Search

CN-122020558-A - Intelligent fusion method for space-sky-earth multi-source heterogeneous data

CN122020558ACN 122020558 ACN122020558 ACN 122020558ACN-122020558-A

Abstract

The invention discloses an aerospace multisource heterogeneous data intelligent fusion method, which relates to the technical field of data processing and comprises the following steps of multi-mode data cleaning and space-time alignment, dynamic knowledge graph construction and semantic association, quality-driven self-adaptive fusion decision, increment optimization and interpretable decision output; the invention adopts wave band self-adaptive filtering and sliding window outlier detection, combines a space-time Kriging interpolation method to unify grid coordinates, reduces multi-source data alignment errors, builds dynamic knowledge graph semantic disambiguation based on a space-time convolution network, eliminates most of cross-source semantic conflicts through a high-confidence verification and manual auditing dual-channel mechanism, dynamically triggers data level, feature level and decision level fusion by building a coverage, definition and continuity quantization index system, can ensure fusion precision under different environments, and is updated in increment based on a flow-type causal tracing engine by Kalman filtering and combined with a causal discovery algorithm to trace back decision paths, thereby reducing response delay.

Inventors

  • WANG WEI

Assignees

  • 江苏城乡建设职业学院

Dates

Publication Date
20260512
Application Date
20260206

Claims (8)

  1. 1. An aerospace-earth multi-source heterogeneous data intelligent fusion method is characterized by comprising the following steps of: the method comprises the steps of firstly, performing cross-modal noise filtering and space-time reference unification on acquired space-time-ground multi-source data to generate a standardized space-time data stream; Step two, based on the standardized space-time data flow, constructing a dynamic knowledge graph of space-space multi-source data, eliminating semantic conflict through cross-modal alignment, and generating a multi-modal feature set with consistent semantics; step three, dynamically selecting a data level, feature level or decision level fusion strategy according to the semantic association degree and the data quality evaluation result to generate a primary data fusion result; and step four, carrying out streaming update on the primary data fusion result to realize incremental optimization, generating an interpretable decision path by combining a causal reasoning model, realizing causal tracing, and outputting final fusion data and a tracing report.
  2. 2. The method for intelligently fusing the space-sky multisource heterogeneous data is characterized by comprising the specific steps of performing cross-modal noise filtering and space-time reference unification, namely performing noise correction on satellite remote sensing data and aerial images by adopting a band adaptive filtering algorithm based on an atmospheric optical thickness attenuation coefficient and a historical data regression result, performing outlier detection on ground sensor data by adopting a sliding window, removing outliers exceeding a preset standard deviation range, and aligning satellite, aerial and ground data of different coordinate systems to a unified space-time grid by a space-time Kriging interpolation method.
  3. 3. The method for intelligently fusing the space-sky multisource heterogeneous data according to claim 1, wherein in the second step, the specific steps of constructing a dynamic knowledge graph are that the ground feature type attribute of a satellite remote sensing data pixel block, the object detection frame boundary of an aerial image and the space-time measurement value of a ground sensor are defined, a space-time graph convolution network is used, the satellite remote sensing data pixel block and a ground sensor node are related according to space proximity and time synchronism, when semantic conflict is detected, a high confidence data source is preferentially adopted for verification, and if the conflict is continuous, a manual auditing process is triggered.
  4. 4. The method for intelligent fusion of space-sky and earth multi-source heterogeneous data according to claim 1, wherein in the third step, dynamic selection rules of fusion strategies are that a multi-dimensional quality evaluation model is built, space coverage of satellite remote sensing data S cover , definition of aerial images R res and time continuity of a ground sensor C cont are calculated, if S cover is more than 0.9 and R res is more than 200dpi, data level fusion is started, if C cont is less than 0.6 or abnormal values of sensors exist, switching to feature level fusion is carried out, and if severe contradiction of cross-source data is detected, decision level fusion is triggered and conflict areas are marked.
  5. 5. The method of claim 4, wherein the satellite remote sensing data has a spatial coverage S cover =effective pixel number/total pixel number, the ground sensor has a time continuity C cont =1-missing period number/total period number, and the aerial image has a resolution R res = (pixel size x fly height)/camera focal length.
  6. 6. The method for intelligently fusing the space-sky multi-source heterogeneous data according to claim 1, wherein in the fourth step, a stream processing engine ‌ is adopted to perform incremental fusion on satellite, aviation and ground data which are transmitted in real time, and the method specifically comprises the following steps: Receiving real-time data flow every 10 seconds, and updating a fusion result through a Kalman filtering algorithm; and (3) performing conflict detection on the updated result, and triggering the fusion strategy of the third step to reevaluate if the difference between the updated result and the historical data exceeds a threshold value.
  7. 7. The method for intelligently fusing the space-sky and ground multi-source heterogeneous data according to claim 1, wherein the implementation of the causal tracing comprises the steps of constructing a causal graph among multi-source data variables based on a causal discovery algorithm ‌ and tracing key influencing factors of a decision result.
  8. 8. The method for intelligently fusing the space-sky and ground multi-source heterogeneous data according to claim 7, wherein the optimization method of the causal discovery algorithm comprises the steps of introducing domain knowledge constraint, defining causal relation search space, carrying out uncertainty quantification on causal side weights, and eliminating weak causal relations with confidence coefficient lower than 0.7.

Description

Intelligent fusion method for space-sky-earth multi-source heterogeneous data Technical Field The invention relates to the technical field of data processing, in particular to an aerospace multi-source heterogeneous data intelligent fusion method. Background The space-earth multi-source heterogeneous data refer to data from different platforms such as satellite remote sensing, aerial remote sensing, ground monitoring and the like, have the characteristics of multiple sources, isomerism and mass, are different in data sources and various in types, comprise satellite images, aerial photography, ground measurement and the like, are integrated through a data fusion technology to improve the accuracy and reliability of monitoring, are widely applied to multiple fields such as environment monitoring, disaster early warning, agricultural management and urban planning, and are difficult to directly use due to the fact that the space-earth multi-source heterogeneous data are complex in sources and various in types, and the data fusion technology plays a key role in the process. The data fusion technology is a technology for extracting effective information and forming unified expression by integrating multi-source heterogeneous data from different sources, different formats and different structures, and has the core targets of eliminating redundancy and conflict through multi-level processing (such as data level, feature level and decision level fusion) of the data to generate more comprehensive, reliable and high-value information. The existing multi-source heterogeneous data fusion method is single in flow, has the problems that data distortion is caused by incomplete noise filtering, fusion dislocation is caused by non-uniform space-time references, a fixed fusion strategy is difficult to adapt to data quality fluctuation, and traceability is lacking in a decision process, so that the fusion precision loss of space-to-ground multi-source heterogeneous data is obvious, and therefore, the invention provides an intelligent fusion method for space-to-ground multi-source heterogeneous data to solve the problems in the prior art. Disclosure of Invention Aiming at the problems, the invention aims to provide an aerospace multi-source heterogeneous data intelligent fusion method which solves the problems of large noise interference, non-uniform space-time reference, strategy rigidification and difficult decision tracing of the existing multi-source heterogeneous data fusion method. In order to achieve the purpose of the invention, the invention is realized by the following technical scheme that the method for intelligently fusing the space-sky multi-source heterogeneous data comprises the following steps: the method comprises the steps of firstly, performing cross-modal noise filtering and space-time reference unification on acquired space-time-ground multi-source data to generate a standardized space-time data stream; Step two, based on the standardized space-time data flow, constructing a dynamic knowledge graph of space-space multi-source data, eliminating semantic conflict through cross-modal alignment, and generating a multi-modal feature set with consistent semantics; step three, dynamically selecting a data level, feature level or decision level fusion strategy according to the semantic association degree and the data quality evaluation result to generate a primary data fusion result; and step four, carrying out streaming update on the primary data fusion result to realize incremental optimization, generating an interpretable decision path by combining a causal reasoning model, realizing causal tracing, and outputting final fusion data and a tracing report. The method is further improved in the first step, the specific steps of cross-modal noise filtering and space-time reference unification are carried out, namely, a band self-adaptive filtering algorithm is adopted for satellite remote sensing data and aerial images, noise correction is carried out based on an atmospheric optical thickness attenuation coefficient and a historical data regression result, outliers exceeding a preset standard deviation range are removed for ground sensor data through sliding window outlier detection, and satellite, aerial and ground data of different coordinate systems are aligned to a unified space-time grid through a space-time Kriging interpolation method. The method is further improved in that in the second step, the specific steps of constructing the dynamic knowledge graph are that ground object category attributes of satellite remote sensing data pixel blocks, object detection frame boundaries of aerial images and space-time measurement values of a ground sensor are defined, a space-time graph convolution network is used, the satellite remote sensing data pixel blocks and ground sensor nodes are associated according to space adjacency and time synchronism, when semantic conflict is detected, a high-confidence data source is p