JP7706833B2 - Apparatus and method for suppressing spurious signals - Google Patents

Description

This invention relates to a device and method for suppressing false image signals, and more specifically to a technique for suppressing false image components caused by multiple reflections when detecting moving objects as targets using radar.

By illuminating a target with radio waves transmitted from a radar, information on distance, direction, speed, and intensity can be obtained. During this measurement, the radio waves not only reflect back directly from the target, but also return after multiple reflections off the ground, walls, etc., resulting in false images (also called "virtual images") caused by the multiple reflections (in other words, caused by multipath). False images caused by multiple reflections have signal position distribution (i.e., point cloud set density), intensity, and speed components similar to those of a real image, so despite being noise, they return results as if the target of interest were present, complicating accurate sensing of the surroundings. For this reason, establishing technology to remove and suppress false images like those mentioned above is a major challenge.

Conventional techniques for distinguishing between real and false images include a method for determining whether the outbound and return paths of radio wave propagation match (see Patent Document 1), a method for distinguishing between clutter (i.e., noise components including false images) and targets through dynamic threshold processing using a CFAR (Constant False Alarm Rate) algorithm (see Patent Document 2), and a method for capturing images of locations corresponding to the positions of detected radar signal points with a camera and determining that radar signal points in positions where the presence of a target cannot be confirmed through image processing are false images.

JP 2020-186943 A JP 2004-271262 A

CFAR algorithms are roughly divided into two types of methods: CA-CFAR and OS-CFAR. The characteristics of each method are that CA-CFAR can be realized with a small amount of calculation, but the target may be included in the threshold calculation, so there is a possibility that the real image may be overlooked. In addition, OS-CFAR can obtain a stable threshold value because the probability that the target will be included in the threshold calculation is low, but the amount of calculation is large. Furthermore, when the signal characteristics of the real image and the false image are significantly different, the clutter components can be effectively suppressed by threshold processing using the CFAR algorithm, but there is a problem in that side lobes caused by high-intensity reflectors and false images caused by multipath can have intensity information and velocity information equivalent to that of the real image, making it difficult to identify and suppress these as false image components.

In addition, when using a radar device and a camera to distinguish between real and false images, there is the problem that the device configuration becomes large-scale.

The present invention aims to provide an artifact suppression device and method that can suppress artifact components with high accuracy using a simple device configuration.

In order to solve the above problem, a false signal suppression device according to the present invention includes a signal processing unit that performs frequency analysis on radar data output from a radar device that transmits radio waves into a detection target area and receives radio waves from the detection target area to perform a radar scan, to generate a set of combined data of position, signal strength, and relative speed; a target position detection unit that uses the set of combined data of position, signal strength, and relative speed to generate a cluster of radar signal points for a target and performs a tracking process for the cluster; a target calculation unit that generates combined data of the relative speed and signal strength information based on the signal strength for each radar scan for a cluster related to a moving object among the clusters and a collection of clusters for a predetermined time length associated with the cluster related to the moving object in the tracking process (referred to as a "moving object time series cluster"); and a feature calculation unit which arranges combination data of values of relative speed and values of the signal strength information in a time series, and while shifting the time series arrangement by one in the time axis direction, calculates a correlation value in comparison with the original time series arrangement to generate combination data of the scan execution number at the head of the arrangement and the correlation value, and performs a Fourier transform process on the combination data of the scan time based on the scan execution number at the head of the arrangement and the correlation value to generate a frequency spectrum; and a false image suppression unit which extracts combinations of moving object time series clusters related to the same target based on the correlation value between the frequency spectra of each of the moving object time series clusters, and identifies and removes, as false images, moving object time series clusters other than the moving object time series cluster closest to the radar device among the combinations of moving object time series clusters related to the same target .

A method for suppressing false signals according to the present invention includes a process of frequency analyzing radar data output from a radar device that transmits radio waves into a detection target area and receives radio waves from the detection target area to perform a radar scan, generating a set of combined data of position, signal strength, and relative velocity, a process of generating a cluster of radar signal points for a target using the set of combined data of position, signal strength, and relative velocity and tracking the cluster, a process of generating combined data of the relative velocity for each radar scan for a cluster related to a moving object among the clusters and a collection of clusters for a predetermined time length associated with the cluster related to the moving object in the tracking process (referred to as a "moving object time series cluster") and signal strength information based on the signal strength, and a process of calculating, for each of the moving object time series clusters, a value of the relative velocity for each radar scan and the process includes a process of arranging combination data with the value of the signal strength information in a time series, and shifting the time series arrangement by one in the time axis direction to calculate a correlation value in comparison with the original time series arrangement to generate combination data of the scan execution number at the head of the arrangement and the correlation value; a process of performing a Fourier transform process on the combination data of the scan time based on the scan execution number at the head of the arrangement and the correlation value to generate a frequency spectrum; a process of extracting combinations of moving object time series clusters relating to the same target based on the correlation value between the frequency spectra of each of the moving object time series clusters; and a process of identifying and removing, as false images, moving object time series clusters other than the moving object time series cluster closest to the radar device among the moving object time series clusters included in the combinations of moving object time series clusters relating to the same target .

The false image suppression device and false image suppression method of the present invention make it possible to accurately extract false images corresponding to real images from radar information alone, without using sensors other than the radar, and to suppress false image components with high precision.

1 is a functional block diagram showing a schematic configuration of a false image signal suppression device according to an embodiment of the present invention; 2 is a flowchart showing a processing procedure in the artifact signal suppression device of FIG. 1 and a processing procedure of a artifact signal suppression method according to an embodiment of the present invention; FIG. 13 is a diagram showing an example of array VT data. 11 is a diagram for explaining how a correlation value is calculated in an autocorrelation calculation unit. FIG. 1 is a diagram for explaining the mechanism by which a false image occurs, (A) is a diagram for explaining the difference between the radio wave propagation path of a real image and that of a false image, and (B) is a diagram showing the distribution of radar signal points (PPI) for explaining the difference in appearance between radar signal points of a real image and radar signal points corresponding to a false image. FIG. 13 is a diagram for explaining an outline of measurements in a verification example. FIG. 13 is a diagram showing VT data (array VT data) for analysis pattern 1 in a verification example. FIG. 13 is a diagram showing VT data (array VT data) for analysis pattern 2 in a verification example. FIG. 13 is a diagram showing combination data of the value of the first scan execution number in the array and the correlation value for analysis pattern 1 in the verification example. FIG. 13 is a diagram showing combination data of the value of the first scan execution number in the array and the correlation value for analysis pattern 2 in the verification example. FIG. 13 is a diagram showing combined data of frequency values and spectral intensity values after Fourier transform for analysis pattern 1 in a verification example. FIG. 13 is a diagram showing combined data of frequency values and spectral intensity values after Fourier transform for analysis pattern 2 in the verification example. 1 is a diagram showing correlation values between frequency spectra of each moving object time series cluster in a verification example, where (A) is a result for analysis pattern 1, (B) is a result for analysis pattern 2, and (C) is a result for analysis pattern 3. 1A is a diagram for explaining the process of suppressing false image components for analysis pattern 1 in a verification example, and FIG. 1B is a diagram showing a distribution (PPI) of radar signal points, FIG. 1C is a diagram showing the suppression result of radar signal points corresponding to false images, and FIG. 1A is a diagram for explaining the process of suppressing false image components for analysis pattern 2 in a verification example, and FIG. 1B is a diagram showing a distribution (PPI) of radar signal points, and FIG. 1C is a diagram showing the suppression result of radar signal points corresponding to false images.

The present invention will be described below based on the illustrated embodiment.

Figure 1 is a functional block diagram showing the schematic configuration of a false image signal suppression device 1 according to an embodiment of the present invention. Figure 2 is a flowchart showing the processing procedure in the false image signal suppression device 1 and the processing procedure of a false image signal suppression method according to an embodiment of the present invention.

This false image suppression device 1 is a device that uses a radar to detect moving objects in a detection area as targets and suppresses false image components by extracting and removing false images corresponding to the real images of the targets that appear in the distribution of radar signal points, and is mainly equipped with a signal processing unit 2, a target position detection unit 3, a target calculation unit 4, a feature calculation unit 5, and a false image suppression unit 6.

The false signal suppression device 1 according to this embodiment includes a signal processing unit 2 that performs frequency analysis on radar data output from a radar device that transmits radio waves to a detection target area and receives radio waves from the detection target area to perform radar scanning, generates a set of combined data of position, signal strength, and relative speed, a target position detection unit 3 that uses the set of combined data of position, signal strength, and relative speed to generate a cluster of radar signal points for a target and performs a tracking process for the cluster, a target calculation unit 4 that generates combined data of signal strength information based on the relative speed and signal strength for each radar scan for a cluster related to a moving object among the clusters and a collection of clusters for a predetermined length of time associated with the cluster related to the moving object by tracking processing (referred to as a "moving object time series cluster"); and a target calculation unit 5 that calculates, for each moving object time series cluster, the value of the relative speed for each radar scan and the signal strength information. The system further includes a feature calculation unit 5 that arranges combination data of the scan execution number at the beginning of the array and the correlation value in a time series, and then shifts the time series array one by one in the time axis direction to calculate a correlation value in comparison with the original time series array to generate combination data of the scan execution number at the beginning of the array and the correlation value, and performs a Fourier transform process on the combination data of the scan time and the correlation value based on the scan execution number at the beginning of the array to generate a frequency spectrum; and a false image suppression unit 6 that extracts combinations of moving object time series clusters related to the same target based on the correlation value between the frequency spectra of each moving object time series cluster, and identifies and removes, as false images, moving object time series clusters included in the combinations of moving object time series clusters related to the same target, other than the moving object time series cluster closest to the radar device (see FIG. 1).

In addition, the method for suppressing false signals according to this embodiment includes a process of frequency analyzing radar data output from a radar device that transmits radio waves to a detection target area and receives radio waves from the detection target area to perform radar scanning, to generate a set of combined data of position, signal strength, and relative velocity (steps S1 to S3), a process of generating a cluster of radar signal points for a target object using the set of combined data of position, signal strength, and relative velocity and tracking the cluster (steps S4 to S5), a process of generating combined data of signal strength information based on the relative velocity for each radar scan and the signal strength for a cluster related to a moving object among the clusters and a collection of clusters for a predetermined length of time associated with the cluster related to the moving object in the tracking process (referred to as a "moving object time series cluster") (steps S6 to S8), and a process of generating combined data of the relative velocity value for each radar scan and the signal strength information for each moving object time series cluster. The method includes a process (step S9) of arranging combination data with the value of the scan execution number at the beginning of the array in a time series, shifting the time series array one by one in the time axis direction, calculating a correlation value in comparison with the original time series array, and generating combination data of the scan execution number at the beginning of the array and the correlation value, a process (step S10) of performing a Fourier transform process on the combination data of the scan time and the correlation value based on the scan execution number at the beginning of the array to generate a frequency spectrum, a process (step S11) of extracting combinations of mobile object time series clusters related to the same target based on the correlation value between the frequency spectra of each of the mobile object time series clusters, and a process (step S12) of identifying and removing, as false images, mobile object time series clusters other than the mobile object time series cluster closest to the radar device among the combinations of mobile object time series clusters related to the same target (see FIG. 2).

In this invention, for example, radar data acquired and output by a radar device performing radar scanning using the FMCW (Frequency Modulated-Continuous Wave) radar method may be used. In the FMCW method, a frequency-modulated continuous wave is transmitted and a reflected wave reflected off the surface of the target is received, and the position and relative speed of the target are detected based on the frequency difference between the transmitted signal and the received signal. In the FMCW method, radio waves (in other words, transmitted signals) whose frequency changes in a triangular or sawtooth waveform are transmitted.

A radar device that outputs radar data and can be used in this invention is, for example, a device that has a transmitter and a receiver and has the ability to transmit and receive radio waves, performs FMCW radar scans, and outputs radar data.

The transmitter of the radar device may include, for example, a mechanism including a voltage generator that generates and outputs a predetermined voltage, a voltage-controlled oscillator that generates and outputs radio waves (transmission signals) having a frequency corresponding to the predetermined voltage output from the voltage generator, and a transmission antenna that transmits (in other words, emits or radiates) the radio waves (transmission signals) output from the voltage-controlled oscillator as transmission waves to the detection target area.

The transmitter generates radio waves having a frequency of, for example, 79 GHz, 76 GHz, or 60 GHz (also called "millimeter waves") and transmits them via the transmitting antenna. In this invention, it is preferable to use radio waves in a high frequency band.

The receiving section of the radar device may include, for example, a receiving antenna that receives radio waves including radio waves transmitted from a transmitting antenna (i.e., transmitted waves) that are reflected off the surface of a target within the detection area and return (i.e., reflected waves; also called "Doppler reflected waves"); a mixer that mixes the radio waves (transmitted signal) supplied from the transmitting section with the radio waves (in other words, received signal) output from the receiving antenna to generate and output a differential signal; and an A/D converter (Analog to Digital converter) that performs sampling processing (in other words, analog-to-digital conversion processing) on the differential signal output from the mixer using a predetermined sampling frequency, converts the differential signal into digital data, and outputs the digital signal.

In the FMCW system, an array antenna is formed by one or more antennas, and radio waves (reflected waves) are received by the array antenna as a receiving antenna. A mixer and an A/D converter are provided for each receiving antenna.

The differential signal output as radar data from the radar device is a signal that has a frequency component that is the difference between the frequency component of the radio wave (transmitted signal) supplied from the transmitter and the frequency component of the radio wave (received signal) output from the receiving antenna (i.e., a signal that has a beat frequency, also known as a "beat signal").

As just one example, a radar device may be attached to a vehicle such as an automobile so that a specified range around the vehicle is the detection target area, or may be attached to a structure such as a building so that a specified range around the structure or within the building site is the detection target area.

The false signal suppression device 1 may be provided with a memory unit that records and stores radar data output from the radar device, and may read and use the radar data from the memory unit, or may read and use radar data that is output from the radar device and stored in an external storage device such as a server.

The processing by the signal processing unit 2, target position detection unit 3, and target calculation unit 4 described below is explained on the assumption that it is performed for each radar scan using the radar data output each time the radar device performs a radar scan.

The signal processing unit 2 calculates the position and relative speed of a target within the detection area using a differential signal as radar data (i.e., radar raw data) output from the radar device. The signal processing unit 2 has a frequency analysis unit 21, a position calculation unit 22, and a speed calculation unit 23.

The frequency analysis unit 21 performs frequency analysis of the differential signal (in other words, a beat signal; this is a digital signal) as radar data (in other words, waveform data) output from the radar device.

Specifically, the frequency analysis unit 21 performs a Fourier transform on the waveform data for the sampling time width (e.g., 100 milliseconds) per radar scan, that is, on the amplitude (i.e., reception level) of the differential signal (beat signal) for the sampling time width per radar scan, to generate a frequency spectrum that indicates the frequency distribution of the amplitude of the differential signal, and generates spectral intensity and phase information. The frequency spectrum generated by the frequency analysis unit 21 represents the amplitude (reception level) of each frequency component contained in the differential signal.

When performing Fourier transform processing, the range of frequencies that may be generated by the movement or movement of the target to be detected is set, taking into consideration the expected moving speed and movement speed of the target to be detected, and the frequency interval is set as the Doppler frequency. In other words, the frequency analysis unit 21 performs frequency analysis on the radar data (waveform data) output from the radar device to determine the Doppler component, that is, the spectral intensity and phase for each Doppler frequency of the differential signal.

The position calculation unit 22 generates distance information regarding the distance between the radar device (specifically, the radar device antenna) and the target, and angle information regarding the azimuth angle of the target relative to the radar device (specifically, the radar device antenna) based on the phase information in the frequency spectrum generated by the frequency analysis unit 21. The distance information and angle information are obtained for each Doppler frequency.

There are known methods for calculating distance and azimuth angle based on the difference signal (beat signal), and this invention is not limited to a specific method, so detailed explanation will be omitted here. For example, the distance between the radar device and the target can be calculated by a known method such as a method that utilizes the fact that the difference between the frequency of the transmitted wave and the frequency of the received wave increases or decreases in proportion to the distance between the radar device and the target. In addition, the azimuth angle of the target relative to the radar device can be calculated by a known angle measurement method such as the MUSIC (short for MULTIPLE SIgnal Classification) method or the ESPRIT (short for Estimation of Signal Parameters via Rotational Invariance Techniques) method.

The velocity calculation unit 23 calculates the relative velocity (specifically, the instantaneous relative velocity when the transmitted wave is reflected by the surface of the target) which is the velocity of the target relative to the radar device (specifically, the antenna of the radar device) using each of the Doppler frequencies used in the frequency analysis in the frequency analysis unit 21. In other words, the relative velocity is calculated for each Doppler frequency.

There are known methods for calculating relative velocity, and this invention is not limited to a specific method, so detailed explanation will be omitted here. For example, the relative velocity can be calculated by known methods such as a method that utilizes the fact that the frequency of radio waves (i.e., received waves) reflected by the surface of a target moving relative to the radar and received by the receiving antenna is affected by the speed of the target when reflected by the surface of the target, and shifts according to the relative speed between the target and the radar device due to the Doppler effect, compared to the frequency of radio waves (i.e., transmitted waves) transmitted from the transmitting antenna.

The signal processing unit 2 performs signal processing/frequency analysis on the radar data (i.e., radar raw data) output from the radar device by the above processing, and generates a set of four-dimensional data (X, Y, I, V) that represents the signal strength distribution in a Cartesian coordinate system by speed component. Note that X and Y are the x and y coordinates (unit: m) in a two-dimensional Cartesian coordinate system in which the x and y axes are mutually orthogonal. I is the signal strength (in other words, the reception level) (unit: dB), and V is the relative speed (unit: km/h). The relative speed V [km/h] of the four-dimensional data (X, Y, I, V) generated by the signal processing unit 2 may be divided into a predetermined pitch, such as 0.1 [km/h], as necessary, and the actual value of the relative speed V may be assigned to the division to obtain the rank of the relative speed.

The target position detection unit 3 calculates information about targets in the detection area using a set of four-dimensional data (X, Y, I, V) generated by the signal processing unit 2. The target position detection unit 3 has a signal point extraction unit 31, a cluster generation unit 32, and a tracking processing unit 33.

The signal point extraction unit 31 extracts data that satisfies predetermined conditions from the set of four-dimensional data (X, Y, I, V) generated by the signal processing unit 2 as radar signal points for targets within the detection area.

The processing by the signal point extraction unit 31 is not limited to a specific method as long as it is a method capable of extracting radar signal points for targets within the detection area. For example, a threshold value (called the "intensity threshold value") for the signal strength I used to detect radar signal points for targets is set in advance, and data in which the signal strength I is equal to or greater than the intensity threshold value is extracted as a radar signal point for the target.

The intensity threshold is not limited to a specific value, but is set appropriately, for example, taking into consideration the ability to accurately detect radio waves that are reflected by the surface of a target and return (i.e., reflected waves). The intensity threshold may be set, for example, using the CFAR (short for Constant False Alarm Rate) algorithm, which is one of the threshold algorithms.

The intensity threshold may be a threshold value related to signal strength I for each relative speed V, or a threshold value related to integrated signal strength ΣI calculated by integrating signal strength I for each relative speed V for each combination of x coordinate X and y coordinate Y of the four-dimensional data (X, Y, I, V) (i.e., for each coordinate).

When a threshold value for signal strength I for each relative speed V is set as the intensity threshold value, it is conceivable that for each combination of x coordinate X and y coordinate Y (i.e., for each coordinate), the four-dimensional data (X, Y, I, V) of the combination (i.e., coordinates) where the signal strength I for any of the combinations for each relative speed V is equal to or greater than the intensity threshold value is extracted as a radar signal point for the target.

When a threshold value for the integrated signal strength ΣI is set as the intensity threshold value, the signal strength I for each combination of x coordinate X and y coordinate Y (i.e., for each coordinate) of four-dimensional data (X, Y, I, V) for each relative speed V is integrated to calculate the integrated signal strength ΣI, and the four-dimensional data (X, Y, I, V) for the combination (i.e., coordinates) for which the integrated signal strength ΣI is equal to or greater than the intensity threshold value is extracted as a radar signal point for the target.

The signal point extraction unit 31 generates a set of four-dimensional data (X, Y, I, V) as radar signal points for targets within the detection area.

The cluster generation unit 32 performs clustering processing using a set of four-dimensional data (X, Y, I, V) as radar signal points for the target generated by the signal point extraction unit 31, and generates a cluster of radar signal points in a coordinate system whose coordinate axes are spatial positions (here, a two-dimensional orthogonal coordinate system in which the x-axis and y-axis are mutually orthogonal).

The method of generating clusters is not limited to a specific method or system, and any method or system may be used as long as it is possible to cluster a collection of radar signal points for a target based on the density of the radar signal points extracted by the signal point extraction unit 31, in other words, as long as it is possible to extract a collection of radar signal points that are close to each other within a certain distance. As a method of generating clusters, for example, DBSCAN (short for Density-Based Spatial Clustering of Applications with Noise), which is one of the clustering algorithms, may be used.

The tracking processing unit 33 performs a process of tracking targets detected in the past for each cluster (i.e., for each target) generated by the cluster generation unit 32. In other words, the tracking processing unit 33 performs a tracking process for clusters detected by past radar scans by chronologically associating clusters related to the same target generated by the cluster generation unit 32 at successive points in time.

There are known methods for tracking clusters/targets, and this invention is not limited to a specific method, so a detailed explanation will be omitted here. For example, tracking of clusters/targets can be performed by known methods such as a method using an α-β filter, a Kalman filter, or a particle filter as a tracking filter.

The tracking processing unit 33 may perform tracking processing after determining a representative point of each cluster generated by the cluster generation unit 32 as necessary. For example, the tracking processing unit 33 may perform tracking processing after determining a representative point of a signal point group recognized as one cluster by calculating a weighted average position using the signal strength (reception level) I [dB] at the position of each signal point included in the signal point group as a weight.

For each cluster generated by the cluster generation unit 32 in the current radar scan, the tracking processing unit 33 generates, as a result of the tracking process, a time-series tracking association with the cluster generated by the cluster generation unit 32 in the previous radar scan.

The processing by the signal processing unit 2 and the target position detection unit 3 may be performed in the same manner as in the past. In other words, the processing by the signal processing unit 2 and the target position detection unit 3 is not limited to the above, and various methods may be used, such as calculating the position (X, Y), signal strength I, and relative velocity V of the radar signal point based on the radar data, generating clusters, and performing tracking.

Here, by performing the same processing as in the past, in the processing by the cluster generation unit 32, a cluster corresponding to a real image of a certain target and a cluster corresponding to a false image of the target are generated as separate clusters (it is assumed). Also, by performing the same processing as in the past, in the processing by the tracking processing unit 33, the clusters corresponding to the real image of a certain target are associated with each other by tracking in a time series, and the clusters corresponding to the false images of the target are associated with each other by tracking in a time series.

The target calculation unit 4 calculates information about each cluster generated by the cluster generation unit 32. The target calculation unit 4 has a cluster extraction unit 41, a signal strength identification unit 42, and a VT data generation unit 43.

The cluster extraction unit 41 first determines, for each cluster generated by the cluster generation unit 32, whether the cluster is related to a moving object. There are known methods for determining whether a generated cluster is related to a moving object (e.g., a pedestrian, a vehicle) or a fixed object (e.g., a building, an immovable structure), and the present invention is not limited to a specific method, so a detailed description is omitted here.

If there is no cluster determined to be related to a moving object, the false image suppression device 1 ends the processing for the radar scan. Also, if there is only one cluster determined to be related to a moving object, the false image suppression device 1 determines that the cluster is a cluster corresponding to the real image of the target, and ends the processing for the radar scan.

If there are multiple clusters that are determined to be related to a moving object, the cluster extraction unit 41 extracts, for each cluster, time series data consisting of a set of four-dimensional data (X, Y, I, V) from the latest (in other words, the current) radar scan for the cluster for a predetermined time length (in other words, a predetermined number of radar scans) based on the results of processing by the tracking processing unit 33, and a set of four-dimensional data (X, Y, I, V) from past (in other words, previous to the previous) radar scans that are associated with the cluster by time series tracking.

The time series data extracted for one processing session is not limited to a specific time length (in other words, the number of radar scans), but is set to an appropriate time length (number of radar scans) taking into consideration that the characteristics of the target to be detected, particularly the moving speed, operating speed, and signal strength, can be adequately grasped. As just one example, the time length (number of radar scans) of the time series data extracted for one processing session could be set to 1000 milliseconds (i.e., 10 scans) when the sampling time width per radar scan is 100 milliseconds.

Here, of the clusters in the latest (in other words, the current) radar scan, the cluster that is determined to be related to a moving object is called a "moving object cluster." Also, a collection of clusters including the moving object cluster, which are associated with the moving object cluster by the tracking processing unit 33 through time series tracking, and which correspond to the time length of the time series data extracted as the object of one processing (the number of radar scans) is called a "moving object time series cluster."

For each mobile object time series cluster, the signal strength identification unit 42 generates strength information by speed component for each of the latest (current) and past (previous) clusters that make up the mobile object time series cluster, using a set of four-dimensional data (X, Y, I, V) for the cluster.

The signal strength identification unit 42 may generate the strength information using a set of four-dimensional data (X, Y, I, V) for each cluster that constitutes the mobile object time series cluster, or may generate the strength information using a set of four-dimensional data (X, Y, I, V) within a predetermined range that includes the cluster. The predetermined range that includes the cluster is not limited to a specific size or shape, but is appropriately set to an appropriate size and shape, taking into consideration that the characteristics of the target to be detected, particularly the operating speed and signal strength, can be appropriately grasped. As one example, the predetermined range that includes the cluster may be set to a range surrounded by a rectangle of ±1 [m] that includes the cluster.

Specifically, for a set of four-dimensional data (X, Y, I, V) for a cluster (including a set of four-dimensional data (X, Y, I, V) in a predetermined range including the cluster), the signal strength identification unit 42 identifies, for example, the maximum value Imax [dB] of the signal strength (reception level) I [dB] for each relative speed V [km/h] (which may be a rank of the relative speed divided at a predetermined pitch) and uses this as strength information, or calculates the average value Iavg [dB] of the signal strength I [dB] and uses this as strength information. The signal strength [dB] with the predetermined characteristics identified or calculated by the signal strength identification unit 42 is called "signal strength information".

The signal strength identification unit 42 generates combination data of the value of the relative velocity V and the value of the signal strength information (e.g., maximum value Imax, average value Iavg [dB]) for each of the latest (current) and past (previous) clusters that make up each mobile object time series cluster, in other words, for each radar scan related to the mobile object time series cluster (including cases related to a predetermined range that includes the cluster).

The VT data generation unit 43 adds information on the elapsed time T based on the time at which the radar scan was performed to the combined data of the relative velocity V value for each radar scan and the signal strength information value for the moving object time series cluster generated by the signal strength identification unit 42.

In this embodiment, the information on the elapsed time T is the radar scan execution number S [th], which increases by one each time a radar scan is performed. Since the sampling time width [seconds] per radar scan (i.e., the time interval [seconds] between the start of each radar scan) is set in advance, the radar scan execution number S [th] can be converted to the elapsed time [seconds] of the radar scan.

The processing by the signal processing unit 2, the target position detection unit 3, and the target calculation unit 4 is performed for each radar scan. That is, each time a radar scan is performed, a set of combined data (called "VT data") is generated for each mobile object time series cluster, consisting of the relative speed V [km/h], signal strength information, and elapsed time T information for each radar scan, for a predetermined time length corresponding to one processing target (in other words, for a predetermined number of radar scans).

The feature calculation unit 5 calculates feature amounts for identifying false images corresponding to real images of targets appearing in the distribution of radar signal points. The feature calculation unit 5 has an autocorrelation calculation unit 51 and a Fourier transform unit 52. The processing by the feature calculation unit 5 described below is explained here on the premise that it is performed for each mobile object time series cluster (in other words, for each mobile object time series cluster).

The autocorrelation calculation unit 51 performs autocorrelation analysis on the VT data. Specifically, the autocorrelation calculation unit 51 first arranges the combined data of the relative speed V [km/h] value and the signal strength information value for the VT data in time series (in this embodiment, in the order of the radar scan execution number S [time]). A collection of VT data arranged in time series is called "arranged VT data."

An example of array VT data is shown in Figure 3. In Figure 3, the horizontal axis represents the radar scan execution number S [time] and the vertical axis represents the relative speed V [km/h], and the maximum value Imax [dB] of the signal strength (reception level) I [dB] corresponding to the combination of the radar scan execution number S [time] and the relative speed V [km/h] is displayed as light and dark. In Figure 3, weak signal strength is displayed as dark, and stronger signal strength is displayed as bright. Note that Figure 3 shows a case where the sampling time width per radar scan is 100 milliseconds and the length of the time series data extracted for one processing target is 10 scans, and one second's worth of data is organized.

The array VT data can be said to be distribution information of instantaneous relative velocities with respect to a target in a time series. The relative velocity information obtained using a radar device not only includes the velocity of the torso of a moving object such as a human, but also simultaneously obtains the velocity components of various parts such as the head, upper arms, forearms and hands, thighs, and lower legs and feet. Therefore, since the velocity information of each part of a moving person, such as the head, torso, upper arms, forearms and hands, thighs, and lower legs and feet, is inherent, by arranging the combined data of the relative velocity V value for each radar scan and the signal strength information value in a time series, the periodic movement/operation of each part appears superimposed.

The autocorrelation calculation unit 51 then calculates the inner product of the time series array of the combined data of the relative velocity V value and the signal strength information value for each radar scan for the array VT data while shifting the time series array by one in the time axis direction (in other words, by one scan), and calculates the correlation value in comparison with the set of combined data in the original time series array state before shifting in the time axis direction (see Figure 4). When the time series array of the combined data is shifted by one in the time axis direction, the combined data arranged at the beginning of the time series array is moved to the end of the time series array (see Figure 4). In the example shown in Figure 4, the length of the time series data extracted as the target for one processing is 10 scans, and therefore the number of combined data arranged in time series is 10.

In the calculation of correlation values in the example shown in Figure 4, the state of the original time series array (called the "original state") is defined as "data with scan execution number 1 at the beginning of the array", the state after being shifted once along the time axis is defined as "data with scan execution number 2 at the beginning of the array", the state after being shifted twice along the time axis is defined as "data with scan execution number 3 at the beginning of the array", and so on. The state after being shifted nine times along the time axis is defined as "data with scan execution number 10 at the beginning of the array", and the state after being shifted ten times further along the time axis is defined as "data with scan execution number 11 at the beginning of the array".

In other words, when the autocorrelation calculation unit 51 calculates the correlation value, the shift operation in the time axis direction is repeated the same number of times as the number of combination data arranged in time series, and when the final shift operation is performed (in the example shown in Figure 4, the state after 10 shifts) the original state is restored and the data with the scan execution number 1 at the beginning of the array is matched.

The autocorrelation calculation unit 51 calculates the correlation value between the array VT data, which is in the original time series array state before being shifted along the time axis, and each of the data at the beginning of the array whose scan execution numbers are 1, 2, 3, ..., 10, and 11.

The autocorrelation calculation unit 51 generates combination data of the value of the scan execution number S at the beginning of the array and the correlation value.

The Fourier transform unit 52 performs Fourier transform processing on the combined data of the value of the scan execution number S at the beginning of the array generated by the autocorrelation calculation unit 51 and the correlation value. At this time, the Fourier transform unit 52 performs Fourier transform processing after converting the scan execution number S at the beginning of the array into time [seconds] according to "sampling time width per radar scan × (S-1)". The time obtained by converting the scan execution number S at the beginning of the array is called the "scan time".

Specifically, the Fourier transform unit 52 performs a Fourier transform process on the combination data (in other words, waveform data) of the scan time and correlation value for the time length of the time series data extracted as the target for one processing, that is, on the magnitude (in other words, amplitude) of the correlation value for each scan time for the time length of the time series data extracted as the target for one processing, to generate a frequency spectrum indicating the frequency distribution of the magnitude (in other words, amplitude) of the correlation value, and generates combination data of the frequency [Hz] value after the Fourier transform and the value of the spectral intensity (in other words, amplitude). The frequency spectrum generated by the Fourier transform unit 52 represents the magnitude (in other words, amplitude) of the correlation value of each frequency component in the combination data of the scan time and the correlation value.

By processing by the feature calculation unit 5, the frequency spectrum for each mobile object time series cluster is generated for multiple mobile object time series clusters, that is, combination data of the frequency [Hz] value after Fourier transform and the spectral intensity (in other words, amplitude) value.

The false image suppression unit 6 suppresses false image components by identifying whether each moving object time series cluster corresponds to a real image of the target or a false image based on the frequency spectrum for each moving object time series cluster generated by the feature calculation unit 5, i.e., the combined data of the frequency [Hz] value after Fourier transform and the spectral intensity value. The false image suppression unit 6 has a pairing unit 61 and a false image identification unit 62.

The pairing unit 61 calculates the correlation value between the frequency spectra of each of the mobile object time series clusters for all combinations of two mobile object time series clusters selected from the multiple mobile object time series clusters. Specifically, the pairing unit 61 calculates the correlation value between the combined data of the frequency [Hz] value after Fourier transform and the spectral intensity value (amplitude value) generated by the feature amount calculation unit 5.

The pairing unit 61 then extracts, as a combination of mobile object time series clusters relating to the same target, a combination of two mobile object time series clusters selected from the multiple mobile object time series clusters, in which the correlation value between the frequency spectra satisfies a predetermined condition. Specifically, a threshold value (called a "correlation threshold value") for the correlation value between the frequency spectra for extracting combinations of mobile object time series clusters relating to the same target is set in advance, and a combination of mobile object time series clusters in which the correlation value between the frequency spectra is greater than the correlation threshold value is extracted as a combination of mobile object time series clusters relating to the same target.

Here, the real image of a target and the false image corresponding to said real image are both images of the same target, and therefore are considered to have matching/similar signal characteristics. In particular, when time-series information is taken into account, the real image and false image are considered to have matching periodicities of movement/action and to have synchronized movements. For example, when multiple moving people are the detection targets, the walking movements of each person are considered to be different, and the walking movements are thought to be scattered and low in synchronism, so there will be no small difference in the periodicities and synchronicities of movements between real images of different people. In contrast, the real image and false image of the same person both represent the same person, so the periodicities of movement will match and the movements will be synchronized.

In other words, when a false image appears in addition to a real image, it is considered that there exists a combination of signal point groups (i.e., clusters) whose signal characteristics related to periodicity and synchrony are the same/similar. Therefore, combinations of moving object time series clusters in which the frequency spectra of each moving object time series cluster are highly correlated are extracted as combinations of moving object time series clusters related to the same target, and they are identified as combinations of moving object time series clusters (particularly, moving object clusters) corresponding to real images related to the same target and moving object time series clusters (particularly, moving object clusters) corresponding to false images.

The correlation threshold is not limited to a specific value, but is set to an appropriate value, taking into consideration, for example, whether the periodicity and synchronicity of the target's motion/operation can be appropriately determined and the combination of moving object time series clusters related to the same target can be accurately detected. The intensity threshold can be set to a value within the range of, for example, 0.90 to 0.99.

When the mobile body time series clusters constituting the combination of mobile body time series clusters relating to the same target are also extracted as a combination of mobile body time series clusters relating to the same target in relation to other mobile body time series clusters, the pairing unit 61 regards all of the mobile body time series clusters that are linked as a combination of mobile body time series clusters relating to the same target as a combination of mobile body time series clusters relating to the same target. In other words, the combination of mobile body time series clusters relating to the same target extracted by the pairing unit 61 may consist of a combination of two mobile body time series clusters or a combination of three or more mobile body time series clusters.

Note that a mobile object time series cluster that is not extracted as a combination of mobile object time series clusters with a high correlation between the frequency spectra of each mobile object time series cluster, i.e., a mobile object time series cluster whose frequency spectrum correlation value with any other mobile object time series cluster is equal to or less than the correlation threshold value, is determined to be a mobile object time series cluster (particularly a mobile object cluster) that corresponds to the real image of the target.

For each combination of moving object time series clusters related to the same target extracted by the pairing unit 61, the false image identification unit 62 identifies, as a false image, the moving object time series clusters included in the combination other than the moving object time series cluster closest to the radar device (specifically, the antenna of the radar device).

Here, when considering the mechanism by which false images occur, radar signal points corresponding to false images always appear farther away than radar signal points of real images (see Figure 5). That is, a real image of a target is an image (in other words, a radar signal point, cluster) of radio waves that propagate along the shortest path (i.e., the shortest distance) in which the transmitted wave travels directly from the radar device antenna to the target and the reflected wave travels directly from the target to the radar device antenna, whereas a false image of a target is an image (in other words, a radar signal point, cluster) of radio waves that propagate along a detour path in which the transmitted wave travels from the radar device antenna to the target via reflections on the ground or a wall, for example, and the reflected wave travels from the target to the radar device antenna via reflections on the ground or a wall, for example (see Figure 5(A)).

As a result, the apparent distance of the false image is longer than the actual distance of the real image, and the radar signal points corresponding to the false image appear farther away than the radar signal points of the real image (see Figure 5(B)). Figure 5(B) is a PPI (short for Plan Position Indicator; shown in a plan view) that shows an example of the distribution of radar signal points output from a radar device in a case such as that shown in Figure 5(A).

For this reason, among the moving object time series clusters included in the combination of moving object time series clusters related to the same target, the moving object time series cluster closest to the radar device (specifically, the moving object cluster) can be determined to be a real image, and the moving object time series clusters (specifically, the moving object clusters) other than the moving object time series cluster closest to the radar device can be determined to be false images.

The false image identification unit 62 removes the moving object clusters of the moving object time series clusters that are identified as false images from the moving object clusters determined/identified by the cluster extraction unit 41. This removes the radar signal points that correspond to false images, and suppresses the false image components.

Next, the operation and function of the artifact suppression device 1 and artifact suppression method configured as described above will be described with reference to FIG. 2 and FIGS. 6 to 15 along with a verification example of the effect of the artifact suppression device and artifact suppression method according to the present invention.

In the verification example, the purpose of verifying the operation and processing of the artifact suppression device and the artifact suppression method under conditions in which multiple moving people were measured simultaneously was to specifically perform measurements according to the following conditions (see also Figure 6).
・Radar height: 3m
・Radar depression angle: 20°
・Frequency used: 79GHz
Acquisition period: 100 milliseconds Processing target: 10 scans (number of radar scans of the time series data to be processed)
Moving objects: 2 subjects Movement patterns: walking (slow), walking (normal), running (fast)

In the verification example, the following three analysis patterns were set and measurements were performed for each.
Analysis pattern 1: Subject A: Running (fast), Subject B: Walking (slow)
<Analysis pattern 2> Subject C: walking (normal), Subject D: walking (slow)
Analysis pattern 3: Subject E: walking (normal), Subject F: running (fast)

Regarding the above analysis patterns, analysis pattern 1 is the analysis pattern in which the speed difference between the subjects' movements is maximized, and analysis pattern 2 is the analysis pattern in which the speed difference between the subjects' movements is minimized.

The operation of the artifact suppression device 1 is as follows: first, the frequency analysis unit 21 of the signal processing unit 2 performs frequency analysis using a Fourier transform on the differential signal (beat signal) serving as radar data output from the radar device to generate a frequency spectrum, and generates spectral intensity and phase information for each Doppler frequency of the differential signal (step S1).

Next, the position calculation unit 22 calculates the distance between the radar device and the target for each Doppler frequency based on the phase information, and calculates the azimuth angle of the target relative to the radar device, thereby generating distance information and angle information (step S2).

The velocity calculation unit 23 also calculates the relative velocity of the target for each Doppler frequency using each of the Doppler frequencies used in the frequency analysis in the frequency analysis unit 21 (step S3).

The processing of steps S1 to S3 by the signal processing unit 2 generates a set of four-dimensional data (X, Y, I, V) for each radar scan.

Next, the signal point extraction unit 31 of the target position detection unit 3 generates a set of four-dimensional data (X, Y, I, V) as radar signal points for the target using the set of four-dimensional data (X, Y, I, V) for each radar scan and the intensity threshold value for the signal intensity I (step S4).

Then, the cluster generation unit 32 performs a clustering process using a set of four-dimensional data (X, Y, I, V) representing radar signal points for the target, and generates a cluster of radar signal points (step S5).

The tracking processing unit 33 also performs cluster tracking processing by associating clusters related to the same target that are generated at successive points in time in time.

Next, the cluster extraction unit 41 of the target calculation unit 4 generates a moving object time series cluster, which is a collection of clusters including the moving object cluster in the latest (current) radar scan, that are associated with the moving object cluster by time series tracking, and whose time length (number of radar scans) of the time series data extracted as the target for one processing session is generated (step S6).

Next, for each mobile object time series cluster, the signal strength identification unit 42 uses a set of four-dimensional data (X, Y, I, V) for the clusters that make up the mobile object time series cluster to generate combination data of the value of the relative velocity V and the value of the signal strength information (e.g., maximum value Imax, average value Iavg [dB]) for each radar scan for the mobile object time series cluster (step S7).

The VT data generation unit 43 also generates VT data by adding information on the elapsed time T to the combined data of the relative velocity V value for each radar scan and the signal strength information value for the moving object time series cluster (step S8).

VT data (particularly, array VT data) for analysis pattern 1 in the verification example is shown in FIG. 7, and VT data (particularly, array VT data) for analysis pattern 2 is shown in FIG. 8. In the verification example, VT data with relative velocity V values in the range of -20 to 20 [km/h] was used. Also, whether the VT data relates to a real image or a false image is information that is determined as a result of subsequent processing, but for ease of understanding, the results of identifying whether the image is real or false are also shown.

From the results shown in Figures 7 and 8, it can be seen that in analysis pattern 1 (Figure 7), where the difference in the speed of the subjects' movements is the largest, a difference is observed in the relative speed V between subjects A and B, whereas in analysis pattern 2 (Figure 8), where the difference in the speed of the subjects' movements is the smallest, no large difference is observed in the relative speed V between subjects C and D. It can also be seen that in both analysis pattern 1 and analysis pattern 2, the distribution patterns of instantaneous relative speeds in the time series are similar between real and false images for the same subject.

Next, the autocorrelation calculation unit 51 of the feature calculation unit 5 shifts the time series array of the array VT data one by one along the time axis (in other words, one scan at a time), and calculates the correlation value in comparison with the original time series array state, to generate combination data of the value of the scan execution number S at the beginning of the array and the correlation value (step S9).

Figure 9 shows the combination data of the value of the scan execution number S at the beginning of the array for analysis pattern 1 in the verification example and the correlation value, and Figure 10 shows the combination data of the value of the scan execution number S at the beginning of the array for analysis pattern 2 and the correlation value.

From the results shown in Figures 9 and 10, it can be seen that in both analysis pattern 1 and analysis pattern 2, the waveforms that appear depending on the combination of the value of the scan execution number S at the beginning of the array and the correlation value are similar for real and false images of the same subject.

Next, the Fourier transform unit 52 performs a Fourier transform process on the magnitude (in other words, the amplitude) of the correlation value for each scan time based on the scan execution number S at the beginning of the array to generate a frequency spectrum, i.e., combination data of the frequency [Hz] value after the Fourier transform and the spectral intensity (in other words, the amplitude) value (step S10).

Figure 11 shows combined data of frequency values and spectral intensity values after Fourier transform for analysis pattern 1 in the verification example, and Figure 12 shows combined data of frequency values and spectral intensity values after Fourier transform for analysis pattern 2.

The results shown in Figure 11 show that for subject A in analysis pattern 1 (i.e., running (fast)), a peak is observed around 2.7 Hz in both the real and false images, while for subject B (i.e., walking (slow)), both the real and false images decrease monotonically.

From the results shown in Figure 12, it can be seen that for subject C (i.e., walking (normal)) in analysis pattern 2, both the real image and the false image have peaks around 1.7 Hz, while for subject D (i.e., walking (slow)), both the real image and the false image decrease monotonically.

The results shown in Figures 11 and 12 confirm that the frequency spectrum obtained by Fourier transforming the autocorrelation correlation value has different characteristics/waveforms when the target is different, and that the characteristics/waveforms are similar between real and false images of the same subject.

Next, the pairing unit 61 of the false image suppression unit 6 calculates the correlation value between the frequency spectra of each of the moving object time series clusters for all combinations of two moving object time series clusters selected from the multiple moving object time series clusters, and extracts combinations of moving object time series clusters related to the same target using the correlation value and a correlation threshold value related to the correlation value between the frequency spectra (step S11).

For all combinations of two mobile object time series clusters selected from multiple mobile object time series clusters, the correlation values between the frequency spectra of each combination of two mobile object time series clusters are shown in Figure 13 (A) for analysis pattern 1 in the verification example, in Figure 13 (B) for analysis pattern 2, and in Figure 13 (C) for analysis pattern 3.

The results shown in Figure 13 confirm that by setting the correlation threshold to, for example, 0.95, it is possible to accurately extract combinations of real and false images for each subject (in other words, relating to the same target).

Next, for each combination of moving object time series clusters related to the same target, the false image identification unit 62 identifies as false images and removes the moving object time series clusters included in the combination other than the moving object time series cluster closest to the radar device (step S12).

The process of suppressing false image components for analysis pattern 1 in the verification example is shown in FIG. 14, and the process of suppressing false image components for analysis pattern 2 is shown in FIG. 15. For both analysis patterns, combinations of moving object time series clusters related to the same target are extracted (see each figure (B)) based on the correlation values between the frequency spectra of the autocorrelation for moving object time series clusters generated from the distribution of radar signal points based on the radar data output from the radar device (see each figure (A); specifically, PPI), and it is confirmed that radar signal points corresponding to false images are appropriately suppressed by removing, as false images, moving object time series clusters other than the moving object time series cluster closest to the radar device among the moving object time series clusters included in each combination (see each figure (C)).

According to the embodiment of the false image suppression device 1 and the false image suppression method, a moving object time series cluster is generated to utilize the coincidence of periodicity of the movements/operations of the real image and the false image and the synchronization of the movements. This makes it possible to accurately extract the false image corresponding to the real image from only the radar information without using sensors other than the radar, and to suppress the false image component with high accuracy.

The above describes an embodiment of the present invention, but the specific configuration is not limited to the above embodiment, and any design changes that do not deviate from the gist of the present invention are also included in the present invention.

For example, in the above embodiment, a moving object time series cluster is generated and a false image corresponding to the real image is extracted by utilizing the coincidence of the periodicity of the motion/operation of the real image and the false image or the synchronization of the motion. However, if the detection target is, for example, a motorcycle or an automobile, a false image corresponding to the real image may be extracted from the motion period of the speed component of the rotational motion of the wheels and the traveling speed component of the body. In particular, if the detection target is an object with a significantly different speed component associated with the motion, there is no need to compare the periodicity in the time series data. By comparing the speed components in the data at a single point in time, a combination of clusters with matching/similar speed components may be extracted as a combination of a cluster corresponding to a real image and a cluster corresponding to a false image, and the false image signal/false image component may be suppressed. In this case, the cluster tracking process by the tracking processing unit 33 and the processes related to the moving object time series cluster by the target calculation unit 4 and the feature calculation unit 5 are not necessary. Specifically, the target calculation unit 4 generates combination data of relative speed and signal strength information for clusters related to moving objects (i.e., moving object clusters) among the clusters generated by the cluster generation unit 32, the pairing unit 61 of the false image suppression unit 6 extracts, as a combination of moving object clusters related to the same target, a combination of two moving object clusters selected from a plurality of moving object clusters, in which the correlation value between the combination data of the relative speed value and the signal strength information value of each moving object cluster is greater than a predetermined threshold value, and the false image identification unit 62 identifies and removes, as false images, the moving object clusters included in the combination of moving object clusters related to the same target, other than the moving object cluster closest to the radar device.

This invention can suppress false image components with high accuracy using a simple device configuration, and can therefore be used as a useful technology in fields such as automobile driving assistance, surrounding environment recognition systems, safety monitoring systems, and intruder detection.

REFERENCE SIGNS LIST 1 False image signal suppression device 2 Signal processing unit 21 Frequency analysis unit 22 Position calculation unit 23 Speed calculation unit 3 Target position detection unit 31 Signal point extraction unit 32 Cluster generation unit 33 Tracking processing unit 4 Target calculation unit 41 Cluster extraction unit 42 Signal strength identification unit 43 VT data generation unit 5 Feature amount calculation unit 51 Autocorrelation calculation unit 52 Fourier transform unit 6 False image suppression unit 61 Pairing unit 62 False image identification unit

Claims (2)

a signal processing unit that performs frequency analysis on radar data output from a radar device that transmits radio waves to a detection target area and receives radio waves from the detection target area to perform a radar scan, and generates a set of combined data of position, signal strength, and relative speed;
a target position detection unit that generates a cluster of radar signal points for a target using a set of combination data of the position, signal strength, and relative velocity, and performs a tracking process for the cluster;
a target calculation unit that generates combination data of the relative speed for each radar scan and signal strength information based on the signal strength for a cluster related to a moving object among the clusters and a collection of clusters for a predetermined time length associated with the cluster related to the moving object in the tracking process (referred to as a "moving object time series cluster");
a feature calculation unit that arranges, for each moving object time series cluster, combination data of the value of the relative speed for each radar scan and the value of the signal strength information in a time series, and then, while shifting the time series arrangement by one in the time axis direction, calculates a correlation value in comparison with the state of the original time series arrangement to generate combination data of the scan execution number at the top of the arrangement and the correlation value, and performs a Fourier transform process on the combination data of the scan time and the correlation value based on the scan execution number at the top of the array to generate a frequency spectrum;
a false image suppression unit that extracts a combination of moving object time series clusters related to the same target based on a correlation value between the frequency spectra of each of the moving object time series clusters, and identifies and removes, as false images, moving object time series clusters other than the moving object time series cluster closest to the radar device among the moving object time series clusters included in the combination of moving object time series clusters related to the same target.
13. A device for suppressing false images, comprising: A process of performing frequency analysis on radar data output from a radar device that transmits radio waves to a detection target area and receives radio waves from the detection target area to perform a radar scan, thereby generating a set of combined data of position, signal strength, and relative speed;
generating a cluster of radar signal points for a target using the set of combined data of position, signal strength, and relative velocity, and tracking the cluster;
A process of generating combination data of the relative speed for each radar scan and signal strength information based on the signal strength for a cluster related to a moving object among the clusters and a collection of clusters for a predetermined time length associated with the cluster related to the moving object in the tracking process (referred to as a "moving object time series cluster");
a process of arranging, for each moving object time series cluster, combination data of the value of the relative speed for each radar scan and the value of the signal strength information in a time series, shifting the time series arrangement by one in the time axis direction, calculating a correlation value in comparison with the state of the original time series arrangement, and generating combination data of the scan execution number at the top of the arrangement and the correlation value;
A process of performing a Fourier transform process on combination data of a scan time and a correlation value based on the first scan execution number of the array to generate a frequency spectrum;
extracting a combination of moving object time series clusters related to the same target based on a correlation value between the frequency spectra of each of the moving object time series clusters;
and a process of identifying and removing, as spurious images, moving object time series clusters other than the moving object time series cluster closest to the radar device among the moving object time series clusters included in the combination of moving object time series clusters related to the same target.
13. A method for suppressing false images, comprising: