Space-air-ground integrated remote sensing information collaborative service method and system

Abstract

This invention relates to the field of remote sensing information service technology, and provides a method and system for integrated space-air-ground remote sensing information collaborative services. The method includes: acquiring multi-source observation requirements; acquiring the real-time operating status and physical attributes of satellite nodes, UAV nodes, and ground observation nodes in a multi-dimensional space-air-ground resource pool; selecting executable nodes from the resource pool and assigning observation tasks to each executable node to generate an optimal set of collaborative observation tasks; collecting multi-source remote sensing data and fusing the collected multi-source remote sensing data to generate and distribute collaborative remote sensing information service products. This invention extracts geographic location named entities and time constraint phrases from unstructured text, and then maps them to a set of latitude and longitude coordinates, an effective time window, and a minimum spatial resolution threshold, thereby automatically constructing a collaborative observation task profile and improving the efficiency of task parsing.

Description

Technical Field

[0001] This invention belongs to the field of remote sensing information service technology, and in particular relates to a method and system for integrated air-space-ground remote sensing information collaborative services. Background Technology

[0002] With the rapid development of Earth observation technology, the demand for remote sensing monitoring in complex scenarios is increasingly characterized by multi-scale, multi-angle, and multi-timeframe features. Single satellites, UAVs, or ground observation platforms, limited by orbital periods, endurance, and sensor performance, struggle to independently meet the requirements of high spatiotemporal resolution collaborative observation. Integrated space-air-ground observation systems, through the coordinated scheduling and data fusion of satellites, UAVs, and ground nodes, have become an important technological direction for enhancing remote sensing information service capabilities.

[0003] However, existing space-air-ground remote sensing collaborative observation systems face several challenges in practical applications. The analysis of observation requirements and the construction of task profiles heavily rely on manual intervention. Multi-source observation requirements are often submitted in unstructured text form, making manual extraction of geographic location named entities, time constraint phrases, and resolution requirements inefficient and prone to errors. The lack of a unified quantitative modeling mechanism for the multi-dimensional physical parameters required for cross-platform collaborative scheduling leads to a lack of scientific basis for node selection and task allocation. Furthermore, in the node scheduling stage, existing methods typically allocate resources based on a single indicator, failing to establish a multi-objective trade-off model between spatial resolution, data transmission latency, and platform energy consumption. When multiple nodes can perform the same observation task, it is impossible to select the optimal set of collaborative observation tasks under physical constraints such as the offset angle of the sensor's maximum effective field of view.

[0004] The aforementioned problems severely restrict the application effectiveness of the space-air-ground collaborative observation system in scenarios with stringent timeliness requirements, such as emergency response and environmental monitoring. Therefore, how to provide an intelligent collaborative service solution that does not rely too much on manual labor to optimize timeliness indicators is the technical problem that this invention aims to solve. Summary of the Invention

[0005] The purpose of this invention is to provide a method and system for integrated air-space-ground remote sensing information collaborative services, aiming to solve the above-mentioned problems.

[0006] This invention is implemented as follows: a method for collaborative service of integrated air-space-ground remote sensing information, the method comprising: S1: Obtain multi-source observation requirements and analyze the spatial location, observation time window, and resolution requirements of each observation target; S2: Obtain the real-time operating status and physical attributes of satellite nodes, UAV nodes, and ground observation nodes in the air-space-ground multi-dimensional resource pool; S3: Based on the spatial location, observation time window, resolution requirements, and the real-time operating status and physical attributes, select executable nodes from the resource pool and assign observation tasks to each executable node to generate the optimal set of collaborative observation tasks; S4: Control each node to execute the optimal set of collaborative observation tasks to collect multi-source remote sensing data, and perform fusion processing on the collected multi-source remote sensing data to generate and distribute remote sensing information collaborative service products.

[0007] As a further aspect of the present invention, the parsing in step S1 includes: extracting geographic location named entities and time constraint phrases from unstructured text; mapping the extracted entities to a set of latitude and longitude coordinates, an effective time window, and a minimum spatial resolution threshold to complete the construction of the collaborative observation task profile.

[0008] As a further aspect of the present invention, in step S3, before node selection and task allocation, the heterogeneous physical parameters participating in scheduling are dimensionless. During the dimensionless processing, two mapping methods are distinguished based on the correlation between the physical parameter values ​​and performance: For physical parameters whose larger values ​​indicate better performance, a forward mapping is used, which involves subtracting the preset minimum threshold of the parameter under system constraints from the actual extracted value, and then dividing by the difference between the preset maximum threshold and the minimum threshold, so that the dimensionless parameter after mapping increases as the original value increases. For physical parameters whose smaller values ​​indicate better performance, an inverse mapping is used. This involves subtracting the actual extracted value from the preset maximum threshold of the parameter under system constraints, and then dividing by the difference between the preset maximum and minimum thresholds. This ensures that the dimensionless parameter after mapping also increases with performance optimization.

[0009] As a further aspect of the present invention, the allocation of observation tasks to each executable node in step S3 specifically includes: A multi-objective cooperative scheduling utility model is constructed. Under the premise of satisfying the physical constraint of the maximum effective field of view offset angle of the node sensors, the optimal set of cooperative observation tasks is selected by maximizing the comprehensive cooperative utility. The physical constraint is that the spatial angle between the line connecting the central axis of the observation field of each node and the center of the target does not exceed the sensor's inherent maximum allowable offset angle. The formula for calculating the comprehensive collaborative utility is: the sum of the following: the dimensionless parameter of the spatial observation resolution of each space-ground node assigned to perform the current observation task multiplied by the spatial resolution weight coefficient, the dimensionless parameter of the data transmission and processing delay multiplied by the data delay weight coefficient, and the dimensionless parameter of the platform's expected energy consumption multiplied by the energy consumption weight coefficient.

[0010] As a further aspect of the present invention, the fusion processing of the acquired multi-source remote sensing data in step S4 specifically includes: Calculate the signal-to-noise ratio (SNR) of each data source image, which is the ratio of effective signal power to noise power, and obtain the normalized SNR by processing it with a dimensionless mapping formula; calculate the feature saliency score of each data source, which is equal to the normalized SNR multiplied by the structural similarity between the data source image and the benchmark reference image. Set a noise floor tolerance threshold, which is equal to a preset scaling factor multiplied by the arithmetic mean of the feature significance scores of all data sources; if the feature significance score of a data source is lower than this threshold, the data source is deemed invalid and removed. The fusion weight of the unremoved data sources is calculated using a normalized exponential function with a temperature smoothing coefficient. This is the exponential value obtained by dividing the significance score of each unremoved data source feature by the temperature smoothing coefficient, and then dividing by the sum of the exponential values ​​of all unremoved data sources. Feature-level adaptive fusion of multi-source remote sensing data is completed by weighting each fusion weight with the corresponding feature vector of the data source. The structural similarity is calculated by comparing the data source image with the geometrically corrected reference image in blocks. The reference image is selected as the satellite data source image with the highest spatial resolution in the current observation area. If the image is determined to be invalid, the next available image is selected as the reference in descending order of spatial resolution.

[0011] As a further aspect of the present invention, the indicators obtained in step S1 also include the maximum allowable cloud cover threshold and the lower limit of the solar altitude angle in the observation area; based on the latitude and longitude coordinate set and the effective time window, the corresponding sensor side swing angle range and wave position working mode are matched, and all extracted spatiotemporal constraints and quality thresholds are assembled into a unified collaborative observation task profile constraint set, which serves as the input boundary conditions for node selection and task allocation in step S3.

[0012] As a further solution of the present invention, before performing feature-level adaptive fusion in step S4, for the heterogeneity of different data sources such as satellites, UAVs and ground nodes in terms of spatial resolution and pixel size, a bicubic interpolation or geographic projection inverse transformation algorithm is used to uniformly resample the feature vectors of each source to the same spatial reference grid to ensure that the feature vectors participating in the fusion have consistent dimensionality. The temperature smoothing coefficient is adaptively and dynamically adjusted according to the total number of valid data sources that have not been eliminated. The adjustment method is 1 plus the logarithm of the total number of valid data sources with base 2 divided by 10. When the total number of valid data sources is greater than or equal to 3, weighted fusion is performed; when the total number of valid data sources is greater than or equal to 1 but less than 3, weighted fusion is abandoned and the single data source feature with the highest signal-to-noise ratio is directly output; when the total number of valid data sources is 0, it is determined that all data sources of the current observation task have failed, triggering an alarm for no available data and terminating the fusion process.

[0013] The present invention also provides an integrated air-space-ground remote sensing information collaborative service system, the system comprising: The requirement analysis module is used to obtain multi-source observation requirements and analyze the spatial location, observation time window, and resolution requirements of each observation target. The resource perception module is used to acquire the real-time operating status and physical attributes of satellite nodes, UAV nodes and ground observation nodes in the air-space-ground multi-dimensional resource pool; The collaborative scheduling module is used to select executable nodes from the resource pool and assign observation tasks to each executable node based on the spatial location, observation time window, resolution requirements, real-time operating status and physical attributes, thereby generating an optimal set of collaborative observation tasks. The data fusion service module is used to control each node to execute the optimal collaborative observation task set to collect multi-source remote sensing data, and to perform fusion processing on the collected multi-source remote sensing data to generate and distribute remote sensing information collaborative service products.

[0014] As a further aspect of the present invention, the parsing in the requirement parsing module includes: extracting geographic location named entities and time constraint phrases from unstructured text; mapping the extracted entities to a set of latitude and longitude coordinates, an effective time window, and a minimum spatial resolution threshold to complete the construction of the collaborative observation task profile.

[0015] As a further aspect of this invention, during the operation of the collaborative scheduling module, before node selection and task allocation, the heterogeneous physical parameters participating in the scheduling are dimensionless. During the dimensionless processing, two mapping methods are distinguished based on the correlation between the physical parameter values ​​and performance: For physical parameters whose larger values ​​indicate better performance, a forward mapping is used, which involves subtracting the preset minimum threshold of the parameter under system constraints from the actual extracted value, and then dividing by the difference between the preset maximum threshold and the minimum threshold, so that the dimensionless parameter after mapping increases as the original value increases. For physical parameters whose smaller values ​​indicate better performance, an inverse mapping is used. This involves subtracting the actual extracted value from the preset maximum threshold of the parameter under system constraints, and then dividing by the difference between the preset maximum and minimum thresholds. This ensures that the dimensionless parameter after mapping also increases with performance optimization.

[0016] Compared with the prior art, the beneficial effects of the present invention are as follows: This invention addresses the low level of automation in observation requirement parsing in existing technologies by extracting geographic named entities and time-constrained phrases from unstructured text. These are then mapped to a set of latitude and longitude coordinates, an effective time window, and a minimum spatial resolution threshold, thereby automatically constructing a collaborative observation task profile and improving the efficiency of task parsing. Attached Figure Description

[0017] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention.

[0018] Figure 1 A flowchart of a method for collaborative remote sensing information services integrating air, space, and ground.

[0019] Figure 2 This is a structural diagram of the integrated air-space-ground remote sensing information collaborative service system.

[0020] Figure 3 This is a schematic diagram of the nodes in the space-air-ground multi-dimensional resource pool. Detailed Implementation

[0021] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0022] The specific implementation of the present invention will be described in detail below with reference to specific embodiments.

[0023] Please refer to Figure 1 and Figure 3 One embodiment of the present invention provides a method for collaborative service of integrated air-space-ground remote sensing information, comprising the following steps: S1: Obtain multi-source observation requirements and analyze them to obtain the spatial location, observation time window, and resolution requirements of each observation target. Multi-source observation requirements refer to remote sensing data acquisition requests from different users or application scenarios. These requirements may be presented in various forms, such as text descriptions, geographic coordinate input, or preset task templates. These requirements typically include specific requirements regarding the observation target, observation time, and spatial resolution.

[0024] S2: Acquire the real-time operational status and physical attributes of satellite nodes, UAV nodes, and ground observation nodes in the space-air-ground multi-dimensional resource pool. The space-air-ground multi-dimensional resource pool refers to a collection integrating various remote sensing observation resources, including satellite nodes, UAV nodes, and ground observation nodes. Each node in this resource pool possesses specific observation capabilities and operational status, and its physical attributes (such as sensor type, orbital parameters, and endurance) are perceived and managed by the system. Satellite nodes refer to remote sensing satellites operating in Earth orbit, characterized by wide coverage and strong observation periodicity. UAV nodes refer to unmanned aerial vehicles equipped with remote sensing payloads, characterized by high mobility and rapid deployment to specific areas for detailed observation. Ground observation nodes refer to various sensors or observation stations deployed on the ground, characterized by long-term fixed-point observation and the provision of high-frequency, high-resolution local data.

[0025] S3: Based on the spatial location, observation time window, resolution requirements, and the real-time operating status and physical attributes, select executable nodes from the resource pool and assign observation tasks to each executable node to generate an optimal collaborative observation task set; wherein, the real-time operating status and physical attributes refer to the current working status of each node in the space-air-ground multi-dimensional resource pool (e.g., power, fuel, communication link status, sensor health status) and its inherent physical characteristics (e.g., sensor field of view, maximum offset angle, data transmission rate, platform energy consumption model). This information is the basis for node selection and task allocation.

[0026] The optimal collaborative observation task set refers to a group of observation tasks jointly executed by multiple space-air-ground nodes after system evaluation and optimization. This set aims to meet the needs of multi-source observation in the best way, while maximizing the overall observation utility, such as minimizing resource consumption or shortening response time while ensuring observation quality.

[0027] S4: Control each node to execute the optimal collaborative observation task set to collect multi-source remote sensing data, and perform fusion processing on the collected multi-source remote sensing data to generate and distribute remote sensing information collaborative service products. Multi-source remote sensing data refers to raw remote sensing images, spectral data, or other observation data collected from different space-air-ground nodes. These data may exhibit heterogeneity in spatial resolution, spectral bands, timestamps, and data formats. Fusion processing refers to the process of integrating, correcting, and enhancing remote sensing data from different data sources. This process aims to eliminate heterogeneity between data, extract complementary information, and generate remote sensing information products that are more comprehensive, accurate, and reliable than those from a single data source. Remote sensing information collaborative service products refer to the final remote sensing information deliverables distributed to users in a specific format (e.g., thematic maps, data reports, 3D models) after fusion processing. This product can directly support user decision-making in areas such as environmental monitoring, disaster emergency response, and agricultural yield estimation.

[0028] The method in this embodiment constructs an efficient and reliable integrated air-space-ground remote sensing information collaborative service system through automated demand analysis, comprehensive node scheduling and task allocation, and multi-source data fusion processing. It provides an effective technical solution to address the problems of low automation, insufficient scheduling optimization, and unstable data fusion in existing technologies.

[0029] In a preferred embodiment of the present invention, the parsing in step S1 includes: extracting geographic location named entities and time-constrained phrases from unstructured text; and mapping the extracted entities to a set of latitude and longitude coordinates. Effective time window and minimum spatial resolution threshold Complete the construction of the collaborative observation mission profile, among which Indicates the first Each independent observation target.

[0030] In this embodiment, unstructured text refers to text data that does not have a predefined data model or fixed format, such as natural language descriptions, emails, social media posts, reports, and documents. It is characterized by high information density but loose organization, making it difficult to parse directly using traditional database queries or fixed rules.

[0031] Geographic name entities refer to proper nouns with geographical significance mentioned in unstructured text, such as city names, landmarks, administrative divisions, and descriptions of specific locations. Extracting these entities can be achieved using natural language processing techniques, such as rule-based pattern matching methods that predefine a dictionary of geographic terms and matching rules for identification; or by employing machine learning-based methods to train a named entity recognition model.

[0032] Time-constrained phrases are natural language expressions in unstructured text that describe a range or point in time of observation, such as "tomorrow morning" or "next Tuesday." These phrases can be extracted using regular expressions to define a series of patterns that match time expressions; or by utilizing time expression parsing libraries, which can identify and standardize various natural language time expressions.

[0033] The extracted geographic location named entities are converted into a standardized set of latitude and longitude coordinates. This is to transform vague geographical descriptions into precise spatial positioning information, in which... , Representing the first The latitude and longitude of each observation target can be determined. This can be achieved through geocoding services, such as calling the Amap API to convert place names into corresponding latitude and longitude; or by using localized geographic information system databases and geographic dictionaries for matching and conversion.

[0034] The extracted time constraint phrases are parsed and mapped to explicit start times. and end time The effective time window is designed to precisely define the execution period of the observation task. This can be achieved through time resolution algorithms, such as resolving "tomorrow morning" into a time period from 08:00 to 12:00 the following day; or by integrating with a calendar system to convert relative time expressions into absolute timestamps.

[0035] Minimum spatial resolution threshold This refers to the minimum requirements for the quality of observation data, representing the actual ground size that a unit pixel can represent. It can be obtained based on descriptions of observation precision in text, such as keywords like "high resolution" or "detailed image," which are then converted into specific numerical values ​​using preset mapping rules. For example, "high resolution" might correspond to a spatial resolution of 0.5 meters or lower. Alternatively, it can be extracted from values ​​directly specified by the user in their requirements.

[0036] The construction of the collaborative observation mission profile refers to the process of converting the latitude and longitude coordinate set obtained from the above analysis and mapping. Effective time window and minimum spatial resolution threshold Key parameters, for each individual observation target j, are integrated to form a structured task description that can be directly used by the subsequent scheduling system. This profile serves as a unified input, ensuring the accuracy and automation of subsequent node selection and task allocation.

[0037] In a preferred embodiment of the present invention, in step S3, before node selection and task allocation, the heterogeneous physical parameters involved in scheduling are dimensionless, and the mapping formula is as follows: in These are the actual extracted values ​​of the corresponding physical parameters. This is the preset minimum threshold for the physical parameter under system constraints. This is the preset maximum threshold value for this physical parameter under system constraints. The parameters are uniformly mapped to the [0,1] interval after processing, and are dimensionless. For parameters where smaller values ​​indicate better performance (including spatial resolution GSD, data transmission latency, and expected energy consumption), an inverse normalization formula is used. , making The larger the value, the better the performance, ensuring that each component converges towards better performance when maximizing the overall synergistic utility U; for parameters whose larger values ​​indicate better performance, a positive normalization formula is used.

[0038] In this embodiment, the aforementioned heterogeneous physical parameters refer to various parameters from different space-air-ground observation nodes used to describe their performance or status, such as spatial resolution, data transmission rate, sensor field of view, platform remaining power, task execution energy consumption, and data processing latency. These parameters have different physical units and numerical ranges due to factors such as node type, model, and current operating mode.

[0039] Dimensionless processing aims to eliminate the influence of dimensions on these parameters and unify their numerical range to a standard interval, so as to facilitate subsequent comprehensive evaluation and optimization calculations.

[0040] In the mapping formula This represents the actual measured or estimated value of the physical parameter to be processed. and These are the minimum and maximum thresholds allowed for the physical parameter under system design or operational constraints. These thresholds can be determined based on historical data statistical analysis, expert experience, or the system's minimum and maximum performance requirements. Through these preset thresholds, actual parameter values ​​can be mapped to the [0,1] interval, where 0 typically represents worst performance or highest cost, 1 represents best performance or lowest cost, or vice versa, depending on the nature of the parameter and the optimization objective.

[0041] In a preferred embodiment of the present invention, the step S3 of assigning observation tasks to each executable node specifically includes: Construct a multi-objective cooperative scheduling utility model, while satisfying the physical constraint of the maximum effective field-of-view offset angle of the node sensors. Under the premise of maximizing comprehensive synergistic utility The optimal set of cooperative observation tasks was selected; wherein For the first The spatial angle between the central axis of the observation field of view of each node and the line connecting the center of the target. The comprehensive cooperative utility is the sensor's inherent maximum allowable bias angle. The calculation formula is as follows: in, The higher the overall performance of the scheduling scheme, the better its overall performance. To allocate the total number of air-space-ground nodes to perform the current observation task; Let be the dimensionless parameter of the spatial observation resolution of the i-th node, derived from the spatial resolution parameter of the i-th node. The mapping formula above was used to calculate the result. For the first The dimensionless parameter for the data transmission and processing delay of each node is derived from the first node. Data transmission and processing latency parameters of each node The mapping formula above was used to calculate the result. For the first The dimensionless parameter of the expected energy consumption of each node platform executing a task is derived from the expected energy consumption parameter of the i-th node platform. Obtained by the aforementioned mapping formula; Spatial resolution weighting coefficients, For data delay weighting coefficient, This is the platform's expected energy consumption weighting coefficient. .

[0042] In this embodiment, the scheme constructs a multi-objective collaborative scheduling utility model, aiming to provide a quantitative decision-making basis for complex resource scheduling problems, thereby enabling the systematic comparison and selection of the optimal scheduling scheme. This model can integrate multiple interrelated or potentially conflicting optimization objectives (e.g., observation resolution, data transmission latency, platform energy consumption, etc.) into a unified evaluation framework.

[0043] When assigning tasks, it is essential to ensure that the spatial angle between the central axis of the field of view of each node sensor and the line connecting the center of the target is consistent when observing the target. Not exceeding the sensor's inherent maximum permissible offset angle This constraint is set based on the sensor's physical characteristics and imaging quality requirements, aiming to ensure that the acquired remote sensing data has acceptable geometric accuracy and image quality. Under the premise of satisfying all physical constraints, an optimization algorithm is used to find the optimal comprehensive synergistic utility. The task allocation scheme that maximizes the overall performance. The comprehensive collaborative utility "U" is a quantitative evaluation of the overall performance of the scheduling scheme, maximizing... This means that an optimal balance has been found among multiple objectives. (Comprehensive synergistic utility) The calculation formula defines its specific calculation method; it is a weighted sum that includes the spatial observation resolution. Data transmission and processing delay and platform expected energy consumption These are the dimensionless parameters of the three key performance indicators. By weighted summing of these parameters, the performance of a scheduling scheme across different dimensions can be comprehensively evaluated. These dimensionless parameters are all calculated from their respective original physical parameters using dimensionless mapping formulas.

[0044] Weighting coefficient , , These correspond to the importance of spatial resolution, data latency, and expected platform power consumption, respectively. They are configurable and can be adjusted according to specific task requirements and system strategies. The constraints ensure that the sum of all weight coefficients is 1, making the calculation results of utility "U" more interpretable and comparable.

[0045] In the aforementioned integrated space-air-ground remote sensing information collaborative service method, to address the potential problems of insufficient resource utilization, low task execution efficiency, and difficulty in balancing multiple optimization objectives during node selection and task allocation, this application further proposes a multi-objective collaborative scheduling utility model. This model, based on dimensionless processing of the heterogeneous physical parameters involved in scheduling, introduces a comprehensive collaborative utility degree. As a core indicator for evaluating the quality of a scheduling scheme, specifically, after receiving the observation task requirements that have been analyzed and profiled, the system first preliminarily filters out executable nodes that meet basic conditions based on the real-time operating status and physical attributes of each node in the space-air-ground multi-dimensional resource pool. Subsequently, for these executable nodes, the system constructs a multi-objective collaborative scheduling utility model. This model incorporates key performance indicators such as spatial observation resolution, data transmission and processing latency, and expected platform energy consumption through preset weighting coefficients. , , Weighted combinations are performed to form a comprehensive synergistic utility. When constructing and evaluating scheduling schemes, the system strictly adheres to the maximum effective field-of-view offset angle of the node sensors. The system addresses physical constraints to ensure the selected solution is physically feasible. Through iterative or optimization algorithms, the system continuously adjusts the task allocation scheme and calculates the corresponding calculations for each scheme. The value is ultimately selected to maximize the value. The optimal set of collaborative observation tasks is determined. This process ensures that the selected task set not only meets the observation requirements but also achieves the best balance between resolution, timeliness, and energy consumption, thereby significantly improving the overall performance and efficiency of the integrated space-air-ground remote sensing information collaborative service.

[0046] In a preferred embodiment of the present invention, the fusion processing of the acquired multi-source remote sensing data in step S4 specifically includes: Let the first The signal-to-noise ratio of the images from the data sources is Its value is calculated by the ratio of the effective signal power to the noise power of the image from the data source; Will After processing with the dimensionless mapping formula, the normalized signal-to-noise ratio is obtained. ; according to Calculate the first Feature significance score of each data source The feature saliency score This is a heuristic scoring model based on engineering experience, where the normalized signal-to-noise ratio characterizes the radiometric quality of the data source itself, and the structural similarity characterizes its geometric consistency with the highest resolution reference image. Multiplying the two together can comprehensively evaluate the effective contribution of each data source to the fusion result under a unified dimension.

[0047] Set the noise floor tolerance threshold ,in The arithmetic mean of the significance scores of features from all data sources. The preset proportional coefficient, and the The value range is [0.5, 0.8]; if If so, the data source is deemed invalid and removed. The fusion weights of the data sources without data removal are calculated using a normalized exponential function with a temperature smoothing coefficient. The calculation formula is as follows: in, This represents the total number of data sources that were not removed. It is the temperature smoothing coefficient, and ;according to Complete feature-level adaptive fusion of multi-source remote sensing data. For the first Feature vectors of one data source; Wherein, the structural similarity The reference image is obtained by comparing it in blocks with a geometrically corrected baseline reference image. The baseline reference image is selected as the satellite data source image with the highest spatial resolution within the current observation area. If the image is determined to be invalid, the next available image is selected as the baseline in descending order of spatial resolution. The selection of the baseline reference image is calculated... Previously, the selection was based independently on the signal-to-noise ratio (SNR) and cloud cover detection results of the data sources. The highest resolution image with an SNR higher than the average SNR of all data sources and a cloud cover lower than Cmax / 2 was selected first to avoid the failure of the benchmark selection caused by the circular dependency of SSIM calculation.

[0048] In this embodiment, the signal-to-noise ratio Signal-to-noise ratio (SNR) is an important metric for measuring the quality of images from a data source. It is defined as the ratio of effective signal power to noise power in an image. A higher SNR generally indicates better image quality, containing more effective information and less noise interference. Its value can be calculated by weighing the effective signal power and noise power of the data source image. For example, it can be estimated by analyzing the statistical characteristics of the image, such as the mean and variance in a uniform region. To eliminate the dimensional differences in SNR between different data sources and to allow for comparison and processing on a uniform scale, the SNR... The signal-to-noise ratio needs to be obtained by processing with a dimensionless mapping formula. .

[0049] Feature significance score The intrinsic quality of the data source was taken into account (through normalized signal-to-noise ratio). (reflected) and structural similarity with the benchmark reference image (through structural similarity) (Reflected). This score quantifies the potential contribution of each data source to the final fusion result, and is calculated using the normalized signal-to-noise ratio. Similar in structure The product of.

[0050] Noise floor tolerance threshold This is a dynamically set threshold used to determine whether a data source should be excluded due to low quality or insufficient relevance to the observation target. The threshold is calculated as the arithmetic mean of the significance scores of all data source features. Multiplied by a preset scaling factor To determine, among which the proportionality coefficient The value range is [0.5, 0.8]. This dynamic threshold setting method can adapt to the overall distribution of data source quality under different observation scenarios. When the feature significance score of a certain data source... Below the set noise floor tolerance threshold When a data source is deemed invalid, it is considered a failed data source. Failed data sources, due to insufficient quality or relevance, may introduce noise or errors if included in the fusion process; therefore, they need to be removed to ensure the quality of the final fusion result. To achieve feature-level adaptive fusion of multi-source remote sensing data, a fusion weight needs to be calculated for each data source that is not removed. .

[0051] These weights are calculated using a temperature-smoothing coefficient. It is obtained by calculating the normalized exponential function, and its calculation formula is: .in, The total number of data sources that were not removed, and the temperature smoothing coefficient. The value range is [0.8, 1.5], used to adjust the influence of significance scores on weight allocation, making the weight allocation smoother or more discriminative. The fusion weights for each data source that was not removed are calculated. Then, by using the feature vectors of each data source Its corresponding fusion weight Perform a weighted summation, that is, according to Complete feature-level adaptive fusion of multi-source remote sensing data.

[0052] This fusion method operates at the feature level, which better preserves the effective information from each data source and generates a comprehensive feature vector. Structural similarity This measure assesses the structural similarity between a data source image and a reference image. The value is calculated by comparing the data source image with the geometrically corrected reference image in blocks. The selection of the reference image is crucial; typically, the satellite data source image with the highest spatial resolution within the current observation area is chosen as the reference. If this image is deemed invalid, the next available image is selected as the reference in descending order of spatial resolution, ensuring a consistently high-quality reference standard.

[0053] In a preferred embodiment of the present invention, the index obtained in step S1 further includes the maximum permissible cloud cover threshold. and the lower limit of the solar altitude angle in the observation area Based on the latitude and longitude coordinate set With effective time window Match the corresponding sensor side-swing angle range and wave position working mode, and assemble all extracted spatiotemporal constraints and quality thresholds into a unified collaborative observation mission profile constraint set. This serves as the input boundary condition for node selection and task allocation in step S3.

[0054] In this embodiment, the maximum permissible cloud cover threshold This refers to the maximum percentage of cloud cover allowed over a target area during remote sensing observation missions. Its purpose is to ensure that the acquired remote sensing data has acceptable quality, avoiding invalid or missing data due to cloud obstruction. This threshold can be set based on cloud cover information for the target area obtained from historical meteorological data, real-time meteorological satellite observation data, or numerical weather prediction models, and compared with the actual cloud cover; alternatively, it can be flexibly configured according to the user's specific data quality requirements or mission priorities.

[0055] Lower limit of solar altitude angle in the observation area In remote sensing observation missions, the solar altitude angle refers to the minimum angle between the sun and the horizon at the time of observation for a target area. Its purpose is to ensure that remote sensing images have sufficient brightness, contrast, and shadow information, avoiding poor image quality, difficulty in target identification, or failure to meet specific application requirements due to excessively low solar altitude angles. This lower limit can be calculated using an astronomical calculation model based on the observation time, location, and date, and then compared with this lower limit; alternatively, it can be determined based on the different sensitivities of different types of sensors (such as optical sensors) to lighting conditions and the image quality requirements of specific application scenarios (such as topographic mapping and vegetation monitoring).

[0056] The tilt angle range of a sensor refers to the maximum tilt angle that the central axis of its field of view can deflect relative to the platform's flight direction or vertical direction when performing an observation task. This parameter determines the observation range that the sensor can cover and its ability to revisit specific targets, directly affecting the flexibility and efficiency of the observation task. This range is usually determined by the sensor's hardware design parameters and provided by the manufacturer; alternatively, it can be dynamically adjusted according to mission requirements. For example, a specific tilt angle range may be needed to acquire stereo or multi-angle observation data.

[0057] The spectral operating mode refers to the specific operating state or configuration adopted by a remote sensing sensor when performing an observation task. This includes different combinations of spectral bands, different imaging modes (such as strip mode, spotlight mode, and scanning mode), or different polarization methods. This mode affects the spectral characteristics, spatial resolution, coverage, and data volume of the acquired data to adapt to different application requirements. The mode can be preset according to the sensor type and task requirements; for example, an optical sensor might select the visible light or near-infrared bands, while a SAR sensor might select the X-band or C-band. Alternatively, it can be dynamically switched according to target characteristics and observation objectives to obtain the most effective data.

[0058] Cooperative observation mission profile constraint set This is a unified set used to encapsulate all spatiotemporal constraints and quality thresholds related to remote sensing observation tasks. Its function is to serve as the input boundary conditions for node selection and task allocation in step S3, ensuring that the scheduling results meet all key constraints, thereby improving the effectiveness and feasibility of the scheduling scheme. This constraint set can be stored and managed using data structures; alternatively, it can be encapsulated in a task object using an object-oriented approach for convenient calling and processing by subsequent scheduling algorithms.

[0059] In a preferred embodiment of the present invention, before performing feature-level adaptive fusion in step S4, considering the heterogeneity of different data sources such as satellites, UAVs, and ground nodes in terms of spatial resolution and pixel size, a bicubic interpolation or inverse geographic projection transformation algorithm is used to transform the feature vectors of each source. Uniform resampling to the same spatial reference grid ensures that the feature vectors participating in the fusion have consistent dimensionality; the temperature smoothing coefficient The system adaptively and dynamically adjusts based on the total number M of valid data sources that have not yet been removed. The adjustment function is as follows: Furthermore, when M < 3, a forced rollback strategy is triggered, abandoning weighted fusion and directly outputting the single data source feature with the highest signal-to-noise ratio as the fusion result; When M≥3, weighted fusion is performed; when 1≤M<3, weighted fusion is abandoned and the single data source feature with the highest signal-to-noise ratio is directly output; when M=0, it is determined that all data sources of the current observation task have failed, triggering an alarm for no available data and terminating the fusion process.

[0060] In this embodiment, to address the heterogeneity of different data sources in terms of spatial resolution and pixel size, bicubic interpolation or inverse geographic projection transformation algorithms are used to transform the feature vectors of each source. Unified resampling to the same spatial reference grid aims to address the spatial scale inconsistencies of multi-source data. Bicubic interpolation, an advanced image resampling technique, generates smoother, more detail-preserving new pixel values ​​by weighted averaging of the surrounding 16 pixels, suitable for scaling up or down remote sensing images. The inverse geographic projection transformation algorithm involves uniformly transforming data from different projection coordinate systems to the same geographic coordinate system and projection, and then performing grid resampling on this basis to ensure precise spatial alignment of all data. Both methods effectively eliminate spatial heterogeneity, ensuring that feature vectors from different platforms maintain consistency in dimension and spatial location, laying the foundation for subsequent feature-level fusion.

[0061] The temperature smoothness coefficient The adaptive dynamic adjustment optimizes key parameters in the fusion weight calculation. Temperature smoothing coefficient. In exponential functions, it plays a role in adjusting the "sharpness" of the weight distribution. Smaller... This will cause the weights to be more concentrated on data sources with high significance scores, while larger ones will be less significant. This will result in a smoother weight distribution, allowing contributions from more data sources. (Through the function...) This coefficient can be dynamically adjusted based on the total number M of valid data sources. When M is small, When M is relatively small, the fusion process tends to highlight a few high-quality data sources; when M is large, The relatively large value allows the fusion process to make fuller use of information from multiple data sources, achieving a more comprehensive fusion. This adaptive mechanism enables the fusion strategy to flexibly respond to changes in data availability.

[0062] Furthermore, different fusion strategies are triggered based on the total number M of valid data sources, aiming to ensure optimal or most reliable results under various data availability conditions. When M ≥ 3, the system considers there are enough data sources for robust weighted fusion, fully utilizing the complementary information from multiple data sources. When 1 ≤ M < 3, the number of data sources is limited, and the statistical robustness of weighted fusion may be insufficient. In this case, the system abandons weighted fusion and directly outputs the feature of the single data source with the highest signal-to-noise ratio. This fallback strategy avoids the instability of fusion results or the introduction of noise that may occur due to data scarcity, ensuring the quality of the output data. When M = 0, it indicates that no valid data sources are available. In this case, the system triggers an alarm and terminates the fusion process, avoiding the output of invalid or empty data and improving system reliability.

[0063] The technical solution of this invention, through the aforementioned technical means, in the process of multi-source remote sensing data fusion processing, firstly, by employing bicubic interpolation or inverse geographic projection transformation algorithms, integrates feature vectors from different data sources such as satellites, UAVs, and ground nodes. The data is uniformly resampled to the same spatial reference grid. This preprocessing step effectively eliminates the heterogeneity of data in terms of spatial resolution and cell size, ensuring that all feature vectors involved in the fusion have consistent dimensionality and spatial alignment, thus providing a unified and high-quality input for subsequent feature-level fusion.

[0064] Secondly, the technical solution of this invention introduces a temperature smoothing coefficient. An adaptive dynamic adjustment mechanism. This coefficient is based on the total number M of currently unremoved valid data sources, adjusted via a function. The calculations are performed. This dynamic adjustment allows the fusion weights to be distributed more smoothly when there are enough data sources, while giving preference to high-quality data sources when data sources are scarce, thereby optimizing the fusion process's responsiveness to differences in data quality.

[0065] Finally, the technical solution of this invention implements a multi-stage fusion strategy based on the total number M of valid data sources. When M ≥ 3, weighted fusion is performed to fully utilize multi-source information; when 1 ≤ M < 3, a forced rollback strategy is triggered, abandoning weighted fusion and directly outputting the feature of the single data source with the highest signal-to-noise ratio to ensure output quality when data is scarce; when M = 0, an alarm for no available data is triggered and the fusion process is terminated. This series of coordinated actions ensures the robustness, intelligence, and reliability of the remote sensing data fusion process, enabling the system to adapt to complex and ever-changing data environments and consistently provide high-quality remote sensing information collaborative service products.

[0066] As a preferred embodiment of the technical solution of the present invention, the method further includes: Based on the observation time window, target state change rate, task urgency, and timeliness requirements of the observation results for each observation target, the dynamic task priority for each observation target is determined, and the observation tasks to be executed are sorted according to the dynamic task priority. Based on the observation coverage, spatial resolution, and mobility of satellite nodes, UAV nodes, and ground observation nodes, the ordered observation tasks are scheduled in a hierarchical manner. First, satellite nodes are invoked to perform wide-area initial observations of the target area, and abnormal areas in the target area are determined based on the results of the wide-area initial observations. For the anomalous area, based on the spatial range, degree of anomalousness, and uncertainty of the target state, corresponding executable UAV nodes are selected from the UAV nodes, and the executable UAV nodes are scheduled to perform local high-resolution fine observation of the anomalous area; based on the results of the local high-resolution fine observation, it is determined whether it is necessary to call ground observation nodes to perform supplementary verification observation of the anomalous area. Acquire new observation data generated by each supplementary observation task, and determine the information gain of the corresponding supplementary observation task based on the degree of improvement of the target identification confidence or target state uncertainty by the new observation data; dynamically adjust the node allocation priority of subsequent observation tasks based on the information gain, reduce the task scheduling priority or cancel the execution of observation tasks with information gain lower than a preset threshold, and increase the task scheduling priority of observation tasks with information gain higher than a preset threshold. Consistency analysis is performed on the multi-source observation results acquired by satellite nodes, UAV nodes, and ground observation nodes respectively, and consistency scores are calculated between different observation results. When the consistency score is lower than a preset consistency threshold, it is determined that there is a conflict between the multi-source observation results of the current observation target. Based on the target area, data type, and observation uncertainty corresponding to the conflict between the multi-source observation results, the corresponding air-space-ground observation nodes are re-selected to perform supplementary observation tasks. Multi-source remote sensing data obtained through multiple rounds of collaborative observation are fused, and the comprehensive observation confidence level corresponding to the target area is determined based on the fusion results. When the comprehensive observation confidence level reaches a preset confidence threshold, the subsequent collaborative observation task for the target area is terminated. When the comprehensive observation confidence level is lower than the preset confidence threshold, the area to be supplemented and the corresponding data type to be supplemented are determined based on the current fusion results, and the supplementary observation task of the UAV node or ground observation node is re-executed until the comprehensive observation confidence level reaches the preset confidence threshold or the preset maximum number of collaborative observations is reached.

[0067] In one example of the technical solution of this invention, the execution process of the observation task is further optimized based on the original integrated air-space-ground remote sensing information collaborative service method. The main purpose is to avoid satellites, UAVs and ground observation nodes from performing indiscriminate and repetitive observations when performing tasks, so that different types of observation equipment can cooperate according to their respective advantages, and dynamically decide whether to continue observation based on the observation results already obtained, thereby improving the overall observation efficiency while ensuring the observation accuracy.

[0068] First, task priorities are determined based on the urgency and importance of different observation tasks. For example, targets that change rapidly and require timely information, such as fires, floods, and geological disasters, are given higher priorities; while tasks with slower-changing conditions, such as forest surveys and land surveys, can be given relatively lower priorities. In this way, when multiple tasks exist simultaneously, the more urgent tasks can be prioritized.

[0069] During specific observations, layered observations are conducted based on the characteristics of satellites, drones, and ground observation nodes. Satellites offer wide coverage, so they are initially used to conduct preliminary observations of a large target area—essentially a "quick look" at the entire region. The preliminary data acquired by the satellites is used to determine if any areas within the target region may be subject to change or exhibit abnormal characteristics. If abnormal areas are found, instead of requiring drones to conduct comprehensive observations of the entire target area, drones are dispatched only to these key areas for high-resolution, detailed observations. This approach fully leverages the advantages of satellite's wide coverage and drone's high spatial resolution, reducing unnecessary drone flights and data collection.

[0070] For areas where precise assessment is still impossible after detailed observation by drones, further on-site verification can be conducted using ground-based observation nodes. For example, if a satellite detects a potential anomaly in a certain area, and further drone imaging fails to determine the cause, ground sensors or other ground-based observation equipment can be used for further investigation. Therefore, the entire process can be understood as "the satellite first identifies the problem, the drone focuses on inspection, and ground equipment verifies as necessary."

[0071] After each observation, the system also determines whether the observation truly brought new and effective information. "Information gain" here can be simply understood as "how much this new observation helps improve the accuracy of target identification." If a new observation makes the target identification result clearer, or significantly reduces the system's doubts about the target's state, then the information gain of this observation is high, and similar observations can be arranged subsequently. If the new observation results are basically the same as existing data and do not significantly add new information, then continuing to conduct similar observations is less meaningful, and the priority of this task can be reduced or even stopped to reduce redundant observations.

[0072] Furthermore, because satellites, drones, and ground equipment acquire data in different ways, different devices may give different assessments of the same target. Therefore, the system also performs consistency analysis on the results obtained from different observation nodes. For example, if a satellite determines that a certain area may have an anomaly, while drone observations show that the area is normal, this indicates that there may be a conflict between the observation results from different data sources. In this case, the system can reselect appropriate observation equipment for supplementary observations based on the location and data type of the conflict to further determine the true situation.

[0073] After multiple observations and data fusions, the system determines the overall observation confidence level of the current target area, that is, whether the existing multi-source observation data is sufficiently reliable. If the overall observation confidence level meets the preset requirements, it means that the currently acquired information is sufficient to meet the mission requirements. At this point, the system stops scheduling satellite, UAV, or ground observation nodes to avoid unnecessary observation costs. If the overall observation confidence level is still low, it means that the existing data is insufficient to support accurate judgment. The system will then redetermine the areas requiring key observations and the types of data to be acquired based on the current results, and reschedule the corresponding observation nodes for supplementary observations until sufficiently reliable observation results are obtained or the preset maximum number of observations is reached.

[0074] Therefore, this implementation method forms a dynamic air-space-ground collaborative observation process, which enables different observation nodes to work collaboratively as needed through "wide-area satellite detection, in-depth UAV inspection, necessary verification of ground equipment, dynamic adjustment based on new information, supplementary observation when conflicting results are found, and timely cessation after the accuracy requirements are met." This reduces invalid observations and duplicate data collection while improving the efficiency and accuracy of remote sensing information acquisition in the target area.

[0075] In summary, the simplified expression of the above process is as follows: Based on the observation time window, target status change rate, task urgency, and timeliness requirements of observation results for each observation target, the dynamic task priority is determined, and the observation tasks to be executed are sorted. Based on the observation coverage, spatial resolution, and mobility of satellite nodes, UAV nodes, and ground observation nodes, the ranked observation tasks are scheduled in a hierarchical manner, and the abnormal areas of the target area are identified through satellite nodes. High-resolution, detailed observations are conducted on the drone nodes in the abnormal area, and based on the detailed observation results, it is determined whether to call ground observation nodes for supplementary verification. The information gain is determined based on the new observation data generated from supplementary observations, and the node allocation priority of subsequent observation tasks is adjusted according to the information gain. Consistency analysis is performed on the multi-source observation results obtained from satellite nodes, UAV nodes and ground observation nodes. When the consistency score is lower than the preset threshold, the air-space-ground observation nodes are re-selected to perform supplementary observation tasks. Multi-source remote sensing data obtained from multiple rounds of collaborative observation are fused and processed, and the comprehensive observation confidence level is determined based on the fusion result. When the comprehensive observation confidence level reaches a preset threshold, the subsequent collaborative observation task is terminated; otherwise, the supplementary observation task continues to be executed.

[0076] As a preferred embodiment of the technical solution of the present invention, the present invention also provides an integrated air-space-ground remote sensing information collaborative service system 10, the integrated air-space-ground remote sensing information collaborative service system 10 comprising: Module 11 for demand analysis is used to obtain multi-source observation requirements and analyze the spatial location, observation time window and resolution requirements of each observation target. Resource perception module 12 is used to acquire the real-time operating status and physical attributes of satellite nodes, UAV nodes and ground observation nodes in the air-space-ground multi-dimensional resource pool; The collaborative scheduling module 13 is used to select executable nodes from the resource pool and assign observation tasks to each executable node according to the spatial location, observation time window, resolution requirements, real-time running status and physical attributes, and generate an optimal set of collaborative observation tasks. The data fusion service module 14 is used to control each node to execute the optimal collaborative observation task set to collect multi-source remote sensing data, and to perform fusion processing on the collected multi-source remote sensing data to generate and distribute remote sensing information collaborative service products.

[0077] The parsing described in the requirement parsing module 11 includes: extracting geographic location named entities and time constraint phrases from unstructured text; mapping the extracted entities to a set of latitude and longitude coordinates, an effective time window, and a minimum spatial resolution threshold to complete the construction of the collaborative observation task profile.

[0078] Specifically, during the operation of the collaborative scheduling module 13, before node selection and task allocation, the heterogeneous physical parameters participating in the scheduling are dimensionless. During dimensionless processing, two mapping methods are distinguished based on the correlation between the physical parameter values ​​and performance: For physical parameters whose larger values ​​indicate better performance, a forward mapping is used, which involves subtracting the preset minimum threshold of the parameter under system constraints from the actual extracted value, and then dividing by the difference between the preset maximum threshold and the minimum threshold, so that the dimensionless parameter after mapping increases as the original value increases. For physical parameters whose smaller values ​​indicate better performance, an inverse mapping is used. This involves subtracting the actual extracted value from the preset maximum threshold of the parameter under system constraints, and then dividing by the difference between the preset maximum and minimum thresholds. This ensures that the dimensionless parameter after mapping also increases with performance optimization.

[0079] The core innovation of this embodiment lies in the systematic integration of the demand analysis module 11, resource perception module 12, collaborative scheduling module 13, and data fusion service module 14, achieving automatic analysis of observation demands, optimization of multi-objective collaborative scheduling, and adaptive fusion of multi-source remote sensing data. Specifically, addressing the low level of automation in observation demand analysis in existing technologies, the demand analysis module 11 can extract geographic location named entities and time constraint phrases from unstructured text and map them to a set of latitude and longitude coordinates, an effective time window, and a minimum spatial resolution threshold, thereby automatically constructing a collaborative observation task profile. Furthermore, to address the deficiency of lacking multi-objective comprehensive utility evaluation in node scheduling, the collaborative scheduling module 13, under the physical constraint of the maximum effective field of view offset angle of the node sensors, selects the optimal set of collaborative observation tasks by maximizing the comprehensive collaborative utility degree U. The calculation of U comprehensively considers the dimensionless parameters of spatial resolution, data transmission and processing delay, and platform expected energy consumption. Furthermore, to address the issues of instability in multi-source remote sensing data fusion processing and insufficient spatial benchmark consistency processing, the data fusion service module 14 adopts an adaptive fusion strategy based on feature saliency scoring. It calculates the feature saliency score by multiplying the signal-to-noise ratio and structural similarity of each data source and dynamically adjusts the fusion weight accordingly. At the same time, it introduces a noise floor tolerance threshold to automatically remove invalid data sources. In addition, to address the heterogeneity of satellite nodes, UAV nodes, and ground observation nodes in terms of spatial resolution and pixel size, the module uses bicubic interpolation or inverse geographic projection transformation algorithms to uniformly resample the feature vectors of each source to the same spatial benchmark grid, effectively avoiding errors introduced by pixel-level fusion.

[0080] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for integrated air-space-ground remote sensing information collaborative services, characterized in that, The method includes: S1: Obtain multi-source observation requirements and analyze the spatial location, observation time window, and resolution requirements of each observation target; S2: Obtain the real-time operating status and physical attributes of satellite nodes, UAV nodes, and ground observation nodes in the air-space-ground multi-dimensional resource pool; S3: Based on the spatial location, observation time window, resolution requirements, and the real-time operating status and physical attributes, select executable nodes from the resource pool and assign observation tasks to each executable node to generate the optimal set of collaborative observation tasks; S4: Control each node to execute the optimal set of collaborative observation tasks to collect multi-source remote sensing data, and perform fusion processing on the collected multi-source remote sensing data to generate and distribute remote sensing information collaborative service products. The parsing described in step S1 includes: extracting geographic location named entities and time constraint phrases from unstructured text; mapping the extracted entities to a set of latitude and longitude coordinates, an effective time window, and a minimum spatial resolution threshold to complete the construction of the collaborative observation task profile.

2. The integrated air-space-ground remote sensing information collaborative service method according to claim 1, characterized in that, In step S3, before node selection and task allocation, the heterogeneous physical parameters participating in scheduling are dimensionless. During the dimensionless processing, two mapping methods are distinguished based on the correlation between the physical parameter values ​​and performance: For physical parameters whose larger values ​​indicate better performance, a forward mapping is used, which involves subtracting the preset minimum threshold of the parameter under system constraints from the actual extracted value, and then dividing by the difference between the preset maximum threshold and the minimum threshold, so that the dimensionless parameter after mapping increases as the original value increases. For physical parameters whose smaller values ​​indicate better performance, an inverse mapping is used. This involves subtracting the actual extracted value from the preset maximum threshold of the parameter under system constraints, and then dividing by the difference between the preset maximum and minimum thresholds. This ensures that the dimensionless parameter after mapping also increases with performance optimization.

3. The integrated air-space-ground remote sensing information collaborative service method according to claim 2, characterized in that, The specific steps in step S3, which involve assigning observation tasks to each executable node, include: A multi-objective cooperative scheduling utility model is constructed. Under the premise of satisfying the physical constraint of the maximum effective field of view offset angle of the node sensors, the optimal set of cooperative observation tasks is selected by maximizing the comprehensive cooperative utility. The physical constraint is that the spatial angle between the line connecting the central axis of the observation field of each node and the center of the target does not exceed the sensor's inherent maximum allowable offset angle. The formula for calculating the comprehensive collaborative utility is: the sum of the following: the dimensionless parameter of the spatial observation resolution of each space-ground node assigned to perform the current observation task multiplied by the spatial resolution weight coefficient, the dimensionless parameter of the data transmission and processing delay multiplied by the data delay weight coefficient, and the dimensionless parameter of the platform's expected energy consumption multiplied by the energy consumption weight coefficient.

4. The integrated air-space-ground remote sensing information collaborative service method according to claim 2, characterized in that, The fusion processing of the acquired multi-source remote sensing data described in step S4 specifically includes: Calculate the signal-to-noise ratio (SNR) of each data source image, which is the ratio of effective signal power to noise power, and obtain the normalized SNR by processing it with a dimensionless mapping formula; calculate the feature saliency score of each data source, which is equal to the normalized SNR multiplied by the structural similarity between the data source image and the benchmark reference image. Set a noise floor tolerance threshold, which is equal to a preset scaling factor multiplied by the arithmetic mean of the feature significance scores of all data sources; if the feature significance score of a data source is lower than this threshold, the data source is deemed invalid and removed. The fusion weight of the unremoved data sources is calculated using a normalized exponential function with a temperature smoothing coefficient. This is the exponential value obtained by dividing the significance score of each unremoved data source feature by the temperature smoothing coefficient, and then dividing by the sum of the exponential values ​​of all unremoved data sources. Feature-level adaptive fusion of multi-source remote sensing data is completed by weighting each fusion weight with the corresponding feature vector of the data source. The structural similarity is calculated by comparing the data source image with the geometrically corrected reference image in blocks. The reference image is selected as the satellite data source image with the highest spatial resolution in the current observation area. If the image is determined to be invalid, the next available image is selected as the reference in descending order of spatial resolution.

5. The integrated air-space-ground remote sensing information collaborative service method according to claim 4, characterized in that, The indicators obtained in step S1 also include the maximum allowable cloud cover threshold and the lower limit of the solar altitude angle in the observation area; based on the latitude and longitude coordinate set and the effective time window, the corresponding sensor side swing angle range and wave position working mode are matched, and all extracted spatiotemporal constraints and quality thresholds are assembled into a unified collaborative observation task profile constraint set, which serves as the input boundary conditions for node selection and task allocation in step S3.

6. The integrated air-space-ground remote sensing information collaborative service method according to claim 4, characterized in that, Before performing feature-level adaptive fusion in step S4, considering the heterogeneity of different data sources such as satellites, UAVs, and ground nodes in terms of spatial resolution and pixel size, bicubic interpolation or inverse geographic projection transformation algorithms are used to uniformly resample the feature vectors of each source to the same spatial reference grid, ensuring that the feature vectors participating in the fusion have consistent dimensionality. The temperature smoothing coefficient is adaptively and dynamically adjusted according to the total number of valid data sources that have not been eliminated. The adjustment method is 1 plus the logarithm of the total number of valid data sources with base 2 divided by 10. Weighted fusion is performed when the total number of valid data sources is greater than or equal to 3; when the total number of valid data sources is greater than or equal to 1 but less than 3, weighted fusion is abandoned and the feature of the single data source with the highest signal-to-noise ratio is directly output. When the total number of valid data sources is 0, it is determined that all data sources for the current observation task have failed, triggering an alarm for no available data and terminating the fusion process.

7. A space-air-ground integrated remote sensing information collaborative service system, characterized in that, The system includes: The requirement analysis module is used to obtain multi-source observation requirements and analyze the spatial location, observation time window, and resolution requirements of each observation target. The resource perception module is used to acquire the real-time operating status and physical attributes of satellite nodes, UAV nodes and ground observation nodes in the air-space-ground multi-dimensional resource pool; The collaborative scheduling module is used to select executable nodes from the resource pool and assign observation tasks to each executable node based on the spatial location, observation time window, resolution requirements, real-time operating status and physical attributes, thereby generating an optimal set of collaborative observation tasks. The data fusion service module is used to control each node to execute the optimal collaborative observation task set to collect multi-source remote sensing data, and to perform fusion processing on the collected multi-source remote sensing data to generate and distribute remote sensing information collaborative service products. The parsing described in the requirements parsing module includes: extracting geographic location named entities and time constraint phrases from unstructured text; mapping the extracted entities to a set of latitude and longitude coordinates, an effective time window, and a minimum spatial resolution threshold to complete the construction of the collaborative observation task profile.

8. The integrated air-space-ground remote sensing information collaborative service system according to claim 7, characterized in that, During the operation of the collaborative scheduling module, before node selection and task allocation, the heterogeneous physical parameters participating in the scheduling are dimensionless. During dimensionless processing, two mapping methods are distinguished based on the correlation between the physical parameter values ​​and performance: For physical parameters whose larger values ​​indicate better performance, a forward mapping is used, which involves subtracting the preset minimum threshold of the parameter under system constraints from the actual extracted value, and then dividing by the difference between the preset maximum threshold and the minimum threshold, so that the dimensionless parameter after mapping increases as the original value increases. For physical parameters whose smaller values ​​indicate better performance, an inverse mapping is used. This involves subtracting the actual extracted value from the preset maximum threshold of the parameter under system constraints, and then dividing by the difference between the preset maximum and minimum thresholds. This ensures that the dimensionless parameter after mapping also increases with performance optimization.

Images