A statistical method and system for dynamically changing forestland resource areas

Abstract

The present invention relates to the technical field of ecological environment monitoring, and in particular to a statistical method and system for the dynamic change area of ​​forest resources. It includes obtaining a forest resource data set covering the statistical area; calculating vegetation index and texture feature measurement of the pre-processed data at different periods, extracting features from geographic information data, and constructing a time series graph training set; constructing an incomplete multi-view clustering model that integrates the time series graph, training the model using the time series graph training set, and clustering the forest resources in the statistical area using the trained clustering model; calculating the area of ​​each cluster area separately, and calculating the area change of each cluster area. By analyzing the change trend of the cluster area, the present invention can analyze the dynamic changes of forest resources from multiple dimensions and multiple regions, and provide a detailed and accurate basis for the management of forest resources.

Description

Technical Field

[0001] The present invention relates to the technical field of ecological environment monitoring, and in particular to a statistical method and system for dynamically changing areas of forest resources. Background Art

[0002] As a vital natural resource, forests play a key role in maintaining ecological balance and providing ecological services such as water resource management, soil conservation, and climate regulation. Understanding forest location, type, area, and distribution is crucial for developing effective conservation policies and management plans. By monitoring the dynamics of forests, we can promptly identify issues such as forest degradation and destruction, enabling appropriate intervention and restoration measures. We can also measure the effectiveness of planting rates, safeguarding the integrity and diversity of forest ecosystems.

[0003] In the existing technology, traditional forest area monitoring and statistical methods have limitations. They usually use manual field surveys, which require a large amount of manpower, material resources and time. The measurement process is long and difficult to obtain information on large areas of forest in a short time. With the emergence and rapid development of remote sensing technology, remote sensing technology has advantages such as wide coverage and short repetition period, and can quickly, accurately and real-timely obtain remote sensing data. However, the types of objects contained in remote sensing images are complex, which can easily lead to mis-extraction and omission when extracting information from forests, resulting in low accuracy. CN 114067203 A calculates the covered area and exposed area of ​​forest plants by taking fixed-point photos and performing color rendering and comparison. This relies on a fixed information collection terminal and a specific processing method, which makes it difficult to monitor large areas of forest. The accuracy of information collection may be affected by factors such as shooting angle and weather conditions, and it is impossible to conduct in-depth analysis and prediction of the complex changes in forest resources. Although CN 118097446 A has certain methods in texture feature extraction and model construction, it still has some shortcomings. Its data source is mainly based on visible light visual remote sensing, which is relatively single and cannot fully obtain various information about forest land. In terms of feature extraction, although texture features are taken into account, the comprehensive utilization of other important vegetation indexes and other features may not be sufficient. Moreover, in model construction, the dynamic changes and complex spatiotemporal relationships of forest resources are not fully considered, which affects the accuracy and reliability of the detection of forest area changes.

[0004] In summary, the existing methods for monitoring the dynamic changes in forestland resources have various limitations, and a more comprehensive, accurate and efficient method and system for statistically analyzing the dynamic changes in forestland resources is needed. Summary of the Invention

[0005] In order to solve the above-mentioned problems, the present invention provides a statistical method and system for dynamically changing areas of forest resources.

[0006] In a first aspect, the present invention provides a statistical method for dynamically changing forestland resource areas, which employs the following technical solutions:

[0007] A statistical method for dynamically changing forestland resource areas, comprising:

[0008] S1. Obtain a forest resource dataset covering the statistical area, the forest resource dataset including remote sensing image data of different periods, geographic information data of the corresponding period, and field measurement data of the corresponding period, and preprocess the forest resource dataset;

[0009] S2. Calculate vegetation indices and measure texture features at different times on the preprocessed data, extract features from the geographic information data, and construct a time series map training set;

[0010] S3. Construct an incomplete multi-view clustering model that integrates the time series graph, train the model using the time series graph training set, and use the trained clustering model to cluster the forest resources in the statistical area;

[0011] S4. Calculate the area of ​​forest resource changes, including calculating the area of ​​each cluster region for clustering results in different periods, obtaining the area value of each cluster region at different time points, and calculating the area change of each cluster region.

[0012] Furthermore, the forest resource data is pre-processed, including obtaining high-resolution and medium-resolution remote sensing image data using satellite remote sensing, converting the remote sensing image data into physical radiation data, and performing radiation correction and geometric correction on the physical radiation data to match the remote sensing image data with the geographic coordinate system. The radiation correction formula is:

[0013]

[0014] in, represents the surface reflectivity, Represents the digital value of remote sensing image, is the quantization coefficient, and is the radiation calibration parameter of the sensor.

[0015] Furthermore, the forest resource data is preprocessed, including format conversion and coordinate unification of geographic information data, and field measurement data is checked and errors and outliers in the measurement data are removed.

[0016] Furthermore, the vegetation index calculation includes normalized difference vegetation index and enhancement of vegetation index. The normalized difference vegetation index formula is:

[0017]

[0018] in, and are the reflectances of the near-infrared and red bands, It represents the difference vegetation index, which is used to indicate the growth status and coverage of vegetation. Its value range is between -1 and 1. Positive values ​​indicate vegetation coverage, and negative values ​​indicate no vegetation coverage.

[0019] Furthermore, the vegetation index is enhanced as follows:

[0020]

[0021] in, 、 and is a constant.

[0022] Furthermore, the texture feature measurement includes calculating the texture features of the remote sensing image data through a gray level co-occurrence matrix, and the texture features include mean, variance, contrast, correlation and entropy.

[0023] Furthermore, the construction of the time series graph training set includes:

[0024] A1. Arrange remote sensing image data, geographic information data, and field measurement data acquired at different times in chronological order to form a time series dataset.

[0025] A2. Associate remote sensing image data, geographic information data, and field measurement data to establish a corresponding relationship in time series;

[0026] A3. Use triples to represent relationships in the time series graph, annotate the valid time range for each triple, and generate a time series graph.

[0027] A4. Combine and annotate vegetation indices, texture features, and geographic information from different periods in the time series atlas to construct a time series atlas training set.

[0028] Furthermore, the construction of an incomplete multi-view clustering model integrating the time series graph includes:

[0029] B1. Constructing a model, the model includes a vegetation index view input layer, a texture feature view input layer, and a geographic information view input layer;

[0030] B2. Constructing and utilizing time-weighted graphs Fusion time series graph, where nodes Represents data samples at different time points, represents the temporal relationship between samples, Represents the weight, and the calculation formula is: ,in represents the weight, and Represents two time points, represents the attenuation coefficient;

[0031] B3. Introducing a time regularization term into the model’s objective function ,in Represents the feature representation matrix of all data samples.

[0032] Furthermore, the model is trained using the time series graph training set, including

[0033] C1. Input the time series atlas training set into the model, extract features through the vegetation index view input layer, texture feature view input layer, and geographic information view input layer, and concatenate the extracted feature representations to obtain a fused feature representation. The formula is: ,in, represents the feature representation obtained from the vegetation index view, represents the feature representation obtained from the texture feature view, Represents the feature representation obtained from the geographic information view;

[0034] C2. Calculate the adjusted feature representation based on the time weighted graph and the time regularization term, the formula is: ,in, is the adjustment coefficient;

[0035] C3. Calculate the relationship score between entities in different regions. The formula is: ,in, and Represent the adjusted feature representations of different regions respectively;

[0036] C4. Constructing cross entropy loss function ,in, Represents a sample Belong to cluster The true label, Represents the model prediction sample Belong to cluster The probability of the model is obtained by using the gradient descent algorithm to optimize the model parameters and repeat the forward propagation calculation until the loss function value of the model converges and the model training is completed.

[0037] Furthermore, for the clustering results of different periods, the area of ​​each cluster region is calculated respectively, including determining the pixel range of each cluster in the remote sensing image, marking the image through the image segmentation algorithm, identifying the pixels belonging to each cluster, calculating the number of pixels in each cluster region, and calculating the area of ​​the cluster region. The formula is:

[0038]

[0039] in, is the image resolution in meters / pixel, is the number of pixels.

[0040] Furthermore, the calculation of the area change of each cluster region includes analyzing the change trend of the cluster region, and the formula is: ,in, Indicates the area change rate. A negative number indicates a decrease in forest area, and a positive number indicates an increase in forest area. Indicates time The area of ​​the clustering region, Indicates that the cluster area is at time and time The change in time, .

[0041] Secondly, a statistical system for dynamically changing forestland areas includes:

[0042] a data acquisition module configured to acquire a forest resource dataset covering a statistical area, the forest resource dataset including remote sensing image data of different periods, geographic information data of corresponding periods, and field measurement data of corresponding periods, and preprocess the forest resource dataset;

[0043] The data processing module is configured to calculate vegetation index and texture feature measurement at different periods of the pre-processed data, extract features from the geographic information data, and construct a time series map training set;

[0044] a model training module configured to construct an incomplete multi-view clustering model fused with the time series graph, train the model using the time series graph training set, and cluster the forestland resources in the statistical area using the trained clustering model;

[0045] The area statistics module is configured to calculate the area change of forest resources, including calculating the area of ​​each cluster area for clustering results in different periods, obtaining the area value of each cluster area at different time points, and calculating the area change of each cluster area.

[0046] In a third aspect, the present invention provides a computer-readable storage medium storing a plurality of instructions, wherein the instructions are suitable for being loaded and executed by a processor of a terminal device, for the statistical method of dynamically changing area of ​​forest resources.

[0047] In a fourth aspect, the present invention provides a terminal device comprising a processor and a computer-readable storage medium, wherein the processor is used to implement various instructions; and the computer-readable storage medium is used to store a plurality of instructions, wherein the instructions are suitable for being loaded and executed by the processor to obtain a statistical method for dynamically changing area of ​​forest resources.

[0048] In summary, the present invention has the following beneficial technical effects:

[0049] 1. The present invention proposes a statistical method and system for dynamically changing forestland areas. By acquiring remote sensing image data, geographic information data, and field measurement data from different periods, and preprocessing and comprehensively utilizing these data, this method overcomes the problem of a single data source in existing technologies. Multi-source data complements and verifies each other, providing a more comprehensive and accurate description of the actual situation of forestland resources.

[0050] 2. This invention calculates vegetation indices, including the Normalized Difference Vegetation Index (NDVI) and the Enhanced Vegetation Index (EVI), which can more comprehensively reflect vegetation growth and coverage. The NDVI formula simply and intuitively reflects vegetation coverage, while the EVI formula further optimizes vegetation assessment by introducing constants. This can particularly reduce the impact of atmospheric and soil background in areas of high vegetation coverage, improving the sensitivity and accuracy of the vegetation index to vegetation changes. Regarding texture feature measurement, the gray-level co-occurrence matrix is ​​used to calculate the texture features of remote sensing image data, including multiple features such as mean, variance, contrast, correlation, and entropy. These texture features can describe the texture characteristics of the image from different perspectives, providing richer information for the classification and analysis of forest resources and facilitating a more detailed distinction between different types of forest land and their changes.

[0051] 3. This invention features an innovative model structure for model construction and training, constructing an incomplete multi-view clustering model that integrates a time-series graph. This model includes a vegetation index view input layer, a texture feature view input layer, and a geographic information view input layer, and is capable of comprehensively considering multiple data features. By constructing a time-weighted graph that integrates the time-series graph and introducing a temporal regularization term into the model's objective function, the model can better utilize time-series information, improving its ability to analyze the dynamic changes in forestland resources and overcoming the problem of incomplete consideration of dynamic changes and complex spatiotemporal relationships in existing technologies.

[0052] 4. Regarding area calculation, the present invention accurately calculates the area value of each cluster region at different time points by clustering results at different time periods, determining the pixel range of the cluster in the remote sensing image, and then calculating the area of ​​the cluster region according to a formula. Furthermore, based on the clustering results, the area change of each cluster region can be calculated and the trend of cluster region change can be analyzed. This allows for a multi-dimensional and multi-regional analysis of the dynamic changes in forest resources, providing a detailed and accurate basis for forest resource management. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] Figure 1 This is a flow chart of a statistical method for dynamically changing areas of forest resources according to Example 1 of the present invention. DETAILED DESCRIPTION

[0054] The present invention will be further described in detail below with reference to the accompanying drawings.

[0055] Example 1

[0056] Reference Figure 1 A statistical method for dynamically changing forestland resource areas in this embodiment includes:

[0057] S1. Obtain a forest resource dataset covering the statistical area, the forest resource dataset including remote sensing image data of different periods, geographic information data of the corresponding period, and field measurement data of the corresponding period, and preprocess the forest resource dataset;

[0058] For large-scale forest monitoring requiring higher spatial resolution, the Landsat series of satellites is recommended. High-resolution images can provide more detailed information on local forest characteristics, such as the individual morphology of trees and stand structure. Medium-resolution images can cover larger areas and are used to obtain the overall distribution of forests.

[0059] The forest resource data is pre-processed, including obtaining high-resolution and medium-resolution remote sensing image data using satellite remote sensing, converting the remote sensing image data into physical radiation data, and performing radiation correction and geometric correction on the physical radiation data to match the remote sensing image data with the geographic coordinate system. The radiation correction formula is:

[0060]

[0061] in, represents the surface reflectivity, Represents the digital value of remote sensing image, is the quantization coefficient, and is the radiation calibration parameter of the sensor.

[0062] Geometric correction involves selecting distinct features on the image as ground control points (GCPs). These features should be easily identifiable in images from different periods, such as road intersections, corners of large buildings, and river confluences. Using the GCPs, a conversion relationship between image coordinates and geographic coordinates is established, and the image is geometrically corrected using a polynomial transformation.

[0063] Preprocessing of forest resource data includes format conversion and coordinate unification of geographic information data, checking of field measured data and removal of errors and outliers in the measured data.

[0064] If there are missing values ​​in the measured data, interpolation methods can be used to fill them. Common interpolation methods include linear interpolation and kriging interpolation. Linear interpolation is suitable for cases where data points are relatively evenly distributed. It uses the values ​​of adjacent data points to make a linear estimate. Kriging interpolation takes into account the spatial correlation of the data and can provide more accurate interpolation results. It is particularly suitable for data with spatial distribution characteristics, such as forest land measured data.

[0065] S2. Calculate vegetation indices and measure texture features at different times on the preprocessed data, extract features from the geographic information data, and construct a time series map training set;

[0066] The vegetation index calculation includes normalized difference vegetation index and enhancement of vegetation index. The normalized difference vegetation index formula is:

[0067]

[0068] in, and are the reflectances of the near-infrared and red bands, It represents the difference vegetation index, which is used to indicate the growth status and coverage of vegetation. Its value range is between -1 and 1. Positive values ​​indicate vegetation coverage, and negative values ​​indicate no vegetation coverage.

[0069] The vegetation index is enhanced as follows:

[0070]

[0071] in, 、 and is a constant, usually taking the value: , , , so that the enhanced vegetation index EVI can effectively reflect the growth status and vegetation coverage of vegetation, reduce the impact of factors such as atmosphere and soil background on the vegetation index, and the value of EVI may be more able to highlight the growth of vegetation than NDVI, especially in areas with high vegetation coverage.

[0072] The texture feature measurement includes calculating the texture features of remote sensing image data through a gray level co-occurrence matrix, and the texture features include mean, variance, contrast, correlation and entropy.

[0073] The step of constructing a time series graph training set includes:

[0074] A1. Arrange remote sensing image data, geographic information data, and field measurement data acquired at different times in chronological order to form a time series dataset.

[0075] Remote sensing imagery data should cover multiple time points. These time points should be representative and reflect the dynamic changes in forest resources. Imagery can be acquired in the same season each year to minimize seasonal influences on vegetation growth. Geographic information data and field measurement data should also be collected for the corresponding time periods to ensure a one-to-one correspondence between the three.

[0076] A2. Associate remote sensing image data, geographic information data, and field measurement data to establish a corresponding relationship in time series;

[0077] Using Geographic Information System (GIS) technology, remote sensing imagery data and geographic information data are linked based on geographic coordinates. This ensures that every pixel in the imagery accurately corresponds to a corresponding location in the geographic information data. This requires georeferencing the imagery to align its coordinate system with the coordinate system of the geographic information data. For the common WGS84 coordinate system, ensure that both the imagery and the geographic information data use this coordinate system. Field measurement data is linked to the remote sensing imagery and geographic information data based on the measurement locations. Field measurement data is typically measured at specific locations, which correspond to locations in the imagery. For example, multiple sampling points are set up in a forest to measure tree height and diameter at breast height, and the geographic coordinates of each sampling point are recorded. These coordinates can be used to find the corresponding pixel location in the imagery and retrieve relevant information about that location in the geographic information data. This establishes a temporal correspondence between the field measurement data, the imagery, and the geographic information data.

[0078] A3. Use triples to represent relationships in the time series graph, annotate the valid time range for each triple, and generate a time series graph.

[0079] A triplet typically consists of a subject, a relationship, and an object. The subject is a forest patch or area, the relationship is words like "has," "is located in," and "belongs to" that describe the relationship between the subject and the object, and the object is a vegetation index or geographic information feature.

[0080] A4. Combine and annotate vegetation indices, texture features, and geographic information from different periods in the time series atlas to construct a time series atlas training set.

[0081] For each time point in the time series, vegetation indices, texture features, and geographic information from different periods are combined. For example, for the year 2018, the Normalized Difference Vegetation Index (NDVI) and Enhanced Vegetation Index (EVI) calculated in 2018, along with texture features (mean, variance, contrast, correlation, and entropy) calculated from 2018 remote sensing imagery, are combined with the corresponding geographic information features from 2018 (such as slope, aspect, land use type, and administrative division code) to form a single feature vector.

[0082] When combining features, it is necessary to ensure that each feature has been properly preprocessed and calculated, and is comparable and consistent, and to label the combined feature vectors. The content of the labeling can be determined based on the research purpose and actual needs. This embodiment provides a specific example:

[0083] Labeling by forestland type: If the research objective is to distinguish different forestland types, labeling can be based on a combination of tree species information, vegetation index characteristics, and texture and geographic information features from field data. For a combined feature vector, if the vegetation index indicates high vegetation cover, relatively uniform and fine texture features, the geographic information features indicate a moderate slope, the land use type is forested, and the administrative area is located in a mountainous area, and field data indicates that the dominant tree species in the area is pine, then the feature vector can be labeled "Coniferous Forest (Pine)." If the vegetation cover is also high, but the texture features are relatively coarse, and field data indicates that the dominant tree species are broadleaf species such as poplar, then the labeling can be "Broadleaf Forest (Poplar)." If a mixed forest consists of multiple tree species and the vegetation index, texture, and geographic information features meet the characteristics of a mixed forest, then the labeling can be "Mixed Forest."

[0084] Forest land health is labeled based on vegetation growth vitality as reflected by the vegetation index, vegetation distribution uniformity as reflected by texture features, and factors in geographic information that may affect vegetation health (such as the relationship between slope, aspect, moisture, and light). If the NDVI and EVI values ​​in a feature vector are within a reasonable range and relatively stable, and the variance of the texture features is small, indicating relatively uniform vegetation growth and distribution, and the geographic information features indicate a gentle slope and suitable light and moisture conditions, then the area can be labeled "Healthy." If the vegetation index is low, the contrast of the texture features is high (possibly indicating uneven vegetation distribution with patchy loss), and the geographic information features indicate a steep slope and a risk of soil erosion (e.g., an aspect that is not conducive to water retention), then the area is labeled "Sub-Healthy (possibly affected by topography)." If the vegetation index is extremely low, the texture features are chaotic, and the geographic information features indicate that the area may be severely disturbed (e.g., near areas with frequent human activity and a trend of changing land use types), then the area is labeled "Unhealthy (possibly affected by human activity)."

[0085] The annotation is carried out according to the changing trend of forest land. The annotation content is determined by comparing the feature vectors at different time points and analyzing the changes in vegetation index, texture characteristics and geographic information characteristics.

[0086] For a region's feature vector from 2016 to 2018, if the vegetation index gradually increases from 2016 to 2018, the mean and variance of the texture features also show corresponding stable changes (e.g., an increase in the mean and a decrease in the variance, possibly indicating an increase in and more uniform vegetation cover), and the land use type and administrative divisions in the geographic information features remain unchanged, then the feature vector can be labeled "Increasing." If the vegetation index remains essentially unchanged, the texture features show no significant changes, and the geographic information features are stable, then the feature vector can be labeled "Stable." If the vegetation index decreases, some texture features show unusual changes (e.g., an increase in contrast, possibly indicating a decrease in vegetation cover and patchy loss), and the geographic information features indicate possible human intervention (e.g., a trend of land use type shifting toward non-forest land), then the feature vector can be labeled "Decreasing (likely affected by human activities)."

[0087] S3. Construct an incomplete multi-view clustering model that integrates the time series graph, train the model using the time series graph training set, and use the trained clustering model to cluster the forest resources in the statistical area;

[0088] The incomplete multi-view clustering model for fusion time series graph is constructed, including:

[0089] B1. Constructing a model, the model includes a vegetation index view input layer, a texture feature view input layer, and a geographic information view input layer;

[0090] B2. Constructing and utilizing time-weighted graphs Fusion time series graph, where nodes Represents data samples at different time points, represents the temporal relationship between samples, Represents the weight, and the calculation formula is: ,in represents the weight, and Represents two time points, represents the attenuation coefficient;

[0091] B3. Introducing a time regularization term into the model’s objective function ,in Represents the feature representation matrix of all data samples.

[0092] The model is trained using the time series graph training set, including

[0093] C1. Input the time series atlas training set into the model, extract features through the vegetation index view input layer, texture feature view input layer, and geographic information view input layer, and concatenate the extracted feature representations to obtain a fused feature representation. The formula is: ,in, represents the feature representation obtained from the vegetation index view, represents the feature representation obtained from the texture feature view, Represents the feature representation obtained from the geographic information view;

[0094] C2. Calculate the adjusted feature representation based on the time weighted graph and the time regularization term, the formula is: ,in, is the adjustment coefficient;

[0095] C3. Calculate the relationship score between entities in different regions. The formula is: ,in, and Represent the adjusted feature representations of different regions respectively;

[0096] C4. Constructing cross entropy loss function ,in, Represents a sample Belong to cluster The true label, Represents the model prediction sample Belong to cluster The probability of the model is obtained by using the gradient descent algorithm to optimize the model parameters and repeat the forward propagation calculation until the loss function value of the model converges and the model training is completed.

[0097] S4. Calculate the area of ​​forest resource changes, including calculating the area of ​​each cluster region for clustering results in different periods, obtaining the area value of each cluster region at different time points, and calculating the area change of each cluster region.

[0098] For the clustering results of different periods, the area of ​​each cluster region is calculated respectively, including determining the pixel range of each cluster in the remote sensing image, marking the image through the image segmentation algorithm, identifying the pixels belonging to each cluster, calculating the number of pixels in each cluster region, and calculating the area of ​​the cluster region. The formula is:

[0099]

[0100] in, is the image resolution in meters / pixel, is the number of pixels.

[0101] The calculation of the area change of each cluster region includes analyzing the cluster region change trend, and the formula is: ,in, Indicates the area change rate. A negative number indicates a decrease in forest area, and a positive number indicates an increase in forest area. Indicates time The area of ​​the clustering region, Indicates that the cluster area is at time and time The change in time, .

[0102] The area change rate is used to interpret the changing trend of the cluster area. If the area change rate is positive, it means that the forest area has increased; if the area change rate is negative, it means that the forest area has decreased; if the area change rate is 0, it means that the forest area has remained unchanged.

[0103] Example 2

[0104] This embodiment provides a statistical system for dynamically changing forestland areas, including:

[0105] a data acquisition module configured to acquire a forest resource dataset covering a statistical area, the forest resource dataset including remote sensing image data of different periods, geographic information data of corresponding periods, and field measurement data of corresponding periods, and preprocess the forest resource dataset;

[0106] The data processing module is configured to calculate vegetation index and texture feature measurement at different periods of the pre-processed data, extract features from the geographic information data, and construct a time series map training set;

[0107] a model training module configured to construct an incomplete multi-view clustering model fused with the time series graph, train the model using the time series graph training set, and cluster the forestland resources in the statistical area using the trained clustering model;

[0108] The area statistics module is configured to calculate the area change of forest resources, including calculating the area of ​​each cluster area for clustering results in different periods, obtaining the area value of each cluster area at different time points, and calculating the area change of each cluster area.

[0109] A computer-readable storage medium stores a plurality of instructions, wherein the instructions are suitable for being loaded and executed by a processor of a terminal device, for a statistical method of dynamically changing area of ​​forest resources.

[0110] A terminal device includes a processor and a computer-readable storage medium, wherein the processor is used to implement various instructions; the computer-readable storage medium is used to store multiple instructions, and the instructions are suitable for being loaded and executed by the processor to obtain a statistical method for dynamically changing areas of forest resources.

[0111] The above are all preferred embodiments of the present invention, and are not intended to limit the scope of protection of the present invention. Therefore, any equivalent changes made based on the structure, shape, and principle of the present invention should be included in the scope of protection of the present invention.

Claims

1. A statistical method for dynamically changing forest resource areas, characterized in that: include: Acquire a forest resource dataset covering the statistical area, the forest resource dataset including remote sensing image data of different periods, geographic information data of the corresponding periods, and field measurement data of the corresponding periods, and preprocess the forest resource dataset; Calculate vegetation index and texture feature measurements at different times for the preprocessed data, extract features from geographic information data, and construct a time series map training set; Construct an incomplete multi-view clustering model that integrates time series graphs, train the model using the time series graph training set, and use the trained clustering model to cluster the forest resources in the statistical area; The constructing of an incomplete multi-view clustering model fused with a time series graph includes constructing a model including a vegetation index view input layer, a texture feature view input layer, and a geographic information view input layer, fusing the time series graph with a time weighted graph, and introducing a time regularization term into the objective function of the model; The model is trained using the time series graph training set, including inputting the time series graph training set into the model, extracting features through the vegetation index view input layer, the texture feature view input layer, and the geographic information view input layer, and concatenating the extracted feature representations to obtain a fused feature representation, calculating an adjusted feature representation based on a time weighted graph and a time regularization term, calculating relationship scores between entities in different regions, constructing a cross entropy loss function, and optimizing model parameters using a gradient descent algorithm until the loss function value of the model converges, thereby completing model training; The area change of forest resources is calculated, including the clustering results of different periods, the area of ​​each cluster area is calculated separately, the area value of each cluster area at different time points is obtained, and the area change of each cluster area is calculated.

2. The statistical method for dynamically changing forest resource areas according to claim 1, characterized in that: The pre-processing of the forest resource data includes converting the remote sensing image data into physical radiation data and performing radiation correction and geometric correction on the physical radiation data to match the remote sensing image data with the geographic coordinate system.

3. The statistical method for dynamically changing forest resource areas according to claim 1, characterized in that: The pre-processed data is subjected to vegetation index calculation and texture feature measurement at different periods, including normalized difference vegetation index and vegetation index enhancement, and texture features of remote sensing image data are calculated through gray level co-occurrence matrix, wherein the texture features include mean, variance, contrast, correlation and entropy.

4. The statistical method for dynamically changing forest resource areas according to claim 3, characterized in that: The construction of the time series atlas training set includes associating remote sensing image data, geographic information data and field measurement data to establish a corresponding relationship in the time series, using triples to represent the relationship in the time series atlas, and marking the effective time range of each triple to generate a time series atlas, and combining and marking the vegetation index, texture features and geographic information of different periods in the time series atlas to construct the time series atlas training set.

5. The statistical method for dynamically changing forest resource areas according to claim 1, characterized in that: For the clustering results of different periods, the area of ​​each cluster region is calculated respectively, including determining the pixel range of each cluster in the remote sensing image, marking the image through an image segmentation algorithm, identifying the number of pixels belonging to each cluster, and using the number of pixels of the cluster to calculate the area of ​​the cluster region.

6. A statistical system for dynamically changing forest resource areas, which executes the statistical method for dynamically changing forest resource areas according to claim 1, characterized in that: include: a data acquisition module configured to acquire a forest resource dataset covering a statistical area, the forest resource dataset including remote sensing image data of different periods, geographic information data of corresponding periods, and field measurement data of corresponding periods, and preprocess the forest resource dataset; The data processing module is configured to calculate vegetation index and texture feature measurement at different periods of the pre-processed data, extract features from the geographic information data, and construct a time series map training set; a model training module configured to construct an incomplete multi-view clustering model fused with the time series graph, train the model using the time series graph training set, and cluster the forestland resources in the statistical area using the trained clustering model; The area statistics module is configured to calculate the area change of forest resources, including calculating the area of ​​each cluster area for clustering results in different periods, obtaining the area value of each cluster area at different time points, and calculating the area change of each cluster area.

7. A computer-readable storage medium storing a plurality of instructions, characterized in that: The instructions are suitable for being loaded and executed by a processor of a terminal device, for example, a statistical method for dynamically changing areas of forest resources as claimed in claim 1.

8. A terminal device comprising a processor and a computer-readable storage medium, wherein the processor is configured to implement various instructions; and the computer-readable storage medium is configured to store a plurality of instructions, wherein: The instructions are suitable for being loaded by a processor and executed by a statistical method for dynamically changing areas of forest resources as claimed in claim 1.

Images