A method and system for aircraft attribute identification based on multi-source data fusion

Abstract

The present application provides an aircraft attribute discrimination method and system based on multi-source data fusion. The method of the present application includes: acquiring multi-source aviation data and preprocessing it to obtain processed multimodal aviation data; performing hierarchical alignment and association on the multimodal aviation data, and performing multi-dimensional feature extraction on the associated multimodal aviation data; performing aircraft attribute discrimination based on the extracted features using an attribute discrimination decision model to obtain an aircraft attribute discrimination result; the method further includes: performing anti-interference verification on the aircraft attribute discrimination result, and feeding the verification result back to the attribute discrimination decision model for iterative optimization; outputting the aircraft attribute discrimination result in the form of JSON structured data; wherein the aircraft attribute discrimination result includes aircraft type, mission type, purpose attribute and discrimination confidence; when the change in discrimination confidence exceeds a preset threshold, the discrimination result is triggered to be recalculated and the historical discrimination results are recorded.

Description

Technical Field

[0001] The present invention relates to the technical field of target recognition, and in particular to an aircraft attribute discrimination method and system based on multi-source data fusion. Background Art

[0002] Existing aircraft attribute recognition technologies rely on a single data source (such as ADS-B), which carries risks such as missing information and deception, and cannot verify the true attributes of an aircraft (such as model, aircraft type, and mission nature). Traditional methods do not consider the correlation analysis between ACARS, dynamic airspace restrictions (such as changes in no-fly zones), and social media events (such as accident witnessing). The spatiotemporal inconsistency of multi-source data makes fusion difficult, and there is a lack of an effective dynamic credibility assessment mechanism. For example, when existing technologies rely on a single data source (such as using only ADS-B track vectors), traditional methods suffer from large trajectory alignment errors due to differences in spatiotemporal references and conflicts in data confidence. They also fail to identify forged signals (such as false ICAO addresses), creating vulnerabilities in flight safety monitoring. Furthermore, with the surge in dynamic adjustments to ACARS and temporary no-fly zones and social media events, traditional offline rule base matching methods have a long response delay, making it difficult to meet the real-time airspace safety warning requirements. Summary of the Invention

[0003] The purpose of the present invention is to provide an aircraft attribute identification method and system based on multi-source data fusion, aiming to solve the above-mentioned problems in the prior art.

[0004] An embodiment of the present invention provides an aircraft attribute identification method based on multi-source data fusion, comprising:

[0005] Acquiring multi-source aviation data, and preprocessing the multi-source aviation data to obtain processed multimodal aviation data;

[0006] performing hierarchical alignment and association on the multimodal aviation data, and performing multi-dimensional feature extraction on the associated multimodal aviation data;

[0007] Based on the extracted features, the attribute discrimination decision model is used to discriminate the attributes of the aircraft and obtain the aircraft attribute discrimination results.

[0008] An embodiment of the present invention provides an aircraft attribute identification system based on multi-source data fusion, comprising:

[0009] A data module is used to acquire multi-source aviation data and pre-process the multi-source aviation data to obtain processed multi-modal aviation data;

[0010] an analysis module, configured to perform hierarchical alignment and association on the multimodal aviation data, and perform multi-dimensional feature extraction on the associated multimodal aviation data;

[0011] The discrimination module is used to discriminate the attributes of the aircraft using the attribute discrimination decision model based on the extracted features to obtain the aircraft attribute discrimination result.

[0012] An embodiment of the present invention further provides an electronic device comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor. When the computer program is executed by the processor, the steps of the above-mentioned method for aircraft attribute identification based on multi-source data fusion are implemented.

[0013] An embodiment of the present invention further provides a computer-readable storage medium storing a program for implementing information transmission. When the program is executed by a processor, the steps of the above-mentioned aircraft attribute identification method based on multi-source data fusion are implemented.

[0014] The following beneficial effects can be achieved by employing embodiments of the present invention: Embodiments of the present invention propose a method for intelligently identifying aircraft attributes based on multi-source data fusion. This method utilizes core technologies and strategies, including a multi-source spatiotemporal alignment algorithm, a dynamic credibility weight allocation strategy, and an airspace rule-social media joint reasoning mechanism, to achieve the goals of cross-source collaborative verification of aircraft attributes and multi-dimensional intelligent analysis of abnormal flight behavior. Compared to traditional single-data source vector matching methods, embodiments of the present invention utilize a dynamic weight allocation strategy to integrate ADS-B, ACARS, and other data, improving the decision weight of airspace rule and social media association analysis and ensuring the continued reliability of intelligent aircraft attribute identification. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] In order to more clearly illustrate one or more embodiments of this specification or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments recorded in this specification. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0016] Figure 1 This is a flow chart of an aircraft attribute identification method based on multi-source data fusion according to an embodiment of the present invention;

[0017] Figure 2 is an overall flow chart of an embodiment of the present invention;

[0018] Figure 3 Schematic diagram of the steps of the spatiotemporal correlation algorithm according to an embodiment of the present invention;

[0019] Figure 4 1 is a schematic diagram of the ADS-B data parsing steps according to an embodiment of the present invention;

[0020] Figure 5Schematic diagram of the steps for extracting semantic features of spatial description text according to an embodiment of the present invention;

[0021] Figure 6 Schematic diagram of the attention weight allocation steps in an embodiment of the present invention;

[0022] Figure 7 Schematic diagram of the weight adaptive algorithm steps of an embodiment of the present invention;

[0023] Figure 8 4 is a schematic diagram of an aircraft attribute identification system based on multi-source data fusion according to an embodiment of the present invention. DETAILED DESCRIPTION

[0024] In order to enable those skilled in the art to better understand the technical solutions in one or more embodiments of this specification, the technical solutions in one or more embodiments of this specification will be clearly and completely described below in conjunction with the drawings in one or more embodiments of this specification. Obviously, the described embodiments are only part of the embodiments of this specification, not all of the embodiments. Based on one or more embodiments of this specification, all other embodiments obtained by ordinary technicians in this field without making creative efforts should fall within the scope of protection of this document.

[0025] Method Example

[0026] According to an embodiment of the present invention, a method for distinguishing aircraft attributes based on multi-source data fusion is provided. Figure 1 FIG. 1 is a flow chart of an aircraft attribute identification method based on multi-source data fusion according to an embodiment of the present invention. Figure 1 As shown, the aircraft attribute identification method based on multi-source data fusion according to an embodiment of the present invention specifically includes:

[0027] Step S101: Acquire multi-source aviation data and pre-process the multi-source aviation data to obtain processed multi-modal aviation data; wherein the multi-source aviation data includes ADS-B data, ACARS data, dynamic airspace data, and news or social media data; specifically, including:

[0028] Performing outlier filtering, altitude logic and coordinate rationality verification, and signal quality correction on the ADS-B data;

[0029] Perform message integrity checks on the ACARS data, complete missing fields, and standardize aircraft model codes;

[0030] Verify the timeliness of the dynamic airspace data to obtain the current valid airspace data, use the Douglas-Peucker algorithm to simplify the airspace boundaries by geo-fencing, and uniformly convert the processed airspace data coordinates into UTM coordinates;

[0031] Performing text denoising and geolocation correction on the news or social media data;

[0032] Step S102, hierarchically aligning and correlating the multimodal aviation data, and extracting multi-dimensional features from the correlated multimodal aviation data, specifically includes:

[0033] Aligning and associating the multimodal aviation data according to hierarchical association rules; wherein the hierarchical association rules include spatiotemporal association layer association rules, identifier association layer association rules, and semantic association layer association rules;

[0034] The temporal and spatial association layer association rules are to perform a preliminary match between the ADS-B data and the ACARS data by temporal and spatial proximity, and to identify special purpose aircraft using airspace rules;

[0035] The identifier association layer association rule is to accurately associate aircraft by their ICAO addresses and call signs;

[0036] The semantic association layer association rules are used to infer the attributes of aircraft through feature fusion and behavior analysis of multimodal aviation data;

[0037] The correlated multimodal aviation data is subjected to source feature extraction to obtain ADS-B data features, ACARS message features, dynamic airspace features, and news or social media data features. The resulting multi-type data features are then fused across modalities to generate fused multi-dimensional features.

[0038] Step S103, based on the extracted features, uses the attribute discrimination decision model to perform attribute discrimination on the aircraft to obtain an aircraft attribute discrimination result, which specifically includes:

[0039] The fused multi-dimensional features are respectively input into a dynamic Bayesian network model and a deep neural network model to obtain the output results of the two models. If the output results of the two models are consistent, the aircraft attribute discrimination result is output; if the output results of the two models are inconsistent, the output results are integrated through a dynamic weight allocation and conflict resolution mechanism, and a global confidence assessment is performed on the integrated results to obtain a discrimination confidence. If the discrimination confidence meets a preset threshold, the corresponding aircraft attribute discrimination result is output; if the discrimination confidence does not meet the preset threshold, a manual review is triggered;

[0040] The dynamic weight allocation is to dynamically adjust the weights of different data sources according to data quality;

[0041] The conflict resolution mechanism is to overwrite the current disagreement results based on the historical judgment results;

[0042] The method further comprises:

[0043] Performing anti-interference verification on the aircraft attribute discrimination result, and feeding the verification result back to the attribute discrimination decision model for iterative optimization;

[0044] Outputting the aircraft attribute discrimination result in the form of JSON structured data; wherein the aircraft attribute discrimination result includes aircraft type, mission type, purpose attribute and discrimination confidence;

[0045] When the change in the judgment confidence exceeds a preset threshold, the judgment result is triggered to be recalculated and the historical judgment results are recorded.

[0046] The above technical solution of the embodiment of the present invention is described in detail below in conjunction with the specific situation of the aircraft attribute discrimination method based on multi-source data fusion according to the embodiment of the present invention.

[0047] The embodiment of the present invention is implemented in seven steps: multi-source data collection, data preprocessing, multi-modal data alignment and association, multi-dimensional feature extraction, attribute discrimination decision model, anti-interference verification, result output and real-time update. Figure 2 The specific contents are as follows:

[0048] Step 1: Multi-source data collection

[0049] Build distributed data interfaces: ADS-B receivers (real-time position / speed), ACARS database (flight plan / maintenance status), dynamic airspace GIS system (airspace restriction layer), social media data, etc.

[0050] ADS-B data features: ICAO address (24 bits), call sign (flight number), speed (m / s), vertical speed (m / s), latitude and longitude, altitude (pressure altitude).

[0051] ACARS message features: flight plan (take-off and landing airports, routes), flight type (passenger / cargo), airline code, mission code, etc.

[0052] Open Source Intelligence (OSINT) text features: crawl Twitter / Facebook text in real time to obtain news / social media data.

[0053] Dynamic airspace data: Airspace types are labeled with letter codes and associated with geographic boundary coordinates (GeoJSON format) and altitude restrictions (e.g., Class A airspace ≥ 18,000 feet). This supports temporary airspace applications for emergency missions (e.g., rescue and exercises), and allows for rapid approval and isolation of conflict areas through the UTM system.

[0054] Step 2: Data preprocessing

[0055] Data cleaning: Kalman filtering is used to correct abnormal ADS-B coordinates, ACARS message structured analysis, and social media text entity extraction (location, time, aircraft model).

[0056] 1) ADS-B data cleaning

[0057] Outlier filtering:

[0058] Speed ​​range: exclude speeds < 50 knots (on the ground);

[0059] Altitude logic check: If the altitude value is > 60,000 feet and is not marked as "UAV" or "High Altitude Balloon", it is marked as suspicious data;

[0060] Coordinate rationality: longitude and latitude are discarded when they are outside the range of the earth;

[0061] Signal quality correction: Dynamic weighting based on RSSI (received signal strength).

[0062] 2) ACARS data cleaning

[0063] Message integrity check: When required fields are missing (such as flight number and aircraft model code), the ICAO code is used to query the civil aviation database for completion. If the completion fails, it is marked as "incomplete data" and is only used for local reasoning.

[0064] Aircraft model code standardization: Map non-standard codes to ICAO standards. For models that cannot be mapped, call a third-party API (such as FlightAware) to query the registration number corresponding to the model.

[0065] 3) FAA dynamic airspace data cleaning

[0066] Airspace validity verification: remove invalid airspace (if the effective time > current time or the expiration time < current time);

[0067] If the time is not specified for a special purpose activity area, the default effective time is 6 hours;

[0068] Geofence simplification: Complex polygons (vertices > 100) are compressed using the Douglas-Peucker algorithm, preserving shape error <5%;

[0069] Coordinate conversion: Convert WGS84 coordinates to UTM projection to accelerate spatial calculations.

[0070] 4) Social media data cleaning

[0071] Text denoising: Filter out meaningless content (such as advertisements and emoticons) and use regular expressions to match keywords such as [#drone|#aircraft|exercise];

[0072] Remove duplicate descriptions: based on SimHash algorithm;

[0073] Geolocation correction: The address is parsed into longitude and latitude using the GeoNames API. If the parsing fails, a weighted average estimate is made based on the user's historical post locations (error < 15 km).

[0074] Step 3: Multimodal data alignment and association

[0075] Hierarchical association rule design

[0076] 1) Spatiotemporal correlation layer (coarse correlation)

[0077] Rule 1: Spatiotemporal Proximity Matching

[0078] A. Define spatiotemporal correlation threshold:

[0079] Time window: ±30 seconds (based on ADS-B)

[0080] Spatial radius: ADS-B positioning error ellipse semi-major axis × 3 (usually ≤ 2000 meters)

[0081] B. Association Algorithm: Based on the temporal synchronization and spatial proximity of multi-source aviation data (ADS-B and ACARS), the following association rules are designed. Figure 3 As shown:

[0082] Temporal and spatial proximity matching of ADS-B data and ACARS data: When the time interval between the ADS-B and ACARS time tags is greater than or equal to 30 seconds, the ADS-B data is determined to be unrelated to the ACARS data. When the time interval between the ADS-B and ACARS data is less than 30 seconds, and the distance between the longitude and latitude coordinates of the ADS-B and the corresponding longitude and latitude coordinates of the ACARS is less than 2000 meters, the ADS-B data is determined to be related to the ACARS data. Otherwise, the ADS-B data and the ACARS data are determined to be unrelated.

[0083] Rule 2: Airspace rule filtering

[0084] By combining airspace usage information and data missing information, aircraft that may be involved in sensitive or special missions (such as military, police, scientific research, etc.) can be quickly identified and their attributes marked.

[0085] For example, if the target is located in an FAA special-use restricted area and the ACARS message is missing, the weighted trigger special-use attribute flag is used.

[0086] 2) Identifier association layer (precise association)

[0087] Rule 3: ICAO Address Consistency

[0088] The ICAO 24-bit address of ADS-B is reverse mapped to the ACARS registration number to verify the uniqueness of the airframe.

[0089] Rule 4: Call Sign Logic Check

[0090] Civil aviation call sign format verification (e.g. Air China CCA123 matches the ICAO airline code CCA);

[0091] Special-purpose call sign anomaly detection (e.g., the pattern "GOLD12" does not conform to civil aviation coding standards).

[0092] 3) Semantic association layer (attribute reasoning)

[0093] Rule 5: Multimodal feature fusion, as shown in Table 1

[0094] Fusion formula: Special model confidence = Σ(feature weight × attribute probability);

[0095] If the special purpose confidence level is > 0.7, it is considered a special aircraft.

[0096] Table 1 Multimodal feature fusion

[0097]

[0098] Rule 6: Dynamic Behavior Pattern Analysis

[0099] Abnormal trajectory detection:

[0100] Typical civil aviation behavior: continuous flight along waypoints with gentle altitude changes.

[0101] Typical behaviors for special purposes: rapid climb, sharp turn, and low-altitude penetration.

[0102] Trajectory intention is predicted based on the LSTM model and semantically aligned with FAA airspace usage.

[0103] Step 4: Multi-dimensional feature extraction

[0104] 1) ADS-B data feature extraction

[0105] (1) Temporal characteristics of flight behavior

[0106] Numerical feature extraction:

[0107] Dynamic characteristics: altitude change rate (Δh / Δt), speed fluctuation standard deviation (σ_v), heading angle dispersion (cosine similarity);

[0108] Statistical features: maximum climb rate, average ground speed, and track curvature radius within a 10-second window.

[0109] Category feature encoding: Model Hex encoding parsing process, such as Figure 4 As shown:

[0110] Aircraft transmit messages containing a 24-bit ICAO address (Hex code) via ADS-B radio. This code uniquely identifies an aircraft and is associated with information such as the aircraft model and registration number. The format is 6 hexadecimal characters (e.g., 780AEF), and the stored content includes metadata such as the country code and aircraft category, but does not directly include the aircraft model, requiring database association. ADS-B parsing first queries the system's local database and returns the result directly if the query is found. If the local database cannot be queried, an authoritative database is called through an external interface. If a match is found, the aircraft registration number and model mapped to the queried Hex code are updated in the local database and the result is returned. For example, the Hex code (780AEF) can be associated with the registration number (B-1234), aircraft model (Airbus A320-214), airline (China Eastern Airlines), etc.

[0111] One-hot encoding: Generates a model category vector (dimension = 32 categories).

[0112] (2) Abnormal pattern detection

[0113] Special flight characteristics marking:

[0114] Feature 1: Speed ​​> 600 knots and altitude > 10,000 meters (typical cruising characteristics of special aircraft);

[0115] Feature 2: The track suddenly turns at a right angle (tactical evasive behavior);

[0116] Signature 3: ADS-B signal is intermittently turned off (covert mode).

[0117] 2) ACARS message feature extraction

[0118] (1) Structured field analysis

[0119] A. Key field extraction:

[0120] Remaining fuel → Discretized into three levels: [low, medium, high] (a passenger plane with low fuel may be on a short-haul flight);

[0121] Manifest weight → Load type classification (passenger aircraft: 1.5-2.5 tons / passenger, cargo aircraft: >10 tons);

[0122] Engine parameters → thrust-to-weight ratio estimation (thrust-to-weight ratio of special models > 1.2).

[0123] B. Flight number semantic segmentation:

[0124] Split the airline code and flight number (e.g. "CA1501" → "CA", "1501").

[0125] (2) Text semantic association

[0126] Flight mission reasoning:

[0127] Keyword matching: Fields such as "MEDEVAC" (medical rescue) and "CARGO" (freight) in the message are directly mapped to the task type;

[0128] Context analysis: If the message contains "NO PAX" (no passengers), it is marked as a cargo plane.

[0129] 3) FAA dynamic airspace feature extraction

[0130] (1) Semantic analysis of airspace rules

[0131] Airspace usage vectorization:

[0132] Use the BERT model to extract the semantic features of the spatial domain description text, such as Figure 5 As shown;

[0133] The rule constraint matrix is ​​shown in Table 2:

[0134] Table 2 Rule constraint matrix

[0135]

[0136] (2) Dynamic topology analysis of airspace

[0137] Airspace polygon spatial characteristics:

[0138] Calculate the distance between the aircraft's current position and the boundary of the no-fly zone (GIS spatial analysis);

[0139] If the aircraft's trajectory crosses a temporary restricted region (TFR), the probability of triggering a special mission increases by +0.25.

[0140] 4) Social Media Data Feature Extraction

[0141] Social media data mainly involves text, images, and location data.

[0142] (1) Visual feature extraction

[0143] Aircraft component inspection:

[0144] Use the YOLOv5 model to detect key components in the image: identify key components and identify the model;

[0145] Model visual matching:

[0146] Feature comparison: The image's ResNet-50 features are compared with a model database (such as the Janes Yearbook) using cosine similarity. If a match is found for "Global X Drone," it is directly marked as a special-purpose drone.

[0147] (2) Text feature extraction

[0148] The BERT model is used to parse social media text data. First, the social media data is cleaned, and irrelevant content (such as advertisements) is filtered out using NLP tools to extract aviation-related posts. BERT is then used to extract aviation event keywords (such as "forced landing" and "drone interference") and generate semantic vectors.

[0149] (3) Spatiotemporal text analysis

[0150] User review keyword extraction:

[0151] Regular expressions match special related words (such as "XX aircraft", "exercise", "low-altitude flight");

[0152] Sentiment Analysis: If the description contains "huge roar", the probability of special use is increased.

[0153] 5) Cross-modal feature fusion

[0154] (1) Feature dimensionality reduction and alignment

[0155] Unified embedding space mapping: Use t-SNE to project different modality features into a 256-dimensional common space.

[0156] (2) Attention weight allocation

[0157] Modal importance score: Calculate the information entropy of each modal feature and normalize it into weight (the higher the entropy, the lower the weight), as follows: Figure 6 shown.

[0158] Step 5: Attribute discrimination decision model

[0159] 1) Primary discrimination based on dynamic Bayesian network

[0160] A. Node definition:

[0161] Observation node: {speed pattern, altitude fluctuation, aircraft model code, image features...}

[0162] Hidden layer nodes: {special purpose probability, civil aviation probability, mission type}

[0163] B. Inference formula:

[0164]

[0165] in, is the normalization factor.

[0166] 2) Fine-grained classification based on deep neural networks

[0167] A. Model structure:

[0168] Input layer: multimodal feature concatenation (dimension = 512+128+64);

[0169] Hidden layer: 3 layers of GRU (processing time series features) + 2 layers of full connection (ReLU activation);

[0170] Output layer: Softmax classification (category: {passenger plane, cargo plane, special model 1, special model 2, drone}).

[0171] B. Loss Function:

[0172] (Bayesian network and DNN output consistency constraints)

[0173] in, =0.7, 0.3.

[0174] 3) Dynamic weight allocation and conflict resolution

[0175] A. Weight adaptive algorithm, such as Figure 7 As shown in the figure, the adaptive weighting scheme adjusts the weights of different data sources based on data quality. For example, when the quality of ADS-B data falls below 0.5, its weight is reduced, while the weights of ACARS and social data are increased; otherwise, the default weights are used. Dynamically adjusting the fusion weights as the quality of different data sources changes improves the robustness and adaptability of the system.

[0176] B. Conflict resolution rules:

[0177] If Bayesian network result != DNN result:

[0178] Activate the meta-learning module to retrieve historical similar cases (k-NN algorithm);

[0179] If the proportion of special purposes in historical decisions is greater than 70%, the current result will be overwritten.

[0180] 4) Global confidence evaluation and output

[0181] Confidence calculation model:

[0182] Confidence = 0.6 × classification probability + 0.2 × data integrity + 0.2 × historical consistency

[0183] Output condition: Confidence level > 0.8 (otherwise manual review is triggered).

[0184] Step 6: Anti-interference verification

[0185] 1) Interference scene construction and injection

[0186] A. Sensor noise simulation: Gaussian noise (speed ±15%, altitude ±200 meters) is injected into ADS-B / ACARS data to simulate signal drift;

[0187] B. Fake data generation: Modifying aircraft Hex codes (e.g., disguising a specific aircraft model as a civil aviation B738), forging fake social media images (AI-generated special aircraft with civil aviation livery);

[0188] C. Rule conflict testing: Dynamically modify FAA airspace status (such as temporarily opening a special restricted area for civilian use) to verify logical consistency under airspace rule conflicts.

[0189] 2) Multi-source data consistency verification

[0190] A. Spatiotemporal alignment verification:

[0191] A sliding time window (window size = 30 seconds) is used to perform spatiotemporal alignment of multi-source data. If the deviation between the ADS-B trajectory and the GPS coordinates of the social media image exceeds 2000 meters, a low confidence flag is triggered.

[0192] B. Cross-modal confidence evaluation:

[0193] Define the confidence weight of each data source (e.g., ADS-B = 0.7, FAA regulations = 0.8, social media = 0.5). When the single-source features conflict with the fusion results, initiate a reweighting strategy (reducing the weight of the conflicting source by 30%).

[0194] 3) Fault-tolerant feature fusion mechanism

[0195] A. Redundant feature screening:

[0196] Redundant information such as ADS-B speed patterns, ACARS maintenance codes, and radar signatures of social media images are voted on, retaining only evidence supported by at least two independent sources.

[0197] B. Abnormal feature isolation:

[0198] The Isolation Forest algorithm is used to detect abnormal features (such as the sudden appearance of non-standard flight levels), isolate them, and then recalculate the fusion results.

[0199] 4) Dynamic weight adjustment algorithm

[0200] A weight distribution model (LSTM+Attention) is trained based on historical data to evaluate the reliability of each data source in real time.

[0201] Step 7: Result output and real-time update

[0202] The judgment results are output to relevant personnel or systems for applications such as flight monitoring, safety management, and flight scheduling.

[0203] A. Output format:

[0204] Generates JSON structured data (including aircraft_type aircraft type, mission_type mission type, is_military purpose attribute, and confidence_score confidence field).

[0205] B. Dynamic update mechanism:

[0206] When new data causes the confidence change to exceed a threshold (Δ>0.15), the result is recalculated and the version history is recorded.

[0207] The output json format example is as follows:

[0208] {

[0209] "Attribute identification result": {

[0210] "Type": "XX machine (XX model)";

[0211] "Mission": "Training Flight";

[0212] "Purpose": "Special Purpose";

[0213] Confidence: 0.91

[0214] };

[0215] "Chain of evidence": ["ADS-B speed pattern matching", "XX pylon detected in social media images"]

[0216] }

[0217] Among them, the evidence chain is an optional field, which lists the multimodal data evidence supporting the judgment result.

[0218] In summary, the present invention significantly reduces multi-source data fusion errors and shortens the response time for identifying violations by introducing advanced technologies such as a multimodal spatiotemporal alignment algorithm (based on interpolation compensation for geographic grids), a dynamic credibility weighting strategy (combining Bayesian networks with historical data), and an airspace rule-social media joint inference mechanism. This combination of technologies enables the present invention to maintain high robustness even in complex scenarios. For example, when an ACARS flight plan is temporarily changed or an ADS-B signal is lost, the real-time association of social media event spatiotemporal clustering with airspace GIS layers enables rapid inference of abnormal flight intent, improving accuracy. Therefore, the present invention not only overcomes the core technical bottlenecks of data silos and delayed dynamic rule responses in traditional methods, but also further reduces the misidentification rate of special and civil aircraft through cross-modal feature collaborative analysis (such as signal fingerprint verification and social media keyword sentiment monitoring). This method boasts high monitoring accuracy and strong real-time performance, making it widely applicable in fields such as air traffic control, counterterrorism reconnaissance, and emergency rescue.

[0219] System Example

[0220] According to an embodiment of the present invention, an aircraft attribute identification system based on multi-source data fusion is provided. Figure 8 FIG. 1 is a schematic diagram of an aircraft attribute identification system based on multi-source data fusion according to an embodiment of the present invention. Figure 8 As shown, the aircraft attribute identification system based on multi-source data fusion according to an embodiment of the present invention specifically includes:

[0221] The data module 80 is used to acquire multi-source aviation data and pre-process the multi-source aviation data to obtain processed multi-modal aviation data;

[0222] An analysis module 82 is configured to perform hierarchical alignment and association on the multimodal aviation data, and perform multi-dimensional feature extraction on the associated multimodal aviation data;

[0223] A discrimination module 84 is configured to discriminate the attributes of the aircraft using an attribute discrimination decision model based on the extracted features to obtain an aircraft attribute discrimination result;

[0224] The system further comprises:

[0225] A verification module, configured to perform anti-interference verification on the aircraft attribute discrimination result and feed the verification result back to the attribute discrimination decision model for iterative optimization;

[0226] An output module, configured to output the aircraft attribute discrimination result in the form of JSON structured data; wherein the aircraft attribute discrimination result includes aircraft type, mission type, purpose attribute, and discrimination confidence;

[0227] The dynamic update module is used to trigger the recalculation of the discrimination result and record the historical discrimination results when the change of the discrimination confidence exceeds a preset threshold.

[0228] The embodiment of the present invention is a system embodiment corresponding to the above-mentioned method embodiment. The specific operations of each module can be understood by referring to the description of the method embodiment, which will not be repeated here.

[0229] In summary, compared with the prior art, the embodiments of the present invention have the following advantages, features and positive effects:

[0230] 1. Improved anti-single-source spoofing capabilities: Through cross-modal comparison of ADS-B signal fingerprint verification and ACARS flight plans, this overcomes the vulnerability of traditional vector matching, which relies on a single signal source (such as using only ADS-B track vectors) and is vulnerable to signal forgery attacks, thereby improving the accuracy of spoofing identification.

[0231] 2. Enhanced spatiotemporal dynamic adaptability: Geographically gridded spatiotemporal interpolation compensation technology is used to address the problem of traditional Euclidean space vector matching being sensitive to spatiotemporal offsets (such as trajectory misalignment caused by ACARS message delays), thereby reducing spatiotemporal alignment errors.

[0232] 3. Multi-dimensional Confidence Fusion Innovation: A dynamic weight allocation model based on historical performance replaces the traditional static cosine similarity calculation. This allows for probabilistic reasoning of conflicting data (such as conflicts between social media sightings and airspace filing information), improving the recognition accuracy of aircraft classification tasks.

[0233] 4. Real-time rule response optimization: By integrating airspace GIS layers with spatiotemporal clustering of social media events, we can achieve second-level early warning of no-fly zone violations, reducing response latency from minutes to seconds compared to traditional offline vector library retrieval methods.

[0234] 5. Accuracy and Reliability: Through core technologies and strategies such as a multi-source spatiotemporal alignment algorithm, a dynamic credibility weight allocation strategy, and an airspace rule-social media joint reasoning mechanism, the embodiments of the present invention create an intelligent aircraft attribute identification method based on multi-source data fusion, achieving the goals of cross-source collaborative verification of aircraft attributes and multi-dimensional intelligent analysis of abnormal flight behavior.

[0235] Device Example 1

[0236] An embodiment of the present invention provides an electronic device, comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the computer program implements the steps described in the method embodiment when executed by the processor.

[0237] Device Example 2

[0238] An embodiment of the present invention provides a computer-readable storage medium, on which a program for implementing information transmission is stored. When the program is executed by a processor, the steps described in the method embodiment are implemented.

[0239] The computer-readable storage medium in this embodiment includes, but is not limited to, ROM, RAM, magnetic disk, or optical disk.

[0240] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the above embodiments, or replace some or all of the technical features therein with equivalents. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for aircraft attribute identification based on multi-source data fusion, characterized by include: Acquiring multi-source aviation data, and preprocessing the multi-source aviation data to obtain processed multimodal aviation data; The multimodal aviation data is hierarchically aligned and associated, and multi-dimensional features are extracted from the associated multimodal aviation data, specifically including: The multimodal aviation data is aligned and associated according to hierarchical association rules, wherein the hierarchical association rules include spatiotemporal association layer association rules, identifier association layer association rules, and semantic association layer association rules; source-specific feature extraction is performed on the associated multimodal aviation data to obtain ADS-B data features, ACARS message features, dynamic airspace features, and news or social media data features, respectively; cross-modal feature fusion is performed on the obtained multi-type data features to generate fused multi-dimensional features; Based on the extracted features, the attribute discrimination decision model is used to discriminate the attributes of the aircraft and obtain the aircraft attribute discrimination results, which include: The fused multi-dimensional features are respectively input into a dynamic Bayesian network model and a deep neural network model to obtain the output results of the two models. If the output results of the two models are consistent, the aircraft attribute discrimination result is output; if the output results of the two models are inconsistent, the output results are integrated through a dynamic weight allocation and conflict resolution mechanism, and a global confidence assessment is performed on the integrated results to obtain a discrimination confidence. If the discrimination confidence meets a preset threshold, the corresponding aircraft attribute discrimination result is output; if the discrimination confidence does not meet the preset threshold, a manual review is triggered; The dynamic weight allocation is to dynamically adjust the weights of different data sources according to data quality; the conflict resolution mechanism is to overwrite the current disagreement results based on historical judgment results; The aircraft attribute discrimination result includes aircraft type, mission type, purpose attribute and discrimination confidence.

2. The method according to claim 1, characterized in that The method further comprises: Performing anti-interference verification on the aircraft attribute discrimination result, and feeding the verification result back to the attribute discrimination decision model for iterative optimization; Outputting the aircraft attribute identification result in the form of JSON structured data; When the change in the judgment confidence exceeds a preset threshold, the judgment result is triggered to be recalculated and the historical judgment results are recorded.

3. The method according to claim 1, characterized in that The multi-source aviation data includes ADS-B data, ACARS data, dynamic airspace data, and news or social media data.

4. The method according to claim 3, characterized in that Preprocessing the multi-source aviation data specifically includes: Performing outlier filtering, altitude logic and coordinate rationality verification, and signal quality correction on the ADS-B data; Perform message integrity checks on the ACARS data, complete missing fields, and standardize aircraft model codes; Verify the timeliness of the dynamic airspace data to obtain the current valid airspace data, use the Douglas-Peucker algorithm to simplify the airspace boundaries by geo-fencing, and uniformly convert the processed airspace data coordinates into UTM coordinates; Perform text denoising and geolocation correction on the news or social media data.

5. The method according to claim 3, characterized in that The temporal and spatial association layer association rules are to perform a preliminary match between the ADS-B data and the ACARS data by temporal and spatial proximity, and to identify special purpose aircraft using airspace rules; The identifier association layer association rule is to accurately associate aircraft by their ICAO addresses and call signs; The semantic association layer association rules are used to infer the attributes of aircraft through feature fusion and behavior analysis of multimodal aviation data.

6. An aircraft attribute identification system based on multi-source data fusion, characterized by include: A data module is used to acquire multi-source aviation data and pre-process the multi-source aviation data to obtain processed multi-modal aviation data; The analysis module is used to perform hierarchical alignment and association on the multimodal aviation data and perform multi-dimensional feature extraction on the associated multimodal aviation data, specifically for: The multimodal aviation data is aligned and associated according to hierarchical association rules, wherein the hierarchical association rules include spatiotemporal association layer association rules, identifier association layer association rules, and semantic association layer association rules; source-specific feature extraction is performed on the associated multimodal aviation data to obtain ADS-B data features, ACARS message features, dynamic airspace features, and news or social media data features, respectively; cross-modal feature fusion is performed on the obtained multi-type data features to generate fused multi-dimensional features; The discrimination module is used to discriminate the attributes of the aircraft using the attribute discrimination decision model based on the extracted features and obtain the aircraft attribute discrimination results. It is specifically used to: The fused multi-dimensional features are respectively input into a dynamic Bayesian network model and a deep neural network model to obtain the output results of the two models. If the output results of the two models are consistent, the aircraft attribute discrimination result is output; if the output results of the two models are inconsistent, the output results are integrated through a dynamic weight allocation and conflict resolution mechanism, and a global confidence assessment is performed on the integrated results to obtain a discrimination confidence. If the discrimination confidence meets a preset threshold, the corresponding aircraft attribute discrimination result is output; if the discrimination confidence does not meet the preset threshold, a manual review is triggered; The dynamic weight allocation is to dynamically adjust the weights of different data sources according to data quality; the conflict resolution mechanism is to overwrite the current disagreement results based on historical judgment results; The aircraft attribute discrimination result includes aircraft type, mission type, purpose attribute and discrimination confidence.

7. The system according to claim 6, characterized in that The system further comprises: A verification module, configured to perform anti-interference verification on the aircraft attribute discrimination result and feed the verification result back to the attribute discrimination decision model for iterative optimization; An output module, configured to output the aircraft attribute identification result in the form of JSON structured data; The dynamic update module is used to trigger the recalculation of the discrimination result and record the historical discrimination results when the change of the discrimination confidence exceeds a preset threshold.

8. An electronic device, characterized in that: include: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the computer program is executed by the processor, the steps of the aircraft attribute identification method based on multi-source data fusion according to any one of claims 1 to 5 are implemented.

9. A computer-readable storage medium, characterized in that The computer-readable storage medium stores an information transmission implementation program, and when the program is executed by a processor, the steps of the aircraft attribute identification method based on multi-source data fusion according to any one of claims 1 to 5 are implemented.

Images