Learning device, learning method, program, and radar device

Abstract

To provide a radar device capable of performing tracking processing with both tracking accuracy and trackability.SOLUTION: A learning device learns a tracking model used for a radar device. A first acquisition section acquires, from the radar device, target detection information indicating a position of a target detected on the basis of a received wave, and a track of the target calculated on the basis of the target detection information. A second acquisition section acquires target position information indicating a position of the target. A learning data generation section generates learning data using the target detection information, the track, and the target position information. A learning processing section learns a tracking model for performing tracking processing of the target on the basis of the target detection information using the learning data.SELECTED DRAWING: Figure 3

Description

[Technical field]

[0001] The present invention relates to radar-based surveillance technology. [Background technology]

[0002] There is a known technique for monitoring moving objects such as aircraft using a radar. Patent Document 1 discloses a technique for monitoring moving targets such as aircraft and vehicles using a radar device. [Prior art documents] [Patent documents]

[0003] [Patent Document 1] JP 2016-151416 A Summary of the Invention [Problem to be solved by the invention]

[0004] In a radar device with a tracking function, in order to track a highly maneuvering target that makes a sharp turn, it is necessary to use a tracking filter that achieves both tracking accuracy and tracking capability, but it is difficult to actually realize such a tracking filter. In addition, when tracking a target, a correlation process is performed to match the target with a plot, but in areas with a lot of clutter, there is a risk that erroneously detected clutter will be matched with the target, which reduces the tracking accuracy.

[0005] An object of the present invention is to provide a radar device capable of performing tracking processing that achieves both tracking accuracy and tracking capability. [Means for solving the problem]

[0006] In one aspect of the present invention, a learning device comprises: a first acquisition means for acquiring, from the radar device, target detection information indicating a position of an aircraft detected based on a received wave received in response to a transmission wave from the radar device, and a flight path of the aircraft calculated based on the target detection information; A second acquisition means for acquiring target position information indicating a position of the aircraft; a learning data generating means for generating learning data using the target detection information, the flight path, and the target position information; a simulation data generating means for generating simulation data including the target detection information and a label for the target detection information by simulation; a learning processing means for learning a tracking model for performing a tracking process of the aircraft based on the target detection information, using the learning data and the simulation data; and a radar control means for controlling the radar device to selectively transmit either a first transmission wave for tracking the aircraft or a second transmission wave for collecting data corresponding to conditions that are lacking in generating the learning data.

[0007] In another aspect of the invention, a method of learning comprises: acquiring, from the radar device, target detection information indicating a position of an aircraft detected based on a received wave received in response to transmission of a transmission wave from the radar device, and a flight path of the aircraft calculated based on the target detection information; acquiring target position information indicating a position of the aircraft; generating learning data using the target detection information, the flight track, and the target position information; generating simulation data including the target detection information and a label for the target detection information through a simulation; Using the learning data and the simulation data, a tracking model is trained to perform a tracking process for the aircraft based on the target detection information; The radar device is controlled to selectively transmit either a first transmission wave for tracking the aircraft or a second transmission wave for collecting data corresponding to conditions that are lacking in generating the learning data.

[0008] In another aspect of the invention, a program includes: acquiring, from the radar device, target detection information indicating a position of an aircraft detected based on a received wave received in response to transmission of a transmission wave from the radar device, and a flight path of the aircraft calculated based on the target detection information; acquiring target position information indicating a position of the aircraft; generating learning data using the target detection information, the flight track, and the target position information; generating simulation data including the target detection information and a label for the target detection information through a simulation; Using the learning data and the simulation data, a tracking model is trained to perform a tracking process for the aircraft based on the target detection information; The computer is caused to execute a process of controlling the radar device to selectively transmit as the transmission wave either a first transmission wave for tracking the aircraft or a second transmission wave for collecting data corresponding to conditions that are lacking in generating the learning data.

[0009] In another aspect of the present invention, a radar system includes: a target detection means for transmitting a transmission wave and generating target detection information indicating a position of an aircraft based on a received wave corresponding to the transmission wave; a tracking processing means for performing a tracking process of the aircraft based on the target detection information, using a tracking model that has been trained using simulated data obtained by a simulation for generating data including the target detection information, a flight path of the aircraft calculated based on the target detection information, and target position information indicating a position of the aircraft, and the target detection information and a label for the target detection information; The radar control means is provided with a first transmission wave for tracking the aircraft and a second transmission wave for collecting data corresponding to conditions that are lacking in generating the learning data, and the radar control means performs control to selectively transmit either of these as the transmission wave. Effect of the Invention

[0010] According to the present invention, it is possible to realize a radar device capable of performing tracking processing that achieves both tracking accuracy and tracking capability. [Brief description of the drawings]

[0011] [Figure 1] The basic configuration of a radar device is shown below. [Diagram 2] 2 shows the configuration of a signal processing unit. [Diagram 3] The configuration for learning a tracking model is shown below. [Figure 4] 2 shows the hardware configuration of a learning device. [Diagram 5] 13 is a flowchart of a learning process performed by the learning device. [Figure 6] 1 shows another example of a learning device. [Figure 7] The configuration of a radar device to which a trained model is applied is shown. [Figure 8] 4 is a flowchart of a target detection process performed by a radar device. [Figure 9] The configuration of another radar device to which a trained model is applied is shown. [Figure 10] The configuration for performing beam control for collecting learning data is shown. [Figure 11] 1 shows a configuration for online learning. [Figure 12] The configuration for evaluating the validity of a trained model is shown below. [Figure 13] This shows a configuration for suppressing behavioral fluctuations using a trained model. [Figure 14] 10 shows the configurations of a learning device and a radar device according to a second embodiment. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0012] A preferred embodiment of the present invention will be described below with reference to the drawings. A radar device in this embodiment can be used in a monitoring system for surrounding moving objects. Specifically, the radar device detects moving objects (hereinafter also referred to as "targets") by emitting transmission waves into the surroundings and receiving the reflected waves, and tracks the targets as necessary. Targets include, for example, aircraft flying in the air, vehicles moving on the ground, and ships moving on the sea. In the following embodiment, for convenience of explanation, it is assumed that the radar device is used for air traffic control, and the targets are mainly aircraft.

[0013] <Basic configuration of radar equipment> First, the basic configuration of a radar device will be described. Fig. 1 is a block diagram showing the basic configuration of a radar device. The radar device 100 includes an antenna unit 101, a transmitting / receiving unit 102, a signal processing unit 103, a beam control unit 104, a target detection unit 105, a tracking processing unit 106, and a display operation unit 107.

[0014] The antenna unit 101 amplifies an electrical signal (hereinafter also referred to as a "transmission signal") input from the transmission / reception unit 102, and emits a transmission wave (hereinafter referred to as a "beam") in a transmission direction instructed by the beam control unit 104. The antenna unit 101 also receives a wave reflected by a target of the emitted transmission wave, converts it into an electrical signal (hereinafter also referred to as a "received signal"), combines it, and outputs it to the transmission / reception unit 102.

[0015] In this embodiment, the radar device 100 emits a beam (called a "scan beam") that constantly scans in all directions (360° around) to monitor the presence of a target in the vicinity. When a target is detected, the radar device 100 emits a beam (called a "tracking beam") for tracking the target and tracks the trajectory (called a "wake") of the target. From this point of view, the antenna unit 101 is configured with an antenna that can instantly change the transmission direction, such as an array antenna having multiple antenna elements. Specifically, multiple planar array antennas are arranged to cover all directions, or a cylindrical array antenna can be used. This makes it possible to constantly emit a scan beam in all directions, while emitting a tracking beam in the direction of the target when the target is detected.

[0016] The transmitting / receiving unit 102 generates an electric signal based on transmission wave specifications (hereinafter also referred to as "beam specifications") instructed by the beam control unit 104, and outputs the signal to the antenna unit 101. The beam specifications include the pulse width and transmission timing of the transmission wave. The transmitting / receiving unit 102 also performs A / D conversion on the reception signal input from the antenna unit 101, removes unnecessary frequency bands, and then outputs the signal to the signal processing unit 103 as a reception signal.

[0017] The signal processing unit 103 performs demodulation processing, integration processing, etc. on the received signal input from the transmitting / receiving unit 102, and outputs the processed received signal (hereinafter also referred to as the "processed signal") to the target detection unit 105. FIG. 2 is a block diagram showing the configuration of the signal processing unit 103. The signal processing unit 103 includes a demodulation processing unit 110 and a coherent integration unit 111. The demodulation processing unit 110 demodulates (pulse compresses) the received signal input from the transmitting / receiving unit 102. Basically, to detect a distant target using a radar, a high-power and sharp transmission wave (transmission pulse) is required, but there is a limit to increasing the power due to restrictions such as hardware. Therefore, when emitting a beam, the transmitting / receiving unit 102 frequency-modulates a transmission signal with a predetermined pulse width to generate a transmission wave with a long duration, and transmits it from the antenna unit 101. In response to this, the demodulation processing unit 110 demodulates the received signal input from the transmitting / receiving unit 102 to generate a sharp received pulse, and outputs it to the coherent integration unit 111 .

[0018] The coherent integrator 111 coherently integrates multiple pulses input from the demodulation processor 110 to remove noise and improve the SNR. The radar device 100 emits multiple pulses in the same direction (same azimuth and elevation angle) to detect targets with high accuracy. The number of pulses emitted in the same direction is called the "number of hits." The coherent integrator 111 integrates the received signals (received pulses) of a predetermined number of beams emitted in the same direction to improve the SNR of the received signals. The number of received pulses integrated by the coherent integrator 111 is also called the "number of integrated pulses." The number of integrated pulses is basically equal to the number of hits of the emitted beam.

[0019] Returning to FIG. 1, target detection unit 105 detects a target from the processed signal input from signal processing unit 103 using a predetermined threshold value. Target detection unit 105 measures the distance, direction, and elevation angle of the target, and outputs these to tracking processing unit 106 as target detection results (hereinafter referred to as "plot"). The plot includes the distance, direction, elevation angle, SNR, etc. of the target. Furthermore, target detection unit 105 sets a threshold value for detecting the target based on a threshold setting value input from display operation unit 107.

[0020] The tracking processing unit 106 performs tracking processing on the multiple plots input from the target detection unit 105, and calculates the trajectory of the target. Specifically, the tracking processing unit 106 estimates the position of the target at the current time (called the "estimated target position") based on the multiple plots, and outputs it to the display operation unit 107. The tracking processing unit 106 also calculates a predicted position of the target (called the "predicted target position") based on the multiple plots, and outputs it to the beam control unit 104. The predicted target position indicates the position where the radar device 100 will next cast a tracking beam.

[0021] In detail, the tracking processing unit 106 performs correlation processing and tracking filter processing. The correlation processing is a processing for associating multiple plots obtained by the target detection unit 105 with targets. When multiple targets are detected simultaneously, it is determined which target each of the multiple obtained plots corresponds to, and each plot is associated with each target. The tracking filter processing uses the plots associated with each target to calculate the trajectory of the target. This makes it possible to obtain an estimated target position indicating the current position of the target, and a predicted target position indicating the predicted position of the target in the future.

[0022] The beam control unit 104 determines the transmission direction and beam specifications of the scan beam according to a preset beam schedule. Also, the beam control unit 104 determines the transmission direction and beam specifications of the tracking beam based on the predicted target position input from the tracking processing unit 106. Then, the beam control unit 104 outputs the transmission direction of the scan beam and the tracking beam to the antenna unit 101, and outputs the beam specifications of the scan beam and the tracking beam to the transmission / reception unit 102.

[0023] The display operation unit 107 includes a display unit such as a display, and an operation unit such as a keyboard, a mouse, and operation buttons. The display operation unit 107 displays the positions of the multiple plots input from the target detection unit 105 and the estimated target positions input from the tracking processing unit 106. This allows the operator to see the current positions and tracks of the detected targets. In addition, the operator can input a threshold value used for target detection to the target detection unit 105 and input a clutter determination result used by the signal processing unit 103 for demodulation processing to the signal processing unit 103 by operating the display operation unit 107. Note that "clutter" is a signal generated when an emitted radar is reflected by an object other than the target. The operator can determine an area that is considered to be clutter from experience among the multiple plots displayed on the display operation unit 107, and specify the area by operating the display operation unit 107. This is called "clutter determination".

[0024] With the above configuration, the radar device 100 steadily emits scan beams in all directions to detect a target, and when a target is detected, emits a tracking beam to a predicted target position to track the target.

[0025] First Embodiment In a radar device having a tracking function, in order to track a highly maneuverable target that makes a sharp turn, it is necessary to use a tracking filter that achieves both tracking accuracy and tracking ability, but it is difficult to actually realize such a tracking filter. In addition, when tracking a target, a correlation process is performed to associate the target with a plot, but in an area with a lot of clutter, there is a risk that erroneously detected clutter will be associated with the target, which reduces the tracking accuracy. Therefore, in this embodiment, a tracking process is performed using a model generated by machine learning. Specifically, a tracking model is learned using a plot obtained by the target detection unit 105 and a teacher label (correct answer label) for the plot, and the learned tracking model is applied to the tracking processing unit. This makes it possible to improve detection performance while suppressing costs.

[0026] [Study configuration] (Overall composition) 3 shows the configuration of a radar device 300 and a learning device 200 according to this embodiment. When learning a tracking model, a radar device 300 including a primary radar (PSR: Primary Surveillance Radar) and a secondary radar (SSR: Secondary Surveillance Radar) is used. As shown in the figure, the radar device 300 includes a PSR antenna unit 301, a PSR transceiver unit 302, a PSR signal processing unit 303, a beam control unit 304, a PSR target detection unit 305, a tracking processing unit 306, and a display operation unit 307. These have the same configuration as the antenna unit 101, the transceiver unit 102, the signal processing unit 103, the beam control unit 104, the target detection unit 105, the tracking processing unit 106, and the display operation unit 107 in FIG. 1, respectively, and operate in the same manner. The radar device 300 further includes an SSR antenna unit 308, an SSR transceiver unit 309, and an SSR target detection unit 310.

[0027] The SSR transmission / reception unit 309 outputs an interrogation signal to the SSR antenna unit 308, and the SSR antenna unit 308 transmits an interrogation wave to the target. The SSR antenna unit 308 also receives a response wave from the target in response to the interrogation wave, and outputs the response signal to the SSR transmission / reception unit 309. The SSR transmission / reception unit 309 performs A / D conversion of the response signal and outputs it to the SSR target detection unit 310. The response signal usually includes position information of the target, and the SSR target detection unit 310 generates a target plot (called an "SSR plot") D2 based on the response signal, and outputs it to the tracking processing unit 306. The tracking processing unit 306 generates a target track D3 using the target plot (called a "PSR plot") D1 detected by the PSR target detection unit 305 and the SSR plot D2. The PSR plot is an example of a primary radar plot, and the SSR plot is an example of a secondary radar plot.

[0028] The learning device 200 is provided to learn a tracking model to be applied to the tracking processing unit. The learning device 200 includes a learning data generating unit 201, a data collecting unit 202, and a learning processing unit 204. The learning data generating unit 201 receives a PSR plot D1 from the PSR target detecting unit 305, an SSR plot D2 from the SSR target detecting unit 310, and a track D3 from the tracking processing unit 306. The learning data generating unit 201 generates a teacher label related to the track of the target using the SSR plot D2 and the track D3. Specifically, the teacher label includes the position, speed, acceleration, and true target / false target (true or false target) of the target. Note that the "true target" refers to a correct target such as an aircraft, and the "false target" refers to an object that is erroneously recognized as a target, such as clutter.

[0029] As described above, the tracking processing unit 306 performs correlation processing and tracking filter processing as tracking processing, so the tracking model to be learned is also configured as a model that performs correlation processing and tracking filter processing. The learning data generating unit 201 first uses the SSR plot D2 to generate a teacher label indicating whether each PSR plot is a true target or a false target. Since the SSR plot D2 is obtained for a target that responded to a query signal, the learning data generating unit 201 basically assigns a teacher label of "true target" to the PSR plot D1 with the corresponding SSR plot D2, and assigns a teacher label of "false target" to the PSR plot D1 without the corresponding SSR plot D2. By learning using the teacher labels of the true target / false target thus generated, the tracking model becomes able to determine whether the input PSR plot D1 is a true target or a false target. This makes it possible to prevent a plot obtained by erroneously detecting clutter, etc., from being associated with the target in the correlation processing, and enables stable tracking.

[0030] Furthermore, the learning data generating unit 201 generates teacher labels such as the position, speed, and acceleration of the target for each PSR plot D1 based on the SSR plot D2 and the trajectory D3. When a target such as an aircraft transmits a response signal including its own position, the self-position can be used as a fairly accurate position of the target. Furthermore, the learning data generating unit 201 can obtain the speed, acceleration, and the like of the target corresponding to the PSR plot D1 based on the trajectory D3. By learning using the teacher labels such as the position, speed, and acceleration of the target thus generated, the tracking model can learn the movements of various targets, enabling tracking processing that achieves both tracking accuracy and tracking ability.

[0031] The learning data generation unit 201 outputs pairs of PSR plots and corresponding teacher labels generated as described above as learning data to the data collection unit 202. The data collection unit 202 stores the learning data input from the learning data generation unit 201. The data collection unit 202 stores learning data to which the teacher labels of the target position, velocity, acceleration, and true target / false target are assigned for each PSR plot. The learning processing unit 204 acquires the learning data from the data collection unit 202, learns a tracking model, and generates a learned tracking model.

[0032] (Hardware configuration of the learning device) Fig. 4 is a block diagram showing a hardware configuration of the learning device 200 shown in Fig. 3. As shown in the figure, the learning device 200 includes an input IF (Interface) 21, a processor 22, a memory 23, a recording medium 24, and a database (DB) 25.

[0033] The input IF 21 inputs and outputs data to and from the radar device 300. Specifically, the input IF 21 acquires a PSR plot D1, an SSR plot D2, and a track D3 from the radar device 300. The processor 22 is a computer including a CPU (Central Processing Unit), a GPU (Graphics Processing Unit), and the like, and controls the entire learning device 200 by executing a program prepared in advance. The processor 22 functions as a learning data generation unit 201 and a learning processing unit 204 shown in FIG. 3.

[0034] The memory 23 is composed of a ROM (Read Only Memory), a RAM (Random Access Memory), etc. The memory 23 stores various programs executed by the processor 22. The memory 23 is also used as a working memory while the processor 22 is executing various processes.

[0035] Recording medium 24 is a non-volatile, non-temporary recording medium such as a disk-shaped recording medium or a semiconductor memory, and is configured to be detachable from learning device 200. Recording medium 24 records various programs executed by processor 22. When learning device 200 executes processing, the programs recorded on recording medium 24 are loaded into memory 23 and executed by processor 22.

[0036] The DB 25 stores data input through the input IF 21 and data generated by the learning device 200. Specifically, the DB 25 stores the PSR plot D1, the SSR plot D2, and the track D3 input from the radar device 300, and the learning data generated by the learning data generation unit 201.

[0037] (Learning process) 5 is a flowchart of the learning process by the learning device 200. This process can be realized by the processor 22 shown in FIG. 4 executing a program prepared in advance and operating as each element shown in FIG.

[0038] First, the learning data generating unit 201 acquires the PSR plot D1 output by the PSR target detecting unit 305 of the radar device 300 (step S11). The learning data generating unit 201 also acquires the SSR plot D2 output by the SSR target detecting unit 310 and the track D3 output by the tracking processing unit 306 (step S12). Next, the learning data generating unit 201 uses the PSR plot D1, the SSR plot D2, and the track D3 to generate learning data including the PSR plot D1 and the corresponding teacher label, and stores the data in the data collecting unit 202 (step S13). Next, the learning processing unit 204 learns a tracking model using the input learning data (step S204).

[0039] Next, the learning processing unit 204 judges whether or not a predetermined learning end condition is met (step 15). One example of the learning end condition is that learning using a predetermined amount of learning data or a predetermined number of learning times has been performed. The learning processing unit 204 repeats learning until the learning end condition is met, and ends the process when the learning end condition is met.

[0040] (Another example of a learning device) FIG. 6 shows another example of the learning device. The learning device 200 shown in FIG. 3 generates learning data using actual data generated in the actual radar device 300, such as PSR plots, SSR plots, and wakes. Instead, learning data may be prepared by simulation and learning may be performed. FIG. 6 shows a learning device 200a using simulated data, and includes a simulated data generating unit 205 and a learning processing unit 204. The simulated data generating unit 205 generates learning data (simulated data) including PSR plots in various situations and teacher labels corresponding to the PSR plots by simulation, and inputs the learning data to the learning processing unit 204. The learning processing unit 204 is basically the same as that shown in FIG. 3, and learns a tracking model using the simulated data. Note that the example of FIG. 3 and this example may be combined to learn one tracking model using both the learning data generated based on the actual data of the radar device 300 and the simulated data.

[0041] [Radar device using a tracking model] (composition) Fig. 7 is a block diagram showing the configuration of a radar device 300x to which a learned tracking model is applied. As can be seen by comparing with Fig. 3, the radar device 300x includes a tracking processing unit 314 instead of the tracking processing unit 306 in Fig. 3. The configuration other than the tracking processing unit 314 is the same as that in Fig. 3, so a description thereof will be omitted.

[0042] A learned tracking model generated by the above learning process is set in the tracking processing unit 314. The tracking processing unit 314 detects a target from the input PSR plot using the learned tracking model. Specifically, the tracking processing unit 314 calculates the trajectory of the target from the PSR plot using the tracking model. Then, the tracking processing unit 314 outputs the estimated target position to the display operation unit 307, and outputs the predicted target position to the beam control unit 304.

[0043] (Tracking process) 8 is a flowchart of the tracking process by the radar device 300x. First, the tracking processor 314 acquires a PSR plot from the PSR target detector 305 (step S21). Next, the tracking processor 314 calculates the target's track based on the acquired PSR plot using a learned tracking model (step S22), and outputs the track of the detected target (step S23).

[0044] As described above, in this embodiment, by learning using teacher labels of true targets / false targets for the PSR plot, the tracking processing unit 314 can determine whether the input PSR plot is a true target or a false target. This makes it possible to prevent a plot obtained by erroneously detecting clutter or the like from being associated with a target in correlation processing. In addition, by learning using teacher labels of the position, speed, acceleration, and the like of the target for the PSR plot, the tracking processing unit 314 can learn the movements of various targets, making it possible to achieve both tracking accuracy and tracking performance.

[0045] In the example of Fig. 7, the trained tracking model is applied to a radar device equipped with an SSR as shown in Fig. 3, but instead, it may be applied to a radar device having only a PSR as shown in Fig. 1. Fig. 9 shows an example in which the trained tracking model is applied to a radar device 100x having only a PSR. As can be understood by comparing with Fig. 1, it is the same as the radar device 100 shown in Fig. 1, except that a trained tracking model is used in the tracking processing unit 114.

[0046] [Generate training data] (1) SNR deterioration The tracking model may be trained using training data with a lowered SNR, for example by intentionally adding noise to the PSR plots used as input data when training the tracking model. This makes it possible to generate a tracking model with high accuracy even in a low SNR environment.

[0047] (2) How to generate teacher labels When generating learning data in the learning device 200, teacher labels indicating the target position and true / false target can be generated by the following method.

[0048] (A) Use of secondary radar As described in the above embodiment, first, when a target responds to an interrogation signal transmitted from a secondary radar, the learning data generating unit 201 can generate a teacher label based on the response signal received from the target. Specifically, the learning data generating unit 201 may use a PSR plot corresponding to the target that has transmitted a response signal as a true target, and use the target's own position included in the response signal as the position of the target.

[0049] When using a secondary radar, it is not guaranteed that responses will be obtained from all targets. In the case of air defense radar, aircraft detected as targets include military aircraft as well as passenger aircraft. Aircraft whose identities have been identified, such as passenger aircraft and one's own country's military aircraft (hereafter referred to as "friendly aircraft"), respond to interrogation signals, but aircraft whose identities cannot be identified, such as military aircraft from other countries (hereafter referred to as "unknown aircraft"), do not respond to interrogation signals, so teacher labels cannot be generated for received signals that include unknown aircraft as targets. However, in the case of air defense radar, it is the unknown aircraft that one really wants to detect and track.

[0050] Therefore, the correct answer for unknown aircraft is given using the following method. As a premise, there are three classes of targets that can be detected as targets from the received signal: "clutter (including noise)", "friendly aircraft", and "unknown aircraft". Note that a target is considered to be "clutter" when the target does not actually exist but is mistakenly detected as clutter. Here, the only class that responds to the secondary radar's interrogation signal and has been assigned a teacher label (given the correct answer) is "friendly aircraft". Furthermore, since both "unknown aircraft" and "friendly aircraft" are actually aircraft, the characteristics of the received signals are similar.

[0051] Based on the above assumptions, the unknown aircraft is detected in the following procedure. First, the learning data generating unit 201 uses the received signals of the "friendly aircraft" to create a model that extracts the received signals of "clutter," "friendly aircraft," and "unknown aircraft" from all received signals (Process 1). Next, the learning data generation unit 201 determines, among the extracted received signals of "clutter," "friendly aircraft," and "unknown aircraft," those that have not been determined to be an "friendly aircraft" (i.e., received signals determined to be "clutter" or an "unknown aircraft") and have characteristics similar to those of an "friendly aircraft," as an "unknown aircraft," and generates an "unknown aircraft label" (Process 2). Then, the learning data generation unit 201 creates a model for detecting an "unknown aircraft" from a received signal, using the received signal determined to be an "unknown aircraft" and the "unknown aircraft label" (process 3). This model makes it possible to detect unknown aircraft that do not respond to the secondary radar's interrogation signals from the received signals.

[0052] In reality, the accuracy of the "unknown aircraft label" generated in the above process 2 may be an issue. In that case, the "unknown aircraft label" may be manually assigned by an operator or other human being. According to this method, it is sufficient to manually assign the label "unknown aircraft" only to "received signals having characteristics similar to those of a "friendly aircraft" among the received signals determined to be "clutter" or "unknown aircraft" extracted in the above process 2. In other words, it is sufficient to narrow down the received signals that are likely to be "unknown aircraft" by processes 1 and 2, and then manually assign the label. Therefore, the amount of manual work can be significantly reduced compared to the case where labels are manually assigned to received signals that include all of "clutter", "friendly aircraft", and "unknown aircraft".

[0053] (B) Use of other radar devices In the above embodiment, the target position is acquired using SSR, and the teacher label is generated. However, when there are multiple radar devices, the learning data generating unit 201 may generate the teacher label using plots and tracks obtained from other radar devices. The learning data generating unit 201 may also generate the teacher label using the track of a passive radar that only receives (passive track). Note that the "passive track" is the result of tracking a jamming transmitter based on jamming waves, and the learning data generating unit 201 can use this to generate the estimated position of the jamming transmitter as a teacher label.

[0054] (C) Use of other measuring devices If the target aircraft is equipped with a positioning device such as GPS, the output may be received to generate the teacher label. The same applies if the target is a drone. Also, a stereo camera or the like may be used to estimate the target's position from a photographed image of the target to generate the teacher label. If the target is a ship, ship information may be received from an Automatic Identification System (AIS) to obtain the target's position and generate the teacher label.

[0055] (D) Manual assignment The operator may assign teaching labels by looking at the plots and tracks displayed on the display operation unit 107.

[0056] (3) In the case of radar mounted on a moving object In the above embodiment, it is assumed that the radar device is installed on the ground, but the method of the present embodiment can also be applied to a radar device mounted on a moving object such as an aircraft or a ship. In that case, moving object information (position, attitude, speed, course, etc. of the moving object itself) on the moving object on which the radar device is mounted may be used as an input parameter used by the tracking model. Specifically, the moving object information is input to the learning data generating device 201, and the learning processing unit 204 learns the model using the moving object information in addition to the PSR plot as learning data. In addition, in the radar device 100x or 300x to which the learned model is applied, the moving object information is input to the tracking processing unit 114 or 314, and the tracking processing unit 114 or 314 performs tracking processing using the moving object information.

[0057] (4) Improving the efficiency of data collection using radar equipment As described above, it is difficult to collect the learning data required for learning the tracking model for a situation that occurs only rarely. Therefore, the radar device 300 performs beam control for collecting learning data between beam schedules. In particular, when a condition specified in advance is met, the radar device 300 performs beam control with a focus. The content of the beam control is changed according to the data to be collected.

[0058] FIG. 10 shows a configuration when performing beam control for collecting learning data. The radar device 300 has the same configuration as FIG. 3. On the other hand, the learning device 200 includes a data collection control unit 215 in addition to the configuration shown in FIG. 3. The data collection control unit 215 stores the conditions under which learning data is insufficient, and outputs a data collection request D5 including the conditions of data to be collected to the beam control unit 304 of the radar device 300. The beam control unit 304 controls the PSR antenna unit 301 during the interval of the beam schedule to emit a beam under the conditions shown in the data collection request D5. The radar device 300 constantly monitors all directions with a scan beam, and when a target is detected, tracks the target with a tracking beam. Therefore, the beam control unit 304 can emit a beam for collecting learning data, for example, when a target is not detected or when there is no need to track the target. The reflected wave corresponding to the emitted beam is received by the PSR antenna unit 301, and the received signal is input to the PSR target detection unit 305 via the PSR transmission / reception unit 302 and the PSR signal processing unit 303. The PSR plot generated by the PSR target detection unit 305 is output to the training data generation unit 201. In this way, the training device 200 can collect data corresponding to missing conditions.

[0059] [Apply trained model] (Online learning) When the trained tracking model (hereinafter simply referred to as the "trained model") generated by the learning device 200 is actually applied to a radar device, program rewriting and the like occurs, so it is necessary to stop the operation of the radar device 100. However, since the operation of a radar device performing important monitoring cannot be stopped, the trained model cannot be applied, making online learning difficult.

[0060] Therefore, the control / data processing unit of the radar device is duplicated in advance. For convenience of explanation, the following describes a case where the trained model is applied to a radar device having only a PSR radar. FIG. 11 shows the configuration of a radar device and a learning device for performing online learning. As shown in the figure, the radar device 100a includes an antenna unit 101, a transmission / reception unit 102, a switching unit 120, and two systems of control / data processing units 121a and 121b. The control / data processing units 121a and 121b are units including the signal processing unit 103, the beam control unit 104, the target detection unit 105, the tracking processing unit 106, and the display operation unit 107 of the radar device shown in FIG. 1. The switching unit 120 selectively connects one of the control / data processing units 121a and 121b to the antenna unit 101 and the transmission / reception unit 102. Furthermore, the switching unit 120 outputs data D6 including received signals, plots, tracks, and the like from the control / data processing unit 121a or 121b in operation to the learning data generating unit 201 of the learning device 200a.

[0061] The learning device 200a includes a learning data generation unit 201, a data collection unit 202, a learning processing unit 204, a learning result evaluation unit 220, and a learning result application unit 221. The learning result evaluation unit 220 evaluates the trained model generated by the learning processing unit 204, and outputs the trained model determined to be applicable to the radar device 100a to the learning result application unit 221. The learning result application unit 221 applies the trained model determined to be applicable to the control / data processing units 121a and 121b.

[0062] Assume now that the control / data processing unit 121a is in an active state (actual monitoring operation in progress), and the control / data processing unit 121b is in a standby state. That is, the switching unit 120 connects the control / data processing unit 121a to the antenna unit 101 and the transmitting / receiving unit 102. In this case, the learning device 200a learns a tracking model using data D6 output from the control / data processing unit 121a in the active state. Meanwhile, the learning result application unit 221 applies the learned model determined to be applicable to the control / data processing unit 121b in the standby state, and rewrites the program.

[0063] Next, the switching unit 120 sets the control / data processing unit 121b to an active state, sets the control / data processing unit 121a to a standby state, and applies the new trained model to the control / data processing unit 121a in the standby state. In this way, it is possible to train a tracking model while continuing the monitoring operation in one of the control / data processing units 121, and apply the trained model to the other. That is, it is possible to apply the trained model and perform online learning.

[0064] (Model validity assessment) In online learning, it is difficult to judge whether the radar has the appropriate function, i.e., its validity, no matter how much learning is done. In addition, a tracking processing unit to which a trained model is applied may perform unexpected operations, such as falsely detecting clutter as a target that would not be falsely detected by conventional processing, and recovery from such operations is required. Therefore, the validity of the trained model is judged by operating a control / data processing unit to which a trained model is applied and a control / data processing unit that performs conventional processing in parallel and comparing the processing results of these.

[0065] FIG. 12 shows the configuration of a radar device and a learning device for evaluating the validity of a trained model. As shown in the figure, the radar device 100b includes an antenna unit 101, a transceiver unit 102, a validity evaluation unit 130, and two control / data processing units 131 and 132. The control / data processing unit 131 performs conventional processing, and the control / data processing unit 132 performs processing using a trained model. The control / data processing units 131 and 132 are units including the signal processing unit 103, beam control unit 104, target detection unit 105, tracking processing unit 106, and display operation unit 107 of the radar device shown in FIG. 1. The learning device 200a is the same as that shown in FIG. 11.

[0066] The validity evaluation unit 130 compares the processing result of the conventional processing by the control / data processing unit 131 with the processing result of the trained model by the control / data processing unit 132, and judges the validity of the processing result of the trained model. When it is judged that the processing result of the trained model is not valid, the validity evaluation unit 130 outputs the processing result of the conventional processing to the antenna unit 101 and the transmission / reception unit 102. On the other hand, when it is judged that the processing result of the trained model is valid, the validity evaluation unit 130 outputs the processing result of the trained model to the antenna unit 101 and the transmission / reception unit 102. Note that even when it is judged that the processing result of the trained model is valid, the validity evaluation unit 130 may interpolate the processing result of the trained model with the processing result of the conventional processing to prevent unexpected operation from occurring. In addition, the validity evaluation unit 130 may be generated using machine learning or the like. Furthermore, the processing of the validity evaluation unit 130 may not be fully automatic, but may be performed with the intervention of an operator. For example, an operator may determine the validity of the processing results of the trained model based on the information displayed on the display operation unit 107.

[0067] (Suppression of behavioral variability when using a trained model) When the trained model is applied to the target detection unit, there are cases where the operation of the radar device 100 changes significantly. Therefore, the control / data processing unit of the radar device 100 is duplicated in advance, the trained model is applied with an intentional time lag, and the integrated processing results of the two control / data processing units are adopted as the official processing result.

[0068] FIG. 13 shows the configuration of a radar device and a learning device for suppressing the operation variation due to the trained model. As shown in the figure, the radar device 100c includes an antenna unit 101, a transceiver unit 102, an integration unit 140, and two control / data processing units 141a and 141b. The control / data processing unit 141a uses the old model, and the control / data processing unit 141b performs processing using the new model. The control / data processing units 141a and 141b are units including the signal processing unit 103, the beam control unit 104, the target detection unit 105, the tracking processing unit 106, and the display operation unit 107 of the radar device shown in FIG. 1. The learning device 200a is the same as that shown in FIG. 11.

[0069] The integrator 140 adopts the integrated processing results from the control / data processors 141a and 141b as a formal processing result. For example, the integrator 140 adds the processing results from the control / data processors 141a and 141b and divides the result by 2 to adopt the result as the processing result. This makes it possible to suppress large fluctuations in the operation of the radar device when a new trained model is applied.

[0070] <Second embodiment> FIG. 14(A) is a block diagram showing the functional configuration of a learning device according to the second embodiment. The learning device 50 of the second embodiment includes a first acquisition unit 51, a second acquisition unit 52, a learning data generation unit 53, and a learning processing unit 54. The first acquisition unit 51 acquires target detection information indicating the position of a target detected based on a received wave, and the track of the target calculated based on the target detection information, from a radar device. The second acquisition unit 52 acquires target position information indicating the position of the target. The learning data generation unit 53 generates learning data using the target detection information, the track, and the target position information. The learning processing unit 54 uses the learning data to learn a tracking model that performs target tracking processing based on the target detection information.

[0071] 14B is a block diagram showing the functional configuration of a radar device according to the second embodiment. The radar device 60 includes a target detection unit 61 and a tracking processing unit 62. The target detection unit 61 transmits a transmission wave and generates target detection information indicating the position of the target based on a received wave corresponding to the transmission wave. The tracking processing unit 62 performs target tracking processing based on the target detection information using a tracking model that has been trained using learning data generated based on the target detection information, the target's trajectory calculated based on the target detection information, and target position information indicating the target's position.

[0072] A part or all of the above-described embodiments can be described as, but is not limited to, the following supplementary notes.

[0073] (Appendix 1) a first acquisition unit that acquires, from a radar device, target detection information indicating a position of a target detected based on a received wave and a trajectory of the target calculated based on the target detection information; A second acquisition unit that acquires target position information indicating a position of the target; a learning data generating unit that generates learning data using the target detection information, the flight path, and the target position information; a learning processing unit that uses the learning data to learn a tracking model that performs tracking processing of the target based on the target detection information; A learning device comprising:

[0074] (Appendix 2) 2. The learning device according to claim 1, wherein the learning data generation unit generates a teacher label indicating whether the target indicated by the target detection information is a true target or a false target, based on the target position information.

[0075] (Appendix 3) 3. The learning device according to claim 1, wherein the learning data generation unit generates a teacher label including a position, a speed, and an acceleration of the target indicated by the target detection information, based on the target position information and the trajectory.

[0076] (Appendix 4) the target detection information is a primary radar plot generated by a primary radar of the radar device; 4. The learning device according to claim 1, wherein the target position information is a secondary radar plot generated by a secondary radar.

[0077] (Appendix 5) 5. The learning device according to claim 4, wherein the secondary radar plot is generated by a radar device that generated the primary radar plot.

[0078] (Appendix 6) 5. The learning device according to claim 4, wherein the secondary radar plot is generated by a radar device different from a radar device that generated the primary radar plot.

[0079] (Appendix 7) 7. The learning device according to claim 1, wherein the target position information is self-position information of the target that is included in a received signal from the target.

[0080] (Appendix 8) 8. The learning device according to claim 1, wherein the target position information is generated based on images of the target captured by a stereo camera.

[0081] (Appendix 9) 10. The learning device according to claim 1, further comprising a request unit that requests the radar device to transmit a transmission wave that meets a predetermined condition and generate the target detection information corresponding to the condition.

[0082] (Appendix 10) The radar device is mounted on a moving object, a third acquisition unit that acquires mobile object information including position and movement information of the mobile object; 10. The learning device according to claim 1, wherein the learning data generation unit generates the learning data by using the moving object information.

[0083] (Appendix 11) acquiring, from a radar device, target detection information indicating a position of a target detected based on a received wave and a trajectory of the target calculated based on the target detection information; acquiring target position information indicating a position of the target; generating learning data using the target detection information, the flight track, and the target position information; A learning method for learning a tracking model that performs tracking processing of the target based on the target detection information, using the learning data.

[0084] (Appendix 12) acquiring, from a radar device, target detection information indicating a position of a target detected based on a received wave and a trajectory of the target calculated based on the target detection information; acquiring target position information indicating a position of the target; generating learning data using the target detection information, the flight track, and the target position information; A recording medium having a program recorded thereon that causes a computer to execute a process of learning a tracking model that performs tracking processing of the target based on the target detection information, using the learning data.

[0085] (Appendix 13) a target detection unit that transmits a transmission wave and generates target detection information indicating a position of a target based on a received wave corresponding to the transmission wave; a tracking processing unit that performs tracking processing of the target based on the target detection information, using a tracking model that has been trained using learning data generated based on the target detection information, the target's trajectory calculated based on the target detection information, and target position information indicating a position of the target; A radar device comprising:

[0086] Although the present invention has been described above with reference to the embodiments and examples, the present invention is not limited to the above-mentioned embodiments and examples. Various modifications that can be understood by those skilled in the art can be made to the configuration and details of the present invention within the scope of the present invention. [Explanation of symbols]

[0087] 200 Learning Device 201 Learning Data Generation Unit 202 Data Collection Department 204 Learning processing unit 300 Radar Equipment 301 PSR Antenna Part 302 PSR Transmitter / Receiver 303 PSR signal processing unit 304 Beam Control Unit 305 PSR target detection unit 306, 114, 314 Tracking processing unit 307 Display operation section 308 SSR antenna part 309 SSR Transmitter / Receiver 310 SSR target detection unit

Claims

1. a first acquisition means for acquiring, from the radar device, target detection information indicating a position of an aircraft detected based on a received wave received in response to a transmission wave from the radar device, and a flight path of the aircraft calculated based on the target detection information; A second acquisition means for acquiring target position information indicating a position of the aircraft; a learning data generating means for generating learning data using the target detection information, the flight path, and the target position information; a simulation data generating means for generating simulation data including the target detection information and a label for the target detection information by simulation; a learning processing means for learning a tracking model for performing a tracking process of the aircraft based on the target detection information, using the learning data and the simulation data; a radar control means for controlling the radar device to selectively transmit, as the transmission wave, either a first transmission wave for tracking the aircraft or a second transmission wave for collecting data corresponding to a condition that is lacking in generating the learning data; and A learning device comprising:

2. 2. The learning device according to claim 1, wherein the learning data generating means generates a teacher label indicating whether the aircraft indicated by the target detection information is a true target or a false target, based on the target position information.

3. 3. The learning device according to claim 1, wherein the learning data generating means generates a teacher label including a position, a speed, and an acceleration of the aircraft indicated by the target detection information, based on the target position information and the flight path.

4. the target detection information is a primary radar plot generated by a primary radar of the radar device; 4. The learning device according to claim 1, wherein the target position information is a secondary radar plot generated by a secondary radar.

5. The learning device according to claim 1 , wherein the target position information is self-position information of the aircraft contained in a signal received from the aircraft.

6. The learning device according to any one of claims 1 to 5, wherein the radar control means, when controlling the radar device to transmit the second transmission wave as the transmission wave, further controls the radar device to generate the target detection information corresponding to the conditions.

7. The radar device is mounted on a moving object, a third acquisition means for acquiring mobile object information including position and movement information of the mobile object; The learning device according to claim 1 , wherein the learning data generating means generates the learning data by using the mobile object information.

8. the radar device includes two data processing means for generating the target detection information and the track by using the tracking model, and a switching means capable of switching one of the two data processing means to an active state while switching the other to a standby state, the first acquisition means acquires the target detection information and the track from one of the two data processing means that is switched to the active state, 2. The learning device according to claim 1, further comprising a learning result application means for applying the tracking model, which has been learned using the learning data corresponding to the target detection information and the wake obtained from one of the data processing means, to the other of the two data processing means that has been switched to the standby state.

9. 2. The learning device according to claim 1, wherein the radar device comprises: a first data processing means for acquiring a first processing result by generating the target detection information and the track without using the tracking model; a second data processing means for acquiring a second processing result by generating the target detection information and the track using the tracking model; and a validity evaluation means for determining validity of learning of the tracking model by comparing the first processing result with the second processing result.

10. The radar device has a first data processing means for acquiring a first processing result by generating the target detection information and the track using the tracking model learned earlier in time, a second data processing means for acquiring a second processing result by generating the target detection information and the track using the tracking model learned later in time, and an integration means for acquiring a third processing result by integrating the first processing result and the second processing result, The learning device according to claim 1 , wherein the first acquisition means acquires the target detection information and the flight track corresponding to the third processing result from the radar device.

11. acquiring, from the radar device, target detection information indicating a position of an aircraft detected based on a received wave received in response to transmission of a transmission wave from the radar device, and a flight path of the aircraft calculated based on the target detection information; acquiring target position information indicating a position of the aircraft; generating learning data using the target detection information, the flight track, and the target position information; generating simulation data including the target detection information and a label for the target detection information through a simulation; Using the learning data and the simulation data, a tracking model is trained to perform a tracking process for the aircraft based on the target detection information; A learning method in which the radar device is controlled to selectively transmit as the transmission wave either a first transmission wave for tracking the aircraft or a second transmission wave for collecting data corresponding to conditions that are lacking in generating the learning data.

12. acquiring, from the radar device, target detection information indicating a position of an aircraft detected based on a received wave received in response to transmission of a transmission wave from the radar device, and a flight path of the aircraft calculated based on the target detection information; acquiring target position information indicating a position of the aircraft; generating learning data using the target detection information, the flight track, and the target position information; generating simulation data including the target detection information and a label for the target detection information through a simulation; Using the learning data and the simulation data, a tracking model is trained to perform a tracking process for the aircraft based on the target detection information; A program that causes a computer to execute a process of controlling the radar device to selectively transmit, as the transmission wave, either a first transmission wave for tracking the aircraft or a second transmission wave for collecting data corresponding to conditions that are lacking in generating the learning data.

13. a target detection means for transmitting a transmission wave and generating target detection information indicating a position of an aircraft based on a received wave corresponding to the transmission wave; a tracking processing means for performing a tracking process of the aircraft based on the target detection information, using a tracking model that has been trained using simulated data obtained by a simulation for generating data including the target detection information, a flight path of the aircraft calculated based on the target detection information, and target position information indicating a position of the aircraft, and the target detection information and a label for the target detection information; a radar control means for performing control to selectively transmit, as the transmission wave, either a first transmission wave for tracking the aircraft or a second transmission wave for collecting data corresponding to a condition that is lacking in generating the learning data; and A radar device comprising:

Images