JP2018205175A - Radar device and radar signal processing method thereof - Google Patents

Abstract

To suppress misdetection and extract only a target.SOLUTION: When performing N(N≥2) frame observation, with the unit of observing a radar observation range M(M≥2) times defined as one frame, a radar device of an embodiment of the present invention arrays values of M times of observation per frame into three or two-dimensional coordinates, performs cluster analysis by a difference between a goal on coordinate axes and the distribution of misdetection density, extracts a cluster as a goal candidate, and outputs an observation value or representative value in a cluster for N frames. This means that misdetection is suppressed and only a goal is extracted by performing cluster analysis by a difference between a goal on coordinate axes and the distribution of misdetection density.SELECTED DRAWING: Figure 1

Description

  The present embodiment relates to a radar apparatus and a radar signal processing method thereof.

  In the conventional radar system, when detecting a low RCS (Radar Cross Section) target, the CFAR (Constant False Alarm Rate) threshold is reduced to increase sensitivity. However, reducing the threshold increases false detection. In addition, when observing a low altitude target, false detection increases due to the effect of clutter. As described above, when the number of false detections increases, a large number of verification beams (dedicated verify beams) are required for discriminating whether or not the target is detected after detection in order to establish a track for tracking. As a result, there is a problem that transmission / reception beam scheduling (beam management) for efficiently observing the target within a predetermined period of time increases, making it impossible to observe all the targets.

CFAR processing, Yoshida, 'Revised radar technology', IEICE, pp.87-89 (1996) Cluster analysis, Sebastian Raschka, Python, 'Machine learning programming', Impress, pp.297-319 (2016) Least-squares estimation, Nakamizo, 'Signal analysis and system identification processing', Corona, pp.10-17 (1987) DBSCAN (Density-based Spatial Clustering of Applications with Noise) Sebastian Raschka, Python Machine Learning Programming, Impress, pp.319-323 (2016) Correlation Tracking, Yoshida, 'Revised Radar Technology', IEICE, pp.254-259 (1996) Neural network, Okaya, 'Deep learning', Kodansha, pp.7-26 (2014) Convolutional Neural Network, Okaya, 'Deep Learning', Kodansha, pp.79-110 (2014) Recursive neural network, Okaya, 'Deep Learning', Kodansha, pp.111-130 (2014)

  As described above, in the conventional radar system, if the detection of CFAR threshold for high sensitivity or the number of false detections due to the influence of clutter in the case of target observation in the low sky increases, it is discriminated whether the target is detected after detection. For this reason, a large number of verification beams are required, and there are problems such as the restriction of beam management becomes large and all targets cannot be observed.

  The present embodiment has been made in view of the above problems, and an object thereof is to provide a radar apparatus capable of suppressing erroneous detection and extracting only a target with high accuracy and a radar signal processing method thereof.

  In order to solve the above problem, according to the present embodiment, when N (N ≧ 2) frames are observed with a unit for observing the radar observation range M (M ≧ 2) times as one frame, each frame is measured. M observations are arranged in three-dimensional or two-dimensional coordinates, cluster analysis is performed based on the difference in density distribution between the target and false detection on the coordinate axis, target candidate clusters are extracted, and observation values in N frames of clusters are extracted. Or representative value is output.

  In other words, the radar apparatus according to the present embodiment can suppress erroneous detection and extract only the target by performing cluster analysis based on the difference in density distribution between the target and the erroneous detection on the coordinate axes.

1 is a block diagram showing a schematic configuration of a radar apparatus according to a first embodiment. The figure which shows a mode that a cluster analysis is carried out for every flame | frame divided and slid for every three-dimensional data of multiple scan in 1st Embodiment. The figure which shows the mode of the false detection suppression by the cluster distribution using the difference of density distribution, and target detection in 1st Embodiment. The block diagram which shows schematic structure of the radar apparatus which concerns on 2nd Embodiment. The figure which shows the mode of correlation tracking in 2nd Embodiment. The figure which shows a mode that a predetermined gate is set centering | focusing on the smooth value continuous for every frame in 2nd Embodiment. The figure which shows a mode that a missing is suppressed when there exists an observation value missing at the time of cluster detection in 2nd Embodiment. The block diagram which shows schematic structure of the radar apparatus which concerns on 3rd Embodiment. The figure which shows a mode that a cluster analysis parameter is changed again in 3rd Embodiment, and a cluster analysis is performed. The block diagram which shows schematic structure of the radar apparatus which concerns on 4th Embodiment. The figure which shows a mode that the target range is set with an interpolation curve in 4th Embodiment. The block diagram which shows schematic structure of the radar apparatus which concerns on 5th Embodiment. The figure for demonstrating a DBSCAN system in 5th Embodiment. The block diagram which shows schematic structure of the radar apparatus which concerns on 6th Embodiment. The figure for demonstrating a neural network system in 6th Embodiment. The figure for demonstrating the unit used for a neural network in 6th Embodiment. The figure which shows the example plotted in the point of three-dimensional coordinate (X, Y, Z) based on the coordinate of the target candidate extracted by cluster distribution detection and target candidate extraction as input data in 6th Embodiment .

  Hereinafter, embodiments will be described with reference to the drawings. In the description of each embodiment, the same portions are denoted by the same reference numerals, and redundant description is omitted.

(First Embodiment)-Cluster Distribution As a radar device, in the case of pulse modulation or continuous wave, or as pulse modulation, LPRF having an ambiguity in speed according to the division of PRI (Pulse Repetition Interval). Low Pulse Repetition Frequency: MPRF (Medium Pulse Repetition Frequency) with ambiguity in distance and speed, HPRF (High Pulse Repetition Frequency: High pulse repetition) with ambiguity in distance In the case of frequency), the case of LPRF pulse modulation will be described here for the sake of simplicity.

  The radar apparatus according to the first embodiment will be described with reference to FIGS. 1 to 3. FIG. 1 is a block diagram showing a schematic configuration thereof, FIG. 2 is a diagram showing a state of performing cluster analysis for each frame divided and divided for each of three-dimensional data of a plurality of scans, and FIG. 3 is based on a cluster distribution using a difference in density distribution. It is a figure which shows the mode of false detection suppression and target detection.

  As shown in FIG. 1, the radar apparatus according to the present embodiment forms a monopulse antenna by a plurality of antenna elements, transmits a transmission signal through the antenna 1 that forms a transmission / reception beam of Σ and Δ, and the antenna 1. A transmitter / receiver 2 that receives a reflected signal from the transmitter, and a target distance / angle is observed from the received signal obtained by the transmitter / receiver 2 to calculate a three-dimensional position coordinate, and a detection process using a cluster distribution is performed. And a signal processor 3 for outputting an observation value.

  The transceiver 2 includes a transceiver 21 and a modulation controller 22. The transmission / reception unit 21 transmits a pulse modulated wave of a monopulse generated by the pulse modulation of LPRF of the modulation control unit 22 through the antenna 1, receives a reflected wave of the transmission output, and receives reception signals (Σ beams and Δ beams) (Hereinafter referred to as Σ signal and Δ signal).

The signal processor 3 includes a Σ observation processing unit 31, a CFAR processing unit 32, a distance measurement unit 33, a Δ observation processing unit 34, a cell extraction unit 35, an angle measurement unit 36, a three-dimensional position coordinate output unit 37, and a cluster analysis unit. 38.
The Σ observation processing unit 31 receives a Σ reception signal using a monopulse beam, performs signal processing such as pulse compression or FFT according to the transmission waveform using CPI data, and obtains range-Doppler (RD) data. .
The CFAR processing unit 32 uses the RD data of the Σ signal generated by the monopulse output of the antenna 1 to set a predetermined threshold by the CFAR and detects a reflection point cell.
The distance measuring unit 33 measures the target distance by converting the time axis to the distance axis for the cell of the reflection point obtained by CFAR (ranging).

The Δ observation processing unit 34 receives a Δ reception signal by a monopulse beam, and performs signal processing such as pulse compression and FFT according to the transmission waveform using the CPI data, similarly to the Σ observation processing unit 31. Range-Doppler (RD) data is obtained.
The cell extraction unit 35 uses the processing result of the CFAR processing unit 33 to extract the same cell as the detection cell of the Σ signal.
The angle measuring unit 36 outputs the AZ and EL angles for each extracted cell by performing monopulse angle measurement processing with the target distance obtained by the distance measuring unit 33 for the cell extracted by the cell extracting unit 35. To do.

  The three-dimensional position coordinate output unit 37 calculates and outputs a target three-dimensional position coordinate from the target distance and angle obtained through the angle measuring unit 36.

  The cluster analysis unit 38 performs cluster analysis using the density difference that the cell density near the target having a large amplitude is high and is small near the false detection, and when the number of reflection points is not enough in one CPI processing, An observation value obtained by adding a plurality of CPI detection cells is output.

  In the above configuration, a radar signal processing method according to the present embodiment will be described below.

  First, a pulse-modulated wave is transmitted / received through the antenna 1 (2), and signal processing such as pulse compression and FFT is performed according to the transmission waveform using the output of the Σ signal by the monopulse antenna beam to obtain a range-Doppler signal. (31), a predetermined threshold is set by CFAR (Non-Patent Document 1) to detect a target (32), and a target distance is output from the detected range cell (33).

  Further, the same signal processing is performed for the Δ signal of the monopulse antenna beam (34), the same cell as the detection cell of the Σ signal is extracted (35), and the angles of AZ and EL by monopulse angle measurement processing are output. (36). If the distance and angle are known, three-dimensional position coordinates can be output (37).

  This three-dimensional coordinate data is subjected to cluster analysis (Non-Patent Document 2) (38). Cluster analysis is a method of grouping and classifying input values according to a predetermined condition. In this embodiment, the cluster analysis is used for discriminating a target from erroneous detection due to clutter or noise and extracting only the target. The discrimination between the target and the erroneous detection uses, for example, that the density of reflection points near the target is larger than that in the case of the erroneous detection.

  In the case of observing a plurality of scans, as shown in FIG. 2, each three-dimensional data of scans M, M + 1,..., M + m is divided into sliding frames N, N + 1,. . FIG. 3 shows the state of false detection suppression and target detection by the cluster distribution using the density distribution difference. That is, as shown in FIG. 3A, cluster analysis based on the density difference is performed for false detection and targets for a plurality of scans, and as shown in FIG. 3B, clusters of only high-density targets are extracted.

When the target cluster is extracted, the observation value in the cluster is output as it is, or the centroid value according to the following equation using the average value and amplitude of the three-dimensional coordinates in the cluster is output as the observation value as the representative value.

  As described above, the radar apparatus according to the present embodiment, when observing N (N ≧ 2) frames with a unit for observing a predetermined observation range M (M ≧ 2) times as one frame, M for each frame. The observed values for the batch are arranged in three-dimensional or two-dimensional coordinates, cluster analysis is performed, target candidate clusters are extracted, and the observed values or representative values in the clusters for N frames are output. That is, erroneous detection can be suppressed and only the target can be extracted by cluster analysis based on the difference in density distribution between the target and the erroneous detection on the three-dimensional coordinate axis.

Second Embodiment—Error Detection Suppression by Correlation Tracking A radar apparatus according to a second embodiment will be described with reference to FIGS. 4 shows a schematic configuration thereof, FIG. 5 shows a state of correlation tracking, FIG. 6 shows a state in which a predetermined gate is set around a continuous smooth value for each frame, and FIG. 7 is missing at the time of cluster detection. The figure shows how to suppress omission when there are observed values.

  In the configuration of the first embodiment, erroneous detection that is an observation value other than the extracted cluster may remain. In this embodiment, the countermeasure is described.

  The radar apparatus according to the present embodiment includes a correlation tracking unit 39 as shown in FIG. 4, and after detecting a cluster by the same method as in the first embodiment, correlation tracking processing is performed using a representative value of the cluster (non-patent). Reference 5) is performed. As shown in FIG. 5, in the correlation tracking, first, the representative value of the first scan is set as a predicted value, and the representative value closest to the predicted value is selected from the representative values in a predetermined gate. Is calculated. A predetermined gate is set as shown in FIG. 6 centering on a continuous smooth value for each frame, and an observed value inside the gate is extracted while suppressing (deleting) an observed value outside the gate as an erroneous detection. In this case, as shown in FIG. 7 (a), even if there is an observation value missing at the time of cluster detection, as shown in FIG. 7 (b), by re-extracting the observation value in the correlation tracking gate, Missing can be suppressed.

  As described above, the radar apparatus according to the second embodiment suppresses erroneous detection by performing correlation tracking processing using clusters of target candidates for N (N ≧ 1) frames. As described above, the residual error detection can be suppressed by performing the correlation tracking process using the extracted cluster.

Third Embodiment—Multiple Cluster Analysis A radar apparatus according to the third embodiment will be described with reference to FIGS. 8 and 9. FIG. 8 shows the schematic configuration, and FIG. 9 shows a state in which the cluster analysis is performed again by changing the parameters of the cluster analysis.

  In the first and second embodiments, erroneous detection may remain. The radar apparatus according to the present embodiment is made for the countermeasure, and, as shown in FIG. 8, the cluster analysis unit 38a performs a plurality of times of cluster analysis by changing parameters. That is, in a state where the distribution as shown in FIG. 9A is obtained, cluster detection is first performed by the same method as in the first or second embodiment. As a result, when there is a residual erroneous detection as shown in FIG. 9B, the density distribution of the observed values changes. Therefore, in this embodiment, as shown in FIGS. 9C and 9D, cluster analysis is sequentially performed by changing parameters. For example, in the case of the DBSCAN system (Non-Patent Document 4) described in the fifth embodiment, the parameters include the gate radius and the number of observation values in the gate. By repeating this cluster analysis a predetermined number of times (≧ 2), erroneous detection can be suppressed. At this time, although it is assumed that the target observation value is suppressed, the missing target observation value can be restored using the method described in the second or fourth embodiment.

  As described above, the radar apparatus according to the present embodiment suppresses erroneous detection by performing cluster analysis processing a plurality of times (P ≧ 2) times. As a result, residual error detection that cannot be suppressed by one cluster analysis can be suppressed.

(Fourth embodiment)-Target extraction (fitting) by cluster interpolation
A radar apparatus according to the fourth embodiment will be described with reference to FIGS. 10 and 11. FIG. 10 shows the schematic configuration, and FIG. 11 shows how the target range is set by the interpolation curve.

  In the first to third embodiments, the technique for suppressing the erroneous detection has been described. At this time, the target cluster may be missing. Although the technique for suppressing the influence of the omission by the correlation tracking process of the second embodiment has been described, in this embodiment, other countermeasures will be described.

  As shown in FIG. 10, the radar apparatus according to this embodiment has a configuration in which a target interpolation unit 3 </ b> A is arranged between the cluster analysis unit 31 a and the correlation tracking unit 39 of the second embodiment. That is, in this embodiment, as shown in FIGS. 11A to 11D, after extracting clusters, a target candidate range is extracted using observation values of L (L ≧ 1) frames. At this time, an interpolation curve is extracted using a least square curve (Non-Patent Document 3) or the like.

Here, a method of calculating an interpolation function by the least square method will be described. When the explanatory variable is expressed by the vector x and the objective variable is expressed by the vector y, the following expression is obtained.

The calculation of the least square curve of the data of equation (2) corresponds to the calculation of the parameter θ that minimizes the following equation.

When f is a polynomial, θ is a coefficient of each term as follows.

[Theta] (ap in the case of a polynomial) of equation (3) can be generally calculated as a nonlinear estimation problem under the conditions of the following equation (Non-Patent Document 3).

The equation (5) can be generally calculated by repeating the following equation.

  A region having a predetermined width shown in FIG. 11D is set around the fitting curve calculated by the above processing, and all observation values in the region are extracted. This makes it possible to restore the target observation value that was lost during cluster detection.

  Note that the method of this embodiment uses (xq, yq) data for L frames when calculating the interpolation curve, but 1 to L, 2 to L + 1, ... while sliding the L frame data. For example, after the data after the first L frame is acquired, a continuous interpolation curve can be obtained and the missing data can be corrected by cluster analysis. The target interpolation value thus extracted is output as an observation value (X, Y, Z) as it is, or the observation value is output after correlation tracking processing (39) in order to further suppress erroneous detection.

  As described above, after extracting the target candidates for N frames, the radar apparatus according to the fourth embodiment calculates a fitting curve using the observation values, and extracts the observation values in the region having a predetermined width. Then, the missing target observation value in the target cluster area is interpolated. As described above, even when the target observation value is missing due to the cluster analysis, the target cluster range is interpolated by the interpolation process, so that the missing target observation value can be suppressed.

(Fifth Embodiment) -DBSCAN System A radar apparatus according to a fifth embodiment will be described with reference to FIGS. FIG. 12 shows the schematic configuration, and FIG. 13 shows processing by the DBSCAN system.

  In this embodiment, a method using DBSCAN (Non-Patent Document 4) as a cluster analysis method will be described.

  As shown in FIG. 12, the radar apparatus according to the present embodiment employs the DBSCAN method in the cluster analysis unit 38b. As shown in FIG. 13, the circle range of radius ε and the number of observations MinPts within this circle range are obtained. Set and classify clusters by density difference. A plurality of clusters are generated for each cluster satisfying the set radius ε and the number of observations MinPts, and the others can be classified as noise.

  Specifically, a point that has at least the set number of observations MinPts within the set radius ε is a core point, and if the number of adjacent points within the radius ε is less than MinPts, it is considered a border point. It is. Points that are neither core points nor border points are noise points. Thereby, the reflection point (core point and border point) by a target and misdetection (noise point) can be distinguished (nonpatent literature 4).

  As described above, the radar apparatus according to the present embodiment uses DBSCAN as the cluster analysis method. Thereby, the cluster analysis by the difference of the density distribution of a target and false detection can be performed efficiently.

(Sixth Embodiment) -Neural Network A radar apparatus according to a sixth embodiment will be described with reference to FIGS. FIG. 14 shows the schematic configuration, FIG. 15 shows the concept of the neural network system, FIG. 16 shows the configuration of units used in the neural network, and FIG. 17 shows cluster distribution detection and target candidate extraction as input data. An example is shown in which plotting is performed at points of three-dimensional coordinates (X, Y, Z) based on the extracted target candidate coordinates.

  In this embodiment, attention is paid to the density difference between the target and the false detection as a cluster detection method. Furthermore, in order to enhance the discrimination ability between the target and the erroneous detection, a method of identifying using a neural network (Non-patent Document 6) by inputting coordinate data in a predetermined range centered on the selected target cluster can be applied. As a specific example of the neural network (NN), a convolutional neural network (Non-Patent Document 6), a recursive neural network (Non-Patent Document 7), or the like can be used. In this case, in addition to the density difference between the target and the false detection, a difference in spatial spread characteristics can be extracted as a feature amount for discrimination.

  As shown in FIG. 14, the radar apparatus according to the present embodiment has a configuration in which a target candidate extraction filter 3B is arranged between the target interpolation unit 3A and the correlation tracking unit 39.

The target candidate extraction filter 3B includes a three-dimensional data generation unit 3B1, a column vector conversion unit 3B2, an NN (neural network) processing unit 3B3, and a filter 3B4. The target interpolation unit 3A extracts the detection result from the cluster distribution. Based on the target candidates, the neural network NN identifies whether it is a target or a false detection, and only the target candidates are extracted. FIG. 15 (multilayer perceptron method) shows a configuration example of the NN used for this purpose, and FIG. 16 shows the unit configuration (Non-Patent Document 6). Each unit constituting the NN is formulated as follows.

As the activation function, various applications such as a sigmoid function (Non-Patent Document 6) can be applied.

  As the input data, based on the coordinates of the target candidates extracted by the cluster distribution detection 31b and the target candidate extraction 33, as shown in FIG. 17, plotted at the points of the three-dimensional coordinates (X, Y, Z), Three-dimensional data (Nx, Ny, Nz) is converted into a one-dimensional column vector (1 to Nx × Ny × Nz) including a point in a space where no target exists, and is input to NN. As the learning of the weight of NN, by using a simulator or an actual measurement value, known three-dimensional learning data of a true value (target or false detection teacher signal) is generated, and the generated signal (X, Y, Z ) And a teacher signal (whether it is a target or a false detection). The identification of the target or false detection corresponds to binary classification or multi-class classification (number of classes = 2).

In general, when classifying into multiple classes, training data for learning can be defined by the following equation.

The error function E is defined by the square error of the following equation.

Select w so that this is the smallest. For this purpose, a stochastic gradient descent method (Non-Patent Document 6) or the like may be used.

  In the actual processing after learning, the NN of FIG. 15 is operated using a fixed weight, and the target or false detection or the target candidate filter 3B4 is passed, and the correlation tracking unit 39 is used only with the target candidate. And the target observed values (X, Y, Z) are output. As described above, it is possible to output the observed value of the target signal with less influence of false detection.

  As described above, the radar apparatus according to the sixth embodiment plots three-dimensional coordinates using the observation value coordinates of the target cluster candidate extracted by the cluster analysis, and receives the three-dimensional plot coordinates as an input. NN) is used to filter whether the target or false detection. Thereby, the target and the erroneous detection can be identified using the NN process, and the process in which the erroneous detection is further suppressed can be performed.

  Note that the present invention is not limited to the above-described embodiment as it is, and can be embodied by modifying the components without departing from the scope of the invention in the implementation stage. In addition, various inventions can be formed by appropriately combining a plurality of components disclosed in the embodiment. For example, some components may be deleted from all the components shown in the embodiment. Furthermore, constituent elements over different embodiments may be appropriately combined.

1 ... antenna,
2 ... transceiver, 21 ... transceiver, 22 ... modulation controller,
DESCRIPTION OF SYMBOLS 3 ... Signal processor, 31 ... Sigma observation processing part, 32 ... CFAR processing part, 33 ... Distance measuring part, 34 ... Delta observation processing part, 35 ... Cell extraction part, 36 ... Angle measuring part, 37 ... Three-dimensional position coordinate Output unit, 38, 38a, 38b ... cluster analysis unit, 39 ... correlation tracking unit, 3A ... target interpolation unit, 3B ... target candidate extraction filter, 3B1 ... three-dimensional data generation unit, 3B2 ... column vector conversion unit, 3B3 ... NN (Neural network) processing unit, 3B4... Filter.

Claims (7)

An array that arranges M observation values for each frame into three-dimensional or two-dimensional coordinates when observing the radar observation range M (M ≧ 2) times as one frame and N (N ≧ 2) frames. Means,
A cluster analysis is performed on the arrayed observation values based on the difference in density distribution between the target and the false detection on the coordinate axis to extract target candidate clusters, and observation values or representative values in clusters for N frames from the target candidate clusters And a cluster analysis means for outputting a radar apparatus.   2. The radar apparatus according to claim 1, further comprising correlation tracking means for performing correlation tracking processing using clusters of target candidates for N (N ≧ 1) frames.   The radar apparatus according to claim 1, wherein the cluster analysis unit performs the cluster analysis process P (P ≧ 1) times.   Further, after extracting the target candidates for the N frames, a fitting curve using the observed values is calculated, the observed values in the region having a predetermined width are extracted, and the target observed values in the target cluster region are extracted. The radar apparatus according to claim 1, further comprising interpolation means for interpolating the missing part.   The radar apparatus according to claim 1, wherein the cluster analysis unit uses a DBSCAN (Density-based Spatial Clustering of Applications with Noise) method as a cluster analysis method.   Further, the target cluster candidate extracted by the cluster analysis is plotted on the three-dimensional coordinates, the three-dimensional plot coordinates are input, and the target or false detection filter processing is performed using the neural network (NN). The radar apparatus according to claim 1, further comprising filter processing means for performing the following. The unit for observing the radar observation range M (M ≧ 2) times is one frame, and when observing N (N ≧ 2) frames, M observation values for each frame are arranged in three-dimensional or two-dimensional coordinates,
A cluster analysis is performed on the arrayed observation values based on the difference in density distribution between the target and the false detection on the coordinate axis to extract target candidate clusters, and observation values or representative values in clusters for N frames from the target candidate clusters Radar signal processing method for a radar device that outputs a signal.

Images