Radar device and radar signal processing method

Abstract

To reduce integration loss due to range walk and Doppler walk with a relatively small processing scale even in the case of long-term integration processing.SOLUTION: According to a radar device of the embodiment, a SIMO transmit beam and a receive beam are formed for the observation range, received signals of the receive beam are input in units of CPI, and data for (ND+MD-1)×CPI (MD>1) (MD is the number of CPIs to be integrated, and ND is the number of processing times for the MD×CPI integration unit) including past CPIs is used to sequentially extract and integrate ND (ND>1) sets of CPI data, extract the range-Doppler data with the maximum SN among the ND sets, perform target detection processing to extract a range cell, perform PGA processing using the slow-time axis FFT processed data of the extracted range cell among the entire range-Doppler data of (ND+MD-1)×CPI, perform angle measurement processing using the signal of the range-Doppler cell with the maximum SN, and output the observation value.SELECTED DRAWING: Figure 1

Description

[Technical Field]

[0001] The present embodiment relates to a radar device and a radar signal processing method. [Background technology]

[0002] In radar observation, when a target has a low RCS (Radar Cross-Section) or is at a long distance, it is necessary to improve the system gain. However, when there are restrictions on transmission power or antenna gain, long-term integration becomes necessary. In a radar device that performs such long-term observation, when there are a large number of hits, integration loss occurs due to range walk and Doppler walk. To address this issue, for example, Patent Document 1 proposes a method of maximizing an integral sequence by searching for velocity and acceleration.

[0003] However, with this method of maximizing integral sequences, when the range walk or Doppler walk is large, the search range increases, further increasing the processing scale. Furthermore, when a target exhibits complex movements, such as a drone, the Doppler walk increases during long-term observation, resulting in large phase changes and increased integration loss. Furthermore, when performing identification processing using MDS (Minimum Detectable Signal: Minimum Receiver Sensitivity), performing processing separately from the detection processing to increase the signal-to-noise ratio increases processing time. Furthermore, when performing long-term integration processing, the target moves during that time, resulting in a shift in target position in the output for each observation frame. Furthermore, when observing highly maneuverable targets, the system gain of long-term integration processing may be insufficient. [Prior art documents] [Patent documents]

[0004] [Patent Document 1] Patent No. 5025403 [Patent Document 2] Patent No. 5072694 [Non-patent literature]

[0005] [Non-Patent Document 1] Pulse compression: Ouchi, 'Fundamentals of Synthetic Aperture Radar for Remote Sensing', Tokyo Denki University Press, pp.131-149 (2003) [Non-patent document 2] CFAR (Constant False Alarm Rate): Yoshida, 'Revised Radar Technology', Institute of Electronics, Information and Communication Engineers, pp.87-89 (1996) [Non-patent document 3] Monopulse Angle Measurement: Yoshida, 'Revised Radar Technology', Institute of Electronics, Information and Communication Engineers, pp.260-264 (1996) [Non-patent document 4] PGA method: Charles V. Jakowatz, 'Spotlight-Mode Synthetic Aperture Radar: A Signal Processing Approach', Springer, pp.251-256(1996) [Non-Patent Document 5] Identification by MDS: Peter Klaer, 'An Investigation of Rotary Drone HERM Line Spectrum under Maneuvering Condit', Sensors 2020,20,5940 [Non-patent document 6] CNN: Saito, 'Deep Learning from Scratch', O'Reilly Japan, pp.205-221(2016) [Non-Patent Document 7] NN correlation processing: Samuel S. Blackman, Design and Analysis of Modern Tracking Systems, ARTECH HOUSE, INC., pp.8-11(1999) [Non-patent document 8] Kalman filter: Samuel S. Blackman, Multiple-Target Tracking with Radar Applications, ARTECH HOUSE, INC., pp.25-28(1986) Summary of the Invention [Problem to be solved by the invention]

[0006] The objective of this embodiment is to provide a radar device and a radar signal processing method that can reduce integration losses due to range walk and Doppler walk, enable identification by MDS, correct target position deviations during long-term integration processing, and suppress a decrease in system gain even when observing highly maneuvering targets, all with a relatively small processing scale, even during long-term integration processing. [Means for solving the problem]

[0007] In order to solve the above problems, the radar device according to the first embodiment forms SIMO (Single Input Multiple Output) transmit and receive beams for the observation range, inputs CPI (Coherent Pulse Interval, sometimes abbreviated as cpi in the following description) as a unit, and uses data for (ND+MD-1)×CPI (MD>1) (MD is the number of CPIs to be integrated, and ND is the number of processing times for the MD×CPI integration unit), including past CPIs, sequentially extracts and integrates ND (ND>1) sets of CPI data, extracts the range-Doppler (RD) data with the maximum signal-to-noise ratio among the ND sets, and performs detection processing using CFAR or the like to extract range cells. Then, among the entire RD data of (ND+MD-1)×CPI, performs PGA (Phase Gradient Autofocus) processing using the slow-time data of the extracted range cells, performs angle measurement processing using the signal of the range-Doppler cell with the maximum signal-to-noise ratio, and outputs the observation value (divided integration + PGA angle measurement).

[0008] That is, according to the radar device of the first embodiment, even if there is a range walk or a Doppler walk within the integration time, it is possible to efficiently integrate, detect a target, and perform angle measurement processing.

[0009] The radar device of the second embodiment generates an MDS (Micro Doppler Spectrum) by arranging the FFT results of ND data for the range cells extracted in the first embodiment, performs MDS identification processing using a CNN (Convolutional Neural Network) or the like, and outputs the identification result (MDS by division and integration).

[0010] That is, according to the radar device of the second embodiment, even in the case of long-term integration, it is possible to generate an MDS using the detection processing result and perform identification processing efficiently.

[0011] The radar device according to the third embodiment performs correlation tracking processing using the observed values ​​of the range and angle measurement results in the first or second embodiment to calculate the target speed and target acceleration, and outputs predicted values ​​at a time that is half the timestamp of the (ND+MD-1)×CPI data, thereby correcting the position deviation of the target movement in the correlation tracking results (position bias correction).

[0012] That is, the radar device according to the third embodiment makes it possible to correct position deviations due to target motion over a long integration time.

[0013] The radar device according to the fourth embodiment inputs data in units of CPI (Coherent Pulse Interval) in a radar that forms transmission and reception beams for an observation range, applies a partial SISO (Single Input Single Output) beam to SIMO (Single Input Multiple Output) in time division in units of CPI, and uses (ND+MD-1) × CPI (MD>1) data, including past CPIs, to sequentially extract and integrate ND (ND>1) CPI data excluding the CPI to which SISO is applied, extract range-Doppler (RD) data with the maximum SN ratio among the ND data, and perform detection processing using CFAR or the like to extract range cells. Then, PGA (Phase Gradient Algorithm) is performed using the slow-time data of the extracted range cells from the entire RD data of (ND+MD-1) × CPI. Autofocus) processing is performed, and the signal from the range-Doppler cell with the maximum SN is used to process the angle. For SISO beams, integration, detection, MDS identification, and angle measurement processing are performed in CPI units (time-division processing of SISO and SIMO).

[0014] That is, the radar device according to the fourth embodiment applies SISO beams to highly maneuvering targets in a time-division manner, thereby enabling efficient integration, MDS identification, angle measurement processing, and correlation tracking processing. [Brief explanation of the drawings]

[0015] [Figure 1] FIG. 1 is a block diagram showing the configuration of a transmission system and a reception system of a radar device according to the first embodiment. [Figure 2] FIG. 2 is a conceptual diagram showing the configuration of Ns divided sub-arrays used in the antenna in the first embodiment. [Figure 3] FIG. 3 is a flowchart showing the flow of transmission and reception processing in the first embodiment. [Figure 4] FIG. 4 is a conceptual diagram showing the shapes of transmission beams and reception beams by SIMO in the first embodiment. [Figure 5]FIG. 5 is a conceptual diagram showing the relationship between the timing of the PGA process and the SN in the first embodiment. [Figure 6] FIG. 6 is a characteristic diagram showing the characteristics of long-time integration using (ND+MD-1)×CPI data on the slow-time axis in the first embodiment. [Figure 7] FIG. 7 is a flowchart showing the flow of the PGA process in the first embodiment. [Figure 8] FIG. 8 is a conceptual diagram showing how the main lobe is improved by the PGA processing in the first embodiment. [Figure 9] FIG. 9 is a characteristic diagram showing the state of angle measurement processing in the first embodiment. [Figure 10] FIG. 10 is a diagram showing a three-dimensional coordinate system indicating the calculated position of the target in the first embodiment. [Figure 11] FIG. 11 is a block diagram showing the configuration of a transmission system and a reception system of a radar device according to the second embodiment. [Figure 12] FIG. 12 is a flowchart showing the flow of transmission and reception processing in the second embodiment. [Figure 13] FIG. 13 is a conceptual diagram showing how an MDS for target identification is generated in the second embodiment. [Figure 14] FIG. 14 is a conceptual diagram showing how an MDS is generated by thinning out the shift amount in the PGA process in the second embodiment. [Figure 15] FIG. 15 is a block diagram showing the configuration of a transmission system and a reception system of a radar device according to the third embodiment. [Figure 16] FIG. 16 is a flowchart showing the flow of transmission and reception processing in the third embodiment. [Figure 17] FIG. 17 is a conceptual diagram showing a situation in which, in the third embodiment, in the case of processing using long-term data, bias components occur in the target position at the time of outputting the results and in the target position of the processing results due to target movement. [Figure 18]FIG. 18 is a block diagram showing a processing system in which NN correlation processing and Kalman tracking processing are used in combination in the third embodiment. [Figure 19] FIG. 19 is a block diagram showing the configuration of a transmission system and a reception system of a radar device according to the fourth embodiment. [Figure 20] FIG. 20 is a flowchart showing the flow of transmission and reception processing in the fourth embodiment. [Figure 21] FIG. 21 is a conceptual diagram showing the state of PGA processing when a SISO beam is applied to a highly maneuvering target in a time-division manner in the fourth embodiment. [Figure 22] FIG. 22 is a conceptual diagram showing how transmission beams and reception beams are switched between SIMO and SISO in the fourth embodiment. DETAILED DESCRIPTION OF THE INVENTION

[0016] Hereinafter, embodiments will be described with reference to the drawings.

[0017] (First embodiment) A radar device according to a first embodiment will be described with reference to FIGS.

[0018] First, the configuration of a radar device according to this embodiment will be described with reference to Figures 1 and 2. Figure 1 shows the configuration of a radar device according to the first embodiment, with (a) being a block diagram showing a transmission system and (b) being a block diagram showing a reception system. Also, Figure 2 is a conceptual diagram showing the configuration of Ns divided subarrays used in the antenna of the reception system of the radar device shown in Figure 1.

[0019] The transmission system shown in FIG. 1(a) includes a signal generator 11 that generates a transmission seed signal, a modulator 12 that generates a modulated signal from the transmission seed signal, a frequency converter 13 that converts the modulated signal into an RF (Radio Frequency) signal, a pulse modulator 14 that pulse-modulates the RF signal to generate a radar signal, and a transmission antenna 15 that transmits the radar signal toward the observation range.

[0020] On the other hand, the receiving system shown in Fig. 1(b) constitutes the receiving antenna shown in Fig. 2 and includes Ns receiving subarrays (1 to Ns) 161 to 16Ns that each receive radar reflection signals, frequency converters 171 to 17Ns that frequency convert the Ns received signals to baseband, AD converters 181 to 18Ns that convert the Ns baseband received signals to digital signals, CPI data storage units 191 to 19Ns that sequentially store the digitally converted CPI-unit data in preparation for processing the next observation frame, and total CPI readers 201 to 20Ns that read data for (ND + MD - 1) × CPI of the observation frame from the CPI data storage units 191 to 19Ns. Here, MD represents the number of CPIs to be integrated, and ND represents the number of integration units of MD × CPI that are processed.

[0021] The receiving system also includes a DBF Σ beam former 21 that forms Σ beams from all CPI data of the observation frame, a per-CPI FFT processor 22 that inputs the CPI data after Σ beam formation and performs slow-time axis FFT processing for each CPI, a per-CPI pulse compressor 23 that performs pulse compression (see Non-Patent Document 1) on the CPI data that has undergone slow-time axis FFT processing, a clutter suppressor 24 that suppresses clutter near the clutter frequency of the Doppler axis in the range-Doppler (RD) data, a maximum CPI selector 25 that selects a CPI with the maximum SN from the clutter-suppressed range-Doppler (RD) data, a CFAR detector 26 that performs target detection processing using CFAR (Non-Patent Document 2) from the selected maximum CPI data, and a target range selector 27 that selects the range of a detected target.

[0022] Furthermore, the receiving system includes a ΔAZ / ΔEL beam former 28 that forms a ΔAZ beam and a ΔEL beam from all the CPI data of the observation frame, a pulse compressor 29 that inputs the CPI data after the formation of the ΔAZ beam and the ΔEL beam and performs pulse compression, a range cell extractor 30 that extracts the range cell selected by the target range selector 27 from the pulse-compressed CPI data, a PGA processor 31 that performs PGA processing (see Non-Patent Document 4) on the extracted range cell, a slow-time axis FFT processor 32 that performs FFT processing on the slow-time axis of the PGA processing result, a CFAR detector 33 that detects targets using CFAR from the output of the FFT processing on the slow-time axis, an AZ / EL monopulse angle measurer 34 that performs monopulse angle measurement (see Non-Patent Document 3) in the AZ axis and EL axis of the detected target, and an observation value output unit 35 that obtains and outputs observation values ​​converted into three dimensions (X, Y, Z) from the range and AZ / EL angle measurement values ​​at which the target is detected.

[0023] The transmission and reception processing operation of the radar device configured as above will be explained with reference to Fig. 3 to Fig. 6. Here, Fig. 3 is a flowchart showing the flow of the transmission and reception processing, Fig. 4 is a conceptual diagram showing the shapes of the transmission beam and reception beam by SIMO, Fig. 5 is a conceptual diagram showing the relationship between the timing of PGA processing and SN, and Fig. 6 is a characteristic diagram showing the characteristics of long-time integration using (ND+MD-1)×CPI data on the slow-time axis.

[0024] The radar device according to this embodiment is designed for operation capable of long-term integration, and when the observation range is wide, it may use SIMO (Single-Input and Multiple-Output) in which the observation range is covered by a single fan beam for transmission and the same range is covered by multiple pencil beams for reception, as shown in Fig. 4. The transmission and reception process of radar signals will be described with reference to the flowchart shown in Fig. 3.

[0025] In the transmission system shown in Fig. 1(a), a transmission seed signal generated by a signal generator 11 is modulated by a modulator 12, converted into an RF signal by a frequency converter 13, pulse-modulated by a pulse modulator 14, and transmitted as a radar signal from a transmitting antenna 14 toward the observation range. In the reception system shown in Fig. 1(b), radar reflection signals received by receiving antennas 161-16Ns are frequency-converted to baseband by frequency converters 171-17Ns and converted into digital signals by AD converters 181-18Ns (step S11). The digitally converted received signals are sequentially converted into data in CPI units and stored in CPI data storage devices 191-19Ns (step S12). Data for (ND+MD-1) × CPI shown in Fig. 5(a) is sequentially read by total CPI readers 201-20Ns as a PGA processing range (step S13).

[0026] Next, the DBF Σ beam former 21 forms a Σ beam from all CPI data within the PGA processing range (step S14), the per-CPI FFT processor 22 performs slow-time axis FFT processing on the input CPI data (step S15), and the per-CPI pulse compressor 23 performs pulse compression (step S16). The clutter suppressor 24 suppresses clutter near the clutter frequency of the Doppler axis in the range-Doppler (RD) data (step S17). The maximum CPI selector 25 selects (extracts) and saves the CPI with the maximum SN shown in Figure 5(b) from the clutter-suppressed RD data (steps S18 and S19). At this point, it is determined whether all CPI data within the PGA processing range has been completed (step S20). If any CPI data remains, the CPI data is changed, and the processes of steps S14 to S19 are executed. Finally, the CPI data with the maximum SN within the PGA processing range is saved. When processing of all CPI data within the PGA processing range is completed, the maximum CPI data is selected (step S22), target detection processing is performed by the CFAR detector 26 (step S23), and the target range selector 27 selects the target range (step S24).

[0027] Meanwhile, the ΔAZ / ΔEL beam former 28 forms a ΔAZ beam and a ΔEL beam from all CPI data in the PGA processing range (step S25), the pulse compressor 29 compresses the pulses (step S26), the range cell extractor 30 selects (extracts) the range cell selected by the target range selector 27 (step S27), and the PGA processor 31 performs PGA processing (step S28). Thereafter, the slow-time axis FFT processor 32 performs slow-time axis FFT processing on the PGA processing result (step S29), the CFAR detector 33 detects the target (step S30), and the AZ / EL monopulse angle finder 34 performs monopulse angle measurement in the AZ axis and EL axis (step S31). Here, it is determined whether processing of all range cells in which targets have been detected within the PGA processing range has been completed (step S32), and if any range cells remain, the range cells are changed (step S33) and the processing of steps S27 to S31 is executed.If processing of all range cells has been completed, the observation value output unit 35 calculates and outputs the observation value converted into three dimensions (X, Y, Z) from the range in which the target was detected and the AZ / EL angle measurement value (step S34), and the series of processes is completed.

[0028] Here, we explain the characteristics of long-term integration using (ND+MD-1)×CPI data on the slow-time axis. When the slow-time axis is long for a highly maneuvering target, range walk and Doppler walk occur, resulting in integration loss and a decrease in signal-to-noise ratio (SN ratio), as shown in the upper part of Figure 6. Performing PGA processing on the slow-time axis allows for highly efficient integration through Doppler correction, but applying PGA requires the extraction of target range cells, making it impossible to apply before detection. Alternatively, there is a method for integrating (ND+MD-1)×CPI data and then performing detection, but this increases the processing scale, so a method that reduces the number of integrations and achieves efficient integration is required. Therefore, we consider performing slow-time axis FFT processing using MD×CPI data, including overlaps, within the (ND+MD-1)×CPI data to obtain ND-based range-Doppler (RD) data, and then extracting range cells using the RD data with the maximum SN ratio. This is because, as shown in the lower part of Figure 6, SN is more easily accumulated by integrating for each division unit and taking the maximum value than by performing FFT processing using the entire (ND+MD-1)×CPI data, and an improvement in SN can be expected compared to when there is no division. As a result, if range cells can be extracted, by extracting range cells from the entire (ND+MD-1)×CPI data and performing PGA processing on the slow-time axis, it is possible to align the phase, correct Doppler walk, and integrate efficiently.

[0029] On the other hand, a ΔAZ beam and a ΔEL beam are formed, pulse compressed to extract range cells, and PGA processing is performed. After that, slow-time axis FFT processing is performed, targets are detected using slow-time axis CFAR, etc., AZ / EL monopulse angle measurement is performed, and observation values ​​converted from the range and AZ / EL angle measurement values ​​to 3D (X, Y, Z) are output.

[0030] The above processing will be formulated. First, there is pulse compression (Non-Patent Document 1). This obtains a signal on the range frequency-slow-time axis by range axis correlation, just like the signal that has been FFT processed on the fast-time axis. This can be formulated as follows:

[0031]

number

[0032]

number

[0033] This range-frequency-slow-time axis signal is used to perform pulse compression using range-axis inverse FFT processing.

[0034] Next, the PGA processing (see Non-Patent Document 4) of the PGA processor 28 will be described with reference to Fig. 7 to Fig. 10. Fig. 7 is a flowchart showing the flow of the PGA processing, and Fig. 8 is a conceptual diagram showing how the main lobe is improved by the PGA processing.

[0035] In FIG. 7, a range-compressed slow-time axis signal (e.g., FIG. 8(a1) or (a2)) is input (step S281), and a maximum value (peak value) exceeding a predetermined amplitude threshold is extracted from the Doppler axis signal of the range cell where a target was detected (step S282). Next, to remove the phase gradient of the maximum value relative to the Doppler axis and extract only the phase shift due to velocity and acceleration, the maximum value is shifted to zero frequency on the Doppler axis (zero shift) and the Doppler axis signal is rearranged (step S283, FIG. 8(b1) or (b2)). Next, to remove the oscillation component of the phase shift and obtain a stable correction component, a window function is multiplied to ±R (R≧1) cells centered on the maximum value (0 Doppler) to generate a signal s0 with zeros outside the window function (step S284), and this signal is then subjected to inverse FFT processing (step S285, FIG. 8(c)).

[0036]

number

[0037]

number

[0038] Next, the angle measurement process will be formulated with reference to Fig. 9 and Fig. 10. Fig. 9 is a characteristic diagram showing the state of the angle measurement process, and Fig. 10 is a diagram showing a three-dimensional coordinate system showing the calculated position of the target.

[0039] When a Doppler cell is detected from the results of PGA processing and slow-time axis FFT processing of the detected target range cell, monopulse angle measurement (see Non-Patent Document 3) can be performed. As shown in Figure 8(a), this is a method of calculating the error voltage ε shown in the following equation using the Σ beam and Δ (ΔAZ and ΔEL) beam, and measuring the angle using an error voltage table obtained in advance based on the error curve characteristics shown in Figure 8(b).

[0040]

number

[0041] From the above-mentioned angle measurement values ​​and distance measurement values ​​using the range cells, the three-dimensional coordinates (X, Y, Z) of the target shown in FIG. 910 can be calculated.

[0042]

number

[0043] As described above, the radar device according to the first embodiment forms transmit and receive beams for the observation range, inputs CPI as a unit, and uses (ND+MD-1)×CPI (MD>1) data, including past CPIs, to sequentially extract and integrate ND (ND>1) sets of CPI data. It then extracts the range-Doppler (RD) data with the maximum signal-to-noise ratio (SN) among the ND sets, performs target detection processing using CFAR or the like, extracts range cells, performs PGA (Phase Gradient Autofocus) processing using the slow-time data of the extracted range cells from the entire (ND+MD-1)×CPI RD data, and performs angle measurement processing using the range-Doppler cell signal with the maximum SN (split integration + PGA angle measurement). This allows for efficient integration to detect targets and perform angle measurement processing, even when range walk or Doppler walk occurs within the integration time.

[0044] (Second embodiment) Next, a second embodiment will be described with reference to FIGS.

[0045] Fig. 11 shows the configuration of a radar device according to the second embodiment, where (a) is a block diagram showing the configuration of the transmission system and (b) is a block diagram showing the configuration of the reception system, Fig. 12 is a flowchart showing the flow of transmission and reception processing in the second embodiment, Fig. 13 is a conceptual diagram showing how an MDS for target identification is generated in the second embodiment, and Fig. 14 is a conceptual diagram showing how an MDS is generated by thinning out the shift amount in PGA processing in the second embodiment. Note that in Figs. 11 and 12, the same parts as in Figs. 1 and 3 are designated by the same reference numerals, and redundant explanations will be omitted here.

[0046] In the first embodiment, a highly efficient integration method using (ND+MD-1)×CPI data was described. For target detection, having identification information offers significant operational benefits, and there are methods for identifying targets using MDS (see Non-Patent Document 5) and CNN (see Non-Patent Document 6). In the second embodiment, a method for creating an MDS with a reduced processing scale will be described.

[0047] In the radar device according to this embodiment, the configuration of the transmission system shown in Fig. 11(a) is the same as that shown in Fig. 1(a), and the configuration of the reception system shown in Fig. 11(b) differs from the configuration shown in Fig. 1(b) in that it adds an MDS Doppler axis rearranger 36 that rearranges target range cells selected by the target range selector 27 on the MDS Doppler axis, an MDS discriminator 37 that identifies an MDS from the rearranged target range, and an observation value discrimination processor 38 that discriminates observation values ​​from the identified MDS. Also, the flowchart of the transmission and reception process shown in Fig. 12 differs from the flowchart shown in Fig. 3 in that it adds range cell selection (step S35) that selects target range cells, MDS Doppler axis rearrangement (step S36) that rearranges the selected target range cells on the MDS Doppler axis, MDS discrimination (step S37) that identifies an MDS from the rearranged target range, and range cell end determination (step S38) and range cell change (step S39) for discriminating observation values ​​from the identified MDS.

[0048] Here, as shown in Fig. 13(a), MDS extracts the slow-time axis for each detection range cell, and uses STFT (Short-Time FFT) processing, which performs short-time FFT processing while sliding, to create the spectrogram shown in Fig. 13(b).It is possible to observe changes in Doppler with respect to the horizontal time-index axis, and since the image differs depending on the target, it can be used for target identification.

[0049] The MDS Doppler axis rearranger 36 sequentially inputs the (ND+MD-1)×CPI data in the PGA processing range shown in Fig. 14(a), extracts the extraction range for each observation frame as shown in Fig. 14(b), rearranges the data so that the Doppler change can be seen on the time-index axis as shown in Fig. 14(c), and generates the MDS as shown in Fig. 13(b). The MDS classifier 37 observes the Doppler change from the generated MDS, and the observation value discriminator 38 discriminates the observation values ​​of the Doppler change identified from the MDS to identify targets. In the processing flow, a range cell is selected (step S35) based on the target range selected in the target range cell selection process (step S24), the (ND+MD-1)×CPI data of the PGA processing range of the selected range cell is input, an MDS is created and sorted on the Doppler axis (step S36), target identification using the MDS is performed using CNN or the like (step S37), the range cell end is determined (step S38), the range cell is changed (step S39) and the range cell is switched sequentially, and the MDS identification result is output.

[0050] That is, to obtain an MDS image during long-term integration, an STFT is performed for each detection range cell, which increases the processing scale. In contrast, in this embodiment, the entire (ND+MD-1)×CPI data is divided and FFT processed for detection purposes, and the results can be utilized. That is, as shown in FIG. 6 of the first embodiment, an MDS can be generated simply by arranging the FFT processing results in MD×CPI units as time-indexes, thereby reducing the processing scale. In this arrangement, the shift amount for each STFT time-index is 1 cpi. However, when rearranging the data as shown in FIG. 14(c), the shift amount can be thinned to P (P≧1)×cpi to generate an MDS. Furthermore, if there are insufficient time-indexes in the MD×CPI data, previous MD×CPI data can be used, taking advantage of the fact that the entire observation range can be continuously observed.

[0051] As described above, the radar device according to the second embodiment generates an MDS by arranging the FFT results of ND data for the range cells extracted in the first embodiment, performs classification processing using CNN or the like, and outputs the classification results (MDS by divided integration). That is, according to the radar device according to this embodiment, even in the case of long-term integration, an MDS is generated using the processing results for detection, so that classification processing can be performed efficiently.

[0052] (Third embodiment) The configuration of the third embodiment will be described with reference to FIGS.

[0053] Fig. 15 shows the configuration of a radar device according to the third embodiment, where (a) is a block diagram showing the configuration of the transmission system and (b) is a block diagram showing the configuration of the reception system, Fig. 16 is a flowchart showing the flow of transmission and reception processing in the third embodiment, Fig. 17 is a conceptual diagram showing a state in which, in the case of processing using long-term data in the third embodiment, a bias component occurs in the target position at the time of result output and the target position of the processing result due to target movement, and Fig. 18 is a block diagram showing the processing system in the third embodiment in which NN correlation processing and Kalman tracking processing are used together. Note that in Fig. 15, the same parts as in Figs. 3 and 11 are designated by the same reference numerals, and in Fig. 16, the same parts as in Figs. 4 and 12 are designated by the same reference numerals, and duplicated explanations will be omitted here.

[0054] In the radar device according to this embodiment, the configuration of the transmission system shown in Fig. 15(a) is the same as that shown in Fig. 11(a), and the configuration of the reception system shown in Fig. 15(b) differs from the configuration shown in Fig. 11(b) in that a correlation tracker 39 and a position error corrector 40 are added. Also, the flowchart of the transmission and reception process shown in Fig. 16 differs from the flowchart shown in Fig. 12 in that correlation tracking (step S40), smoothed value correction (step S41), target number completion determination (step S42), target number change (step S43), and smoothed value output (step S44) are added.

[0055] In the first embodiment, a method for efficient integration using long-term data was described. When processing using long-term data, the target moves, resulting in a bias component between the target position at the time of outputting the results and the target position in the processing result. For simplicity, FIG. 17 shows a case where the target distance changes linearly. As shown in FIG. 17(a), the observation result is data from a point roughly in the center of the processing time compared to the true target value, resulting in a bias of approximately half the processing time. To correct this, as shown in FIG. 17(b), the predicted processing results are output for only half the processing time using velocity and acceleration components. To achieve this, the correlation tracker 39 performs correlation tracking processing (step S40) based on the observed values, and the position error corrector 40 corrects the position error by smoothing the correlation tracking results (step S41). This processing is performed for all detected targets (steps S42 and S43), and the corrected smoothed value is finally output (step S44).

[0056] Representative correlation tracking methods include NN (Nearest Neighbor) (see Non-Patent Document 7) and Kalman filter (see Non-Patent Document 8). Other methods can also be used as long as they can calculate velocity and acceleration. The Kalman filter sets a model using a state equation and an observation equation that represent a motion model.

[0057]

number

[0058] If there are multiple observed values ​​for the observation vector, one of the observed values ​​must be selected through correlation processing, and this device uses NN (Nearest Neighbor) processing. Figure 18 shows a processing system that combines NN correlation processing and Kalman tracking processing. NN correlation processing selects the observed value that is closest to the predicted value from the observed values ​​within a specified gate size centered on the predicted value, and performs tracking filter processing.

[0059] The NN correlation Kalman filter can be expressed as a recursive formula as follows: The residual vector is expressed as the Mahalanobis distance (normalized squared distance) normalized by the covariance matrix S of the residual.

[0060]

number

[0061] The observed value shown in equation (8) in the first embodiment is used as y in equation (9), and the smoothed values ​​xs of the position, velocity, and acceleration are calculated using equation (10). Then, with half the long-term integration time (MD × CPI) as L frames, the bias-corrected target smoothed value can be calculated using the following equation.

[0062]

number

[0063] For simplicity, the reference time for bias correction (Time_cal) has been described above as half the processing time, but the reference time for bias correction may be set to a time closer to the true value based on the target motion and the actual measured or calculated values ​​of the processing. Also, since correlation tracking processing is performed in observation frame units, if the number of observation frames is N × CPI (N ≥ 1), Time_cal = Reference time / (N × CPI), rounded off.

[0064] This makes it possible to output a smoothed value close to the target true value even in the case of long-term integration.

[0065] As described above, the radar device according to the third embodiment performs correlation tracking processing using the observed values ​​of the range and angle measurement results in the first or second embodiment, calculates the target speed and target acceleration, and outputs predicted values ​​at a time that is half the timestamp of the (ND+MD-1)×CPI data, thereby correcting the position deviation of the target motion in the correlation tracking results (position bias correction). This makes it possible to correct the position deviation due to the target motion over a long integration time.

[0066] (Fourth embodiment) The configuration of the fourth embodiment will be described with reference to FIGS.

[0067] Fig. 19 shows the configuration of a radar device according to the fourth embodiment, where (a) is a block diagram showing the configuration of the transmission system and (b) is a block diagram showing the configuration of the reception system, Fig. 20 is a flowchart showing the flow of transmission and reception processing in the fourth embodiment, Fig. 21 is a conceptual diagram showing the state of PGA processing when a SISO beam is applied to a highly maneuvering target in a time-division manner in the fourth embodiment, and Fig. 22 is a conceptual diagram showing the state of switching between a transmission beam and a reception beam using SIMO and a transmission beam and a reception beam using SISO in the fourth embodiment. Note that in Fig. 19, the same parts as in Figs. 3, 11, and 15 are designated by the same reference numerals, and in Fig. 20, the same parts as in Figs. 4, 12, and 16 are designated by the same reference numerals, and redundant explanations will be omitted here.

[0068] In the radar device according to this embodiment, the configuration of the transmission system shown in Fig. 19(a) is the same as that shown in Fig. 15(a). The configuration of the reception system shown in Fig. 19(b) and the processing flow of the reception system shown in Fig. 20 differ from the configuration and processing flow shown in Fig. 15(b) and Fig. 16 in that the SISO CPI readers 411-41Ns read CPI data for each subarray from the CPI data storages 191-19Ns for SISO (step S45), the DBF Σ beam former 42 forms Σ beams from all SISO CPI data of the observation frame (step S46), the per-CPI FFT processor 43 inputs the CPI data after Σ beam formation and performs slow-time axis FFT processing for each CPI (step S47), the per-CPI pulse compressor 44 performs pulse compression on the CPI data that has been subjected to slow-time axis FFT processing (step S48), and the range-Doppler (RD) data the clutter suppressor 45 suppressing clutter near the clutter frequency of the Doppler axis in the radar (step S49); a CFAR detector 46 performing target detection processing by CFAR from the clutter-suppressed range-Doppler (RD) data (step S50); a target range selector 47 selecting the range of the detected target (step S51); an MDS generator 48 selecting a cell of the target range selected by the target range selector 47 (step S52) and generating an MDS from the selected target range cell (step S53); an MDS discriminator 49 identifying an observed value from the generated MDS (step S54); and an observed value discriminator 50 sequentially changing the range cell during the observation period and discriminating the observed value from all identified MDS (steps S55 and S56).Furthermore, the receiving system includes a ΔAZ / ΔEL beam former 51 that forms a ΔAZ beam and a ΔEL beam from all SISO CPI data of the observation frame (step S57), a pulse compressor 52 that inputs the CPI data after the formation of the ΔAZ beam and the ΔEL beam and pulse compresses them (step S58), a range cell extractor 53 that extracts range cells of the observation values ​​discriminated by the observation value discriminator 50 from the pulse-compressed CPI data (steps S61 and S62), a PGA processor 54 that performs PGA processing on the extracted range cells when a target can be detected and a range cell can be extracted using the divided CPI, and a FF of the extracted range cell (or PGA processing result) on the slow-time axis. The system is equipped with a slow-time axis FFT processor 55 that performs T processing (step S59), a CFAR detector 56 that detects targets using CFAR from the FFT processing output of the slow-time axis (step S59), an AZ / EL monopulse angle measurer 57 that measures the monopulse angles of the AZ and EL axes for the detected targets (step S60), an observation value output unit 58 that calculates, stores, and outputs observation values ​​converted from the range and AZ / EL angle measurement values ​​at which the target is detected into three dimensions (X, Y, Z) (step S63), and a correlation tracker 59 that calculates the correlation of the observation value output for each detected target, tracks the target (steps S64 to S66), and outputs each smoothed value (step S67).

[0069] The radar device according to this embodiment will be described in comparison with the first, second, and third embodiments. In the first, second, and third embodiments, as shown in FIG. 4, a case where long-term integration is performed by taking advantage of the fact that observation is always ongoing in the case of SIMO operation using a transmit fan beam and a receive pencil beam within the observation range is described. Since the gain is significantly reduced with a transmit fan beam, long-term integration is essential for low RCS (Radar Cross-Section) targets, and the measures described in the first, second, and third embodiments are necessary. In this case, if the target becomes even more maneuverable, there is a possibility that long-term integration will not be sufficient, and therefore, in this embodiment, a countermeasure for this will be described.

[0070] To facilitate processing in the case of highly maneuvering targets, it is desirable to use SISO operation of the receiving pencil with the transmitting pencil, which does not require the application of long-term integration. The first, second, and third embodiments are characterized by the process of dividing the (ND+MD-1)×CPI observation data into segments for target detection, and extracting the maximum value of the RD data for each segment. In contrast, this embodiment uses some CPIs as SISO beams and integrates them in segments excluding those CPIs, thereby minimizing the impact on the results. In this case, since there is no missing data on the slow-time axis with the divided CPIs, the side lobes on the Doppler axis do not deteriorate, and false detections can be reduced.

[0071] On the other hand, if targets can be detected and range cells extracted using the divided CPI, PGA can be applied, as shown in Figures 21(a) and 21(b). In this case, the SISO CPI is zero-filled and processing is performed using all of the (ND+MD-1)×CPI observation data. The Doppler axis side lobes of the slow-time axis FFT processing results after PGA processing are slightly degraded by the zero-filled data, but because the range cell data after detection is used, no false detections occur.

[0072] 21(a) and (b), for simplicity, the number of targets in SISO operation is set to 1, and the case of 1 CPI is shown, but L CPIs can be put into SISO operation according to the number of targets L (L≧1). By performing the above processing, by switching to and applying SISO operation in a time-division manner during SIMO operation as shown in FIGS. 22(a) and (b), efficient processing is possible even for highly maneuvering targets.

[0073] The receiving system for SISO operation is as shown in Figure 18(b), and the processing flow is as shown in Figure 20. That is, while the input CPI data is saved, in the case of CPI reading for SISO processing 4b, processing is switched from SIMO processing to SISO processing in a time-division manner. Compared to SIMO, processing related to long-term integration is eliminated. The MDS can also be generated by a normal short-time Fourier transform, since processing is performed using only 1 CPI.

[0074] As described above, the radar device according to this embodiment inputs data in units of CPI to a radar that forms transmit and receive beams for an observation range. Among the data obtained by applying a partial SISO beam to SIMO in time division on a CPI basis, (ND+MD-1)×CPI (MD>1) data, including past CPIs, sequentially extracts and integrates ND (ND>1) sets of CPI data excluding the CPI to which SISO is applied. Then, it extracts the range-Doppler (RD) data with the maximum S / N ratio among the ND sets and performs target detection processing using CFAR or the like to extract range cells. Then, it performs PGA processing using the slow-time data of the extracted range cells from the entire (ND+MD-1)×CPI RD data. Then, it performs angle measurement processing using the signal of the range-Doppler cell with the maximum S / N ratio. For the SISO beam, it performs integration, detection, MDS identification, and angle measurement processing in units of CPI. That is, according to the radar device of the fourth embodiment, by applying SISO beams in a time-division manner to highly maneuvering targets, it is possible to efficiently perform integration, MDS identification, angle measurement processing, and correlation tracking processing.

[0075] The present invention is not limited to the above-described embodiments, and the components can be modified and embodied in practice without departing from the spirit of the invention. Furthermore, various inventions can be created by appropriately combining multiple components disclosed in the above-described embodiments. For example, some components may be omitted from all the components shown in the embodiments. Furthermore, components from different embodiments may be appropriately combined. [Explanation of symbols]

[0076] 11...signal generator, 12...modulator, 13...frequency converter, 14...pulse modulator, 15...transmitting antenna, 161-16Ns...subarray, 171-17Ns...frequency converter, 181-18Ns...AD converter, 191-19Ns...CPI data storage, 201-20Ns...total CPI reader, 21...DBFΣ beam former, 22...per CPI FFT processor, 23...per CPI pulse compressor, 24...clutter suppressor, 25...maximum CPI selector, 26...CFAR detector, 27...target range selector, 28...ΔAZ / ΔEL beam former, 29...pulse compressor, 30...range cell extractor, 31...PGA processor, 32...slow-time axis FFT processor, 33...CFAR detector, 34...AZ / EL monopulse angle finder, 35...observation value output unit, 36...MDS Doppler axis rearranger, 37...MDS classifier, 38...observation value discriminator, 39...correlation tracker, 40...position error corrector, 411~41Ns...SISO CPI reader, 42...DBFΣ beamformer, 43...per CPI FFT processor, 44...per CPI pulse compressor, 45...clutter suppressor, 46...CFAR detector, 47...target range selector, 48...MDS generator, 49...MDS classifier, 50...observation value discriminator, 51...ΔAZ / ΔEL beamformer, 52...pulse compressor, 53...range cell extractor, 54...PGA processor, 55...slow-time axis FFT processor, 56...CFAR detector, 57...AZ / EL monopulse angle finder, 58...observation value output unit, 59...correlation tracker.

Claims

1. a transmitting / receiving means for forming a transmitting beam and a receiving beam of SIMO (Single Input Multiple Output) for the observation range; a signal processing means for receiving a signal received from the transmitting / receiving means in units of CPI (Coherent Pulse Interval), sequentially extracting and integrating ND (ND>1) sets of CPI data using data for (ND+MD-1) x CPI (MD>1) (MD is the number of CPIs to be integrated, and ND is the number of processing times for the integral unit of MD x CPI) including past CPIs, extracting range-Doppler data with the maximum SN from the ND sets, performing target detection processing to extract a range cell, performing PGA (Phase Gradient Autofocus) processing using slow-time axis FFT processing data of the extracted range cell from the entire range-Doppler data of (ND+MD-1) x CPI, performing angle measurement processing using the signal of the range-Doppler cell with the maximum SN, and outputting the observed value; A radar device comprising:

2. 2. The radar device according to claim 1, wherein the signal processing means generates an MDS (Micro Doppler Spectrum) by division and integration by arranging the slow-time axis FFT results of ND types of data for the extracted range cells, performs MDS identification processing, and outputs the identification result.

3. 2. The radar device according to claim 1, wherein the signal processing means performs correlation tracking processing using observed values ​​of range and angle measurement results, calculates the velocity and acceleration of the detected target, and outputs predicted values ​​at a time that is half the timestamp of the (ND+MD-1)×CPI data, thereby correcting positional deviation of the target movement in the correlation tracking results.

4. 2. The radar device according to claim 1, wherein the signal processing means applies a partial SISO (Single Input Single Output) beam to a SIMO (Single Input Multiple Output) in a time-division manner in CPI units, and performs integration, detection, MDS identification, and angle measurement processing in CPI units for the SISO beam.

5. It is used in a radar device that forms a SIMO (Single Input Multiple Output) transmission beam and a reception beam for an observation range, The received signal of the receiving beam is input in units of CPI (Coherent Pulse Interval), and using data for (ND+MD-1) x CPI (MD>1) (MD is the number of CPIs to be integrated, and ND is the number of processing times for the MD x CPI integration unit), including past CPIs, ND (ND>1) CPI data is extracted in order and integrated, and the range-Doppler data with the maximum SN among the ND data is extracted, target detection processing is performed, and the range cell is extracted. Of the entire range-Doppler data of (ND+MD-1) x CPI, PGA (Phase Gradient Autofocus) processing is performed using the slow-time axis FFT processing data of the extracted range cell, and angle measurement processing is performed using the signal of the range-Doppler cell with the maximum SN, and the observation value is output. Radar signal processing method.

Images