An airport low-altitude flight track playback analysis method and system

Abstract

The application provides an airport low-altitude flight track playback analysis method and system, and relates to the technical field of air traffic management. The method comprises the following steps: acquiring low-altitude flight track data, constructing a hybrid neural network model to filter and process the low-altitude flight track data to generate intermediate flight track data; identifying key track feature data and non-key track data in the intermediate flight track data through an SVM classifier and respectively performing lossless compression and lossy compression to generate playback flight track data; designing a collaborative computing framework of CPU, GPU and FPGA, and distributing the playback flight track data to multiple track computing nodes for collaborative processing; adopting a storage architecture based on consistent hashing to partition and store the playback flight track data, and designing a multi-level index mechanism to support the retrieval of the playback flight track data, so that the track playback precision can be improved, the data processing efficiency can be optimized, and the system storage and computing cost can be reduced.

Description

Technical Field

[0001] This invention relates to the field of air traffic management technology, and in particular to a method and system for analyzing low-altitude flight tracks at airports. Background Technology

[0002] In the current air traffic management system, the number of low-altitude aircraft continues to increase, and effective management and monitoring of flight paths have become the key to ensuring flight safety.

[0003] However, the low-altitude environment is complex and changeable, and the raw track data collected by sensors is often affected by various factors, resulting in noise and errors. This directly affects the accuracy of track data analysis and playback, and reduces the efficiency of air traffic management.

[0004] Existing solutions typically rely on general digital filtering techniques and standard data compression algorithms to process track data. For example, Kalman filtering is used for data preprocessing, and Huffman coding or other lossless compression algorithms are used for data compression. However, traditional filtering techniques struggle to maintain the accuracy and integrity of track data when faced with high-noise data in complex low-altitude environments. While commonly used data compression algorithms can save storage space to some extent, high compression ratios often sacrifice data recovery quality and playback speed, impacting subsequent track analysis. Furthermore, when dealing with large volumes of data, existing technologies are slow and cannot meet practical needs; as data volume increases, data storage and management efficiency becomes inefficient.

[0005] Therefore, it is necessary to provide a method and system for analyzing low-altitude flight tracks at airports to solve the above-mentioned technical problems. Summary of the Invention

[0006] To address the aforementioned technical problems, this invention provides a method and system for analyzing low-altitude flight tracks at airports, which solves the problems of insufficient data accuracy, imbalance between compression quality and speed, and low efficiency in big data processing in existing technologies.

[0007] This invention provides a method for playing back and analyzing low-altitude flight tracks at airports, the method comprising:

[0008] Acquire low-altitude flight trajectory data, construct a hybrid neural network model to filter the low-altitude flight trajectory data, and generate intermediate flight trajectory data.

[0009] The key trajectory feature data and non-key trajectory data in the intermediate flight trajectory data are identified by the SVM classifier and then subjected to lossless compression and lossy compression respectively to generate playback flight trajectory data.

[0010] Design a collaborative computing framework for CPU, GPU and FPGA to distribute the replayed flight trajectory data to multiple trajectory computing nodes for collaborative processing;

[0011] A consistent hash-based storage architecture is adopted to partition and store the replay flight track data, and a multi-level indexing mechanism is designed to support the retrieval of the replay flight track data.

[0012] Preferably, the step of acquiring low-altitude flight track data and constructing a hybrid neural network model to filter the low-altitude flight track data to generate intermediate flight track data specifically includes:

[0013] The low-altitude flight trajectory data is acquired, and the hybrid neural network model containing bidirectional LSTM and dilated convolution is constructed.

[0014] The hybrid neural network model is trained by acquiring historical flight track samples, and the historical flight track samples cover 12 typical low-altitude flight scenarios;

[0015] After performing sliding window preprocessing on the low-altitude flight track data, the time and spatial features in the low-altitude flight track data are extracted through the hybrid neural network model to generate fused flight track features.

[0016] By dynamically adjusting the model's frequency band filtering coefficients through the residual learning structure of the hybrid neural network model, the impulse noise and periodic interference in the fused flight track features are eliminated, thereby generating the intermediate flight track data.

[0017] Preferably, the hybrid neural network model includes:

[0018] A bidirectional LSTM layer is used to extract the temporal dependency features from the low-altitude flight track data, i.e., the time features, and the time features include velocity and acceleration;

[0019] A dilated convolutional layer is used to capture multi-scale coordinate features, i.e., spatial features, in the low-altitude flight trajectory data, and the spatial features include longitude, latitude, and altitude.

[0020] An attention mechanism layer is used to weightedly fuse the temporal and spatial features to generate spatiotemporal flight trajectory features;

[0021] The residual connection structure is used to connect and fuse the low-altitude flight track data with the spatiotemporal flight track features to generate the fused flight track features;

[0022] The training process of the hybrid neural network model is as follows:

[0023] Obtain historical flight track samples that include different weather conditions, geographical environments, and aircraft types;

[0024] The normal flight segments, flight change points, and noise areas in the historical flight track samples are marked;

[0025] The labeled historical flight track samples are divided into a track training set, a track verification set, and a track test set.

[0026] The hybrid neural network model is trained based on the aforementioned trajectory training set, and the training parameters are updated using the cross-entropy loss function and the stochastic gradient descent optimization algorithm.

[0027] The trained hybrid neural network model is validated based on the track validation set, and the model parameters of the hybrid neural network model are adjusted.

[0028] The adjusted hybrid neural network model is tested based on the aforementioned track test set to evaluate its filtering effect and generalization ability.

[0029] Preferably, the dynamic adjustment process of the model frequency band filtering coefficients is as follows:

[0030] Real-time monitoring of the noise intensity and spectral characteristics of the fused flight trajectory features;

[0031] When the detected change in noise intensity exceeds a preset change threshold, the adaptive coefficient update mechanism of the hybrid neural network model is triggered;

[0032] The frequency band filtering coefficients are calculated and adjusted using the residual learning structure of the hybrid neural network model.

[0033] Preferably, the step of identifying key trajectory feature data and non-key trajectory data in the intermediate flight trajectory data using an SVM classifier and performing lossless compression and lossy compression respectively to generate replay flight trajectory data specifically includes:

[0034] The SVM classifier identifies the key trajectory feature data and the non-key trajectory data in the intermediate flight trajectory data. The key trajectory feature data includes flight trajectory turning points and trajectory altitude change points. The non-key trajectory data is the flight cruise segment.

[0035] The key trajectory feature data is losslessly compressed, and the non-key trajectory feature data is lossily compressed in combination with the aircraft kinematics prediction model to generate the replay flight trajectory data.

[0036] The replay flight track data is deduplicated by hashing using spatiotemporal correlation.

[0037] Preferably, the identification process of the key trajectory feature data is as follows:

[0038] Extract the kinematic features from the intermediate flight track data;

[0039] The SVM classifier is used to classify the kinematic features, identify the turning points of the flight path and the points of change in flight path altitude, and generate the key flight path feature data;

[0040] Spatiotemporal correlation verification is performed on the identified key trajectory feature data to eliminate false change feature points in the key trajectory feature data;

[0041] The lossy compression of the non-critical trajectory feature data by combining the aircraft kinematics prediction model specifically includes:

[0042] A kinematic prediction model for the aircraft is established based on Newton's laws of motion, and kinematic prediction is performed on the flight cruise segment to generate a predicted flight trajectory.

[0043] The actual flight trajectory is obtained, and the trajectory deviation between the actual flight trajectory and the predicted flight trajectory is calculated. Flight waypoints whose trajectory deviation exceeds a preset trajectory deviation threshold are then selected.

[0044] The flight waypoints are subjected to lossy compression encoding to complete the lossy compression operation of the non-critical waypoint feature data.

[0045] Preferably, the design of the collaborative computing framework of CPU, GPU, and FPGA distributes the replayed flight trajectory data to multiple trajectory computing nodes for collaborative processing, specifically including:

[0046] The design incorporates a collaborative computing framework for the CPU, GPU, and FPGA. A data sharding module divides the replay flight trajectory data into multiple trajectory data sub-blocks according to a three-dimensional spatial grid and assigns them to the corresponding trajectory computing nodes for collaborative processing. The dimensions of the three-dimensional spatial grid include longitude, latitude, and altitude.

[0047] Specifically, the CPU is used for task scheduling and global parameter updates, the GPU is used to perform matrix operations for filtering, and the FPGA is used to accelerate bitwise operations for lossless compression.

[0048] Preferably, the step of dividing the replayed flight trajectory data into multiple trajectory data sub-blocks according to a three-dimensional spatial grid using a data sharding module and assigning them to corresponding trajectory computing nodes for collaborative processing specifically includes:

[0049] Based on the dimensions of the three-dimensional spatial grid, the data sharding module maps the replay flight trajectory data onto the three-dimensional spatial grid to generate multiple trajectory data sub-blocks;

[0050] Aggregate and index the replay flight trajectory data within each trajectory data sub-block to generate a corresponding trajectory space index table;

[0051] Multiple trajectory computing nodes are set up. Based on the load and network topology of each trajectory computing node, each trajectory data sub-block is allocated to the corresponding trajectory computing node for collaborative processing according to the trajectory space index table, and a node data transmission channel is established between the trajectory computing nodes.

[0052] Preferably, the storage architecture based on consistent hashing is used to partition and store the replay flight track data, and a multi-level indexing mechanism is designed to support the retrieval of the replay flight track data, specifically including:

[0053] Using the consistent hash-based storage architecture, the replay flight track data is partitioned into multiple hash data blocks by two dimensions: timestamp and geographical region.

[0054] For each hash data block, multiple replica data blocks are expanded, and the hash data block and the corresponding replica data block are stored in the consistent hash-based storage architecture;

[0055] The multi-level indexing mechanism is designed to support the retrieval of the replay flight track data, and the multi-level indexing mechanism includes a global index based on date, a regional index based on region code, and a feature index based on track type.

[0056] An airport low-altitude flight trajectory playback and analysis system, the system comprising:

[0057] The trajectory filtering module is used to acquire low-altitude flight trajectory data and construct a hybrid neural network model to filter the low-altitude flight trajectory data to generate intermediate flight trajectory data.

[0058] The trajectory compression module is used to identify key trajectory feature data and non-key trajectory data in the intermediate flight trajectory data through an SVM classifier and perform lossless compression and lossy compression respectively to generate playback flight trajectory data.

[0059] The collaborative processing module is used to design a collaborative computing framework for CPU, GPU and FPGA, and to distribute the replayed flight trajectory data to multiple trajectory computing nodes for collaborative processing.

[0060] The storage and retrieval module is used to partition and store the replay flight track data using a consistent hashing-based storage architecture, and to design a multi-level indexing mechanism to support the retrieval of the replay flight track data.

[0061] Compared with related technologies, the airport low-altitude flight trajectory playback analysis method and system provided by the present invention has the following beneficial effects:

[0062] This invention can acquire low-altitude flight track data, construct a hybrid neural network model to filter the low-altitude flight track data to generate intermediate flight track data; use an SVM classifier to identify key and non-key track features in the intermediate flight track data and perform lossless and lossy compression respectively to generate replay flight track data; design a collaborative computing framework of CPU, GPU, and FPGA to distribute the replay flight track data to multiple track computing nodes for collaborative processing; adopt a consistent hash-based storage architecture to partition and store the replay flight track data, and design a multi-level indexing mechanism to support the retrieval of the replay flight track data, thereby significantly improving the accuracy of track playback, optimizing data processing efficiency, and reducing system storage and computing costs.

[0063] In terms of data processing accuracy, this invention employs a deep learning-based adaptive filtering algorithm to accurately identify various types of noise in complex low-altitude environments, dynamically optimize filtering parameters, and effectively improve the purity and reliability of flight track data, laying the foundation for high-precision flight track playback. In the field of data compression, a hierarchical compression algorithm intelligently distinguishes between critical and non-critical data, significantly improving data compression efficiency while ensuring the integrity of key flight track features, thus guaranteeing the real-time and continuous nature of flight track playback. Regarding the computational processing architecture, the organic combination of heterogeneous computing and parallel processing technologies fully leverages the advantages of different computing units, significantly improving the processing speed of large-scale flight track data and effectively addressing the data processing challenges brought about by the surge in the number of low-altitude aircraft. In terms of storage management, a dual-dimensional hash partitioning and multi-level indexing mechanism not only enhances the security and reliability of data storage but also enables rapid retrieval of flight track data across different time periods, significantly improving data management efficiency and reducing system storage and computing costs. Attached Figure Description

[0064] Figure 1 A flowchart of an airport low-altitude flight track playback analysis method provided in an embodiment of the present invention;

[0065] Figure 2 A system block diagram of an airport low-altitude flight trajectory playback and analysis system provided in an embodiment of the present invention;

[0066] Figure 3 This is a schematic diagram of the hardware structure of an electronic device provided in an embodiment of the present invention. Detailed Implementation

[0067] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0068] like Figure 1 The diagram shown is a flowchart of a method for analyzing low-altitude flight tracks at airports, provided by an embodiment of the present invention. Figure 1 The execution entity of the method shown can be a software and / or hardware device. The execution entity of this application can include, but is not limited to, at least one of the following: user equipment, network equipment, etc. User equipment can include, but is not limited to, computers, smartphones, personal digital assistants (PDAs), and the aforementioned electronic devices. Network equipment can include, but is not limited to, a single network server, a server group consisting of multiple network servers, or a cloud based on cloud computing consisting of a large number of computers or network servers. Cloud computing is a type of distributed computing, consisting of a super virtual computer composed of a group of loosely coupled computers. This embodiment does not limit this. Steps S1 to S4 are detailed as follows:

[0069] S1, acquire low-altitude flight trajectory data, construct a hybrid neural network model to filter the low-altitude flight trajectory data and generate intermediate flight trajectory data;

[0070] Low-altitude flight track data refers to real-time position, speed, and heading data collected by sensors or navigation systems when an aircraft is flying in low-altitude areas, typically below 3000 meters. Hybrid neural network models are models composed of bidirectional LSTM and dilated convolutions, which combine the advantages of different network structures to achieve noise filtering and feature extraction from low-altitude flight track data. Intermediate flight track data refers to the filtered flight track data.

[0071] In practical applications, various parameter information generated by aircraft flying in low-altitude areas, including position, speed, and heading, can be collected to generate low-altitude flight track data. The accuracy and completeness of this data directly affect the effectiveness of subsequent processing. Next, a hybrid neural network model can be constructed, integrating the advantages of bidirectional LSTM and dilated convolution, to filter the original low-altitude flight track data. This removes noise interference and transforms the data into smoother, more accurate intermediate flight track data.

[0072] S2, using an SVM classifier to identify key trajectory feature data and non-key trajectory data in the intermediate flight trajectory data, and performing lossless compression and lossy compression respectively to generate playback flight trajectory data;

[0073] It is understandable that the SVM (Support Vector Machine) classifier is a supervised learning model based on statistical learning theory. It divides intermediate flight track data into key track feature data and non-key track data by constructing an optimal hyperplane. Replay flight track data refers to track data that has been compressed and can be used for later flight process reproduction, data analysis, or simulation verification.

[0074] Next, an SVM classifier can be used to identify features in the intermediate flight track data. Specifically, the SVM classifier, based on statistical learning theory, can accurately distinguish between critical and non-critical flight track features in the intermediate flight track data. Critical flight track features refer to data on sudden changes in the aircraft's state during flight. Non-critical features refer to data on the aircraft maintaining a stable flight state during flight. Different compression strategies can be adopted for these two types of data: lossless compression is used for critical flight track features to ensure data integrity and no loss for subsequent accurate analysis; lossy compression is used for non-critical flight track data, which significantly reduces the amount of data stored while allowing for some data loss, thereby generating replay flight track data and achieving efficient data simplification.

[0075] S3, Design a collaborative computing framework for CPU, GPU and FPGA, and distribute the replayed flight trajectory data to multiple trajectory computing nodes for collaborative processing;

[0076] It should be noted that the CPU, GPU, and FPGA collaborative computing framework can leverage the logical control capabilities of the CPU (Central Processing Unit), the parallel computing advantages of the GPU (Graphics Processing Unit), and the hardware acceleration features of the FPGA (Field-Programmable Gate Array) to construct a heterogeneous computing architecture, enabling efficient distributed processing of trajectory data. A trajectory computing node refers to a logical unit within the collaborative computing framework that undertakes specific trajectory data processing tasks.

[0077] In practical applications, a collaborative computing framework integrating CPU, GPU, and FPGA can be designed. The CPU handles overall logic control and task scheduling; the GPU, with its powerful parallel computing capabilities, can process large amounts of data simultaneously; and the FPGA provides flexible hardware acceleration features. Working together, these three components can distribute replay flight trajectory data across multiple trajectory computing nodes. These nodes can form a distributed processing system, and through collaborative operations, significantly improve the speed and efficiency of data processing, meeting the real-time processing needs of large-scale trajectory data.

[0078] S4. A consistent hash-based storage architecture is adopted to partition and store the replay flight track data, and a multi-level indexing mechanism is designed to support the retrieval of the replay flight track data.

[0079] Among them, the consistent hashing-based storage architecture is a distributed storage algorithm used to partition and store replay flight track data. The multi-level indexing mechanism refers to a multi-dimensional index structure with multiple dimensions, used for fast querying and locating of replay flight track data.

[0080] For data storage management, a consistent hashing-based storage architecture can be adopted. This architecture uses a unique hash mapping algorithm to rationally partition and store replay flight track data, and possesses good scalability and stability. Even with the addition or removal of storage nodes, it can minimize data migration and ensure data storage reliability. Simultaneously, a multi-level indexing mechanism can be designed to build an index structure from multiple dimensions. This multi-level indexing design enables rapid location and retrieval of target data, greatly improving data query efficiency and facilitating users to quickly obtain the required flight track information, providing strong support for applications such as flight monitoring, track analysis, and safety assessment.

[0081] In the specific implementation process, the acquisition of low-altitude flight trajectory data and the construction of a hybrid neural network model to filter the low-altitude flight trajectory data to generate intermediate flight trajectory data specifically include:

[0082] The low-altitude flight trajectory data is acquired, and the hybrid neural network model containing bidirectional LSTM and dilated convolution is constructed.

[0083] The hybrid neural network model is trained by acquiring historical flight track samples, and the historical flight track samples cover 12 typical low-altitude flight scenarios;

[0084] After performing sliding window preprocessing on the low-altitude flight track data, the time and spatial features in the low-altitude flight track data are extracted through the hybrid neural network model to generate fused flight track features.

[0085] By dynamically adjusting the model's frequency band filtering coefficients through the residual learning structure of the hybrid neural network model, the impulse noise and periodic interference in the fused flight track features are eliminated, thereby generating the intermediate flight track data.

[0086] This involves acquiring low-altitude flight track data, which records the real-time status information of aircraft flying in low-altitude areas. Subsequently, a hybrid neural network model can be constructed, integrating two core structures: a bidirectional LSTM (Long Short-Term Memory) network and dilated convolution. The bidirectional LSTM processes sequential data in both forward and backward directions, effectively capturing the temporal correlations within the flight track data; while dilated convolution expands the network's perception range of data features without significantly increasing computational cost, enabling effective extraction of spatial features.

[0087] Then, historical flight track samples covering 12 typical low-altitude flight scenarios can be obtained. These samples contain flight track data under different environments and flight conditions, such as low-altitude urban navigation, mountain terrain avoidance, and flight under complex weather conditions, and are widely representative. Using these samples as training data, the constructed hybrid neural network model is optimized and trained, enabling the model to learn and master the inherent laws and characteristic patterns of flight track data in different scenarios, thereby improving the model's ability to process and adapt to actual low-altitude flight track data.

[0088] Furthermore, a sliding window method can be used to preprocess the acquired low-altitude flight track data, dividing the continuous track data into fixed-length segments, with each segment serving as an independent processing unit. This transforms the original long-sequence data into a form easily processed by the model while preserving local features and contextual information. After preprocessing, the powerful feature extraction capabilities of hybrid neural network models can be leveraged to deeply mine temporal and spatial features from the low-altitude flight track data. Temporal features reflect the changing trends of flight status over time, while spatial features describe the aircraft's trajectory in space. Finally, the temporal and spatial features can be fused to form a fused flight track feature.

[0089] When eliminating noise in fused flight track features, the residual learning structure in a hybrid neural network model can be utilized. By introducing an identity mapping, the model can automatically learn the difference between data features and noise. Specifically, the residual learning structure can dynamically adjust the model's frequency band filtering coefficients, accurately identify the suddenness of impulse noise for fused flight track features, and effectively eliminate the regularity of periodic interference. This removes interference information from the original data, generating smoother and more accurate intermediate flight track data. This provides high-quality data support for subsequent track analysis and feature extraction, ensuring the effectiveness and reliability of the entire low-altitude flight track data processing workflow.

[0090] The hybrid neural network model includes:

[0091] A bidirectional LSTM layer is used to extract the temporal dependency features from the low-altitude flight track data, i.e., the time features, and the time features include velocity and acceleration;

[0092] A dilated convolutional layer is used to capture multi-scale coordinate features, i.e., spatial features, in the low-altitude flight trajectory data, and the spatial features include longitude, latitude, and altitude.

[0093] An attention mechanism layer is used to weightedly fuse the temporal and spatial features to generate spatiotemporal flight trajectory features;

[0094] The residual connection structure is used to connect and fuse the low-altitude flight track data with the spatiotemporal flight track features to generate the fused flight track features;

[0095] Understandably, hybrid neural network models can adopt a hierarchical structure design and achieve feature extraction and fusion of trajectory data through the collaborative work of multiple modules.

[0096] Specifically, the first layer is a bidirectional LSTM layer. This layer uses a bidirectional LSTM structure to simultaneously process forward and backward temporal information, accurately capturing temporal dependency features in low-altitude flight track data, specifically covering dynamic temporal features such as speed change trends and acceleration fluctuation patterns. The second layer is a dilated convolutional layer. This layer uses a dilated convolutional structure to extract coordinate information from the track data at different scales, effectively identifying spatial distribution patterns and geometric features in longitude, latitude, and altitude data. The third layer is an attention mechanism layer. This layer introduces an attention allocation mechanism to adaptively weight the extracted temporal and spatial features. This mechanism can dynamically adjust the weight allocation to achieve the organic fusion of spatiotemporal features, generating a spatiotemporal feature expression containing complete flight state information. The fourth layer is a residual connection structure. This layer superimposes the original track data with the fused spatiotemporal features through residual connections. It retains the basic information in the original data while incorporating deeply mined spatiotemporal features, forming a fused flight track feature containing multi-dimensional information, providing a rich and reliable feature foundation for subsequent filtering processing.

[0097] The training process of the hybrid neural network model is as follows:

[0098] Obtain historical flight track samples that include different weather conditions, geographical environments, and aircraft types;

[0099] The normal flight segments, flight change points, and noise areas in the historical flight track samples are marked;

[0100] The labeled historical flight track samples are divided into a track training set, a track verification set, and a track test set.

[0101] The hybrid neural network model is trained based on the aforementioned trajectory training set, and the training parameters are updated using the cross-entropy loss function and the stochastic gradient descent optimization algorithm.

[0102] The trained hybrid neural network model is validated based on the track validation set, and the model parameters of the hybrid neural network model are adjusted.

[0103] The adjusted hybrid neural network model is tested based on the aforementioned track test set to evaluate its filtering effect and generalization ability.

[0104] Furthermore, historical flight track samples covering various flight scenarios can be collected first. These samples include flight data under different weather conditions, geographical environments, and aircraft types. Weather conditions include sunny, cloudy, and rainy conditions; geographical environments include plains, mountains, and cities; and aircraft types include fixed-wing and rotary-wing aircraft. This ensures the diversity and representativeness of the sample set. Subsequently, the samples can be finely labeled to distinguish normal flight phases, key points of flight state changes, and noise interference areas, providing clear target labels for supervised learning.

[0105] After labeling, the sample set can be divided into training, validation, and test sets in a 4:1:1 ratio. During training, the hybrid neural network model is trained using the training set. The cross-entropy loss function is used to measure the difference between the model's predictions and the labeled values, and stochastic gradient descent is used to dynamically adjust the model parameters, gradually converging the loss function value. During validation, model performance is evaluated using data independent of the training set. Based on the validation results, the model's hyperparameters are adjusted and optimized to improve its generalization ability. Finally, in the testing phase, the model is comprehensively evaluated using a test set that was not used in training and validation. The evaluation focuses on the model's filtering effect in real-world application scenarios and its adaptability to unknown flight scenarios, ensuring that the model maintains high-precision trajectory data processing capabilities even in complex environments.

[0106] Through this hierarchical model design and rigorous training and verification process, the hybrid neural network model can effectively extract key features from low-altitude flight trajectory data, providing solid support for subsequent trajectory analysis and applications.

[0107] The dynamic adjustment process of the model's frequency band filtering coefficients is as follows:

[0108] Real-time monitoring of the noise intensity and spectral characteristics of the fused flight trajectory features;

[0109] When the detected change in noise intensity exceeds a preset change threshold, the adaptive coefficient update mechanism of the hybrid neural network model is triggered;

[0110] The frequency band filtering coefficients are calculated and adjusted using the residual learning structure of the hybrid neural network model.

[0111] Specifically, the first step is to continuously monitor the fused flight track characteristics, simultaneously collecting two key indicators: noise intensity and spectral characteristics. Noise intensity reflects the energy level of interference signals in the data, while spectral characteristics reveal the distribution of noise across different frequency components. Real-time changes in these two indicators can intuitively reflect the noise status of the flight track data.

[0112] When monitoring data shows that the sudden change in noise intensity exceeds the system's preset threshold, it indicates that the flight track data may be affected by sudden interference, such as electromagnetic pulses or weather anomalies. At this point, the adaptive coefficient update mechanism of the hybrid neural network model is immediately triggered, initiating a dynamic adjustment process for the frequency band filtering coefficients.

[0113] The adaptive adjustment process relies on the residual learning structure within the hybrid neural network model. This structure accurately identifies the noise frequency bands that require focused processing by calculating the residual between the original input and the network output. During the adjustment process, filtering weights for each frequency band are dynamically allocated based on the current noise's spectral characteristics, increasing filtering intensity for high-noise bands and reducing interference for high-signal-quality bands. This intelligent adjustment mechanism can automatically optimize filtering strategies for different types of noise.

[0114] Through the aforementioned closed-loop control involving real-time monitoring and dynamic adjustment, stable filtering performance can be maintained in complex and ever-changing flight environments. When the noise environment changes, the frequency band filtering coefficient can respond quickly and adjust to its optimal state, ensuring the accuracy and reliability of intermediate flight trajectory data. This process not only improves the system's robustness to sudden noise but also provides a high-quality data foundation for subsequent trajectory analysis and decision-making.

[0115] The process of identifying key and non-key trajectory features in the intermediate flight trajectory data using an SVM classifier and performing lossless and lossy compression respectively to generate replay flight trajectory data specifically includes:

[0116] The SVM classifier identifies the key trajectory feature data and the non-key trajectory data in the intermediate flight trajectory data. The key trajectory feature data includes flight trajectory turning points and trajectory altitude change points. The non-key trajectory data is the flight cruise segment.

[0117] The key trajectory feature data is losslessly compressed, and the non-key trajectory feature data is lossily compressed in combination with the aircraft kinematics prediction model to generate the replay flight trajectory data.

[0118] The replay flight track data is deduplicated by hashing using spatiotemporal correlation.

[0119] Understandably, the first step is to use an SVM classifier to identify features in the intermediate flight track data, dividing it into critical track feature data and non-critical track data. Critical track feature data contains important information reflecting changes in flight status, such as flight track turning points and altitude change points. These data points play a crucial role in the accurate reproduction and analysis of the flight process. Non-critical track data mainly consists of the cruise segment, during which the flight status is relatively stable, and the data changes exhibit strong regularity.

[0120] Differentiated compression strategies can be adopted for different types of data. For critical flight path feature data, lossless compression technology is used to ensure that no information is lost during the compression process, thereby guaranteeing the accurate reconstruction of critical flight states. For non-critical flight path data, lossy compression is performed in conjunction with an aircraft kinematic prediction model. This model, based on the aircraft's motion laws, predicts and approximates the cruise segment data, significantly reducing the data volume within a certain allowable error range, effectively balancing data accuracy and storage efficiency.

[0121] After data classification and compression, spatiotemporal correlation can be further utilized to optimize the replay flight track data. Hash deduplication technology is used to identify and eliminate duplicate data blocks. Because similar flight patterns or environmental characteristics may exist during flight, some data blocks may exhibit spatiotemporal repetition. Hash deduplication technology can quickly locate and remove these redundant data, further reducing data storage volume.

[0122] Through the aforementioned classification, differential compression, and deduplication processes, the resulting replay flight track data retains the integrity of key flight characteristics while significantly reducing data storage costs. This provides an efficient and reliable data foundation for subsequent flight playback, data analysis, and decision support. This processing workflow optimizes storage resource utilization and improves overall system performance while ensuring data availability.

[0123] The process for identifying the key trajectory feature data is as follows:

[0124] Extract the kinematic features from the intermediate flight track data;

[0125] The SVM classifier is used to classify the kinematic features, identify the turning points of the flight path and the points of change in flight path altitude, and generate the key flight path feature data;

[0126] Spatiotemporal correlation verification is performed on the identified key trajectory feature data to eliminate false change feature points in the key trajectory feature data;

[0127] Kinematic features can be extracted from intermediate flight track data, including trends in motion parameters such as speed, acceleration, and angular velocity, as well as the magnitude of changes in position coordinates in the horizontal and vertical directions. These kinematic features comprehensively reflect the aircraft's motion state, providing a data foundation for subsequent identification of key feature points.

[0128] Next, an SVM classifier can be used to classify the extracted kinematic features. Based on statistical learning theory, the SVM classifier constructs an optimal classification hyperplane to perform pattern recognition and classification on the kinematic feature data. In this process, the SVM classifier can divide the kinematic feature data into different categories, accurately identifying flight path turning points and altitude change points. These feature points are key nodes reflecting changes in flight status and are of great significance for flight process analysis and safety assessment. The identified feature points will form a set of key flight path feature data.

[0129] Finally, to ensure the accuracy of key flight path feature data, the identified feature points can be verified for spatiotemporal correlation. This verification process considers the continuity and logic of the flight path in terms of time sequence and spatial location. By analyzing the sequence of changes in feature points in the time dimension and their positional relationships in the spatial dimension, it determines whether the feature points conform to the actual motion patterns of the aircraft. Feature points that do not conform to spatiotemporal correlation, i.e., false change feature points, will be eliminated, thereby ensuring the reliability and authenticity of the key flight path feature data.

[0130] The lossy compression of the non-critical trajectory feature data by combining the aircraft kinematics prediction model specifically includes:

[0131] A kinematic prediction model for the aircraft is established based on Newton's laws of motion, and kinematic prediction is performed on the flight cruise segment to generate a predicted flight trajectory.

[0132] The actual flight trajectory is obtained, and the trajectory deviation between the actual flight trajectory and the predicted flight trajectory is calculated. Flight waypoints whose trajectory deviation exceeds a preset trajectory deviation threshold are then selected.

[0133] The flight waypoints are subjected to lossy compression encoding to complete the lossy compression operation of the non-critical waypoint feature data.

[0134] In practical applications, a kinematic prediction model for the aircraft can first be constructed based on Newton's laws of motion. This model comprehensively considers the aircraft's motion characteristics during the cruise phase, taking into account physical factors such as aerodynamics, gravity, and engine thrust. Combined with the aircraft's own parameters, it performs kinematic prediction of the flight trajectory during the cruise phase. By dynamically extrapolating parameters such as speed, acceleration, and flight attitude, a predicted flight trajectory is generated, which reflects the aircraft's motion trend under ideal conditions.

[0135] Secondly, actual flight trajectory data recorded during the actual flight can be obtained, containing the aircraft's true position information at each moment. Then, the actual flight trajectory can be compared and analyzed with the predicted flight trajectory to calculate the trajectory deviation between the two. This deviation value quantifies the difference between the aircraft's actual motion and the theoretical prediction, and can intuitively reflect abnormal changes during flight. Based on a preset trajectory deviation threshold, the deviation values ​​of all waypoints are filtered, retaining those whose deviations exceed the threshold. These filtered waypoints often represent key state changes that deviate from the normal motion pattern during flight, and have high information value.

[0136] Finally, lossy compression encoding can be performed on the selected flight waypoints. The encoding process uses a data compression algorithm to simplify the waypoint information while allowing for some data loss. By removing redundant information and approximate quantization, the data storage volume is effectively reduced while preserving the main features of the flight path, ultimately completing the lossy compression operation on non-critical flight path feature data.

[0137] This compression strategy, based on model prediction and bias filtering, reduces the amount of data while retaining key changes during flight, ensuring data availability while improving the storage and transmission efficiency of flight track data.

[0138] The aforementioned collaborative computing framework, integrating CPU, GPU, and FPGA, distributes the replayed flight trajectory data to multiple trajectory computing nodes for collaborative processing. Specifically, it includes:

[0139] The design incorporates a collaborative computing framework for the CPU, GPU, and FPGA. A data sharding module divides the replay flight trajectory data into multiple trajectory data sub-blocks according to a three-dimensional spatial grid and assigns them to the corresponding trajectory computing nodes for collaborative processing. The dimensions of the three-dimensional spatial grid include longitude, latitude, and altitude.

[0140] Specifically, the CPU is used for task scheduling and global parameter updates, the GPU is used to perform matrix operations for filtering, and the FPGA is used to accelerate bitwise operations for lossless compression.

[0141] When constructing a collaborative computing framework, a data sharding module can be introduced as a key component for data allocation. This module performs fine-grained partitioning of replay flight track data based on three-dimensional spatial grid rules. A three-dimensional spatial grid system is constructed using longitude, latitude, and altitude as dimensions, dividing the entire low-altitude flight area into multiple spatial sub-regions. Correspondingly, the track data is also cut into multiple track data sub-blocks and assigned to corresponding track computing nodes. This spatial grid-based data sharding strategy makes data processing more targeted, reduces data transmission and computation redundancy, and improves the feasibility of parallel processing.

[0142] In collaborative computing, the CPU acts as the control center of the entire collaborative computing framework, responsible for overall management of task scheduling and global parameter updates. Based on data sharding, it rationally allocates task priorities for each track computing node and coordinates data interaction and synchronization between different nodes. Simultaneously, it dynamically adjusts and optimizes global parameters during the computing process, ensuring the stability and accuracy of the computation flow. The GPU, with its powerful parallel computing capabilities, undertakes matrix operations in the filtering process. Track data filtering involves numerous matrix multiplications and additions. Through its many-core architecture, the GPU can process multiple data units simultaneously, significantly improving matrix operation speed and accelerating the filtering process. The FPGA focuses on bitwise operations in lossless compression. Lossless compression algorithms typically involve bit-level logical operations and data reconstruction. Customized hardware logic circuits can accelerate bitwise operations, significantly improving the processing efficiency of lossless compression compared to traditional software computing methods, ensuring the efficiency and integrity of key track feature data during compression.

[0143] Through this spatial grid-based data sharding strategy and precise division of heterogeneous computing resources, the collaborative computing framework achieves efficient parallel processing of replay flight trajectory data, fully leveraging the performance advantages of different computing devices and providing powerful computing support for the real-time processing and analysis of low-altitude flight trajectory data.

[0144] The process of dividing the replayed flight trajectory data into multiple trajectory data sub-blocks according to a three-dimensional spatial grid using a data sharding module and assigning them to corresponding trajectory computing nodes for collaborative processing specifically includes:

[0145] Based on the dimensions of the three-dimensional spatial grid, the data sharding module maps the replay flight trajectory data onto the three-dimensional spatial grid to generate multiple trajectory data sub-blocks;

[0146] Aggregate and index the replay flight trajectory data within each trajectory data sub-block to generate a corresponding trajectory space index table;

[0147] Multiple trajectory computing nodes are set up. Based on the load and network topology of each trajectory computing node, each trajectory data sub-block is allocated to the corresponding trajectory computing node for collaborative processing according to the trajectory space index table, and a node data transmission channel is established between the trajectory computing nodes.

[0148] It should be noted that, based on the dimensional system of a three-dimensional spatial grid, the replay flight track data can be mapped onto this spatial grid through a data sharding module. By quantizing and segmenting the three dimensions of longitude, latitude, and altitude, continuous track data can be discretized into multiple track data sub-blocks with well-defined spatial boundaries. This mapping mechanism not only achieves spatial partitioning of the data but also preserves the topological relationships of the track data in three-dimensional space, providing a structured foundation for subsequent parallel processing.

[0149] Then, aggregation and indexing operations can be performed on the replay flight track data within each track data sub-block. Aggregation integrates track point data within the same spatial region, extracts key features, and generates a statistical summary. Indexing constructs a track spatial index table, enabling rapid location and retrieval of track data through spatial filling curves or multi-dimensional index structures. This indexing mechanism not only accelerates data access efficiency within a single node but also provides a logical basis for cross-node data collaboration.

[0150] Finally, based on the load status of computing resources and the network topology, track data sub-blocks can be dynamically allocated to corresponding track computing nodes. By monitoring the computing resource utilization, memory usage, and network bandwidth of each node in real time, and combining this with the data distribution information in the spatial index table, a load balancing algorithm is used to achieve optimal allocation of data sub-blocks. Simultaneously, to ensure efficient data transmission between nodes, dedicated data transmission channels can be established between track computing nodes. By optimizing network protocols, caching strategies, and data routing algorithms, data transmission latency can be reduced, thereby improving the overall throughput of distributed computing.

[0151] This three-dimensional spatial grid partitioning and distributed processing strategy can transform large-scale flight track data into subtasks that can be processed in parallel, giving full play to the advantages of multi-node collaborative computing, achieving high efficiency and scalability in flight track data processing, and providing strong technical support for applications such as low-altitude flight monitoring and situational awareness.

[0152] The system employs a consistent hashing-based storage architecture to partition and store the replay flight track data, and designs a multi-level indexing mechanism to support the retrieval of the replay flight track data, specifically including:

[0153] Using the consistent hash-based storage architecture, the replay flight track data is partitioned into multiple hash data blocks by two dimensions: timestamp and geographical region.

[0154] For each hash data block, multiple replica data blocks are expanded, and the hash data block and the corresponding replica data block are stored in the consistent hash-based storage architecture;

[0155] The multi-level indexing mechanism is designed to support the retrieval of the replay flight track data, and the multi-level indexing mechanism includes a global index based on date, a regional index based on region code, and a feature index based on track type.

[0156] By combining a consistent hashing-based storage architecture with a multi-level indexing mechanism, a solution for efficient storage and fast retrieval of replay flight track data can be provided. This process ensures data reliability and access efficiency through dual-dimensional data partitioning, redundant replica storage, and multi-level index construction.

[0157] Specifically, a consistent hashing-based storage architecture can be adopted to partition the replay flight track data. Using timestamps and geographical regions as dual-dimensional partitioning criteria, a hash algorithm maps the data to different storage areas, generating multiple hash data blocks. The timestamp dimension segments the data according to the flight time series, while the geographical region dimension divides the data based on the aircraft's spatial location. Combining these two dimensions enables fine-grained data classification and storage. This dual-dimensional hashing partitioning strategy not only meets the need for retrieving historical flight tracks by time but also facilitates data statistics and analysis based on geographical regions, improving the structure and manageability of data storage.

[0158] To ensure the reliability and fault tolerance of data storage, multiple replica data blocks can be generated for each hash data block. These replica data blocks are stored together with the original hash data block in a consistent hash storage architecture and distributed across different physical nodes using a distributed storage approach. This replica redundancy mechanism ensures that even if some storage nodes fail, data can still be accessed normally through other replica nodes, effectively avoiding the risk of data loss and enhancing the stability and availability of the storage system.

[0159] Finally, a multi-level indexing mechanism can be constructed to support efficient data retrieval. This indexing system comprises three levels: a date-based global index, which uses the flight date as the index key to achieve rapid time positioning of the entire dataset, facilitating users to retrieve historical flight tracks by time period; a region-code-based regional index, which is indexed based on geographic region codes and is suitable for querying flight data within specific airspaces; and a feature-based index based on flight track type, which categorizes and indexes according to flight mission type, such as training flights and route flights, meeting users' needs for precise retrieval based on flight track features. The collaborative work of these multi-level indexes forms a three-dimensional retrieval network from macro to micro and from time to space, significantly improving the response speed and accuracy of data queries.

[0160] Through data partitioning and replication redundancy in the consistent hash storage architecture, as well as efficient retrieval support from the multi-level indexing mechanism, the reliability and scalability of low-altitude flight track data storage can be ensured, providing a solid data management foundation for applications such as aviation data analysis and flight monitoring.

[0161] like Figure 2 The diagram shown is a system block diagram of an airport low-altitude flight trajectory playback and analysis system provided in an embodiment of the present invention. The system includes:

[0162] The trajectory filtering module is used to acquire low-altitude flight trajectory data and construct a hybrid neural network model to filter the low-altitude flight trajectory data to generate intermediate flight trajectory data.

[0163] The trajectory compression module is used to identify key trajectory feature data and non-key trajectory data in the intermediate flight trajectory data through an SVM classifier and perform lossless compression and lossy compression respectively to generate playback flight trajectory data.

[0164] The collaborative processing module is used to design a collaborative computing framework for CPU, GPU and FPGA, and to distribute the replayed flight trajectory data to multiple trajectory computing nodes for collaborative processing.

[0165] The storage and retrieval module is used to partition and store the replay flight track data using a consistent hashing-based storage architecture, and to design a multi-level indexing mechanism to support the retrieval of the replay flight track data.

[0166] Figure 2 The apparatus of the illustrated embodiment can be used to perform corresponding actions. Figure 1 The steps in the method embodiments shown are implemented in a similar manner and have similar technical effects, and will not be repeated here.

[0167] An electronic device includes a memory and a processor, wherein the memory stores a computer program, and when the processor runs the computer program stored in the memory, the processor performs the steps of the airport low-altitude flight track playback analysis method as described in any of the above.

[0168] like Figure 3 The diagram shown is a hardware structure schematic of an electronic device according to an embodiment of the present invention. The electronic device 30 includes: a processor 31, a memory 32, and a computer program; wherein...

[0169] The memory 32 is used to store the computer program, and the memory may also be flash memory. The computer program is, for example, an application program or functional module that implements the above method.

[0170] The processor 31 is configured to execute the computer program stored in the memory to implement the various steps performed by the device in the above method. For details, please refer to the relevant descriptions in the preceding method embodiments.

[0171] Alternatively, the memory 32 can be either standalone or integrated with the processor 31.

[0172] When the memory 32 is a device independent of the processor 31, the device may further include:

[0173] Bus 33 is used to connect the memory 32 and the processor 31.

[0174] A readable storage medium storing a computer program, which, when executed by a processor, is used to implement the steps of an airport low-altitude flight track playback analysis method as described in any of the above claims.

[0175] The readable storage medium can be a computer storage medium or a communication medium. A communication medium includes any medium that facilitates the transfer of computer programs from one location to another. A computer storage medium can be any available medium accessible to a general-purpose or special-purpose computer. For example, a readable storage medium is coupled to a processor, enabling the processor to read information from and write information to the readable storage medium. Of course, the readable storage medium can also be a component of the processor. The processor and the readable storage medium can reside in an Application-Specific Integrated Circuit (ASIC). Alternatively, the ASIC can be located in a user equipment. Of course, the processor and the readable storage medium can also exist as discrete components in a communication device. The readable storage medium can be a read-only memory (ROM), random access memory (RAM), CD-ROM, magnetic tape, floppy disk, and optical data storage device, etc.

[0176] The present invention also provides a program product including executable instructions stored in a readable storage medium. At least one processor of the device can read the executable instructions from the readable storage medium, and the at least one processor executes the executable instructions to cause the device to implement the methods provided in the various embodiments described above.

[0177] In the embodiments of the above-described device, it should be understood that the processor can be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), etc. The general-purpose processor can be a microprocessor or any conventional processor. The steps of the method disclosed in this invention can be directly manifested as execution by a hardware processor, or execution by a combination of hardware and software modules within the processor.

[0178] Through the above embodiments, this invention, through an airport low-altitude flight track playback analysis method and system, acquires low-altitude flight track data, constructs a hybrid neural network model to filter the low-altitude flight track data to generate intermediate flight track data; uses an SVM classifier to identify key track feature data and non-key track data in the intermediate flight track data and performs lossless compression and lossy compression respectively to generate playback flight track data; designs a collaborative computing framework of CPU, GPU, and FPGA to distribute the playback flight track data to multiple track computing nodes for collaborative processing; adopts a consistent hash-based storage architecture to partition and store the playback flight track data, and designs a multi-level indexing mechanism to support the retrieval of the playback flight track data, thereby significantly improving the accuracy of track playback, optimizing data processing efficiency, and reducing system storage and computing costs.

[0179] In terms of data processing accuracy, this invention employs a deep learning-based adaptive filtering algorithm to accurately identify various types of noise in complex low-altitude environments, dynamically optimize filtering parameters, and effectively improve the purity and reliability of flight track data, laying the foundation for high-precision flight track playback. In the field of data compression, a hierarchical compression algorithm intelligently distinguishes between critical and non-critical data, significantly improving data compression efficiency while ensuring the integrity of key flight track features, thus guaranteeing the real-time and continuous nature of flight track playback. Regarding the computational processing architecture, the organic combination of heterogeneous computing and parallel processing technologies fully leverages the advantages of different computing units, significantly improving the processing speed of large-scale flight track data and effectively addressing the data processing challenges brought about by the surge in the number of low-altitude aircraft. In terms of storage management, a dual-dimensional hash partitioning and multi-level indexing mechanism not only enhances the security and reliability of data storage but also enables rapid retrieval of flight track data across different time periods, significantly improving data management efficiency and reducing system storage and computing costs.

[0180] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features; and these modifications or substitutions 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 analyzing low-altitude flight tracks at airports, characterized in that, The method includes: Acquire low-altitude flight trajectory data, construct a hybrid neural network model to filter the low-altitude flight trajectory data, and generate intermediate flight trajectory data. The key trajectory feature data and non-key trajectory data in the intermediate flight trajectory data are identified by the SVM classifier and then subjected to lossless compression and lossy compression respectively to generate playback flight trajectory data. Design a collaborative computing framework for CPU, GPU and FPGA to distribute the replayed flight trajectory data to multiple trajectory computing nodes for collaborative processing; A consistent hash-based storage architecture is adopted to partition and store the replay flight track data, and a multi-level indexing mechanism is designed to support the retrieval of the replay flight track data; The process of acquiring low-altitude flight trajectory data and constructing a hybrid neural network model to filter the low-altitude flight trajectory data to generate intermediate flight trajectory data specifically includes: The low-altitude flight trajectory data is acquired, and the hybrid neural network model containing bidirectional LSTM and dilated convolution is constructed. The hybrid neural network model is trained by acquiring historical flight track samples, and the historical flight track samples cover 12 typical low-altitude flight scenarios; After performing sliding window preprocessing on the low-altitude flight track data, the time and spatial features in the low-altitude flight track data are extracted through the hybrid neural network model to generate fused flight track features. By dynamically adjusting the model's frequency band filtering coefficients through the residual learning structure of the hybrid neural network model, the impulse noise and periodic interference in the fused flight track features are eliminated, thereby generating the intermediate flight track data.

2. The airport low-altitude flight trajectory playback analysis method according to claim 1, characterized in that, The hybrid neural network model includes: A bidirectional LSTM layer is used to extract the temporal dependency features from the low-altitude flight track data, i.e., the time features, and the time features include velocity and acceleration; A dilated convolutional layer is used to capture multi-scale coordinate features, i.e., spatial features, in the low-altitude flight trajectory data, and the spatial features include longitude, latitude, and altitude. An attention mechanism layer is used to weightedly fuse the temporal and spatial features to generate spatiotemporal flight trajectory features; A residual learning structure is used to connect and fuse the low-altitude flight trajectory data with the spatiotemporal flight trajectory features to generate the fused flight trajectory features; The training process of the hybrid neural network model is as follows: Obtain historical flight track samples that include different weather conditions, geographical environments, and aircraft types; The normal flight segments, flight change points, and noise areas in the historical flight track samples are marked; The labeled historical flight track samples are divided into a track training set, a track verification set, and a track test set. The hybrid neural network model is trained based on the aforementioned trajectory training set, and the training parameters are updated using the cross-entropy loss function and the stochastic gradient descent optimization algorithm. The trained hybrid neural network model is validated based on the track validation set, and the model parameters of the hybrid neural network model are adjusted. The adjusted hybrid neural network model is tested based on the aforementioned track test set to evaluate its filtering effect and generalization ability.

3. The airport low-altitude flight trajectory playback analysis method according to claim 1, characterized in that, The dynamic adjustment process of the model's frequency band filtering coefficients is as follows: Real-time monitoring of the noise intensity and spectral characteristics of the fused flight trajectory features; When the detected change in noise intensity exceeds a preset change threshold, the adaptive coefficient update mechanism of the hybrid neural network model is triggered; The frequency band filtering coefficients are calculated and adjusted using the residual learning structure of the hybrid neural network model.

4. The airport low-altitude flight trajectory playback analysis method according to claim 1, characterized in that, The process of identifying key and non-key trajectory features in the intermediate flight trajectory data using an SVM classifier and performing lossless and lossy compression respectively to generate replay flight trajectory data specifically includes: The SVM classifier identifies the key trajectory feature data and the non-key trajectory data in the intermediate flight trajectory data. The key trajectory feature data includes flight trajectory turning points and trajectory altitude change points. The non-key trajectory data is the flight cruise segment. The key trajectory feature data is losslessly compressed, and the non-key trajectory feature data is lossily compressed in combination with the aircraft kinematics prediction model to generate the replay flight trajectory data. The replay flight track data is deduplicated by hashing using spatiotemporal correlation.

5. The airport low-altitude flight trajectory playback analysis method according to claim 4, characterized in that, The process for identifying the key trajectory feature data is as follows: Extract the kinematic features from the intermediate flight track data; The SVM classifier is used to classify the kinematic features, identify the turning points of the flight path and the points of change in flight path altitude, and generate the key flight path feature data; Spatiotemporal correlation verification is performed on the identified key trajectory feature data to eliminate false change feature points in the key trajectory feature data; The lossy compression of the non-critical trajectory feature data by combining the aircraft kinematics prediction model specifically includes: A kinematic prediction model for the aircraft is established based on Newton's laws of motion, and kinematic prediction is performed on the flight cruise segment to generate a predicted flight trajectory. The actual flight trajectory is obtained, and the trajectory deviation between the actual flight trajectory and the predicted flight trajectory is calculated. Flight waypoints whose trajectory deviation exceeds a preset trajectory deviation threshold are then selected. The flight waypoints are subjected to lossy compression encoding to complete the lossy compression operation of the non-critical waypoint feature data.

6. The airport low-altitude flight trajectory playback analysis method according to claim 1, characterized in that, The aforementioned collaborative computing framework, integrating CPU, GPU, and FPGA, distributes the replayed flight trajectory data to multiple trajectory computing nodes for collaborative processing. Specifically, it includes: The design incorporates a collaborative computing framework for the CPU, GPU, and FPGA. A data sharding module divides the replay flight trajectory data into multiple trajectory data sub-blocks according to a three-dimensional spatial grid and assigns them to the corresponding trajectory computing nodes for collaborative processing. The dimensions of the three-dimensional spatial grid include longitude, latitude, and altitude. Specifically, the CPU is used for task scheduling and global parameter updates, the GPU is used to perform matrix operations for filtering, and the FPGA is used to accelerate bitwise operations for lossless compression.

7. The airport low-altitude flight trajectory playback analysis method according to claim 6, characterized in that, The process of dividing the replayed flight trajectory data into multiple trajectory data sub-blocks according to a three-dimensional spatial grid using a data sharding module and assigning them to corresponding trajectory computing nodes for collaborative processing specifically includes: Based on the dimensions of the three-dimensional spatial grid, the data sharding module maps the replay flight trajectory data onto the three-dimensional spatial grid to generate multiple trajectory data sub-blocks; Aggregate and index the replay flight trajectory data within each trajectory data sub-block to generate a corresponding trajectory space index table; Multiple trajectory computing nodes are set up. Based on the load and network topology of each trajectory computing node, each trajectory data sub-block is allocated to the corresponding trajectory computing node for collaborative processing according to the trajectory space index table, and a node data transmission channel is established between the trajectory computing nodes.

8. The airport low-altitude flight trajectory playback analysis method according to claim 1, characterized in that, The system employs a consistent hashing-based storage architecture to partition and store the replay flight track data, and designs a multi-level indexing mechanism to support the retrieval of the replay flight track data, specifically including: Using the consistent hash-based storage architecture, the replay flight track data is partitioned into multiple hash data blocks by two dimensions: timestamp and geographical region. For each hash data block, multiple replica data blocks are expanded, and the hash data block and the corresponding replica data block are stored in the consistent hash-based storage architecture; The multi-level indexing mechanism is designed to support the retrieval of the replay flight track data, and the multi-level indexing mechanism includes a global index based on date, a regional index based on region code, and a feature index based on track type.

9. An airport low-altitude flight trajectory playback and analysis system, applied to the airport low-altitude flight trajectory playback and analysis method as described in any one of claims 1-8, characterized in that, The system includes: The trajectory filtering module is used to acquire low-altitude flight trajectory data and construct a hybrid neural network model to filter the low-altitude flight trajectory data to generate intermediate flight trajectory data. The trajectory compression module is used to identify key trajectory feature data and non-key trajectory data in the intermediate flight trajectory data through an SVM classifier and perform lossless compression and lossy compression respectively to generate playback flight trajectory data. The collaborative processing module is used to design a collaborative computing framework for CPU, GPU and FPGA, and to distribute the replayed flight trajectory data to multiple trajectory computing nodes for collaborative processing. The storage and retrieval module is used to partition and store the replay flight track data using a consistent hashing-based storage architecture, and to design a multi-level indexing mechanism to support the retrieval of the replay flight track data.

Images