Aircraft individual identification method and device based on multi-dimensional track characteristics

Abstract

The embodiment of the invention relates to the technical field of aviation target recognition, in particular to an aircraft individual recognition method and device based on multi-dimensional track features. A specific embodiment of the method comprises the steps of performing data preprocessing on acquired original track data to generate standard track data; performing feature extraction processing on the standard track data to generate track feature data; performing format conversion processing on the track characteristic data to generate track conversion data; and inputting the flight path conversion data into a pre-trained aircraft individual identification model to obtain aircraft individual identification information. According to the embodiment, the track characteristic analysis method for effectively processing the dynamic characteristics and the multi-dimensional relevance of the time sequence track data is constructed.

Description

TECHNICAL FIELD

[0001] Embodiments of the present application relate to the technical field of aerial target recognition, and in particular to a method and device for aircraft individual recognition based on multi-dimensional track features. BACKGROUND

[0002] Implementing aircraft individual recognition usually requires a large amount of feature data such as images, radars and radio frequencies. On the one hand, due to the limited resolution or detection capability of sensors, it is not necessarily possible to provide rich and complete feature data. For example, the visible light image-based recognition method has the following challenges: it is difficult to obtain a clear image when the aircraft is moving at high speed, the detection distance affects the imaging resolution of the aircraft, and the weather can significantly affect the quality of the aircraft image. Therefore, when visible light images are difficult to obtain, individual recognition based on other radiation sources is more important. On the other hand, there is less deep analysis of individual recognition features based on measured track data. In the field of aerial target recognition, existing researches mainly focus on aircraft model recognition, aircraft action recognition, aerial target intention recognition, and aerial target track pattern recognition.

[0003] The existing aircraft individual recognition technology has the problems of difficulty in obtaining images and less research on measured track data as individual recognition features. Traditional classifiers are difficult to effectively process the dynamic characteristics and multi-dimensional correlation of time series track data. SUMMARY

[0004] The summary portion of the present application is used to introduce the concepts in a brief form, which will be described in detail in the specific embodiments section. The summary portion of the present application is not intended to identify key or essential features of the claimed technical solutions, nor is it intended to limit the scope of the claimed technical solutions.

[0005] Some embodiments of the present application propose a method and device for aircraft individual recognition based on multi-dimensional track features, a computer device and a computer readable storage medium to solve one or more of the technical problems mentioned in the background section.

[0006] In a first aspect, some embodiments of the present application provide a method for aircraft individual recognition based on multi-dimensional track features, which comprises: performing data preprocessing on the obtained original track data to generate standard track data; performing feature extraction processing on the standard track data to generate track feature data; performing format conversion processing on the track feature data to generate track conversion data; inputting the track conversion data into a pre-trained aircraft individual recognition model to obtain aircraft individual recognition information.

[0007] In a second aspect, some embodiments of the present application provide an aircraft individual identification device based on multi-dimensional track features, the device comprising: an acquisition unit configured to acquire raw track data; a data preprocessing unit configured to perform data preprocessing on the acquired raw track data to generate standard track data; a feature extraction unit configured to perform feature extraction processing on the standard track data to generate track feature data; a format conversion unit configured to perform format conversion processing on the track feature data to generate track conversion data; and an input unit configured to input the track conversion data into a pre-trained aircraft individual identification model to obtain aircraft individual identification information.

[0008] In a third aspect, the present application also provides a computer device, comprising a processor, a memory, and a computer program stored in the memory and executable by the processor, wherein when the computer program is executed by the processor, the method described in any of the implementations of the first aspect is implemented.

[0009] In a fourth aspect, the present application also provides a computer-readable storage medium, which stores a computer program, wherein when the computer program is executed by a processor, the method described in any of the implementations of the first aspect is implemented.

[0010] The above various embodiments of the present application have the following beneficial effects: by the aircraft individual identification method based on multi-dimensional track features of some embodiments of the present application, a track feature analysis method for effectively processing the dynamic characteristics and multi-dimensional correlation of time series track data is constructed. Specifically, the reason why it is difficult to construct a track feature analysis method for effectively processing the dynamic characteristics and multi-dimensional correlation of time series track data is that: the existing aircraft individual identification technology has problems such as difficulty in obtaining images, less research on measured track data as individual identification features, and traditional classifiers are difficult to effectively process the dynamic characteristics and multi-dimensional correlation of time series track data. Based on this, the aircraft individual identification method based on multi-dimensional track features of some embodiments of the present application first performs data preprocessing on the obtained original track data to generate standard track data. Secondly, the standard track data is processed for feature extraction to generate track feature data. Then, the track feature data is processed for format conversion to generate track conversion data. Then, the track conversion data is input into the pre-trained aircraft individual identification model to obtain aircraft individual identification information. Therefore, the present application proposes an aircraft individual GRU identification algorithm using multi-dimensional track data as feature quantities, which first uses six-dimensional track data of longitude, latitude, altitude, speed, climb and heading as aircraft individual identification features from the perspectives of dynamics and geographic information space, then constructs an evaluation model for track feature selection based on the area under the curve (Area Under the Curve, AUC), and finally establishes an aircraft individual identification algorithm combining attention mechanism (Attention Mechanism) and gate recurrent unit (Gate Recurrent Unit, GRU). BRIEF DESCRIPTION OF DRAWINGS

[0011] The above and other features, advantages and aspects of the embodiments of the present application will become more apparent upon reading the following detailed description in conjunction with the accompanying drawings, in which like references refer to like elements. It is to be understood that the drawings are diagrammatic and schematic representations of such embodiments and elements, and are not limiting of the scope of the present application unless otherwise specifically stated.

[0012] Figure 1 is a flowchart of some embodiments of the aircraft individual identification method based on multi-dimensional track features according to the present application;

[0013] Figure 2 is a schematic diagram of the feature structure of track feature data of the aircraft individual identification method based on multi-dimensional track features according to the present application;

[0014] Figure 3 is a model structure schematic diagram of the first initial gate model of the aircraft individual identification method based on multi-dimensional track features according to the present application;

[0015] Figure 4 is a schematic diagram of a label matrix of a method for aircraft individual identification based on multi-dimensional track features according to the present application;

[0016] Figure 5 is a schematic diagram of a probability matrix of a method for aircraft individual identification based on multi-dimensional track features according to the present application;

[0017] Figure 6 is a schematic diagram of the structure of some embodiments of a device for aircraft individual identification based on multi-dimensional track features according to the present application;

[0018] Figure 7 is a schematic diagram of the structure of a computer device suitable for implementing some embodiments of the present application. DETAILED DESCRIPTION

[0019] Embodiments of the present application will be described in more detail with reference to the drawings. Although certain embodiments of the present application are shown in the drawings, it is understood that the present application can be embodied in various forms and should not be interpreted in a limiting sense. Rather, these embodiments are provided so that this disclosure will be thorough and complete. It is understood that the drawings and embodiments are for illustrative purposes only and should not be construed as limiting the scope of the present application.

[0020] It should also be noted that, for the sake of brevity, only some of the various embodiments of the present application are specifically described herein. It is understood that the embodiments of the present application specifically described herein can be used in combination with each other, unless specifically noted otherwise.

[0021] It should be noted that the terms "first", "second", and the like in the present application are used only to distinguish different devices, modules or units, and do not imply the order of execution or the mutual dependency of the functions performed by these devices, modules or units.

[0022] It should be noted that the terms "one", "multiple", etc. in the present application are illustrative and not restrictive, and those skilled in the art should understand that, unless otherwise explicitly stated in the context, "one" or "multiple" should be understood as "one or more".

[0023] The names of the messages or information exchanged between the devices in the embodiments of the present application are only for illustrative purposes, and are not intended to limit the scope of the messages or information.

[0024] The present application will be described in detail below with reference to the drawings and in conjunction with the embodiments.

[0025] Figure 1 A flowchart 100 of some embodiments of a method for aircraft individual identification based on multi-dimensional track features according to the present application is shown. The method for aircraft individual identification based on multi-dimensional track features comprises the following steps:

[0026] In step 101, the acquired original track data is preprocessed to generate standard track data.

[0027] In some embodiments, the execution subject of the aircraft individual identification method based on multi-dimensional track features can preprocess the acquired original track data to generate standard track data. The original track data can be acquired from a storage terminal through wired or wireless connection. The original track data can include, but is not limited to, longitude, latitude and altitude, ground speed, climb rate and heading angle. The original track data can represent the data of the aircraft in high-speed motion.

[0028] It should be noted that the wireless connection can include, but is not limited to, 3G / 4G connection, WiFi connection, Bluetooth connection, WiMAX connection, Zigbee connection, UWB (ultra wideband) connection, and other now known or future developed wireless connection methods.

[0029] In practice, the execution subject can preprocess the acquired original track data to generate standard track data by the following steps:

[0030] First, the longitude, latitude and altitude included in the original track data are converted to generate a position vector. In practice, the execution subject can convert the longitude, latitude and altitude representing the geodetic coordinate system into a position vector in the Earth Centered Fixed (ECF) system.

[0031] Second, the ground speed, climb rate and heading angle included in the original track data are converted to generate a velocity vector. In practice, the execution subject can convert the original track data into a velocity vector in the East North Up (ENU) system.

[0032] Third, the position vector and the velocity vector are determined as equivalent measurement data.

[0033] Fourth, the equivalent measurement data is removed to generate noise data. In practice, the execution subject can remove the trend item of the equivalent measurement data by a preset removal algorithm to obtain noise data. For example, the preset removal algorithm can be a spline fitting algorithm.

[0034] In the fifth step, the noise data is calibrated to generate noise statistical feature data. In practice, the execution subject can calibrate the noise data by a preset calibration algorithm to generate the noise statistical feature data. For example, the preset calibration algorithm can be: first, determine the trend item of the noise data by a cubic spline fitting algorithm to obtain a spline fitting error. Second, determine the covariance matrix of the spline fitting error by a least square noise variance algorithm to obtain the noise statistical feature data.

[0035] In the sixth step, a state equation is constructed based on the equivalent measurement data. In practice, the execution subject can construct a uniform acceleration model described by a VTC (Velocity Turn Climb, VTC) coordinate system acceleration as the state equation by the following steps:

[0036] In the first sub-step, a state vector is determined. In practice, the state vector is composed of a position vector in the earth-fixed system, a velocity vector, and VTC system acceleration, and the state vector can refer to the following formula:

[0037]

[0038] wherein, represents the state vector, represents the position vector in the earth-fixed system, represents the longitude, represents the latitude, represents the altitude, represents the velocity vector, represents the ground speed, represents the climb rate, represents the heading angle, represents the VTC system acceleration, represents the velocity component, represents the turning component, represents the climbing component.

[0039] In the second sub-step, the state equation is determined based on the state vector. In practice, within a measurement interval, the change of the state vector is a process with constant VTC system acceleration, and the discretized state equation can refer to the following formula:

[0040]

[0041] wherein, represents the serial number, represents the state equation, represents the coordinate conversion matrix from the VTC system to the earth-fixed system, represents the measurement interval, represents a 3-order unit matrix, ​​represents a 3-by-3 zero matrix, represents process noise, which is unknown zero-mean Gaussian white noise.

[0042] In the seventh step, a measurement equation is determined based on the above state equation. The measurement equation can refer to the following formula:

[0043]

[0044] wherein, represents a measurement equation, represents an x-axis component of the, represents a y-axis component of the, represents a z-axis component of the, represents a z-axis component of the, represents a z-axis component of the, represents a z-axis component of the, represents a z-axis component of the, , represents a coordinate conversion matrix from the earth-fixed system to the local ENU system, , represents a longitude, represents a latitude. Herein, can embody a functional relationship from the state equation to the measurement equation, wherein, represents a coefficient, .

[0045] In the eighth step, the original track data is processed for track smoothing based on the above state equation and the above measurement equation to generate track smoothed data. In practice, based on the above state equation and the above measurement equation, the execution subject can process the original track data for track smoothing by a preset track smoothing algorithm to generate track smoothed data. For example, the preset track smoothing algorithm can be an extended Kalman filter. Thus, the Kalman filter is used to perform state prediction based on the uniform motion model, and according to the validity of the measurement data, it is selected whether to fuse the measurement information for state update.

[0046] In the ninth step, the track smoothed data is processed for bad point rejection based on the above noise statistical characteristic data to generate track rejection data. In practice, based on the above noise statistical characteristic data, the execution subject can reject the bad points in the track smoothed data by a preset bad point rejection algorithm to obtain track rejection data. For example, the preset bad point rejection algorithm can be a chi-square test algorithm of innovation. Thus, the Kalman filter is used to perform state prediction based on the uniform motion model, to calculate the innovation (prediction error), to define the Mahalanobis distance of the innovation, and if it is greater than a threshold, it is determined that the measurement data is a bad point, the measurement update is skipped, and the predicted value is directly used as the state update.

[0047] ​In the tenth step, the track elimination data is converted into standard track data. In practice, the execution subject can convert the track elimination data into standard track data in the geodetic coordinate system. Thus, the position of the original data is geodetic latitude and longitude, and the coordinate system used in the track preprocessing is the earth-fixed coordinate system. After preprocessing, the earth-fixed coordinate system needs to be converted back to the geodetic latitude and longitude.

[0048] In the second step, the standard track data is processed to generate track feature data.

[0049] In some embodiments, the execution subject can perform feature extraction processing on the standard track data to generate track feature data. In practice, the execution subject can perform feature extraction processing on the standard track data through a one-dimensional convolution layer to generate track feature data. The one-dimensional convolution layer can have a convolution kernel size of 3, a convolution layer number of 1, a step size of 1, and a padding method of padding 1 unit on both sides of the input sequence to ensure that the output length after the convolution operation is consistent with the input length. Figure 2 Figure 2 Fig. 1 shows the structure of the track feature data of the aircraft individual identification method based on multi-dimensional track features according to the present application. The track feature data can include geographical space information features (corresponding to position vectors) and dynamic features (corresponding to velocity vectors). The geographical space information features can include longitude, latitude, and altitude. The dynamic features can include speed (ground speed), heading (heading angle), and climb (climb rate). Figure 2 In Fig. 1, track 1 and track 2 represent discrete track points in each track data.

[0050] In the third step, the track feature data is converted into track conversion data.

[0051] In some embodiments, the execution subject can perform format conversion processing on the track feature data to generate track conversion data. In practice, the execution subject can perform format conversion processing on the track feature data through a preset format conversion algorithm to generate track conversion data. For example, the preset format conversion algorithm can be a permute (dimension rearrangement) algorithm. Thus, the convolution layer does not use an activation function, and directly uses the permute method in Python to convert the dimension of the convolution tensor into the input data format of the subsequent GRU layer.

[0052] In the fourth step, the track conversion data is input into a pre-trained aircraft individual identification model to obtain aircraft individual identification information.

[0053] ​In some embodiments, the execution subject can input the track conversion data into a pre-trained aircraft individual identification model to obtain aircraft individual identification information. The pre-trained aircraft individual identification model can be a model pre-trained to take track conversion data as input and aircraft individual identification information as output.

[0054] Optionally, before the track conversion data is input into the pre-trained aircraft individual identification model to obtain the aircraft individual identification information, the method further includes:

[0055] First, a training sample set is obtained. The training samples in the training sample set include sample track data and sample aircraft individual identification information.

[0056] Second, an initial aircraft individual identification model is determined. The initial aircraft individual identification model includes a first initial gating model, a second initial gating model, an initial attention model, a first initial fully connected model, a first initial linear model, an initial activation model, an initial dropout model, a second initial linear model, and a second initial fully connected model.

[0057] The first initial gating model can be a gating recurrent model that takes sample track data as input and outputs first initial gating data. For example, the first initial gating model can be a GRU (Gated Recurrent Unit). For example, the first initial gating model can refer to Figure 3 , Figure 3 A model structure diagram of the first initial gating model of the aircraft individual identification method based on multi-dimensional track features according to the present application is shown. The first initial gating model can include a hidden state update formula, an update gate, a reset gate, and a candidate hidden state.

[0058] The hidden state update formula is:

[0059] ,

[0060] wherein, t represents a time step, h t represents a hidden state at the current time step, o u represents an output of the update gate, h t-1 represents a hidden state at the previous time step, h t represents a candidate hidden state at the current time step, represents element-wise multiplication. Here, the output of the first initial gating model is first initial gating data, and the first initial gating data includes each hidden state .

[0061] The update gate is:

[0062] ,

[0063] wherein, denotes a sigmoid activation function, denotes a weight matrix of the update gate, denotes an input of the current time step.

[0064] wherein, the reset gate is:

[0065] ,

[0066] wherein, denotes an output of the reset gate, denotes a weight matrix of the reset gate.

[0067] wherein, the candidate hidden state is:

[0068] ,

[0069] wherein, denotes a weight matrix.

[0070] The second initial gated model can be a gated recurrent model taking the first initial gated data as input and taking the second initial gated data as output. For example, the second initial gated model can be a GRU (Gated Recurrent Unit). Here, the second initial gated model can refer to the first initial gated model described above, and will not be repeated here.

[0071] The initial attention model can be an attention model taking the second initial gated data as input and taking the initial attention data as output. For example, the initial attention model can be:

[0072] ,

[0073] wherein, denotes a time point, denotes the initial attention data, denotes a first time point hidden state included in the second initial gated data, denotes an attention score (i.e. an unnormalized weight) determined by the first time point hidden state vector , denotes an attention probability distribution value (i.e. a normalized weight) determined by the first time point hidden state vector , and denotes a weight coefficient matrix of the initial attention model at the first time point, denotes the offset corresponding to the time instant, denotes the offset corresponding to the time instant, denotes the normalized exponential function.

[0074] Here, the core logic of the initial attention model includes: 1) learnable attention weight parameters self.attention_weight (i.e., weight coefficient matrix and ); 2) calculating attention scores by matrix multiplication torch.matmul (matrix dot product); 3) normalizing the scores using Softmax (i.e., normalized exponential function) to obtain attention weights; 4) aggregating features by weighted summation x*attention_weights (i.e., ) and sum (dim=1) (denoting summing along the row direction of the matrix).

[0075] The first initial fully connected model can be a model that takes initial attention data as input and outputs initial fully connected data. For example, the first initial fully connected model can be a fully connected layer. Thus, through the first initial fully connected model, the output of the previous Attention layer can be passed through a fully connected network to convert the output of the GRU into the output of the classification task.

[0076] The first initial linear model can be a linear model that takes initial fully connected data as input and outputs first initial linear data. For example, the first initial linear model can be the first linear layer in the FNN (feedforward neural network). Thus, through the first initial linear model, the output of the GRU can be mapped.

[0077] The initial activation model can be an activation model that takes first initial linear data as input and outputs initial activation data. For example, the initial activation model can be a Tanh (hyperbolic tangent function) activation function. Thus, through the initial activation model, nonlinearity can be added.

[0078] The initial dropout model can be a model that takes initial activation data as input and outputs initial dropout data. For example, the initial dropout model can be a Dropout (dropping units in a neural network). Thus, through the initial dropout model, overfitting can be prevented.

[0079] The second initial linear model can be a linear model that takes initial dropout data as input and outputs second initial linear data. For example, the second initial linear model can be the second linear layer in the FNN. Thus, through the second initial linear model, the features can be mapped to the number of classes to obtain the original scores for each class.

[0080] The second initial full connection model can be a model taking the second initial linear data as input and outputting the initial aircraft individual identification information. For example, the second initial full connection model can be a full connection layer. Thus, the initial aircraft individual identification information representing the original scores of each sample belonging to each category can be obtained through the second initial full connection model.

[0081] Thus, the output of the last Attention layer is converted into the output of the classification task through the full connection network. The first linear layer in the FNN is used to map the output of the GRU, and the Tanh activation function is used to increase the nonlinearity. The Dropout layer is added to prevent overfitting. The second linear layer is used to map the features to the number of categories to obtain the original scores of each category. Finally, the output tensor of the full connection network represents the original scores of each sample belonging to each category.

[0082] Step 3: Select target training samples from the above training sample set.

[0083] Step 4: Input the sample track data included in the selected target training samples into the first initial gating model to obtain first initial gating data.

[0084] Step 5: Input the first initial gating data into the second initial gating model to obtain second initial gating data.

[0085] Step 6: Input the second initial gating data into the initial attention model to obtain initial attention data.

[0086] Step 7: Input the initial attention data into the first initial full connection model to obtain initial full connection data.

[0087] Step 8: Input the initial full connection data into the first initial linear model to obtain first initial linear data.

[0088] Step 9: Input the first initial linear data into the initial activation model to obtain initial activation data.

[0089] Step 10: Input the initial activation data into the initial dropout model to obtain initial dropout data.

[0090] Step 11: Input the initial dropout data into the second initial linear model to obtain second initial linear data.

[0091] Step 12: Input the second initial linear data into the second initial full connection model to obtain initial aircraft individual identification information.

[0092] In the thirteenth step, a difference value between the initial aircraft individual identification information and the sample aircraft individual identification information included in the target training sample is determined based on a preset loss function. The preset loss function can be, but is not limited to, a mean square error loss function (MSE), a hinge loss function, a cross entropy loss function (CrossEntropy), a 0-1 loss function, an absolute value loss function, a log loss function, a square loss function, an exponential loss function, and the like.

[0093] In the fourteenth step, in response to determining that the difference value is greater than or equal to a preset difference value, the network parameters of the initial aircraft individual identification model are adjusted.

[0094] Optionally, in response to determining that the difference value is less than the preset difference value, the initial aircraft individual identification model is determined as the aircraft individual identification model.

[0095] Here, the track data set (i.e., the standard track data) constructed by the present application is a multi-classification data set, which contains 4 aircraft targets, namely B737 (short-range twin-jet passenger aircraft), C17-1 (transport aircraft), C17-2 (transport aircraft), and E35L (business aircraft), and each target contains 110 complete tracks, a total of more than 100,000 track points. The number of test samples is m, the number of classifications is 4 (i.e., the classification labels label correspond to one aircraft individual respectively), and these samples and corresponding labels are expressed in matrix form, with each row representing a sample and each column representing whether the sample belongs to the classification, thereby forming a [m, 4] label matrix L as shown in Figure 4 .A schematic diagram of a label matrix of the aircraft individual identification method based on multi-dimensional track features according to the present application is shown in Figure 4 .The ID in the above formula represents the track point sequence number. Figure 4 The sequence number is represented by . The 1 in the label matrix indicates that the track point, i.e., the test sample, belongs to the classification. The 0 in the label matrix indicates that the track point, i.e., the test sample, does not belong to the classification.

[0096] After the model test is completed, the probabilities of the test samples under each classification are calculated, and the model test score is also expressed in a [m, 4] matrix, with each row representing the probability of the sample under each classification, denoted as a probability matrix P as shown in Figure 5 .A schematic diagram of a probability matrix of the aircraft individual identification method based on multi-dimensional track features according to the present application is shown in Figure 5 . The probability value of the i-th track point under the j-th classification is represented by . The sequence number is represented by . .

[0097] For any one test sample, the label is only composed of 0 and 1, and the position of 1 indicates the category of the test sample (i.e. positive in the binary classification problem, i.e. positive sample), and 0 indicates other types (i.e. negative in the binary classification problem, i.e. negative sample). If the classifier (i.e. the aircraft individual identification model) classifies the test sample correctly, the value of the position corresponding to 1 in the sample label in the probability matrix P is greater than the probability value of the position corresponding to 0 (for example, track point 1 is a track point of category 1, and in the case of correct classification, the probability value of track point 1 under category 1 is greater than the probability value under category 2, greater than the probability value under category 3, etc.). Then the label matrix L and the probability matrix P are respectively expanded by row, and after transposition, the final ROC curve (receiver operating characteristic curve, receiver operating characteristic curve) can be directly obtained by calculation.

[0098] After obtaining the ROC curve, the area under the ROC curve, i.e. AUC (Area Under the Curve, Area Under the Curve), is calculated. The value of AUC can intuitively evaluate the goodness of the classifier (i.e. the aircraft individual identification model), the larger the AUC value, the more likely the current classification algorithm will arrange the positive samples in front of the negative samples, thereby better classification. Among them, AUC can refer to the following formula:

[0099] ,

[0100] Among them, represents the serial number, represents the area under the curve, represents the sum of the serial numbers of only positive samples, represents the serial number of the th sample, represents the number of positive samples, represents the number of negative samples.

[0101] According to different feature combinations D1={longitude latitude height speed}, feature combination D2={longitude latitude height speed climb heading}, the AUC value is calculated respectively, and the feature combination with the largest AUC value is selected as the final selected feature.

[0102] The above various embodiments of the present application have the following beneficial effects: by the aircraft individual identification method based on multi-dimensional track features of some embodiments of the present application, a track feature analysis method for effectively processing the dynamic characteristics and multi-dimensional correlation of time series track data is constructed. Specifically, the reason why it is difficult to construct a track feature analysis method for effectively processing the dynamic characteristics and multi-dimensional correlation of time series track data is that: the existing aircraft individual identification technology has problems such as difficulty in obtaining images, less research on measured track data as individual identification features, and traditional classifiers are difficult to effectively process the dynamic characteristics and multi-dimensional correlation of time series track data. Based on this, the aircraft individual identification method based on multi-dimensional track features of some embodiments of the present application first performs data preprocessing on the obtained original track data to generate standard track data. Secondly, the standard track data is processed for feature extraction to generate track feature data. Then, the track feature data is processed for format conversion to generate track conversion data. Then, the track conversion data is input into the pre-trained aircraft individual identification model to obtain aircraft individual identification information. Therefore, the present application proposes an aircraft individual GRU identification algorithm using multi-dimensional track data as feature quantities, which first uses six-dimensional track data of longitude, latitude, height, speed, climb and heading as aircraft individual identification features from the perspective of dynamics and geographic information space, then constructs an evaluation model for track feature selection based on the area under the curve (Area Under the Curve, AUC), and finally combines the attention mechanism (Attention Mechanism) and the gate recurrent unit (Gate Recurrent Unit, GRU) to establish an aircraft individual identification algorithm.

[0103] Further reference Figure 6 As an implementation of the methods shown in the above figures, the present disclosure provides some embodiments of an aircraft individual identification device based on multi-dimensional track features, which correspond to the method embodiments shown in Figure 1 The aircraft individual identification device based on multi-dimensional track features can be applied to various electronic devices.

[0104] As Figure 6As shown, the aircraft individual identification device 600 based on multi-dimensional track features of some embodiments includes a data preprocessing unit 601, a feature extraction unit 602, a format conversion unit 603, and an input unit 604. Among them, the data preprocessing unit 601 is configured to perform data preprocessing on the obtained original track data to generate standard track data; the feature extraction unit 602 is configured to perform feature extraction processing on the above-mentioned standard track data to generate track feature data; the format conversion unit 603 is configured to perform format conversion processing on the above-mentioned track feature data to generate track conversion data; and the input unit 604 is configured to input the above-mentioned track conversion data into a pre-trained aircraft individual identification model to obtain aircraft individual identification information.

[0105] It can be understood that the units described in the aircraft individual identification device 600 based on multi-dimensional track features correspond to the respective steps in the method described above. Figure 1 Therefore, the operations, features and beneficial effects described above for the method also apply to the aircraft individual identification device 600 based on multi-dimensional track features and the units contained therein, which will not be described here.

[0106] The present application also provides a computer device 700. As shown, Figure 7 The computer device 700 includes a bus 701, a processor 702, a memory 703, and a communication interface 704. The processor 702, the memory 703, and the communication interface 704 communicate through the bus 701. The computer device 700 can be a server or a terminal device. It should be understood that the present application does not limit the number of processors and memories in the computer device 700.

[0107] The bus 701 can be a peripheral component interconnect (PCI) bus or an extended industry standard architecture (EISA) bus, etc. The bus can be divided into an address bus, a data bus, a control bus, etc. For ease of representation, Figure 7 only one line is used in the figure, but it does not mean that there is only one bus or one type of bus. The bus 701 can include a path for transmitting information between various components (e.g., the memory 703, the processor 702, the communication interface 704) of the computer device 700.

[0108] The processor 702 can include any one or more of a central processing unit (CPU), a graphics processing unit (GPU), a microprocessor (MP), or a digital signal processor (DSP), among other processors.

[0109] The memory 703 can include volatile memory, such as random access memory (RAM), and non-volatile memory, such as read-only memory (ROM), floppy disks, mechanical hard disks, or solid state hard disks, among others.

[0110] The memory 703 stores executable program code, which the processor 702 executes to implement the functions of the aforementioned data preprocessing unit, feature extraction unit, format conversion unit, and input unit, respectively, thereby implementing the aforementioned aircraft individual identification method based on multi-dimensional track features. That is, the memory 703 stores instructions for executing the aforementioned aircraft individual identification method based on multi-dimensional track features.

[0111] The communication interface 704 uses a transceiver module, such as but not limited to a network interface card or a transceiver, to enable communication between the computer device 700 and other devices or communication networks.

[0112] The embodiments of the present application also provide a chip including a processor and a data interface, the processor reading instructions stored on a memory through the data interface to execute the aforementioned aircraft individual identification method based on multi-dimensional track features.

[0113] The embodiments of the present application also provide a computer readable storage medium. The aforementioned computer readable storage medium can be any available medium or data storage device that can be accessed by a computer device, or a data center including one or more available media, among other data storage devices. The aforementioned available medium can be a magnetic medium, such as a floppy disk, a hard disk, or a magnetic tape, an optical medium, such as a DVD, or a semiconductor medium, such as a solid state hard disk, among others. The computer readable storage medium includes instructions, which instruct the computer device to execute the aforementioned aircraft individual identification method based on multi-dimensional track features.

[0114] Any combination of the technical features in the above-described embodiments can be made, and for the sake of brevity, not all possible combinations are described, however, it is understood that the scope of protection includes all possible combinations.

[0115] The above embodiments are only used to illustrate the technical solutions of the present application, rather than limit the technical solutions; although the present application has been described in detail with reference to the foregoing embodiments, it should be understood by those skilled in the art that the technical solutions recorded in the foregoing embodiments can still be modified, or some technical features can be replaced by equivalents; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the protection scope of the technical solutions of the embodiments of the present application.

Claims

1. An aircraft individual identification method based on multi-dimensional track features, comprising: The acquired raw track data is preprocessed to generate standard track data; The standard track data is subjected to feature extraction processing to generate track feature data; The trajectory feature data is processed to convert its format, thereby generating trajectory conversion data; The trajectory conversion data is input into a pre-trained aircraft individual identification model to obtain aircraft individual identification information.

2. The aircraft individual identification method based on multi-dimensional track features according to claim 1, wherein, The original track data includes: latitude, longitude, altitude, ground speed, rate of climb, and heading angle; The process of preprocessing the acquired raw track data to generate standard track data includes: The latitude, longitude, and altitude data included in the original track data are converted to a new format to generate a position vector. The original track data, including ground speed, rate of climb, and heading angle, are converted to generate a velocity vector. The position vector and the velocity vector are determined as equivalent measurement data; The equivalent measurement data is then processed to remove noise data. The noise data is calibrated to generate noise statistical feature data; Based on the equivalent measurement data, a state equation is constructed; Based on the state equation, the measurement equation is determined; Based on the state equation and the measurement equation, the original track data is smoothed to generate smoothed track data. Based on the noise statistical feature data, the bad point removal process is performed on the smoothed track data to generate track removal data. The track removal data is then converted to a new format to generate standard track data.

3. The aircraft individual identification method based on multi-dimensional track features according to claim 1, wherein, Before inputting the trajectory conversion data into a pre-trained aircraft individual recognition model to obtain aircraft individual recognition information, the method further includes: Obtain a training sample set, wherein the training samples in the training sample set include: sample flight track data and sample aircraft individual identification information; An initial aircraft individual recognition model is determined, which includes: a first initial gating model, a second initial gating model, an initial attention model, a first initial fully connected model, a first initial linear model, an initial activation model, an initial dropout model, a second initial linear model, and a second initial fully connected model. Select target training samples from the training sample set; The selected target training samples, including the sample trajectory data, are input into the first initial gating model to obtain the first initial gating data; The first initial gating data is input into the second initial gating model to obtain the second initial gating data; The second initial gating data is input into the initial attention model to obtain the initial attention data; The initial attention data is input into the first initial fully connected model to obtain the initial fully connected data; The initial fully connected data is input into the first initial linear model to obtain the first initial linear data; The first initial linear data is input into the initial activation model to obtain the initial activation data; The initial activation data is input into the initial dropout model to obtain the initial dropout data; The initial discarded data is input into the second initial linear model to obtain the second initial linear data; The second initial linear data is input into the second initial fully connected model to obtain the initial aircraft individual identification information; Based on a preset loss function, the difference between the initial aircraft individual identification information and the sample aircraft individual identification information included in the target training samples is determined; In response to the determination that the difference value is greater than or equal to the preset difference value, the network parameters of the initial aircraft individual recognition model are adjusted.

4. The aircraft individual identification method based on multi-dimensional track features according to claim 3, wherein, The method further includes: In response to the determination that the difference value is less than the preset difference value, the initial aircraft individual identification model is determined as the aircraft individual identification model.

5. The aircraft individual identification method based on multi-dimensional track features according to claim 3, wherein, The first initial gating model includes: hidden state update formula, update gate, reset gate, and candidate hidden states. The hidden state update formula is: ， in, Indicates a time step. This indicates the hidden state at the current time step. This indicates that the output of the updated gate is being updated. This indicates the hidden state of the previous time step. This represents the candidate hidden state at the current time step. This indicates element-wise multiplication. The update gate is: ， in, This represents the sigmoid activation function. This represents the weight matrix of the updated gate. This represents the input at the current time step. The reset door is: ， in, This indicates that the output of the reset door is being reset. This represents the weight matrix of the reset gate. The candidate hidden states are: ， in, This represents the weight matrix.

6. The aircraft individual identification method based on multi-dimensional track features according to claim 3, wherein, The initial attention model is: ， in, Indicates time, This represents the initial attention data. This indicates that the second initial gating data includes the first... The state vector is hidden at all times. Indicates the first Hidden state vector at all times The determined attention score, Indicates the first Hidden state vector at all times The determined attention probability distribution value, and Indicates the first The initial weight coefficient matrix of the attention model at time step [time]. Indicates the first The corresponding offset at any given time. This represents the normalized exponential function.

7. The aircraft individual identification method based on multi-dimensional track features according to claim 3, wherein, The aircraft individual identification model includes an AUC, where the AUC is: ， in, Indicates the serial number. Represents the area under the curve. This indicates that only the indices of positive samples are added. Indicates the first The serial number of the sample. Indicates the number of positive samples. This indicates the number of negative samples.

8. An aircraft individual identification device based on multi-dimensional track features, comprising: The data preprocessing unit is configured to preprocess the acquired raw track data to generate standard track data; The feature extraction unit is configured to perform feature extraction processing on the standard track data to generate track feature data; The format conversion unit is configured to perform format conversion processing on the track feature data to generate track conversion data; The input unit is configured to input the trajectory conversion data into a pre-trained aircraft individual identification model to obtain aircraft individual identification information.

9. A computer device, wherein, The computer device includes a processor, a memory, and a computer program stored in the memory and executable by the processor, wherein when the computer program is executed by the processor, it implements the steps of the method as described in any one of claims 1-8.

10. A computer-readable storage medium, wherein, The computer-readable storage medium stores a computer program, wherein when the computer program is executed by a processor, it implements the steps of the method as described in any one of claims 1-8.

Images