JP2016223780A - Method for determining discriminant, and discrimination device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To improve the accuracy of determining whether there exists an object that a vehicle could collide with.SOLUTION: A method for determining a discriminant has: a first acquisition step (S11) for acquiring, in association with the type of radio wave reflector, reflected wave intensity data indicating the intensity of a reflected wave that occurs as a radio wave transmitted from a vehicle is reflected at a radio wave reflector; a generation step (S12) for generating characteristic data indicating relationship between the distance from the vehicle to the radio wave reflector and the intensity of the reflected wave on the basis of the reflected wave intensity data; a second acquisition step (S13) for acquiring a feature value indicating the characteristic of the reflected wave; an expression generation step (S15) for generating a plurality of first expressions in which a discriminant including a feature value and coefficient that correspond to an upper structure at a height that the vehicle will not collide with is defined as a first value, and generating a plurality of second expressions in which a discriminant including a feature value and coefficient that correspond to a radio wave reflector other than the upper structure is defined as a second value; and a determination step (S18) for determining a coefficient on the basis of probability that the plurality of first expressions and the plurality of second expressions are satisfied.SELECTED DRAWING: Figure 5

Description

  The present invention relates to a method for determining a discriminant for discriminating the type of an object in front of a vehicle, and a discriminating apparatus for discriminating the type of an object using the discriminant determined by this method.

  2. Description of the Related Art Conventionally, there is known a method of detecting an object that has a risk of colliding with a vehicle by mounting a device that emits radar on the vehicle and detecting a reflected wave of the radar transmitted from the vehicle. Patent Document 1 discloses a lower structure in which a detected object has a risk of colliding with a vehicle based on the characteristic of reflected wave intensity (hereinafter referred to as “reflected wave intensity”) detected after transmitting a radar. There is disclosed a method for discriminating whether an object is an object or an upper structure that has no risk of collision with a vehicle.

JP2012-18058A

  In the method described in Patent Document 1, it is determined whether or not the detected object is an upper structure based on whether or not the characteristic of the reflected wave intensity satisfies a predetermined condition. . For example, in the conventional method, the number of maximum values detected in a curve indicating the relationship between distance and reflected wave intensity is specified, and a table indicating the relationship between the number of detected maximum values and the height of an object prepared in advance is referred to. By doing so, the height of the object was specified.

  However, the characteristic of the reflected wave intensity greatly varies depending on the environment around the vehicle. Therefore, in an environment where the vehicle actually travels, the relationship between the number of detected maximum values prepared assuming an ideal state and the height of the target object often does not apply, and a discrimination criterion is created. It was difficult. In addition, when a discrimination criterion that relies on human experience and intuition is used, there is a problem that the probability of erroneous discrimination is high.

  Therefore, the present invention has been made in view of these points, and an object of the present invention is to improve the determination accuracy of whether or not there is an object at risk of collision with a vehicle.

In the first aspect of the present invention, reflected wave intensity data indicating the intensity of a reflected wave generated by reflection of a radio wave transmitted from a vehicle by a radio wave reflector, which is executed by a computer, is classified into the type of the radio wave reflector. A first acquisition step of acquiring in association; a generation step of generating characteristic data indicating a relationship between a distance from the vehicle to the radio wave reflector and the intensity of the reflected wave based on the reflected wave intensity data; A second acquisition step of acquiring a characteristic amount indicating the characteristic of the reflected wave, which is calculated based on characteristic data; and the characteristic amount corresponding to the upper structure which is the radio wave reflector having a height at which the vehicle does not collide. And a plurality of first expressions having a first value as a discriminant including a coefficient multiplied by the feature quantity, and multiplying the feature quantity and the feature quantity corresponding to the radio wave reflector other than the upper structure. The coefficient generation step for generating a plurality of second expressions having the discriminant including the coefficient as a second value, and the probability that the plurality of first expressions and the plurality of second expressions are satisfied. A discriminant determination method characterized by comprising:
By doing in this way, the discrimination | determination precision of a target classification can be improved compared with the case where a discriminant is determined based on a designer's experience and intuition.

  The feature amount includes, for example, a first feature amount corresponding to the distance and a second feature amount indicating the magnitude of the change amount of the reflected wave intensity with respect to the predetermined change amount of the distance. The feature amount may include a third feature amount indicating a frequency with which the intensity of the reflected wave decreases by a predetermined amount or more with respect to a predetermined change amount of the distance.

  The discriminant is, for example, a regression model including the first feature quantity, a first coefficient multiplied by the first feature quantity, the second feature quantity, and a second coefficient multiplied by the second feature quantity. is there.

  According to a second aspect of the present invention, there is provided a discriminating device for discriminating the type of the radio wave reflector based on the reflected wave generated by the radio wave transmitted from the vehicle reflected by the radio wave reflector, the discriminant described above Storage means for storing the discriminant determined by the determination method, transmitting means for transmitting radio waves, receiving means for receiving reflected waves of the radio waves transmitted by the transmitting means, and reflected waves received by the receiving means And calculating a feature value of the radio wave reflector based on the value of the discriminant calculated by the second calculator and a calculating unit that calculates the value of the discriminant using the calculated feature amount. And a discriminating means for discriminating the type.

  According to the present invention, it is possible to improve the accuracy of determining whether or not there is an object at risk of a vehicle collision.

It is a figure for demonstrating the outline | summary of one Embodiment of this invention. It is a figure for demonstrating the principle in which the vehicle V discriminate | determines the kind of target. It is a figure which shows the characteristic of the reflected wave intensity | strength which considered the reflection with respect to the ground. It is a figure for demonstrating the flow which determines a discriminant. It is a flowchart which shows the procedure in which a computer determines the coefficient of a discriminant. It is a figure for demonstrating the characteristic of the reflected wave intensity used for learning of a discriminant. FIG. 10 is a distribution diagram of discriminant values obtained by inputting a feature amount based on characteristic data into a discriminant including a determined coefficient. It is a figure which shows the structure of the discrimination | determination apparatus 1 which discriminate | determines the classification of a target.

[Overview of this embodiment]
FIG. 1 is a diagram for explaining an outline of an embodiment of the present invention. FIG. 1A shows a state where an obstacle TS (for example, a vehicle traveling in front) as a radio wave reflector (hereinafter referred to as “target”) that reflects radio waves exists in front of a vehicle V traveling on a road. Is shown. When the vehicle V transmits a radio wave such as a millimeter wave radar, the transmitted radio wave reaches the obstacle TS via various routes such as A → B and A → C → B shown in FIG. Reflected by the obstacle TS. The vehicle V receives a radio wave reflected by the obstacle TS (hereinafter referred to as “reflected wave”), and detects the presence of the obstacle TS based on the characteristics of the received reflected wave. When the vehicle V detects the presence of the obstacle TS, it activates the emergency brake system and automatically stops.

  FIG. 1B shows a state where the upper structure TJ exists above the front of the vehicle V. The upper structure TJ is a target installed at a height at which the vehicle V does not collide, such as a road sign and a footbridge. The radio wave transmitted by the vehicle V reaches the upper structure TJ through various paths and is reflected by the upper structure TJ, as in FIG. The vehicle V receives the reflected wave reflected by the upper structure TJ, and detects the presence of the upper structure TJ based on the characteristics of the received reflected wave.

  In order to prevent the emergency brake system from being activated when the vehicle V detects the upper structure TJ, the vehicle V determines whether the target that reflected the radio wave is an obstacle TS based on the characteristics of the received reflected wave. It is determined whether or not the object is TJ. The vehicle V according to the present embodiment determines whether the target is the obstacle TS or the upper structure TJ by using a discriminant that can improve the discrimination accuracy as compared with the related art.

[Determination method of target type]
FIG. 2 is a diagram for explaining the principle by which the vehicle V determines the type of target. FIG. 2A is a diagram showing the relationship between the distance from the vehicle V to the obstacle TS and the intensity of the reflected wave (hereinafter referred to as “reflected wave intensity”) when the target is the obstacle TS. As shown in FIG. 2A, when the target is an obstacle TS existing on the road, the reflected wave intensity increases as the distance from the vehicle V to the target decreases.

  FIG. 2B is a diagram showing the relationship between the distance from the vehicle V to the upper structure TJ and the reflected wave intensity when the target is the upper structure TJ. As shown in FIG. 2B, when the target is an upper structure TJ existing above, the reflected wave intensity decreases as the distance from the vehicle V to the target decreases, and when the distance further decreases, the reflected wave Cannot be detected.

By the way, the path traveled by the millimeter wave radar is not only a path directly traveling from the vehicle V to the target, but also reflected by the ground or the like. The reflected wave intensity has a characteristic that it changes depending on the distance to the target due to interference between the wave reflected on the ground and the directly traveling wave. The reflected wave intensity P can be expressed by the following formula (1).

In the above equation, ρ is the reflectance of the road surface, κ = 2π / λ is the wave number, λ is the wavelength, h _s is the height of the millimeter wave radar transmitter, _ht is the height of the target, and R is the height to the target. The distance, A, is a coefficient determined by the transmission antenna gain, the reflection cross section, and the like. It can be seen from the cos term in Equation (1) that the reflected wave intensity periodically changes with respect to the reciprocal of the distance R to the target.

FIG. 3 is a diagram illustrating the relationship between the distance from the vehicle V to the target and the reflected wave intensity corresponding to the equation (1). That is, FIG. 3 is a diagram showing the characteristics of the reflected wave intensity in consideration of reflection on the ground. 3A shows the characteristic of reflected wave intensity when the target is an obstacle TS, and FIG. 3B shows the characteristic of reflected wave intensity when the target is an upper structure TJ. ing. As shown in FIG. 3, as the vehicle V approaches the target, the cycle in which the reflected wave intensity changes becomes shorter. The period becomes longer as the height h _t of the target is small, the period is shortened as the height h _t of the target is large. That is, the period when the target is the upper structure TJ is shorter than the period when the target is the obstacle TS.

  The vehicle V determines whether or not the target is the upper structure TJ based on the difference in the reflected wave intensity characteristics depending on the target type as shown in FIG. However, in the actual environment, the characteristic of the reflected wave intensity is different from the ideal characteristic as shown in FIG. 3 due to the influence of external noise, multipath, and the like. Therefore, a discriminant for discriminating the target type with high accuracy is required based on the characteristic of reflected wave intensity that is not ideal. In the present embodiment, using a computer, the value of the discriminant is calculated while changing the coefficient of the discriminant using the measured data, and the optimum coefficient is learned. The discrimination accuracy can be improved by discriminating the type of the target using the discriminant including the optimum coefficient obtained by learning.

[Method of determining discriminant]
FIG. 4 is a diagram for explaining the flow of determining the discriminant. First, in various environments, the characteristic of the reflected wave from the obstacle TS and the characteristic of the reflected wave from the upper structure TJ are measured, and actual measurement data obtained by the measurement is collected as learning data. The actual measurement data is data in which the target type is associated with the data indicating the relationship between the distance and the reflected wave intensity as shown in FIG.

  The actual measurement data is input to a computer that executes a program for learning the coefficient of the discriminant. The actual measurement data may be input to the computer via the storage medium, or may be input to the computer via the Internet from the vehicle V that collects the actual measurement data.

  By executing the program, the computer uses the learning data to determine the coefficient so that a different tendency appears in the discriminant value depending on whether or not the target is the superstructure. The discriminant thus obtained is stored in a storage medium of the vehicle V at the time of manufacturing the vehicle V or after manufacturing. The vehicle V transmits a radar while traveling, and calculates the value of the discriminant by substituting the received data indicating the characteristic of the reflected wave intensity into the discriminant. The vehicle V can determine whether the target is an upper structure based on whether the value of the discriminant is greater than or less than the threshold.

  FIG. 5 is a flowchart showing a procedure for determining the coefficient of the discriminant by the computer. First, the computer acquires reflected wave intensity data indicating the intensity of a reflected wave generated by reflecting a radio wave transmitted from the vehicle V with a target (step S11).

  Subsequently, the computer generates characteristic data indicating the relationship between the distance from the vehicle V to the target and the intensity of the reflected wave based on the reflected wave intensity data (step S12). The computer, for example, shows the relationship between the distance and the intensity by specifying the distance from the vehicle V to the target based on the delay time from the time when the radar is transmitted to the time when the reflected wave is received. Generate characteristic data as shown.

Subsequently, the computer acquires a feature amount indicating the feature of the characteristic data calculated based on the characteristic data (step S13). The computer acquires a large number of feature quantities as learning data for determining the coefficient of the discriminant. Computer, for example, to obtain the distance x ₁ to the target as the first feature amount. The computer, for example, obtains the frequency x ₃ which acquires the tilt x ₂ of the curve showing the relationship between the distance and the reflected wave intensity to the target as the second feature quantity, the reflected wave intensity decreases as the third feature amount To do. The computer may acquire these feature values from the outside, or may acquire them as a result calculated by the computer based on the characteristic data.

Slope x ₂ of the curve showing the relationship between the distance and the reflected wave intensity to the target, for example, is calculated by first order approximation of a plurality of points of the characteristic data, the reflected wave intensity for a given variation of the distance change amount of Expressed by size. Frequency x ₃ which reflected wave intensity is decreased, for example, the reflected wave intensity is represented by the size of the frequency to be reduced more than a predetermined amount for a given change in distance, corresponding to the period of the reflected wave intensity fluctuates .

FIG. 6 is a diagram for explaining the characteristics of reflected wave intensity used for discriminant learning. FIG. 6A shows the characteristics of the ideal reflected wave intensity shown in FIG. When the period in which the reflected wave intensity changes with respect to the time interval for sampling the reflected wave received by the vehicle V is short, the characteristic of the reflected wave intensity acquired by the vehicle V is as shown by the solid line in FIG. . The computer acquires characteristic data indicating the relationship between the distance from the vehicle V to the target and the reflected wave intensity, as indicated by the solid line in FIG. 6B, and is calculated based on the acquired characteristic data. acquiring an inclination x ₂ of the characteristic curve of the reflected wave intensity shown).

Further, the computer calculates a ratio of the number of data (for example, data a and data b in FIG. 6D) smaller than a predetermined value relative to the “distance to the target” by comparison with the reflected wave intensity indicated by the adjacent data. reflected wave intensity is obtained as the frequency x ₃ to decrease. If the time interval for sampling the reflected wave is sufficiently small, the computer obtains the frequency at which the reflected wave intensity becomes the minimum value in the characteristics shown in FIG. 6A as the frequency x _{3 at} which the reflected wave intensity decreases. May be.

Returning to FIG. 5, the processing following step S13 will be described. The computer sequentially inputs the feature values acquired in step S13 into the discriminant, and sets the discriminant value to the first value “1” or the second value “0” according to the target type. A plurality of formulas including coefficients are generated (step S14). As the discriminant, the computer uses, for example, a logistic regression model that is one of generalized linear models represented by the following formula (2).

The coefficient a ₁ in Expression (2) is a coefficient that is multiplied by the first feature quantity x ₁ , and the coefficient a ₂ is a coefficient that is multiplied by the second feature quantity x ₂ , and the coefficient a ₃ is a coefficient multiplied to x _3. The computer, when the target is the upper structure TJ, that is, when the radio wave is reflected by the upper structure TJ, the first feature value x ₁ , the second feature value x _2, and the third feature value of the reflected wave obtained. By setting the value of the discriminant p (x) when x ₃ is substituted into the equation (2) to the first value “1”, the first including the coefficient a ₁ , the coefficient a ₂ , and the coefficient a ₃ as variables Generate an expression. Further, when the target is the obstacle TS, that is, when the radio wave is reflected by the obstacle TS, the first feature quantity x ₁ , the second feature quantity x _2, and the third feature quantity x ₃ of the reflected wave are obtained. By setting the value of the discriminant p (x) when substituted into the equation (2) to the second value “0”, the second equation including the coefficients a ₁ , a ₂ , and a ₃ as variables is generated. To do.

The computer repeats step S13 and step S14 until the generation of the first expression or the second expression is completed for all feature amounts (learning data). Then, when the generation of the first expression or the second expression is completed for all feature amounts (learning data) (Yes in step S15), the computer calculates coefficients a ₁ , The coefficient a ₂ and the coefficient a ₃ are determined (step S16). The computer determines the coefficient a ₁ , the coefficient a ₂ , and the coefficient a ₃ using the maximum likelihood method, for example, so that the probability that all the expressions are satisfied is maximized.

  FIG. 7 is a distribution diagram of values obtained by inputting feature amounts based on characteristic data into discriminants including determined coefficients. ○ indicates the discriminant value calculated using the feature value when the target is the superstructure TJ, and Δ indicates the feature value when the target is the obstacle TS. The discrimination value is shown. FIG. 7A shows a case where the discriminant is not appropriate, where the discriminant value indicated by ○ and the discriminant value indicated by Δ are mixed, and the target is the upper structure TJ. The case where the target is the obstacle TS cannot be separated. Therefore, it is difficult to determine the type of target using such a discriminant.

  On the other hand, FIG. 7B shows a case where the discriminant is appropriate, and there are regions where there are many o and regions where there is a lot of Δ. The computer sets a threshold value of the discriminant value for discriminating whether the target is the obstacle TS or the upper structure TJ based on the distribution state of the discriminant value shown in FIG. decide. In order to prevent the computer from misidentifying that the target is the superstructure TJ even though the target is the obstacle TS, for example, as shown in FIG. 7B, the target is the obstacle TS. A determination threshold value is determined so that a determination value in a certain case is not included. In order to reduce the probability that the computer erroneously determines that the target is the upper structure TJ even though the target is the obstacle TS, the computer sets the determination threshold as shown in FIG. You may set to the area | region where the discriminant value in case of TJ is included.

[Configuration of Discriminating Device 1]
FIG. 8 is a diagram showing the configuration of the discriminating apparatus 1 that discriminates the type of the target based on the reflected wave intensity received during traveling using the discriminant determined by the above method. The determination device 1 includes a radar transmitter 11, a reflected wave receiver 12, a storage unit 13, and a controller 14. The determination device 1 is provided and used in the front part of the vehicle V, for example.

The radar transmitter 11 transmits a millimeter wave radar in the direction in which the vehicle V travels.
The reflected wave receiving unit 12 receives the reflected wave reflected by the radar transmitted by the radar transmitting unit 11 and notifies the control unit 14 of the value of the received reflected wave intensity.

  The storage unit 13 is a storage medium such as a ROM, a RAM, and a hard disk. The storage unit 13 stores a program executed by the control unit 14. The storage unit 13 stores a discriminant including an optimum coefficient calculated by the computer.

  The control unit 14 is a CPU, for example. The control unit 14 functions as a transmission control unit 141, a calculation unit 142, and a determination unit 143 by executing a program stored in the storage unit 13.

The transmission control unit 141 controls the radar transmission unit 11 to transmit a radar.
Based on the reflected wave intensity of the radio wave transmitted from the vehicle V, the calculating unit 142 calculates the distance x _{1 to} the target, the slope x _{2 of} the curve indicating the relationship between the distance to the target and the reflected wave intensity, and the reflected wave intensity. to calculate the frequency _{x 3} to decrease. The calculation unit 142 calculates the value of the discriminant including the coefficient determined in step S18 of FIG. 5 using these feature amounts. The calculation unit 142 notifies the determination unit 143 of the calculated value.

  The determination unit 143 determines the type of the target based on the discriminant value calculated by the calculation unit 142. Specifically, the determination unit 143 determines that the target is an obstacle TS when the value of the discriminant is equal to or less than a predetermined threshold, and the target is the superstructure when the value of the discriminant is larger than the threshold. It is determined that the object is TJ. When the determination unit 143 determines that the target is the obstacle TS, the determination unit 143 displays warning information on the instrument panel of the vehicle V or sounds a warning sound.

[Modification]
The computer may acquire actual measurement data in association with the surrounding environment (for example, weather, brightness) when the vehicle V is traveling, and may determine an optimum discriminant in association with the surrounding environment. The discriminating apparatus 1 has a measuring unit (not shown) that measures the state of the surrounding environment, stores a plurality of discriminants in association with the measured state of the surrounding environment, and detects the state of the surrounding environment detected during traveling The target type may be determined using a discriminant corresponding to the surrounding environment based on the information indicating the state of the surrounding environment input by the driver or the information indicating the state of the surrounding environment.

[Experimental example]
The computer uses 1507 samples of reflected wave intensity when the target type is “superstructure” and 10515 samples of reflected wave intensity when the target type is “stopped vehicle”. The optimum coefficient in equation (2) was determined. Then, the type of the target was discriminated while the vehicle V was running using the discriminating device 1 that stores the discriminant including the determined coefficient. The vehicle V equipped with the discriminating device 1 travels on 59 roads where the superstructure is present, and the obtained reflected wave intensity data is substituted into the discriminant to discriminate the target type. The probability that the target was correctly identified as the superstructure was 67.8%.

As a comparative example, discrimination was performed using the following discriminant without using a coefficient determined by learning.
(Discriminant)
75 ≦ x ₁ <100 and x ₂ ≧ 0.20 and x ₁ + 50x ₃ ≧ 200, or x ₁ ≧ 100 and x ₂ ≧ 0.15 and x ₁ + 50x ₃ ≧ 200
In this case, the probability that the discriminating apparatus 1 could correctly discriminate that the target was the superstructure was 42.4%. As described above, it was confirmed that the determination accuracy of the target type can be improved by determining the optimum coefficient using the method according to the present embodiment.

[Effect in this embodiment]
As described above, in the present embodiment, the type of the target is determined using the discriminant determined based on the actually measured data of the reflected wave intensity when the radar is transmitted from the vehicle V. Therefore, the discrimination accuracy of the target type can be improved as compared with the case where the discriminant is determined based on the designer's experience and intuition.

  As mentioned above, although this invention was demonstrated using embodiment, the technical scope of this invention is not limited to the range as described in the said embodiment. It will be apparent to those skilled in the art that various modifications or improvements can be added to the above embodiment. It is apparent from the scope of the claims that the embodiments added with such changes or improvements can be included in the technical scope of the present invention.

DESCRIPTION OF SYMBOLS 1 Discriminating device 11 Radar transmission part 12 Reflected wave receiving part 13 Storage part 14 Control part 141 Transmission control part 142 Calculation part 143 Discrimination part

Claims (5)

Executed by the computer,
A first acquisition step of acquiring, in association with the type of the radio wave reflector, reflected wave intensity data indicating the intensity of the reflected wave generated by the radio wave transmitted from the vehicle reflected by the radio wave reflector;
Based on the reflected wave intensity data, a generation step for generating characteristic data indicating a relationship between the distance from the vehicle to the radio wave reflector and the intensity of the reflected wave;
A second acquisition step of acquiring a feature amount indicating the characteristic of the reflected wave, which is calculated based on the characteristic data;
Generating a plurality of first expressions having a first value as a discriminant including the feature quantity corresponding to the superstructure that is the radio wave reflector of a height at which the vehicle does not collide and a coefficient multiplied by the feature quantity; And a formula generation step for generating a plurality of second formulas having a second value of the discriminant formula including the feature quantity corresponding to the radio wave reflector other than the superstructure and the coefficient multiplied by the feature quantity;
A determining step for determining the coefficient based on a probability that the plurality of first expressions and the plurality of second expressions are satisfied;
A discriminant determination method characterized by comprising: The feature amount includes a first feature amount corresponding to the distance and a second feature amount indicating a magnitude of a change amount of the intensity of the reflected wave with respect to a predetermined change amount of the distance.
The discriminant determination method according to claim 1. The feature amount includes a third feature amount that indicates a frequency with which the intensity of the reflected wave decreases by a predetermined amount or more with respect to a predetermined change amount of the distance.
The discriminant determination method according to claim 2. The discriminant is a regression model including the first feature quantity, a first coefficient multiplied by the first feature quantity, the second feature quantity, and a second coefficient multiplied by the second feature quantity. Characterized by the
The discriminant determination method according to claim 2 or 3. A discriminating device for discriminating a type of the radio wave reflector based on a reflected wave generated by reflecting the radio wave transmitted from the vehicle by the radio wave reflector;
Storage means for storing a discriminant determined by the discriminant determining method according to any one of claims 1 to 4,
A transmission means for transmitting radio waves;
Receiving means for receiving a reflected wave of the radio wave transmitted by the transmitting means;
Calculating means for calculating the feature quantity of the reflected wave received by the receiving means, and calculating a value of the discriminant using the calculated feature quantity;
Discriminating means for discriminating the type of the radio wave reflector based on the value of the discriminant calculated by the second calculating means;
A discriminating apparatus characterized by comprising:

Images