Radar system and radar signal processing method

Abstract

To ensure that, even when observed values are available in abundance, correlation tracking is hardly lost.SOLUTION: According to an embodiment, when the number of observed values exceeds a prescribed threshold, regarding the observed values in the N(N≥1)-th cycle, a clustering process is carried out and the observed values are replaced with the cluster, when the prescribed threshold is not exceeded, P (P≥1) particles are generated for each flight track in a tracking process, by using the observed values not having been subjected to clustering, a correlation gate is set for each particle, and correlation processing with M (M≥2) observed values in the correlation gate or the clustered observed values is carried out, so as to calculate the likelihood of each particle, then conducting a tracking process by a particle filter.SELECTED DRAWING: Figure 2

Description

[Technical Field]

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

[0002] Many conventional radar systems that perform correlation tracking (Non-Patent Document 2) using particle filters (Non-Patent Document 1) are based on the assumption that the number of observation points is small. In particular, the correlation between each particle and the observation value is often calculated by performing correlation processing using the nearest observation value, such as NN (Nearest Neighbor) processing of the observation values within a gate (Non-Patent Document 3). In this case, when there are multiple observation values, such as a group of targets, the particle distribution cannot be accurately determined, and correlation tracking may be lost. There is also a method that applies cluster analysis to the observation values (Patent Document 1). However, in order to compensate for the lack of reflection points, this method considers one observation as one cycle, synthesizes the three-dimensional positions of multiple cycles, and then performs cluster analysis. This differs from the processing when a large number of reflection points occur within one cycle. [Prior art documents] [Patent documents]

[0003] [Patent Document 1] Patent No. 6773606 [Non-patent literature]

[0004] [Non-Patent Document 1] Particle Filter, Katayama, ''Nonlinear Kalman Filter'', Asakura Publishing, pp.141-152(2011) [Non-patent document 2] Correlation Tracking, Yoshida, 'Revised Radar Technology', Institute of Electronics, Information and Communication Engineers, pp.254-259 (1996) [Non-patent document 3] NN correlation processing, Samuel S. Blackman, 'Design and Analysis of Modern Tracking Systems', Artech House, pp.8-11. (1999) [Non-patent document 4] Cluster Analysis, Sebastian Raschka, 'Python Machine Learning Programming', Impress, pp.297-319 (2016) [Non-patent document 5] DBSCAN (Density-based Spatial Clustering of Applications with Noise), Sebastian Raschka, 'Python Machine Learning Programming', Impress, pp.319-323 (2016) [Non-patent document 6] CFAR (Constant False Alarm Rate) Processing, Yoshida, 'Revised Radar Technology', Institute of Electronics, Information and Communication Engineers, pp.87-89 (1996) [Non-Patent Document 7] Pulse Compression, Yoshida, 'Revised Radar Technology', Institute of Electronics, Information and Communication Engineers, pp.278-280 (1996) [Non-patent document 8] Monopulse, Yoshida, 'Revised Radar Technology', Institute of Electronics, Information and Communication Engineers, pp.260-264 (1996) Summary of the Invention [Problem to be solved by the invention]

[0005] As described above, a conventional radar system that performs correlation tracking using a particle filter has a problem in that correlation tracking is easily lost when there are a large number of observations, such as for a group of targets.

[0006] An object of this embodiment is to provide a radar system and a radar signal processing method that are less likely to cause correlation tracking loss even when there are a large number of observations such as a group of targets. [Means for solving the problem]

[0007] In order to solve the above problems, the radar system according to this embodiment receives, in a receiving system, a reflected wave of a single pulse or modulated N (N≧1) pulse signal transmitted from a transmitting system, and performs correlation tracking processing using observation values that have been signal-processed using signals in range cell units within a PRI (Pulse Repetition Interval) interval for each pulse transmitted at the PRI interval, and outputs target information. The receiving system determines whether the number of observation values in the N (N≧1)th cycle exceeds a predetermined threshold, and if the number of observation values exceeds the threshold, performs cluster processing to replace the observation value with a cluster and output it as a cluster observation value, and if the number of observation values does not exceed the threshold, outputs an observation value without performing cluster processing, generates P (P≧1) particles for each track of the correlation tracking processing, sets a correlation gate for each particle, performs correlation processing on M (M≧2) observation values or cluster observation values within the correlation gate to calculate the likelihood of each particle, and performs tracking processing using a particle filter. The error voltage is obtained by monopulse angle measurement processing of the observed value, and the absolute value of the imaginary part of the error voltage is Predetermined value If it is greater, the likelihood of the particle for said observation is multiplied by a predetermined coefficient. [Brief explanation of the drawings]

[0008] [Figure 1] FIG. 1 is a block diagram showing the configuration of a transmission system and a reception system of a radar system according to the first embodiment. [Figure 2] FIG. 2 is a flowchart showing the flow of processing in the reception system of the first embodiment. [Figure 3] FIG. 3 is a diagram showing three-dimensional coordinates for three-dimensional conversion processing in three-dimensional position identification in the first embodiment. [Figure 4] FIG. 4 illustrates an example of the cluster analysis process according to the first embodiment. [Figure 5] FIG. 5 is a diagram illustrating an example of DBSCAN processing according to the first embodiment. [Figure 6] FIG. 6 is a diagram illustrating an example of correlation tracking processing according to the first embodiment. [Figure 7]FIG. 7 is a diagram illustrating an example of processing by the particle filter according to the first embodiment. [Figure 8] FIG. 8 is a flowchart showing the flow of processing in the reception system of the second embodiment. [Figure 9] FIG. 9 is a flowchart showing the flow of processing in the reception system of the third embodiment. [Figure 10] FIG. 10 is a flowchart showing the flow of processing in the reception system of the fourth embodiment. DETAILED DESCRIPTION OF THE INVENTION

[0009] Hereinafter, embodiments will be described with reference to the drawings.

[0010] (First embodiment) (cluster analysis + correlation processing) FIG. 1 is a block diagram showing the configuration of a radar system according to the first embodiment, where (a) is a block diagram showing the configuration of a transmission system, and (b) is a block diagram showing the configuration of a reception system.

[0011] In the transmission system shown in Figure 1(a), a signal generator 11 generates a transmission seed signal, a modulator 12 modulates and multiplexes transmission information onto the transmission seed signal, a frequency converter 13 converts the modulated signal into a high-frequency signal, a pulse modulator 14 pulse-modulates the high-frequency signal to generate a transmission pulse train, and a transmission antenna 15 transmits N (N≧2) hit pulses.

[0012] In the receiving system shown in Fig. 1(b), a reflected wave of a pulse signal transmitted from a transmitting antenna 15 is received by a receiving antenna 21, and the received signal is frequency-converted to baseband by a frequency converter 22 and converted to a digital signal by an AD converter 23. Next, a signal processor 24 performs pulse compression (PC) (Non-Patent Document 7) and an FFT (Fast Fourier Transform) on the slow-time axis (N-hit PRI (Pulse Repetition Interval) axis), and an observation value is detected by a CFAR (Constant False Alarm Rate; see Non-Patent Document 6) 25.

[0013] Next, distance and speed meter 26 measures the distance and speed of the detected observation values, angle meter 27 measures the angle using monopulse angle measurement (Non-Patent Document 8) or the like, and three-dimensional position identifier 28 calculates the (X, Y, Z) values of the observation values. Next, if the number of observation values obtained by three-dimensional position identifier 28 exceeds its threshold, cluster analyzer 29 analyzes the presence or absence of clusters on the (X, Y, Z) axes and determines whether the detection is a target or a false positive based on the presence or absence of clusters. If a cluster is present, the cluster is replaced as the observation value, correlation processing is performed by correlation processor 2A, likelihood is calculated by likelihood calculator 2B, and the observation value is tracked by tracking processor 2C, and target information such as a smoothed value for each track is output.

[0014] In the above radar system, the transmission system and the reception system may be integrated or may be installed at separate locations.

[0015] The processing operation of the radar system having the above configuration will be described with reference to FIG.

[0016] 2 is a flowchart showing the flow of processing in the reception system in the first embodiment. In this embodiment, a long-time integration method is adopted, and data is acquired in range cell units within the PRI for each pulse transmitted at PRI intervals, and signal processing is performed using the acquired data.

[0017] First, a received signal is acquired (step S11), which is received by the receiving antenna 21, frequency-converted to baseband by the frequency converter 22, and converted to a digital signal by the AD converter 23. Next, signal processing such as pulse compression (PC) and FFT on the slow-time axis is performed (step S22), and an observation value is detected by CFAR (step S13).

[0018] Next, the detected observation values are subjected to distance measurement processing (step S14), speed measurement processing (step S15), and angle measurement processing (step S16), and (X, Y, Z) values are calculated by position identification processing using three-dimensional conversion (step S17).

[0019] Next, it is determined whether the number of observed values exceeds the threshold (step S18), and if it exceeds the threshold (large), cluster analysis is performed on the (X, Y, Z) axes (step S19), but if the number of observed values does not exceed the threshold (small), cluster analysis is not performed. Thereafter, in order to distinguish between target and false detection based on the presence or absence of clusters, the clusters are replaced as observed values and output together with observed values other than clusters (step S20).

[0020] Here, the observed values within the gate are extracted in range cell units within the PRI (step S21), correlation processing (step S22), likelihood calculation (step S23), and tracking processing (step S24) are performed, followed by particle filtering (step S25), particle changes are measured (step S26), and the processing of steps S21 to S26 is repeated until the changes disappear. When the particle filtering processing is completed, target information such as smoothed values for each track is output as the result of the tracking processing (step S27), completing the series of processing steps.

[0021] In the above process, an example of the process applied in this embodiment will be specifically described with reference to FIGS.

[0022] FIG. 3 is a diagram showing three-dimensional coordinates for the three-dimensional conversion process of the target in step S17 (R: distance, AZ: azimuth angle (azimuth angle), EL: elevation angle (elevation angle)).

[0023] First, we will formulate the 3D position identification method. Phase monopulse angle measurement can be performed by calculating the error voltage ε shown in the following equation using the Σ beam and Δ beam, and then using a previously acquired error voltage table.

[0024]

number

[0025]

number

[0026] If the number of observed values is large, cluster analysis (see Non-Patent Document 4) is performed in step S19. Figure 4 shows an example of cluster analysis processing. Figure 4(a) shows the state before cluster analysis, and Figure 4(b) shows the state after cluster analysis. That is, the observed values may include false detections due to thermal noise in cluttered environments or when the signal-to-noise ratio (SN) of the observed values is low. Compared to false detections, the observed values of a group of targets are often observed as clustered reflection points, so cluster analysis may enable the separation of targets and false detections into clusters or other entities. Furthermore, in the case of a group of multiple targets in close proximity, cluster analysis allows them to be separated and processed.

[0027] There are various cluster analysis methods, but as one example, the DBSCAN method (Non-Patent Document 5) will be described. Figure 5 shows an example of DBSCAN processing. As shown in Figure 5, the DBSCAN method sets a radius ε and a number of observations within the radius ε, MinPts, and classifies clusters based on differences in density. Multiple clusters are generated for each cluster that satisfies the set radius ε and number of observations, MinPts, and others can be classified as noise. Specifically, a point that has at least the set number of observations, MinPts, of neighboring points within the set radius ε is considered a core point, and a point that has fewer neighboring points within the radius ε than MinPts is considered a border point. Points that are neither core points nor border points are considered noise points. This makes it possible to distinguish between reflection points due to targets (core points and border points) and false detections (noise points).

[0028] The clusters calculated by cluster analysis are used as new M (M≧2) observed values, and correlation tracking processing is performed. If the number of observed values is small, correlation tracking processing is performed using the original observed values as they are. Figure 6 is a diagram showing a typical correlation tracking processing. In correlation tracking processing, as shown in Figure 6, a correlation gate is set for each cycle, centered on the predicted value of the previous cycle, and observed values within the correlation gate are extracted. A representative extraction method is NN (Nearest Neighbor) processing (see Non-Patent Document 3), which extracts the observed value closest to the predicted value. A smoothed value is calculated using this NN value and the predicted value, and a predicted value for the next cycle is calculated.

[0029] Here, the particle filter (see Non-Patent Document 1) will be explained using Fig. 7. Fig. 7 is a diagram showing an example of processing by the particle filter. Fig. 7 is an explanatory diagram for one wake, and in the case of multiple wakes, similar processing is performed for each wake.

[0030] First, as shown in Figure 7(a), P (P ≥ 1) particles are generated at random positions using random numbers for the smoothed value (t-1) of one wake. Then, as shown in Figure 7(b), a predicted value (t) is calculated for the P particles (prior distribution). A correlation gate is set for each particle of this predicted value, and observation values are extracted, for example, using neural network processing, to calculate the likelihood of each particle. Based on this likelihood, the posterior distribution can be calculated by multiplying the prior distribution by the likelihood, as shown in Figure 7(c). In the posterior distribution diagram in Figure 7(c), the vertical axis is extended when the likelihood is large. Based on this posterior distribution, a resampling process is performed, setting a large number of particles in the large distribution areas and a small number of particles in the small distribution areas, as shown in Figure 7(d). A smoothed value for the next cycle is generated, and the above process is repeated for the following cycles. To output the smoothed value for each cycle using this particle filter, the average value of the positions of the P particles for each wake is calculated.

[0031] The particle filter has the advantage that it functions effectively even when the error distribution of the target motion model or observation model does not follow a Gaussian distribution, and is effective when there are multiple targets, such as a group of targets.

[0032] As described above, in this embodiment, for the observation values in the Nth (N≧1) cycle, if the number of observation values exceeds a predetermined threshold, cluster processing is performed to replace the observation value with a cluster. If the number of observation values does not exceed the predetermined threshold, observation values that are not subjected to cluster processing are used to generate P (P≧1) particles for each tracking track, set a correlation gate for each particle, and perform correlation processing with M (M≧2) observation values (or cluster observation values) within the gate to calculate the likelihood of each particle, and then perform tracking processing using a particle filter. In other words, in the case of a large number of observation values, isolated false positives are first suppressed by cluster analysis, and multiple group targets are separated before correlation processing is performed. During correlation processing, instead of extracting only a single observation value as in NN correlation, multiple observation values are used, thereby enabling accurate extraction of group targets, thereby enabling highly accurate correlation tracking processing.

[0033] (Second embodiment) Calculation of likelihood based on all observed values FIG. 8 is a flowchart showing the processing flow of the receiving system of the second embodiment. In FIG. 8, the same parts as in FIG. 2 are denoted by the same reference numerals. The receiving system shown in FIG. 8 differs from the first embodiment in the processing of step S23a, which is characterized by the likelihood calculation processing being performed using the sum of all observation values within the gate. That is, in the first embodiment, an example was described in which NN correlation values were used during correlation processing for each particle. This method has low processing costs, but in the case of a group of targets, there is a possibility that a false positive may be detected by chance, resulting in the calculation of a likelihood with a large error, which may degrade the accuracy of correlation tracking and lead to a loss. Therefore, this embodiment takes measures to address this issue.

[0034] In this embodiment, the likelihood of a particle is calculated for all observed values within the correlation gate of each particle, and the sum of these values is used as the likelihood of the particle.

[0035]

number

[0036] This likelihood λ(p) is used to perform processing such as resampling. The added value may include multiplication by a coefficient. Also, an average value may be included.

[0037] As described above, in this embodiment, the correlation process involves calculating the likelihood of M observation values within the gate of the Pth particle, and the sum of these is used as the likelihood of the Pth particle. That is, likelihoods are calculated for all observation values within the correlation gate of each particle and added together to reduce the influence of specific observation values, including erroneous detections, thereby making it easier to continue correlation tracking.

[0038] (Third embodiment) Calculation of likelihood using the center of gravity of all observed values Fig. 9 is a flowchart showing the processing flow of the reception system of the third embodiment. In Fig. 9, the same parts as in Fig. 2 and Fig. 8 are designated by the same reference numerals. The reception system shown in Fig. 9 differs from the first and second embodiments in the processing of step S23b, and this embodiment is characterized in that the likelihood is calculated using the center of gravity of all observation values extracted within the gate.

[0039] In order to use all observed values, the center of gravity (center of gravity value) of all observed values is calculated. This is formulated as follows:

[0040]

number

[0041] As described above, in this embodiment, the correlation process calculates the center of gravity of M observations within the gate of the Pth particle, and the likelihood of the Pth particle is determined as the likelihood of the Pth particle with respect to that center of gravity. In other words, the likelihood is calculated by calculating the center of gravity of all observations within the correlation gate of each particle, which reduces the influence of specific observations, including false positives, and thereby makes it easier to continue correlation tracking.

[0042] (Fourth embodiment) Likelihood coefficient based on the imaginary part of the error voltage Fig. 10 is a flowchart showing the flow of processing in the receiving system of the fourth embodiment. In Fig. 10, the same parts as in Fig. 2 are assigned the same reference numerals. The receiving system shown in Fig. 10 differs from the first embodiment in the processing of step S28, where the correlation processing result obtained in step S22 and the error voltage obtained in the angle measurement processing of step S16 are used to calculate a likelihood coefficient for the imaginary part of the error voltage (step S28), and the process proceeds to the next likelihood calculation step S23.

[0043] This embodiment is characterized by the fact that likelihood calculation is performed using the sum of all observation values within a gate. That is, this embodiment describes the case where there are multiple reflection points at one observation point. If the radar's range and angle resolution is insufficient, even if multiple targets exist, they will be combined and observed as a smaller number of targets. Even in this case, it is necessary to recognize that there are multiple targets in order to correctly calculate the likelihood. Therefore, attention is focused on the error voltage during monopulse angle measurement.

[0044]

number

[0045] This error voltage is observed as a real value in the case of a single target, but an imaginary part occurs in the case of multiple targets, so if the imaginary part is large, it can be determined that there are multiple targets. By utilizing this property, if the absolute value of the imaginary part is large, it is recognized as multiple targets, and by multiplying it by a correction coefficient after calculating the likelihood, this can be reflected in the likelihood.

[0046]

number

[0047] Here, we have described phase monopulse angle measurement, but squint angle measurement using a squint beam squinted on the AZ axis (EL axis) can also be used instead of the difference beam. In this case, the complex signal of the following equation can be used to calculate the imaginary part.

[0048]

number

[0049] As described above, in this embodiment, if the imaginary part of the error voltage of the monopulse angle measurement (phase monopulse, squint monopulse) of the observation value is greater than a predetermined threshold, the likelihood of a particle for that observation value is multiplied by a predetermined coefficient.

[0050] That is, even when the observation value is one point, multiple reflection points may be included due to the constraints of the resolution of the radar device or the receiving device. Therefore, the presence or absence of multiple reflection points is observed using the imaginary part of the monopulse error voltage, and when the imaginary part is large, the number of reflection points is large. By using this information, the likelihood is increased, which equivalently increases the number of particles, making it easier to continue correlation tracking.

[0051] Furthermore, 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 formed 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]

[0052] 11... signal generator, 12... modulator, 13... frequency converter, 14... pulse modulator, 15... transmitting antenna, 21...receiving antenna, 22...frequency converter, 23...AD converter, 24...signal processor, 25...CFAR, 26...range and speed finder, 27...angle finder, 28...3D position identifier, 29...cluster analyzer, 2A...correlation processor, 2B...likelihood calculator, 2C...tracking processor.

Claims

1. A radar system in which a single pulse or a reflected wave of a modulated N (N≧1) pulse signal transmitted from a transmission system is received by a reception system, and for each pulse transmitted at a PRI (Pulse Repetition Interval) interval, a correlation tracking process is performed using an observation value obtained by signal processing using a signal in range cell units within the PRI interval, and target information is output, a means for determining whether or not the number of observed values in the Nth (N≧1)th cycle exceeds a predetermined threshold in the receiving system; means for performing cluster processing to replace the observed value with a cluster when the number of observed values exceeds the threshold, and outputting the replaced observed value as a cluster observed value; means for outputting the observations without clustering if the number of observations does not exceed the threshold; means for generating P particles (P≧1) for each track of the correlation tracking process and setting a correlation gate for each particle; means for performing correlation processing on M (M≧2) observation values or cluster observation values within the correlation gate to calculate likelihood of each particle and performing tracking processing using a particle filter; means for acquiring an error voltage by monopulse angle measurement processing of the observed value, and multiplying the likelihood of a particle for the observed value by a predetermined coefficient when the absolute value of the imaginary part of the error voltage is greater than a predetermined value; A radar system comprising:

2. 2. The radar system according to claim 1, wherein the correlation tracking process calculates and adds the likelihoods of M observation values within the correlation gate of the Pth particle, and the added value of the likelihoods of the M observation values is the likelihood of the Pth particle.

3. 2. The radar system according to claim 1, wherein the correlation tracking process calculates the centroid of the M observation values within the correlation gate of the Pth particle, and calculates the likelihood of the Pth particle with respect to the centroid of the M observation values.

4. The radar system receives a reflected wave of a single pulse or modulated N (N≧1) pulse signal transmitted from a transmission system in a reception system, and performs correlation tracking processing using observation values that are signal-processed using signals in range cell units within a PRI (Pulse Repetition Interval) interval for each pulse transmitted at the PRI interval, and outputs target information. In the receiving system, it is determined whether or not the number of observed values exceeds a predetermined threshold for the observed values in the Nth (N≧1) cycle; If the number of observed values exceeds the threshold, a clustering process is performed to replace the observed values with clusters and output them as cluster observed values; If the number of observed values does not exceed the threshold, output the observed values without performing cluster processing; generating P particles (P≧1) for each track of the correlation tracking process and setting a correlation gate for each particle; performing correlation processing on M (M≧2) observation values or cluster observation values within the correlation gate to calculate likelihood of each particle, and performing tracking processing using a particle filter; An error voltage is obtained by monopulse angle measurement processing of the observed value, and if the absolute value of the imaginary part of the error voltage is greater than a predetermined value, the likelihood of the particle for the observed value is multiplied by a predetermined coefficient. A radar signal processing method for a radar system.

Images