JP2014006122A - Object detector - Google Patents

Abstract

PROBLEM TO BE SOLVED: To improve accuracy in tracking a target in a case with a large difference between a detection point and a tracking point of a tracking filter.SOLUTION: An object detector 100 for detecting an object in the circumference of a vehicle by updating a tracking point of an object includes: position detection means 11 for detecting an object position; position prediction means for predicting the predicted position of an object on the basis of a tracking point and an object movement rule (e.g. S300, a formula (1)); tracking point updating means for updating a tracking point on the basis of an object position detected by the position detection means and a predicted position predicted by the position prediction means (e.g. S500, a formula (5)); and correction means for correcting an object position detected by the position detection means with the use of distance information of an object detected by the position detection means when an object detected by the position detection means exists within a first predetermined range from a tracking point (e.g. S504, a dummy detection point generation unit 22). The tracking point updating means updates a tracking point on the basis of an object position corrected by the correction means and a predicted position predicted by the position prediction means.

Description

  The present invention relates to an object detection device that detects the position of an object around a vehicle.

  There is known an object detection device that informs the driver of the presence of other features such as a rear side traveling in an adjacent traveling lane or a parallel vehicle by monitoring a blind spot from the driver's seat with a radar. . For example, when the driver turns on the blinker switch, when the object detection device detects that there is an obstacle on the rear side, the driver alerts the driver by sounding an alarm sound.

  The position (detection point) of the target detected by the radar is known to vary slightly. Conventionally, the detection point is input to the tracking filter to generate the tracking point, and the target is tracked by the tracking point. And the alarm etc. are emitted based on this tracking result (for example, refer to patent documents 1). In Patent Document 1, a fixed position P1, which is the position of the observation target object at the current sampling time, is obtained from the position R1 indicating the observation result at the current sampling time and the estimated position P0 predicted based on the information on the previous sampling time. A vehicle-mounted periphery monitoring device for calculation is disclosed.

Japanese Patent Laid-Open No. 2003-217099

  However, when a plurality of targets are present within the radar detection range (for example, when they are close to each other), interference occurs due to reflected waves from the plurality of targets, and the variation in detection point position increases. As a result, there are no detection points in the vicinity of the tracking point, and the detection point for the tracking filter to update the tracking point may not be obtained.

  FIG. 14A is an example of a diagram illustrating a detection result of a radar detection point. The radar position (vehicle position) is shown at the origin of the coordinates, and the relative position of the detection point is plotted with “x”. The width direction of the vehicle amount is the X axis (left direction is negative, right direction is positive), and the direction perpendicular to the axle is the Y axis (forward is positive and rear is negative).

  FIG. 14A shows the detection result of the detection point in a situation where the bicycle is traveling inside the lane (in the lane). Since the detection result in FIG. 14A includes the passage of time, the bicycle is moving. When there are a plurality of targets within the radar detection range, detection points vary in the circumferential direction around the radar position. It can be seen that the detection point may vary up to just behind the host vehicle.

FIG. 14B is an example of a diagram schematically showing variation in detection points. The tracking points of black circles are connected to each other by a straight line, and detection points are detected around the tracking points. In the figure, two tracking points are tracked. The tracking point closer to the radar position is newer in time and the two targets are closer to each other.
When a detection point is detected within a predetermined distance from the tracking point, the tracking point at the next timing can be determined using the detection point. However, since the detection point greatly varies in the circumferential direction of the circle centered on the radar position due to interference of the reflected wave of the radar, the detection point is not detected within a predetermined distance from the last tracking point. There may be one detection point.
When the detection point no longer exists near the tracking point due to variations in the detection point, the object detection device can calculate the estimated position using the motion model of the target. However, since the target movement obtained from the motion model may deviate from the actual movement of the target, the detection accuracy of the target position may be reduced.

  In view of the above problems, an object of the present invention is to provide an object detection device capable of improving target tracking accuracy when a difference between a detection point and a tracking filter tracking point is large.

  The present invention is an object detection device that detects an object around a vehicle by updating the tracking point of the object, and includes position detection means (for example, a detection unit 11) that detects the position of the object, and the tracking point and the object. Based on the position predicting means (for example, the equation (1)) for predicting the predicted position of the object from the movement rule, the position of the object detected by the position detecting means, and the predicted position predicted by the position predicting means, A tracking point update unit (e.g., Equation (5)) for updating the tracking point and the position detection unit detects when the object detected by the position detection unit is within a first predetermined range from the tracking point. Correction means (for example, dummy detection point creation unit 22) for correcting the position of the object detected by the position detection means using the object distance information, and the tracking point update means includes: Corrected object position, and Based on the predicted position of the position prediction means predicts and updates the tracking point, and wherein the.

  When the difference between the detection point and the tracking filter tracking point is large, it is possible to provide an object detection device capable of improving the target tracking accuracy.

It is an example of the figure explaining the characteristic characteristic of the object detection apparatus of this embodiment. It is a figure which shows the structural example of the vehicle-mounted system containing a radar apparatus. It is an example of the schematic block diagram of a radar apparatus. It is an example of the functional block diagram of the object detection apparatus of this embodiment. It is an example of the figure which illustrates typically the frequency of a transmission signal, a reception signal, and a beat signal. It is an example of the figure explaining determination of an azimuth | direction. It is an example of the figure explaining the electric power of the beat signal obtained by Fourier transform. It is an example of the flowchart figure which shows the procedure in which a tracking part tracks a tracking point by a Kalman filter (prior art). It is an example of the flowchart figure which shows the procedure in which a tracking part tracks a tracking point. It is an example of the figure which illustrates typically tracking by a dummy detection point. It is an example of the figure which shows the list of tracking points typically. It is an example of the flowchart figure which shows the procedure in which a tracking part tracks a tracking point. It is an example of the figure which illustrates typically tracking by a dummy detection point. It is an example of the figure which shows the detection result of the detection point of a radar.

Hereinafter, embodiments for carrying out the present invention will be described with reference to the drawings.
FIG. 1 is an example of a diagram illustrating a characteristic outline of the object detection device of the present embodiment. As shown in FIG. 1 (a), the object detection apparatus receives and analyzes the reflected wave obtained by the radar reflecting on the target, and periodically measures the position of the detection point (triangle in the figure). Further, every time the detection point is measured, for example, the tracking point (circle in the figure) is calculated by applying the detection point to the Kalman filter.

  If the detection point is not found within the predetermined range of the tracking point (dotted circle in the figure) due to the presence of multiple targets in the radar detection range, the object detection device of this embodiment is a dummy A detection point (hereinafter referred to as a dummy detection point) is created, and the calculation of the tracking point is continued by applying the dummy detection point to the Kalman filter.

  FIG. 1B is an example of a diagram illustrating a method for creating a dummy detection point. In the present embodiment, when the dummy detection points are created, it is considered that the detection points vary in the circumferential direction with respect to the radar position. The variation in the circumferential direction means that the displacement of the x coordinate is large, but the y coordinate relatively reflects the position of the target. Accordingly, the x coordinate (= X ′) of the dummy detection point is obtained using the past x coordinate of the tracking point, and the y coordinate (= Y ′) of the dummy detection point is obtained using X ′ and the distance R.

The dummy detection point A ′ (X ′, Y ′ ′) of the tracking point A is obtained as follows, for example.
X ′: average of past X values of tracking points and the latest value Y ′: the intersection of the circle of distance R from the radar position to the detection point and the straight line of x coordinate = X ′ is Y ′ .

Y ′ = √ (R ² −X ′ ² )
That is, X ′ is determined first, and Y ′ that is the Y coordinate is obtained from the distance R to the detection point. In this way, the x-coordinate having a large variation among the detection points can be estimated from the tracking point, and the y-coordinate can be obtained using the points on the circumference where the detection points are distributed. Since the tracking detection point is calculated by applying the dummy detection point estimated in this way to the Kalman filter, the accuracy of the tracking point can be maintained rather than using the motion model.

[Configuration example]
FIG. 2 is a diagram illustrating a configuration example of an in-vehicle system 300 including a radar device. The radar apparatus 100 corresponds to the object detection apparatus recited in the claims. The radar device 100 and a driving support ECU (Electronic Control Unit) 200 are connected via an in-vehicle network such as a CAN (Controller Area Network). As described later, the radar apparatus 100 periodically calculates the distance to the target (the target may be referred to as a target), the relative speed, and the direction (hereinafter, these may be collectively referred to as target information). It is transmitted to the driving support ECU 200. The relative position of the target with respect to the host vehicle is obtained from the direction and the distance. In addition, since the target can be a target of target information as long as it reflects the radar, there may be a planar object such as a manhole, a road surface, or a wall in addition to a three-dimensional object.

  The radio wave transmission / reception unit of the radar apparatus 100 according to the present embodiment is disposed inside the left end corner and the right end corner of the rear bumper of the vehicle made of a material that allows radio waves to pass, such as resin. The center of the radar transmission direction is arranged to be about 45 to 70 degrees with respect to the direction parallel to the axle. The irradiation angle (radar detection range) can be designed to be 90 to 120 degrees, for example. The elevation angle is almost zero (parallel to the road surface).

  In addition to the radar device 100 that detects a parallel running vehicle, the radar device 100 may be attached in front of the host vehicle. The tracking method of this embodiment can be applied regardless of the object detection range (radar irradiation range of the radar apparatus).

  The driving assistance ECU 200 provides various driving assistances such as BSM (Blind Spot Monitoring-System) and CTW (Cross Traffic Warning) based on the target information. A blinker switch 201, a wheel speed sensor 202, a steering angle sensor 203, and an operation device 204 are connected to the driving support ECU 200. In addition, it has a general sensor mounted on the vehicle such as a yaw rate sensor. The blinker switch 201 detects the operation direction of the blinker lever. The wheel speed sensor 202 detects the rotation of the rotor arranged on each wheel as a wheel pulse by a sensor on the vehicle body side taking out from a magnetic flux change or the like. The rotational speed is obtained from the number of wheel pulses per unit time, and the vehicle speed can be obtained by further considering the tire diameter. The steering angle sensor 203 is a sensor that detects the rotation angle of the steering shaft. There are various detection principles. For example, S pole and N pole magnetic bodies are arranged on the steering shaft side, the periphery of the steering shaft is enclosed in a ring shape, and a change in magnetism is detected on the ring side. Thus, the rotation angle is detected.

  The operation device 204 is various in-vehicle devices used for driving support for avoiding abnormal approach to the target. Examples of the operation device 204 include an alarm buzzer, an audio output device, and a steering motor. The alarm buzzer warns that changing the lane by blowing the buzzer on the meter panel may cause abnormal approach to the target, and the audio output device will give a message (for example, “There is another vehicle on the rear side”) ) Is output from the speaker. When a driver steers the steering wheel when there is another vehicle on the rear side, the steering motor gives a rotational torque to the steering shaft and generates a reaction force in the direction opposite to the steering direction. Since the driver requires a larger steering force than before, the driver can be alerted to the presence of the parallel running vehicle.

  FIG. 3 shows an example of a schematic configuration diagram of the radar apparatus 100. The radar apparatus 100 is configured to include a VCO (voltage controlled oscillator) 16, an ASIC 17, and a microcomputer 18 on a substrate 11 connected to a plurality of antennas. The configuration diagram shown is an example, and the function of the ASIC 17 may be implemented by the microcomputer 18 or the function of the microcomputer 18 may be implemented by the ASIC 17. Moreover, you may provide DSP other than showing in figure.

  The antenna includes a transmitting antenna 12 and a plurality of receiving antennas 14-1 to 14-n. The transmission antenna 12 transmits a radar wave modulated to a predetermined frequency by the transmission circuit 13. As will be described later, the frequency of the transmission wave is controlled so as to increase or decrease like a triangular wave, and the beat frequency is measured in each of the rising and falling intervals. Such a radar is called FMCW (Frequency-modulated continuous-wave), but in this embodiment, a pulse radar that switches the irradiation direction of the radar every short time may be used.

  The transmission circuit 13 converts, for example, a high-frequency signal in the millimeter wave band generated by the VCO 16 into a voltage and supplies the voltage to the transmission antenna 12 as a transmission signal. The transmission circuit 13 has a distribution circuit that distributes the transmission signal (high-frequency signal) generated by the VCO 16 to the reception circuits 15-1 to 15-n of the n reception antennas 14-1 to 14-n. .

  Although the number of receiving antennas is n in the figure, it is sufficient to have at least two receiving antennas in order to detect the azimuth. The receiving antennas 14-1 to 14-n have the same configuration including the receiving circuits 15-1 to 15-n. The receiving antennas 14-1 to 14-n receive the reflected waves reflected by the target and output them to the receiving circuits 15-1 to 15-n. The reception circuits 15-1 to 15-n mix and amplify the reception signals acquired from the reception antennas 14-1 to 14-n and the transmission signals acquired from the transmission circuit 13 with a mixer, and then remove unnecessary frequency components. Remove and output to the substrate 11 side. The mixed signal is a beat signal generated by the difference in frequency between the transmission signal and the reception signal. The frequency of the beat signal is called the beat frequency.

  Each of the receiving circuits 15-1 to 15-n generates a beat signal and sends it to the substrate 11 side. Each of the receiving antennas 14-1 to 14-n and the receiving circuits 15-1 to 15-n may be referred to as a channel. Although not shown in the figure, each receiving circuit 15-1 to 15-n and the substrate 11 are connected via a switch, and at the same time, one receiving circuit 15-1 to 15-n is connected to a beat signal. Is supposed to generate.

  The VCO 16 on the substrate 11 side is controlled by the microcomputer 18 and the high frequency signal is modulated by a triangular wave having an ascending section and a descending section. Further, the ASIC 17 is provided at the subsequent stage of the A / D conversion circuits 21-1 to 21-n and the A / D conversion circuits 21-1 to 21-n provided corresponding to the receiving circuits 15-1 to 15-n. The provided buffers 22-1 to 22-n and the FFT processing unit 23 are provided.

  The A / D conversion circuits 21-1 to 21-n convert the beat signals into digital data and store them in the subsequent buffers 22-1 to 22-n for each channel. Each of the buffers 22-1 to 22-n is divided into an ascending section buffer and a descending section buffer. Beat signals received by the receiving antennas 14-1 to 14-n in the rising section are stored in the rising section buffer, and beat signals received by the receiving antennas 14-1 to 14-n in the falling section are stored in the falling section buffer. .

  The FFT processing unit 23 performs FFT processing on the beat signal for each channel. The beat signal is converted into data of a relationship between frequency and power by FFT processing. The radio waves received by the receiving antennas 14-1 to 14-n contain components other than the reflected wave, but the frequency at which the power is peak can be estimated as the beat frequency by FFT processing.

  The microcomputer 18 is an information processing apparatus including a CPU, a ROM, a RAM, an input / output interface, and other general circuits. The microcomputer 18 performs a characteristic process of this embodiment described later.

  FIG. 4 shows an example of a functional block diagram of the object detection apparatus of the present embodiment. The object detection apparatus includes a detection unit 11, a tracking unit 12, and an alarm unit 13. The detection unit 11 corresponds mainly to the antenna to the ASIC in FIG. 3, the tracking unit 12 corresponds mainly to the microcomputer, and the alarm unit 13 corresponds to the operation device of FIG.

  The detection unit 11 performs FFT (Fast Fourier Transform) on the beat signal to calculate the distance to the target, the relative speed, and the direction. For example, the tracking unit 12 tracks a detection point using a Kalman filter, and estimates the target position with high accuracy. The tracking unit 12 includes a list creation unit 21 and a dummy detection point creation unit 22. These will be described later. The warning unit 13 alerts the driver of the presence of a rear side vehicle or a parallel armored vehicle. Hereinafter, the detection unit 11, the tracking unit 12, and the alarm unit 13 will be described in order.

<Calculation of distance, relative speed, direction by detector>
The radar apparatus 100 calculates the distance and the relative speed based on the processing result processed by the FFT processing unit 23, and calculates the azimuth based on the phase of the beat signal having the peak frequency obtained by the FFT processing.

  FIG. 5 is an example of a diagram schematically illustrating frequencies of a transmission signal, a reception signal, and a beat signal. FIG. 5A shows a case where the relative speed of the host vehicle and the target is zero, and FIG. 5B shows a case where the speeds of the host vehicle and the target are different from each other.

The transmission signal S repeatedly increases and decreases in frequency. A frequency variation amount is ΔF, a center frequency is f ₀ , and 1 / fm is a repetition period of an ascending section and a descending section. The frequency increases in the rising section and decreases in the falling section. In addition, until the receiving antennas 14-1 to 14-n receive the reception signal R after the transmission antenna 12 transmits the transmission signal S, the radio wave reciprocates at the speed of light C through the distance to the target. Time is needed. Therefore, when the relative speed is zero, there is a difference between the frequency of the transmission signal and the reception signal according to the distance to the target and fm. When this difference is the beat frequency fb and the relative speed is zero, the beat frequency fb1 in the rising section and the beat frequency fb2 in the falling section are equal.

  When the relative speed is not zero, the frequency of the transmission signal S is Doppler shifted according to the relative speed when reflected by the target, so the frequency of the reception signal R changes more than the change due to the distance to the target and fm. (Or change is reduced). Let the Doppler frequency be fd. When the host vehicle is approaching the target, the frequency of the received signal is shifted (increased) by the Doppler frequency fd in the ascending section, so that the difference between the frequency of the receiving signal and the transmitting signal is reduced, and in the descending section, the reception signal is received. The difference in frequency between the signal and the transmission signal becomes large. Therefore, the beat frequency fb1 in the rising section and the beat frequency fb2 in the falling section are not equal.

If the beat frequency when the relative speed is zero is fr, the beat frequencies fb1 and fb2 can be expressed as follows.
fb1 = fr−fd
fb2 = fr + fd
When this is transformed, the following equation is obtained.
fr = (fb1 + fb2) / 2
fd = (fb2-fb1) / 2
The distance R and the relative speed V of the target can be obtained from the following expressions.
R = (C / (4 · ΔF · fm)) · fr
V = (C / (2 · f ₀ )) · fd
FIG. 6 is an example of a diagram illustrating the determination of the azimuth. FIG. 6A is a diagram schematically illustrating a received signal when the target is present in front of the radar apparatus 100. When the target is present in front, there is almost no path difference between the target and the two receiving antennas 14-1 and 14-2, so that the phases of the received signals of the receiving antenna 14-1 and the receiving antenna 14-2 are the same. It becomes a phase. Note that the radar apparatus 100 processes the beat signal, but if the received signal has the same phase, the beat signal also has the same phase.

FIG. 6B shows a path difference in the case where the target is present at an angle θ with respect to the front surface of the radar apparatus 100. A path difference x occurs between the target and the receiving antennas 1 and 2. When the distance between the antennas is d, the path difference x is d · sin θ. There is the following relationship between the phase difference Δφ and the direction θ. λ is the wavelength of the received signal.
Δφ = 2π × (d · sinθ / λ)
Therefore, if the phase difference Δφ is obtained, since λ and d are fixed values, the azimuth θ can be obtained from the following equation.
θ = arcsin (λ · Δφ / (2 · π · d))
The phase difference Δφ is obtained from the result of Fourier transform performed by the FFT processing unit 23 for each reception channel.

  As shown in FIG. 7, a power peak appears at a frequency with many frequency components by performing Fourier transform for each rising and falling interval of each reception channel. The frequency at which this peak is obtained is the beat frequency. As described above, the distance R and the relative speed V are obtained from the beat frequency fb1 in the rising section and the beat frequency fb2 in the falling section.

Further, the frequency function of the beat frequency is obtained as a complex number including a real number and an imaginary number by Fourier transform. Since the amplitude of the complex number (Z = a + ib) is √ (a ² + b ² ) and the phase is arctan (b / a), the phase of the beat frequency can be calculated for each receiving antenna. The difference between the phase φ of the reception channel 14-1 and the phase φ of the reception channel 14-2 is the phase difference Δφ.

  This method of obtaining the orientation is called a monopulse method. In this embodiment, the determination of the azimuth by the monopulse method has been described. However, the method of obtaining the azimuth is known such as DBF (Digital Beam Forming) processing, MUSIC (Multiple Signal Classification) analysis, Capon analysis, etc., and is limited to the monopulse method. is not.

<Estimation of target position by tracking unit>
In this embodiment, as an example, the target position is tracked by a Kalman filter. In order to indicate the position by appropriate two-dimensional coordinates, the tracking unit 12 represents the position of the detection point by (x, y) from the distance R and the direction θ. The Kalman filter is a technique for obtaining the most probable estimated value of the state variable x (vector) for optimal control. In the Kalman filter, the state of the system is described by the equation of state (1) and the observation equation (2).

x is a vector of state variables, and in this embodiment, the position x in the x direction and the velocity v _{x in the} x direction (first derivative of x), and the position y in the y direction and the velocity v _{y in the} y direction (first order of y). It has four elements (differentiation).

z is a vector indicating the observed value, and in this embodiment, the position (x, y) obtained from the distance R and the direction is used as an element. w _t is process noise, and v _t is observation noise. In the Kalman filter, it is assumed that the noise follows a Gaussian distribution, the average value of the noise is zero, the standard deviation of the process noise is S, and the standard deviation of the observed noise is Q.

  F is a matrix relating the state of the system at time t and the state of the system at time t + 1. H is a matrix that links the state variable x and the observation value z.

Assuming that the motion model of the target within a minute time is constant velocity linear motion, F and H can be obtained as follows. tau is the sampling time of the detection point, the velocity v _x, v advanced in _y position is the position at time t + 1 between tau (x, y). Note that F and H are examples, and can be appropriately designed, for example, considering the speed with H.

In the Kalman filter, there are two state estimates. One is an estimated value called prior estimation

And the state at time t estimated before the observation value at time t is obtained (hereinafter referred to as pre-estimation or pre-estimation value). The other is post hoc estimation

And the state at time t estimated from the observed value at time t (hereinafter referred to as post-estimation or post-estimation value).

  The covariance matrix of the error between the prior estimation value and the observation value is defined as Pt, and the covariance matrix of the error between the posterior estimation value and the observation value is defined as P′t.

  The pre-estimation value and the post-estimation value are considered to be not significantly different, assuming that the error contained in the observation value is not so large. Therefore, in the Kalman filter, using a coefficient (matrix) called Kalman gain, an equation for estimating the posterior estimation value from the pre-estimation value and the observed value is described as follows.

Equation (5) shows that the difference between the posterior estimation value and the prior estimation value is equal to the difference between the value estimated by H from the prior estimation value and the observed value z _t multiplied by the Kalman gain K. Means. Therefore, it means that the value of the posterior estimation can be corrected by adding the second term on the right side to the value of the prior estimation of the first term. The Kalman gain K is a correction coefficient of the second term on the right side, and determines how much the value of the posterior estimation is corrected.

  The Kalman gain K is obtained so that the covariance matrix P′t of the error between the posterior estimation value and the observation value is minimized. Although the process is omitted, the Kalman gain is obtained by this, and the covariance matrix P′t is obtained as follows. I is a unit matrix.

FIG. 8 is an example of a flowchart illustrating a procedure in which the tracking unit 12 tracks a tracking point using a Kalman filter (prior art). The process of FIG. 8 is repeatedly performed for each new target detected while each target is detected. Thereby, the target is tracked.

  The tracking unit 12 is given in advance the initial value of the pre-estimation value and the initial value of the co-variance matrix Pt of the pre-estimation error (S100). Any initial value may be zero.

  The tracking unit 12 first calculates the Kalman gain K using Expression (6) (S200).

  Next, the tracking unit 12 updates the pre-estimated value (S300). As a result, the value of the post-estimation can be calculated in the next step S500. In addition, since the value of the posterior estimation is not obtained in the initial state, an observation value (constant) of an assumed target such as a guardrail or a bicycle is given.

Next, when the observation value z _t is detected (Yes in S400), the tracking unit 12 uses the Kalman gain K, the pre-estimation value, and the observation value z _{t according} to Equation (5) to perform the post-estimation. The value is updated (S500). The observation value z _t is searched for within a predetermined distance from the tracking point. _{Alternatively} , the latest prior estimation value may be used as the observation value z _t . The post-estimation value is the newest tracking point, and the next observed value (detection point) is searched for around the new tracking point or the latest prior estimation value. When there is no detection point within a certain range, tracking is performed by the motion model, not by the Kalman filter.

  Next, the tracking unit 12 updates the covariance matrix P′t of the posterior estimation error using Expression (7) (S600). By updating the covariance matrix P′t, it is possible to update the covariance matrix Pt of the error of the prior estimation.

The tracking unit 12 updates the covariance matrix Pt of the pre-estimation error (S700). Thereby, the Kalman gain K can be calculated in the next step S200.
Thus, for the next observed value z _t , the Kalman gain is calculated from the covariance matrix Pt of the error of the pre-estimation updated in S700 (S200), and the Kalman gain and the pre-estimation updated in S300 are calculated. The value of the posterior estimation is calculated from the value of (S500).

  In the present embodiment, the Kalman filter has been described. However, any filter suitable for object tracking, such as an αβ filter, an αβγ filter, or a particle filter, can be used similarly.

<Awareness of the presence of the target by the alarm unit>
Since the tracking unit 12 estimates the most probable target position from the detection point (observed value z _t ), the alarm unit 13 can call attention more accurately than using the detection point itself.

  For example, when a target (tracking point) enters a warning area within a predetermined distance from the host vehicle, the warning unit 13 lights a warning lamp to notify the target of the presence. Further, the driver is alerted that there is a possibility of abnormally approaching the target by blowing an alarm sound at the timing when the driver turns on the blinker.

[Tracking 1 of this embodiment]
FIG. 9 shows an example of a flowchart showing a procedure for the tracking unit 12 to track a tracking point, and FIG. 10 shows an example of a diagram for schematically explaining tracking by a dummy detection point.

  The tracking unit 12 repeats the procedure shown in FIG. 9 for each radar scan (one beat signal sampling). As shown in FIG. 10, it is assumed that two tracking points A and B have been tracked by the radar scanning so far.

  When the tracking point is calculated in step S500 of FIG. 8 (or after S300 when using the latest prior estimation value), the list creation unit 21 calculates the tracking point or the previous estimation value from A tracking point that is likely to be affected by interference is extracted and a list is created (S501). When a plurality of targets are present in the vicinity, the detection points vary in the circumferential direction of a circle with a constant distance from the radar apparatus 100. Therefore, among the plurality of tracking points (targets), detection points of a plurality of tracking points that satisfy conditions such as a short distance R and a short interval between the tracking points may be affected by interference. For example, the list creation unit 21 extracts a tracking point whose difference in distance R is a threshold value.

  FIG. 11 is an example of a diagram schematically showing a list of tracking points. A distance R, a relative speed V, an azimuth θ, and a power P are associated with the tracking point. Since the tracking point is obtained as the position (x, y), the distance R is obtained from the position (x, y). The relative velocity V may be obtained from the state variable x, or an observation value (detection point) used for calculating the tracking point may be used. The azimuth θ is obtained from the position (x, y). The electric power P uses the observed value (detection point) used to calculate the tracking point.

  Next, the dummy detection point creation unit 22 performs the processing of S502 to S507 in FIG. 9 for all the tracking points including the tracking points in the tracking point list.

  First, the dummy detection point creation unit 22 determines whether or not the tracking point of interest is a tracking point included in the list of tracking points (S502).

  When focusing on the tracking point A, the determination in S502 is Yes, and the dummy detection point creation unit 22 determines whether or not the detection point is within the radial threshold ΔR with respect to the tracking point A (S503). ΔR corresponds to the first width (first predetermined range) of the claims. That is, in consideration of the fact that the detection points vary in the circumferential direction, a detection point having a short distance R is searched even if the deviation in the X-axis direction is large. When there are a plurality of detection points within the threshold ΔR, the detection point with the closest relative distance or the detection point with the closest distance R is selected.

If the detection point is within the radial threshold ΔR with respect to the tracking point A (Yes in S503), the dummy detection point creation unit 22 creates a dummy detection point (S504). First, the X ′ coordinate of the dummy detection point is obtained as follows.
X ′: final X value of tracking point A, average value of X value history, etc. The intersection of a circle of radius R of the detection point found within threshold ΔR and X ′ = a constant straight line is obtained. The intersection is a dummy detection point. Since the x coordinate of the intersection point is X ′, Y ′ which is the y coordinate of the dummy detection point is as follows.
Y ′: √ (R ² −X ′ ² )
In FIG. 10, the dummy detection point A ′ of the tracking point A is obtained at a position where the y coordinate is slightly larger than the tracking point A.

  The tracking unit 12 performs the tracking process of the Kalman filter of FIG. 8 using the dummy detection point A ′ as the observation value zt (S505). As a result, the dummy detection point A ′ is applied to the observation value zt of the equation (5), and the tracking point A can be updated.

  Next, the dummy detection point creation unit 22 returns to step S502 and repeats the same processing for the tracking point to be focused next. In FIG. 10, a dummy detection point B ′ of the tracking point B is created slightly on the X axis side from the tracking point B. In the example of FIG. 10, the dummy detection point B ′ is created using the same detection point as the tracking point A for the tracking point B.

  In this way, when there is a possibility that reflected waves from a plurality of targets may affect the tracking point, the detection point distance R using the x coordinate of the tracking point without using the x coordinate where the deviation of the detection point is large. A point on the same distance R as a dummy detection point. Thereby, it is possible to update the Kalman filter by extracting information (radius R) that can be expected to have high accuracy from detection points that may be affected by interference.

  Therefore, since there are a plurality of targets, even when the detection point is affected by the interference of the radar, it is possible to extract the information with a small deviation from the information held by the detection point and continue the tracking.

  In step S502, when the tracking point of interest is not registered in the list of tracking points (No in S502), the tracking unit 12 performs conventional pairing (S506). Conventional pairing refers to determining whether or not a detection point is within a predetermined distance of the tracking point of interest, and determining the detection point for updating the Kalman filter. If there is no detection point near the tracking point of interest, the position of the tracking point of interest is updated based on a motion model (for example, constant velocity linear motion).

  In step S503, when there is no detection point within the threshold ΔR in the radial direction with respect to the tracking point A (No in S503), since the deviation from the detection point is large, the tracking unit 12 performs conventional pairing (S506). ).

And the tracking part 12 performs the tracking process of the Kalman filter of FIG. 8 using the detection point specified by the conventional pairing (S507). As a result, the detection point is applied to the observation value z _t in the equation (5), and the tracking points A and B can be updated.

  When the interference of the radar decreases while repeating this processing, the next tracking point A updated at the tracking point A and the dummy detection point A ′, the next updated at the tracking point B and the dummy detection point B ′, and so on. The tracking point B is not included in the list of tracking points. Therefore, normal pairing (S506) becomes possible.

[Tracking 2 of this embodiment]
In the tracking 1 of this embodiment described above, a dummy detection point is created if a detection point is found in step S503 for the tracking point in the list of tracking points. However, even if the radar interference may affect the detection point, it is expected that the detection points are likely to be concentrated near the target having a high reflection intensity (in FIG. 14, the detection points are close to the guard rail). doing.).

  Therefore, even if there is a possibility that the interference of the radar may affect the detection point, if there is a detection point near the tracking point, it can be estimated that the error of the detection point is small. Therefore, in this case, the radar apparatus 100 that updates the Kalman filter without creating a dummy detection point will be described. By using the detection point, it is possible to suppress an increase in processing load and improve the accuracy of the tracking point, rather than creating a dummy detection point.

  FIG. 12 shows an example of a flowchart showing a procedure for the tracking unit 12 to track the tracking point, and FIG. 13 shows an example of a diagram for schematically explaining tracking by the dummy detection point.

  The processing from steps S501 to S503 in FIG. 12 is the same as that in FIG. As shown in FIG. 13, it is assumed that two tracking points A and B are registered in the list of tracking points.

  In step S503, when the detection point is within the radial threshold ΔR with respect to the tracking point A (Yes in S503), the dummy detection point creation unit 22 does not immediately create a dummy detection point, and the dummy detection point creation unit 22 Determines whether there is a detection point in the vicinity region E with respect to the tracking point A (S511). The neighborhood region E corresponds to a second predetermined range in the claims. The neighboring region E is represented by a fan shape surrounding the tracking point A in FIG. The length in the radial direction of the vicinity region E is a threshold value ΔR ′, and the angle in the azimuth direction is g [degrees]. That is, in FIG. 13, the detection point is detected within g degrees in the azimuth direction from the tracking point A and within ΔR ′ in the radial direction from the tracking point A. A detection point is not detected in the region near the tracking point B. By selecting a detection point in the vicinity as the tracking point in this way, it is possible to update the Kalman filter using the detection point even when the detection point may be affected by radar interference.

  If there is a detection point in the vicinity of the tracking point (Yes in S511), the Kalman filter is updated using the detection point detected in Step S511 (S512). Therefore, the tracking point can be updated with a detection point near the tracking point without creating a dummy detection point.

If there is no detection point in the vicinity of the tracking point (No in S511), the process proceeds to step S504, and the dummy detection point creation unit 22 creates a dummy detection point (S504). Similarly to FIG. 10, the tracking unit 12 performs the tracking process of the Kalman filter of FIG. 8 using the dummy detection point B ′ (S505). As a result, the dummy detection point B ′ is applied to the observation value z _t in the equation (5), and the tracking point B can be updated.

  Therefore, according to the radar apparatus 100 of FIG. 12, if there is a detection point near the detection point that is likely to be affected by the interference of the radar, the Kalman filter is updated by giving priority to the detection point. Therefore, the tracking point can be updated using a detection point with high reflection intensity, and the accuracy of the tracking point can be maintained. If there is no detection point in the vicinity, a dummy detection point is created from the detection point within ΔR for the tracking point and the Kalman filter is updated, so that the accuracy of the tracking point can be maintained.

DESCRIPTION OF SYMBOLS 11 Detection part 12 Tracking part 13 Alarm part 21 List preparation part 22 Dummy detection point preparation part 100 Radar apparatus

Claims (7)

An object detection device that detects an object around a vehicle by updating a tracking point of the object according to the position of the detected object,
Position detecting means for detecting the position of the object;
Position prediction means for predicting the predicted position of the object from the tracking point and the movement rule of the object;
Tracking point update means for updating the tracking point based on the position of the object detected by the position detection means and the predicted position predicted by the position prediction means;
When the object detected by the position detection means exists within a first predetermined range from the tracking point, the position information detected by the position detection means is used to determine the position of the object detected by the position detection means. Correction means for correcting,
The tracking point update unit updates the tracking point based on the position of the object corrected by the correction unit and the predicted position predicted by the position prediction unit.
An object detection apparatus characterized by that. When the object detected by the position detecting means is within the first predetermined range from the tracking point,
Further, when the object detected by the position detection unit exists within a second predetermined range from the tracking point, the tracking point update unit detects the position within the second predetermined range from the tracking point. Updating the tracking point based on the position of the object detected by the means and the predicted position predicted by the position prediction means;
The object detection apparatus according to claim 1. The first predetermined range is a circumferential region having a first width in the radial direction centered on a circle having a constant distance from the object detection device to the tracking point.
The object detection apparatus according to claim 1, wherein the object detection apparatus is an object detection apparatus. The second predetermined range is centered on the circumference of a circle having a constant distance from the object detection device to the tracking point, and out of the circumferential region having a second width in the radial direction. It is a fan-shaped region having a predetermined angle around the azimuth,
The object detection apparatus according to claim 2. Excluding stationary objects with respect to the road surface, having an object specifying means for specifying objects whose distance to the object is within a threshold value,
The correcting means corrects the position of the object only for the object specified by the object specifying means.
The object detection apparatus according to claim 1. When the vehicle width direction is the x direction and the vehicle length direction is the y direction,
The correction means determines the x coordinate of the object detected by the position detection means from the past position information of the tracking point;
determine the y coordinate of the point where the distance to the object detection device is equal to the distance to the object detected by the position detection means on a line where x is constant;
The x-coordinate and the y-coordinate are the corrected positions of the object detected by the position detecting means,
The object detection apparatus according to claim 1, wherein the object detection apparatus is an object detection apparatus. The position detection means transmits radio waves to the vehicle periphery, detects the distance to the object and the azimuth of the object based on the reflected wave reflected from the object around the vehicle, and identifies the position of the object from the distance and azimuth.
The object detection apparatus according to any one of claims 1 to 6.