Land space monitoring method and system based on satellite remote sensing

Abstract

The invention relates to the technical field of territorial space monitoring, and particularly discloses a territorial space monitoring method and system based on satellite remote sensing, and the method comprises the steps: carrying out the periodic observation of a target territorial space region through employing a plurality of different types of satellites, and recording the metadata information of the satellites; the collected data are preprocessed, and quality evaluation is carried out; constructing a feature extraction model based on a convolutional neural network by adopting a deep learning algorithm, performing feature extraction on the preprocessed satellite remote sensing image, and dividing territorial space ground features into different categories; the classification results of different periods are compared, the change condition of territorial space ground features is detected, a change area is analyzed, and the possible change trend in the future is predicted; fusing the processed data, and carrying out comprehensive analysis of territorial space; and outputting monitoring and analysis results in a visual form. According to the invention, the accuracy and reliability of territorial space monitoring are improved, and dynamic and comprehensive monitoring of territorial space is realized.

Description

Technical Field

[0001] The present invention relates to the field of land and space monitoring technology, and specifically to a land and space monitoring method and system based on satellite remote sensing. Background Art

[0002] Traditional land space monitoring mainly relies on manual field surveys, aerial photogrammetry and other means. These methods have many limitations: manual surveys are inefficient and have a small coverage area, making it difficult to meet the needs of dynamic monitoring of large-scale land space; although aerial photogrammetry can obtain image data over a larger area, it is restricted by factors such as weather conditions and flight costs, and cannot achieve high-frequency and normalized monitoring.

[0003] Satellite remote sensing technology has gradually become an important technical means for national land space monitoring due to its advantages such as wide coverage, short observation period, and data acquisition not restricted by geographical conditions. However, existing monitoring technologies based on satellite remote sensing still face many challenges:

[0004] First, single-type satellite data has limitations. For example, high-resolution optical satellites are easily obscured by clouds, and synthetic aperture radar satellites have limited imaging resolution and ground object recognition capabilities, making it difficult to comprehensively and accurately obtain complex national land and spatial information.

[0005] Second, in terms of data processing, traditional data preprocessing methods lack the accuracy for radiometric calibration, geometric correction, and atmospheric correction, leading to the accumulation of errors in subsequent object classification and change detection.

[0006] Third, existing land feature classification algorithms are mostly based on traditional machine learning models, which are difficult to adapt to the complex and changing land feature characteristics in terms of feature extraction and classification accuracy, especially the ability to distinguish similar land feature types;

[0007] Fourth, in the process of change detection and trend prediction, there is a lack of in-depth mining of spatiotemporal data, making it impossible to effectively predict the evolution trend of national land space;

[0008] Fifth, the multi-source data fusion method is still immature, and it is difficult to fully integrate the complementary information of satellite remote sensing data with other data such as terrain, planning, and meteorology, which limits the comprehensive analysis and decision-making support capabilities of monitoring results.

[0009] Therefore, a land space monitoring method and system based on satellite remote sensing has become an urgent problem to be solved. Summary of the Invention

[0010] The technical problem to be solved by the present invention is to provide a land space monitoring method and system based on satellite remote sensing, so as to improve the accuracy and reliability of land space monitoring and realize dynamic and comprehensive monitoring of land space.

[0011] To solve the above technical problems, the present invention provides a technical solution: a land space monitoring method based on satellite remote sensing, comprising the following steps:

[0012] S1. Utilize multiple satellites of different types, including high-resolution optical satellites, synthetic aperture radar satellites, and hyperspectral satellites, to periodically observe the target national territory space area according to a preset observation plan, and record satellite metadata information, including orbital parameters, observation time, and imaging angle;

[0013] S2. Preprocess the collected satellite remote sensing data, including radiometric calibration, geometric correction, and atmospheric correction. Then, perform quality assessment on the preprocessed data and eliminate data of substandard quality.

[0014] S3. Use deep learning algorithms to build a feature extraction model based on convolutional neural networks to extract features from pre-processed satellite remote sensing images, and use classification algorithms to classify land features into different categories;

[0015] S4. Compare the classification results of different periods and use a combination of difference analysis and threshold segmentation to detect changes in land features. Analyze the type, area, and trend of changes in the changed areas. Use spatiotemporal data mining algorithms to analyze the spatiotemporal distribution patterns of the changed areas and predict possible future trends.

[0016] S5. Fuse satellite remote sensing data with terrain data, land use planning data, and meteorological data, using feature-level and decision-level data fusion methods, and conduct comprehensive land space analysis based on the fused data;

[0017] S6. Output the monitoring and analysis results in a visual form and realize interactive display of monitoring results through the geographic information system platform.

[0018] Furthermore, the radiation calibration converts the grayscale value of the satellite observation data into the actual radiation brightness value; the geometric correction eliminates the geometric deformation caused by the satellite attitude and the earth's curvature; and the atmospheric correction removes the influence of the atmosphere on radiation transmission to obtain the real surface reflectivity or radiation brightness data.

[0019] Furthermore, in step S3, the feature extraction model based on convolutional neural network is trained by a large amount of labeled satellite remote sensing image data (x i ,y i ) for training, where x i is the image data, y i is the corresponding category label; the cross entropy loss function Loss is used to optimize the network parameters:

[0020]

[0021] Among them, N is the number of samples, C is the number of categories, and y ij is the true label of sample i belonging to category j, p ij The model predicts the probability that sample i belongs to category j;

[0022] The Softmax function is used to convert the network output into probability during classification: where z ij is the output value of the network when sample i belongs to category j.

[0023] Furthermore, in step S4, the method for detecting changes in land space features is as follows:

[0024] Difference analysis: Calculate the difference D between the corresponding pixels of the classification result images A and B in the two periods: D(x,y) = |A(x,y)-B(x,y|;

[0025] Threshold segmentation: Set the threshold T. If D(x,y)>T, it is determined that the area where the pixel is located has changed.

[0026] Furthermore, in step S4, the spatiotemporal data mining algorithm uses the spatiotemporal autoregressive model to predict future change trends. For a certain location (x, y) at time t+1, the feature category y x,y,t+1 :

[0027]

[0028] Among them, α i and β j is the model parameter, N(x,y) is the spatial neighborhood of (x,y), ∈ x,y,t+1 is the error term.

[0029] Furthermore, in step S5, a feature matching method is used to extract the feature vector F from the satellite remote sensing data. rs and the feature vector F extracted from other data other , matching fusion is performed by calculating the cosine similarity Sim: Merge feature vectors based on similarity;

[0030] Decision-level fusion: Using a weighted average method, the decision results D1, D2, ..., D m , the corresponding weights are w1,w2,…,w m , the fused decision result D final for:

[0031] The present invention also provides a land space monitoring system based on satellite remote sensing, which is used to implement the above method, including:

[0032] The satellite data acquisition module communicates with multiple satellites of different types, controls the satellites to conduct periodic observations of the target land space area according to the preset observation plan, records the satellite metadata information and obtains satellite remote sensing data;

[0033] The data preprocessing module performs radiometric calibration, geometric correction, and atmospheric correction on the collected satellite remote sensing data, performs quality assessment on the preprocessed data, and eliminates data of substandard quality;

[0034] The feature extraction and classification module builds a feature extraction model based on convolutional neural networks, extracts features from pre-processed satellite remote sensing images, and uses classification algorithms to classify land features into different categories.

[0035] The change detection and trend prediction module compares the classification results of different periods and uses a combination of difference analysis and threshold segmentation to detect changes in land features. It analyzes the change type, area, and trend of the changed areas, and uses a spatiotemporal autoregressive model to analyze the spatiotemporal distribution of the changed areas and predict possible future change trends.

[0036] The data fusion and analysis module integrates satellite remote sensing data with terrain data, land use planning data, and meteorological data at the feature and decision levels, and conducts comprehensive analysis of national land space based on the fused data;

[0037] The result output and display module outputs the monitoring and analysis results in a visual form and realizes the interactive display of monitoring results through the geographic information system platform.

[0038] Furthermore, in the data fusion and analysis module, a feature-level fusion method based on feature matching is used for vector data, and a decision-level fusion method based on weighted average is used for raster data.

[0039] The advantages of the present invention compared with the prior art are:

[0040] This invention leverages the advantages of each type of satellite through collaborative observation using high-resolution optical satellites, synthetic aperture radar satellites, and hyperspectral satellites. Optical satellites provide high-definition visual information of the Earth's surface, synthetic aperture radar satellites enable all-weather and all-day observation, and hyperspectral satellites capture the spectral characteristics of ground objects. The combination of these three can comprehensively cover the needs of land and space monitoring in different geographical environments and climate conditions, effectively avoiding data loss and providing a complete and accurate data foundation for land and space monitoring.

[0041] The present invention adopts scientific and rigorous radiometric calibration, geometric correction and atmospheric correction methods to accurately convert satellite observation data into real and reliable surface information.

[0042] This method uses a convolutional neural network to construct a feature extraction model, utilizing a cross-entropy loss function and a softmax function for optimization training. It can automatically learn the complex and ever-changing characteristic patterns of land features. Compared with traditional classification algorithms, this method significantly improves the ability to distinguish similar features, effectively improving the accuracy and reliability of feature classification and providing precise data support for refined land management.

[0043] This innovative approach, combining difference analysis, threshold segmentation, and a spatiotemporal autoregressive model, not only accurately detects areas of change in land and spatial features but also predicts future trends by mining the spatiotemporal dimensions of historical data. This dynamic monitoring and prediction capability helps identify potential issues in land and spatial utilization in advance, providing a forward-looking basis for scientific planning and decision-making.

[0044] This paper uses a data fusion method based on both feature and decision levels to achieve a deep fusion of satellite remote sensing data with multi-source data such as topography, land use planning, and meteorology. By using algorithms such as cosine similarity calculation and weighted averaging, it fully integrates the complementary information from various data sources and conducts a comprehensive multi-dimensional analysis of national land space. This method can more comprehensively and deeply reveal the evolution patterns and internal connections of national land space, providing more valuable decision-making support for national land space planning, ecological protection, and disaster warning.

[0045] This invention uses a geographic information system platform to visualize and interactively display monitoring results, transforming complex monitoring data into intuitive and easy-to-understand maps, charts, and 3D models. Users can quickly query and analyze land and space information for different regions and time periods through interactive operations, significantly improving the intuitiveness and efficiency of decision-making, reducing the cost of information comprehension, and facilitating the timely formulation and adjustment of land and space management strategies by relevant departments. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] Figure 1 The present invention is a flow chart of a land space monitoring method based on satellite remote sensing.

[0047] Figure 2 This is a system block diagram of a land space monitoring system based on satellite remote sensing of the present invention. DETAILED DESCRIPTION

[0048] Various exemplary embodiments of the present invention will be described in detail below with reference to the accompanying drawings. It should be noted that unless otherwise specifically stated, the relative arrangements of components and steps, numerical expressions and numerical values ​​set forth in these embodiments do not limit the scope of the present invention.

[0049] The following description of at least one exemplary embodiment is merely illustrative in nature and is in no way intended to limit the invention, its application, or uses.

[0050] Technologies, methods, and equipment known to ordinary technicians in the relevant art may not be discussed in detail, but where appropriate, the technologies, methods, and equipment should be considered part of the specification.

[0051] In all examples shown and discussed herein, any specific values ​​should be interpreted as merely exemplary and not limiting. Therefore, other examples of the exemplary embodiments may have different values.

[0052] The following is a detailed description of a land space monitoring method and system based on satellite remote sensing according to the present invention in conjunction with the accompanying drawings.

[0053] Combined with attachment Figure 1-2 , the present invention is introduced in detail.

[0054] A land space monitoring method based on satellite remote sensing comprises the following steps:

[0055] S1. Data Collection: Utilizing multiple satellites of varying types, including high-resolution optical satellites, synthetic aperture radar satellites, and hyperspectral satellites, we periodically observe the target national territory according to a pre-set observation plan and record satellite metadata, including orbital parameters, observation time, and imaging angle. Different types of satellites can provide different types of remote sensing data, enriching monitoring information.

[0056] S2. Data preprocessing: Preprocess the collected satellite remote sensing data, including radiation calibration, geometric correction, and atmospheric correction. The quality of the preprocessed data is evaluated and data with unqualified quality is eliminated.

[0057] Radiometric calibration: Convert the grayscale value of satellite observation data into the actual radiometric brightness value. The formula is: L = Gain × DN + Bias, where L is the radiometric brightness value, Gain is the gain coefficient, DN is the grayscale value, and Bias is the offset coefficient.

[0058] Geometric correction: Eliminate geometric deformation caused by satellite attitude and earth curvature factors, using the polynomial transformation formula:

[0059] where a ij and b ij are the polynomial coefficients, and n is the order of the polynomial.

[0060] Atmospheric correction: removes the influence of the atmosphere on radiation transmission to obtain true surface reflectance or radiation brightness data. Taking the 6S model as an example, the surface reflectance calculation formula is: Among them L p is the atmospheric radiation, T s is the atmospheric transmittance in the direction of the sun, Tv is the atmospheric transmittance in the observation direction, E s is the solar irradiance at the upper boundary of the atmosphere, θ s is the solar zenith angle, and d is the distance between the sun and the earth.

[0061] S3. Feature extraction and classification: Using deep learning algorithms, we build a feature extraction model based on convolutional neural networks to extract features from pre-processed satellite remote sensing images, and use classification algorithms to classify land features into different categories.

[0062] The feature extraction model based on convolutional neural network is trained through a large amount of labeled satellite remote sensing image data (x i ,y i ) for training, where x i is the image data, y i is the corresponding category label.

[0063] The cross entropy loss function Loss is used to optimize network parameters: Among them, N is the number of samples, C is the number of categories, and y ij is the true label of sample i belonging to category j, p ij The model predicts the probability that sample i belongs to category j.

[0064] The Softmax function is used to convert the network output into probability during classification: where z ij is the output value of the network when sample i belongs to category j.

[0065] S4. Change detection and trend prediction: Compare the classification results of different periods, use a combination of difference analysis and threshold segmentation to detect changes in land features, analyze the type, area and trend of changes in the changed areas, use spatiotemporal data mining algorithms to analyze the spatiotemporal distribution patterns of the changed areas, and predict possible future trends.

[0066] Difference analysis: Calculate the difference D between the corresponding pixels of the classification result images A and B in the two periods: D(x,y) = |A(x,y)-B(x,y|;

[0067] Threshold segmentation: Set the threshold T. If D(x,y)>T, it is determined that the area where the pixel is located has changed.

[0068] The spatiotemporal data mining algorithm uses the spatiotemporal autoregressive model to predict future change trends. For a certain location (x, y) at time t+1, the feature category y x,y,t+1 for: Among them, α i and β j is the model parameter, N(x,y) is the spatial neighborhood of (x,y), ∈x,y,t+1 is the error term.

[0069] S5. Data fusion and comprehensive analysis: Integrate satellite remote sensing data with terrain data, land use planning data, and meteorological data, adopt data fusion methods based on feature level and decision level, and conduct comprehensive analysis of national land space based on the fused data.

[0070] Feature-level fusion: Using a feature matching method, the feature vector F extracted from satellite remote sensing data is rs and the feature vector F extracted from other data other , matching fusion is performed by calculating the cosine similarity Sim: Merge feature vectors based on similarity;

[0071] Decision-level fusion: Using a weighted average method, the decision results D1, D2, ..., D m , the corresponding weights are w1,w2,…,w m , the fused decision result D final for:

[0072] S6. Result output and display: Output the monitoring and analysis results in a visual form, and realize interactive display of monitoring results through the geographic information system platform.

[0073] To implement the above method, the present invention also provides a land space monitoring system based on satellite remote sensing, comprising:

[0074] Satellite data acquisition module: Communicates with multiple satellites of different types, controls satellites to conduct periodic observations of target land space areas according to preset observation plans, records satellite metadata information and obtains satellite remote sensing data.

[0075] Data preprocessing module: performs radiometric calibration, geometric correction, and atmospheric correction on the collected satellite remote sensing data, performs quality assessment on the preprocessed data, and eliminates data of unqualified quality.

[0076] Feature extraction and classification module: Build a feature extraction model based on convolutional neural network, extract features from preprocessed satellite remote sensing images, and use classification algorithms to classify land and space objects into different categories.

[0077] Change detection and trend prediction module: Compare the classification results of different periods, use a combination of difference analysis and threshold segmentation to detect changes in land space objects, analyze the change type, area and change trend of the changed area, use the spatiotemporal autoregressive model to analyze the spatiotemporal distribution pattern of the changed area and predict possible future change trends.

[0078] Data Fusion and Analysis Module: This module fuses satellite remote sensing data with terrain data, land use planning data, and meteorological data at the feature and decision levels, and conducts comprehensive spatial analysis of the national territory based on the fused data. For vector data, a feature-matching-based approach is used for feature-level fusion, while for raster data, a weighted average-based approach is used for decision-level fusion.

[0079] Result output and display module: Output the monitoring and analysis results in a visual form, and realize interactive display of monitoring results through the geographic information system platform.

[0080] The specific implementation process of the land space monitoring method and system based on satellite remote sensing of the present invention is as follows:

[0081] Example 1: Land and Space Monitoring Method Based on Satellite Remote Sensing

[0082] Data collection:

[0083] Select appropriate high-resolution optical satellites, synthetic aperture radar satellites, and hyperspectral satellites, and develop a pre-set observation plan based on the scope of the target national land area and monitoring requirements. During the observation process, metadata such as the satellite's orbital parameters, observation time, and imaging angle are recorded. For example, for national land area monitoring of a specific city, a high-resolution optical satellite can be selected to conduct visible light observations during the day, while a synthetic aperture radar satellite can be used to conduct microwave observations at night or in inclement weather conditions to obtain more comprehensive information.

[0084] Data preprocessing:

[0085] Radiometric calibration: Converts the grayscale values ​​of collected satellite remote sensing data into actual radiometric brightness values ​​based on the satellite's calibration parameters. For example, for an optical satellite with a gain factor of 0.1 and a bias factor of 1, if the grayscale value of a pixel at DN = 100, then its radiometric brightness value L = 0.1 × 100 + 1 = 11.

[0086] Geometric correction: Using the digital elevation model and ground control points, a method combining polynomial transformation and affine transformation is used for geometric correction. First, a certain number of ground control points are selected and their coordinates in the image coordinate system and the geographic coordinate system are determined. Then, the polynomial coefficient a is calculated according to the polynomial transformation formula. ij and b ij , perform geometric correction on the image.

[0087] Atmospheric correction: Use the method based on radiation transfer model for atmospheric correction, such as 6S model. Input the radiation brightness L received by the sensor, the atmospheric path radiation L p , atmospheric transmittance T in the direction of the suns , atmospheric transmittance T in the observation direction v , the upper limit of the solar irradiance E s , solar zenith angle θ s , the distance between the sun and the earth d and other parameters are used to calculate the surface reflectivity ρ.

[0088] Quality Assessment: Data quality is assessed by calculating indicators such as the signal-to-noise ratio and contrast. If the data indicators are below the preset threshold, the data is considered to be of substandard quality and is discarded. For example, if the preset signal-to-noise ratio threshold is 10 and the signal-to-noise ratio of the data in a certain area is 8, the data in that area is considered to be of substandard quality.

[0089] Feature extraction and classification:

[0090] Data preparation: Collect a large amount of labeled satellite remote sensing image data (x i ,y i ), and divide it into training set, validation set and test set.

[0091] Model construction: Build a feature extraction model based on convolutional neural networks. For example, classic convolutional neural network architectures such as ResNet and VGG can be used.

[0092] Model training: The cross entropy loss function is used to optimize network parameters, and the stochastic gradient descent algorithm is used for training. During the training process, the network parameters are continuously adjusted to gradually reduce the loss function.

[0093] Classification: Use the trained model to extract features from the preprocessed satellite remote sensing images, use the Softmax function to convert the network output into probability, and determine the ground object category based on the principle of maximum probability.

[0094] Change detection and trend prediction:

[0095] Difference analysis: Calculate the difference D between the corresponding pixels of the classification result images A and B in the two periods. For example, for a certain pixel, if A(x,y) = 1 and B(x,y) = 2, then D(x,y) = |1-2| = 1.

[0096] Threshold segmentation: The Otsu algorithm is used to automatically determine the threshold T. If D(x,y)>T, it is determined that the area where the pixel is located has changed.

[0097] Trend prediction: Determine the model parameter α based on the spatiotemporal autoregressive model i and β j , for a certain location (x, y) at time t+1, the feature category y x,y,t+1 Make predictions.

[0098] Data fusion and comprehensive analysis:

[0099] Feature-level fusion: The feature vector F extracted from satellite remote sensing data rs and the feature vector F extracted from other data other , calculate the cosine similarity Sim. If Sim is greater than the preset threshold, the feature vectors are merged. For example, if the preset threshold is 0.8, if Sim=0.9, then F rs and F other merge.

[0100] Decision-level fusion: Use the analytic hierarchy process to determine the weights w1, w2, ..., w of different data sources m , calculate the fusion decision result D according to the weighted average formula final .

[0101] Result output and display:

[0102] The monitoring and analysis results are output in the form of maps, charts, and other visual displays, and interactively displayed through the GIS platform. Users can query monitoring results, change trends, and other information for different regions through the GIS platform.

[0103] Example 2: National Land Space Monitoring System Based on Satellite Remote Sensing

[0104] A land space monitoring system based on satellite remote sensing includes a satellite data acquisition module, a data preprocessing module, a feature extraction and classification module, a change detection and trend prediction module, a data fusion and analysis module, and a result output and display module. The specific implementation of each module is as follows:

[0105] Satellite data acquisition module: receives remote sensing data and metadata information sent by satellites through satellite communication connection, and stores them in the database.

[0106] Data preprocessing module: reads satellite remote sensing data from the database, performs preprocessing operations such as radiation calibration, geometric correction, atmospheric correction and quality assessment, and stores the preprocessed data in a new database table.

[0107] Feature extraction and classification module: Extract satellite remote sensing images from preprocessed data, build a feature extraction model based on convolutional neural network, perform model training and classification, and store the classification results in the database.

[0108] Change detection and trend prediction module: reads classification results of different periods from the database, performs difference analysis, threshold segmentation and trend prediction, and stores the change detection and trend prediction results in the database.

[0109] Data fusion and analysis module: reads satellite remote sensing data, terrain data, land use planning data, meteorological data, etc. from the database, performs feature-level and decision-level fusion, conducts comprehensive analysis of the national land space based on the fused data, and stores the analysis results in the database.

[0110] Result output and display module: reads monitoring and analysis results from the database, outputs them in a visual form, and realizes interactive display through the geographic information system platform.

[0111] The present invention and its embodiments are described above. This description is not restrictive. The drawings show only one embodiment of the present invention, and the actual structure is not limited thereto. In short, if a person skilled in the art is inspired by this and, without departing from the purpose of the present invention, designs structures and embodiments similar to this technical solution without inventiveness, they shall fall within the scope of protection of the present invention.

Claims

1. A land space monitoring method based on satellite remote sensing, characterized in that: The following steps are involved: S1. Utilize multiple satellites of different types, including high-resolution optical satellites, synthetic aperture radar satellites, and hyperspectral satellites, to periodically observe the target national territory space area according to a preset observation plan, and record satellite metadata information, including orbital parameters, observation time, and imaging angle; S2. Preprocess the collected satellite remote sensing data, including radiometric calibration, geometric correction, and atmospheric correction. Then, perform quality assessment on the preprocessed data and eliminate data of substandard quality. S3. Use deep learning algorithms to build a feature extraction model based on convolutional neural networks to extract features from pre-processed satellite remote sensing images, and use classification algorithms to classify land features into different categories; S4. Compare the classification results of different periods and use a combination of difference analysis and threshold segmentation to detect changes in land features. Analyze the type, area, and trend of changes in the changed areas. Use spatiotemporal data mining algorithms to analyze the spatiotemporal distribution patterns of the changed areas and predict possible future trends. S5. Fuse satellite remote sensing data with terrain data, land use planning data, and meteorological data, using feature-level and decision-level data fusion methods, and conduct comprehensive land space analysis based on the fused data; S6. Output the monitoring and analysis results in a visual form and realize interactive display of monitoring results through the geographic information system platform.

2. The method for land space monitoring based on satellite remote sensing according to claim 1, characterized in that: The radiation calibration converts the grayscale value of satellite observation data into actual radiation brightness value; the geometric correction eliminates the geometric deformation caused by satellite attitude and earth curvature factors; the atmospheric correction removes the influence of the atmosphere on radiation transmission to obtain true surface reflectivity or radiation brightness data.

3. The method for land space monitoring based on satellite remote sensing according to claim 2, characterized in that: In step S3, the feature extraction model based on convolutional neural network is trained by a large amount of labeled satellite remote sensing image data (x i ,y i ) for training, where x i is the image data, y i is the corresponding category label; the cross entropy loss function Loss is used to optimize the network parameters: Among them, N is the number of samples, C is the number of categories, and y ij is the true label of sample i belonging to category j, p ij The model predicts the probability that sample i belongs to category j; The Softmax function is used to convert the network output into probability during classification: where z ij is the output value of the network when sample i belongs to category j.

4. The method for land space monitoring based on satellite remote sensing according to claim 3, characterized in that: In step S4, the method for detecting changes in land space features is as follows: Difference analysis: Calculate the difference D between the corresponding pixels of the classification result images A and B in the two periods: D(x,y) = |A(x,y)-B(x,y|; Threshold segmentation: Set the threshold T. If D(x,y)>T, it is determined that the area where the pixel is located has changed.

5. The method for land space monitoring based on satellite remote sensing according to claim 4, characterized in that: In step S4, the spatiotemporal data mining algorithm uses the spatiotemporal autoregressive model to predict future change trends. For a certain location (x, y) at time t+1, the feature category y x,y,t+1 : Among them, α i and β j is the model parameter, N(x,y) is the spatial neighborhood of (x,y), ∈ x,y,t+1 is the error term.

6. The method for land space monitoring based on satellite remote sensing according to claim 5, characterized in that: In step S5, a feature matching method is used to extract the feature vector F from the satellite remote sensing data. rs and the feature vector F extracted from other data other , matching fusion is performed by calculating the cosine similarity Sim: Merge feature vectors based on similarity; Decision-level fusion: Using a weighted average method, the decision results D1, D2, ..., D m , the corresponding weights are w1,w2,…,w m , the fused decision result D final for:

7. A land and space monitoring system based on satellite remote sensing, used to implement the land and space monitoring method based on satellite remote sensing according to any one of claims 1 to 6, characterized in that: include: The satellite data acquisition module communicates with multiple satellites of different types, controls the satellites to conduct periodic observations of the target land space area according to the preset observation plan, records the satellite metadata information and obtains satellite remote sensing data; The data preprocessing module performs radiometric calibration, geometric correction, and atmospheric correction on the collected satellite remote sensing data, performs quality assessment on the preprocessed data, and eliminates data of substandard quality; The feature extraction and classification module builds a feature extraction model based on convolutional neural networks, extracts features from pre-processed satellite remote sensing images, and uses classification algorithms to classify land features into different categories. The change detection and trend prediction module compares the classification results of different periods and uses a combination of difference analysis and threshold segmentation to detect changes in land features. It analyzes the change type, area, and trend of the changed areas, and uses a spatiotemporal autoregressive model to analyze the spatiotemporal distribution of the changed areas and predict possible future change trends. The data fusion and analysis module integrates satellite remote sensing data with terrain data, land use planning data, and meteorological data at the feature and decision levels, and conducts comprehensive analysis of national land space based on the fused data; The result output and display module outputs the monitoring and analysis results in a visual form and realizes the interactive display of monitoring results through the geographic information system platform.

8. The land space monitoring system based on satellite remote sensing according to claim 7, characterized in that: In the data fusion and analysis module, a feature-level fusion method based on feature matching is used for vector data, and a decision-level fusion method based on weighted average is used for raster data.

Images