A forest fire monitoring method based on low earth orbit satellites

Abstract

The application discloses a kind of forest fire monitoring methods based on low-orbit satellite, comprising: obtaining data by multiple sensors, and generating fusion dataset by denoising and smoothing processing.Utilize convolutional neural network to combine multi-scale analysis to enhance fire point features, form feature set.Multiple satellite cooperative observation data are matched with feature set, and enhanced dataset of high frequency observation is generated by time series alignment.Adopt deep learning and transfer learning optimization model, extract key area features, share multiple satellite data through distributed computing, obtain coverage enhanced dataset.Based on timeliness evaluation model, dynamically adjust multiple satellite scheduling strategy, output scheduling result.Through reinforcement learning optimization algorithm robustness, combined with complex weather condition simulation, obtain robustness enhanced model, and utilize multi-source data verification technology to verify data fusion precision, output final monitoring result.The application significantly improves the accuracy, timeliness and coverage integrity of fire point detection.

Description

Technical Field

[0001] This invention belongs to the field of forest fire monitoring, and in particular relates to a forest fire monitoring method based on low-orbit satellites. Background Technology

[0002] Forest fire monitoring is a crucial issue in ecological protection and disaster prevention, its importance being self-evident. It directly relates to the stability of ecosystems, the safety of human life and property, and the sustainable development of society and the economy. The frequency and destructive power of forest fires have increased significantly, urgently requiring efficient and accurate monitoring technologies to address this severe challenge. However, current monitoring methods based on ground sensors or traditional remote sensing technologies have significant limitations. Ground monitoring is limited by narrow coverage, high deployment costs, and difficulty in implementation in remote areas; traditional high-orbit satellite remote sensing, due to insufficient resolution and low observation frequency, cannot meet the demands for real-time and high-precision monitoring. These shortcomings lead to difficulties in identifying fire points in the early stages of a fire, delayed early warnings, inaccurate fire assessments, and a lack of effective guidance for firefighting decisions.

[0003] Against this backdrop, the core technological challenges facing forest fire monitoring have become increasingly prominent, with the accuracy of data fusion, the real-time performance of intelligent analysis, and the coverage capability of multi-source collaboration becoming key bottlenecks. First, heterogeneous data acquired from multiple sensors (such as different bands and resolutions) are susceptible to noise interference during fusion, affecting the accuracy of fire detection. Second, real-time analysis of image data is limited by the algorithm's adaptability to complex scenarios, especially in areas with variable weather conditions or complex terrain, leading to significant false alarms and missed alarms. Finally, observations from a single satellite are limited by orbit and time, making high-frequency, all-weather coverage impossible, while the data sharing and complementarity mechanisms for multi-satellite collaboration are not yet mature, resulting in insufficient timeliness and completeness of monitoring. These technical challenges collectively restrict the transformation of forest fire monitoring from passive response to proactive early warning. Therefore, how to achieve accurate fusion of multi-sensor data on low-Earth orbit satellite platforms, improve the real-time performance and robustness of intelligent fire detection algorithms, and optimize high-frequency coverage capabilities through multi-satellite collaboration has become a key issue in improving the effectiveness of dynamic monitoring and early warning of forest fires. Solving this problem will directly promote the advancement of monitoring technology towards intelligence and systematization, providing more reliable technical support for forest fire prevention and control. Summary of the Invention

[0004] To address the aforementioned technical problems, this invention provides a forest fire monitoring method based on low-Earth orbit satellites, comprising:

[0005] Multi-sensor data is acquired from a low-Earth orbit satellite platform. The heterogeneous sensor data is processed using a pre-set denoising model. Adaptive filtering technology is used to smooth the data in different bands to obtain a fused dataset.

[0006] For the fused dataset, if the noise interference exceeds the preset threshold during complex scene recognition, features are extracted through a convolutional neural network, and the fire point features are enhanced by multi-scale analysis technology to obtain a feature set.

[0007] Acquire multi-satellite collaborative observation data, and match the multi-source collaborative coverage data with the feature set using time series alignment technology to obtain an enhanced dataset with improved high-frequency observation capabilities;

[0008] For augmented datasets, if the demand for complex scene recognition increases, the image analysis adaptability is optimized through deep learning algorithms, and the model parameters are adjusted by combining transfer learning techniques to output optimized results.

[0009] Key regional features are extracted from the optimization results, and a distributed computing framework is used to perform data sharing processing on the multi-satellite collaborative complementarity to obtain a coverage enhancement dataset.

[0010] Based on the coverage enhancement dataset, the observation frequency and completeness are analyzed using a preset timeliness evaluation model. If the timeliness and completeness are lower than expected, the multi-satellite collaborative scheduling strategy is dynamically adjusted, and the scheduling results are output.

[0011] Based on the scheduling results, reinforcement learning techniques are used to iteratively optimize the robustness of the algorithm. By simulating the fire detection task under complex weather conditions, a robustness enhancement model is obtained.

[0012] The final fire detection data is obtained from the robustness enhancement model, and the accuracy of data fusion is verified by multi-source data verification technology to output the final monitoring results.

[0013] Preferably, the process of obtaining the fused dataset includes:

[0014] Sensor data generated by multiple sensors is acquired from low-Earth orbit satellites, and heterogeneous characteristics are processed through a preset denoising model. Multiple sensors are carried on a low-Earth orbit satellite platform to acquire multi-source data of forest areas.

[0015] The acquired sensor data is preprocessed to eliminate data heterogeneity between different sensors, resulting in the fused dataset; wherein, the preprocessing includes data format unification, time synchronization, and spatial registration.

[0016] Preferably, the process of obtaining the feature set includes:

[0017] Feature extraction is performed on the fused dataset using a convolutional neural network, and feature maps of different scales are fused using a spatial pyramid transformation or feature pyramid network structure.

[0018] Meanwhile, during the feature extraction process, an attention mechanism is introduced to enhance the basic features and assign weights to different features to highlight the fire point features and suppress noise interference.

[0019] The enhanced features are fused with multi-scale features to obtain the feature set.

[0020] Preferably, the process of obtaining the augmented dataset includes:

[0021] Multi-source data was acquired through multi-satellite observations, and time-series alignment technology was used to process the data alignment, resulting in the first aligned dataset.

[0022] Cooperative coverage information is extracted from the aligned first dataset and fused with the feature set features through matching to obtain the fused second dataset;

[0023] For the fused second dataset, coverage analysis is used to determine the spatial distribution characteristics of the multi-source data and obtain the distribution characteristic set;

[0024] If the distribution characteristic set meets the preset threshold condition, the distribution characteristic set is classified by the random forest algorithm to determine the high-frequency observation area and obtain the region set.

[0025] Based on the correspondence between the region set and the time series, the temporal distribution characteristics of high-frequency observations are obtained, resulting in a time feature set;

[0026] By mapping the time feature set to the enhanced dataset, the enhanced regions of observation capability are determined, and the enhanced region set is obtained.

[0027] The augmented region set is optimized using data augmentation techniques to obtain an optimized augmented dataset.

[0028] Preferably, the process of optimizing image analysis adaptability using deep learning algorithms, adjusting model parameters using transfer learning techniques, and outputting the optimization results includes:

[0029] Preliminary analysis results are obtained by performing initial processing on the augmented dataset using deep learning algorithms.

[0030] Based on the preliminary analysis results, transfer learning techniques were used to adjust the model parameters for complex scenarios, resulting in an optimized parameter set.

[0031] If the optimized parameter set meets the preset threshold condition, then feature extraction is performed on the complex scene using image analysis technology to obtain the feature set;

[0032] The augmented dataset is matched using a feature set to obtain a matched dataset;

[0033] The distribution characteristics of scene recognition are determined by the matching dataset, and a distribution feature set is obtained;

[0034] Based on the mapping relationship between the distribution feature set and the optimization parameter set, obtain the mapping dataset;

[0035] The mapped dataset is optimized using data augmentation techniques to obtain the optimized result.

[0036] Preferably, the process of extracting key regional features from the optimization results and using a distributed computing framework to perform data sharing processing on multi-satellite collaborative complementarity to obtain a coverage enhancement dataset includes:

[0037] The distribution range of regional features is determined by identifying key regions, and a preliminary feature set is obtained by using feature extraction techniques.

[0038] Based on the preliminary feature set, multi-star collaborative task units are divided to obtain a collaborative and complementary allocation scheme;

[0039] A distributed computing framework is used to perform data sharing processing on the allocation scheme to obtain a shared data group;

[0040] If the shared data group meets the preset threshold conditions, the data range is adjusted through coverage enhancement technology to obtain the enhanced data group;

[0041] The enhanced data group is processed to optimize the workflow, resulting in optimized data units;

[0042] By analyzing the mapping relationship between the optimized data units and regional features, the complementary distribution of multi-star collaboration is determined, and a coverage enhancement dataset is obtained.

[0043] A computational framework is used to validate the coverage-enhanced dataset and determine the final distribution characteristics.

[0044] Preferably, based on the coverage enhancement dataset, the observation frequency and completeness are analyzed using a preset timeliness evaluation model. If the timeliness and completeness are lower than expected, the multi-satellite collaborative scheduling strategy is dynamically adjusted, and the scheduling results are output. The process includes:

[0045] By covering the enhanced dataset, the frequency and completeness of observations are analyzed using a pre-defined timeliness assessment model to obtain the analysis data set;

[0046] Based on the analyzed data set, determine whether the integrity is below a preset threshold condition, and obtain the judgment data unit;

[0047] The collaborative scheduling scheme is processed using dynamic adjustment technology through the aforementioned judgment data unit to obtain the adjustment data group;

[0048] For the adjusted data group, obtain multi-star collaborative task allocation information and determine the task allocation set;

[0049] By using the task allocation set and employing scheduling strategy optimization techniques to process the data range, an optimized scheduling unit is obtained.

[0050] Based on the optimized scheduling unit, the resource distribution characteristics of multi-satellite cooperation are determined, and the scheduling result is obtained;

[0051] Based on the scheduling results, a distributed computing framework is used to verify data consistency and determine the final scheduling data group.

[0052] Preferably, based on the scheduling results, reinforcement learning techniques are used to iteratively optimize the robustness of the algorithm. The process of obtaining a robustness-enhanced model by simulating fire detection tasks under complex weather conditions includes:

[0053] For the final scheduling data set, reinforcement learning techniques are used to process the feature distribution under complex meteorological conditions to obtain a feature-enhanced dataset;

[0054] By using the feature enhancement dataset, the boundary conditions of the fire detection task are obtained, and the boundary feature set is determined.

[0055] For the aforementioned boundary feature set, task simulation technology is used to generate multi-scenario detection data to obtain a simulated data set;

[0056] Based on the simulated data set, if the distribution characteristics of the detection task exceed a preset threshold, the algorithm robustness parameter is adjusted to obtain an adjustment parameter set;

[0057] By adjusting the parameter set, the iteratively optimized model structure is obtained, and the model framework is determined.

[0058] For the aforementioned model framework, distributed computing is used to verify the stability of fire detection, resulting in a robustness-enhanced model.

[0059] Preferably, the process of obtaining the final fire detection data from the robustness enhancement model, verifying the data fusion accuracy through multi-source data verification technology, and outputting the final monitoring result includes:

[0060] Fire detection data is extracted from the robustness enhancement model, and data features are fused using multi-source data verification technology to obtain a fused data set;

[0061] Based on the fused data set, distributed computing technology is used to process the detection task distribution to obtain task distribution data;

[0062] If the distribution characteristics of the task distribution data exceed a preset threshold, the verification technical parameters are adjusted to determine the adjusted parameter set.

[0063] The adjusted parameter set is used to obtain the accuracy-fused detection data and output the accuracy-enhanced data.

[0064] Based on the enhanced accuracy data, cluster analysis technology is used to divide the fire detection area, and the area division result is obtained;

[0065] Based on the area division results, monitoring result data is generated using spatial mapping technology to determine the final monitoring data;

[0066] Key features are extracted from the final monitoring data to obtain the final monitoring results, which include a set of fire point distribution features.

[0067] Compared with the prior art, the present invention has the following advantages and technical effects:

[0068] This invention constructs a complete process from data acquisition from a low-Earth orbit satellite platform to the final output of fire point monitoring results through multi-sensor data acquisition, adaptive filtering, and multi-source data fusion. Addressing the fire point identification problem in complex scenarios, this invention employs convolutional neural networks and multi-scale analysis techniques to extract and enhance fire point features, combined with real-time stream processing and intelligent analysis models to improve detection accuracy. Simultaneously, this invention utilizes multi-satellite collaborative observation data, employing methods such as time series alignment, deep learning, and transfer learning to optimize image analysis adaptability and enhance coverage. Furthermore, this invention uses reinforcement learning techniques to improve algorithm robustness and ensures fusion accuracy through multi-source data verification. This method significantly improves the accuracy, timeliness, and coverage integrity of fire point detection, providing an effective solution for fire monitoring under complex meteorological conditions. Attached Figure Description

[0069] The accompanying drawings, which form part of this application, are used to provide a further understanding of this application. The illustrative embodiments and descriptions of this application are used to explain this application and do not constitute an undue limitation of this application. In the drawings:

[0070] Figure 1 This is a schematic diagram of the method flow according to an embodiment of the present invention. Detailed Implementation

[0071] It should be noted that, unless otherwise specified, the embodiments and features described in this application can be combined with each other. This application will now be described in detail with reference to the accompanying drawings and embodiments.

[0072] It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions, and although a logical order is shown in the flowchart, in some cases the steps shown or described may be executed in a different order than that shown here.

[0073] like Figure 1 As shown, this embodiment provides a forest fire monitoring method based on low-Earth orbit satellites, including:

[0074] Multi-sensor data is acquired from a low-Earth orbit satellite platform. The heterogeneous sensor data is processed using a pre-set denoising model. Adaptive filtering technology is used to smooth the data in different bands to obtain a fused dataset.

[0075] For the fused dataset, if the noise interference exceeds the preset threshold during complex scene recognition, features are extracted through a convolutional neural network, and the fire point features are enhanced by multi-scale analysis technology to obtain a feature set.

[0076] By acquiring multi-satellite collaborative observation data and matching the multi-source collaborative coverage data with the feature set through time series alignment technology, an enhanced dataset with improved high-frequency observation capabilities is obtained.

[0077] For augmented datasets, if the demand for complex scene recognition increases, the image analysis adaptability is optimized through deep learning algorithms, and the model parameters are adjusted by combining transfer learning techniques to output optimized results.

[0078] Key regional features are extracted from the optimization results, and a distributed computing framework is used to perform data sharing processing on the multi-satellite collaborative complementarity to obtain a coverage enhancement dataset.

[0079] Based on the coverage enhancement dataset, the observation frequency and completeness are analyzed through a pre-set timeliness evaluation model. If the timeliness and completeness are lower than expected, the multi-satellite collaborative scheduling strategy is dynamically adjusted and the scheduling results are output.

[0080] Based on the scheduling results, reinforcement learning techniques are used to iteratively optimize the robustness of the algorithm. By simulating the fire detection task under complex weather conditions, a robustness enhancement model is obtained.

[0081] The final fire detection data is obtained from the robustness enhancement model, and the accuracy of data fusion is verified by multi-source data verification technology to output the final monitoring results.

[0082] Furthermore, the process of obtaining the fused dataset includes:

[0083] Sensor data generated by multiple sensors is acquired from low-Earth orbit satellites, and heterogeneous characteristics are processed through a preset denoising model. Multiple sensors are carried on a low-Earth orbit satellite platform to acquire multi-source data of forest areas.

[0084] The acquired sensor data is preprocessed to eliminate data heterogeneity between different sensors and obtain a fused dataset; the preprocessing includes data format unification, time synchronization and spatial registration.

[0085] Specifically, acquiring sensor data generated by multiple sensors from low-orbit satellites is an important technical means in the field of forest monitoring.

[0086] For example, low-Earth orbit satellites can carry optical sensors, infrared sensors, and radar sensors to collect visible light images, thermal imaging data, and topographic information of forests, respectively. Taking a 500-square-kilometer forest area as an example, optical sensors can capture canopy coverage, infrared sensors can detect fire hazards, and radar sensors can penetrate clouds to obtain topographic relief. This multi-source data acquisition method can comprehensively reflect the forest's ecological condition.

[0087] In one possible implementation, the pre-defined denoising model is particularly crucial for handling heterogeneous characteristics. Data from different sensors exhibit inconsistencies in resolution and sampling frequency.

[0088] For example, an optical sensor might cover 10 meters per pixel, an infrared sensor 50 meters, and radar data might have a resolution of 30 meters. The denoising model uses interpolation and filtering to standardize the resolution to 30 meters, while simultaneously removing cloud and fog interference or signal noise. This processing significantly improves data usability and lays the foundation for subsequent analysis.

[0089] It should be noted that preliminary data preprocessing includes data format standardization, time synchronization, and spatial registration.

[0090] Specifically, format unification converts optical imagery (TIFF), infrared imagery (DAT), and radar imagery (HDF) into a unified NetCDF format, facilitating integrated analysis. Time synchronization ensures that the data corresponds to the same moment in time.

[0091] For example, the 5-minute acquisition time difference caused by satellite orbit variations is adjusted. Spatial registration is performed using geographic coordinate system calibration, with pixel deviation controlled within 1 meter. This preprocessing eliminates heterogeneity and ensures spatial consistency of the data.

[0092] In one embodiment, the configuration can be optimized when the low-Earth orbit satellite platform carries multiple sensors.

[0093] For example, the satellite operates in an orbit at an altitude of 500 kilometers, carrying an optical camera with a resolution of 10 meters, an infrared spectrometer with a detection range of 50 kilometers, and a synthetic aperture radar with a coverage of 100 kilometers. When collecting data, the camera records tree density, the infrared spectrometer monitors temperature anomalies, and the radar delineates terrain contours. After fusing multi-source data, it can accurately identify forest fire risk areas.

[0094] Preferably, the first dataset after denoising provides a high-quality foundation for subsequent applications.

[0095] For example, in a 200-kilometer forest belt, the first dataset showed that the temperature in a certain area was 5 degrees Celsius above the average, and the canopy cover decreased by 20%. Combined with radar data analysis, it was found that low-lying terrain led to water accumulation, affecting vegetation growth. This multi-dimensional analysis not only improves monitoring accuracy but also provides a basis for decision-making in fire prevention and ecological protection.

[0096] Understandably, eliminating data heterogeneity brings significant technical benefits.

[0097] For example, traditional single sensors may miss fire hazards in rainy weather, while multi-sensor collaboration, after preprocessing, can supplement the deficiencies of optical images with infrared data, reducing the false alarm rate to below 5%. This method can effectively reduce resource waste and improve response speed in forest management.

[0098] In one embodiment, the preprocessed dataset can also support dynamic monitoring.

[0099] For example, data is collected every three days, and vegetation change trends are analyzed continuously over 30 days. If the canopy cover in a certain area continues to decline, the possibility of pests, diseases, or human logging can be inferred by combining temperature and topographic data. This long-term monitoring capability provides continuous support for ecological protection.

[0100] For example, the technologies described above allow forest management departments to quickly locate problem areas, reduce the cost of manual patrols, and improve the timeliness of disaster warnings. The collaboration and noise reduction of multi-source data together ensure the reliability and usability of the data.

[0101] Furthermore, the process of obtaining the feature set includes:

[0102] Feature extraction of the fused dataset is performed using a convolutional neural network, and feature maps of different scales are fused using spatial pyramid transformation or feature pyramid network structure.

[0103] Meanwhile, during the feature extraction process, an attention mechanism is introduced to enhance the basic features and assign weights to different features to highlight the fire point features and suppress noise interference.

[0104] The enhanced features are fused with multi-scale features to obtain a feature set.

[0105] In one possible implementation, the convolutional neural network (CNN) employs spatial pyramid transformation to capture multi-scale information when extracting features from the fused dataset. Spatial pyramid transformation generates feature representations at multiple scales by pooling the feature maps at different levels.

[0106] For example, in a forest monitoring scenario, the first fused dataset contains information on canopy cover, temperature distribution, and topographic relief. Spatial pyramid transformation can extract coarse-grained features at the global scale (such as the average temperature of the entire forest) and fine-grained features at the local scale (such as canopy details in a small area). This method allows the model to focus on both large-scale trends and small-scale outliers simultaneously; for example, features of areas with concentrated temperature anomalies in a forest will be amplified.

[0107] Specifically, the Feature Pyramid Network (FPN) structure fuses feature maps of different depths through top-down and lateral connections.

[0108] For example, when processing 500 square kilometers of forest data, shallow feature maps may retain details at a resolution of 10 meters, such as the shape of tree edges; deep feature maps, on the other hand, extract more abstract information, such as the trend of terrain undulation. FPN merges these feature maps to form a feature set that combines both detail and semantics.

[0109] Understandably, this fusion method is particularly suitable for forest fire monitoring because it can simultaneously preserve the local high-temperature characteristics of the fire point and the overall distribution of the surrounding environment.

[0110] It should be noted that the attention mechanism plays a crucial role in feature enhancement. The global attention module highlights important information by calculating the weight of each location in the feature map.

[0111] For example, in infrared data, the temperature characteristics of fire points are given higher weight, while noise from clouds and fog is suppressed.

[0112] In one embodiment, assuming a temperature anomaly of 38 degrees Celsius in a certain area, significantly higher than the average of 32 degrees Celsius, the global attention module enhances the feature representation of this area while reducing the weight of irrelevant areas. This enhanced feature can more accurately reflect potential fire risks.

[0113] In one possible implementation, when fusing multi-scale features, the enhanced features can be weighted and superimposed first, and then the number of channels can be adjusted through a convolutional layer.

[0114] For example, the feature map generated by the spatial pyramid transformation may have three scales: 10 meters, 30 meters, and 50 meters resolution. After attention enhancement, they are merged into a unified 30-meter resolution feature set.

[0115] Preferably, this method ensures that the characteristics of different sensor data are preserved and work together, such as the canopy details in optical data complementing the temperature anomalies in infrared data.

[0116] For example, in a 200-kilometer-long forest belt, the feature set shows that a certain low-lying area exhibits both a rising temperature trend and sparse tree canopy characteristics. Combined with topographic data analysis, this area may have water accumulation affecting vegetation growth, thereby increasing the risk of fire.

[0117] Understandably, this multi-dimensional feature fusion can provide a more reliable foundation for subsequent classification or prediction tasks, such as quickly locating areas that need to be monitored.

[0118] In one embodiment, the attention mechanism can also dynamically adjust the weights.

[0119] For example, if a region experiences consistently higher temperatures and a decrease in canopy cover over 10 consecutive days of data, the model will gradually increase its focus on characteristics of that region. This dynamic approach helps capture gradual trends in forest ecology, providing support for long-term monitoring.

[0120] For example, this method can also reduce false alarms because it integrates data from multiple days rather than a single point in time.

[0121] Specifically, the feature set generation process emphasizes the efficient integration of multi-source information. In forest monitoring, a single feature may be distorted due to weather or sensor limitations, but after multi-scale fusion and attention enhancement, the feature set can more comprehensively reflect the actual situation.

[0122] For example, the ability of radar data to penetrate clouds, combined with the temperature sensitivity of infrared data, makes fire point identification more reliable in rainy weather. This technology significantly improves the robustness of monitoring, providing a solid guarantee for ecological management and disaster early warning.

[0123] Furthermore, the process of obtaining augmented datasets includes:

[0124] Multi-source data was acquired through multi-satellite observations, and time-series alignment technology was used to process the data alignment, resulting in the first aligned dataset.

[0125] Cooperative coverage information is extracted from the aligned first dataset and fused with the feature set features through matching to obtain the fused second dataset;

[0126] For the fused second dataset, coverage analysis is used to determine the spatial distribution characteristics of the multi-source data and obtain the distribution characteristic set;

[0127] If the distribution characteristic set meets the preset threshold condition, the distribution characteristic set is classified by the random forest algorithm to determine the high-frequency observation area and obtain the region set.

[0128] Based on the correspondence between the regional set and the time series, the temporal distribution characteristics of high-frequency observations are obtained, resulting in a time feature set;

[0129] By mapping the temporal feature set to the augmented dataset, the enhanced regions of observation capability are determined, and the enhanced region set is obtained.

[0130] Data augmentation techniques are used to optimize the augmented region set, resulting in an optimized augmented dataset.

[0131] Specifically, acquiring multi-source data through multi-satellite observations is fundamental to the analysis of complex scenarios. For example, multiple remote sensing satellites can be used to observe the same area at different times, acquiring various types of data, such as optical and infrared data. Due to differences in satellite orbits and observation times, the data may have temporal deviations. When using time-series alignment techniques to align the data, interpolation adjustments can be made based on timestamps.

[0132] For example, suppose a region is observed by two satellites at 8:00 and 8:05 on March 15th. The two sets of data are aligned to the same time reference, such as 8:03, using linear interpolation, resulting in the first aligned dataset. This alignment method ensures temporal consistency in subsequent analyses. Extracting cooperative coverage information from the first aligned dataset involves spatial matching of multi-source data.

[0133] Specifically, the observation ranges of different satellites can be overlaid and analyzed using a geographic coordinate system.

[0134] For example, satellite A covers the area from 120° to 121° east longitude, and satellite B covers the area from 120.5° to 121.5° east longitude. The cooperative coverage information is the data of the overlapping part from 120.5° to 121° east longitude.

[0135] After matching, when fusing with the features of the feature set, a weighted averaging method is used to combine the collaborative coverage information with the existing feature set to generate a fused second dataset. This fusion improves the spatial integrity of the data.

[0136] For the fused second dataset, when determining spatial distribution characteristics using coverage analysis, a grid partitioning method can be employed. The observation area is divided into 1 km × 1 km grids, and the data density within each grid is calculated to obtain a distribution characteristic set. For example, if a grid contains 50 data points, exceeding a preset threshold of 30, it is marked as a high-density area. If the distribution characteristic set meets the conditions, it is then classified using the random forest algorithm.

[0137] In one embodiment, the random forest divides the region into high-frequency observation areas and low-frequency observation areas based on features such as data density and coverage frequency, resulting in a region set. This classification can quickly identify key areas. When obtaining temporal distribution characteristics based on the correspondence between the region set and the time series...

[0138] It should be noted that the number of observations in a high-frequency observation area within 24 hours can be counted.

[0139] For example, a region is observed 10 times between 6:00 and 18:00, and the temporal feature set indicates that its active period is during the daytime. Mapping this temporal feature set to an augmented dataset identifies regions with enhanced observation capabilities. In one possible implementation, if the augmented dataset records device performance parameters, the augmented region set might be areas with a 50% increase in observation frequency. When optimizing using data augmentation techniques, such as increasing data point density through interpolation, an optimized augmented dataset is generated. This optimization improves data usability.

[0140] In one embodiment, the collaborative coverage information extraction from multi-satellite observations can also be dynamically adjusted by incorporating historical data. For example, based on observation records from the past 7 days, features of high-frequency coverage areas are prioritized to improve fusion efficiency.

[0141] Understandably, the combination of time series alignment and feature fusion provides a more reliable data foundation for subsequent analysis. Random forest classification, on the other hand, ensures the accuracy of region segmentation through comprehensive judgment of multi-dimensional features. The synergistic effect of these methods ultimately provides strong support for data processing in complex scenarios.

[0142] Furthermore, the process of optimizing image analysis adaptability through deep learning algorithms, adjusting model parameters by combining transfer learning techniques, and outputting the optimized results includes:

[0143] Preliminary analysis results are obtained by performing initial processing on the augmented dataset using deep learning algorithms.

[0144] Based on the preliminary analysis results, transfer learning techniques were used to adjust the model parameters for complex scenarios, resulting in an optimized parameter set.

[0145] If the optimized parameter set meets the preset threshold conditions, then feature extraction is performed on the complex scene using image analysis technology to obtain the feature set;

[0146] The augmented dataset is matched using a feature set to obtain a matched dataset;

[0147] The distribution characteristics of scene recognition are determined by matching the dataset, and a distribution feature set is obtained;

[0148] Based on the mapping relationship between the distribution feature set and the optimization parameter set, obtain the mapping dataset;

[0149] The mapped dataset was optimized using data augmentation techniques to obtain the optimized result.

[0150] Specifically, when initially processing augmented datasets using deep learning algorithms, for example, in the field of satellite collaborative observation, the augmented dataset can be viewed as remote sensing data acquired from multiple satellites, containing both temporal and spatial information. One possible implementation involves using a convolutional neural network to extract features from the data. The input data might be 1000 images with a resolution of 256×256, and the output would be preliminary analysis results such as the probability distribution of land cover types. This approach can quickly process high-dimensional data, laying the foundation for subsequent analysis.

[0151] When using transfer learning techniques to adjust model parameters for complex scenarios, such as complex terrain with low forest cover, a model pre-trained on a large-scale general dataset, such as a ResNet-based network, can be selected, and then the parameters can be fine-tuned using 200 samples from an augmented dataset. The optimized parameter set may be a set of adjusted weights that meets threshold conditions, such as achieving an accuracy of 85%, thereby adapting to the specific scenario.

[0152] When extracting features from complex scenes using image analysis techniques, specifically, edge detection algorithms can be used to extract feature boundaries. In one embodiment, a feature set is extracted from the image of the complex scene, such as values ​​like a river width of 50 meters and vegetation density of 0.7, to generate a feature vector. This method highlights the key attributes of the scene, facilitating subsequent matching.

[0153] When matching an augmented dataset with a feature set, similarity calculations are used to map the feature set to samples in the augmented dataset. For example, feature vectors are compared one by one with 500 samples in the dataset, and matching datasets with a similarity higher than 0.9 are selected. This matching effectively integrates multi-source information. When determining the distribution characteristics of scene recognition using the matching dataset, in one embodiment, if the matching dataset covers an area of ​​1000 square kilometers, the distribution feature set may show 70% plains and 20% mountains. This distribution characteristic reflects the spatial heterogeneity of the data. When obtaining the mapping dataset based on the mapping relationship between the distribution feature set and the optimization parameter set, the data weights are adjusted through the parameter set to generate a mapping dataset that better matches the distribution characteristics. Assuming that the weight of plain areas is increased by 10% and that of mountains is decreased by 5%, the new dataset can more accurately reflect the characteristics of the target area.

[0154] When optimizing a mapped dataset using data augmentation techniques, 300 images in the dataset can be rotated, flipped, or have noise added to generate an augmented dataset. This approach enriches data diversity and improves the reliability of subsequent applications.

[0155] Therefore, in this embodiment, transfer learning to adjust parameters reduces training costs, image analysis to extract features enhances detail representation, and data augmentation optimizes data quality. These technologies support each other, jointly improving the efficiency and accuracy of multi-satellite collaborative observation data processing. The optimized augmented dataset can be used for disaster monitoring, quickly locating high-risk areas, demonstrating significant practical value.

[0156] Furthermore, the process of extracting key regional features from the optimization results and using a distributed computing framework to perform data sharing processing on multi-satellite collaborative complementarity to obtain the coverage enhancement dataset includes:

[0157] The distribution range of regional features is determined by identifying key regions, and a preliminary feature set is obtained by using feature extraction techniques.

[0158] Based on the preliminary feature set, multi-star collaborative task units are divided to obtain a collaborative and complementary allocation scheme;

[0159] A distributed computing framework is used to process the data sharing of the allocation scheme, resulting in a shared data group.

[0160] If the shared data group meets the preset threshold conditions, the data range is adjusted through coverage enhancement technology to obtain the enhanced data group;

[0161] The processing flow of the enhanced data group is optimized to obtain optimized data units;

[0162] By optimizing the mapping relationship between data units and regional features, the complementary distribution of multi-star collaboration is determined, and a coverage enhancement dataset is obtained.

[0163] A computational framework is used to validate the coverage-enhanced dataset and determine the final distribution characteristics.

[0164] Specifically, multi-satellite collaboration may involve multiple satellites being responsible for feature extraction tasks in different regions or of different types.

[0165] For example, one satellite focuses on high-resolution image acquisition, while another satellite is responsible for a wide range of spectral analysis. The two complement each other through the division of mission units, thereby improving the comprehensiveness of data acquisition.

[0166] A distributed computing framework is used to process the data sharing of the allocation scheme, resulting in a shared data group. This distributed computing framework can be deployed in the cloud, allowing satellite-collected data to be uploaded and shared in real time via the network.

[0167] For example, each satellite fragments its collected 5GB data and uploads it to the computing node. After parallel processing, a shared data set containing boundary information is generated. This method effectively improves data processing efficiency. If the shared data set meets a preset threshold condition, the data range is adjusted using coverage enhancement technology to obtain an enhanced data set.

[0168] It should be noted that the threshold condition may be set to a data coverage rate of 90% or higher.

[0169] In one embodiment, if the coverage of the shared data set is only 85%, the missing parts can be filled in using interpolation or data fusion techniques to ultimately form an enhanced data set with 95% coverage. This ensures the completeness of subsequent analysis.

[0170] The processing flow is optimized for the augmented data set to obtain optimized data units. This optimization process may include data cleaning and denoising. For example, for noisy boundary data in the augmented data set, outliers are removed using smoothing filtering techniques, resulting in clearer optimized data units. This optimization helps improve the accuracy of subsequent analyses. By optimizing the mapping relationship between data units and regional features, the complementary distribution of multi-satellite collaboration is determined, resulting in a covered augmented dataset.

[0171] Furthermore, based on the coverage enhancement dataset, the observation frequency and completeness are analyzed using a pre-defined timeliness assessment model. If the timeliness and completeness are lower than expected, the multi-satellite collaborative scheduling strategy is dynamically adjusted, and the process of outputting the scheduling results includes:

[0172] By covering the enhanced dataset, the frequency and completeness of observations are analyzed using a pre-defined timeliness assessment model to obtain the analysis data set;

[0173] Based on the analyzed data set, determine whether the integrity is below a preset threshold condition, and obtain the judgment data unit;

[0174] By determining the data unit, a dynamic adjustment technique is used to process the collaborative scheduling scheme, resulting in an adjusted data group;

[0175] To adjust the data group, obtain task allocation information for multi-satellite collaboration, and determine the task allocation set;

[0176] By using a task allocation set and employing scheduling strategy optimization techniques to process the data range, an optimized scheduling unit is obtained.

[0177] Based on the optimized scheduling unit, the resource distribution characteristics of multi-satellite cooperation are determined, and the scheduling result is obtained;

[0178] Based on the scheduling results, a distributed computing framework is used to verify data consistency and determine the final scheduling data group.

[0179] Specifically, by enhancing the coverage dataset, a pre-defined timeliness assessment model is used to analyze observation frequency and completeness, resulting in an analysis data set. This process can be understood as quantifying the time sensitivity of the data. For example, in multi-satellite collaborative operations, assuming a certain area requires 10 observations per day but only 8 are actually completed, the timeliness assessment model will give a score based on observation frequency and data completeness, such as 80 points, which is below the pre-defined threshold of 90 points. The timeliness assessment model will then combine historical data and real-time feedback to analyze whether the missing observations affect the overall coverage.

[0180] Based on the analyzed data set, a judgment data unit is obtained to determine whether the integrity score is below a preset threshold. Specifically, if the integrity score of a certain area is only 75 points, which is below the threshold, the judgment data unit will mark that area as needing adjustment. For example, this judgment can be based on the data fluctuations over three consecutive days to determine whether it is an accidental missing data or a systemic problem, thereby providing a basis for subsequent scheduling.

[0181] By identifying data units and employing dynamic adjustment techniques to process the collaborative scheduling scheme, an adjusted data set is obtained. In one possible implementation, if a satellite cannot cover the target area due to orbital limitations, the mission time of another satellite can be dynamically adjusted. For example, if satellite A was originally scheduled to perform observations at 9:00, it can now be adjusted to satellite B covering the area at 9:30, and the adjusted data set records this change. This approach can effectively improve resource utilization.

[0182] To adjust the data set, task allocation information for multi-satellite collaboration is obtained to determine the task allocation set. For example, for a certain area, the task allocation set might include satellite A responsible for morning observations, satellite B responsible for afternoon observations, and satellite C providing backup support. Specifically, this allocation is generated based on the performance parameters and geographical location of each satellite to ensure efficient task allocation.

[0183] By using a task allocation set and employing scheduling strategy optimization techniques to process the data range, an optimized scheduling unit is obtained. In one embodiment, if the coverage area of ​​a certain region is too large, the task is split into multiple sub-regions using optimization techniques. For example, if the original coverage area is 1000 square kilometers, it is optimized into 3 sub-regions, each approximately 330 square kilometers, each executed by a different satellite. This splitting significantly improves observation accuracy. Based on the optimized scheduling unit, the resource distribution characteristics of multi-satellite cooperation are determined, and the scheduling result is obtained.

[0184] It should be noted that the resource distribution characteristics may manifest as an excessively high load on a certain satellite.

[0185] For example, satellite A undertakes 60% of the tasks, while satellite B only undertakes 20%. The sixth scheduling result will reflect this distribution and provide data support for subsequent equalization. For the sixth scheduling result, a distributed computing framework is used to verify data consistency and determine the final scheduling data set.

[0186] For example, the distributed framework distributes the scheduling results to multiple computing nodes for verification. Each node checks whether timestamps are synchronized and whether coverage overlaps, ultimately confirming that the data groups are conflict-free. This verification ensures the reliability of the scheduling scheme during actual execution.

[0187] Furthermore, based on the scheduling results, reinforcement learning techniques are used to iteratively optimize the robustness of the algorithm. The process of obtaining the robustness-enhanced model by simulating fire detection tasks under complex weather conditions includes:

[0188] For the final scheduling data set, reinforcement learning techniques are used to process the feature distribution under complex meteorological conditions to obtain a feature-enhanced dataset;

[0189] By using the feature enhancement dataset, the boundary conditions of the fire detection task are obtained, and the boundary feature set is determined.

[0190] For the boundary feature set, task simulation technology is used to generate multi-scene detection data to obtain a simulated data set;

[0191] Based on the simulated data set, if the distribution characteristics of the detection task exceed the preset threshold, the algorithm robustness parameter is adjusted to obtain the set of adjustment parameters;

[0192] By adjusting the parameter set, the iteratively optimized model structure is obtained, and the model framework is determined.

[0193] For the model framework, distributed computing is used to verify the stability of fire detection, resulting in a robustness-enhanced model.

[0194] Specifically, when using reinforcement learning to process the feature distribution under complex weather conditions for the final scheduling data set, reinforcement learning gradually optimizes the feature extraction strategy through continuous trial and error and a reward mechanism. For example, under cloudy or strong wind conditions, the original data may be distorted due to occlusion or noise. In one embodiment, the reward function can be set to focus on high-confidence features, with an initial learning rate of 0.01. After training 100 times, the clarity of the feature distribution is significantly improved, resulting in a feature-enhanced dataset. This provides a more reliable foundation for subsequent tasks. The feature-enhanced dataset is then used to obtain the boundary conditions for the fire detection task.

[0195] Specifically, by using a feature-enhanced dataset, the boundary conditions for the fire detection task are obtained, and the boundary feature set is determined. For example, in a 500-square-kilometer forest area, the boundary conditions can be set as temperature anomalies exceeding 50 degrees Celsius for more than 10 minutes to determine the boundary feature set. This approach ensures that the detection task focuses on key areas, improving efficiency.

[0196] When generating multi-scene detection data using task simulation techniques for boundary feature sets, one possible implementation can simulate three scenarios: sunny, cloudy, and nighttime. Preferably, the fire point brightness value is set to 80 in the sunny scenario, reduced to 50 due to cloud interference in the cloudy scenario, and increased to 90 at night due to enhanced background contrast, thus obtaining a set of simulated data. This multi-scene design helps to comprehensively evaluate detection capabilities.

[0197] When determining whether the distribution characteristics of the detection task exceed a preset threshold based on the simulated data set and adjusting the algorithm's robustness parameters, for example, if the preset threshold is a detection rate of 90%, and the detection rate is only 85% in a multi-cloud scenario, the parameters can be increased from the default value of 0.5 to 0.7 by increasing the convolutional layer depth or adjusting the loss weights, thus obtaining an adjusted parameter set. This adjustment can effectively handle abnormal distributions under complex conditions. The adjusted parameter set is then used to obtain the iteratively optimized model structure.

[0198] It should be noted that iterative optimization can gradually approach optimal performance through multiple verifications. In one embodiment, after 5 iterations, the model's false alarm rate for fire points decreased from 10% to 3%, thus determining the seventh model framework. This optimization process significantly improves the model's practicality. Distributed computing was then used to verify the stability of fire point detection for the seventh model framework.

[0199] Understandably, distributed computing distributes tasks across multiple nodes for parallel processing. For example, running 1000 fire detection tasks on 10 computing nodes, with 98% consistency across nodes, yields a robust, enhanced model. This approach ensures the model's reliability in large-scale applications while improving processing speed. Throughout the process, from feature enhancement to model optimization, every step is closely aligned with the fire detection task.

[0200] For example, reinforcement learning improves feature quality, boundary conditions focus the task scope, multi-scenario simulation enriches data diversity, and parameter tuning and distributed verification ensure stability and robustness. These elements support each other, forming a complete solution that provides strong technical support for fire detection in complex environments.

[0201] Furthermore, the process of obtaining the final fire detection data from the robustness enhancement model, verifying the data fusion accuracy through multi-source data verification technology, and outputting the final monitoring results includes:

[0202] Fire detection data is extracted from the robustness enhancement model, and data features are fused using multi-source data verification technology to obtain a fused data set;

[0203] Based on the fused data set, distributed computing technology is used to process the distribution of detection tasks to obtain task distribution data.

[0204] If the distribution characteristics of the task distribution data exceed the preset threshold, the verification technical parameters are adjusted to determine the adjusted parameter set.

[0205] By using the adjusted parameter set, the detection data after accuracy fusion is obtained, and the accuracy-enhanced data is output.

[0206] Based on the enhanced accuracy data, cluster analysis technology was used to divide the fire detection area, and the area division results were obtained.

[0207] Based on the regional division results, spatial mapping technology is used to generate monitoring result data to determine the final monitoring data.

[0208] Key features are extracted from the final monitoring data to obtain the final monitoring results, which include a set of fire point distribution features.

[0209] Specifically, when extracting fire detection data from robustness enhancement models, for example, under complex weather conditions, fire detection data may contain multi-dimensional features such as temperature, humidity, and wind speed.

[0210] In one embodiment, it is assumed that the fire point data of a certain area includes a temperature of 35 degrees Celsius, humidity of 20%, and wind speed of 5 meters per second. After being extracted by the model, an initial dataset is formed, which provides a basis for subsequent fusion.

[0211] By fusing data features through multi-source data verification techniques, multi-source data can be obtained from satellite remote sensing, ground sensors, and weather stations. For example, satellite data might show the coordinates of a fire point, ground sensors might detect an abnormal temperature, and weather stations might provide wind direction data. During fusion, the timestamps and spatial locations of these data are first aligned, and then features are integrated through weighted averaging or a voting mechanism to obtain the fused data set. This approach effectively reduces the bias from a single data source.

[0212] When using distributed computing technology to process the distribution of detection tasks in a fused data set, one possible implementation involves sharding the data across multiple computing nodes. For example, a forest area might be divided into 10 sub-regions, with each node processing 1000 data records to statistically analyze the frequency and intensity of fire points, ultimately aggregating the data to form the task distribution data. This distributed processing significantly improves efficiency. If the distribution characteristics of the task distribution data exceed a preset threshold, the verification parameters are adjusted.

[0213] Preferably, assuming the threshold is set to a fire density of 0.5 fire points per square kilometer, if a certain area reaches 0.7 fire points per square kilometer, an adjusted parameter set can be generated by increasing the verification frequency or adjusting the feature weights. This ensures the reliability of the data.

[0214] When obtaining the fused detection data using the adjusted parameter set, outliers in the fused data set are filtered out. In one embodiment, if a fire point temperature suddenly jumps to 100 degrees Celsius, significantly exceeding the normal range, it is discarded, the average value is recalculated, and the enhanced accuracy data is output. This method helps improve data accuracy.

[0215] When using cluster analysis to divide fire detection areas based on enhanced accuracy data, a density-based clustering method is employed. For example, assuming an area has 50 fire data points, it can be divided into three densely populated fire areas by setting a radius of 1 kilometer and a minimum number of points of 5. This division clearly reflects the distribution pattern of fire points.

[0216] Monitoring data is generated based on the regional division results using spatial mapping technology. Specifically, the divided regions are mapped onto a two-dimensional grid. For example, if a region is divided into three sub-regions, A, B, and C, with 20 fire points in region A, 15 in region B, and 10 in region C, the gridding process generates intuitive monitoring data. This mapping facilitates subsequent analysis and visualization.

[0217] When extracting key features from the final monitoring data, attention should be paid to indicators such as fire point density, spread rate, and duration. For example, a certain fire point distribution feature set shows a density of 0.6 fire points per square kilometer, a spread rate of 2 kilometers per hour, and a duration of 3 hours.

[0218] In this embodiment, multi-source data fusion improves feature integrity, distributed computing ensures large-scale data processing capabilities, and clustering analysis and spatial mapping optimize regional management efficiency. These components support each other, forming a complete chain for analyzing fire point distribution characteristics.

[0219] The above are merely preferred embodiments of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A forest fire monitoring method based on low-Earth orbit satellites, characterized in that, include: Multi-sensor data is acquired from a low-Earth orbit satellite platform. The heterogeneous sensor data is processed using a pre-set denoising model. Adaptive filtering technology is used to smooth the data in different bands to obtain a fused dataset. For the fused dataset, if the noise interference exceeds the preset threshold during complex scene recognition, features are extracted through a convolutional neural network, and the fire point features are enhanced by multi-scale analysis technology to obtain a feature set. Acquire multi-satellite collaborative observation data, and match the multi-source collaborative coverage data with the feature set using time series alignment technology to obtain an enhanced dataset with improved high-frequency observation capabilities; For augmented datasets, if the demand for complex scene recognition increases, the image analysis adaptability is optimized through deep learning algorithms, and the model parameters are adjusted by combining transfer learning techniques to output optimized results. Key regional features are extracted from the optimization results, and a distributed computing framework is used to perform data sharing processing on the multi-satellite collaborative complementarity to obtain a coverage enhancement dataset. Based on the coverage enhancement dataset, the observation frequency and completeness are analyzed using a preset timeliness evaluation model. If the timeliness and completeness are lower than expected, the multi-satellite collaborative scheduling strategy is dynamically adjusted, and the scheduling results are output. Based on the scheduling results, reinforcement learning techniques are used to iteratively optimize the robustness of the algorithm. By simulating the fire detection task under complex weather conditions, a robustness enhancement model is obtained. The final fire detection data is obtained from the robustness enhancement model, and the accuracy of data fusion is verified by multi-source data verification technology to output the final monitoring results. The process of obtaining augmented datasets includes: Multi-source data was acquired through multi-satellite observations, and time-series alignment technology was used to process the data alignment, resulting in the first aligned dataset. Cooperative coverage information is extracted from the aligned first dataset and fused with the feature set features through matching to obtain the fused second dataset; For the fused second dataset, coverage analysis is used to determine the spatial distribution characteristics of the multi-source data and obtain the distribution characteristic set; If the distribution characteristic set meets the preset threshold condition, the distribution characteristic set is classified by the random forest algorithm to determine the high-frequency observation area and obtain the region set. Based on the correspondence between the region set and the time series, the temporal distribution characteristics of high-frequency observations are obtained, resulting in a time feature set; By mapping the time feature set to the enhanced dataset, the enhanced regions of observation capability are determined, and the enhanced region set is obtained. The augmented region set is optimized using data augmentation techniques to obtain an optimized augmented dataset.

2. The method according to claim 1, characterized in that, The process of obtaining the fused dataset includes: Sensor data generated by multiple sensors is acquired from low-Earth orbit satellites, and heterogeneous characteristics are processed through a preset denoising model. Multiple sensors are carried on a low-Earth orbit satellite platform to acquire multi-source data of forest areas. The acquired sensor data is preprocessed to eliminate data heterogeneity between different sensors, resulting in the fused dataset; wherein, the preprocessing includes data format unification, time synchronization, and spatial registration.

3. The method according to claim 1, characterized in that, The process of obtaining the feature set includes: Feature extraction is performed on the fused dataset using a convolutional neural network, and feature maps of different scales are fused using a spatial pyramid transformation or feature pyramid network structure. Meanwhile, during the feature extraction process, an attention mechanism is introduced to enhance the basic features and assign weights to different features to highlight the fire point features and suppress noise interference. The enhanced features are fused with multi-scale features to obtain the feature set.

4. The method according to claim 1, characterized in that, The process of optimizing image analysis adaptability using deep learning algorithms, adjusting model parameters using transfer learning techniques, and outputting the optimized results includes: Preliminary analysis results are obtained by performing initial processing on the augmented dataset using deep learning algorithms. Based on the preliminary analysis results, transfer learning techniques were used to adjust the model parameters for complex scenarios, resulting in an optimized parameter set. If the optimized parameter set meets the preset threshold condition, then feature extraction is performed on the complex scene using image analysis technology to obtain the feature set; The augmented dataset is matched using a feature set to obtain a matched dataset; The distribution characteristics of scene recognition are determined by the matching dataset, and a distribution feature set is obtained; Based on the mapping relationship between the distribution feature set and the optimization parameter set, obtain the mapping dataset; The mapped dataset is optimized using data augmentation techniques to obtain the optimized result.

5. The method according to claim 1, characterized in that, The process of extracting key regional features from the optimization results and using a distributed computing framework to perform data sharing processing on multi-satellite collaborative complementarity to obtain a coverage enhancement dataset includes: The distribution range of regional features is determined by identifying key regions, and a preliminary feature set is obtained by using feature extraction techniques. Based on the preliminary feature set, multi-star collaborative task units are divided to obtain a collaborative and complementary allocation scheme; A distributed computing framework is used to perform data sharing processing on the allocation scheme to obtain a shared data group; If the shared data group meets the preset threshold conditions, the data range is adjusted through coverage enhancement technology to obtain the enhanced data group; The enhanced data group is processed to optimize the workflow, resulting in optimized data units; By analyzing the mapping relationship between the optimized data units and regional features, the complementary distribution of multi-star collaboration is determined, and a coverage enhancement dataset is obtained. A computational framework is used to validate the coverage-enhanced dataset and determine the final distribution characteristics.

6. The method according to claim 1, characterized in that, Based on the coverage enhancement dataset, the observation frequency and completeness are analyzed using a preset timeliness evaluation model. If the timeliness and completeness are lower than expected, the multi-satellite collaborative scheduling strategy is dynamically adjusted, and the process of outputting the scheduling results includes: By covering the enhanced dataset, the frequency and completeness of observations are analyzed using a pre-defined timeliness assessment model to obtain the analysis data set; Based on the analyzed data set, determine whether the integrity is below a preset threshold condition, and obtain the judgment data unit; The collaborative scheduling scheme is processed using dynamic adjustment technology through the aforementioned judgment data unit to obtain the adjustment data group; For the adjusted data group, obtain multi-star collaborative task allocation information and determine the task allocation set; By using the task allocation set and employing scheduling strategy optimization techniques to process the data range, an optimized scheduling unit is obtained. Based on the optimized scheduling unit, the resource distribution characteristics of multi-satellite cooperation are determined, and the scheduling result is obtained; Based on the scheduling results, a distributed computing framework is used to verify data consistency and determine the final scheduling data group.

7. The method according to claim 1, characterized in that, Based on the scheduling results, reinforcement learning techniques are used to iteratively optimize the robustness of the algorithm. The process of obtaining the robustness enhancement model by simulating fire detection tasks under complex weather conditions includes: For the final scheduling data set, reinforcement learning techniques are used to process the feature distribution under complex meteorological conditions to obtain a feature-enhanced dataset; By using the feature enhancement dataset, the boundary conditions of the fire detection task are obtained, and the boundary feature set is determined. For the aforementioned boundary feature set, task simulation technology is used to generate multi-scenario detection data to obtain a simulated data set; Based on the simulated data set, if the distribution characteristics of the detection task exceed a preset threshold, the algorithm robustness parameter is adjusted to obtain an adjustment parameter set; By adjusting the parameter set, the iteratively optimized model structure is obtained, and the model framework is determined. For the aforementioned model framework, distributed computing is used to verify the stability of fire detection, resulting in a robustness-enhanced model.

8. The method according to claim 1, characterized in that, The process of obtaining the final fire detection data from the robustness enhancement model, verifying the data fusion accuracy through multi-source data verification technology, and outputting the final monitoring results includes: Fire detection data is extracted from the robustness enhancement model, and data features are fused using multi-source data verification technology to obtain a fused data set; Based on the fused data set, distributed computing technology is used to process the detection task distribution to obtain task distribution data; If the distribution characteristics of the task distribution data exceed a preset threshold, the verification technical parameters are adjusted to determine the adjusted parameter set. The adjusted parameter set is used to obtain the accuracy-fused detection data and output the accuracy-enhanced data. Based on the enhanced accuracy data, cluster analysis technology is used to divide the fire detection area, and the area division result is obtained; Based on the area division results, monitoring result data is generated using spatial mapping technology to determine the final monitoring data; Key features are extracted from the final monitoring data to obtain the final monitoring results, which include a set of fire point distribution features.

Images