Land resource monitoring method and system

Abstract

The invention provides a land resource monitoring method and system, and the method comprises the steps: carrying out high-altitude shooting through employing a multi-mode sensor loaded by an unmanned plane or a satellite, obtaining land surface and atmospheric environment data, carrying out the preliminary cleaning and standardization processing of the data, and carrying out the data fusion through employing a convolutional neural network. And then feature recognition is performed through unsupervised deep clustering fusion UDCF, and spatial positioning and analysis are performed on recognized feature point data in combination with a geographic information system (GIS) and a double-frequency global positioning system (GPS) so as to excavate the relationship between geography, climate and time factors and land resources. And finally, all data are integrated to form a visual land resource distribution report, and the report comprises various features, spatial distribution and time change information of land resources. According to the method, comprehensive, efficient and accurate land resource monitoring can be realized, and the scientific and technological level of land resource monitoring and management can be improved.

Description

Technical Field

[0001] The present invention belongs to the field of land and resources, and more specifically relates to a land resource monitoring method and system. Background Art

[0002] Traditional land resource monitoring methods mainly include field surveys, remote sensing image interpretation, etc. Field surveys have high accuracy, but have a long operation cycle, high cost, and are greatly restricted by topography. Remote sensing image interpretation can obtain large-scale land resource information simultaneously, with high efficiency and low cost. However, since professional personnel are required for manual interpretation, the efficiency is low and there is a large degree of subjectivity.

[0003] With the development of information technology and remote sensing technology, remote sensing image automatic interpretation technology has emerged. By applying machine learning and deep learning technologies, the accuracy and efficiency of obtaining land resource information can be further improved. However, since the existing remote sensing image automatic interpretation technology still has some problems, such as the recognition accuracy for complex terrains and complex land types is not high enough, and the processing ability for time series data is weak, etc.

[0004] Therefore, in view of the above problems, the present invention provides a method and system for obtaining multi-modal sensor data based on drones or satellites, and for automatically monitoring and analyzing land resources by using big data processing and deep learning technologies, aiming to more accurately and efficiently achieve land resource monitoring and management. Summary of the Invention

[0005] The present invention relates to a land resource monitoring method and system, especially an automatic land resource monitoring method and system implemented by using modern remote sensing technology and information technology. In current social production activities, as the most basic production factor, the reasonable development, scientific management, protection and utilization of land resources have extremely important significance. And quickly, accurately and comprehensively obtaining land resource information is an important means to realize the optimal allocation and sustainable utilization of land resources.

[0006] To achieve the above object, the present invention is implemented by adopting the following technical solutions: The method includes:

[0007] Data acquisition: Using a drone or satellite equipped with a multi-modal sensor to take aerial photos to obtain land surface and atmospheric environment data;

[0008] Data preprocessing: Conducting preliminary cleaning and standardization processing on the obtained multi-source data;

[0009] Data fusion: Using a convolutional neural network for multi-source data fusion to fuse the preprocessed data;

[0010] Feature recognition: Unsupervised Deep Clustering Fusion (UDCF) is used to recognize features from the fused data;

[0011] Spatial positioning and analysis: Combining Geographic Information System (GIS) and Dual-frequency Global Positioning System (GPS) to perform spatial positioning and analysis on the identified feature point data, and to explore the relationships between geographical, climatic, temporal factors and land resources;

[0012] Data integration and visualization: Integrate all data to form a visualized land resource distribution report, which includes various features, spatial distribution and temporal change information of land resources.

[0013] In one solution, sensors for visible light, infrared, polarization, and hyperspectral multi-spectrum.

[0014] In one solution, the data preprocessing includes: Noise removal is achieved through a median filter, expressed as:

[0015]

[0016] where represents the filtered pixel value, x(i,j) is the original pixel value, and k and l define the size of the filter window;

[0017] Missing values are processed by mean filling, and the formula is:

[0018]

[0019] where x filled is the filled value, N is the number of non-missing values, and x i are non-missing values.

[0020] In one solution, the data fusion is achieved through a Convolutional Neural Network (CNN), integrating data from multi-source sensors and extracting high-level feature representations; separate CNN modules are designed for each data source; these modules will extract features of each data source through convolution and pooling operations.

[0021] In one solution, the feature recognition is performed on the fused data by Unsupervised Deep Clustering Fusion (UDCF); UDCF learns the intrinsic structure and patterns of the data in an unsupervised manner, classifies the features, and generates labels; throughout the process, UDCF achieves deep feature fusion by self-generating labels and integrating multi-view information.

[0022] In one solution, the spatial positioning and analysis are as follows: obtaining the precise geographical coordinates of each feature point through GPS technology; obtaining the geographical coordinates of the feature points and importing this data into GIS for spatial analysis; and conducting correlation analysis between the feature points and other geographical information.

[0023] In one solution, the specific steps of the spatial positioning and analysis are as follows:

[0024] Obtaining the precise geographical coordinates of each feature point through GPS technology. GPS positioning is achieved by solving the pseudorange equation:

[0025]

[0026] where ρi is the pseudorange to the i-th satellite, (x i , y i , z i ) are the coordinates of the satellite, (x, y, z) are the coordinates of the receiver, c is the speed of light, dT and dT i are the clock biases of the receiver and the satellite;

[0027] Once the geographical coordinates of the feature points are obtained, this data is imported into GIS for spatial analysis. Spatial analysis involves the following steps:

[0028] Data integration and preprocessing: overlaying and matching the feature point data with other geographical layers to ensure data consistency and integrity;

[0029] Spatial interpolation and modeling: using spatial interpolation methods to model the feature point data to estimate the land resource characteristics of unobserved areas, with the mathematical expression:

[0030]

[0031] where is the estimated value at location s 0 , Z(s i ) is the observed value at location s i , and λ i is the weight coefficient;

[0032] Relationship analysis and visualization: evaluating the relationships between geographical, climatic, and temporal factors and land resource characteristics through spatial statistical analysis; using the mapping function of GIS to display the analysis results graphically.

[0033] On the other hand, a land resource monitoring system, the system is applicable to the method, and the system includes:

[0034] Data acquisition module: using an unmanned aerial vehicle or satellite equipped with multi-modal sensors to take aerial photos to obtain land surface and atmospheric environment data;

[0035] Data preprocessing module: perform preliminary cleaning and standardization on the acquired multi-source data;

[0036] Data fusion module: use a convolutional neural network for multi-source data fusion to fuse the preprocessed data;

[0037] Feature recognition module: perform feature recognition on the fused data through unsupervised deep clustering fusion UDCF;

[0038] Spatial positioning and analysis module: combine Geographic Information System GIS and dual-frequency Global Positioning System GPS to perform spatial positioning and analysis on the identified feature point data, and explore the relationships between geographical, climatic, temporal factors and land resources;

[0039] Data integration and visualization module: integrate all data to form a visualized land resource distribution report.

[0040] Advantages of the present invention:

[0041] When preprocessing and fusing the acquired data, the present invention adopts an advanced convolutional neural network algorithm, which improves the efficiency and accuracy of data deep processing and fusion, and can better mine and utilize the data collected by various modal sensors.

[0042] The present invention uses unsupervised deep clustering fusion UDCF to perform feature recognition on the fused data, which can deeply learn the internal structure and patterns of the data, understand various complex land features, and generate useful labels, improving the accuracy and stability of recognition.

[0043] The present invention performs spatial positioning analysis through Geographic Information System GIS and dual-frequency Global Positioning System GPS, which can accurately capture the relationships between geographical, climatic, temporal factors and land resources, providing an important scientific basis for the protection, utilization and management of land resources. Description of the drawings

[0044] Figure 1 is the flowchart of the method of the present invention;

[0045] Figure 2 is the flowchart of the spatial analysis of the present invention;

[0046] Figure 3 is the system block diagram of the present invention. Detailed implementation manners

[0047] To facilitate the understanding of the present invention, the present invention will be described more comprehensively below with reference to the relevant accompanying drawings. Typical embodiments of the present invention are shown in the drawings. However, the present invention can be implemented in many different forms and is not limited to the embodiments described herein. On the contrary, these embodiments are provided to make the disclosure of the present invention more thorough and comprehensive.

[0048] Unless otherwise defined, all technical and scientific terms used in the present invention have the same meaning as understood by those skilled in the technical field to which the present invention belongs. The terms used in the description of the present invention in this specification are only for the purpose of describing specific embodiments and are not intended to limit the present invention. To facilitate the understanding of the present invention, the present invention will be described more comprehensively below with reference to the relevant accompanying drawings. Typical embodiments of the present invention are shown in the drawings. However, the present invention can be implemented in many different forms and is not limited to the embodiments described herein. On the contrary, these embodiments are provided to make the disclosure of the present invention more thorough and comprehensive.

[0049] As Figure 1 shown, a land resource monitoring method includes: The method comprises the following steps: First, use a drone or satellite to take high-altitude photos to obtain image data of the land surface. Second, analyze and identify the image data through a deep learning algorithm to identify feature points related to land resources, including soil type, vegetation cover, topography, etc. Then, combine the geographic information system (GIS) and the dual-frequency global positioning system (GPS) to perform spatial positioning and analysis on the identified feature point data. Finally, integrate all the data through professional software to form a visual land resource distribution map and provide it to relevant decision-makers for accurate decision-making.

[0050] S1. Data acquisition: Use a drone or satellite equipped with multi-modal sensors, including sensors with multi-spectrum such as visible light, infrared, polarization, and hyperspectral, to obtain multi-source heterogeneous land surface and atmospheric environment data.

[0051] First, it is necessary to select a suitable drone or satellite platform according to the specific requirements and environmental conditions of the monitoring area. These platforms need to have high stability and flexibility to ensure high-quality data collection under various weather and terrain conditions.

[0052] The visible light sensor is used to capture the basic image information of the land surface, while the infrared sensor can detect temperature changes and vegetation health conditions. The polarization sensor provides additional surface information by measuring the polarization characteristics of light, and the hyperspectral sensor can capture subtle spectral changes to identify different soil and vegetation types.

[0053] After the sensor installation and calibration are completed, the drone or satellite conducts high-altitude photography according to the predetermined flight path and schedule. During the photography process, it is necessary to ensure the synchronous operation of the sensors to obtain multi-source heterogeneous data. This data not only includes the image information of the land surface but also covers the relevant data of the atmospheric environment, such as humidity, temperature, and aerosol concentration. These atmospheric data are crucial for subsequent image correction and analysis.

[0054] S2. Data preprocessing: First, perform preliminary cleaning and standardization on the obtained multi-source data, including removing noise, filling missing values, and unifying the data format for subsequent processing steps.

[0055] In step S2, data preprocessing is a crucial link to ensure the accuracy and reliability of subsequent analysis. First, conduct preliminary cleaning on the obtained multi-source data. The cleaning process includes removing noise and outliers. Noise removal can be achieved through a median filter, and the median filter is especially suitable for removing isolated noise points. Mathematically, it can be expressed as:

[0056]

[0057] where, represents the filtered pixel value, x(i,j) is the original pixel value, and k and l define the size of the filtering window.

[0058] Next, dealing with missing values is an important step in data preprocessing. Mean filling is adopted, and the formula is:

[0059]

[0060] where, x filled is the filled value, N is the number of non-missing values, and x i are non-missing values.

[0061] Unifying the data format and scale is a necessary step to ensure the consistency of different data sources. For format unification, data conversion tools can be used to convert all data into a unified format (such as GeoTIFF or NetCDF). Scale unification is achieved through normalization or standardization to eliminate the scale differences between different data sources. The commonly used standardization formula is:

[0062]

[0063] where, z is the standardized value, x is the original data value, μ is the mean, and σ is the standard deviation.

[0064] Finally, the cleaned and standardized data needs to be checked for format to ensure that the dimensions and formats of all datasets are consistent. This can be done automatically by writing scripts to ensure that there are no format errors or incompatibilities when the data is input into subsequent processing steps.

[0065] Through systematic data cleaning, missing value imputation, and standardization, a solid foundation is laid for subsequent data fusion and feature recognition. By applying appropriate mathematical tools and algorithms, the quality and consistency of multi-source data are ensured, thereby enhancing the performance and accuracy of the entire monitoring system.

[0066] S3. Data Fusion: Use a convolutional neural network for multi-source data fusion to fuse the preprocessed data. For each type of data, feature extraction is performed through convolution and pooling operations, and then a multi-layer deep convolutional neural network is used to fuse these features in order to obtain a unified and high-level feature representation.

[0067] In step S3, data fusion is achieved through a convolutional neural network (CNN) to integrate data from multi-source sensors and extract high-level feature representations. First, separate convolutional neural network modules need to be designed for each data source (such as visible light, infrared, hyperspectral, etc.). These modules will extract features of each data source through convolution and pooling operations.

[0068] The convolution operation is the core of feature extraction. By sliding the convolution kernel (filter) over the data, the weighted sum of the local region is calculated to extract local features. The convolution operation can be expressed as:

[0069]

[0070] where f(i, j) is the convolution result, x(i, j) is the input data, k(m, n) is the convolution kernel, and m and n are the dimensions of the kernel.

[0071] The pooling operation is used to reduce the dimension of the data while retaining important features. The commonly used pooling methods are max pooling and average pooling. Max pooling can be expressed as:

[0072] p(i, j) = max{x(i + m, j + n)}

[0073] where p(i, j) is the pooling result, x(i, j) is the input data, and m and n define the size of the pooling window.

[0074] For each data source, the feature maps extracted through several layers of convolution and pooling operations will be flattened and input into the fully connected layer. These feature maps are converted into one-dimensional feature vectors in the fully connected layer.

[0075] Next, the feature vectors from different data sources are fused. Using the feature concatenation method, all feature vectors are concatenated into a long vector:

[0076] F = [f 1 , f 2 , …, f n

[0077] where F is the fused feature vector, and f i is the feature vector of the i-th data source.

[0078] In a multi-layer deep convolutional neural network, these fused feature vectors will be input into subsequent convolutional layers and fully connected layers to further extract and optimize high-level feature representations. Through the backpropagation algorithm, the parameters of the network will be continuously adjusted to minimize the loss function, thereby improving the accuracy and robustness of the feature representation.

[0079] Effective fusion of multi-source data is achieved through a convolutional neural network, extracting unified and high-level feature representations. This process not only integrates the data advantages of different sensors but also improves the recognition ability for complex surface and atmospheric conditions, providing a more accurate basis for subsequent analysis and decision-making.

[0080] S4. Feature Recognition: Feature recognition is achieved by combining unsupervised learning and supervised learning methods. This method aims to explore the internal structure of the data using unsupervised learning and precisely classify the identified features through supervised learning. This hybrid method fully utilizes the exploration ability of unsupervised learning and the classification accuracy of supervised learning.

[0081] First, the core of the unsupervised learning stage lies in constructing a deep neural network to learn the latent representation of the data. Usually, this network can adopt a multi-layer autoencoder, whose purpose is to extract a compact and meaningful feature representation by encoding and decoding the input data. Mathematically, the objective of the autoencoder is to minimize the reconstruction error, and the formula is:

[0082]

[0083] where x is the input data, is the reconstructed data, and L is the loss function.

[0084] After the unsupervised learning stage, preliminary clustering analysis is performed using the deep feature representation. Based on the K-means or other clustering algorithms, pseudo-labels are generated. Its optimization objective is:

[0085]

[0086] where z i ​is the feature representation of the sample, is the clustering center, N is the number of samples.

[0087] Next, enter the supervised learning stage. Here, the pseudo-labels generated in the unsupervised stage are used as the initial labels to construct a classifier for training. The goal of the classifier is to optimize the model parameters by minimizing the classification loss function. The classification loss function is the cross-entropy loss, defined as:

[0088]

[0089] where y is the true label, is the predicted probability, C is the number of classes.

[0090] By alternately optimizing unsupervised feature learning and supervised classifier training, the classification accuracy of the model is gradually improved. In this process, multi-view information fusion is introduced to enhance the richness of the feature representation and the robustness of classification. The multi-view information can come from different data sources or the outputs of network layers. By integrating this information, more comprehensive feature patterns can be captured.

[0091] Finally, through the combination of unsupervised and supervised learning, the labels output by the network provide a structured classification result for the data. These labels not only reflect the internal structure of the data but also are precisely adjusted through supervised learning and are suitable for subsequent analysis and applications.

[0092] In summary, step S4 identifies the features of the data through the combination of unsupervised learning and supervised learning. This method uses unsupervised learning to explore the data structure and achieves precise classification through supervised learning, providing an effective solution for the identification and classification of complex data.

[0093] S5, Spatial Location and Analysis: Combine the Geographic Information System (GIS) and the dual-frequency Global Positioning System (GPS) to perform spatial location and analysis on the identified feature point data, mine the relationships between factors such as geography, climate, and time and land resources, and present the dynamic distribution of land resources in a graphical manner.

[0094] Spatial location and analysis process the identified feature point data by combining the Geographic Information System (GIS) and the dual-frequency Global Positioning System (GPS). The goal of this step is to mine the relationships between factors such as geography, climate, and time and land resources and present the dynamic distribution of land resources in a graphical manner.

[0095] First, the precise geographical coordinates of each feature point are obtained through GPS technology. Dual-frequency GPS can provide higher positioning accuracy, especially in environments with obvious multipath effects. Mathematically, GPS positioning can be achieved by solving the pseudorange equation:

[0096]

[0097] where ρi is the pseudorange to the i-th satellite, (x i ,y i ,z i ) are the coordinates of the satellite, (x, y, z) are the coordinates of the receiver, c is the speed of light, dT and dT i are the clock biases of the receiver and the satellite.

[0098] Once the geographical coordinates of the feature points are obtained, the next step is to import this data into GIS for spatial analysis. GIS provides powerful tools for processing and analyzing geographical data, and can perform correlation analysis between feature points and other geographical information (such as land use types, climate data, temporal changes, etc.).

[0099] As Figure 2 shown, in GIS, spatial analysis involves the following steps:

[0100] S501. Data integration and preprocessing: Overlay and match the feature point data with other geographical layers (such as climate, soil type) to ensure data consistency and integrity.

[0101] S502. Spatial interpolation and modeling: Use spatial interpolation methods (such as Kriging interpolation) to model the feature point data to estimate the land resource characteristics of unobserved areas. The mathematical expression of Kriging interpolation is:

[0102]

[0103] where is the estimated value at location s 0 , Z(s i ) is the observed value at location s i , and λ i are the weight coefficients.

[0104] S503. Relationship analysis and visualization: Through spatial statistical analysis, evaluate the relationships between geographical, climate, temporal and other factors and land resource characteristics. Use the mapping function of GIS to display the analysis results graphically, such as generating maps of dynamic changes in land resources, time series graphs, etc.

[0105] Through these steps, S5 realizes the spatial positioning and analysis of the identified feature points. Combining GIS and GPS technologies can not only accurately locate the feature points but also reveal the complex relationships between land resources and various environmental factors through comprehensive analysis. This graphical dynamic distribution display provides intuitive basis for decision-makers and supports the sustainable management and planning of land resources.

[0106] S6. Data integration and visualization: All data are integrated through professional software to form a visual land resource distribution report that can be provided to decision-makers. The report includes various features, spatial distribution, and temporal changes of land resources, etc., facilitating accurate decision-making by decision-makers.

[0107] The goal of data integration and visualization is to integrate all analysis results into a comprehensive land resource distribution report through professional software. This report aims to provide intuitive and detailed information support for decision-makers, including various features, spatial distribution, and temporal changes of land resources.

[0108] First of all, the process of data integration requires unified management of all data obtained in the previous steps (such as feature recognition, spatial positioning and analysis). It involves using a database management system (DBMS) to store and manage data to ensure data consistency and integrity. Data from different sources, such as geographical coordinates, climate information, land use types, etc., need to be standardized in the database for subsequent analysis and visualization.

[0109] Next, use professional software (such as ArcGIS, QGIS or Tableau) for data visualization. The key to visualization lies in selecting appropriate chart and map types to clearly present data features and relationships.

[0110] Import the integrated data into the visualization software, clean, filter, and transform the data as needed to ensure that the data format meets the visualization requirements.

[0111] Select suitable visualization types according to data features and report objectives. For example, use a heatmap to display the spatial density of land resources, use a time series chart to show the dynamic changes of land resources, and use a pie chart or bar chart to compare the proportions of different land types.

[0112] If the display of geographical information is involved, creating an interactive map is an effective way. The map can integrate multi-level information, such as terrain, land use, climate data, etc., and achieve multi-dimensional display through layer control. The layout design needs to consider the logical order and visual effect of information to ensure the readability and professionalism of the report.

[0113] Finally, all the visualizations are integrated into a complete report. This report not only includes static charts and maps but also can generate dynamic reports (such as PDF or web applications) through digital tools, allowing decision-makers to access and interact on different platforms.

[0114] Through these steps, the integration and visualization of data are achieved, transforming the results of complex data analysis into reports that are easy to understand and use. This report provides comprehensive land resource information support for decision-makers, helping them make accurate decisions in resource management and planning. Generally speaking, through this method, we can effectively integrate and utilize multi-source data, accurately identify and analyze the characteristics of land resources, thus achieving efficient and precise land resource monitoring.

[0115] As Figure 3 shown, a land resource monitoring system applicable to the method includes: a data acquisition module: using a drone or satellite equipped with multi-modal sensors to take aerial photos to obtain land surface and atmospheric environment data; a data preprocessing module: performing preliminary cleaning and standardization processing on the acquired multi-source data; a data fusion module: using a convolutional neural network for multi-source data fusion to fuse the preprocessed data; a feature recognition module: performing feature recognition on the fused data through unsupervised deep clustering fusion UDCF;

[0116] a spatial positioning and analysis module: combining a geographic information system GIS and a dual-frequency global positioning system GPS to perform spatial positioning and analysis on the identified feature point data, and exploring the relationships between geographical, climatic, and temporal factors and land resources; a data integration and visualization module: integrating all data to form a visualized land resource distribution report.

[0117] Those of ordinary skill in the art can understand that all or part of the processes of implementing the methods in the above embodiments can be completed by instructing relevant hardware through a computer program. The program can be stored in a computer-readable storage medium. When the program is executed, it can include the processes of the embodiments of the above methods. Among them, the storage medium can be a magnetic disk, an optical disc, a read-only memory (ROM), or a random access memory (RAM), etc.

[0118] It should be understood that the detailed description of the technical solutions of the present invention with the aid of the preferred embodiments above is illustrative rather than restrictive. Those of ordinary skill in the art can modify the technical solutions recorded in each embodiment or perform equivalent substitution on some of the technical features based on reading the specification of the present invention; and these modifications or substitutions do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of each embodiment of the present invention.

Claims

1. A land resource monitoring method, characterized in that: The method includes: Data acquisition: Use drones or satellites equipped with multimodal sensors to take aerial photos and obtain land surface and atmospheric environment data; Data preprocessing: preliminary cleaning and standardization of the acquired multi-source data; Data fusion: Use convolutional neural network for multi-source data fusion to fuse the pre-processed data; Feature recognition: Use unsupervised learning to learn the intrinsic structure of the data and supervised learning to accurately classify the identified features; Spatial positioning and analysis: Combined with the Geographic Information System (GIS) and the dual-frequency Global Positioning System (GPS), the identified feature point data is spatially positioned and analyzed to explore the relationship between geographical, climatic, temporal factors and land resources; Data integration and visualization: Integrate all data to form a visualized land resource distribution report, which includes various characteristics, spatial distribution and temporal change information of land resources.

2. A land resource monitoring method according to claim 1, characterized in that: The multimodal sensor includes: visible light, infrared, polarization, and hyperspectral multi-spectrum sensors.

3. A land resource monitoring method according to claim 1, characterized in that: The data preprocessing includes: noise removal is achieved by a median filter, which is expressed as: in, represents the pixel value after filtering, x(i,j) is the original pixel value, k and l define the size of the filtering window; The missing values ​​are filled with mean value, the formula is: Among them, x filled is the value after padding, N is the number of non-missing values, x i is a non-missing value.

4. A land resource monitoring method according to claim 1, characterized in that: The data fusion is achieved through convolutional neural networks, integrating data from multi-source sensors and extracting high-level feature representations; a separate convolutional neural network module is designed for each data source; These modules will extract features from each data source through convolution and pooling operations.

5. A land resource monitoring method according to claim 1, characterized in that: Speaking of feature recognition, first of all, the core of the unsupervised learning stage is to build a deep neural network to learn the potential representation of the data, using a multi-layer autoencoder, whose purpose is to extract compact and meaningful feature representation by encoding and decoding the input data.

6. A land resource monitoring method according to claim 1, characterized in that: The spatial positioning and analysis described herein: obtaining the precise geographic coordinates of each feature point through GPS technology; obtaining the geographic coordinates of the feature points, and importing these data into GIS for spatial analysis; and correlating the feature points with other geographic information for analysis.

7. A land resource monitoring method according to claim 6, characterized in that: The specific steps of the spatial positioning and analysis are as follows: The precise geographic coordinates of each feature point are obtained through GPS technology. GPS positioning is achieved by solving the pseudo-range equation: Where ρi is the pseudorange to the i-th satellite, (x i ,y i ,z i ) are the coordinates of the satellite, (x,y,z) are the coordinates of the receiver, c the speed of light, dT and dT i is the clock deviation between the receiver and the satellite; Once the geographic coordinates of the feature points are obtained, these data are imported into GIS for spatial analysis; spatial analysis involves the following steps: Data integration and preprocessing: Overlay and match feature point data with other geographic layers to ensure data consistency and integrity; Spatial interpolation and modeling: Use spatial interpolation methods to model feature point data to estimate the land resource characteristics of unobserved areas. The mathematical expression is: in, is the estimated value of position s0, Z(s i ) is the position s i The observed value of i is the weight coefficient; Relationship analysis and visualization: Evaluate the relationship between geographic, climatic, temporal factors and land resource characteristics through spatial statistical analysis; use GIS mapping capabilities to display analysis results in a graphical manner.

8. A land resource monitoring system, the system being applicable to the method according to any one of claims 1 to 7, characterized in that: The system comprises: Data acquisition module: Use drones or satellites equipped with multimodal sensors to take aerial photos and obtain land surface and atmospheric environment data; Data preprocessing module: performs preliminary cleaning and standardization on the acquired multi-source data; Data fusion module: Use convolutional neural network for multi-source data fusion to fuse the pre-processed data; Feature recognition module: It performs feature recognition on the fused data through unsupervised deep clustering and fusion of UDCF; Spatial positioning and analysis module: Combined with the Geographic Information System (GIS) and the dual-frequency global positioning system (GPS), the identified feature point data is spatially positioned and analyzed to explore the relationship between geographical, climatic, temporal factors and land resources; Data integration and visualization module: Integrate all data to form a visual land resource distribution report.

Images