Radar device, signal processing device, and signal processing method

Abstract

To provide a radar device that reduces noise and improves the detection performance of a moving target.SOLUTION: According to an embodiment, a radar device includes an antenna unit, a transmission / reception unit, and a signal processing unit. The transmission / reception unit transmits a radar pulse from the antenna unit, receives radio waves arriving at the antenna unit, and generates a reception signal. The signal processing unit processes the reception signal. The signal processing unit includes a pulse compression processing unit, a first high-sensitivity processing unit, a second high-sensitivity processing unit, and a detection processing unit. The pulse compression processing unit pulse-compresses the reception signal and generates a compression pulse signal. The first high-sensitivity processing unit reduces noise in the compression pulse signal and generates a first high-sensitivity signal. The second high-sensitivity processing unit reduces noise of the first high-sensitivity signal and generates a second high-sensitivity signal. The detection processing unit detects a target from the second high-sensitivity signal.SELECTED DRAWING: Figure 2

Description

Technical Field

[0001] Embodiments of the present invention relate to a radar device, a signal processing device, and a signal processing method.

Background Art

[0002] For example, a radar device mounted on an aircraft for detecting an approaching target is known. In this type of radar device, for example, it has been proposed to suppress clutter using a pattern discriminator to which machine learning is applied.

Prior Art Documents

Patent Documents

[0003]

Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0004] When the target is present at a distance or has a small Radar Cross Section (RCS), the Signal to Noise ratio (S / N) of the received signal becomes small. Increasing the aperture area of the radar device's antenna or increasing the transmission power is effective for increasing the S / N, but the scale of the hardware increases. Also, increasing the number of pulses used for coherent integration or the like can increase the S / N, but the update rate may increase or the detection area may be limited. Moreover, in the case of a target moving at high speed, the detection performance significantly deteriorates. Therefore, it is necessary to improve the detection performance of a target moving at high speed without increasing the hardware scale of the radar antenna by a signal processing method using machine learning different from the conventional method. However, in order to handle targets with various speeds, it is necessary to train a pattern discriminator capable of handling various speeds, and the scale of machine learning becomes large.

[0005] Therefore, an object is to provide a radar device, a signal processing device, and a signal processing method that reduce noise and have high detection performance for a moving target.

Means for Solving the Problems

[0006] According to an embodiment, a radar device includes an antenna unit, a transmission / reception unit, and a signal processing unit. The transmission / reception unit transmits a radar pulse from the antenna unit, receives the radio wave that has arrived at the antenna unit, and generates a reception signal. The signal processing unit processes the reception signal. The signal processing unit includes a pulse compression processing unit, The integration processing unit, a first sensitivity improvement processing unit, a second sensitivity improvement processing unit, and a detection processing unit. The pulse compression processing unit pulse-compresses the reception signal to generate a compressed pulse signal. The integration processing unit pulse-integrates the compressed pulse signal to generate an integrated pulse signal. The first sensitivity improvement processing unit reduces noise in the compressed pulse signal to generate a first sensitivity-improved signal. The second sensitivity improvement processing unit reduces noise in the first sensitivity-improved signal to generate a second sensitivity-improved signal. The detection processing unit detects a target from the second sensitivity-improved signal. The second sensitivity improvement processing unit generates a second sensitivity improvement signal from the integrated pulse signal and the first sensitivity improvement signal.

Brief Description of the Drawings

[0007]

Figure 1

Figure 2

Figure 3

Figure 4

Figure 5

Figure 6

Figure 7

Figure 8

Figure 9

Figure 10

Embodiments for Carrying Out the Invention

[0008] Hereinafter, embodiments will be described with reference to the drawings. FIG. 1 is a functional block diagram showing an example of a radar device according to each embodiment. In the embodiments, a pulse radar is assumed. As shown in FIG. 1, the radar device 1 includes an antenna unit 10, a transmission / reception unit 11, and a signal processing unit 12. Further, the radar device 1 may include a display unit 13 for visually displaying the output of the signal processing unit 12 to present it to the user. Here, the user is, for example, a monitor who monitors targets by the radar device 1. The target is, for example, an object that moves relatively fast in the air, such as a flying object or an aircraft. A person may or may not be on the target. Also, the target is not necessarily limited to an object moving in the air.

[0009] The antenna unit 10 is, for example, an array antenna having a plurality of antenna elements arranged regularly. The antenna unit 10 transmits radio waves (radar waves) into space and receives radio waves including reflected waves (received signals) from the target.

[0010] The transmission / reception unit 11 generates a radar pulse and sends it to the antenna unit 10. Also, the transmission / reception unit 11 generates a received signal by performing processing such as amplification and analog / digital (A / D) conversion on the radio waves that have arrived at the antenna unit 10. The transmission / reception unit 11 sends the generated received signal to the signal processing unit 12.

[0011] The signal processing unit 12 is a computer including, for example, a memory and a processor. The signal processing unit 12 performs various signal processes on the received signal from the transmission / reception unit 11. This signal processing includes, for example, signal processes such as clutter suppression, generation of radar video, and target detection. The signal processing unit 12 sends the information of the detected target to the display unit 13.

[0012] The display unit 13 visually presents to the user based on the information of the target input from the signal processing unit 12.

[0013] [First Embodiment] The first embodiment will be described. FIG. 2 is a functional block diagram showing an example of the signal processing unit according to the first embodiment. The signal processing unit 12 includes a pulse compression processing unit 121, a sensitivity improvement processing unit 122, a sensitivity improvement processing unit 123, and a detection processing unit 124.

[0014] The pulse compression processing unit 121 pulse-compresses the received signal from the transmission / reception unit 11 to generate a compressed pulse signal. The pulse compression processing unit 121 outputs the compressed pulse signal to the sensitivity improvement processing unit 122. The compressed pulse signal is, for example, three-dimensional data including distance, azimuth, and elevation as axes. The compressed pulse signal may be four-dimensional or higher-dimensional data including other elements such as Doppler.

[0015] The sensitivity improvement processing unit 122 performs a first sensitivity improvement process on the compressed pulse signal from the pulse compression processing unit 121 to generate a first sensitivity improvement signal. The sensitivity improvement processing unit 122 outputs the first sensitivity improvement signal to the sensitivity improvement processing unit 123. The first sensitivity improvement signal is a signal in which signals other than the target signal component in the compressed pulse signal are reduced as noise. As will be described later, the sensitivity improvement processing unit 122 performs a process of reducing noise in the compressed pulse signal using a learned model. Here, the compressed pulse signal of each pulse may be input to the sensitivity improvement processing unit 122, or a plurality of compressed pulse signals such as in units of CPI (Coherent Processing Interval) may be input together.

[0016] The high-sensitivity processing unit 123 performs a second high-sensitivity process on the first high-sensitivity signal from the high-sensitivity processing unit 122 to generate a second high-sensitivity signal. The high-sensitivity processing unit 123 outputs the second high-sensitivity signal to the detection processing unit 124. The second high-sensitivity signal is a signal in which signals other than the target signal component in the first high-sensitivity signal are reduced as noise. As will be described later, the high-sensitivity processing unit 123 performs a process of reducing noise in the first high-sensitivity signal using a learned model in the same manner as the high-sensitivity processing unit 122.

[0017] The detection processing unit 124 detects a target from the second high-sensitivity signal from the high-sensitivity processing unit 123. In the second high-sensitivity signal output from the high-sensitivity processing unit 123, the level of signal components other than the target is reduced. Therefore, the detection processing unit 124 can extract only the target.

[0018] FIG. 3 is a block diagram showing an example of the configuration of the high-sensitivity processing unit 122 and the high-sensitivity processing unit 123. The high-sensitivity processing unit 122 and the high-sensitivity processing unit 123 are computers having a physical entity as hardware, for example. The high-sensitivity processing unit 122 and the high-sensitivity processing unit 123 may be configured by a single computer or may be configured by separate computers. Hereinafter, it will be described on the assumption that the high-sensitivity processing unit 122 and the high-sensitivity processing unit 123 are configured by a single computer. The computer constituting the high-sensitivity processing unit 122 and the high-sensitivity processing unit 123 includes a processor 21 and a storage unit 24. Further, the computer includes a ROM 22, a RAM 23, and a communication unit 25.

[0019] The processor 21 is an arithmetic device such as a CPU (Central Processing Unit) or an MPU (Micro Processing Unit), for example. The processor 21 reads a program stored in the ROM 22 or the storage unit 24 into the RAM 23 and operates as the high-sensitivity processing unit 122 and the high-sensitivity processing unit 123 by executing the program read into the RAM 23.

[0020] The ROM 22 is a non-volatile memory such as a flash memory and operates as a program memory. The ROM 22 stores programs executed by the processor 21 and control data and the like used by the processor 21.

[0021] The RAM 23 is a volatile memory such as a DRAM and operates as a working memory that temporarily holds data. Programs are loaded into the RAM 23. Also, the RAM 23 temporarily holds data processed by the processor 21.

[0022] The storage unit 24 is a storage configured by a non-volatile memory such as an SSD (Solid State Drive). The storage unit 24 stores a program 24a executed by the processor 21, data 24b used by the processor 21, a learned model 24c, and a learned model 24d. The learned model 24c is a learned model for the high-sensitivity processing unit 122. The learned model 24c is a learned model in which the weights of each layer are set to input a compressed pulse signal and output a first high-sensitivity signal with noise components other than the target component included in the compressed pulse signal reduced. That is, the learning of the learned model 24c is performed to identify the target signal component and the noise signal component included in the input compressed pulse signal, and output a signal with the intensity of the signal component identified as noise reduced. Learning with restrictions may be performed so that a margin is included in the signal range identified as the target in the compressed pulse signal. The learned model 24d is a learned model in which the weights of each layer are set to input the first high-sensitivity signal and output a second high-sensitivity signal with noise components other than the target component included in the first high-sensitivity signal reduced. That is, the learning of the learned model 24d is performed to identify the target signal component and the noise signal component included in the input first high-sensitivity signal, and output a signal with the intensity of the signal component identified as noise reduced. Learning with restrictions may be performed so that a margin is included in the signal range identified as the target in the first high-sensitivity signal. The margin of the signal range identified as the target in the first high-sensitivity signal may be narrower than the margin of the signal range identified as the target in the compressed pulse signal.

[0023] The communication unit 25 is connected to the pulse compression processing unit 121 and acquires a compressed pulse signal from the pulse compression processing unit 121. Further, the communication unit 25 is connected to the detection processing unit 124 and outputs the second high-sensitivity signal to the detection processing unit 124.

[0024] Further, the processor 21 includes a learning unit 21a and a learning unit 21b. The processor 21 operates as the learning unit 21a and the learning unit 21b by executing processing according to the instructions described in the program 24a.

[0025] The learning unit 21a repeatedly performs machine learning using the compressed pulse signal acquired from the pulse compression processing unit 121 via the communication unit 25 as learning data, and generates a learned model. The generated learned model is stored in the storage unit 24 as the learned model 24c.

[0026] The learning unit 21b repeatedly performs machine learning using the first high-sensitivity signal, which is the output of the high-sensitivity processing unit 122 generated by the learning unit 21a, as learning data, and generates a learned model. As the learning data, a signal simulating the first high-sensitivity signal may be used. The generated learned model is stored in the storage unit 24 as the learned model 24d.

[0027] For example, the learned model 24c and the learned model 24d can be generated by repeatedly providing learning data to a Convolutional Neural Network (CNN). This type of machine learning is called deep learning. In the first embodiment, the learning unit 21a repeatedly performs supervised machine learning using the compressed pulse signal as learning data, and generates the learned model 24c. The teacher data in this case is, for example, data of a compressed pulse signal including only a target. Also, in the first embodiment, the learning unit 21b repeatedly performs supervised machine learning using the first high-sensitivity signal as learning data, and generates the learned model 24d. The teacher data in this case is, for example, data of the first high-sensitivity signal including only a target.

[0028] Here, the learning data used for the learning to generate the learned model 24c may be created in consideration of the movement of the target between the compressed pulse signals. Similarly, the learning data used for the learning to generate the learned model 24d may be created in consideration of the movement of the target between the first high-sensitivity signals.

[0029] Also, the training data used for training to generate the learned models 24c and 24d may be data for each assumed target speed for each Doppler filter bank.

[0030] The sensitivity improvement processing units 122 and 123 can also be implemented as software. In this case, the sensitivity improvement processing units 122 and 123 can be configured as programs that are expanded in the RAM and executed by the processor.

[0031] Also, the pulse compression processing unit 121 and the detection processing unit 124 may each be configured as dedicated hardware configured to execute the above-described operations, or may be configured as software that causes the processor to execute the above-described operations.

[0032] Next, the operation of the signal processing unit 12 of the first embodiment will be described. FIG. 4 is a flowchart showing an example of target detection processing by the signal processing unit 12 in the first embodiment. In step S1, the pulse compression processing unit 121 of the signal processing unit 12 pulse-compresses the received signal from the transmission / reception unit 11 to generate a compressed pulse signal. When a high S / N signal in the compressed pulse signal is detected by prior threshold processing or the like, this high S / N target signal may be replaced with noise.

[0033] In step S2, the sensitivity improvement processing unit 122 generates a first sensitivity improvement signal from the compressed pulse signal. Specifically, the sensitivity improvement processing unit 123 inputs the compressed pulse signal to the learned model 24c to obtain the first sensitivity improvement signal.

[0034] In step S3, the sensitivity improvement processing unit 123 generates a second sensitivity improvement signal from the first sensitivity improvement signal. Specifically, the sensitivity improvement processing unit 123 inputs the first sensitivity improvement signal to the learned model 24d to obtain the second sensitivity improvement signal.

[0035] In step S4, the detection processing unit 124 detects a target from the second sensitivity-enhanced signal. For example, the detection processing unit 124 detects, as a target, a signal range having an intensity equal to or greater than a threshold value in the second sensitivity-enhanced signal.

[0036] FIG. 5 is a diagram for explaining the operations of the sensitivity enhancement processing units 122 and 123. In FIG. 5, a compressed pulse signal is input to the sensitivity enhancement processing unit 122 in time series. The sensitivity enhancement processing unit 122 groups the input compressed pulse signals into a plurality of units. The plurality of units of compressed pulse signals are, for example, M compressed pulse signals X(T), X(T−1),..., X(T−M+1) including the compressed pulse signal X(T) at time T shown in FIG. 5. After thus grouping the compressed pulse signals into a plurality of units, the sensitivity enhancement processing unit 122 performs sensitivity enhancement processing on each unit of the compressed pulse signals. (Equation 1) shows the first sensitivity-enhanced signal Y1(T) that is the output of the sensitivity enhancement processing unit 122 for the M compressed pulse signals from time T. D1(X(T), X(T−1),..., X(T−M+1)) in (Equation 1) indicates that noise reduction processing as sensitivity enhancement processing using the learned model 24c is performed on the input compressed pulse signals.

Equation

[0037] As described above, the sensitivity enhancement processing unit 122 may perform noise reduction processing as sensitivity enhancement processing on each of the input compressed pulse signals without grouping the input compressed pulse signals into a plurality of units.

[0038] After the sensitivity enhancement processing in the sensitivity enhancement processing unit 122, the sensitivity enhancement processing unit 123 performs sensitivity enhancement processing on, for example, the first sensitivity-enhanced signals Y1(T), Y1(T−T s1 ),..., Y1(T−T sN-1 ) over the past N times. (Equation 2) shows the second sensitivity-enhanced signal Y2(T) that is the output of the sensitivity enhancement processing unit 123 for the first sensitivity-enhanced signals over the past N times. D2(Y1(T), Y1(T−T s1), …, Y1(T - T sN-1 (),…) performs noise reduction processing as a sensitivity improvement process using the learned model 24d on the input first sensitivity improvement signal. [Number]

[0039] As described above, according to the first embodiment, noise reduction processing as a sensitivity improvement process is performed on the compressed pulse signal, and noise reduction processing as a sensitivity improvement process is further performed on the first sensitivity improvement signal sensitized by the noise reduction processing. Thus, in the first embodiment, by performing two-stage sensitivity improvement processing on the compressed pulse signal, a signal with higher sensitivity than the existing technology can be generated. By performing target detection processing using the sensitized signal, the target can be detected with a smaller number of pulse hits than in the existing technology. Since the target can be detected with a smaller number of pulse hits, it is also suitable for detecting a fast-moving target. Furthermore, target detection processing is performed by improving the sensitivity of a low S / N signal. Therefore, it is not necessary to increase the S / N by increasing the aperture area or transmission power of the antenna. For this reason, the ease of mounting is not impaired. That is, in the first embodiment, when detecting a target with the same RCS, the detection accuracy equivalent to the existing technology can be maintained even if the scale of the hardware is smaller than that of the existing technology.

[0040] Here, in the first embodiment, padding processing may be performed on the boundary portions of each compression pulse signal. The padding processing is a process of expanding the boundary portion of the compression pulse signal with a predetermined value, for example, data of 0. However, when the padding processing is performed, there is a possibility that the performance of the high-sensitivity processing for the compression pulse signal may differ between the central portion and the boundary portion of the compression pulse signal. Therefore, instead of the boundary portion being expanded with data of 0, the boundary portion may be expanded with noise data. By expanding the boundary portion with noise data, the performance of the high-sensitivity processing for the compression pulse signal is made uniform between the central portion and the boundary portion of the compression pulse signal. Here, the noise data may be generated, for example, by pseudo-random numbers adjusted to the noise level of the received signal. Also, the noise data may be data of a fixed pattern stored in advance in the ROM 22 or the like. Furthermore, the noise data may be noise data obtained by performing noise measurement.

[0041] [Second Embodiment] The second embodiment will be described. FIG. 6 is a functional block diagram showing an example of the signal processing unit according to the second embodiment. The signal processing unit 12 in the second embodiment includes a pulse compression processing unit 121, a high-sensitivity processing unit 122, a high-sensitivity processing unit 123, a detection processing unit 124, and an integration processing unit 125. Here, for elements similar to those in FIG. 2 in FIG. 6, the same reference numerals as those in FIG. 2 are given. For elements given the same reference numerals as those in FIG. 2, the description will be simplified or omitted.

[0042] In the second embodiment, the pulse compression processing unit 121 outputs the compression pulse signal to the high-sensitivity processing unit 122 and the integration processing unit 125.

[0043] The high-sensitivity processing unit 122 is the same as in the first embodiment. That is, the high-sensitivity processing unit 122 performs the first high-sensitivity processing on the compression pulse signal from the pulse compression processing unit 121 to generate a first high-sensitivity signal. The high-sensitivity processing unit 122 outputs the first high-sensitivity signal to the high-sensitivity processing unit 123.

[0044] The integration processing unit 125 generates an integrated pulse signal by pulse-integrating the compressed pulse signal from the pulse compression processing unit 121. The integration processing unit 125 outputs the integrated pulse signal to the high-sensitivity processing unit 123. For the integration processing in the integration processing unit 125, for example, coherent integration can be used.

[0045] The high-sensitivity processing unit 123 performs a second high-sensitivity process on the first high-sensitivity signal from the high-sensitivity processing unit 122 and the integrated pulse signal from the integration processing unit 125 to generate a second high-sensitivity signal. The high-sensitivity processing unit 123 outputs the second high-sensitivity signal to the detection processing unit 124. The high-sensitivity processing unit 123 performs a process of reducing noise in the first high-sensitivity signal using a learned model in the same manner as the high-sensitivity processing unit 122. Note that the high-sensitivity processing unit 123 in the second embodiment may also be realized by the computer shown in FIG. 3.

[0046] In the second embodiment, the learned model 24d used in the high-sensitivity processing unit 123 is generated by repeating supervised machine learning using the first high-sensitivity signal and the integrated pulse signal as learning data. The teacher data in this case is, for example, data of the first high-sensitivity signal and the integrated pulse signal including only the target.

[0047] The detection processing unit 124 is the same as in the first embodiment. That is, the detection processing unit 124 detects the target from the second high-sensitivity signal from the high-sensitivity processing unit 123.

[0048] As described above, according to the second embodiment, noise reduction processing and integration processing as high-sensitivity processing are performed on the compressed pulse signal, and noise reduction processing as high-sensitivity processing is further performed with the first high-sensitivity signal and the integrated pulse signal as inputs. Similar to the first high-sensitivity signal, it can be said that the integrated pulse signal is also a compressed pulse signal with increased sensitivity. Such two types of high-sensitivity signals are used as learning data, and learning in the high-sensitivity processing unit 123 is performed. That is, the learning data can be increased compared to the first embodiment. Further, by performing high-sensitivity processing in the learned model 24d according to such two types of high-sensitivity signals, further high-sensitivity is expected compared to the first embodiment.

[0049] Here, also in the second embodiment, padding processing may be performed on the boundary portions of the respective compressed pulse signals, or the boundary portions may be extended by noise data.

[0050] [Third Embodiment] The third embodiment will be described. FIG. 7 is a functional block diagram showing an example of a signal processing unit according to the third embodiment. The signal processing unit 12 in the third embodiment includes a pulse compression processing unit 121, a high-sensitivity processing unit 122, a detection processing unit 124, and a probability output unit 126. Here, for elements similar to those in FIG. 2 in FIG. 7, the same reference numerals as in FIG. 2 are given. For elements given the same reference numerals as in FIG. 2, the description will be simplified or omitted.

[0051] In the third embodiment, the high-sensitivity processing unit 123 is replaced by the probability output unit 126. Further, in the third embodiment, the compressed pulse signal generated by the pulse compression processing unit 121 is output not only to the high-sensitivity processing unit 122 but also to the detection processing unit 124. The high-sensitivity processing unit 122 and the detection processing unit 124 may be input with the compressed pulse signal one pulse at a time, or may be input with a plurality of compressed pulse signals, such as in CPI units, in a batch.

[0052] The probability output unit 126 performs probability output processing on the first high-sensitivity signal from the high-sensitivity processing unit 122 to generate a probability signal. The probability signal is a signal that represents the distribution of the probability that the signal component corresponding to the distribution of the target-likeness in the first high-sensitivity signal is the target signal component. The probability output unit 126 performs processing to generate a probability signal using a learned model. Note that the probability output unit 126 may also be implemented by the computer shown in FIG. 3.

[0053] In the third embodiment, the learning unit 21b repeatedly performs machine learning using, as learning data, the first high-sensitivity signal that is the output of the high-sensitivity processing unit 122 generated by the learning unit 21a, and generates a learned model 24d. The teacher data in this case is, for example, data of a target map indicating the target signal range in the input first high-sensitivity signal and data of a noise map indicating the noise signal range in the input first high-sensitivity signal.

[0054] The detection processing unit 124 detects a target by extracting the target signal component from the compressed pulse signal from the pulse compression processing unit 121 based on the probability signal from the probability output unit 126. The detection processing unit 124 generates a binarized signal by binarizing the probability signal, for example, using a threshold value. The binarized signal is a signal that is 1 when the probability of being a target is equal to or greater than the threshold value and 0 when it is less than the threshold value. Then, the detection processing unit 124 matches the binarized signal and the compressed pulse signal by calculating the product of the binarized signal and the compressed pulse signal. The detection processing unit 124 detects, as a target, a signal range where the product of the binarized signal and the compressed pulse signal is not zero.

[0055] Here, the threshold value when the probability signal is binarized may be set appropriately. For example, if the threshold value is a small value, the signal range detected as a target is likely to be wide. Therefore, it becomes easier to extract the target from the compressed pulse signal. On the other hand, if it is necessary to improve the detection accuracy of the target, the threshold value may be set to a large value.

[0056] Next, the operation of the signal processing unit 12 of the third embodiment will be described. FIG. 8 is a flowchart showing an example of target detection processing by the signal processing unit 12 in the third embodiment. In step S11, the pulse compression processing unit 121 of the signal processing unit 12 pulse-compresses the received signal from the transmission / reception unit 11 to generate a compressed pulse signal.

[0057] In step S12, the sensitivity enhancement processing unit 122 generates a first sensitivity-enhanced signal from the compressed pulse signal. Specifically, the sensitivity enhancement processing unit 123 inputs the compressed pulse signal to the learned model 24c to obtain the first sensitivity-enhanced signal.

[0058] In step S13, the probability output unit 126 generates a probability signal from the first sensitivity-enhanced signal. Specifically, the sensitivity enhancement processing unit 123 inputs the first sensitivity-enhanced signal to the learned model 24d to obtain the probability signal.

[0059] In step S14, the detection processing unit 124 detects a target from the compressed pulse signal and the probability signal. The detection processing unit 124 calculates, for example, the product of the compressed pulse signal and the binary signal of the probability signal. Then, the detection processing unit 124 detects a signal range where the product is not zero as the target.

[0060] As described above, according to the third embodiment, noise reduction processing as high-sensitivity processing is performed on the compressed pulse signal, and output processing of a probability signal representing the distribution of the target-likeness in the first high-sensitivity signal that has been made highly sensitive by the noise reduction processing is performed. Then, the target is detected using the compressed pulse signal and the probability signal. In this way, in the third embodiment, probability output processing is performed instead of the second-stage high-sensitivity processing on the compressed pulse signal. Even in this case, as in the first embodiment, the target can be detected with a smaller number of pulse hits compared to existing technologies. Since the target can be detected with a smaller number of pulse hits, it is also suitable for detecting a target moving at high speed. Furthermore, the target detection process is performed by increasing the sensitivity of a low S / N signal. Therefore, it is not necessary to increase the S / N by increasing the antenna aperture area or transmission power. For this reason, the ease of mounting is not impaired. That is, in the first embodiment, when detecting a target with the same RCS, even if the scale of the hardware is smaller than that of existing technologies, detection accuracy equivalent to that of existing technologies can be maintained.

[0061] Here, also in the third embodiment, padding processing may be performed on the boundary portions of each compressed pulse signal, or the boundary portions may be extended by noise data.

[0062] [Fourth Embodiment] The fourth embodiment will be described. FIG. 9 is a functional block diagram showing an example of a signal processing unit according to the fourth embodiment. The signal processing unit 12 in the fourth embodiment includes a pulse compression processing unit 121, a high-sensitivity processing unit 122, a detection processing unit 124, and a probability output unit 126a. Here, for elements similar to those in FIG. 2 in FIG. 9, the same reference numerals as in FIG. 2 are given. For elements given the same reference numerals as in FIG. 2, the description will be simplified or omitted.

[0063] In the fourth embodiment, the high-sensitivity processing unit 122 and the probability output unit 126a are provided between the pulse compression processing unit 121 and the detection processing unit 124.

[0064] In the fourth embodiment, the high-sensitivity processing unit 122 generates a first high-sensitivity signal from the compressed pulse signal. The high-sensitivity processing unit 122 outputs the first high-sensitivity signal to the detection processing unit 124.

[0065] The probability output unit 126a performs probability output processing on the compressed pulse signal from the pulse compression processing unit 121 to generate a probability signal. The probability output unit 126a performs processing to generate a probability signal using a learned model. Note that the probability output unit 126a may be realized by the computer shown in FIG. 3.

[0066] The learning unit 21b in the fourth embodiment repeatedly performs machine learning using the compressed pulse signal as learning data to generate a learned model 24d. The teacher data in this case is, for example, data of a target map indicating a target signal range in the input compressed pulse signal and data of a noise map indicating a noise signal range in the input compressed pulse signal.

[0067] The detection processing unit 124 detects a target by extracting a target signal component from the first high-sensitivity signal from the high-sensitivity processing unit 122 based on the probability signal from the probability output unit 126a. The detection processing unit 124 generates a binarized signal by binarizing the probability signal with, for example, a threshold value. Then, the detection processing unit 124 matches the binarized signal and the first high-sensitivity signal by calculating the product of the binarized signal and the first high-sensitivity signal. The detection processing unit 124 detects as a target a signal range where the product of the binarized signal and the first high-sensitivity signal is not zero.

[0068] Here, the threshold value when the probability signal is binarized may be set appropriately. For example, if the threshold value is a small value, the signal range detected as a target is likely to be wide. Therefore, it becomes easier to extract a target from the compressed pulse signal. On the other hand, if it is necessary to further improve the detection accuracy, the threshold value may be set to a large value.

[0069] As described above, according to the fourth embodiment, even in the case where the target is detected by the matching between the first sensitivity-enhanced signal and the probability signal, the target can be detected with a smaller number of pulse hits compared to the existing technology as in the first embodiment. Since the target can be detected with a smaller number of pulse hits, it is also suitable for detecting a target moving at high speed. Furthermore, the target detection process is performed by enhancing the sensitivity of a low S / N signal. Therefore, it is not necessary to increase the S / N by increasing the aperture area of the antenna or the transmission power. For this reason, the ease of mounting is not impaired. That is, in the first embodiment, when detecting a target with the same RCS, even if the scale of the hardware is smaller than that of the existing technology, the detection accuracy equivalent to that of the existing technology can be maintained.

[0070] Here, in the fourth embodiment, the probability output unit 126a generates a probability signal from the compressed pulse signal. In contrast, the probability output unit 126a may be configured to generate a probability signal from the first sensitivity-enhanced signal in the same manner as in the third embodiment. Further, the probability output unit 126a may be configured to generate a probability signal from the integrated pulse signal generated by the integration processing unit 125 described in the second embodiment.

[0071] Also, the detection processing unit 124 may detect the target by taking the matching between the second sensitivity-enhanced signal and the probability signal instead of the first sensitivity-enhanced signal. In this case, a sensitivity-enhanced processing unit 123 is provided between the sensitivity-enhanced processing unit 122 and the detection processing unit 124.

[0072] Furthermore, also in the fourth embodiment, padding processing may be performed on the boundary portions of the respective compressed pulse signals, or the boundary portions may be extended by noise data.

[0073] [Fifth Embodiment] The fifth embodiment will be described. FIG. 10 is a functional block diagram showing an example of a signal processing unit according to the fifth embodiment. The signal processing unit 12 in the fifth embodiment includes a pulse compression processing unit 121, a detection processing unit 124, and a probability output unit 126a. Here, for elements similar to those in FIG. 2 in FIG. 10, the same reference numerals as in FIG. 2 are given. For elements given the same reference numerals as in FIG. 2, the description will be simplified or omitted.

[0074] In the fifth embodiment, the probability output unit 126a is provided between the pulse compression processing unit 121 and the detection processing unit 124.

[0075] Similar to the probability output unit 126a in the fourth embodiment, the probability output unit 126a performs probability output processing on the compressed pulse signal from the pulse compression processing unit 121 to generate a probability signal.

[0076] Based on the probability signal from the probability output unit 126a, the detection processing unit 124 detects a target by extracting a target signal component from the compressed pulse signal from the pulse compression processing unit 121. The detection processing unit 124 may also detect a target by performing threshold processing on the probability signal from the probability output unit 126a.

[0077] Although several embodiments of the present invention have been described, these embodiments are presented as examples and are not intended to limit the scope of the invention. These novel embodiments can be implemented in various other forms, and various omissions, replacements, and changes can be made without departing from the gist of the invention. These embodiments and their modifications are included in the scope and gist of the invention, and are also included in the invention described in the claims and its equivalent scope.

Description of Reference Numerals

[0078] 1…Radar device, 10…Antenna unit, 11…Transceiver unit, 12…Signal processing unit, 13…Display unit, 21…Processor, 21a…Learning unit, 21b…Learning unit, 22…ROM, 23…RAM, 24…Storage unit, 24a…Program, 24b…Data, 24c…Trained model, 24d…Trained model, 25…Communication unit, 121…Pulse compression processing unit, 122…Sensitivity improvement processing unit, 123…Sensitivity improvement processing unit, 124…Detection processing unit, 125…Integration processing unit, 126…Probability output unit, 126a…Probability output unit.

Claims

1. An antenna unit, A transceiver unit that transmits a radar pulse from the antenna unit and receives an electromagnetic wave arriving at the antenna unit to generate a received signal, A signal processing unit that processes the received signal, comprising: The signal processing unit includes a pulse compression processing unit that pulse-compresses the received signal to generate a compressed pulse signal, an integration processing unit that pulse-integrates the compressed pulse signal to generate an integrated pulse signal, a first sensitivity improvement processing unit that reduces noise in the compressed pulse signal to generate a first sensitivity-improved signal, a second sensitivity improvement processing unit that reduces noise in the first sensitivity-improved signal to generate a second sensitivity-improved signal, and a detection processing unit that detects a target from the second sensitivity-improved signal, comprising: The second sensitivity improvement processing unit generates the second sensitivity-improved signal from the integrated pulse signal and the first sensitivity-improved signal, A radar device.

2. The first sensitivity improvement processing unit includes a first learning unit that repeatedly performs machine learning using the compressed pulse signal to generate a first learned model, The second sensitivity improvement processing unit includes a second learning unit that repeatedly performs machine learning using the first sensitivity-improved signal to generate a second learned model, The radar device according to claim 1.

3. In the radar device according to claim 1 or 2, a boundary portion of the compressed pulse signal is extended by a noise signal.

4. An antenna unit, A transceiver unit that transmits a radar pulse from the antenna unit and receives an electromagnetic wave arriving at the antenna unit to generate a received signal, and a signal processing unit that processes the received signal comprising: The signal processing unit includes a pulse compression processing unit that pulse-compresses the received signal to generate a compressed pulse signal, a sensitivity improvement processing unit that reduces noise in the compressed pulse signal to generate a sensitivity-improved signal, a probability output unit that calculates the probability of a target-likeness in the sensitivity-improved signal to generate a probability signal, and a detection processing unit that detects a target from the compressed pulse signal and the probability signal in a radar device.

5. The sensitivity improvement processing unit includes a first learning unit that repeatedly performs machine learning using the compressed pulse signal to generate a first learned model, and the probability output unit includes a second learning unit that repeatedly performs machine learning using the sensitivity-improved signal to generate a second learned model. The radar device according to claim 4.

6. An antenna unit, a transmission / reception unit that transmits a radar pulse from the antenna unit and receives a radio wave that has arrived at the antenna unit to generate a received signal, and a signal processing unit that processes the received signal in a radar device. The signal processing unit includes a pulse compression processing unit that pulse-compresses the received signal to generate a compressed pulse signal, a sensitivity improvement processing unit that reduces noise in the compressed pulse signal to generate a sensitivity-improved signal, a probability output unit that calculates the probability of a target-likeness in the compressed pulse signal to generate a probability signal, and a detection processing unit that detects a target from the sensitivity-improved signal and the probability signal in a radar device.

7. The sensitivity improvement processing unit includes a first learning unit that repeatedly performs machine learning using the compressed pulse signal to generate a first learned model, The probability output unit includes a second learning unit that repeatedly performs machine learning using the compressed pulse signal to generate a second learned model. The radar device according to claim 6.

8. The machine learning is deep learning that repeatedly provides learning data to a convolutional neural network to generate the first learned model and the second learned model. The radar device according to any one of claims 2, 5, and 7.

9. In a signal processing device that processes a received signal of a radar device, A pulse compression processing unit that pulse-compresses the received signal to generate a compressed pulse signal; A first sensitivity improvement processing unit that reduces noise in the compressed pulse signal to generate a first sensitivity-improved signal; An integration processing unit that pulse-integrates the compressed pulse signal to generate an integrated pulse signal; A second sensitivity improvement processing unit that reduces noise in the first sensitivity-improved signal to generate a second sensitivity-improved signal; A detection processing unit that detects a target from the second sensitivity-improved signal and is provided with The second sensitivity improvement processing unit generates the second sensitivity-improved signal from the integrated pulse signal and the first sensitivity-improved signal. Signal processing device.

10. In a signal processing device that processes a received signal of a radar device, A pulse compression processing unit that pulse-compresses the received signal to generate a compressed pulse signal; A sensitivity improvement processing unit that reduces noise in the compressed pulse signal to generate a sensitivity-improved signal; A probability output unit that calculates the probability of target-likeness in the sensitivity-improved signal to generate a probability signal; A detection processing unit that detects a target from the compressed pulse signal and the probability signal and is provided with a signal processing device.

11. In a signal processing method for processing a received signal of a radar device by a computer, the computer pulse-compresses the received signal to generate a compressed pulse signal, the computer pulse-integrates the compressed pulse signal to generate an integrated pulse signal, the computer reduces noise in the compressed pulse signal to generate a first high-sensitivity signal, the computer reduces noise of the first high-sensitivity signal from the integrated pulse signal and the first high-sensitivity signal to generate a second high-sensitivity signal, and the computer detects a target from the second high-sensitivity signal A signal processing method comprising the steps of:

12. In a signal processing method for processing a received signal of a radar device by a computer, the computer pulse-compresses the received signal to generate a compressed pulse signal, the computer reduces noise in the compressed pulse signal to generate a high-sensitivity signal, the computer calculates a probability of a target-likeness in the compressed pulse signal to generate a probability signal, and the computer detects a target from the high-sensitivity signal and the probability signal A signal processing method comprising the steps of:

Images