JP6858067B2 - Radar device and control method of radar device - Google Patents

Description

The present invention relates to a radar device and a control method for the radar device.

Conventionally, a radar device installed on the front side of a vehicle body or the like outputs a transmission wave to a transmission range outside the vehicle, receives a reflected wave from the target, and outputs target data including position information of the target. It is derived and the stationary vehicle or the like located in front of the vehicle is determined from the target data. Then, the vehicle control device provided in the vehicle acquires information about the stationary vehicle or the like from the radar device, controls the behavior of the vehicle based on this information, avoids collision with the stationary vehicle or the like, and avoids collision with the stationary vehicle or the like. It provides the user with safe and comfortable driving (see, for example, Patent Document 1).

Japanese Unexamined Patent Publication No. 2016-006383

However, the above-mentioned conventional technique has a problem that the accuracy of discriminating between a stationary vehicle and an upper object is insufficient, and the upper object is erroneously detected as a stationary vehicle.

An example of an embodiment of the present application is intended to provide, for example, a radar device for accurately discriminating a stationary vehicle and an upper object, and a control method for the radar device.

An example of the embodiment of the present application is, for example, that the radar device receives a reflected wave that is reflected by a target that exists in the vicinity of the radar transmission wave transmitted to the vicinity of the own vehicle and obtains a target based on a received signal. A plurality of parameters related to the above and the detection distance of the target are derived. Then, each parameter, a target existing in the traveling direction of the own vehicle, the likelihood of the target colliding when the own vehicle advances in the traveling direction, and a target existing in the traveling direction of the own vehicle. However, a collision corresponding to each parameter and detection distance derived from a likelihood model in which a known correlation with the likelihood of a target that does not collide when the own vehicle advances in the traveling direction is modeled for each detection distance. A first index is calculated based on each likelihood ratio of the target to be used and the target to be non-collision. Then, based on the calculated first index, it is determined whether the target is a colliding target or a non-colliding target.

According to an example of the embodiment of the present application, for example, a stationary vehicle and an upper object can be accurately discriminated.

FIG. 1 is a schematic diagram showing an outline of target detection by the radar device according to the first embodiment. FIG. 2 is a diagram showing a configuration of a radar device according to the first embodiment. FIG. 3 is a diagram showing the relationship between the transmitted wave and the reflected wave and the beat signal. FIG. 4A is a diagram illustrating peak extraction in the up section. FIG. 4B is a diagram illustrating peak extraction in the down section. FIG. 5 is a diagram conceptually showing the angle estimated by the directional calculation process as an angle spectrum. FIG. 6A is a diagram illustrating pairing based on the azimuth angle and the angular power of each of the up section and the down section. FIG. 6B is a diagram for explaining the pairing result. FIG. 7A is a diagram illustrating a method of calculating the lateral position difference according to the first embodiment. FIG. 7B is a diagram illustrating a background for calculating the lateral position difference according to the first embodiment. FIG. 8 is a diagram showing a lateral position difference model according to the first embodiment. FIG. 9A is a diagram illustrating a method of calculating the relative speed difference according to the first embodiment. FIG. 9B is a diagram illustrating a background for calculating the relative speed difference according to the first embodiment. FIG. 10 is a diagram showing a relative speed difference model according to the first embodiment. FIG. 11 is a diagram illustrating the calculation of the upper beam inclination and the lower beam inclination, which are parameters for calculating the independent beam inclination difference according to the first embodiment. FIG. 12 is a diagram showing an independent beam inclination difference model according to the first embodiment. FIG. 13 is a diagram illustrating a method of calculating the vertical beam power difference inclination according to the first embodiment. FIG. 14 is a diagram showing a vertical beam power difference inclination model according to the first embodiment. FIG. 15A is a diagram showing the relationship between the angular power and the distance of a stationary vehicle. FIG. 15B is a diagram showing the relationship between the angular power and the distance of the upper object. FIG. 16 is a diagram illustrating a method of calculating the vertical beam surface integral according to the first embodiment. FIG. 17 is a diagram showing a vertical beam surface integral differential model according to the first embodiment. FIG. 18 is a diagram showing an extrapolation factor type likelihood table according to the first embodiment. FIG. 19A is a flowchart showing the target information output process according to the first embodiment. FIG. 19B is a flowchart showing a subroutine for removing unnecessary objects according to the first embodiment. FIG. 20 is a diagram showing overlapping distributions of probability ratios of stationary vehicles and upper objects. FIG. 21 is a diagram showing a configuration of a radar device according to the second embodiment. FIG. 22A is a diagram showing a table A for storing weighting coefficients and intercepts for regressing each probability ratio according to the initial detection distance according to the second embodiment. FIG. 22B is a diagram showing a table B for storing weighting coefficients and intercepts for regressing each probability ratio according to the initial detection distance according to the second embodiment. FIG. 23A is a diagram illustrating a method of calculating the attenuation amount area according to the third embodiment. FIG. 23B is a diagram illustrating a method of calculating the attenuation amount area according to the third embodiment. FIG. 24A is a diagram illustrating a method of calculating the increased amount area according to the third embodiment. FIG. 24B is a diagram illustrating a method of calculating the increased amount area according to the third embodiment. FIG. 25 is a flowchart showing a subroutine for removing unnecessary objects according to the third embodiment.

The radar device and the control method of the radar device according to the embodiment of the present application will be described below with reference to the accompanying drawings. The embodiments shown below are merely examples, and do not limit the present application. In addition, the embodiments shown below mainly show the configurations and processes related to the disclosed technology, and the description of other configurations and processes will be omitted. Then, each embodiment and modification may be appropriately combined within a consistent range. Further, in each embodiment, the same reference numerals are given to the same configurations and processes, and the description of the existing configurations and processes will be omitted.

[Embodiment 1]
(Outline of target detection by the radar device according to the first embodiment)
In the first embodiment, a stationary vehicle and an upper object to be detected by a radar device are discriminated from a relatively long distance. That is, when the stationary vehicle and the upper object are discriminated from a single target data, if the distributions of the stationary vehicle and the upper object are overlapped, it becomes difficult to discriminate between the two based on the threshold value.

Therefore, in the first embodiment, a statistical model is constructed from a known correlation between the values that a stationary vehicle and an upper object can take for a certain parameter and the probability based on the frequency of occurrence, and the likelihood of the stationary vehicle and the likelihood of the stationary vehicle and the likelihood of each scan by the radar device The upper object likelihood is calculated based on the statistical model. Then, using the Bayesian filter method, the likelihood is updated for each scan to improve the reliability, and the stationary vehicle and the upper object are discriminated based on the ratio (logit) of the stationary vehicle likelihood and the upper object likelihood. Do. Here, since the discrimination is not sufficient with only one parameter, in the first embodiment, the performance can be complemented with each other by using a plurality of parameters for which the corresponding statistical models are defined.

FIG. 1 is a schematic diagram showing an outline of target detection by the radar device according to the first embodiment. The radar device 1 according to the first embodiment is mounted on a front portion such as in the front grill of the own vehicle A, and detects the target T (targets T1 and T2) existing in the traveling direction of the own vehicle A. The target T includes a moving target and a stationary target. The target T1 shown in FIG. 1 is, for example, a preceding vehicle moving along the traveling direction of the own vehicle A or a stationary stationary object (including a stationary vehicle). Further, the target T2 shown in FIG. 1 is, for example, an upper object other than a vehicle that is stationary above the traveling direction of the own vehicle A, for example, a traffic light, an overpass, a road sign, a guide sign, or the like.

In order to guarantee the performance of the radar device 1 even when the vertical axis on which the radar is mounted is tilted due to the load or suspension in the own vehicle A, as shown in FIG. 1, the lower transmission wave TW1 and the upper transmission wave TW2 are used, for example. It is a scan radar that alternately transmits every 5 msec. The lower transmission wave TW1 is transmitted from the lower transmission unit TX1 of the radar device 1 toward the lower side in the traveling direction of the own vehicle A. The upper transmission wave TW2 is transmitted from the upper transmission unit TX2 of the radar device 1 toward the upper side in the traveling direction of the own vehicle A. The lower transmission unit TX1 and the upper transmission unit TX2 are, for example, antennas.

As shown in FIG. 1, the radar device 1 has the downward transmission wave TW1 or the upper transmission wave TW2 alone because a part of the scan range by the lower transmission wave TW1 and the upper transmission wave TW2 overlaps in the direction perpendicular to the own vehicle A. The target T is detected in a wider range in the vertical direction than. The radar device 1 detects the target T by receiving the reflected wave obtained by reflecting the lower transmission wave TW1 and the upper transmission wave TW2 on the target T at the receiving unit RX.

(Configuration of Radar Device According to Embodiment 1)
FIG. 2 is a diagram showing a configuration of a radar device according to the first embodiment. The radar device 1 according to the first embodiment is, for example, an object existing in the vicinity of the own vehicle A by using FM-CW (Frequency Modulated-Continuous Wave) which is a frequency-modulated continuous wave among various methods of millimeter wave radar. The mark T is detected.

As shown in FIG. 2, the radar device 1 is connected to the vehicle control device 2. The vehicle control device 2 is connected to the brake 3 and the like. In the vehicle control device 2, for example, the distance until the transmitted wave emitted by the radar device 1 is reflected by the target T1 and the reflected wave reflected by the target T1 is received by the receiving antenna of the radar device 1 becomes a predetermined distance or less, and the own vehicle A When there is a risk of colliding with the target T1, the brake 3, the throttle, the gear, and the like are controlled to control the behavior of the own vehicle A, and the own vehicle A is prevented from colliding with the target T1. As an example of a system that performs such vehicle control, there is, for example, an ACC (Adaptive Cruise Control) system.

The distance from the transmitted wave irradiated by the radar device 1 to the reflected wave reflected by the target T1 until it is received by the receiving antenna of the radar device 1 is called the "vertical distance", and is the left-right direction (vehicle width) of the own vehicle A. The distance of the target T in the direction) is called "horizontal distance". The left-right direction of the own vehicle A is also the direction of the lane width of the road on which the own vehicle A travels. The "lateral distance" is expressed with a positive value on the right side of the own vehicle A and a negative value on the left side of the own vehicle A with the center position of the own vehicle A as the origin. The "horizontal distance" is a "horizontal position" described later.

Further, as shown in FIG. 2, the radar device 1 includes a transmission unit 4, a reception unit 5, and a signal processing unit 6.

The transmission unit 4 includes a signal generation unit 41, an oscillator 42, a switch 43, a lower transmission unit TX1, and an upper transmission unit TX2. The signal generation unit 41 generates a modulated signal whose voltage changes in a triangular wave shape and supplies it to the oscillator 42. The oscillator 42 frequency-modulates a continuous wave signal based on the modulated signal generated by the signal generation unit 41, generates a transmission signal whose frequency changes with the passage of time, and generates a transmission signal whose frequency changes with the passage of time, and the lower transmission unit TX1 and the upper transmission unit TX2. Output to.

The switch 43 connects either the lower transmission unit TX1 or the upper transmission unit TX2 to the oscillator 42. The switch 43 operates at a predetermined timing (for example, every 5 msec) under the control of the transmission control unit 61 described later, and switches the connection between either the lower transmission unit TX1 or the upper transmission unit TX2 and the oscillator 42. That is, the switch 43 switches the connection with the oscillator 42 in the order of, for example, ... → lower transmission unit TX1 → upper transmission unit TX2 → lower transmission unit TX1 → upper transmission unit TX2 ....

The lower transmission unit TX1 and the upper transmission unit TX2 transmit the lower transmission wave TW1 and the upper transmission wave TW2 to the outside of the own vehicle A based on the transmission signal. Hereinafter, the lower transmission unit TX1 and the upper transmission unit TX2 may be collectively referred to as "transmission unit TX". In FIG. 2, the lower transmission unit TX1 and the upper transmission unit TX2 are illustrated one by one, but the number thereof can be appropriately changed in design. The transmission unit TX is composed of a plurality of antennas, and outputs the downward transmission wave TW1 and the upward transmission wave TW2 in different directions via the plurality of antennas to cover the scan range. Hereinafter, the downward transmission wave TW1 and the upward transmission wave TW2 may be collectively referred to as "transmission wave TW".

The lower transmission unit TX1 and the upper transmission unit TX2 are connected to the oscillator 42 via the switch 43. Therefore, either the lower transmission wave TW1 or the upper transmission wave TW2 is output from one transmission unit TX of the transmission unit TX according to the switching operation of the switch 43. Further, the output transmission wave TW is also sequentially switched by the switching operation of the switch 43.

The receiving unit 5 includes an individual receiving unit 52 connected to each of the receiving units RX, which are four antennas forming the array antenna. In FIG. 2, four receiving units RX are illustrated, and the number of receiving units RX can be appropriately changed in design. Each receiving unit RX receives the reflected wave RW from the target T. Each individual receiving unit 52 processes the reflected wave RW received via the corresponding receiving unit RX.

Each individual receiver 52 includes a mixer 53 and an A / D (Analog / Digital) converter 54. The received signal obtained from the reflected wave RW received by the receiving unit RX is sent to the mixer 53. Corresponding amplifiers may be arranged between the receiving unit RX and the mixer 53, respectively.

The transmission signal distributed from the oscillator 42 of the transmission unit 4 is input to the mixer 53, and the transmission signal and the reception signal are mixed in the mixer 53, respectively. As a result, a beat signal indicating a beat frequency, which is a difference frequency between the frequency of the transmission signal and the frequency of the reception signal, is generated. The beat signal generated by the mixer 53 is converted into a digital signal by the A / D converter 54 and then output to the signal processing unit 6.

The signal processing unit 6 is a microcomputer including a CPU (Central Processing Unit), a storage unit 63, and the like, and controls the entire radar device 1. The signal processing unit 6 stores various data to be calculated, information on a target detected by the data processing unit 7, and the like in the storage unit 63. Further, the storage unit 63 includes a lateral position difference model 63a, a relative velocity difference model 63b, an independent beam tilt difference model 63c, a vertical beam power difference tilt model 63d, a vertical beam surface integral model 63e, and an extrapolation factor type likelihood, which will be described later. The table 63f is stored. The storage unit 63 can use, for example, an EPROM (Erasable Programmable Read Only Memory), a flash memory, or the like, but is not limited thereto.

The signal processing unit 6 includes a transmission control unit 61, a Fourier transform unit 62, and a data processing unit 7 as functions realized by software in a microcomputer. The transmission control unit 61 controls the signal generation unit 41 of the transmission unit 4 and also controls the switching of the switch 43. The data processing unit 7 includes a peak extraction unit 70, an angle estimation unit 71, a pairing unit 72, a continuity determination unit 73, a filtering unit 74, a target classification unit 75, an unnecessary target removal unit 76, a grouping unit 77, and an object. The target information output unit 78 is included.

The Fourier transform unit 62 performs a fast Fourier transform (FFT) on the beat signals output from each of the plurality of individual receiving units 52. As a result, the Fourier transform unit 62 converts the beat signals related to the received signals of the plurality of receiving units RX into a frequency spectrum which is data in the frequency domain. The frequency spectrum generated by the Fourier transform unit 62 is output to the data processing unit 7.

In the frequency spectrum generated by the Fourier transform unit 62, the peak extraction unit 70 sets a peak exceeding a predetermined signal level in each section of the up section in which the frequency of the transmitted signal rises and the down section in which the frequency falls. Extract.

Here, the processing of the peak extraction unit 70 will be described with reference to FIGS. 3, 4A, and 4B. FIG. 3 is a diagram showing the relationship between the transmitted wave and the reflected wave and the beat signal. FIG. 4A is a diagram illustrating peak extraction in the up section. FIG. 4B is a diagram illustrating peak extraction in the down section. For the sake of simplicity, the reflected wave RW shown in FIG. 3 is an ideal reflected wave from one target T. Further, in FIG. 3, the transmitted wave TW is shown by a solid line, and the reflected wave RW is shown by a broken line.

In the upper view of FIG. 3, the vertical axis represents frequency [GHz] and the horizontal axis represents time [msec]. In FIG. 3, the lower transmission wave TW1 is output in the section of timings t1 to t2, and the upper transmission wave TW2 is output in the section of timings t2 to t3.

As shown in FIG. 3, the lower transmission wave TW1 and the upper transmission wave TW2 are continuous waves whose frequencies go up and down in a predetermined period around a predetermined frequency, and the frequencies change linearly with time. .. Here, the center frequency of the lower transmission wave TW1 and the upper transmission wave TW2 is f0, the displacement width of the frequency is ΔF, and the reciprocal of one cycle in which the frequency goes up and down is fm.

Since the reflected wave RW is the one in which the lower transmission wave TW1 and the upper transmission wave TW2 are reflected by the target T, the frequency is a frequency centered on a predetermined frequency in a predetermined period like the lower transmission wave TW1 and the upper transmission wave TW2. Becomes a continuous wave that goes up and down. However, the reflected wave RW has a delay with respect to the downward transmission wave TW1 and the like. The delay time τ corresponds to the vertical distance from the own vehicle A to the target T.

Further, in the reflected wave RW, a frequency shift of the frequency fd occurs with respect to the transmitted wave TW due to the Doppler effect according to the relative speed of the target T with respect to the own vehicle A.

As described above, the reflected wave RW has a frequency shift according to the relative speed as well as a delay time according to the vertical distance with respect to the downward transmission wave TW1 and the like. Therefore, as shown in the lower view of FIG. 3, the beat frequency of the beat signal generated by the mixer 53 decreases in the up section (hereinafter, may be referred to as “UP”) in which the frequency of the transmission signal increases. The value will be different depending on the down section (hereinafter, may be referred to as "DN").

The beat frequency is the frequency of the difference between the frequency of the downward transmission wave TW1 and the like and the frequency of the reflected wave RW. Hereinafter, the beat frequency in the up section is referred to as pup, and the beat frequency in the down section is referred to as fdn. In the lower view of FIG. 3, the vertical axis represents frequency [kHz] and the horizontal axis represents time [msec].

Then, as shown in FIGS. 4A and 4B, after the Fourier transform in the Fourier transform unit 62, waveforms in the respective frequency regions of the beat frequency up in the up section and the beat frequency fdn in the down section are obtained. In FIGS. 4A and 4B, the vertical axis represents the signal power [dB] and the horizontal axis represents the frequency [kHz].

The peak extraction unit 70 extracts peak Pu and peak Pd that exceed a predetermined signal power Pref in the waveforms shown in FIGS. 4A and 4B. The peak extraction unit 70 extracts peaks Pu and Pd for each of the lower transmission wave TW1 and the upper transmission wave TW2 shown in FIG. The predetermined signal power Pref may be constant or variable. Further, the predetermined signal power Pref may be set to a different value in the up section and the down section.

In the frequency spectrum of the up section shown in FIG. 4A, peak Pu appears at the positions of the three frequencies pup1, pup2, and pup3, respectively. Further, in the frequency spectrum of the down section shown in FIG. 4B, peaks Pd appear at the positions of the three frequencies fdn1, fdn2, and fdn3, respectively. In FIGS. 4A and 4B, three peak Pu and three peak Pd are illustrated, but one or more peak Pu and one peak Pd appear. Hereinafter, the frequency may be referred to as another unit, bin. 1 bin corresponds to about 467 Hz.

If the relative velocity is not taken into consideration, the frequency at the position where the peak appears in the frequency spectrum corresponds to the vertical distance of the target T. One bin corresponds to a vertical distance of about 0.36 m. Then, for example, paying attention to the frequency spectrum in the up section, the target exists at the position of the vertical distance corresponding to the frequency up in which the peak Pu appears. Therefore, the peak extraction unit 70 extracts frequencies at which peak Pu and peak Pd having a power exceeding a predetermined signal power Pref appear in both the frequency spectrum of the up section and the down section. Hereinafter, the frequency extracted in this way is referred to as a "peak frequency".

The frequency spectrum of the up section and the down section as shown in FIGS. 4A and 4B is obtained from the received signal received by one receiving unit RX. Therefore, the Fourier transform unit 62 derives the frequency spectrum of each of the up section and the down section from each of the received signals received by the four receiving units RX.

Since the four receiving units RX receive the reflected wave RW from the same target, the peak frequencies extracted are the same between the frequency spectra of the four receiving units RX. However, since the positions of the four receiving units RX are different from each other, the phase of the reflected wave RW is different for each receiving unit RX. Therefore, the phase information of the received signals having the same bin is different for each receiving unit RX. Further, when a plurality of targets exist at different angles of the same bin, the signal of one peak frequency in the frequency spectrum contains information about the plurality of targets.

The angle estimation unit 71 separates information about a plurality of targets existing in the same bin from a signal of one peak frequency by directional calculation processing for each of the up section and the down section, and each of the plurality of targets. Estimate the angle. The targets existing in the same bin are the targets whose vertical distances are substantially the same. The angle estimation unit 71 pays attention to the reception signals of the same bin in all the frequency spectra of the four reception units RX, and estimates the angle of the target based on the phase information of the reception signals.

Well-known angle estimation methods such as ESPRIT (Estimation of Signal Parameters via Rotational Invariance Techniques), MUSIC (Multiple Signal Classification), and PRISM (Panchromatic Remote-sensing Instrument for Stereo Mapping) are used to estimate the angle of such a target. The method can be used. As a result, the angle estimation unit 71 calculates the power of the signals of the plurality of peak angles and the signals of the plurality of angles from the signals of one frequency.

FIG. 5 is a diagram conceptually showing the angle estimated by the directional calculation process as an angle spectrum. In FIG. 5, the vertical axis represents the signal power [dB] and the horizontal axis represents the angle [deg]. In the angle spectrum, the angle estimated by the directional calculation process appears as a peak Pa exceeding a predetermined signal power Pref. Hereinafter, the angle estimated by the directional calculation process is referred to as a "peak angle". The plurality of peak angles derived from the signals of one peak frequency at the same time indicate the angles of a plurality of targets existing in the same bin.

The angle estimation unit 71 derives such a peak angle for all peak frequencies in the frequency spectrum of the up section and the down section.

Through the above processing, the peak extraction unit 70 and the angle estimation unit 71 derive peak data corresponding to each of the plurality of targets existing in front of the own vehicle A in each of the up section and the down section. The peak data includes parameters such as the above-mentioned peak frequency, peak angle, and signal power of the peak angle (hereinafter, referred to as “angle power”).

The pairing unit 72 has a peak Pu in the up section and a peak in the down section based on the degree of agreement between the peak angle and the angle power of the up section calculated by the angle estimation unit 71 and the peak angle and the angle power of the down section. Perform pairing to associate Pd. FIG. 6A is a diagram illustrating pairing based on the azimuth angle and the angular power of each of the up section and the down section. FIG. 6B is a diagram for explaining the pairing result. In FIGS. 6A and 6B, the horizontal axis represents "angle (direction)" and the vertical axis represents "distance (vertical distance)".

As shown in FIG. 6A, the pairing unit 72 pairs peaks whose peak angles and angular powers are close to each other within a predetermined range among the directional calculation results of the peaks of UP and DN. That is, the pairing unit 72 calculates the Mahalanobis distance using, for example, the peak angle and the angular power of the frequency peaks of UP and DN, respectively. A well-known technique is used to calculate the Mahalanobis distance. The pairing unit 72 associates two peaks, UP and DN, at which the Mahalanobis distance is the minimum value.

In this way, the pairing unit 72 associates the peaks related to the same target T with each other. As a result, the pairing unit 72 derives the target data related to each of the plurality of target Ts existing in front of the own vehicle A. This target data is also called "pair data" because it is obtained by associating two peaks.

Then, as shown in FIG. 6B, the pairing unit 72 calculates the relative speed and distance of each target T with respect to the own vehicle A from the peaks of the paired UP and DN (indicated by “◯” in FIG. 6B). To do. For example, the pairing unit 72 uses the two peak data of the up section and the down section, which are the sources of the target data (pair data), to use the parameters (vertical distance, horizontal distance, relative velocity) of the target data. ) Can be derived. The radar device 1 will detect the presence of the target T by pairing.

The processing by the peak extraction unit 70, the angle estimation unit 71, and the pairing unit 72 as described above is performed alternately by the lower transmission unit TX1 and the upper transmission unit TX2 each time the reflected wave RW is received for each beam irradiation (every scan). This is a process for deriving the instantaneous values of the target data parameters (vertical distance, horizontal distance, relative velocity).

The continuity determination unit 73 determines the temporal continuity between the target data derived in the past process and the target data derived in the latest process. That is, the continuity determination unit 73 determines whether or not the target data derived in the past process and the target data derived in the latest process are the same target. For example, the past process is the previous target data derivation process, and the latest process is the current target data derivation process. Specifically, the continuity determination unit 73 predicts the position of the current target data based on the target data derived in the previous target data derivation process, and is derived in the current target data derivation process. It is assumed that the target data closest to the predicted position within a predetermined range is the target data having continuity with the target data derived in the past processing.

In the latest processing, the continuity determination unit 73 has not derived the target data having continuity with the target data derived in the past processing, that is, the continuity of the target data derived in the past processing. If it is determined that there is no target data, the target data that has not been derived in the latest processing is virtually derived based on the parameters (vertical distance, horizontal distance, relative velocity) of the target data derived in the past processing. "Insert processing" is performed.

The extrapolated data derived by the extrapolation process is treated as the target data derived by the latest process. When the extrapolation process is performed on the target data a plurality of times in succession or at a relatively high frequency, the target data is deleted from the predetermined storage area of the storage unit 63 as if the target was lost. To. Specifically, the information of the parameter of the target number indicating the target is deleted, and a value indicating that the parameter has been deleted (a value indicating the deletion flag OFF) is set in the target number. The target number is an index for identifying each target data, and a different number is assigned to each target data.

The filtering unit 74 smoothes the parameters (vertical distance, horizontal distance, relative velocity) of the two target data derived in each of the past processing and the latest processing in the time axis direction to derive the target data. The target data after such filtering is also called "internal filter data" with respect to the pair data representing the instantaneous value.

The target classification unit 75 classifies each target into a preceding vehicle, a stationary object (including a stationary vehicle), and an oncoming vehicle based on the relative speed. For example, the target classification unit 75 classifies a target having a relative speed that is in the same direction as the speed of the own vehicle A and is larger than the magnitude of this speed as a "preceding vehicle". Further, for example, the target classification unit 75 classifies a target having a relative speed that is substantially opposite to the speed of the own vehicle A as a “stationary object”. Further, for example, the target classification unit 75 classifies a target having a relative speed that is opposite to the speed of the own vehicle A and having a relative speed larger than the magnitude of this speed as an “oncoming vehicle”. The "preceding vehicle" may be a target having a relative speed that is in the same direction as the speed of the own vehicle A and is smaller than the magnitude of this speed. Further, the "oncoming vehicle" may be a target having a relative speed that is opposite to the speed of the own vehicle A and is smaller than the magnitude of this speed.

The unnecessary target removal unit 76 determines, among the targets, upper objects, lower objects, rain, received wave ghosts, etc. as unnecessary targets, and excludes them from the output targets. The process of discriminating the upper object among the unnecessary object markers will be described in detail later.

The grouping unit 77 performs grouping that integrates a plurality of target data as target data of the same object into one. For example, the grouping unit 77 reduces the number of output of the target data by collecting the target data of the same object into one output for the target data whose detection positions and velocities are close to each other within a predetermined range. ..

The target information output unit 78 selects a predetermined number (for example, 10) of target data as an output target from a plurality of derived target data or derived by extrapolation, and selects the selected target data as a vehicle. Output to control device 2. The target information output unit 78 preferentially selects target data related to a target existing in the own lane and closer to the own vehicle A based on the vertical distance and the horizontal distance of the target data. Here, the "own lane" is a traveling lane assuming a width of about 1.8 m at both ends of the lane when the own vehicle A travels substantially in the center of the lane. The width that defines the "own lane" can be changed in design as appropriate.

The target data derived by the above target data derivation process is stored in a predetermined storage area of the storage unit 63 as a parameter corresponding to the target number indicating each target data, and the target data can be derived from the next time onward. In the processing, it is used as the target data derived in the past processing.

That is, the target data derived in the past target data derivation process is saved as a "history". For example, the peak extraction unit 70 refers to the "peak frequency" stored in the predetermined storage area of the storage unit 63 as the "history", predicts the "peak frequency" having temporal continuity with the "history", and predicts the "peak frequency". A frequency within, for example, ± 3 bin of the predicted “peak frequency” is extracted. As a result, the radar device 1 can quickly select the "peak frequency" corresponding to the target that needs to be preferentially output to the vehicle control device 2. The "peak frequency" of the predicted target data this time is called "predicted bin".

(Distinguishing process of stationary vehicle and upper object according to the first embodiment)
Hereinafter, with reference to FIGS. 7A to 18, the details of the process of discriminating between the stationary vehicle and the upper object performed by the unnecessary object target removing unit 76 according to the first embodiment will be described in the order of STEP1 to STEP3.

<STEP1: Calculation of probability ratio>
The unnecessary target removal unit 76 obtains six probability ratios (log-likelihoods) related to the target from the parameters based on the target data acquired for each scan and each corresponding probability distribution model based on the following equation (1). calculate. The probability distribution model used for calculating each of the six probability ratios is defined or constructed in advance, for example, every 10 m based on actual measurement data, and a model in which less than 10 m is linearly interpolated is used.

As the probability distribution model used for calculating the probability ratio, as described above with reference to FIG. 2, the lateral position difference model 63a, the relative velocity difference model 63b, the independent beam tilt difference model 63c, and the vertical beam power difference tilt model 63d , Upper and lower beam area differential model 63e, and extrapolation factor type probability table 63f. Details of the lateral position difference model 63a will be described later with reference to FIG. Details of the relative velocity difference model 63b will be described later with reference to FIG. Details of the independent beam tilt difference model 63c will be described later with reference to FIG. Details of the vertical beam power difference inclination model 63d will be described later with reference to FIG. Details of the vertical beam area differential model 63e will be described later with reference to FIG. Details of the extrapolation factor type likelihood table 63f will be described later with reference to FIG.

In the above equation (1), the "probability ratio 1" when n = 1 is the log-likelihood of a stationary vehicle and an upper object based on the parameter "lateral position difference" described later. In the above equation (1), the "probability ratio 2" when n = 2 is the log-likelihood of a stationary vehicle and an upper object based on the parameter "relative velocity difference" described later. In the above equation (1), the "probability ratio 3" when n = 3 is the log-likelihood of a stationary vehicle and an upper object based on the parameter "independent beam inclination difference" described later. In the above equation (1), the "probability ratio 4" when n = 4 is the log-likelihood of the stationary vehicle and the upper object based on the parameter "vertical beam power difference slope" described later. In the above equation (1), the “probability ratio 5” when n = 5 is the log-likelihood of the stationary vehicle and the upper object based on the parameter “upper and lower beam surface integral” described later. In the above equation (1), the "probability ratio 6" when n = 6 is the log-likelihood of a stationary vehicle and an upper object based on the parameter "extrapolation factor type" described later.

-STEP1-11: Logarithmic likelihood calculation based on "horizontal position difference" The unnecessary target removal unit 76 is described below (2) from the difference between the current horizontal position acquired by the current scan and the previous horizontal position acquired by the previous scan. Calculate the lateral position difference based on the formula.

FIG. 7A is a diagram illustrating a method of calculating the lateral position difference according to the first embodiment. In FIG. 7A, the horizontal axis is the distance (detection distance) and the vertical axis is the horizontal position. As shown in FIG. 7A, among the horizontal positions of the target acquired every 50 msec, for example, from the current horizontal position (for example, corresponding to (1) in FIG. 7A) to the previous horizontal position (for example, (2) in FIG. 7A). (Equivalent to) is subtracted to calculate the lateral position difference.

FIG. 7B is a diagram illustrating a background for calculating the lateral position difference according to the first embodiment. As shown in FIG. 7B, the lateral position difference is calculated by (a) the position of the reflection point is stable in the vehicle, while (b) the position of the reflection point is in the upper object due to the lateral width. It is based on the fact that the lateral position difference tends to be large because it becomes unstable and the lateral position can move.

Then, the unnecessary object target removing unit 76 calculates the “probability ratio 1” based on the above equation (1) from the lateral position difference model 63a illustrated in FIG. 8 and the lateral position difference calculated based on the above equation (2). .. FIG. 8 is a diagram showing a lateral position difference model according to the first embodiment. The lateral position difference model 63a is a probability distribution model in which the horizontal axis is the lateral position difference [m] and the vertical axis is the likelihood, and the relationship between the lateral position difference and the likelihood of the stationary vehicle and the upper object is shown. The probability distribution model of a stationary vehicle and an upper object shown in FIG. 8 is a model based on, for example, a normal distribution, which is constructed in advance by the maximum likelihood estimation method and the design of experiments method. In the lateral position difference model 63a, the parameters that characterize the model are adjusted for each of the stationary vehicle and the upper object in order to improve the determination accuracy.

Note that FIG. 8 illustrates a lateral position difference model when the distance from the own vehicle A to the target is 80 m as the lateral position difference model 63a, and the distance from the own vehicle A to the target is 10 to 80 to 150 m. The illustration of the lateral position difference model of each distance in units of 10 m up to the degree is omitted.

For example, consider the case where the "horizontal position difference" calculated based on the above equation (2) is "1". In this case, referring to FIG. 8, when the lateral position difference on the horizontal axis is "1", the likelihood of the stationary vehicle on the vertical axis is about "0.13" and the likelihood of the upper object is about "0". It becomes .27 ”. Therefore, in the above equation (1), when n = 1, the probability ratio 1 = log (still vehicle likelihood 1) -log (upper object likelihood 1) = log (0.13) -log (0.27). ), The probability ratio 1 can be calculated.

-STEP 1-2: Log likelihood calculation based on "relative velocity difference" The unnecessary target removal unit 76 is based on the derivative of the distance difference between the current distance acquired by the current scan and the previous distance acquired by the previous scan, and the relative velocity this time. , Calculate the relative speed difference based on the following equation (3).

FIG. 9A is a diagram illustrating a method of calculating the relative speed difference according to the first embodiment. As shown in FIG. 9A, among the distances of the target acquired every 50 msec, for example, the distance from the current distance (for example, corresponding to (1) in FIG. 9A) to the previous distance (for example, corresponding to (2) in FIG. 9A). Is subtracted, (this time distance-previous distance) is calculated, and (this time distance-previous distance) is further divided by the update cycle (for example, ΔT = 50 msec in FIG. 9A) to divide the distance between the current distance and the previous distance. Calculate the differential of the difference. Then, the relative velocity (for example, corresponding to (3) in FIG. 9A) is added to the reverse sign of the derivative of the distance difference between the current distance and the previous distance to calculate the relative velocity difference.

FIG. 9B is a diagram illustrating a background for calculating the relative speed difference according to the first embodiment. As shown in FIG. 9B, the relative velocity difference is calculated because the position of the reflection point becomes unstable due to the vertical and horizontal widths of the upper object and the vertical and horizontal positions can move, so that the distance from the radar device 1 is large. It is based on the fact that it tends to change and the time derivative of the detection distance becomes large.

Then, the unnecessary object target removing unit 76 calculates the “probability ratio 2” based on the above equation (1) from the relative velocity difference model 63b illustrated in FIG. 10 and the relative velocity difference calculated based on the above equation (3). .. FIG. 10 is a diagram showing a relative speed difference model according to the first embodiment. The relative speed difference model 63b is a probability distribution model in which the horizontal axis is the relative speed difference [m / s] and the vertical axis is the likelihood, and the relationship between the relative speed difference and the likelihood of the stationary vehicle and the upper object is shown. The probability distribution model of a stationary vehicle and an upper object shown in FIG. 10 is a model based on, for example, a normal distribution, which is constructed in advance by the maximum likelihood estimation method and the design of experiments method. In the relative speed difference model 63b, the parameters that characterize the model are adjusted for each of the stationary vehicle and the upper object in order to improve the determination accuracy.

Note that FIG. 10 illustrates a relative speed difference model when the distance from the own vehicle A to the target is 80 m as the relative speed difference model 63b, and the distance from the own vehicle A to the target is 10 to 80 to 150 m. The illustration of the relative velocity difference model of each distance in units of 10 m up to the degree is omitted.

For example, consider the case where the "relative velocity difference" calculated based on the above equation (3) is "2". In this case, referring to FIG. 10, when the relative speed difference on the horizontal axis is "2", the likelihood of the stationary vehicle on the vertical axis is about "0.11" and the likelihood of the upper object is about "0". It becomes .75 ". Therefore, in the above equation (1), when n = 2, the probability ratio 2 = log (still vehicle likelihood 2) -log (upper object likelihood 2) = log (0.11) -log (0.75). ), The probability ratio 2 can be calculated.

-STEP1-3: Logarithmic likelihood calculation based on "independent beam tilt difference" The unnecessary target removal unit 76 is described below from the current distance and angular power acquired by the current scan and the previous distance and angular power acquired by the previous scan. The independent beam tilt difference is calculated based on the equation (4). Since the independent beam tilt difference is related to the power fluctuation of multipath, it is based on the fact that a stationary vehicle and an upper object can be discriminated by looking at the independent beam tilt difference.

FIG. 11 is a diagram illustrating the calculation of the upper beam inclination and the lower beam inclination, which are parameters for calculating the independent beam inclination difference according to the first embodiment. As shown in FIG. 11, for example, among the angular powers of the targets of the upper and lower beams acquired every 50 msec, the lower beam has the previous angular power from the current angular power difference (for example, corresponding to (1) in FIG. 11). The difference (for example, corresponding to (2) in FIG. 11) is subtracted, and the subtraction result is divided by (previous distance-current distance) to calculate the lower beam inclination. Further, for the upper beam, the previous angular power difference (for example, corresponding to (4) in FIG. 11) is subtracted from the current angular power difference (for example, corresponding to (3) in FIG. 11), and the subtraction result is obtained (previous distance-). Divide by (distance this time) to calculate the upper beam tilt difference. Then, the previous upper beam inclination is subtracted from the current upper beam inclination to calculate the independent beam inclination difference of the upper beam. In addition, the independent beam inclination of the lower beam is calculated by subtracting the previous lower beam inclination from the lower beam inclination of this time.

Then, the unnecessary object target removing unit 76 obtains a “probability ratio 3” based on the above equation (1) from the independent beam inclination difference model 63c illustrated in FIG. 12 and the independent beam inclination difference calculated based on the above equation (4). calculate. FIG. 12 is a diagram showing an independent beam inclination difference model according to the first embodiment. The independent beam tilt difference model 63c is a probability distribution model in which the horizontal axis is the independent beam tilt difference [dB / m] and the vertical axis is the likelihood, and the relationship between the independent beam tilt difference and the likelihood of the stationary vehicle and the upper object is shown. Is. The probability distribution model of a stationary vehicle and an upper object shown in FIG. 12 is a model based on, for example, a normal distribution, which is constructed in advance by the maximum likelihood estimation method and the design of experiments method. In the independent beam tilt difference model 63c, the parameters that characterize the model are adjusted for each of the stationary vehicle and the upper object in order to improve the determination accuracy.

Note that FIG. 12 illustrates a case where the distance from the own vehicle A to the target is 80 m as the independent beam inclination difference model 63c, and the distance from the own vehicle A to the target is 10 m from 10 to 80 to 150 m. The illustration of the independent beam tilt difference model for each distance of the unit is omitted.

For example, consider the case where the "independent beam inclination difference" calculated based on the above equation (4) is "0". In this case, referring to FIG. 12, when the independent beam inclination difference on the horizontal axis is "0", the likelihood of the stationary vehicle on the vertical axis is about "0.14" and the likelihood of the upper object is about "0". It becomes 0.05 ". Therefore, in the above equation (1), when n = 3, the probability ratio 3 = log (still vehicle likelihood 3) -log (upper object likelihood 3) = log (0.14) -log (0.05). ), The probability ratio 3 can be calculated.

STEP1-4: Logarithmic likelihood calculation based on "upper and lower beam power difference slope" The unnecessary target removal unit 76 calculates the slope of the power difference in the upper and lower beams based on the following equation (5). Since the vertical beam power difference slope is related to the power fluctuation of multipath, it is based on the fact that a stationary vehicle and an upper object can be discriminated by looking at the vertical beam power difference slope.

FIG. 13 is a diagram illustrating a method of calculating the vertical beam power difference inclination according to the first embodiment. As shown in FIG. 13, among the angular powers of the targets of the upper and lower beams acquired every 50 msec, for example, the lower beam previous angle power (for example, corresponding to (1) in FIG. 13) and the lower beam current angle power ( For example, (corresponding to (3) in FIG. 13) is linearly interpolated, and the interpolated value of the lower beam angle power (for example, corresponding to (2) in FIG. 13) at the distance d1 corresponding to the previous angle power of the upper beam (for example, corresponding to (2) in FIG. 13) is interpolated. (Corresponding to (2') in) is calculated. Further, the upper beam previous angle power (for example, corresponding to (2) in FIG. 13) and the upper beam current angle power (for example, corresponding to (4) in FIG. 13) are linearly interpolated, and the lower beam current angle power (for example, FIG. The interpolated value of the upper beam angle power at the distance d2 corresponding to (corresponding to (3) in 13) (for example, corresponding to (3') in FIG. 13) is calculated.

The reason why the interpolated value is used as described above is that the upper beam and the lower beam are transmitted alternately at a cycle of 50 msec, and there is a time lag in the timing of the angular powers of both. This is to calculate the difference.

Then, the unwanted object removal unit 76 sets the "current angle power difference" as "interpolated value of the upper beam angle power (for example, corresponding to (3') in FIG. 13) -lower beam this time angle power (for example, in FIG. 13). (Equivalent to (3)) ”is calculated. Further, the unwanted object removal unit 76 sets the "previous angle power difference" as "upper beam previous angle power (for example, corresponding to (2) in FIG. 13) -lower beam angle power interpolated value (for example, (for example, in FIG. 13). 2') corresponds to) "is calculated. Then, the unnecessary object target removing unit 76 calculates the vertical beam power difference inclination based on the above equation (5). However, "d1" and "d2" in the above equation (5) are "d1" and "d2" in FIG.

Then, the unnecessary object target removing unit 76 has a “probability ratio 4” based on the above equation (1) from the vertical beam power difference inclination model 63d illustrated in FIG. 14 and the vertical beam power difference inclination calculated based on the above equation (5). "Calculate. FIG. 14 is a diagram showing a vertical beam power difference inclination model according to the first embodiment. In the vertical beam power difference slope model 63d, the horizontal axis is the vertical beam difference slope [dB / m] and the vertical axis is the likelihood, and the probability distribution showing the relationship between the vertical beam difference slope and the likelihood of each of the stationary vehicle and the upper object. It is a model. The probability distribution model of a stationary vehicle and an upper object shown in FIG. 14 is a model based on, for example, a normal distribution, which is constructed in advance by the maximum likelihood estimation method and the design of experiments method. In the vertical beam power difference inclination model 63d, the parameters that characterize the model are adjusted for each of the stationary vehicle and the upper object in order to improve the determination accuracy.

Note that FIG. 14 illustrates a vertical beam power difference tilt model when the distance from the own vehicle A to the target is 80 m as the vertical beam power difference tilt model 63d, and the distance from the own vehicle A to the target is 10. The illustration of the vertical beam power difference inclination model for each distance in units of 10 m from 80 to 150 m is omitted.

For example, consider the case where the "upper and lower beam difference inclination" calculated based on the above equation (5) is "0". In this case, referring to FIG. 14, when the vertical beam difference inclination on the horizontal axis is "0", the likelihood of the stationary vehicle on the vertical axis is about "0.2" and the likelihood of the upper object is about "0". It becomes 0.06 ”. Therefore, in the above equation (1), when n = 4, the probability ratio 4 = log (still vehicle likelihood 4) -log (upper object likelihood 4) = log (0.2) -log (0.06). ), The probability ratio 4 can be calculated.

-STEP1-5: Log-likelihood calculation based on "upper and lower beam surface integral" The multipath of the reflected wave affects the angular power, and the effect changes depending on the height of the target. At long distances, the effect of multipath appears more strongly on objects above than stationary vehicles. In addition, the distance (frequency per distance) at which the convex Null point due to multipath appears is not easily affected by the vertical mounting angle of the radar device. It is not easily affected by the radar mounting conditions. The "convex Null" is a curve that is convex upward in the vicinity of the maximum point, and refers to a curve that takes a shape similar to the vicinity of the minimum point of a cycloid curve, for example, in the vicinity of the minimum point.

FIG. 15A is a diagram showing the relationship between the angular power and the distance of a stationary vehicle. FIG. 15B is a diagram showing the relationship between the angular power and the distance of the upper object. From the frame-enclosed portion of FIG. 15A, it can be seen that the stationary vehicle is affected by multipath at a distance of 70 to 80 m or less, and the distribution of the angular power varies and the amount of change is large. On the other hand, from the frame-enclosed portion of FIG. 15B, it can be seen that the upper object is affected by the multipath regardless of the distance, and the distribution of the angular power varies and the amount of change is large.

Therefore, the unwanted object removal unit 76 calculates the vertical beam surface integral derivative as shown in FIG. FIG. 16 is a diagram illustrating a method of calculating the vertical beam surface integral according to the first embodiment. As shown in FIG. 16, for example, the angular power of the same distance is calculated by linear interpolation for the angular power of each target of the upper and lower beams acquired every 50 msec, and the "angle power difference" of the upper and lower beams of the same distance is calculated. (Equivalent to (1) to (6) in (a) of FIG. 16).

As described above, the "angle power difference" of the upper and lower beams is calculated by using the angular power of the same distance by linear interpolation. The upper beam and the lower beam are alternately transmitted at a cycle of 50 msec, and the angular power of both is calculated. This is because there is a time lag in the timing of, and the timing is aligned to calculate a more accurate angular power difference.

Then, as shown in FIG. 16B, the unwanted object removal unit 76 sets | (1) the angular power difference − (2) the angular power difference | as the “vertical beam angular power difference”, and {( The area of (1') is obtained by calculating "upper and lower beam angle power difference" x "distance difference" with the distance of 1) − the distance of (2)} as the “distance difference”. | * | Is the absolute value of *.

Similarly, the area of (2') = | the angular power difference of (2)-the angular power difference of (3) | x {the distance of (2)-the distance of (3)}, the area of (3') = | Angle power difference of (3)-angle power difference of (4) | × {distance of (3)-distance of (4)}, area of (4') = | angle power difference of (5)-(6) Angle power difference | × {distance of (5) − distance of (6)} is calculated.

FIG. 16 (c) shows the area in each cycle calculated as described above. Then, the unnecessary target removal unit 76 sets the total value of the area (vertical beam angle power difference) in each cycle of the same target calculated as described above based on the following equation (6) as the total of the distance differences. Divide by to calculate the "upper and lower beam area derivative".

Then, the unnecessary object target removing unit 76 obtains a “probability ratio 5” based on the above equation (1) from the vertical beam area differential model 63e illustrated in FIG. 17 and the “vertical beam area differential” calculated as described above. calculate. FIG. 17 is a diagram showing a vertical beam surface integral differential model according to the first embodiment. The vertical beam surface integral model 63e is a probability distribution model in which the horizontal axis is the vertical beam surface integral [dB] and the vertical axis is the likelihood, and the relationship between the vertical beam surface integral and the likelihood of each of the stationary vehicle and the upper object is shown. .. The probability distribution model of a stationary vehicle and an upper object shown in FIG. 17 is a model based on, for example, a strain normal distribution, which is constructed in advance by the maximum likelihood estimation method and the design of experiments method. In the vertical beam area differential model 63e, the parameters that characterize the model are adjusted for each of the stationary vehicle and the upper object in order to improve the determination accuracy.

Note that FIG. 17 illustrates a vertical beam area differential model when the distance from the own vehicle A to the target is 80 m as the vertical beam area differential model 63e, and the distance from the own vehicle A to the target is 10 to 80. The illustration of the vertical beam power difference inclination model for each distance in units of 10 m up to about 150 m is omitted.

For example, consider the case where the "upper and lower beam surface integral" calculated based on the above equation (6) is "1". In this case, referring to FIG. 17, when the vertical beam area derivative on the horizontal axis is "1", the likelihood of the stationary vehicle on the vertical axis is about "0.52" and the likelihood of the upper object is about "1". It becomes 0.1 ”. Therefore, in the above equation (1), when n = 5, the probability ratio 5 = log (still vehicle likelihood 5) -log (upper object likelihood 5) = log (0.52) -log (0.1). ), The probability ratio 5 can be calculated.

-STEP1-6: Logarithmic likelihood calculation based on "extrapolation factor type" The upper object may be extrapolated due to unstable reflection due to the effects of multipoint reflection, reflection point movement, power attenuation due to multipath, etc. There are many. Therefore, judging from the characteristics of extrapolation, the likelihood of a stationary vehicle and an upper object is calculated. That is, with respect to the same target, the likelihood logarithm is determined from the extrapolation factor type likelihood table 63f shown in FIG. 18 based on the presence or absence of extrapolation in this scan and the factors of extrapolation. FIG. 18 is a diagram showing an extrapolation factor type likelihood table according to the first embodiment. In FIG. 18, the description of specific numerical values is omitted by "...". For example, the continuity determination unit 73 performs extrapolation processing, and stores the presence / absence of extrapolation and the type of extrapolation factor in the case of having extrapolation in a predetermined storage area of the storage unit 63.

There are seven types of extrapolation factors, for example, "no history", "no peak", "no angle", "no continuity", "predicted bin deviation", "Mahalanobis distance NG", and "no pair".

"No history" means that the "history" corresponding to the "peak frequency" extracted this time cannot be acquired, or the "history" does not exist. “No peak” means that the peak extraction unit 70 cannot extract the peak from the frequency spectrum generated by the Fourier transform unit 62. “No angle” means that the peak extraction unit 70 can extract the peak, but the angle estimation unit 71 cannot estimate the angle of the target.

"No continuity" means that the pairing unit 72 was able to perform pairing, but the continuity determination unit 73 determined that there was no temporal continuity with the target data derived in the latest processing. To say that.

“Predicted bin deviation” means that the actual position of the current target data does not exist within a predetermined range (for example, within ± 3 bin) of the predicted position of the current target data predicted by the continuity determination unit 73. To say.

“Mahalanobis distance NG” means that pairing cannot be performed by the pairing unit 72 because the minimum value of the Mahalanobis distance is equal to or greater than a predetermined value. “No pair” means that pairing cannot be performed by the pairing unit 72 due to factors other than “no history”, “no peak”, “no angle”, “no continuity”, “predicted bin deviation”, and “Mahalanobis distance NG”.

In the case of no extrapolation in this scan, that is, in the case of normal detection, the unnecessary object target removing unit 76 sets each “normal detection” of the extrapolation factor type likelihood table 63f in FIG. 18 as each likelihood logarithm of the stationary vehicle and the upper object. Read the likelihood logarithm corresponding to the detection distance in the "likelihood logarithm" column. For example, when the detection distance R is 100 m, the likelihood logarithm of the stationary vehicle is "-0.04" and the likelihood logarithm of the upper object is "-0.16" with reference to the line "95 <R≤105". It becomes. Therefore, in the above equation (1), when n = 6, the probability ratio 6 = log (still vehicle likelihood 6) -log (upper object likelihood 6) = (-0.04)-(-0.16). ), The probability ratio 6 can be calculated.

Further, when the unnecessary object target removing unit 76 has extrapolation in this scan and the extrapolation factor type is "no history", the extrapolation factor type shown in FIG. 18 is set as each likelihood logarithm of the stationary vehicle and the upper object. The likelihood logarithm corresponding to the detection distance in each "historyless likelihood logarithm" column of the likelihood table 63f is read out. For example, when the detection distance R is 100 m, the likelihood logarithm of the stationary vehicle is "-2.48" and the likelihood logarithm of the upper object is "-1.13" with reference to the line "95 <R≤105". It becomes. Therefore, in the above equation (1), when n = 6, the probability ratio 6 = log (still vehicle likelihood 6) -log (upper object likelihood 6) = (-2.48)-(-1.13). ), The probability ratio 6 can be calculated. The same applies to other extrapolation factor types.

<STEP2: Calculation of probability ratio Overall>
The unwanted object removal unit 76 calculates the probability ratio Overall, which is the sum of the probability ratios 1 to 6 calculated in STEP1-1 to STEP1-6 described above, based on the following equation (7).

<STEP3: Discrimination process for stationary vehicles and objects above>
The unnecessary target removal unit 76 determines whether the target is a stationary vehicle or an upper object by determining the threshold value of the probability ratio Overall all calculated in STEP 2 described above. That is, the unnecessary target removal unit 76 determines that the target is a stationary vehicle when the probability ratio Overall is above the predetermined threshold, and determines that the target is an upper object when the probability ratio is less than the predetermined threshold.

(Target information output processing according to the first embodiment)
FIG. 19A is a flowchart showing the target information output process according to the first embodiment. The signal processing unit 6 periodically repeats the target information derivation process at regular time intervals (for example, 5 msec seconds). At the start of the target information derivation process, the beat signal converted from the reflected wave RW is input from the four receiving units RX to the signal processing unit 6.

First, the Fourier transform unit 62 of the signal processing unit 6 performs a fast Fourier transform on the beat signals output from each of the plurality of individual receiving units 52 (step S11). Next, the peak extraction unit 70 extracts peaks exceeding a predetermined signal level from the frequency spectrum generated by the Fourier transform unit 62 in an up section in which the frequency of the transmitted signal rises and a down section in which the frequency falls. Extract in the section (step S12).

Next, the angle estimation unit 71 separates information about a plurality of targets existing in the same bin from a signal of one peak frequency by directional calculation processing for each of the up section and the down section, and those plurality of objects. The angle of each mark is estimated (step S13).

Next, the pairing unit 72 associates peaks related to the same target T with each other, and derives pairing data related to each of the plurality of target T existing in front of the own vehicle A (pairing, step S14). .. Next, the continuity determination unit 73 determines the continuity of whether or not the target data derived in the past process and the target data derived in the latest process are the same target (step S15).

Next, the filtering unit 74 smoothes the parameters (vertical distance, horizontal distance, relative velocity) of the two target data derived in each of the past processing and the latest processing in the time axis direction, and the target data (target data (vertical distance, horizontal distance, relative velocity)). (Internal filter data) is derived (filtering, step S16). Next, the target classification unit 75 classifies each target into a preceding vehicle, a stationary object (including a stationary vehicle), and an oncoming vehicle based on the relative speed (step S17).

Next, the unnecessary target removing unit 76 determines the upper object, the lower object, the rain, and the like as unnecessary objects among the targets, and excludes them from the output target (step S18). Of the processes in step S18, the process of excluding the upper object from the output target will be described later with reference to FIG. 19B.

Next, the grouping unit 77 performs grouping that integrates a plurality of target data as target data of the same object into one (step S19). Next, the target information output unit 78 selects a predetermined number of target data as output targets from a plurality of target data derived or derived by extrapolation, and selects the selected target data as a vehicle control device. Output to 2 (step S20). When the step S20 is completed, the signal processing unit 6 ends the target information derivation process.

(Removal of unnecessary objects according to the first embodiment)
FIG. 19B is a flowchart showing a subroutine for removing unnecessary objects according to the first embodiment. FIG. 19B shows a flow of a process for removing an upper object according to the first embodiment among the unnecessary object removal in step S18 shown in FIG. 19A.

First, the unnecessary target removal unit 76 calculates the lateral position difference of the target based on the above equation (2) (step S18-1). Next, the unwanted object removal unit 76 calculates the probability ratio 1 based on the lateral position difference based on the lateral position difference calculated in step S18-1 and the above equation (1) (step S18-2).

Next, the unnecessary target removal unit 76 calculates the relative velocity difference of the target based on the above equation (3) (step S18-3). Next, the unwanted object removal unit 76 calculates the probability ratio 2 based on the relative velocity difference based on the relative velocity difference calculated in step S18-3 and the above equation (1) (step S18-4).

Next, the unwanted object removal unit 76 calculates the independent beam inclination difference based on the above equation (4) (step S18-5). Next, the unwanted object target removing unit 76 calculates the probability ratio 3 based on the independent beam inclination difference based on the independent beam inclination difference calculated in step S18-5 and the above equation (1) (step S18-6).

Next, the unwanted object removal unit 76 calculates the vertical beam power difference inclination based on the above equation (5) (step S18-7). Next, the unwanted object removal unit 76 calculates the probability ratio 4 based on the vertical beam power difference slope calculated in step S18-7 and the above equation (1) (step S18-8). ..

Next, the unwanted object removal unit 76 calculates the upper and lower beam surface integrals based on the above equation (6) (step S18-9). Next, the unwanted object removal unit 76 calculates the probability ratio 5 based on the vertical beam area differential and the above equation (1) based on the vertical beam area differential calculated in step S18-9 (step S18-10).

Next, the unwanted object removal unit 76 specifies the presence / absence of extrapolation and the type of extrapolation factor (step S18-11). Next, the unwanted object removal unit 76 calculates the probability ratio 6 based on the presence / absence of extrapolation specified in step S18-11, the type of extrapolation factor, and the above equation (1) (step S18-12).

Next, the unwanted object removal unit 76 calculates the probability ratio Overall All based on the above equation (7) (step S18-13). Next, the unwanted object removal unit 76 determines whether or not the probability ratio Overall calculated in step S18-13 is equal to or greater than the threshold value. When the probability ratio Overall is equal to or greater than the threshold value (step S18-14: Yes), the unwanted object removal unit 76 determines that the target is a stationary vehicle (step S18-15). On the other hand, when the probability ratio Overall is less than the threshold value (step S18-14: No), the unnecessary object target removing unit 76 determines that the target is an upper object (step S18-16). When step S18-15 or step S18-16 is completed, the unwanted object removal unit 76 shifts the process to step S19 in FIG. 19A.

In the first embodiment, for a plurality of parameters, the values that the stationary vehicle and the upper object can take and the frequency of occurrence thereof are regarded as the probability density function, and the stationary vehicle probability (static vehicle likelihood) and the upper object probability (upper object likelihood) of the current values are regarded as the probability density functions. Degree) is calculated based on each probability density function of multiple parameters. Further, in the first embodiment, Bayesian update is performed to update the prior probability to the posterior probability every time data is acquired. This is repeated, and it is determined whether the vehicle is a stationary vehicle or an upper object in the manner of a Bayesian filter based on the logarithmic ratio of the stationary vehicle probability and the upper object probability.

Therefore, according to the first embodiment, whether or not the target detected in the traveling direction of the own vehicle collides with the own vehicle (for example, whether or not the target requires vehicle control such as brake control) is determined. It can be identified from a relatively long distance (for example, about 80 m in front of the target), the detection rate is improved, and the vehicle control based on the target detection can be operated at an appropriate timing and an appropriate instruction.

Further, in the first embodiment, the probability ratios 1 to 6 are parameters that are not easily affected by the mounting height of the radar device 1 on the own vehicle A and the vertical mounting angle, and the stationary vehicle and the stationary vehicle and the radar mounting conditions do not depend on the radar mounting conditions. It is possible to accurately identify the upper object.

[Embodiment 2]
In the above-described first embodiment, it may not be possible to sufficiently distinguish between a stationary vehicle and an upper object. That is, as shown in FIG. 20, there is a range in which the value of the probability ratio Overall based on a plurality of parameters overlaps between a stationary vehicle and an upper object. FIG. 20 is a diagram showing overlapping distributions of probability ratios of stationary vehicles and upper objects. Therefore, in the following second embodiment, a method of separating the probability ratio Overall is described in order to improve the determination accuracy.

Hereinafter, the causes ( _Hi ) (i = 1, 2) are defined as shown in the following (Table 1).

Further, the parameters (D _j ) (j = 1 to 7) are defined as shown in the following (Table 2). It is assumed that _{D 1 to} D ₇ in the following (Table 2) are all independent.

_{At this time, the posterior probability P (H 1} | D), which is the probability that the target is a stationary vehicle based on the parameters D (D _{1 to} D ₇ ), is expressed by the following equation (8-1). _{The posterior probability P (H 2} | D), which is the probability that the target is an upper object based on the parameters D (D _{1 to} D ₇ ), is expressed by the following equation (8-2).

Using a naive Bayes filter, the determination of a stationary vehicle and an upper object is made by a logit, which is a logarithm of the probability ratio of the stationary vehicle and the upper object, as shown in the following equation (9).

Here, from the assumption that the parameters D (D _{1 to} D ₇ ) are independent, P (D | H ₁ ) is the following equation (10-1), and P (D | H ₂ ) is the following (10-2). ) Can be expressed as an expression.

When the above equation (9) is modified by using the above equations (10-1) to (10-2), the following equations (11-1) to (11-2) are obtained.

Then, as shown in the above equation (11-2), the first term of the above equation (11-2) is x ₁ , the second term is x ₂ , ..., The seventh term is x ₇ , the eighth term. Assuming that x ₀ , the above equation (11-2) can be expressed as the following equation (12).

The part enclosed in parentheses in the above equation (12) can be interpreted as a generalized linear model with _{x 1 to x 7 as predictive variables.} Therefore, in the second embodiment, the regression coefficient and the intercept are determined from the multivariate analysis for the portion enclosed in parentheses in the above equation (12) to improve the discrimination accuracy between the stationary vehicle and the upper object.

The target data handled in the discrimination between the stationary vehicle and the upper object is qualitative data like the stationary vehicle and the upper object. In addition, all normalities cannot be assumed for the predictive variables (probability ratio 1 (x ₁ ) to probability ratio 7 (x _7)). Therefore, logistic regression analysis is adopted as a method for multivariate analysis, predictive variables are selected and integrated in consideration of multicollinearity and AIC (Akaike's Information Criterion), and regression coefficients (weighting factors) and sections are detected. Determined every 10 m. However, if the detection distance is less than 10 m, the regression coefficient (weighting coefficient) and intercept are calculated by linear interpolation.

In the second embodiment, the weighting coefficient / intercept table A63g described later is switched when the initial detection distance of the target is, for example, 80 to 110 m, and the weighting coefficient / intercept table B63h described later is switched when the initial detection distance is other than the first distance. By using the above, the accuracy of distinguishing between stationary vehicles and objects above is improved.

(Outline of target detection by the radar device according to the second embodiment)
FIG. 21 is a diagram showing a configuration of a radar device according to the second embodiment. FIG. 22A is a diagram showing a table A for storing weighting coefficients and intercepts for regressing each probability ratio according to the initial detection distance according to the second embodiment. FIG. 22B is a diagram showing a table B for storing weighting coefficients and intercepts for regressing each probability ratio according to the initial detection distance according to the second embodiment.

As shown in FIG. 21, the radar device 1A according to the second embodiment includes a signal processing unit 6A, a data processing unit 7A, and a storage unit 63A. The data processing unit 7A of the signal processing unit 6A includes an unnecessary target removal unit 76A. Further, the storage unit 63A further stores the weighting coefficient / intercept table A63g shown in FIG. 22A and the weighting coefficient / intercept table B63h shown in FIG. 22B. Regarding other configurations, the radar device 1A according to the second embodiment is the same as the radar device 1 according to the first embodiment.

The weighting coefficient / intercept table A63g shown in FIG. 22A is a weighting coefficient that corrects the probability ratio i (i = 1-3, 5-6) when the initial detection distance of the target is, for example, the first distance of 80 to 110 m. α _i and intercept β are stored for each initial detection distance. Further, in the weighting coefficient / intercept table B63h shown in FIG. 22B, when the initial detection distance of the target is other than the above-mentioned first distance, the weighting coefficient i (i = 1-3, 5-6) is corrected, respectively. α _i and intercept β are stored for each initial detection distance.

_{In FIGS. 22A and 22B, parameters D 1 to} D ₆ , that is, probability ratios 1 to probability ratio 6 are selected by selecting the predictive variables (parameters) based on the multicollinearity and AIC described above, and correspond _{to the parameters D 7.} The probability ratio to do is excluded. Further, in FIGS. 22A and 22B, the parameters D _{3 to} D ₄ , that is, the probability ratio 3 and the probability ratio 4 are integrated into one by integrating the predictive variables (parameters). _{Therefore, in FIGS. 22A and 22B, the weighting coefficients α i} and the intercept β (i = 1-3, 5-6) corresponding to the probability ratio 1 to the probability ratio 3 and the probability ratio 5 to the probability ratio 6 are detected for the first time. Store by distance.

The unwanted object removal unit 76A uses the weighting coefficient / intercept table A63g or the weighting coefficient / intercept table B63h read from the weighting coefficient / intercept table A63g or the weighting coefficient / intercept table B63h according to the initial detection distance based on the following equation (13), and uses the above-described embodiment. The corrected probability ratio OverAll is calculated after regression-correcting the probability ratios i (i = 1 to 6) calculated in the same manner as in 1.

When the initial detection distance is the above-mentioned first distance, the unwanted object removal unit 76A has a weight coefficient α i and an intercept β ( _{weight coefficient α i) of each probability ratio i corresponding to the distance at each scan from the weight coefficient / intercept table A63 g.} i = 1-6) is read. On the other hand, when the initial detection distance is other than the above-mentioned first distance, the unnecessary object target removing unit 76A has a weighting coefficient α _i and a weighting coefficient α i of each probability ratio i corresponding to the distance at each scan from the weighting coefficient / intercept table B63h. Read the intercept β (i = 1-6). Weight coefficient when the initial detection distance is the first distance ・ The weight coefficient α _i and intercept β (i = 1 to 6) read from the intercept table A63g are the weight coefficient when the initial detection distance is other than the first distance. _{-Compared with the weighting coefficients α i} and the intercept β (i = 1 to 6) read from the intercept table A63h, the change in the values according to the detection distance of the target is relatively large. Then, the unnecessary object target removing unit 76A is _{corrected based on the weight coefficient α i} of each probability ratio i read out, the intercept β (i = 1 to 6), and each probability ratio i based on the above equation (13). The ratio OverAll is calculated.

In the second embodiment, when calculating the probability ratio OverAll from each parameter (probability ratio 1 to 6) for the probability ratio updated by Bayes for each scan, the weight obtained by logistic regression analysis differs depending on the initial detection distance of the target. Weight by coefficient and intercept. Therefore, in the second embodiment, the accuracy of discriminating between a stationary vehicle and an upper object can be improved.

As described above, the "probability ratio 1" is the log-likelihood of a stationary vehicle and an upper object based on the parameter "lateral position difference". The "probability ratio 2" is the log-likelihood of a stationary vehicle and an upper object based on the parameter "relative velocity difference". The “probability ratio 3” is the log-likelihood of a stationary vehicle and an upper object based on the parameter “independent beam inclination difference”. The "probability ratio 4" is the log-likelihood of a stationary vehicle and an upper object based on the parameter "vertical beam power difference slope". The “probability ratio 5” is the log-likelihood of a stationary vehicle and an upper object based on the parameter “upper and lower beam surface integral”. The “probability ratio 6” is the log-likelihood of a stationary vehicle and an upper object based on the parameter “extrapolation factor type”.

Here, the characteristics of the weighting coefficient / weighting coefficient of the probability ratio of the intercept table A63g shown in FIG. 22A will be described. _{The weight coefficient α 1} having a probability ratio of 1 has a larger value when the distance is relatively large (for example, when it exceeds 80 m) than when the distance is relatively small (for example, 80 m or less). This is important as a parameter for determining whether the target is a stationary vehicle or an upper object because the lateral position difference of the upper object becomes larger when the distance of the target is relatively large than when the distance is relatively small. The degree becomes high. Therefore, when the distance between the targets is relatively large compared to the case where the distance is relatively small _{, the value of the weighting coefficient α 1 with} the probability ratio 1 is set to a large value.

Further, the weight coefficient α ₂ having a probability ratio of 2 has a larger value when the distance is relatively small than when the distance is relatively large. This is because the smaller the distance, the greater the reflected intensity of the reflected wave from the target, the more accurately the relative speed difference of the target can be calculated, and the more important it is as a parameter for distinguishing between a stationary vehicle and an upper object. However, when the distance between the targets becomes smaller (for example, 60 m) or less, the weighting coefficient α ₂ becomes a small value. This is because as the distance becomes smaller, the reflected wave is not received outside the transmission range of the transmitted wave of the radar device 1.

Further, the weighting coefficient α ₅ having a probability ratio of 5 has a larger value when the distance is relatively small than when the distance is relatively large. This is because the smaller the distance, the more the result of the surface integral can be obtained, and the more important it is as a parameter for discriminating between a stationary vehicle and an upper object.
In this way, paying attention to the change in the value of various parameters according to the distance of the stationary vehicle and the change in the value of various parameters according to the distance of the upper object, the weighting coefficient of various parameters is changed according to the distance of the target. By doing so, it has become possible to accurately distinguish between a stationary vehicle and an upper object.

Further, when the distance (first detection distance) at which the radar device 1 first detects the target target is a distance other than the first distance (for example, 80 to 110 m), the weight is weighted by the weighting coefficient / intercept table B63 shown in FIG. 22B. The coefficient α _i is calculated. When the initial detection distance of the target is the first distance, the weighting coefficient α _i is calculated from the weighting coefficient / intercept table A63g shown in FIG. 22A. Thus, in the calculation of the weighting factor alpha _i, when calculating the weight coefficient alpha _i by the weighting factor, slice table B63 shown in FIG. 22B, the distance of the target becomes the specified distance as a first distance, _{The weighting factor α i} is calculated from another table (weighting factor / section table A63g shown in FIG. 22A) having the same type of parameters but different weighting factor values. As a result, the parameters used for determining the type of the target can be optimally weighted according to the distance of the target, and the type of the target can be reliably determined.

[Modifications of Embodiments 1 and 2]
(1) Probability ratio Overall In the first embodiment, the target is determined as a stationary vehicle when the probability ratio Overall (the same applies to the corrected probability ratio Overall in the second embodiment) is equal to or more than the threshold value, and the target is determined when the probability ratio is less than the threshold value. Is determined to be an upper object. However, not limited to this, when determining whether or not the target is a stationary vehicle by comparing the "reliability of the stationary vehicle" with the threshold value, the probability ratio is set as a multiplication factor C for multiplying the "reliability of the stationary vehicle". OverAll may be converted and used. That is, when the "reliability of the stationary vehicle used for the threshold value = C × (reliability of the stationary vehicle)" is equal to or higher than a predetermined threshold value, the target is determined to be a stationary vehicle.

Here, the "reliability of a stationary vehicle" is an index indicating whether or not the target data is data related to a stationary vehicle, for example, taking a value in the range of 0 to 100, and the higher the value, the more the stationary vehicle. Indicates that there is a high probability. The "reliability of a stationary vehicle" is calculated using a plurality of information (for example, "longitudinal distance", "angle power", "extrapolation frequency", etc.) included in the target data.

For example, it is assumed that two threshold values of threshold value 1> threshold value 2 are provided. When the probability ratio Overall ≧ threshold value 1, the magnification C = 1. This indicates that since it can be determined that the "reliability of the stationary vehicle" is high, the "reliability of the stationary vehicle" is used as it is for the threshold value determination as to whether or not the vehicle is a stationary vehicle. When the threshold value 2 ≧ probability ratio Overall, the magnification C = 0. This indicates that since it can be determined that the "reliability of the stationary vehicle" is low, the "reliability of the stationary vehicle" is set to 0 so that the vehicle is not determined to be a stationary vehicle.

When the threshold value 1> the probability ratio Overall> the threshold value 2, the magnification C = (probability ratio Overall-threshold value 2) / (threshold value 1-threshold value 2). That is, the magnification C indicates how much the probability ratio Overall exceeds the threshold value 2 between the threshold value 1 and the threshold value 2. For example, when C = 0.5, whether or not the "reliability of the stationary vehicle used for the threshold value determination" calculated by multiplying the "reliability of the stationary vehicle" by 0.5 is the stationary vehicle. Indicates that it is used for threshold value determination.

In this way, by converting the probability ratio Overall into a magnification C that is multiplied by the "reliability of the stationary vehicle", the judgment line as to whether or not the vehicle is a stationary vehicle is widened, and various factors are taken into consideration to make it more comprehensive. Can identify a stationary vehicle.

(2) Discrimination between stationary vehicle and non-stationary vehicle In the first and second embodiments, various parameters related to the stationary vehicle and the upper object are calculated for each target detection, and the various parameters and the likelihood of the stationary vehicle and the upper object are calculated. Based on the likelihood model in which the correlation is modeled for each detection distance, the Bayesian filter method is used to discriminate between the stationary vehicle and the upper object by updating the log-likelihood ratio of the stationary vehicle and the upper object corresponding to various parameters. However, the present invention is not limited to this, and the same Bayesian filter method may be applied to the determination of a stationary vehicle and a target other than the stationary vehicle (for example, a lower object).

For example, for each target detection, various parameters related to the stationary vehicle and the target other than the stationary vehicle were calculated, and the correlation between the various parameters and the likelihood of the stationary vehicle and the target other than the stationary vehicle was modeled for each detection distance. Build a likelihood model. Then, based on this likelihood model, the log-likelihood ratio of the target other than the stationary vehicle and the stationary vehicle corresponding to various parameters may be updated to discriminate the target other than the stationary vehicle and the stationary vehicle.

[Embodiment 3]
Next, the radar device 1 according to the third embodiment will be described. In the following, the parts different from those of the first embodiment will be mainly described, and the description of the overlapping parts will be omitted.

In the above-described first embodiment, in the unnecessary target removal process (see FIG. 19B), the determination of whether the target is a stationary vehicle or an upper object (hereinafter, may be referred to as “target determination”) is determined by the probability ratio Overall. Is used as an index. The probability ratio Overall is an example of the first index.

It has already been described with reference to FIG. 20 that the above-mentioned value of the probability ratio Overall has an overlapping range between a stationary vehicle and an upper object. Therefore, in the third embodiment, the target determination is performed by further using an index different from the probability ratio Overall.

As a result, in the third embodiment, for example, in the probability ratio Overall, even a target for which it is difficult to clearly determine whether it is a stationary vehicle or an upper object can be accurately determined.

Hereinafter, in detail, in the radar device 1 (see FIG. 1) mounted on the own vehicle A, when the target (target T) is, for example, a stationary vehicle (target T1), as it approaches the stationary vehicle, Since the reflected wave from the stationary vehicle is easily received in front of the receiving unit RX, the angular power, which is the received power of the reflected wave, tends to increase.

On the other hand, in the radar device 1, when the target is, for example, an upper target (target T2), the angular power is attenuated or changed because the target gradually deviates from the main transmission region of the transmitted wave as it approaches the upper target. Tends to occur in a state where

Therefore, in the third embodiment, paying attention to the characteristic that the angular power is easily attenuated when the target is an upward object, the amount of attenuation of the angular power according to the detection distance (vertical distance to the target) of the target is determined. The second index shown was calculated, and the target determination was performed based on the second index. As the second index, for example, the attenuation area A1 (see FIG. 23A and the like) can be used, which will be described later.

Further, in the third embodiment, paying attention to the characteristic that the angular power tends to increase when the target is a stationary vehicle, a third index indicating the amount of increase in the angular power according to the detection distance of the target is calculated. Therefore, the target is determined based on the third index. As the third index, for example, the increased area A2 (see FIG. 24A and the like) described later can be used.

In the following, first, the attenuation amount area A1 which is the second index will be described. The unwanted object removal unit 76 calculates the attenuation area A1 based on the following equation (14).

23A and 23B are diagrams for explaining the calculation method of the attenuation amount area A1 according to the third embodiment. 23A and 23B are graphs in which the horizontal axis is the distance (detection distance) and the vertical axis is the angular power. Further, FIG. 23A is an example of a graph when the target is an upper object, and FIG. 23B is an example of a graph when the target is a stationary vehicle.

When the angular power of the current processing corresponds to (1) in FIG. 23A, for example, the unnecessary target removal unit 76 sets the maximum value of the angular power up to the current processing as the reference power. In the example shown in FIG. 23A, since the angular power Pa (corresponding to (2) in FIG. 23A) of the previous processing is the maximum value, the unnecessary object target removing unit 76 sets such an angular power Pa as a reference power.

Then, as shown in Eq. (14), the unwanted object removal unit 76 is based on the reference power ((2) in FIG. 23A) from the angular power (corresponding to (1) in FIG. 23A) acquired in this process. The value obtained by multiplying the difference obtained by subtracting (corresponding to) by the distance difference between the previous processing distance and the current processing distance (previous current time distance difference) is calculated and accumulated to calculate the attenuation area A1. To do. That is, the unwanted object removal unit 76 calculates the attenuation amount area A1 by using, for example, the segmented quadrature method.

Therefore, as shown in FIG. 23A, when the target is an upper object, the angular power is attenuated as the upper object approaches. Therefore, the attenuation area A1 increases every time the calculation process of the equation (14) is repeated. It becomes. In FIG. 23A, the attenuation amount area A1 when the target approaches the distance Ya is shown surrounded by diagonal lines.

On the other hand, as shown in FIG. 23B, the damping amount area A1 when the target is a stationary vehicle is less likely to increase than in the case of an upper object. Specifically, in the unwanted object removal unit 76, when the angular power of the current processing corresponds to (1) in FIG. 23B, for example, the angular power Pb of the previous processing (corresponding to (2) in FIG. 23B) is the maximum value. Therefore, the angular power Pb is set as the reference power.

Then, the unnecessary object target removing unit 76 subtracts the reference power (corresponding to (2) in FIG. 23B) from the angular power (corresponding to (1) in FIG. 23B) of this processing, and the distance difference between the previous time and this time. The attenuation area A1 is calculated by calculating and accumulating the values obtained by multiplying by.

Here, when the target is a stationary vehicle, the angular power may be temporarily attenuated due to the influence of multipath or the like, but basically increases as the stationary vehicle approaches. Therefore, in the example of FIG. 23B, for example, when the angle power Pc (corresponding to (3) in FIG. 23B) is obtained in this processing, it is larger than the angle power Pb which was the reference power, so that the unnecessary target removal unit 76 Updates the angle power Pc, which is the new maximum value, as the reference power.

As a result, the value of "current angle power (here, angle power Pc) -reference power (here, angle power Pc)" in the equation (14) becomes zero, so that the attenuation area A1 is not accumulated and does not increase.

Further, in the example of FIG. 23B, since the angular power continues to increase until the angular power Pd shown in FIG. 23B is reached, the reference power also continues to be updated, and as a result, the attenuation area A1 does not increase. It will be. Therefore, when the target is a stationary vehicle, in the example of FIG. 23B, the attenuation area A1 slightly increases when the target is temporarily attenuated due to the influence of multipath or the like after the time when the angular power Pd of (4) is reached. Just do it.

Then, the unnecessary object target removing unit 76 compares the attenuation amount area A1 calculated as described above with the threshold value A1a, and when the attenuation amount area A1 becomes the threshold value A1a or more, it is determined that the target is an upper object. judge. Further, when the damping amount area A1 is less than the threshold value A1a, the unnecessary object target removing unit 76 determines that the target is not an upper object, in other words, the possibility of a stationary vehicle is high.

As described above, in the third embodiment, by using the attenuation amount area A1 indicating the attenuation amount of the angular power according to the detection distance of the target, it is determined whether the target is a stationary vehicle or an upper object. It can be done with high accuracy. That is, by judging by the integrated value of the signal power attenuation related to the target in the multiple target detection processes, it is judged by the instantaneous value of the signal power attenuation related to the target in one target detection process. However, it is possible to accurately determine whether the target is a stationary vehicle or an upper object.

Further, in the third embodiment, when the angular power of the current processing is larger than the reference power used in the previous processing, the angular power of the current processing is updated as the reference power.

Thereby, for example, when the target is a stationary vehicle and the angular power tends to continue to increase (see FIG. 23B), it is possible to make it difficult to increase the damping amount area A1. Therefore, it is possible to compare the damping amount area A1 and the threshold value A1a, and if the damping amount area A1 is less than the threshold value A1a, it is possible to determine that the target is likely to be a stationary vehicle.

Next, the increased area A2, which is the third index, will be described. The unwanted object removal unit 76 calculates the increased area A2 based on the following equation (15).

24A and 24B are diagrams for explaining the calculation method of the increased amount area A2 according to the third embodiment. 24A and 24B are graphs in which the horizontal axis is the distance (detection distance) and the vertical axis is the angular power. Further, FIG. 24A is an example of a graph when the target is an upper object, and FIG. 24B is an example of a graph when the target is a stationary vehicle. Note that FIG. 24A shows an example of a state in which the angular power hardly changes when the target is an upper object.

In the following, a method of calculating the increased area A2 will be described with reference to FIG. 24B first. The unnecessary target removal unit 76 sets a reference value based on the angular power when the detection distance of the target is the specified distance Yb or more.

Specifically, the unwanted object removal unit 76 calculates the average value of the angular powers corresponding to the distances from (1) to (2) in FIG. 24B when the target distance is equal to or greater than the specified distance Yb. Then, the unnecessary object target removing unit 76 sets such an average value (here, the angle power Pe) as a reference value. In FIG. 24B and FIG. 24A described later, the reference value is indicated by a white square mark.

The above-mentioned specified distance Yb can be set to an arbitrary value, but may be set to a relatively long value, specifically, a value such that the distance from the radar device 1 to the target is a long distance.

Then, the unnecessary object target removing unit 76 calculates the difference obtained by subtracting the reference value (here, the angular power Pe) from the angular power (for example, corresponding to (3) in FIG. 24B) acquired in this process. The unwanted object removal unit 76 calculates and accumulates a value obtained by multiplying the difference by the distance difference between the distance of the previous processing and the distance of the current processing (the distance difference of the previous time and this time), and accumulates the increased area. Calculate A2.

Therefore, as shown in FIG. 24B, when the target is a stationary vehicle, the angular power increases as the stationary vehicle approaches, so that the increased area A2 increases each time the calculation process of Eq. (15) is repeated. It becomes. In FIG. 24B, the area of increase A2 when the target approaches the distance Yc is shown surrounded by diagonal lines.

On the other hand, as shown in FIG. 24A, the increased amount area A2 when the target is an upper object is less likely to increase than in the case of a stationary vehicle because the angular power hardly changes. Specifically, the unwanted object removal unit 76 is based on the average value (angle power Pe) of the angular power corresponding to the distance between (1) and (2) in FIG. 24A when the target distance is equal to or greater than the specified distance Yb. Set as a value.

Then, the unnecessary object removal unit 76 calculates the difference obtained by subtracting the reference value Pe from the angular power (for example, corresponding to (3) in FIG. 24A) of this processing.

Here, the increased amount area A2 is an index indicating the increased amount of the angular power, as described above. Therefore, when the difference is a negative value, in other words, when the angular power is attenuated, it may not be reflected in the increase amount area A2.

Specifically, when the difference obtained by subtracting the reference value from the angular power of the current processing is a positive value, the unnecessary target removal unit 76 accumulates the value obtained by multiplying the difference by the distance difference last time. , The increased area A2 may be calculated.

Therefore, since the angular power of (3) in FIG. 24A is smaller than the reference value (here, the angular power Pe), the difference becomes a negative value and is therefore not reflected in the increased amount area A2. In the example of FIG. 24A, since the difference obtained by subtracting the reference value is a positive value for the angular power of (4), the unnecessary object target removing unit 76 is a value obtained by multiplying the difference by the distance difference last time. Is accumulated and reflected in the increased amount area A2. As a result, for the increased amount area A2, only the increased amount of the angular power is accumulated, and thus the accuracy of the target determination described later can be improved.

In the example of FIG. 24A, since the angular power hardly changes, the increased area A2 only slightly increases when the angular power temporarily increases due to the influence of multipath and exceeds the reference value. ..

Then, the unnecessary object target removing unit 76 compares the increased amount area A2 calculated as described above with the threshold value A2a, and if the increased amount area A2 is less than the threshold value A2a, determines that the target is an upper object. .. Further, when the increased amount area A2 is equal to or greater than the threshold value A2a, the unnecessary object target removing unit 76 determines that the target is not an upper object, in other words, the possibility of a stationary vehicle is high.

As described above, in the third embodiment, by using the increased amount area A2 indicating the increased amount of the angular power according to the detection distance of the target, it is determined whether the target is a stationary vehicle or an upper object. It can be done with high accuracy.

By using an instantaneous value as the angular power in this processing in the calculation of the attenuation amount area A1 and the increase amount area A2, the responsiveness of the target determination can be improved. The instantaneous value of the angular power may vary, but since the attenuation area A1 and the increase area A2 are cumulative values, it is possible to absorb such variations.

(Removal of unnecessary objects according to the third embodiment)
FIG. 25 is a flowchart showing a subroutine for removing unnecessary objects according to the third embodiment. Note that FIG. 25 corresponds to the flowchart of FIG. 19B. Further, in FIG. 25, the processes for calculating the probability ratios 1 to 6 from step S18-1 to step S18-12 in FIG. 19B are collectively shown as step S18-A.

As shown in FIG. 25, the unwanted object removal unit 76 calculates the probability ratio Overall All based on the probability ratios 1 to 6 calculated in steps S18-A (step S18-13). Next, the unwanted object removal unit 76 calculates the attenuation amount area A1 and the increase amount area A2 (step S18-13a).

Next, the unwanted object removal unit 76 determines whether or not the calculated attenuation amount area A1 is equal to or greater than the threshold value A1a (step S18-13b). When it is determined that the attenuation area A1 is equal to or greater than the threshold value A1a (step S18-13b: Yes), the unwanted object removal unit 76 determines that the target is an upper object (step S18-16).

When the attenuation area A1 is less than the threshold value A1a (step S18-13b: No), the unnecessary target removal unit 76 estimates that the target is likely to be a stationary vehicle, and subsequently the increase area A2 is the threshold value A2a. It is determined whether or not it is less than (step S18-13c).

When it is determined that the increased amount area A2 is less than the threshold value A2a (step S18-13c: Yes), the unwanted object removal unit 76 determines that the target is an upper object (step S18-16).

When it is determined that the increased amount area A2 is equal to or greater than the threshold value A2a (step S18-13c: No), the unwanted object removal unit 76 estimates that the target is likely to be a stationary vehicle, and subsequently steps S18-. It is determined whether or not the probability ratio Overall all calculated in 13 is equal to or greater than the threshold value (step S18-14).

When it is determined that the probability ratio Overall is equal to or greater than the threshold value (step S18-14: Yes), the unnecessary target removal unit 76 determines that the target is a stationary vehicle (step S18-15). On the other hand, when it is determined that the probability ratio Overall is less than the threshold value (step S18-14: No), the unnecessary object target removing unit 76 determines that the target is an upper object (step S18-16).

As described above, in the third embodiment, when the attenuation amount area A1 and the increase amount area A2 are used for the target determination, it is shown that the target is clearly above the detected angular power. , The target can be determined to be an upper object regardless of the probability ratio Overall.

Therefore, in the third embodiment, for example, in the probability ratio Overall, even if it is difficult to clearly determine whether the target is a stationary vehicle or an upper object, the target can be accurately determined.

Further, in the third embodiment, the target determination is performed using both the attenuation amount area A1 and the increase amount area A2. As a result, the target determination can be reliably performed.

That is, the value of the angular power may fluctuate drastically due to the influence of, for example, multipath. Therefore, even if the target is an upper object, depending on the value of the angular power, for example, while the attenuation amount area A1 and the increase amount area A2, a value indicating that the target is an upper object may not be obtained.

Even in such a case, if the other of the attenuation amount area A1 and the increase amount area A2 has a value indicating that it is an upper object, it is possible to determine the target as an upper object. That is, in the third embodiment, the influence of the multipath can be reduced, and thus the target determination can be reliably performed.

Specifically, even when the target is an upper object, if the height of the upper object from the road surface is less than a predetermined value, the damping amount area A1 may be less than the threshold value A1a. In such a case, the value of the increased amount area A2 of the upper object whose height from the road surface is less than a predetermined value becomes less than the threshold value A2a. This is because the upper object is easily affected by multipath because the height from the road surface is higher than that of the stationary vehicle, and the increased area A2 is smaller than that of the stationary vehicle. In this way, when the target is an upper object, even if the value of the damping amount area A1 shows a tendency of a stationary vehicle, the type of the target is accurately determined by making a judgment using the value of the increased amount area A2. can do.

Here, if it is necessary to accurately determine that the target is an upper object, it is conceivable to determine only by the value of the increased amount area A2. However, even when the target is a stationary vehicle, when the height (vehicle height) of the vehicle body from the road surface of the vehicle (for example, a truck or the like) is equal to or higher than a predetermined value, it is easily affected by multipath. Therefore, not only the determination using the increased amount area A2 in which the characteristic appears prominently when the target is a stationary vehicle, but also the determination using the damping amount area A1 in which the characteristic appears prominently when the target is an upper object is performed. I have to. As a result, a target that satisfies all the conditions for a stationary vehicle is determined to be a stationary vehicle according to a plurality of determination criteria.

[Modified Example of Embodiment 3]
Next, a modified example of the third embodiment will be described. In the modified example of the third embodiment, the damping amount area A1 and the increased amount area A2 are converted and used as a magnification C to be multiplied by the "reliability of the stationary vehicle" described in the modified examples of the first and second embodiments. You may.

Specifically, for example, when the attenuation area A1 is equal to or greater than the threshold value A1a, the target can be determined to be an upper object, so the magnification C may be set to 0. As a result, the "reliability of the stationary vehicle" becomes 0, and it is possible to prevent the target from being determined to be a stationary vehicle.

Similarly, for example, when the increased area A2 is less than the threshold value A2a, the target can be determined to be an upper object. Therefore, the magnification C is set to 0 and the "reliability of the stationary vehicle" is set to 0, and thus the target is determined to be a stationary vehicle. It can be prevented from being done.

In the above, the attenuation amount area A1 and the increase amount area A2 are converted to the magnification C, but the present invention is not limited to this, and is added or subtracted with respect to, for example, the "reliability of a stationary vehicle". You may want to convert it to a value.

In the above-described third embodiment, the calculation process of the attenuation amount area A1 and the increase amount area A2 and the process of comparing the attenuation amount area A1 and the increase amount area A2 with the threshold values A1a and A2a are performed in the calculation process of the probability ratio OverAll. It was configured to be done later, but it is not limited to this. That is, the calculation processing of the attenuation amount area A1 and the increase amount area A2 may be performed prior to or at the same time as the calculation process of the probability ratio Overall.

Further, in the third embodiment, the target determination is performed using both the attenuation amount area A1 and the increase amount area A2, which are the second and third indexes, but the target determination is not limited to this, for example. It may be configured to use either of the second and third indicators.

The peak extraction unit 70, the angle estimation unit 71, and the pairing unit 72 described above are examples of the derivation unit. Further, the unnecessary object removal unit 76 is an example of a calculation unit and a determination unit. Further, a stationary vehicle is an example of a target that collides (for example, a target that requires vehicle control such as brake control), and an upper object is a target that does not collide (for example, a target that does not require vehicle control such as brake control). This is an example.

Of each of the processes described in the embodiments, all or part of the processes described as being automatically performed can also be performed manually. Alternatively, of the respective processes described in the embodiments, all or part of the processes described as being manually performed can be automatically performed by a known method.

Further, the integration and distribution of each part described in the embodiment can be appropriately changed based on the processing load and the processing efficiency. In addition, the above-mentioned and illustrated processing procedures, control procedures, specific names, and information including various data and parameters can be appropriately changed unless otherwise specified.

Further effects and variations can be easily derived by those skilled in the art. For this reason, the broader aspects of the disclosed art are not limited to the particular details and representative embodiments expressed and described as described above. Therefore, various modifications can be made without departing from the spirit or scope of the general concept of the invention as defined by the appended claims and their equivalents.

A Own vehicle TX transmitter TX1 Lower transmitter TX2 Upper transmitter RX receiver 1, 1A Radar device 2 Vehicle control device 3 Brake 4 Transmitter 41 Signal generator 42 Oscillator 43 Switch 5 Receiver 52 Individual receiver 53 Mixer 54 A / D Converter 6, 6A Signal processing unit 61 Transmission control unit 62 Fourier transform unit 63, 63A Storage unit 63a Horizontal position difference model 63b Relative velocity difference model 63c Independent beam tilt difference model 63d Vertical beam power difference tilt model 63e Vertical beam area Differential model 63f External factor type likelihood table 63g Weight coefficient / intercept table A
63h Weighting factor / intercept table B
7, 7A Data processing unit 70 Peak extraction unit 71 Angle estimation unit 72 Pairing unit 73 Continuity determination unit 74 Filtering unit 76, 76A Unnecessary target removal unit 77 Grouping unit 78 Target information output unit

Claims (9)

A plurality of parameters related to the target and the detection distance of the target based on the received signal obtained by receiving the reflected wave reflected by the target existing in the vicinity of the radar transmission wave transmitted to the vicinity of the own vehicle. And the derivation part that derives
Each parameter, a target existing in the traveling direction of the own vehicle, the likelihood of the target colliding when the own vehicle advances in the traveling direction, and the target existing in the traveling direction of the own vehicle. Each derived unit is derived from a likelihood model that models a known correlation with the likelihood of a target that does not collide when the own vehicle moves in the traveling direction for each detection distance. A calculation unit that calculates a first index based on the likelihood ratios of the colliding target and the non-collising target corresponding to the parameters and the detection distance.
A radar including a determination unit for determining whether the target is a colliding target or a non-collision target based on the first index calculated by the calculation unit. apparatus. The plurality of parameters are parameters calculated for each acquisition of the received signal.
The first parameter regarding the lateral position of the target with respect to the traveling direction of the own vehicle, and
A second parameter relating to the relative speed of the target with respect to the own vehicle,
A third parameter relating to the received power of the reflected wave and
A fourth parameter relating to the reception power difference of each reflected wave corresponding to the first transmission wave and the second transmission wave having different transmission directions among the radar transmission waves, and
A fifth parameter relating to the received power difference leveled for each detection distance, and
The radar device according to claim 1, further comprising a sixth parameter relating to normal detection and abnormality detection factor of the target based on the received signal. The calculation unit
When the initial detection distance of the target is a predetermined distance, the first index is calculated based on each likelihood ratio corrected by using the first coefficient that linearly regresses each of the plurality of parameters. When the initial detection distance of the target is other than the predetermined distance, the first index is calculated based on each likelihood ratio corrected by using a second coefficient that linearly regresses each of the plurality of parameters. And
The first coefficient is
The radar device according to claim 1 or 2, wherein the change in the value according to the detection distance of the target is relatively large as compared with the second coefficient. The calculation unit
A second index showing the amount of attenuation of the received power of the reflected wave according to the detection distance of the target, and a third index showing the amount of increase in the received power of the reflected wave according to the detection distance of the target. And calculate
The determination unit
Based on at least one of the second index and the third index calculated by the calculation unit, it is determined whether the target is the colliding target or the non-colliding target. The radar device according to any one of claims 1 to 3. The calculation unit
The difference between the detection distance of the previous processing and the detection distance of this processing is the difference obtained by subtracting the reference power, which is the maximum value of the reception power of the reflected wave up to this processing, from the reception power of the reflected wave acquired in this processing. The radar device according to claim 4, wherein the second index is calculated by accumulating the values obtained by multiplying the detection distance differences. The calculation unit
When the received power of the reflected wave acquired in the present process is larger than the reference power used in the previous process, the received power of the reflected wave acquired in the present process is updated as the reference power. The radar device according to claim 5. The calculation unit
A reference value is set based on the received power of the reflected wave when the detection distance of the target is equal to or longer than the specified distance, and the difference obtained by subtracting the reference value from the received power of the reflected wave acquired in this process is used. Any of claims 4 to 6, wherein the third index is calculated by accumulating the value obtained by multiplying the detection distance difference between the detection distance of the previous processing and the detection distance of the current processing. The radar device described in one. The calculation unit
When the difference obtained by subtracting the reference value from the received power of the reflected wave acquired in this process is a positive value, the third value is accumulated by multiplying the difference by the detection distance difference. The radar device according to claim 7, wherein the index is calculated. It is a control method of the radar device executed by the control device of the radar device.
A plurality of parameters related to the target and the detection distance of the target based on the received signal obtained by receiving the reflected wave reflected by the target existing in the vicinity of the radar transmission wave transmitted to the vicinity of the own vehicle. And the derivation step to derive
Each parameter, a target existing in the traveling direction of the own vehicle, the likelihood of the target colliding when the own vehicle advances in the traveling direction, and the target existing in the traveling direction of the own vehicle. Each of the likelihood models derived by the derivation step from a likelihood model that models a known correlation with the likelihood of a target that is a target and does not collide when the own vehicle advances in the traveling direction for each detection distance. A calculation step of calculating a first index based on each likelihood ratio of the colliding target and the non-collising target corresponding to the parameters and the detection distance, and
A radar characterized by including a determination step of determining whether the target is a colliding target or a non-collision target based on the first index calculated by the calculation step. How to control the device.

Images