Radar device and method for processing its radar signals
The radar system addresses the S/N ratio decrease in large targets by using modulated pulses, range compression, and machine learning to enhance detection accuracy and precision in target range cells and SAR imaging.
Patent Information
- Authority / Receiving Office
- JP · JP
- Patent Type
- Applications
- Current Assignee / Owner
- KK TOSHIBA
- Filing Date
- 2024-12-25
- Publication Date
- 2026-07-07
AI Technical Summary
Conventional radar systems face challenges in detecting large targets with high accuracy due to decreased signal-to-noise ratio (S/N) caused by range resolution, leading to inaccurate detection of target range cells.
The radar device employs modulated pulse transmission, range compression, and FFT processing to improve S/N ratio by dividing the range frequency band and integrating amplitude, combined with machine learning for target range selection and range walk correction to enhance accuracy.
Enables high-precision detection of large targets by improving signal-to-noise ratio and correcting range walk, allowing for accurate target range cell detection and high-precision SAR imaging.
Smart Images
Figure 2026112846000001_ABST
Abstract
Description
Technical Field
[0001] This embodiment relates to a radar device and a radar signal processing method thereof.
Background Art
[0002] In a conventional radar device, when observing a target such as a large ship at a high SN compared to the range resolution after pulse compression and further selecting the range cell range of the target detected by the observation, since it is separated by the range resolution, the signal SN decreases, and there is a problem that the target range cell range cannot be detected with high accuracy.
Prior Art Documents
Non-Patent Documents
[0003]
Non-Patent Document 1
Non-Patent Document 2
Non-Patent Document 3
Non-Patent Document 4
Non-Patent Document 5
Summary of the Invention
Problems to be Solved by the Invention
[0004] As described above, conventional radar systems have the problem that, for targets that are large compared to their range resolution, the signal-to-noise ratio (S / N) decreases because they are separated by the range resolution, and furthermore, they cannot detect the target range cell range with high accuracy.
[0005] This embodiment has been made in view of the above problems, and aims to provide a radar device and a radar signal processing method that can detect targets with a high signal-to-noise ratio even when the target is large compared to the range resolution, and can further detect the range cell range of the target with high accuracy. [Means for solving the problem]
[0006] To solve the above problems, the first embodiment is a radar device that transmits and receives modulated N (N≧1) pulses and range-compresses the received signal along the fast-time axis. In this device, the received signal is subjected to an FFT (Fast Fourier Transform) along the fast-time axis to obtain a signal along the range frequency axis. This signal is then pulse-compressed while remaining a broadband signal. Meanwhile, the range frequency axis is divided into L (L≧1) sections, and each section is narrow-band pulse-compressed. After amplitude integration of the pulse-compression results for each L section, a target is detected, and a broadband pulse-compressed signal with a predetermined width (target range cell ±P, P≧1) is extracted centered on the target range cell. The target range is then selected as the range where the signal-to-noise (SN) value exceeds a predetermined threshold. With this configuration, by dividing the range frequency band and pulse-compressing it, the range resolution is brought closer to the target length. Furthermore, by amplitude integration of the processing for each section, the SN is improved, enabling target detection. The range range in which the target exists is limited, centered on the detected range cell, and the target range is selected as the range where the amplitude level exceeds a predetermined threshold.
[0007] Furthermore, the second embodiment, in the configuration of the first embodiment, uses the extracted pulse compression results as input and employs machine learning inference processing using the results learned with the target range range as a training signal to select the target range cell range. With this configuration, the target range can be limited to the detected range cell and selected by machine learning processing.
[0008] Furthermore, in the third embodiment, using N pulses sequentially from the modulated Nall (Nall=M×N, N≧1, M>1) pulses of the first and second embodiments, the target range range is extracted using the method of the first or second embodiment. Then, the center position of the target range range range is extracted, and using the slow-time axis-range axis data of the Nall pulse, the range walk of the target center position is corrected, and an FFT of the slow-time axis is performed on the range-slow-time axis data to obtain a SAR image. With this configuration, by extracting the center position of the target existence range, the range walk of the range-slow-time axis can be corrected with high accuracy even when the target length is large, thus enabling the acquisition of a high-precision SAR image.
[0009] As described above, the radar system according to this embodiment can detect targets with a high signal-to-noise ratio even when the target is large compared to the range resolution, and can further detect targets with high accuracy by selecting the range cell range of the target. [Brief explanation of the drawing]
[0010] [Figure 1] Figure 1 is a block diagram showing the configuration of a radar device according to the first embodiment. [Figure 2] Figure 2 is a flowchart showing the flow of the target detection process after signal transmission and reception in the first embodiment. [Figure 3] Figure 3 shows waveform diagrams illustrating the narrowband compression and broadband compression processes in the target detection process in the first embodiment. [Figure 4] Figure 4 is a block diagram showing the configuration of a radar system according to the second embodiment. [Figure 5]Figure 5 is a flowchart showing the flow of the target detection process after signal transmission and reception in the second embodiment. [Figure 6] Figure 6 shows waveform diagrams illustrating the narrowband compression and wideband compression processes in the target detection process in the second embodiment. [Figure 7] Figure 7 shows a method for selecting the target length using machine learning in the second embodiment. [Figure 8] Figure 8 is a block diagram showing the configuration of a radar system according to the third embodiment. [Figure 9] Figure 9 is a flowchart showing the flow of the target detection process after signal transmission and reception in the third embodiment. [Figure 10] Figure 10 is a waveform diagram showing how the target image is acquired from range walk correction and ISAR processing in the third embodiment. [Modes for carrying out the invention]
[0011] The embodiments will be described below with reference to the drawings. In the description of each embodiment, the same parts will be denoted by the same reference numerals, and redundant descriptions will be omitted.
[0012] (First embodiment) A radar device according to the first embodiment will be described with reference to Figures 1 to 3. Figure 1 is a block diagram showing the configuration of a radar system. In Figure 1, in the transmission system, a transmission signal generator 11 modulates the transmission seed signal with a reference signal to generate a transmission signal such as a modulated pulse, which is then converted into an analog signal by a DA converter 12, converted into a high-frequency (RF) signal by a frequency converter 13, and pulse modulation at PRI (Pulse Repetition Interval) intervals is performed by a pulse modulator 14, and the signal is transmitted as a radar signal by a transmission antenna 15.
[0013] In the receiving system, the radar signal reflected by the target, etc., is captured by the receiving antenna 16, the frequency is converted to baseband by the frequency converter 17, and the signal is converted to a digital signal by the AD converter 18, and the following target detection process is performed.
[0014] First, for the received signal converted into a digital signal, FFT processing in the slow-time axis is performed by slow-time axis FFT 19, and FFT processing in the fast-time axis is performed by received signal fast-time axis FFT 20. Also, for the reference signal used for transmission signal generation, FFT processing in the fast-time axis is performed by reference signal fast-time axis FFT 21. Subsequently, the received signal that has undergone FFT processing in the fast-time axis by received signal range frequency divider 22 is divided into range frequencies at a predetermined interval, and similarly, the reference signal that has undergone FFT processing in the fast-time axis by reference signal range frequency divider 23 is divided into range frequencies at the same interval as the received signal. Next, narrow-band pulse compression is performed for each division unit by narrow-band pulse compressor 24, the pulse compression effect for each division unit is amplitude-added (integrated) by inter-band amplitude integrator 25, and a target is detected by CFAR (see Non-Patent Document 2) or the like by target detector 26 to extract the target range cell. On the other hand, wide-band pulse compression is performed in a wide band using the received signal and the reference signal before frequency division by wide-band pulse compressor 27 to obtain the signal in the range (fast-time) axis, a target range candidate range for extracting the target range from the target detection result and the signal in the range axis is extracted by target range candidate range extractor 28, range cells with a wide-band pulse compression effect exceeding a predetermined amplitude threshold are extracted by target range selector 29, and the target range is selected from the minimum amplitude value and the maximum amplitude value in that range, and the target information within the target range is output.
[0015] The processing operation of the radar device with the above configuration will be described with reference to FIGS. 2 and 3. FIG. 2 is a flowchart showing the flow of target detection processing after signal transmission and reception, and FIG. 3 is a waveform diagram showing the states of narrowband compression processing and broadband compression processing in the target detection processing. Generally, for each pulse transmitted at the PRI (Pulse Repetition Interval) interval, data on the slow-time axis and fast-time axis are acquired in units of range cells within the PRI. However, in this embodiment, assuming the case where the number of hits N = 1, the processing on the slow-time axis is omitted, and the processing on the fast-time axis will be described. Generally, in the case of N (N>1) hits, for example, an FFT on the slow-time axis (19) may be performed after the AD conversion (18) in FIG. 1.
[0016] In FIG. 2, the radar device according to this embodiment transmits and receives broadband signals (step S11). Specifically, a transmission type signal is generated by the signal generator 11, a modulation signal is generated by the modulator 12, converted into a high-frequency signal by the frequency converter 13, pulse-modulated by the pulse modulator 14, and the pulse generated from the transmission antenna 5 is transmitted. On the other hand, the reception antenna 16 receives the reflected signal of the transmission pulse, the received signal is frequency-converted to the baseband by the frequency converter 17, and converted into a digital signal by the AD converter 18.
[0017] Next, the radar device performs an FFT process on the slow-time axis (19, step S12, FIG. 3(b)) on the received signal converted into a digital signal (FIG. 3(a)), performs an FFT process on the fast-time axis (20, step S13), and further performs an FFT process on the fast-time axis on the reference signal which is the transmission modulation signal (21, step S14). In the following formulations, for the sake of clarity, the description of the signal on the slow-time axis is omitted.
[0018]
Equation
[0019] In this case, if the chirp bandwidth is insufficient and it is necessary to artificially improve the resolution of the range cell, the received signal Sin and the reference signal Sref can be replaced using zero-padding.
[0020]
number
[0021] Next, to prevent signal-to-noise ratio degradation due to the target length being greater than the range resolution, we consider reducing the range resolution. To do this, we divide the range frequency bandwidth of both the received signal Sin(ff) and the reference signal Sref(ff) (22,23, steps S15,S16, Figure 3(c)).
[0022]
number
[0023] Narrowband pulse compression is performed for each of these division units (24, steps S17, S18, S19, S20, Figure 3(d)).
[0024]
number
[0025] Next, the amplitude summation (amplitude integration) of the narrowband pulse compression results for each division unit is performed (25, step S21).
[0026]
number
[0027] The above describes narrowband pulse compression by dividing the frequency range. Next, we will perform broadband pulse compression using the entire frequency range (27, step S23). This can be done using the received signal and reference signal from equation (1) before frequency division.
[0028]
number
[0029]
number
[0030]
number
[0031] Once the target range cell is identified from the narrowband pulse compression results, a candidate target range is extracted from the wideband pulse compression results (28, step S24), a target range is selected from the signals within the candidate target range using an amplitude threshold (29, step S25), and the selected target range is saved (step S26). Here, the target range is changed until a predetermined number of targets is reached (steps S27, S28), and the processes in steps S23 to S26 are repeatedly executed. As the number of targets reaches the target range within the target range where the target range cell obtained by the narrowband pulse compression effect is saved, a candidate range for extracting the target range can be set as shown in Figure 3(g). By setting a candidate range, false detection of the target range can be reduced. Among the range cells in the candidate range, the range cell in which the broadband pulse compression result is the maximum amplitude value is extracted, a predetermined amplitude threshold is set based on that maximum amplitude value, and range cells with broadband pulse compression results exceeding that amplitude threshold are extracted, and the target range can be extracted (selected) from the minimum and maximum values of the range (Step S29, Figure 3(h)).
[0032] Therefore, with the radar device configured as described above, when transmitting and receiving modulated N (N≧1) pulses and compressing the received signal along the fast-time axis, the received signal is first subjected to an FFT along the fast-time axis to obtain a signal along the range frequency axis. This signal is then pulse-compressed while remaining a broadband signal. Meanwhile, the range frequency axis is divided into L (L≧1) sections, and each section is subjected to narrowband pulse compression. After amplitude integration of the pulse compression results for each L section, a target is detected, and a broadband pulse-compressed signal with a predetermined width (target range cell ±P, P≧1) is extracted centered on the target range cell. The target range is then selected as the range where the amplitude level exceeds a predetermined threshold from the maximum signal-to-noise ratio (SN) value. Thus, by dividing the range frequency band and pulse-compressing it, the range resolution can be brought closer to the target length. Furthermore, by amplitude integration of the processing for each section, the SN can be improved to detect the target. This allows the target range to be limited to the range where the target exists, centered on the detected range cell, and the target range can be selected as the range where the amplitude level exceeds a predetermined threshold.
[0033] (Second embodiment) A radar device according to the second embodiment will be described with reference to Figures 4 to 7. Here, Figure 4 is a block diagram showing the configuration of the radar device according to the second embodiment, Figure 5 is a flowchart showing the flow of target detection processing after signal transmission and reception in the second embodiment, Figure 6 is a waveform diagram showing the narrowband compression processing and wideband compression processing in the target detection processing in the second embodiment, and Figure 7 is a diagram showing a method for selecting the target length using machine learning in the second embodiment. Note that in Figures 4 and 5, the same processing parts as in Figures 1 and 2 are denoted by the same reference numerals, and only the different parts will be described here.
[0034] In the first embodiment, a method was described for extracting candidate target ranges based on narrowband pulse compression results and selecting a target length (target range) using an amplitude threshold. In this embodiment, a method for selecting a target length using machine learning is described.
[0035] In this embodiment, the radar device has a transmission system that is the same as that of the first embodiment, as shown in Figure 4(a), and a reception system that employs machine learning processing instead of an amplitude threshold in the target range selector 30, as shown in Figure 4(b). Furthermore, as shown in step S24 of Figure 5 and Figure 6(g), the extraction result of the target range candidate range (step S24) is subjected to target range selection processing (step S30) using machine learning, and the extracted target range range is saved (step S26). For machine learning, as shown in Figure 7, it is necessary to define and learn the input signals (range cells (1~Q) and output signals (1~Q)). The input signals are the amplitude signals (range cells 1~range cell Q) (Q>1) of the range candidate range signals extracted by the method of the first embodiment. The output is a teacher signal in which the range cell range of the known target range range (target selection range) is set to a value of 1, and everything else (non-target selection range) is set to a value of 0. The number of range cells will be the same as the input range cells, from 1 to Q.
[0036] Therefore, with the radar device configured as described above, the offline trained results of the neural network (see Non-Patent Documents 3 and 4), which has been trained using training data that generates input signals for various target lengths, can be stored, and by inputting the candidate range extracted in the first embodiment, the target range can be output.
[0037] (Third embodiment) A radar system according to the third embodiment will be described with reference to Figures 8 to 10. Here, Figure 8 is a block diagram showing the configuration of the radar system according to the third embodiment, Figure 9 is a flowchart showing the flow of target detection processing after signal transmission and reception in the third embodiment, and Figure 10 is a waveform diagram showing how the target image is acquired from range walk correction to ISAR processing in the target detection processing of the third embodiment. Note that in Figures 8 and 9, the same processing parts as in Figures 4 and 5 are denoted by the same reference numerals, and only the different parts will be described here.
[0038] In radar systems, when observing target data as an image using range-Doppler data, range walk correction is crucial. To achieve this, it is necessary to correct the range walk at the center of the target range. However, when the target length is large, the reflection point changes along the slow-time axis, making it difficult to correct the range walk with high accuracy. As a countermeasure, the first and second embodiments describe a method for correcting the range walk with high accuracy using a method for selecting the target range.
[0039] As shown in Figure 8, the radar system according to this embodiment selects a target range range using a target range range selector 29 or 30, then performs an inverse FFT on the slow-time axis signal using a slow-time axis inverse FFT 31, extracts the center of the target range range using a target range range center extractor 32, corrects the range walk with respect to the slow-time axis using a range walk corrector 33, performs an FFT on the range walk corrected slow-time axis signal using a slow-time axis FFT 34, and acquires and transmits a target image using a target image outputter 35.
[0040] In this embodiment, we consider the target detection process of a radar device that transmits and receives Nall (Nall=N×M, N≧1, M>1) pulses. First, as shown in Figure 9, if we use N hits sequentially in the 1 to M target detections performed by the processes in steps S31 and S32, we can extract the target range for M times using the method of the first or second embodiment. Using this, we can calculate the center position (distance) for M times, and assuming that the center position for each of the N hits is the same, we can extract the center position for Nall=N×M hits. After extracting the center position for each of the M processes, in step S33, we save the result of an inverse FFT (31) on the slow-time axis for processing the Nall hits, and after the completion of the M processes, we concatenate the results to calculate the data for Nall=N×M hits and extract the center position of the target range for each of the M processes (32). As described above, as shown in Figure 10(a), the range walk with respect to the slow-time axis for Nall hits can be determined. By correcting this range walk as shown in Figure 10(b) (33, step S34), and making the range the same with respect to the slow-time axis, an FFT process is performed on the slow-time axis for Nall hits (step S35), resulting in a high-precision SAR image (Non-Patent Literature 5) as shown in Figure 10(c) and outputting it (step S36).
[0041] The above describes a method for generating high-precision target images by accurately extracting the center position of the target range. When this method is applied to correlation tracking, where N hits constitute one frame and the observed values for each frame are used, the observed position of the target stabilizes at the center of the target, thereby improving the accuracy of correlation tracking. Furthermore, if only correlation tracking is required, stable correlation tracking can be achieved by extracting the target center position every N hits and performing M correlation tracking cycles using that range.
[0042] It should be noted that the present invention is not limited to the above embodiments, and the components can be modified and implemented in practice without departing from the spirit of the invention. Furthermore, various inventions can be formed by appropriately combining the multiple components disclosed in the above embodiments. For example, some components may be deleted from all the components shown in the embodiments. Moreover, components from different embodiments may be appropriately combined. [Explanation of Symbols]
[0043] 11...Transmit signal generator, 12...DA converter, 13...Frequency converter, 14...Pulse modulator, 15...Transmit antenna, 16...Receive antenna, 17...Frequency converter, 18...AD converter, 19...Slow-time axis FFT, 20...Received signal fast-time axis FFT, 21...Reference signal fast-time axis FFT, 22...Received signal range frequency divider, 23...Reference signal range frequency divider 23, 24...Narrowband pulse compressor, 25...Interband amplitude integrator, 26...Target detector, 27...Broadband pulse compressor, 28...Target range candidate range extractor, 29...Target range range selector (amplitude threshold), 30...Target range range selector (machine learning), 31...Slow-time axis inverse FFT, 32...Target range range center extractor, 33...Range walk corrector, 34...Slow-time axis FFT, 35...Target image outputter.
Claims
1. In a radar device that transmits and receives modulated N (N≧1) pulses and range-compresses the received signal along the fast-time axis, A means for obtaining a signal on the range frequency axis by processing the received signal on the fast-time axis using FFT (Fast Fourier Transform), A means for pulse-compressing the signal in the aforementioned range frequency axis while maintaining it as a broadband signal, The means for dividing the range frequency axis into L (L≧1) and performing narrowband pulse compression on each division unit, A means for detecting a target by amplitude integration of the L-way narrowband pulse compression results obtained by the L-division, A means for extracting a signal of a predetermined width wideband pulse compression result centered on the target range cell in which the aforementioned target was detected, means for selecting a target range from the signal obtained from the broadband pulse compression result of a predetermined width, and A radar device equipped with the following.
2. The radar device according to claim 1, wherein the selection of the target range is a range in which the signal obtained by compressing a broadband pulse of a predetermined width is taken as input and exceeds a predetermined level threshold above the maximum value of the signal to noise (SN).
3. The radar device according to claim 1, wherein the selection of the target range is performed using a machine learning inference process that takes the signal of the result of broadband pulse compression of a predetermined width as input and uses the target range as a training signal.
4. A means for extracting a target range using N pulses in sequence from the modulated Nall (Nall = M × N, N ≥ 1, M > 1) pulses transmitted and received, and for extracting the center position of that target range, A means for correcting the range walk of the target center position using the slow-time axis-range axis data of the Nall pulse, A means for obtaining a SAR image by performing an FFT on the slow-time axis using the range-walk corrected range-slow-time axis data, and A radar device according to claim 1, comprising the above. With this configuration, by extracting the central position of the target presence range, the range walk along the range-slow-time axis can be corrected with high accuracy even when the target length is large, thereby enabling the acquisition of a high-precision SAR image. The radar device according to claim 1.
5. It is used in radar equipment that transmits and receives modulated N (N≧1) pulses and range-compresses the received signal along the fast-time axis. The received signal is processed using FFT (Fast Fourier Transform) on the fast-time axis to obtain the signal on the range frequency axis. The signal in the aforementioned range frequency axis is pulse-compressed while remaining a broadband signal. The aforementioned range frequency axis is divided into L (L≧1), and narrowband pulse compression is performed on each division unit. The target is detected by amplitude integration of the L-way narrowband pulse compression results obtained by the L-division method. The signal obtained from the broadband pulse compression result of a predetermined width is extracted, centered on the target range cell where the aforementioned target was detected. The target range is selected from the signal obtained by compressing a broadband pulse of the predetermined width. A method for processing radar signals from a radar system.