensing image data of a research region; preprocessing of geometric correction, radiation correction and noise removal is carried out on the remote sensing image; carrying out preliminary classification on the image data by adopting an unsupervised classification method, and correcting a classification result in combination with field collected data; constructing a transfer matrix according to a classification result, performing change detection on the classification result of the multi-temporal image, and analyzing a conversion relation and area change among different land classes; and generating a thematic map to display land class distribution and change trend, and outputting a statistical analysis result. According to the method, quantitative analysis and visual display are combined, the land utilization change rule can be efficiently and accurately reflected, a scientific basis is provided for regional land resource management and planning, the applicability is high, and the implementation cost is low.

Description

Technical Field

[0001] The present invention relates to the technical fields of remote sensing technology and land resource management technology, and specifically provides a method for detecting changes in regional land use data based on remote sensing images. Background Art

[0002] With the rapid development of urbanization and regional economy, the study of land use / cover change (LUCC) has become an important direction in the fields of environmental science, geography, and land resource management. As an efficient and accurate means of obtaining spatial information, remote sensing technology has been widely used in the monitoring of regional land use changes. However, there are still some deficiencies in the existing technologies in practical applications, which affect the effectiveness and practicality of land use change detection.

[0003] Existing land use change analysis methods mostly rely on the image classification results of a single time point, lacking a systematic quantitative analysis of the land type conversion relationship between multi-temporal data, and it is difficult to comprehensively reveal the dynamic change characteristics. At the same time, traditional classification methods are sensitive to the problems of misclassification and wrong classification of complex surface types. Especially when the spectral characteristics of ground objects are similar, the classification accuracy is low, and a large amount of manpower is often required for correction. In addition, in the expression of change results, existing technologies rely more on simple statistical reports or text descriptions, lacking an intuitive spatial expression method, which is not conducive to user understanding and application.

[0004] There are also limitations in the data processing efficiency of existing technologies. Remote sensing image data usually has a wide coverage range and a large amount of data. Especially in multi-temporal analysis, the data processing complexity increases significantly. Many traditional methods lack an efficient processing flow for large-scale image data and are difficult to meet the needs of regional and multi-temporal land use change research. Generally speaking, existing technologies are difficult to achieve a comprehensive analysis and efficient expression of land use changes under the premise of ensuring accuracy and efficiency, which limits their wide application in land resource management, ecological protection, and planning decision-making. Summary of the Invention

[0005] Aiming at the deficiencies of the existing technologies, the present invention provides a method for detecting changes in regional land use data based on remote sensing images, which solves the problems of low classification accuracy of multi-temporal land use, inaccurate change detection, and non-intuitive result expression in the existing technologies.

[0006] To achieve the above objectives, the present invention is realized through the following technical solutions: A method for detecting changes in regional land use data based on remote sensing images, comprising the following steps:

[0007] Obtain multi-temporal remote sensing image data of the research area;

[0008] Preprocess the remote sensing image data;

[0009] Classify the preprocessed image data. Use unsupervised classification methods to preliminarily classify the image data, and combine the data collected in the field to correct the classification results;

[0010] Construct a transfer matrix based on the classification results to perform change detection on the classification results of multi-temporal images, and analyze the conversion relationships and area changes between different land classes;

[0011] Generate a thematic map to display the distribution and change trends of each land class, and output the statistical analysis results of land use changes.

[0012] Preferably, the multi-temporal remote sensing image data includes Landsat satellite images at different time nodes.

[0013] Preferably, the preprocessing includes:

[0014] Eliminate the projection deformation and spatial offset of the image through geometric correction;

[0015] Use a filtering method to remove periodic noise and spike noise;

[0016] Process the thin cloud area using the spectral weakening method, and correct the shadow area using the ratio method.

[0017] Preferably, the classification method uses the ISODATA and K-means unsupervised classification algorithms, and adjusts the classification results in combination with national standards.

[0018] Preferably, the classification method includes the following steps:

[0019] Use the K-means clustering algorithm for preliminary classification. The algorithm realizes classification by minimizing the objective function. The objective function is:

[0020]

[0021] where k is the number of clusters, n is the number of samples, is the feature vector of the j-th sample in the i-th class, c i is the cluster center of the i-th class, ∥·∥ 2 represents the Euclidean distance;

[0022] Use the ISODATA dynamic clustering algorithm to optimize the classification results, specifically including:

[0023] Dynamically adjust the number of classes k. Update the number of classes k by merging classes with Euclidean distances less than a preset threshold or splitting classes with sample variances greater than a preset threshold;

[0024] Iteratively update the cluster center c i of each class to minimize the total distance between samples within the class;

[0025] Match the preliminary classification results with the field-collected data, and adjust the classification categories according to the actual land use distribution to ensure that the classification accuracy meets the preset standards.

[0026] Preferably, the transition matrix is used for quantitative analysis of the conversion direction of land use types, including the change relationships between categories of farmland, construction, forest, water area, grassland, unused cultivated land, and urban and rural construction land.

[0027] Preferably, the calculation method of the transition matrix includes the following steps:

[0028] Perform spatial overlay on the classification results of multi-temporal images, and extract the intersection of land use types at different time nodes through GIS tools;

[0029] Construct a transition matrix T, and the element T of the transition matrix ij represents the conversion area from land type i at time t 1 to land type j at time t 2 , and the specific calculation formula is:

[0030]

[0031] where S(x, y) is the unit area of the position coordinate (x, y) in the study area R. If (x, y) belongs to category i at time t 1 and belongs to category j at time t 2 , then S(x, y) ≠ 0, otherwise S(x, y) = 0;

[0032] Analyze each element in the transition matrix and extract the main characteristics of land use change, including:

[0033] The unchanged part of the area of the same type of ground objects;

[0034] The conversion area between different categories.

[0035] Preferably, the thematic map is output through GIS software, including the visualization display of the area, proportion, and change trend of different land use types.

[0036] The present invention also provides a device for detecting changes in regional land use data based on remote sensing images, including:

[0037] A data acquisition module for acquiring multi-temporal remote sensing image data;

[0038] A preprocessing module for performing geometric correction, denoising processing, and abnormal feature correction on the remote sensing image data;

[0039] A classification module for classifying the preprocessed image data and correcting the classification results in combination with the field-collected data;

[0040] A change detection module, which is used to construct a transition matrix based on the classification results and analyze the change trend of land use types in multi-temporal images;

[0041] A result output module, which is used to generate a thematic map and output the statistical analysis results of land use changes.

[0042] The present invention provides a method for detecting changes in regional land use data based on remote sensing images. It has the following beneficial effects:

[0043] 1. By combining the unsupervised classification method with field-collected data for classification correction, the present invention significantly improves the accuracy of the classification results. The preliminary classification realizes efficient clustering through spectral characteristics, and then combines field data to correct misclassification and mixed classification phenomena, ensuring that the classification results are consistent with the actual land type distribution. This method is applicable to land use analysis of large areas and multiple land types, especially having significant advantages in areas with complex ground object distributions.

[0044] 2. By constructing a transition matrix, the present invention quantitatively analyzes the conversion relationship and area change between multi-temporal land use types, and reveals the spatial dynamic characteristics of land type changes. Through the analysis of the diagonal and non-diagonal elements of the transition matrix, the stability and conversion direction of each land type are clarified. This method provides a scientific basis for studying the adjustment of regional land use structure, ecological protection and urbanization development.

[0045] 3. The thematic map generated by the present invention presents the spatial distribution, change trend and hot spots of each land type in a visual form, making the change results intuitive and easy to read. Combined with the annotation of colors and symbols, it clearly shows the land use pattern and change characteristics of the research area. The high-resolution output of the thematic map can be widely applied to decision support, academic research and public communication. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] Figure 1 It is a schematic flow chart of the method of the present invention;

[0047] Figure 2 It is a schematic structural diagram of the device of the present invention.

[0048] Among them, 10 is a data acquisition module; 20 is a preprocessing module; 30 is a classification module; 40 is a change detection module; 50 is a result output module. DETAILED DESCRIPTION OF THE INVENTION

[0049] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0050] Please refer to the attached Figure 1 , the present invention provides a method for detecting changes in regional land use data based on remote sensing images, which can dynamically analyze land use types at different time nodes, reveal land use change laws, and provide data support for the scientific management and planning of regional land resources.

[0051] As Figure 1 shown, the method for detecting changes in regional land use data based on remote sensing images may include the following steps:

[0052] S1. Obtain multi-temporal remote sensing image data of the study area;

[0053] S2. Preprocess the remote sensing image data;

[0054] S3. Classify the preprocessed image data;

[0055] S4. Construct a transition matrix based on the classification results;

[0056] S5. Generate a thematic map and output the analysis results.

[0057] The following will elaborate on each step of the method of the present invention in detail.

[0058] For step S1, in this embodiment, step S1 obtains multi-temporal remote sensing image data of the study area, laying a data foundation for subsequent data processing and land use change analysis. The specific content includes the following:

[0059] In this embodiment, the obtained remote sensing image data is sourced from a publicly available satellite remote sensing image resource library, such as the Landsat series image data jointly released by the National Aeronautics and Space Administration (NASA) and the United States Geological Survey (USGS). Preferably, Landsat TM (Thematic Mapper) and Landsat8 (Operational Land Imager, OLI) images are selected as the data sources.

[0060] In this embodiment, to ensure the coverage and continuity of the data, the selected Landsat data should meet the following conditions:

[0061] The spatial resolution is preferably 30 meters to ensure that the image has sufficient details to meet the requirements of regional land use analysis;

[0062] The time nodes are preferably representative to ensure that the images can reflect the long-term dynamics of land use change. In this embodiment, the image data of 2000 and 2021 are selected as the main analysis data;

[0063] The geographical coverage of the data should completely cover the study area (taking Lanzhou City as an example) to avoid data loss problems in the boundary areas.

[0064] In this embodiment, to ensure the quality of the data and the consistency of multi-temporal images, it is preferably to select image data with similar imaging seasons to avoid the influence of seasonal surface feature differences (such as vegetation cover changes) on the analysis results. At the same time, to reduce the radiation bias caused by sensor differences, it is recommended in this embodiment to preferentially select remote sensing data of the same series or data that has been corrected for consistency.

[0065] In this embodiment, during the acquisition of image data, field-collected data is also combined to provide a reference basis for the distribution of ground objects for the subsequent correction of classification results and accuracy evaluation. The field-collected data should include the specific distribution and characteristics of the main land use types (such as farmland, forest, grassland, etc.) within the study area.

[0066] In this embodiment, to achieve unified management and subsequent processing of the data, the image data is uniformly stored in the GeoTIFF format, which has the ability to store spatial information and can directly support subsequent GIS analysis and processing.

[0067] In this embodiment, when conducting a quality check on the remote sensing image data, if cloud cover, shadows, or other noise problems are found in the images, alternative images can be selected for supplementation during the data acquisition process. In addition, the preprocessing methods in the subsequent steps can also be used to correct related problems.

[0068] In this embodiment, the basic objective of obtaining image data is to provide a complete, reliable, and continuous remote sensing data set, thereby providing a scientific basis for multi-temporal land use change detection.

[0069] For step S2, in this embodiment, step S2 preprocesses the acquired remote sensing image data to eliminate noise, geometric deviations, and other abnormal features in the image data, providing a reliable data input for subsequent classification and change detection.

[0070] In this embodiment, the preprocessing of remote sensing images includes but is not limited to the following steps: geometric correction, radiometric correction, noise removal, correction processing of thin clouds and shadows, etc.

[0071] In this embodiment, geometric correction aims to solve the problem of projection distortion of remote sensing images caused by factors such as the curvature of the earth, sensor offset, and the rotation of the earth. Geometric correction is performed by registering high-precision topographic control points (such as obvious feature points on topographic maps) with the remote sensing image. Preferably, bilinear interpolation or nearest neighbor interpolation algorithms are used to recalculate pixel values, so that the image coordinates are transformed into a unified geographic coordinate system (such as the WGS84 coordinate system) to ensure the spatial consistency and superposability of the images.

[0072] In this embodiment, radiometric correction is mainly used to correct the spectral value deviation in the image caused by atmospheric scattering and absorption. The atmospheric correction methods used include but are not limited to the FLAASH (Fast Line-of-sight Atmospheric Analysis of Spectral Hypercubes) tool or the surface reflectance model correction method. During this process, the reflectance of the image is adjusted according to the specific acquisition conditions of the image (such as solar altitude angle, atmospheric water vapor content, etc.) to restore the true spectral characteristics of the earth's surface.

[0073] In this embodiment, to remove periodic noise (such as strip noise and spike noise) in the image, the Fourier transform (FFT) filtering method is preferably used. Specifically, the image data is transformed into the frequency domain, and high-frequency interference signals are identified and smoothed. Strip noise is weakened by low-pass filtering technology; spike noise is eliminated by frequency domain filtering combined with inverse transformation. The above methods can significantly improve the visual quality of the image and the accuracy of subsequent classification.

[0074] In this embodiment, for the thin cloud-covered areas that may exist in the image, the spectral weakening method is used to correct the influence of thin clouds. Specifically, based on the spectral characteristics of the cloud-covered area, its reflectance value is adjusted or cloud-free images of other time phases are used to replace the thin cloud area, so as to restore the surface information blocked by the thin clouds.

[0075] In this embodiment, for the shadow areas that may exist in the image, the spectral value is corrected by combining the ratio method. The specific implementation method is to calculate the spectral ratio of the shadow area to the non-shadow area to restore the spectral characteristics of the shadow area. At the same time, digital elevation model (DEM) data can also be combined to correct the shadow area by simulating solar illumination conditions.

[0076] In this embodiment, to ensure the quality and consistency of the image data after preprocessing, the image is inspected for quality after preprocessing, mainly checking whether the image meets geometric consistency, radiometric consistency, and whether there are no obvious noises and abnormal features. If problems are found, return to the previous steps for reprocessing or replace the original image data.

[0077] The preprocessing of remote sensing images in this embodiment can be achieved through publicly available algorithms and existing software tools (such as ENVI, PIE-Basic 6.3, etc.).

[0078] For step S3, in this embodiment, step S3 classifies the preprocessed remote sensing image data, specifically including using an unsupervised classification method to preliminarily classify the image data and correcting the classification results in combination with field-collected data to obtain an accurate land use classification result.

[0079] In this embodiment, the classification method uses unsupervised classification techniques, and preferably uses two algorithms, ISODATA and K-means, to perform clustering analysis on the image data. This method analyzes the spectral characteristics of pixels, divides the ground objects in the image into several categories, and generates a preliminary classification result.

[0080] In this embodiment, the implementation of the K-means clustering algorithm includes the following specific steps

[0081] First, randomly initialize k clustering centers c i , where k is the preset number of categories;

[0082] Then, assign each pixel point x j to the category corresponding to the nearest clustering center c i and calculate its Euclidean distance:

[0083] d ij =∥x j -c i ∥

[0084] where d ij represents the Euclidean distance from the pixel point x j to the clustering center c i ;

[0085] Next, update the clustering center of each category by minimizing the objective function J:

[0086]

[0087] where k is the number of clusters, n is the number of samples, is the feature vector of the jth sample in the ith category, c i is the clustering center of the ith category, and ∥·∥ 2 represents the Euclidean distance.

[0088] Repeat the above steps until the clustering centers no longer change or reach the preset convergence condition.

[0089] In this embodiment, the ISODATA dynamic clustering algorithm adds the functions of class merging and splitting on the basis of K-means, thereby optimizing the clustering effect. The specific implementation is as follows:

[0090] When the distance between the clustering centers of two classes is less than the set threshold, they are merged into one class;

[0091] When the internal variance of a certain class is greater than the preset threshold, it is split into two classes;

[0092] By dynamically adjusting the number of classes k, the final classification result more conforms to the actual distribution of ground objects.

[0093] In this embodiment, the preliminary classification result is only based on the pixel spectral characteristics, so there may be misclassification or wrong classification phenomena. In order to further improve the classification accuracy, the classification result is corrected by combining the field-collected data. The field-collected data includes the distribution and characteristic spectral values of each main land use type (such as farmland, forest, grassland, water area, etc.) in the study area.

[0094] In this embodiment, the specific method of classification correction is as follows:

[0095] For the misclassified pixels in the classification result, reassign the classes according to the spectral characteristics and geographical distribution of the field-collected data;

[0096] For the areas with blurred classification boundaries, accurately divide them by overlaying auxiliary data (such as DEM or multi-temporal images);

[0097] According to the requirements of the "Classification Specification for Current Land Use" (GB / T 21010-2007), the classification result is adjusted to seven major categories: farmland, construction, forest, water area, grassland, unused cultivated land, and urban and rural construction land.

[0098] In this embodiment, to verify the accuracy of the classification result, the confusion matrix method is used for evaluation. The confusion matrix can quantify the matching situation between the classification result and the actual land use type, and reflect the reliability of the classification result by calculating the overall accuracy and Kappa coefficient.

[0099] The classification method in this embodiment combines the high efficiency of unsupervised classification and the accuracy of field data, and can significantly improve the classification accuracy while ensuring efficiency.

[0100] For step S4, in this embodiment, step S4 constructs a transfer matrix based on the classification result to perform change detection on the classification result of the multi-temporal image, so as to quantitatively analyze the conversion relationship and area change between different land use types.

[0101] In this embodiment, the transition matrix is a commonly used quantitative analysis method for characterizing the mutual conversion between multi-temporal land use types. Specifically, the transition matrix performs pixel-level spatial superposition on the land use classification results at two time points, and statistically analyzes the conversion area of each land type between the two time nodes. The form of the transition matrix is as follows:

[0102]

[0103] Where T ij represents the conversion area from land type i at time t 1 to land type j at time t 2 , and the unit is usually square kilometers or the number of pixels.

[0104] In this embodiment, the construction of the transition matrix includes the following key steps:

[0105] First, perform a spatial superposition operation on the land use classification results at two time points. Use the Intersect tool of GIS software (such as ArcGIS) to perform pixel-level spatial superposition on the classification data at the two time nodes to generate a superimposed layer with attribute records. The superimposed layer records the classification results of each pixel at the two time nodes. For example, a pixel belongs to "farmland" at time t 1 and is converted to "construction land" at time t 2 .

[0106] Then, according to the attribute records of the superimposed layer, statistically analyze the conversion area of each land type between the two time points. The specific calculation formula is:

[0107]

[0108] Where S(x,y) is the unit area of the location coordinate (x,y) in the study area R. If (x,y) belongs to category i at time t 1 and belongs to category j at time t 2 , then S(x,y)≠0; otherwise, S(x,y) = 0.

[0109] In this embodiment, the diagonal element T of the transition matrix ii represents the area where the same land type remains unchanged between the two time nodes. For example, the area where farmland remains as farmland in 2000 and 2021 is the diagonal element. The non-diagonal element T ij represents the conversion area from land type i to land type j. For example, the area where unused land is converted to grassland is the corresponding non-diagonal element in the transition matrix.

[0110] In this embodiment, by analyzing the transition matrix, the main characteristics of land use change can be extracted, including but not limited to the following:

[0111] Stability analysis: Calculate the stable area ratio of each land type between two time points through the sum of the diagonal elements of T ii For each land type, the stable area ratio between two time points is calculated through the sum of the diagonal elements of T

[0112] Conversion direction analysis: Identify the main conversion directions of each land type through the distribution of the non - diagonal elements of T ij For example, whether farmland is mainly converted into construction land. That is, through the distribution of the non - diagonal elements of T, the main conversion directions of each land type are identified

[0113] Area change analysis: Calculate the net increase or decrease area of each land type through the sum of rows and columns

[0114] In this embodiment, to improve the accuracy of change detection, abnormal data can be checked in combination with the land type characteristics of the actual research area. For example, if a certain area was originally a water area, but the classification result shows that it has been converted into construction land, it is necessary to further verify whether this result conforms to the actual situation

[0115] In this embodiment, the construction of the transfer matrix and the change detection process combine the time - series analysis of remote sensing images and GIS spatial analysis techniques, which can accurately reveal the spatio - temporal characteristics of regional land - use changes

[0116] For step S5, in this embodiment, step S5 generates a thematic map based on the classification and change detection results to visually display the spatial distribution and change trends of each land type, and outputs the statistical analysis results of land - use changes, providing a scientific basis for regional land - use planning and management

[0117] In this embodiment, the production of the thematic map is based on GIS software (such as ArcGIS), and the classification results of multiple time phases and their change information are spatially visualized. Specifically, the classification results and conversion relationships of each land type are represented by different colors or symbols, so as to visually reflect the land - use pattern and its dynamic changes in the research area

[0118] In this embodiment, to generate a classified thematic map, first, the format of the classification results needs to be converted. The classification data (such as image data in GeoTIFF format) is loaded into the GIS software, and each land type is assigned a unique color or legend through symbolization. For example, farmland is represented by green, grassland by light yellow, forest by dark green, water area by blue, and construction land by red

[0119] In this embodiment, the generation of the thematic map of changes is based on the analysis results of the transfer matrix. Specifically, the conversion relationships are represented in the form of arrows or overlay layers. For example, the areas where farmland is converted into construction land are marked with red arrows, and the areas where unused land is converted into grassland are marked with yellow arrows. In addition, the hotspots of changes are presented in the form of a heat map, such as marking the high-frequency areas of construction land expansion.

[0120] In this embodiment, the statistical analysis results are calculated through the spatial analysis tools (such as the attribute statistics tool) of GIS software and output in the form of tables or charts. The statistical contents include:

[0121] The area change of each land use type, and the calculation formula is:

[0122]

[0123] where ΔA i represents the area change of land use type i, and represent the total areas of land use type i at times t 1 and t 2 respectively;

[0124] The conversion ratio between land use types, and the calculation formula is:

[0125]

[0126] where P ij represents the ratio of land use type i converted into land use type j, and T ij is the corresponding conversion area;

[0127] Statistics of hotspots of changes. Use the spatial clustering analysis tool (such as the hotspot analysis tool) of GIS to locate the high-frequency areas of land use type changes and generate a summary of regional changes in combination with actual geographical information.

[0128] In this embodiment, the finally generated thematic map includes but is not limited to the following contents:

[0129] The thematic map of the spatial distribution of each land use type at two time nodes;

[0130] The dynamic change thematic map of the conversion relationships between land use types;

[0131] The clustering analysis map of hotspots of changes.

[0132] In this embodiment, to ensure the expression accuracy of the thematic map, the generated map is reviewed and corrected. The main inspection contents include: whether the land type color markings are consistent, whether the spatial distribution conforms to the geographical reality, and whether the conversion direction and ratio are consistent with the results of the transfer matrix. After the correction is completed, the thematic map is exported as a high-resolution image or in PDF format for subsequent display or application.

[0133] The statistical analysis results and thematic maps in this embodiment can provide data support for the scientific management of regional land resources. For example, by analyzing the reduction of farmland and the expansion of construction land, the ecological problems that may be faced during the urbanization process can be determined; by analyzing the changes in unused land, the potential and limitations of regional land development can be evaluated.

[0134] The method described in this embodiment can be implemented through existing GIS software and publicly available statistical analysis tools, and combines the quantitative analysis results of the change matrix, which can accurately reflect the dynamic change characteristics of land use.

[0135] Generally speaking, the present invention comprehensively reveals the dynamic change characteristics of land use in the study area by acquiring multi-temporal remote sensing image data, performing preprocessing, classification, change detection, and visualization analysis on it. By using unsupervised classification methods combined with field-collected data to optimize the classification results, and quantitatively analyzing the conversion relationship and area change trend of land use types based on the transfer matrix. The present invention further generates thematic maps and statistical analysis results, intuitively displaying the spatial distribution and change rules of each land type, providing a scientific basis for regional land resource management, planning, and ecological protection, and possessing high efficiency, accuracy, and universality.

[0136] The regional land use data change detection device based on remote sensing images described below can be correspondingly referred to the regional land use data change detection method based on remote sensing images described above.

[0137] Please refer to the attached Figure 2 , the present invention also provides a regional land use data change detection device based on remote sensing images, including:

[0138] A data acquisition module 10 for acquiring multi-temporal remote sensing image data;

[0139] A preprocessing module 20 for performing geometric correction, denoising processing, and abnormal feature correction on the remote sensing image data;

[0140] A classification module 30 for classifying the preprocessed image data and correcting the classification results in combination with field-collected data;

[0141] A change detection module 40 for constructing a transfer matrix based on the classification results and analyzing the change trend of land use types in multi-temporal images;

[0142] A result output module 50 is configured to generate a thematic map and output the statistical analysis results of land use changes.

[0143] The device in this embodiment can be used to execute the above method embodiment, and its principle and technical effects are similar, which will not be elaborated here.

[0144] Although the embodiments of the present invention have been shown and described, those of ordinary skill in the art can understand that various changes, modifications, substitutions, and variations can be made to these embodiments without departing from the principle and spirit of the present invention. The scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A method for detecting changes in regional land use data based on remote sensing images, characterized in that: The following steps are involved: Obtain multi-temporal remote sensing image data of the study area; Preprocess remote sensing image data; Classify the preprocessed image data, use unsupervised classification methods to perform preliminary classification on the image data, and correct the classification results in combination with field collected data; Based on the classification results, a transfer matrix is ​​constructed to perform change detection on the classification results of multi-temporal images, and the transformation relationship and area change between different land types are analyzed; Generate thematic maps to show the distribution and changing trends of various land types, and output statistical analysis results of land use changes.

2. The method for detecting changes in regional land use data based on remote sensing images according to claim 1, characterized in that: The multi-temporal remote sensing image data includes Landsat satellite images at different time points.

3. The method for detecting changes in regional land use data based on remote sensing images according to claim 1, characterized in that: The pre-processing comprises: Eliminate projection distortion and spatial offset of images through geometric correction; Use filtering methods to remove periodic noise and spike noise; The spectral attenuation method is used to process the thin cloud area, and the ratio method is used to correct the shadow area.

4. The method for detecting changes in regional land use data based on remote sensing images according to claim 1, characterized in that: The classification method adopts ISODATA and K-means unsupervised classification algorithms, and adjusts the classification results in combination with national standards.

5. The method for detecting changes in regional land use data based on remote sensing images according to claim 4, characterized in that: The classification method comprises the following steps: The K-means clustering algorithm is used for preliminary classification. The algorithm achieves classification by minimizing the objective function, which is: Where k is the number of clusters, n is the number of samples, is the feature vector of the jth sample in the i-th class, c i is the cluster center of the i-th class, ∥·∥ 2 represents the Euclidean distance; The ISODATA dynamic clustering algorithm is used to optimize the classification results, including: Dynamically adjust the number of categories k, by merging categories whose Euclidean distance is less than a preset threshold, or splitting categories whose sample variance is greater than a preset threshold, to update the number of categories k; Iteratively update the cluster center c of each category i , so that the total distance between samples within the category is minimized; Match the preliminary classification results with the field-collected data, adjust the classification categories according to the actual land distribution, and ensure that the classification accuracy meets the preset standards.

6. The method for detecting changes in regional land use data based on remote sensing images according to claim 1, characterized in that: The transfer matrix is ​​used to quantitatively analyze the transformation direction of land use types, including the change relationship between farmland, buildings, forests, water areas, grasslands, unused cultivated land and urban and rural construction land categories.

7. The method for detecting changes in regional land use data based on remote sensing images according to claim 6, characterized in that: The calculation method of the transfer matrix comprises the following steps: The classification results of multi-temporal images are spatially superimposed, and the intersection of land use types at different time nodes is extracted through GIS tools; Construct the transfer matrix T, the elements of the transfer matrix T ij It represents the conversion area from land type i at time t1 to land type j at time t2. The specific calculation formula is: Where S(x,y) is the unit area of ​​the location coordinate (x,y) in the study area R. If (x,y) belongs to category i at time t1 and belongs to category j at time t2, then S(x,y)≠0, otherwise S(x,y)=0; Analyze each element in the transfer matrix and extract the main characteristics of land use change, including: The unchanged portion of the area of ​​similar features; The conversion area between different categories.

8. The method for detecting changes in regional land use data based on remote sensing images according to claim 1, characterized in that: The thematic map is output through GIS software, including a visual display of the area, proportion and change trend of different land types.

9. A device for detecting changes in regional land use data based on remote sensing images, used to execute the method for detecting changes in regional land use data based on remote sensing images as claimed in any one of claims 1 to 8, characterized in that: include: Data acquisition module, used to acquire multi-temporal remote sensing image data; Preprocessing module, used to perform geometric correction, denoising and abnormal feature correction on remote sensing image data; The classification module is used to classify the pre-processed image data and correct the classification results in combination with the field collected data; Change detection module, which is used to construct a transfer matrix based on the classification results and analyze the changing trend of land use types in multi-temporal images; The result output module is used to generate thematic maps and output the statistical analysis results of land use changes.

Images