JP2008128998A - Smoothing filter system and speed detector - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a smoothing filter device and a speed detector using it capable of smoothing exactly time-series data which have a dispersion into a value near the center of data dispersion amplitude. <P>SOLUTION: The smoothing filter device 1 which smoothes and outputs data of physical quantity X input discretely serially comprises an addition fraction decision means 2 to decide addition fraction P, a smoothing means 3 to output calculated value Xout(t), by weighting addition of input value Xin(t) of physical quantity X input this time and output value Xout(t-1) prescribed previous period carrying out with addition fraction P, a time-derivative smoothing value calculation means 4 which calculates smoothing value dXs(t) of the time-derivative of output value through calculating time-derivative dXout(t) of the output value. The addition fraction decision means 2 calculates time-derivative dXin(t) of the input value, and decides addition fraction P based on time-derivative dXin(t) of the calculated input value and smoothing value dXs(t-1) of the time-derivative of output value calculated from the time-derivative smoothing value calculation means 4. <P>COPYRIGHT: (C)2008,JPO&INPIT

Description

  The present invention relates to a smoothing filter device and a speed detection device, and more particularly to a smoothing filter device that smoothes input physical quantity data and a speed detection device that detects the speed of an object using the same.

  In recent years, in vehicles such as passenger cars and trucks, the distance and relative speed of the preceding vehicle is detected by analyzing the image captured in front of the host vehicle and measuring with a radar device, etc., and automatically controls acceleration / deceleration of the host vehicle. Development of a preceding vehicle follow-up device and a collision prevention device to be performed is underway (see, for example, Patent Documents 1 and 2).

  In order to reliably perform such control, it is necessary to accurately detect the distance between the host vehicle and the preceding vehicle, the relative speed, and the like without delay. However, in actuality, for example, when the relative speed is detected using a vehicle speed sensor or is detected from a temporal difference in distance detected as time series data, the error or accuracy of the sensor itself or the accuracy of the detected distance As a result of detection or erroneous detection, etc., as shown schematically in FIG. 23, a so-called rampage in which the value greatly increases or decreases occurs in the relative speed detected as time-series data.

Since time series data whose values greatly fluctuate up and down in this way cannot be used for control as they are, for the purpose of suppressing such time series data fluctuations, data such as Kalman filter, moving average filter, time constant filter, etc. Smoothing with a smoothing filter is generally performed (see, for example, Patent Document 3).
JP-A-10-69598 JP 2006-188155 A JP 2001-242242 A

  By the way, in the data having the fluctuation exemplified in FIG. 23, the smoothing filter may be used to smooth the data to pass through the vicinity of the center of the amplitude of the data fluctuation as shown by the curve A in FIG. desired. However, in the smoothing using the smoothing filter as described above, the smoothing having the respective characteristics is usually performed according to the characteristics of the filter.

  For example, when a moving average filter or a time constant filter is used among the above three types of smoothing filters, the curve B that changes with time delay compared to the ideal smoothing curve A, that is, in FIG. In the example, the data is smoothed like a curve B that decreases with a delay of ΔT in comparison with a smoothing curve A that decreases with time series data.

  When the own vehicle is automatically controlled based on the smoothed result with time delay as shown by curve B, for example, when the time-series data is relative speed data, the relative speed at a certain time is smoothed. It is determined that the relative speed is larger than the actual relative speed represented by the curve A, and a large braking is automatically applied, which gives the driver a sense of incongruity and forces unnecessary acceleration to close the distance between the vehicles. .

  Further, when the time series data is data on the distance between the host vehicle and the preceding vehicle, it is determined that the distance at a certain time is a distance larger than the actual distance represented by the smoothing curve A, On the contrary, since the necessary amount of braking is not applied, the driver may suddenly step on the brake or, in the worst case, may collide with the preceding vehicle.

  On the other hand, when the Kalman filter is applied to the time-series data of FIG. 24 and smoothed, the data is usually smoothed like a curve C in which the increase / decrease in data is more emphasized than the ideal smoothing curve A. Is done. In this case, when the own vehicle is automatically controlled based on the curve C, if the time-series data is distance or relative speed data, it is recognized that the value has changed abruptly more rapidly than the actual value. As a result of braking, the driver may feel uncomfortable, or in the worst case, the vehicle may collide with a preceding vehicle.

  As described above, when smooth data is smoothed and automatic control or the like is performed based on the smoothed data, a conventional smoothing filter such as a Kalman filter, a moving average filter, or a time constant filter is used as it is. That is not always effective. As shown in the ideal smoothing curve A shown in FIG. 24, it is desired to develop a smoothing filter capable of accurately smoothing time-series data having a fluctuation to a value near the center of the amplitude of the data fluctuation. .

  The present invention has been made in view of such circumstances, and provides a smoothing filter device capable of accurately smoothing time-series data having a fluctuation to a value near the center of the amplitude of the data fluctuation. The purpose is to do. It is another object of the present invention to provide a speed detection device that can accurately detect the speed of an object using it.

In order to solve the above problem, the first invention provides:
In a smoothing filter device for smoothing and outputting data of physical quantities that are discretely input in time series,
An addition rate determining means for determining an addition rate;
Smoothing means for outputting, as an output value of the smoothed physical quantity data, a value calculated by weighting and adding the input value of the physical quantity data input this time and the output value of a predetermined period before by the addition ratio; ,
A time derivative smoothing value calculating means for calculating a time derivative of the output value and calculating a smoothing value of the time derivative of the output value;
The addition ratio determining means calculates a time derivative of the input value, and based on the calculated time derivative of the input value and the smoothed value of the time derivative of the output value calculated by the time derivative smoothed value calculating means. The addition ratio is determined.

  According to a second aspect of the present invention, in the smoothing filter device according to the first aspect of the invention, the addition ratio determining unit is configured to input the input based on a smoothed value of a time derivative of the output value calculated by the time derivative smoothed value calculating unit. A change tendency of the value is determined, and the addition ratio is determined based on a relationship between the determined change tendency of the input value and time differentiation of the input value.

  According to a third aspect of the present invention, in the smoothing filter device according to the first aspect of the invention, the addition ratio determining means uses a predetermined coefficient to obtain the smoothed value of the time derivative of the output value calculated by the time derivative smoothed value calculating means. A difference between the amplified value and the time derivative of the input value is calculated, and the addition ratio is determined based on the difference.

  According to a fourth aspect of the present invention, in the smoothing filter device according to the third aspect, the predetermined coefficient is a value obtained by amplifying the smoothed value of the time derivative of the output value by the predetermined coefficient. It is characterized by being set so as to be within the amplitude.

  According to a fifth invention, in the smoothing filter device of the third or fourth invention, the predetermined coefficient is a positive value, and the smoothed value of the time derivative of the output value is amplified by the predetermined coefficient. The value is set so as to be a value near the end of the amplitude of the time derivative of the input value.

  According to a sixth aspect of the present invention, in the smoothing filter device according to any one of the third to fifth aspects, the addition rate determination means increases the addition rate weighted to the input value as the absolute value of the difference decreases. It is determined to be a value.

  According to a seventh aspect, in the smoothing filter device according to the sixth aspect, the addition rate determination means includes a table in which an addition rate is assigned in advance for each absolute value range divided into a plurality of ranges. The addition ratio is determined based on the table.

  According to an eighth aspect of the present invention, in the smoothing filter device according to the sixth aspect of the invention, the addition ratio determining means includes a monotone decreasing function for the absolute value of the difference, and determines the adding ratio based on the monotonic decreasing function. It is characterized by that.

  According to a ninth aspect of the present invention, in the smoothing filter device according to any one of the first to eighth aspects of the invention, the addition ratio determining means sets the addition ratio in advance while the number of inputs of the input value is less than a predetermined number. It is determined to be a set constant value.

  According to a tenth aspect of the present invention, in the smoothing filter device according to any one of the first to ninth aspects, the time differential smoothed value calculating means includes the time differential of the output value and the time of the output value before the predetermined period. A differential smoothing value is weighted and added at a predetermined ratio to calculate a time differential smoothing value of the output value of the current cycling period.

  According to an eleventh aspect of the present invention, in the smoothing filter device according to any one of the first to tenth aspects, the physical quantity data is data of a change rate of the other physical quantity calculated by time-differentiating the other physical quantity. It is characterized by being.

  According to a twelfth aspect of the present invention, in the smoothing filter device according to any one of the third to tenth aspects of the invention, the physical quantity data is data of a change rate of the other physical quantity obtained by time differentiation of the other physical quantity. The predetermined coefficient is varied so as to be smaller as the value of the other physical quantity is smaller.

A thirteenth aspect of the present invention is a velocity detection hand device that detects the velocity of an object using the smoothing filter device of any one of the first to tenth aspects of the invention,
The physical quantity data is speed data input from a measuring unit that measures the speed of the object, and the smoothed speed is output.

A fourteenth aspect of the invention is a speed detection device that detects the speed of an object using the smoothing filter device of the eleventh or twelfth aspect of the invention,
The other physical quantity is a distance input from a measuring unit that measures the distance of the object, and the physical quantity data is data of the speed of the object obtained by time differentiation of the distance, and the smoothed data It is characterized by outputting a speed.

  According to a fifteenth aspect, in the speed detection device according to the thirteenth or fourteenth aspect, the speed is a vehicle speed of the host vehicle.

  According to a sixteenth aspect, in the speed detection device according to the thirteenth or fourteenth aspect, the speed is a relative speed between the host vehicle and a preceding vehicle.

  According to a seventeenth aspect, in the speed detection device according to the sixteenth aspect, the time differential smoothed value calculating means calculates the time differential of the output value and the smoothed value of the time differential of the output value before the predetermined period. When calculating the smoothed value of the time derivative of the output value by weighted addition at a predetermined ratio, the smoother of the time derivative of the output value before the predetermined period as the distance between the host vehicle and the preceding vehicle increases. The predetermined ratio for weighting the value is increased.

An eighteenth invention is the speed detection device according to any one of the fifteenth to seventeenth inventions,
Means for detecting that the vehicle is traveling on a highway,
The time differential smoothing value calculation means, when the means detects that the host vehicle is traveling on a highway, the time differentiation of the output value and the time differentiation of the output value before the predetermined period. When calculating the smoothed value of the time derivative of the output value by weighting the smoothed value of the output value at a predetermined ratio and weighting the smoothed value of the time derivative of the output value before the predetermined period. It is characterized by increasing the ratio.

  According to the first invention, the smoothing filter device is configured to accurately smooth the value near the center of the amplitude of the fluctuation of the time-series data without causing a time delay unlike the conventional smoothing filter. Therefore, it is necessary to accurately grasp the change tendency of the input value of the data and accurately reflect the change tendency of the data in the smoothed output value.

  In the present invention, the time derivative of the output value smoothed by the time derivative smoothed value calculation means is further smoothed, and the overall change tendency of the input value is grasped from the smoothed value. Then, if the addition ratio determining means is in the same direction as the overall change tendency of the input value ascertained from the smoothed value of the time derivative of the output value, that is, if the change tendency of the input value is decreasing, for example, the physical quantity data The addition ratio is determined so that the addition ratio of the input values in the decreasing direction increases. Finally, time smoothing filtering is performed by the smoothing means based on the addition ratio, the input value is emphasized, and the output value before a predetermined period is weighted and added.

  In this way, it becomes possible to accurately reflect the change tendency of the input value of the physical quantity data to the output value to be smoothed, and to increase or decrease the smoothed output value accurately according to the change tendency of the input value. Is possible. For this reason, time series data rampage does not occur without causing a time delay as in the case of conventional smoothing filters such as moving average filters and time constant filters, and without emphasizing the increase or decrease in data more than necessary as in the case of the Kalman filter. It is possible to obtain an ideal smoothing curve that has been accurately smoothed to a value near the center of the amplitude of.

  According to the second invention, it is determined from the smoothed value of the time derivative of the output value whether the change tendency of the input value tends to increase, decreases or does not change so much. The addition ratio is determined based on a table or the like to which an addition ratio is assigned in advance corresponding to the trend classification. Therefore, in addition to the effects of the present invention, it is possible to easily obtain the addition ratio and to reduce the calculation load.

  According to the third invention, the addition ratio based on the difference between the value obtained by amplifying the smoothed value of the time derivative of the output value representing the overall change tendency of the input value by the predetermined coefficient and the time derivative of the input value. It is possible to determine the addition ratio representing the degree of emphasizing the input value in the time constant filter processing for calculating the output value based on a specific numerical value called the difference, and the input value is determined as the output value. Therefore, the effects of the above-described inventions can be exhibited more accurately.

  According to the fourth invention, if the predetermined coefficient for amplifying the smoothed value of the time derivative of the output value is set, for example, to a too large value, the smoothing value itself of the smoothed value of the time derivative of the output value has At the same time, the value obtained by amplifying the smoothed value of the time derivative of the output value by a predetermined coefficient and the value of the time derivative of the input value are greatly different, and the usefulness of the difference is weakened.

  Therefore, the smoothed value of the time derivative of the output value is set appropriately so that the value obtained by amplifying the smoothed value of the time derivative of the output value by the predetermined coefficient falls within the amplitude of the time derivative of the input value. It is possible to determine the addition ratio based on the difference in a state where the expansion of the rampage of the device itself is within an appropriate range and the usefulness of the difference is maintained, and the effects of the above-described inventions are more accurately exhibited.

  According to the fifth invention, the value obtained by amplifying the smoothed value of the time derivative of the output value by a predetermined positive value coefficient is set to be a value near the end of the amplitude of the time derivative of the input value. . In this case, the input value for which the absolute value of the difference between the time derivative of the input value and the value obtained by amplifying the smoothed value of the time derivative of the output value and the time derivative of the input value becomes small It is an input value that should be emphasized in order to increase or decrease accurately according to the changing tendency. Therefore, it becomes easy to associate the absolute value of the difference with the addition ratio, and the effects of the third invention and the fourth invention are more accurately exhibited.

  According to the sixth invention, the difference and the addition ratio can be associated with each other by a very simple association by determining the addition ratio to be weighted to the input value so as to be larger as the absolute value of the difference is smaller. It becomes possible, and the effect of each said invention is exhibited correctly and easily.

  According to the seventh aspect, the addition ratio determining means includes a table in which an addition ratio is assigned in advance for each absolute value range of the difference divided into a plurality of ranges, and the addition ratio for weighting the input value is determined. By assigning an addition ratio so that the smaller the absolute value of the difference is, the larger the value is, and determining the addition ratio based on the table, the effects of each of the inventions can be exhibited more accurately and easily. Is possible.

  According to the eighth aspect of the present invention, if the addition rate determining means is provided with a function that monotonously decreases the absolute value of the difference, and the addition rate is determined based on the monotonically decreasing function, each of the above inventions It becomes possible to exhibit the effect of this more accurately and more easily.

  According to the ninth invention, in the smoothing filter device according to any one of the first to eighth inventions, the smoothed value of the time derivative of the output value before the predetermined period is stored in the memory when the smoothing filter device is activated. In some cases, the smoothing value of the time derivative of the output value is not stable even if the information such as is not stored. Therefore, in addition to the effects of the inventions described above, when the number of input values is less than the predetermined number of times, the addition ratio is determined to be a predetermined constant value without being varied, and the smoothing filter device is activated. Therefore, it is possible to stabilize the output value and output an appropriate value, and it is possible to improve the reliability of the apparatus.

  According to the tenth aspect of the invention, when calculating the smoothed value of the time derivative of the output value representing the change tendency of the input value, the smoothed value of the time derivative of the output value and the time derivative of the output value before the predetermined period. Is calculated by time constant filtering that weights and adds at a predetermined ratio, so that the smoothed value of the time derivative of the output value can be smoothed without following the rampage of the input value or the time derivative of the input value. It will change. Therefore, it is possible to accurately grasp the change tendency of the input value from the smoothly changing output value and determine the addition ratio, and it is possible to exhibit the effects of the inventions more accurately.

  According to the eleventh aspect of the invention, it is needless to say that the smoothing filter device can smooth and output the physical quantity data measured and inputted by the measuring means according to the respective inventions. It is also possible to smooth and output the data of the change rate of the physical quantity formed by time-differentiating the measured physical quantity. Even in such a configuration, it is possible to obtain an ideal smoothing curve that is smoothed accurately to a value near the center of the amplitude of the fluctuation of the physical quantity change rate data, as in each of the inventions. It becomes possible, and the effect of each of the above-mentioned inventions is accurately exhibited with respect to the data of the change rate of the physical quantity.

  According to the twelfth aspect, for example, when the relative speed between the host vehicle and the preceding vehicle, which is a change speed obtained by time differentiation of the measured value of the distance between the host vehicle and the preceding vehicle, is smoothed, When the distance to the vehicle is large, the relative speed rampage is large and the relative acceleration rampage, which is the time derivative, is also large.However, as the distance between the host vehicle and the preceding vehicle decreases, the relative speed rampage decreases and the relative acceleration Rampage may also be reduced. In such a case, the smoothed value of the time derivative of the output value amplified by a predetermined coefficient set along the end portion of the relative acceleration rampage is increased from the end portion of the rampage as the distance becomes smaller. The usefulness of the difference may be weakened.

  Therefore, in such a case, if the predetermined coefficient is made to be variable so as to become smaller as the distance between the host vehicle and the preceding vehicle, which is the “other physical quantity”, is smaller, the predetermined coefficient It is possible to prevent the smoothed value of the time derivative of the output value amplified in step 1 from moving away from the end of the relative acceleration fluctuation, and to maintain the usefulness of the difference. Is more effective.

  According to the thirteenth invention, the smoothing filter device according to the first to tenth inventions is a vehicle such as an automobile or a truck, a railway vehicle, an aircraft, a transportation means such as a ship, a robot, various machines or devices, and the like. The present invention can be applied to a speed detection device that detects the speed of an object that is considered by moving itself, moving another object, or moving its own internal components.

  Then, if the speed detection device is configured to perform arithmetic processing on the speed data input from the measuring means for measuring the speed of the object and output the smoothed speed, it originally has a rampage. The speed data can be output as a speed that is smoothed to a value near the center of the amplitude of the fluctuation of the data, and the speed of the object can be accurately detected by the speed detection device. .

  According to the fourteenth invention, the smoothing filter device of the eleventh or twelfth invention is a vehicle such as an automobile or truck, a railway vehicle, an aircraft, a transportation means such as a ship, a robot, various machines or devices, and the like. The present invention can be applied to a speed detection device that grasps a distance from another object and detects the speed of the object by differentiating the grasped distance with respect to time.

  Then, the speed detection device calculates the speed of the object by time-differentiating the distance data input from the measuring means as another physical quantity, performs arithmetic processing on the speed, and outputs a smoothed speed. With this configuration, it becomes possible to output the speed data that originally had the rampage as a speed that is accurately smoothed to a value near the center of the amplitude of the rampage of the data. It is possible to accurately detect the speed.

  According to the fifteenth aspect, in the speed detection device, even when the speed to be smoothed is the vehicle speed of the host vehicle, the effects of the inventions are effectively exhibited.

  According to the sixteenth aspect of the invention, in the speed detection device described above, the effects of the inventions are effectively exhibited even when the speed to be smoothed is the relative speed between the host vehicle and the preceding vehicle.

  According to the seventeenth aspect, when the smoothed speed is the relative speed between the host vehicle and the preceding vehicle, the relative acceleration is greater when the distance between the host vehicle and the preceding vehicle is large than when the distance is small. If the predetermined ratio is kept constant, the smoothed value of the time derivative of the output value may be relatively large.

  Therefore, with the configuration of the seventeenth invention, it is possible to reduce the fluctuation of the smoothed value of the time derivative of the output value as compared with the case where the predetermined ratio is kept constant, and the time of the output value It becomes possible to accurately grasp the change tendency of the data of the relative speed by smoothly changing the smoothing value of the differentiation. Therefore, the function of the smoothing filter device is accurately exhibited, and the effects of the respective inventions regarding the speed detection device using the smoothing filter device are more accurately exhibited.

  According to the eighteenth aspect, when the host vehicle is traveling on a highway, the degree of increase or decrease in relative speed between the host vehicle and the preceding vehicle is much greater than when traveling on a normal road. The relative acceleration tends to change relatively stably. Therefore, if the speed detection device is provided with means for detecting that the host vehicle is traveling on the highway, and the configuration of the eighteenth aspect of the invention, the fluctuation of the smoothed value of the time derivative of the output value can be reduced. Thus, the smoothing value of the time derivative of the output value can be changed smoothly, and the change tendency of the relative speed data can be accurately grasped. Therefore, the function of the smoothing filter device is accurately exhibited, and the effects of the respective inventions regarding the speed detection device using the smoothing filter device are more accurately exhibited.

  Embodiments of a smoothing filter device and a speed detection device using the same according to the present invention will be described below with reference to the drawings.

[First Embodiment]
First, the smoothing filter device will be described. The smoothing filter device 1 is composed of a microcomputer in which a CPU, ROM, RAM and the like (not shown) are connected to a bus.

  As shown in FIG. 1, the smoothing filter device 1 according to the present embodiment mainly includes an addition ratio determination unit 2, a smoothing unit 3, a time differential smoothed value calculation unit 4, and a memory 5. It is configured. The smoothing filter device 1 will be described below when data of a predetermined physical quantity X is input via the I / O interface 6 as an input value Xin (t) in a time-sequential and discrete manner at a predetermined sampling period. After performing the arithmetic processing, the smoothed value of the physical quantity X data is output as an output value Xout (t).

  Here, in the present embodiment, the variable t is a variable representing the sampling timing, and in the following, t represents the sampling timing at which the input value Xin (t) of the time series data of the predetermined physical quantity X is input this time. , T−1 represents the sampling timing one cycle before the previous sampling timing, that is, the current sampling timing t.

  Hereinafter, the configuration and arithmetic processing of the addition ratio determining means 2, the smoothing means 3, and the time differential smoothed value calculating means 4 of the smoothing filter device 1 will be described with reference to the flowchart shown in FIG.

  First, the flow of the arithmetic processing of the entire apparatus will be outlined. The smoothing filter device 1 determines the addition ratio P based on the time differentiation of the data of the physical quantity X by the addition ratio determination means 2 (steps S2 to S5). The data of the physical quantity X is smoothed by weighting addition of the current input value Xin (t) and the previous output value Xout (t-1) using the addition ratio P in the conversion means 3, and the current output value Xout (t). Is calculated (step S6).

  That is, the smoothing means 3 smoothes the data of the physical quantity X by using the time constant filter method as shown in the following formula (3). At this time, the time constant is the same as that of the conventional time constant filter. That is, instead of fixing the addition ratio P to a constant value, the addition ratio P is made variable by the addition ratio determination means 2 based on the time differentiation of the data of the physical quantity X input this time, and is made variable at each sampling timing t. Therefore, smoothing is appropriately performed.

  Further, the time differential smoothed value calculating means 4 of the smoothing filter device 1 determines the smoothed value dXs (t−1) of the time differential of the previous output value Xout (t−1) in order to determine the next addition ratio P. ) Is calculated (steps S7 to S9).

  This will be described in detail below. As will be described later, the memory 5 stores the previous output value Xout (t-1) and the smoothed value dXs (t-1) of the time derivative of the output value calculated in the previous calculation process. Yes.

When the input value Xin (t) of the data of the physical quantity X is input at the current sampling timing t (step S1: YES), the addition ratio determination unit 2 obtains the previous output value Xout (t-1) from the memory 5. After reading, the time derivative dXin (t) of the input value is calculated according to the following equation (1) (step S2). Here, Δt represents a sampling period.
dXin (t) = (Xin (t) −Xout (t−1)) / Δt (1)

  When the input physical quantity X data is data having time-series fluctuation as shown in FIG. 23, for example, the time derivative dXin (t) of the calculated input value is also time-series as shown in FIG. It becomes a waveform with a rough rampage.

  Subsequently, the addition rate determination means 2 reads the smoothed value dXs (t-1) of the time derivative of the output value calculated at the previous sampling timing from the memory 5, and obtains the smoothed value dXs (t-1). Amplification is performed with a predetermined coefficient K (step S3 in FIG. 2). Here, in the present embodiment, the coefficient K is a positive value.

  In the present embodiment, the predetermined coefficient K is set in advance. The degree of fluctuation of the input value Xin (t) of the physical quantity X data input to the smoothing filter device 1 varies depending on the measuring device such as a sensor that measures the physical quantity X data, and the smoothing filter device 1 is applied. It is necessary to set appropriately according to the measurement device to be used.

  As data of the physical quantity X, for example, data having a time series fluctuation as shown in FIG. 23 is input from a certain measuring device, and the time differential dXin (t) of the input value Xin (t) as shown in FIG. 3 is obtained. Suppose that it is obtained. In this case, as shown in FIG. 4, the value dXs (t) × K obtained by amplifying the smoothed value dXs (t) of the time derivative of the output value Xout (t−1) by the coefficient K is the time derivative dXin (t It is preferable that the value of the coefficient K is set so as to be within the upper and lower ends of the fluctuation amplitude of

  In the present embodiment, in particular, the smoothed value dXs (t) multiplied by the coefficient K, that is, dXs (t) × K is in a state substantially along the lower end of the amplitude of the fluctuation of the time derivative dXin (t) of the input value. The value of the coefficient K is set.

Subsequently, the addition ratio determining means 2 continuously calculates the time derivative dXin (t) of the input value calculated by the above expression (1) according to the following expression (2) and the smoothed value dXs (t−t) of the time derivative of the previous output value. A difference D (t) from a value dXs (t−1) × K obtained by amplifying 1) by a predetermined coefficient K is calculated (step S4 in FIG. 2).
D (t) = dXin (t) −dXs (t−1) × K (2)

  Subsequently, the addition ratio determining means 2 determines an addition ratio P used for weighted addition in the smoothing means 3 described later based on the calculated difference D (t) (step S5). In the present embodiment, the addition rate determination means 2 is configured to determine the addition rate P to be a larger value as the absolute value | D (t) | of the difference D (t) is smaller.

  Specifically, for example, when the difference D (t) has an acceleration dimension, the addition rate determination means 2 divides the difference into absolute values | D (t) | as shown in FIG. If the table to which the addition ratio P is assigned for each range is provided, the addition ratio P can be determined based on the calculated difference D (t) and this table. In the present embodiment, the addition rate determination means 2 includes such a table.

  In FIG. 5, G represents standard gravity acceleration. Further, instead of preparing a table as in the present embodiment, for example, a monotone decreasing function having the absolute value | D (t) | of the difference as a variable is provided, and the addition rate P is based on the monotonic decreasing function. It can also be configured to determine

  The addition ratio determination unit 2 transmits the addition ratio P determined as described above to the smoothing unit 3.

The smoothing means 3 uses the addition ratio P determined by the addition ratio determination means 2 to weight and add the current input value Xin (t) and the previous output value Xout (t-1) according to the following equation (3). Thus, the data of the physical quantity X is smoothed to calculate the current output value Xout (t) (step S6 in FIG. 2).
Xout (t) = Xout (t-1) * (1-P) + Xin (t) * P (3)

  The smoothing means 3 outputs the current output value Xout (t) calculated in this way (step S6) and sends it to the time differential smoothed value calculating means 4.

  The time differential smoothing value calculation means 4 is used for the next calculation processing in the addition ratio determination means 2, and based on the output value Xout (t) calculated by the smoothing means 3, the time differentiation of the output value described above. The smoothed value dXs (t-1) is calculated.

Specifically, the time differential smoothed value calculating unit 4 firstly outputs the current output value Xout (t) transmitted from the smoothing unit 3 and the previous output value Xout (t−1) read from the memory 5. Are used to calculate the time derivative dXout (t) of the output value according to the following equation (4) (step S7). Note that Δt represents the sampling period as in the case of the above-described equation (1).
dXout (t) = (Xout (t) −Xout (t−1)) / Δt (4)

  Subsequently, the time differential smoothed value calculating means 4 is based on the calculated time differential dXout (t) of the output value and the smoothed value dXs (t-1) of the time differential of the previous output value read from the memory 5. The smoothed value dXs (t) of the time differentiation of the current output value is calculated (step S8).

  In this calculation of the smoothed value dXs (t) of the time differentiation of the output value, various smoothing methods can be used. Smoothing filters such as the above-described Kalman filter, moving average filter, time constant filter, etc. Can be used.

In the present embodiment, the time differential smoothed value calculating means 4 is based on the time differential dXout (t) of the output value calculated this time and the smoothed value dXs (t−1) of the time differential of the previous output value. The smoothing value dXs (t) of the time derivative of the current output value is calculated by weighted addition with a time constant filter in which the time constant shown in the following equation (5) is fixed.
dXs (t) = dXs (t−1) × 0.95 + dXout (t) × 0.05 (5)

  The time differential smoothing value calculation means 4 calculates the current output value Xout (t) and the calculated time differential smoothing value dXs (t) of the current output value for the calculation processing in the next addition ratio determination means 2. Therefore, “previous output value Xout (t−1)” and “smooth value dXs (t−1) of time derivative of previous output value” are overwritten and saved in the memory 5 respectively (step S9). ).

  As can be seen from the configuration of the smoothing filter device 1 of the present embodiment, the previous output value Xout (t−1) or the previous output value is stored in the memory 5 when the smoothing filter device 1 is started. In some cases, the smoothed value dXs (t-1) of the time derivative is not stored, or the smoothed value dXs (t-1) of the time derivative of the previous output value is not stable.

  Therefore, in the present embodiment, the addition rate determination means 2 sets the addition rate P to 0, for example, while the number of input values Xin (t) of the input physical quantity X data is less than a predetermined number of times set in advance. It is determined to be a predetermined constant value such as .1.

  Next, the operation of the smoothing filter device 1 according to this embodiment will be described.

  In the smoothing filter device 1 according to the present embodiment, the smoothing means 3 smoothes the data of the physical quantity X using the time constant filter technique as shown in the above equation (3) (step S6). ), The time constant of the filter used at that time, that is, the addition ratio P is varied by the addition ratio determination means 2 based on the time differentiation of the data of the physical quantity X (steps S2 to S5). Further, the time differential smoothed value calculating means 4 calculates the time differential smoothed value dXs (t) of the output value for the next calculation process (steps S7 to S9).

  As described above, in the conventional smoothing filter such as the moving average filter and the time constant filter, the time delay with respect to the ideal smoothing curve A passing near the center of the amplitude of the data fluctuation as shown in FIG. Data of physical quantity X is smoothed with ΔT. In order to eliminate such a time delay and ideally smooth the data showing a decreasing tendency or an increasing tendency as shown by the smoothing curve A in FIG. 24, (i) the input value Xin (t ) Is a decreasing trend, an increasing trend, or a tendency that does not change so much, it accurately captures the changing trend of the input value Xin (t), and (ii) the changing trend of the data is smoothed It is necessary to accurately reflect the output value Xout (t).

  In the smoothing filter device 1 according to the present embodiment, first, the change tendency of the input value Xin (t) of (i) is determined by the arithmetic processing of steps S7 and S8 in the time differential smoothed value calculation means 4 described above. The time derivative dXout (t) of the output value Xout (t) output from is calculated, and is further smoothed to calculate and track the smoothed value dXs (t-1).

  It seems seemingly contradictory to determine the change tendency of the input value Xin (t) by the time differentiation of the output value Xout (t), but as shown in FIG. 23, the input value Xin (t) shows a time-series rampage. If it has, the time differential dXin (t) also becomes positive or negative at each sampling timing t as shown in FIGS. The time derivative dXin (t) cannot be used as it is for determining the change tendency of the input value Xin (t).

  Therefore, in this embodiment, the output value Xout (t) obtained by smoothing the value of the input value Xin (t) is used, and the input value Xin (t) is obtained from the smoothed value dXs (t) of the time differentiation dXout (t). The overall trend of change is determined.

  Next, as a method of accurately reflecting the change tendency of the input value Xin (t) determined as described above in (ii) on the output value Xout (t), the smoothing filter device 1 according to the present embodiment uses A technique is used in which the value of the addition ratio P in the above equation (3) is varied in consideration of the change tendency of the input value, that is, the smoothed value dXs (t) of the time derivative of the output value representing it.

  Specifically, the smoothed value dXs (t) representing the overall change tendency of the input value Xin (t) described in (i) is amplified by a predetermined coefficient K, and the current input value Xin (t) The difference D (t) with respect to the time derivative dXin (t) representing the temporal change of the difference is calculated, and the addition ratio P is determined according to the absolute value | D (t) | of the difference.

  In this embodiment, in particular, as shown in FIG. 4, the coefficient K is set so that K times the smoothed value dXs (t) is along the lower end of the amplitude of the fluctuation of the time derivative dXin (t) of the input value. The value is set, and the smaller the | D (t) |, the larger the addition ratio P, and the larger | D (t) |, the smaller the addition ratio P.

  That is, for example, as in the region R2 in FIG. 4, the smoothed value dXs (t) representing the overall change tendency of the input value Xin (t) becomes a negative value, and the input value Xin (t) tends to decrease overall. In this case, the time derivative dXin (t) of the current input value is close to dXs (t) × K, and is largely close to the lower end of the ramping amplitude of the time derivative of the input value dXin (t). When the input value Xin (t) is increased, the contribution ratio P of the input value Xin (t) to the output value Xout (t) is increased.

  If the time derivative dXin (t) of the input value is a negative value, the current input value Xin (t) is often smaller than the previous output value Xout (t-1), so this input value Xin (t) To increase the contribution of the input value Xin (t) to the output value Xout (t), thereby increasing the current output value Xout ( t) can be reduced with respect to the previous output value Xout (t-1).

  That is, in the present embodiment, the smoothed value dXs (t) representing the overall change tendency of the input value Xin (t) is a negative value, and the input value Xin (t) shows a general decreasing tendency. In addition, by increasing the addition ratio P of the input value Xin (t) to increase the contribution of the input value Xin (t) to the output value Xout (t), the decreasing tendency of the output value Xout (t) is further increased. To be emphasized. Therefore, as shown in FIG. 6, the output value Xout (t) is pushed down to the decreasing side from the output value B smoothed when the addition ratio P is fixed, and an ideal smoothing curve A is drawn. It becomes like this.

  On the other hand, when the input value Xin (t) has a tendency not to change as a whole as in the region R1 of FIG. 4, for example, the smoothed value dXs (t) of the time derivative of the output value is a value close to zero. DXs (t) × K is also close to 0. Therefore, as the time differential dXin (t) of the current input value is closer to 0, the absolute value | D (t) | of the difference becomes smaller and the addition ratio P becomes larger. That is, the contribution of the current input value Xin (t), which is not so different from the previous output value Xout (t-1), to the output value Xout (t) is increased.

  Further, even if the time derivative dXin (t) of the input value is greatly increased in the positive or negative direction, | D (t) | is large and the contribution of the input value Xin (t) to the output value Xout (t) is small. Since the appearance frequency is about the same in both positive and negative directions, the influence of the input value Xin (t) on the output value Xout (t) where the time derivative dXin (t) is greatly increased in the positive or negative direction is small. It becomes.

  Therefore, when the input value Xin (t) shows a tendency not to change as a whole as in the region R1 in FIG. 4, even if the input value Xin (t) is violent in time series, the output value Xout (t) continues to take on a value that is less rampant and does not change much overall.

  On the other hand, when the overall change tendency of the input value Xin (t) changes from a decreasing tendency to a tendency that does not change so much as in the region R3 of FIG. 4, for example, the time derivative dXin (t) increases in the positive direction. The value of the addition ratio P of the input value Xin (t) that comes out is not so large.

  However, since the frequency of inputting the input value Xin (t) having a value larger than the previous output value Xout (t-1) is increased as compared with the region R2 that showed a decreasing tendency, the above equation (3) Accordingly, the output value Xout (t) gradually changes from a decreasing tendency to a tendency that does not change so much, and the smoothed value dXs (t) of the time derivative of the output value approaches zero.

  Thus, the smoothing filter device 1 according to the present embodiment shows a tendency that the smoothed output value Xout (t) does not change so much when the input value Xin (t) does not change so much. When the input value Xin (t) is in a decreasing trend, the smoothed output value Xout (t) also shows a decreasing trend, and the smoothing that decreases following the decreasing trend of the input value Xin (t) ideally It is possible to output the output value Xout (t).

  In the above example, the case where the overall change tendency of the input value Xin (t) of the data of the physical quantity X is decreasing as shown in FIG. 23 has been described. For example, as shown in FIG. Even when the overall change tendency of the input value Xin (t) of the X data tends to increase, ideal smoothing of the output value Xout (t) can be similarly performed without changing the configuration. .

  That is, when the overall change tendency of the input value Xin (t) is increasing as shown in FIG. 7, the smoothed value dXs (t) of the time derivative of the output value is also positive as shown in FIG. Takes the value of At this time, as shown in FIG. 4, when the input value Xin (t) has a tendency to decrease as a whole and the smoothed value dXs (t) takes a negative value as in the present embodiment, dXs (t). If the value of the coefficient K is set so that xK is substantially along the lower end of the amplitude of the fluctuation of the time derivative dXin (t) of the input value, the smoothed value dXs (t) also takes a positive value. In this case, as shown in FIG. 8, dXs (t) × K is almost in line with the upper end of the amplitude of the fluctuation of the time derivative dXin (t) of the input value.

  If the difference D (t) is calculated in the same manner as described above and the relationship between the absolute value | D (t) | and the addition ratio P is set, the input value Xin (t) tends to increase overall. In this case, as shown in FIG. 9, the output value Xout (t) is pushed upward from the output value B smoothed when the addition ratio P is fixed, and the ideal smoothing curve A To draw. Further, the case where the input value Xin (t) tends to change little or the case where the input value Xin (t) that tends to increase tends to change little is explained in the same manner as in the case of the decreasing tendency. The input value Xin (t) can be smoothly smoothed.

  As described above, according to the smoothing filter device 1 according to the present embodiment, the smoothed value dXs (t) obtained by smoothing the time derivative dXout (t) of the output value is calculated, and the data of the physical quantity X is based on the calculated value. It is possible to accurately grasp the changing tendency of the input value Xin (t). At the same time, for example, the addition ratio P is determined based on the difference between the value obtained by amplifying the smoothed value dXs (t) by a predetermined coefficient K and the time derivative dXin (t) of the input value. By determining the addition ratio P used to calculate the output value Xout (t) based on the time derivative dXin (t) and the smoothed value dXs (t), the change tendency of the data input value Xin (t) is smoothed. The output value Xout (t) can be accurately reflected.

  As a result, the smoothed output value Xout (t) can be accurately increased / decreased in accordance with the changing tendency of the input value Xin (t), which is similar to a smoothing filter such as a conventional moving average filter or time constant filter. As shown in the ideal smoothing curve A shown in FIG. 24, the time-series data having a fluctuation can be accurately smoothed to a value near the center of the amplitude of the data fluctuation without causing a time delay. It becomes possible.

  Further, if the predetermined coefficient K is appropriately set or the relationship between the absolute value | D (t) | of the difference and the addition ratio P is appropriately set, the data increase / decrease more than necessary as in the conventional Kalman filter. The smoothing that prevents the smoothing that the output value Xout (t) does not exactly match the changing tendency of the input value Xin (t) and that can output an ideal smoothing curve A It becomes possible to make a filter device.

  Further, in the moving average filter, it is necessary to store the past input values Xin (t−1) to Xin (tn) for a predetermined number of samplings for averaging in the memory 5, but in this embodiment, The memory 5 stores only the previous output value Xout (t-1) and the smoothed value dXs (t-1), which is a secondary effect that the usage efficiency of the memory 5 can be reduced. There is also an effect.

  The method for determining the addition ratio P, that is, the method for setting the predetermined coefficient K, and the method for setting the relationship between the difference D (t) calculated based on the coefficient K and the addition ratio P are the methods of this embodiment. It is not limited to.

  As can be seen from the above description, in extreme terms, the change of the input value Xin (t) of the data of the physical quantity X is increasing, decreasing, or even not changing so much. For example, it is possible to perform the smoothing accurately by changing the way of changing the addition ratio P accordingly.

  Therefore, for example, the addition ratio determination means 2 is used to change the input value Xin (t) based on the time differentiation smoothing value dXs (t−1) of the output value previously calculated by the time differentiation smoothing value calculation means 4. It is also possible to make a configuration such that the addition ratio P is determined by changing the method of changing the addition ratio P based on, for example, a table as shown in FIG.

  In the present embodiment, the smoothing filter device 1 that smoothes the input physical quantity X data itself has been described. The smoothing filter device 1 includes, for example, speed and relative speed data input from a measuring unit that measures the speed of the host vehicle or a measuring unit that measures a relative speed between a device equipped with the smoothing filter device and another object. X is smoothed, and the smoothed speed and relative speed are output as an output value Xout (t).

  However, for example, distance Y data is input from a measuring unit that measures the movement distance of a certain movable device, and the input value Yin (t) of the input distance Y data is time-differentiated to speed Xin (t). It is also possible to constitute a smoothing filter device that obtains the calculated velocity Xin (t) and outputs the smoothed velocity as the output value Xout (t).

  Further, for example, distance Y data is input from a measuring unit that measures the distance between a certain device and another object, and the input value Yin (t) of the input distance Y data is time-differentiated to obtain a relative velocity Xin ( It is also possible to configure as a smoothing filter device that obtains t), smoothes the calculated relative velocity Xin (t), and outputs the smoothed relative velocity as the output value Xout (t).

  That is, the smoothing filter device forms the data of the physical quantity X to be smoothed from the data of the physical quantity Y different from the data of the physical quantity X to be output after being smoothed, and smoothes the data of the physical quantity X It is also possible to configure as described above. For example, when another physical quantity Y data is input, the data Y is time-differentiated to calculate the physical quantity data X that is the changing speed of the physical quantity Y, and the physical quantity X data that is the changing speed of the physical quantity Y is smoothed. It is also possible to configure so as to output in the form.

  In this case, the flowchart shown in FIG. 2 is modified like the flowchart shown in FIG. That is, the previous input value Yin (t-1) is stored in the memory 5 in addition to the previous output value Xout (t-1) and the smoothed value dXs (t-1) of the time differentiation of the output value. When the input value Yin (t) of the physical quantity Y at the current sampling timing t is input in step S10 instead of step 1 (step S10: YES), the addition ratio determination means 2 of the smoothing filter device. The previous input value Yin (t−1) is read from the memory 5.

Then, the input value Yin (t) is time-differentiated according to the following formula (6) to calculate the change rate dYin (t) of the data of the physical quantity Y, and the change rate dYin (t) is the target of subsequent processing. The physical quantity is Xin (t) (step S11). Note that Δt represents a sampling period.
Xin (t) = dYin (t) = (Yin (t) −Yin (t−1)) / Δt (6)

  Xin (t) is calculated and smoothed by the addition ratio determining means 2, the smoothing means 3, and the time differential smoothed value calculating means 4 by the above-described arithmetic processing (steps S2 to S9), and the smoothed output. The value Xout (t) is output.

  On the other hand, the addition ratio determination means 2 overwrites and saves the input value Yin (t) of the data of the physical quantity Y input this time as “previous input value Yin (t−1)” in the memory 5 (step S12), and next time. Is used for calculating the time derivative dYin (t) of the input value, that is, the physical quantity Xin (t).

  As described above, even when the data Xin (t) of the physical quantity X is the data Yin (t) of the change rate of the other physical quantity Y calculated by time differentiation of the other physical quantity Y, basically, the above-described implementation is performed. The addition ratio determining means 2 and smoothing means 3 are simply added to the form by performing an arithmetic process of differentiating another physical quantity Y with respect to time and using the calculated change rate Yin (t) as the physical quantity Xin (t) to be processed. Since the other calculation processing in the time differential smoothed value calculation means 4 is performed in exactly the same manner as described above, the effect of the above-described embodiment is also exhibited in the present modification.

  In the above example, the configuration in which the output value Xout (t) is calculated every time the data Xin (t) of the physical quantity X is input has been described. However, the sampling period in which the data Xin (t) of the physical quantity X is input. However, for example, the output value Xout (t) may be calculated every two sampling periods.

[Second Embodiment]
In the present embodiment, a speed detection device using the smoothing filter device according to the first embodiment will be described.

  In the present embodiment, the relative speed between the host vehicle and the preceding vehicle is obtained by time-differentiating the distance between the host vehicle and the preceding vehicle, which is measured by the measuring means described later, using the modification of the first embodiment described above. A speed detection device that calculates, smoothes and detects and outputs the relative speed by a smoothing filter device will be described.

  As shown in FIG. 12, the speed detection apparatus 10 according to the present embodiment includes a measurement unit 18 mainly including an imaging unit 11, a conversion unit 12, an image processing unit 15, and a smoothing filter unit 22. And means 19.

  The configuration from the image pickup means 11 to the three-dimensional object detection means 20 of the detection means 19 is disclosed in Japanese Patent Laid-Open Nos. H5-114099, H5-265547, and H6-266828 previously filed by the applicant of the present application. Detailed descriptions are given in Japanese Patent Laid-Open No. 10-283461, Japanese Patent Laid-Open No. 10-283477, Japanese Patent Laid-Open No. 2006-72495, and the like, and detailed descriptions thereof are left to those publications. A brief description is given below.

  In this embodiment, the image pickup means 11 includes a pair of main cameras each including a built-in image sensor such as a CCD or a CMOS sensor that are synchronized with each other, and attached to the vicinity of a room mirror at a predetermined interval in the vehicle width direction, for example. 11a and a sub camera 11b. The stereo camera is configured to capture a landscape including a road ahead of the vehicle at a predetermined sampling period and output a pair of images. Of the pair of cameras, the main camera 11a is a camera closer to the driver, and captures a reference image T as shown in FIG. Note that an image captured by the sub camera 11b is referred to as a comparative image.

  Image data output from the main camera 11a and the sub camera 11b is converted from analog images into digital images having luminance values for each pixel by the A / D converters 12a and 12b, which are conversion means 12, and the image correction unit 13 Image correction such as displacement and noise removal is performed and stored in the image data memory 14 and transmitted to the detecting means 19 at the same time.

  The image processor 16 of the image processing means 15 performs a stereo matching process and a filtering process on the image data of the reference image T and the comparison image to calculate the parallax dp corresponding to the distance in the real space. Hereinafter, an image to which the parallax dp is assigned is referred to as a distance image. The parallax dp information thus calculated, that is, the distance image, is stored in the distance data memory 17 of the image processing means 15.

Note that the parallax dp, the point (i, j) on the distance image, and the point on the road surface just below the center of the pair of cameras 11a and 11b are the origin, the vehicle width direction of the host vehicle is the X-axis direction, The point (X, Y, Z) in the real space when the direction is the Y-axis direction and the vehicle length direction is the Z-axis direction is uniquely determined by coordinate transformation represented by the following equations (7) to (9) It is matched. In the following equations, CD is the distance between the pair of cameras, PW is the viewing angle per pixel, CH is the height of the pair of cameras, and IV and JV are on the distance image at the infinity point in front of the host vehicle. The i coordinate, j coordinate, and DP represent the vanishing point parallax.
X = CD / 2 + Z * PW * (i-IV) (7)
Y = CH + Z × PW × (j−JV) (8)
Z = CD / (PW × (dp−DP)) (9)

  In the present embodiment, the distance Z from the host vehicle to the three-dimensional object existing in the predetermined area ahead of the host vehicle, from the above imaging unit 11 to the image processing unit 15 including the image processor 16 and the distance data memory 17. The measuring means 18 for measuring the parallax dp uniquely associated with the distance Z by the equation (9) is configured.

  In the present embodiment, the measuring means 18 may be any means as long as it can measure the distance Z to the three-dimensional object existing in front of the host vehicle. In addition to the present embodiment, for example, laser light, infrared light, etc. in front of the host vehicle. And a radar device that measures the distance Z to the three-dimensional object based on the information of the reflected light, and the detection method is not limited to a specific method.

  The detecting means 19 is composed of a microcomputer in which a CPU, ROM, RAM, input / output interface and the like (not shown) are connected to a bus. The detection means 19 is connected to sensors Q such as a vehicle speed sensor, a yaw rate sensor, and a steering angle sensor for measuring the steering angle of the steering wheel. It should be noted that a device for estimating the yaw rate from the vehicle speed of the host vehicle or the like can be used instead of the yaw rate sensor.

  As shown in FIG. 12, the detection means 19 includes a three-dimensional object detection means 20, a preceding vehicle detection means 21, and the smoothing filter device 22 described above, and further includes a memory (not shown). In addition, necessary data from the sensors Q is input to each means of the detection means 19.

  In the present embodiment, the three-dimensional object detection means 20 is configured based on a vehicle exterior monitoring device described in JP-A-10-283461 as described above. Detailed explanations are left to those publications. The configuration will be briefly described below.

  The three-dimensional object detection means 20 reads the distance image described above from the distance data memory 17 and divides the distance image into strip-shaped sections extending in the vertical direction with a predetermined pixel width. Then, each parallax dp belonging to each strip-shaped section is converted into a distance Z according to the equation (9), and a histogram is created with respect to a distance that is positioned above each road Z as being above the road surface. The distance to the maximum section is the distance to the three-dimensional object in the strip-shaped section. This is done for all categories. Hereinafter, a distance representing each section is referred to as a distance Z.

  For example, when the distance Z is calculated for the distance image created from the reference image T shown in FIG. 13 and the calculated distance Z for each section is plotted on the real space, as shown in FIG. The three-dimensional object ahead of the host vehicle is plotted as each point with some variation in the portion corresponding to the portion facing the host vehicle MC.

  The three-dimensional object detection means 20 groups the points plotted in this way into groups G1 to G7 based on the distance and directionality of the adjacent points as shown in FIG. As shown in FIG. 16, sub-groups in which each point in each group is arranged substantially parallel to the vehicle width direction of the host vehicle MC, that is, the X-axis direction, are labeled “objects” O1 to O3. Subgroups arranged side by side in the traveling direction, that is, substantially parallel to the Z-axis direction, are classified as “sidewalls” S1 to S4. Further, the intersection point of the “object” and “side wall” of the same three-dimensional object is labeled as a corner point C.

  In this way, the three-dimensional object detection means 20 sets [object O1, corner point C, sidewall S1], [sidewall S2], [object O2], [object O3], [sidewall S3], and [sidewall S4], respectively. It is detected as a three-dimensional object. The three-dimensional object detection means 20 stores information on the three-dimensional object detected in this way, the coordinates of the end points of each subgroup, and the like in a memory.

  The preceding vehicle detection means 21 detects the three-dimensional object closest to the own vehicle as the preceding vehicle among the three-dimensional objects existing on the traveling path of the own vehicle or on the traveling lane in which the own vehicle is traveling. Yes. In the present embodiment, the preceding vehicle detection means 21 is configured to detect a preceding vehicle on the traveling path of the own vehicle. First, a vehicle speed sensor, a yaw rate sensor, and a steering angle sensor, which are sensors Q, are input. Based on the behavior of the vehicle, that is, the vehicle speed V and yaw rate γ of the host vehicle, the steering angle δ of the steering wheel, etc., the turning curvature Cua of the host vehicle is calculated, and the travel locus Lest of the host vehicle MC is calculated as shown in FIG. It is like that.

The turning curvature Cua is obtained by using, for example, the vehicle speed V and the yaw rate γ.
Cua = γ / V (10)
Is calculated according to Further, for example, using the vehicle speed V and the steering angle δ,
Re = (1 + Asf · V ² ) · (Lwb / δ) (11)
Cua = 1 / Re (12)
It can also be calculated according to Here, Re is a turning radius, Asf is a vehicle stability factor, and Lwb is a wheelbase.

  The preceding vehicle detection means 21 grasps a region for the vehicle width of the host vehicle centered on the travel locus calculated in this way as the traveling path Rest of the host vehicle, and the host vehicle MC among the three-dimensional objects existing thereon. The object O2 closest to is detected.

  The preceding vehicle detection means 21 further reads out the information of the preceding vehicle detected last time from the memory, the positional relationship between the preceding vehicle detected last time and the three-dimensional object closest to the own vehicle detected this time, that is, the object O2, the moving speed, etc. Based on the above, the probability that the previous preceding vehicle and the detected three-dimensional object are the same three-dimensional object is calculated, and if the calculated probability is greater than or equal to a preset threshold value, the detected three-dimensional object is determined as the preceding vehicle. By labeling, detecting the preceding vehicle, updating the information on the preceding vehicle with the information of the three-dimensional object detected this time and continuously registering it in the memory, the information on the preceding vehicle is updated while tracking it Yes.

  As described above, the preceding vehicle detection means 21 may be configured to detect the preceding vehicle as a three-dimensional object that exists on the traveling lane in which the host vehicle MC is traveling and is closest to the host vehicle MC. It is. As an apparatus for detecting the position of a lane such as the left and right lane markings of the host vehicle MC necessary for detecting the traveling lane of the host vehicle MC, for example, Japanese Patent Laid-Open No. 2001-92970 filed earlier by the applicant of the present application. The lane recognition device described in the publication can be used.

  The smoothing filter device 22 is inputted with the preceding vehicle detected by the preceding vehicle detection means 21, that is, in this case, the distance Z between the object O2 and the host vehicle MC. As the distance Z of the object O2 in this case, for example, the distance Z in the Z-axis direction from the host vehicle MC at the center position of the object O2 extending in the vehicle width direction, that is, the X-axis direction is input.

  The smoothing filter device 22 performs the arithmetic processing detailed in the first embodiment, and finally the smoothed relative speed between the host vehicle MC and the preceding vehicle O2 is finally detected by the speed detection device 10. The relative speed is output outside the apparatus. Specifically, according to the flowchart shown in FIG. 11, the addition ratio determining means 2, the smoothing means 3, and the time differential smoothed value calculating means 4 perform arithmetic processing, and the vehicle MC that is the output value Xout (t) and the preceding vehicle MC A relative speed with respect to the vehicle O2 is output.

In this case, the distance Z between the object O2 and the host vehicle MC corresponds to the other physical quantity Yin (t) in the first embodiment, and the relative speed ΔV that is the time derivative of the distance Z is the other physical quantity Yin (t ) Corresponding to the time differential dYin (t), that is, Xin (t), and the output smoothed relative velocity ΔV corresponds to the output value Xout (t). In the present embodiment, the time derivative dXin (t) of the input value used to calculate the output value Xout (t), the time derivative dXout (t) of the output value, and the smoothed value dXs (t) of the time derivative of the output value. ) Each have a dimension of acceleration [m / s ² ].

  Next, the operation of the speed detection device 10 according to the present embodiment will be described.

  In the configuration of the speed detection apparatus 10 according to the present embodiment as described above, the distance Z between the host vehicle and the preceding vehicle is actually continuously smooth due to the influence of the resolution of the imaging unit 11 such as a CCD camera. Despite the change, the distance Z = Yin (t−1) detected at a certain sampling timing t−1 and the distance Z = Yin (t) detected at the next sampling timing t are relatively large. Often detected as changed. In such a case, the relative speed ΔV at which the time differentiation Xin (t) = dYin (t) changes relatively greatly, and the data of the relative speed ΔV becomes data having a rampage in time series.

  Actually, when the distance Z = Yin (t) between the host vehicle and the preceding vehicle is measured by the speed detection device 10 having the configuration of the present embodiment, data that changes relatively smoothly is obtained as shown in FIG. However, when the time differential dYin (t) = Xin (t), which is the relative velocity ΔV, is calculated according to the above equation (6), time series data Xin (t) having a relatively large fluctuation as shown in FIG. 19 is obtained. .

  Further, when the addition ratio determining means 2 of the smoothing filter device 22 calculates the time derivative dXin (t) of Xin (t) corresponding to the relative acceleration between the host vehicle and the preceding vehicle according to the above equation (1), FIG. As shown in FIG. 4, time series data dXin (t) having a large fluctuation is obtained. Further, the K times of the smoothed value dXs (t) of the time derivative of the calculated output value is substantially along the lower end of the amplitude of the fluctuation of the time derivative dXin (t) of Xin (t) as shown in FIG. Transition to.

  In the addition ratio determining means 2 of the smoothing filter device 22, as shown in FIG. 21, the addition increases or decreases sharply according to the absolute value of the difference D (t) between dXin (t) and dXs (t−1) × K. The ratio P is determined. Then, based on the addition rate P that fluctuates at each sampling timing t, that is, the time constant, the smoothing means 3 of the smoothing filter device 22 sets the relative speed ΔV by the time constant filter expressed by the above equation (3). Xin (t), which is time series data having a corresponding fluctuation, is smoothed, and an output value Xout (t) smoothed to a value near the center of the amplitude of the data fluctuation as shown in FIG. 22 is output. .

  In FIG. 22, Xcom (t) represents, as a comparative example, the calculation result of a conventional time constant filter in which the addition ratio P in the time constant filter of equation (3) is fixed to 0.1. .

  As described above, according to the speed detection device 10 according to the present embodiment, the smoothing filter device 22 accurately exhibits the effect of the first embodiment, and is obtained from the distance Z between the host vehicle and the preceding vehicle. The time series data of the relatively large relative speed ΔV = Xin (t) thus obtained is effectively smoothed, and as shown in FIG. 22, the smoothness is accurately smoothed to a value near the center of the data fluctuation amplitude. It is possible to output the relative speed ΔV = Xout (t).

  Therefore, if the output value Xout (t) of the speed detection device 10 is applied to, for example, a preceding vehicle following device, appropriate preceding vehicle following control can be performed based on an appropriate relative speed ΔV between the preceding vehicle and the host vehicle. For example, when control is performed on the basis of the output value Xcom (t) of the conventional time constant filter shown in FIG. It is possible to effectively prevent forcing unnecessary acceleration in order to keep up.

  In the second embodiment, the case where the relative speed ΔV is obtained from the distance Z between the own vehicle and the preceding vehicle when outputting the smoothed relative speed ΔV between the own vehicle and the preceding vehicle has been described. However, the present invention is not limited to this case. For example, the relative speed ΔV between the host vehicle and the preceding vehicle is directly measured by measurement means such as Doppler radar, and the measured relative speed ΔV is smoothed by the smoothing filter device 22. This also applies to the case of detecting and outputting.

  In that case, the arithmetic processing in the smoothing filter device 22 of the speed detection device is performed according to the flowchart shown in FIG. In the flowchart of FIG. 2, the input value Xin (t) is the relative speed ΔV between the host vehicle and the preceding vehicle, and the output value Xout (t) is the relative speed ΔV between the smoothed host vehicle and the preceding vehicle. .

  The present invention is not limited to the smoothing of the relative speed ΔV between the host vehicle and the preceding vehicle as described above, and can be applied to the smoothing of the vehicle speed V of the host vehicle, for example. As a means for measuring the vehicle speed V of the host vehicle, for example, a pulse detection sensor for calculating the vehicle speed V by detecting the rotation of the wheel of the host vehicle described in Japanese Patent Application Laid-Open No. 2004-317206 previously filed by the present applicant. Can be used.

  Further, in the second embodiment, the case where the speed detection device is mounted on a vehicle such as an automobile or a truck has been described. Any means that can measure the speed of an object by moving itself or moving other objects, such as a transportation means such as a ship, a robot, various machines and devices, etc. Is also applicable.

  By the way, when it is set as the structure which calculates | requires relative speed (DELTA) V from the distance Z of the own vehicle and a preceding vehicle like said 2nd Embodiment, the distance per pixel of a distance image is so large that the distance Z of the own vehicle and a preceding vehicle is large. The difference in distance tends to increase and the detection error of the detected distance Z tends to increase. Therefore, as shown in FIG. 18 and FIG. 19, as the distance Z, which is “another physical quantity”, is larger, the “physical quantity data”, that is, the time derivative Xin (t), which is to be smoothed, that is, the host vehicle The relative speed ΔV between the vehicle and the preceding vehicle shows a characteristic tendency that the rampage increases, and that the relative speed ΔV decreases as the distance Z decreases.

  As a result, as shown in FIG. 20, the dXin (t) representing the relative acceleration also decreases as the distance Z decreases, and when the coefficient K is constant as in the case of FIG. At the sampling timing t when the amplitude of the relative acceleration dXin (t) starts to decrease and the amplitude of the relative acceleration dXin (t) starts to decrease, dXs (t) × K is slightly separated from the lower end of the amplitude of the relative acceleration dXin (t).

  As described above, when the amplitude of the relative acceleration dXin (t) becomes smaller as the distance Z, which is the “other physical quantity”, is smaller, the predetermined coefficient K is made smaller as the value of the other physical quantity is smaller. It is preferable to make it variable so as to be a small value. As a fitting method, as described above, the distance Z is divided into predetermined ranges, and a table in which the value of the coefficient K is set for each section is prepared, or the value of the coefficient K is calculated as a function of the distance Z. It can be configured as follows.

  Further, as shown in FIG. 20, when the distance Z is large, the relative acceleration dXin (t) becomes relatively large, and therefore the smoothed value dXs of the time derivative of the output value serving as a reference for determining the addition ratio P. As for (t), when the distance Z is large, it may have a relatively large rampage compared with the case where the distance Z is small.

  For this reason, in this embodiment, the time constant in the calculation of the smoothed value dXs (t) of the time differentiation of the output value is constant as shown in the above equation (5), but the time constant is variable and As the distance Z to the vehicle increases, the first term on the right side in the equation (5), that is, the coefficient of dXs (t-1) is further increased to reduce the fluctuation of the smoothed value dXs (t). The smoothing value dXs (t) is preferably calculated.

  Further, as another viewpoint, when the host vehicle is traveling on a highway, the degree of increase or decrease in relative speed between the host vehicle and the preceding vehicle is not so large compared to when traveling on a normal road. The relative acceleration dXin (t) tends to change relatively stably. In such a case, in the same manner as described above, in the above equation (5) for calculating the smoothed value dXs (t) of the time derivative of the output value which is the reference for determining the addition ratio P, the first on the right side The term, that is, the coefficient of dXs (t-1) can be modified to be further increased to reduce the fluctuation of the smoothed value dXs (t).

  Therefore, the speed detection device 10 is provided with means for detecting that the host vehicle is traveling on the highway, and when this means detects that the host vehicle is traveling on the highway, It is preferable that the smoothing value dXs (t) is calculated by correcting the coefficient of dXs (t-1) to be larger in the equation (5).

  Examples of the means include detection from map information based on information from a navigation device equipped with a GPS (Global Positioning System) receiver or the like, or the vehicle speed of the host vehicle is equal to or higher than a predetermined speed such as 80 km / h. The vehicle travels on the highway by recognizing that the condition has continued for a predetermined time or by measuring the distance between the left and right lane positions of the vehicle and recognizing that the distance is greater than the predetermined distance. It can be configured to detect

It is a block diagram which shows the structure of the smoothing filter apparatus which concerns on 1st Embodiment. It is a flowchart which shows the procedure of the arithmetic processing in a smoothing filter apparatus. It is a graph showing the time differentiation dXin (t) of the calculated input value. It is a graph showing the smoothed value dXs (t) of the time differentiation of the calculated output value, and the value multiplied by K times. It is a figure showing the example of the table by which the addition ratio was allocated to each range of the absolute value of a difference. It is a graph showing the output value at the time of fixing an addition ratio, and an ideal smoothing curve. It is a graph showing the input value of the data of the physical quantity which is increasing. It is a graph showing the smoothed value dXs (t) of the time differentiation of the output value based on the input value of FIG. It is a graph showing the output value and ideal smoothing curve at the time of fixing the addition ratio in case an input value tends to increase. It is a figure showing the example of the table which matches the change tendency of an input value, and an addition rate. It is a flowchart which shows the procedure of the arithmetic processing for smoothing the time differentiation of another physical quantity. It is a block diagram which shows the structure of the speed detection apparatus which concerns on 2nd Embodiment. It is a figure which shows an example of a reference | standard image. It is a figure showing each point which plotted the distance for every division on real space. It is a figure showing each group at the time of grouping each point of FIG. It is a figure showing the object and side wall which were detected based on each group of FIG. It is a figure explaining the driving | running | working locus | trajectory and traveling path of the own vehicle. It is a graph showing the example of the measured distance. It is a graph showing the relative speed calculated based on the distance of FIG. FIG. 20 is a graph showing dXin (t) corresponding to the relative acceleration calculated based on the relative velocity in FIG. 19 and K times the smoothed value dXs (t) of the time derivative of the output value. It is a graph showing the addition ratio determined based on FIG. It is a graph showing the output value calculated based on the addition ratio of FIG. It is a graph showing the time series data of the relative speed which is in the decreasing tendency. FIG. 24 is a graph showing a smoothing curve when the data of FIG. 23 is ideally smoothed and a curve when smoothing by a conventional smoothing filter.

Explanation of symbols

1, 22 Smoothing filter device 2 Addition ratio determining means 3 Smoothing means 4 Time differential smoothed value calculating means 10 Speed detecting device 18 Measuring means D (t) Difference dXin (t) Time differential dXout (t) output of input value Time derivative of value dXs (t) Smoothed value K of time derivative of output value K Coefficient MC Own vehicle O2 Preceding vehicle P Addition ratio Δt Sampling period V Vehicle speed ΔV Relative speed X Physical quantity Xin (t) Input value Xout of physical quantity data t) Output value Y of physical quantity data Other physical quantity Z Distance

Claims (18)

In a smoothing filter device for smoothing and outputting data of physical quantities that are discretely input in time series,
An addition rate determining means for determining an addition rate;
Smoothing means for outputting, as an output value of the smoothed physical quantity data, a value calculated by weighting and adding the input value of the physical quantity data input this time and the output value of a predetermined period before by the addition ratio; ,
A time derivative smoothing value calculating means for calculating a time derivative of the output value and calculating a smoothing value of the time derivative of the output value;
The addition ratio determining means calculates a time derivative of the input value, and based on the calculated time derivative of the input value and the smoothed value of the time derivative of the output value calculated by the time derivative smoothed value calculating means. A smoothing filter device that determines the addition ratio.   The addition ratio determining means determines a change tendency of the input value based on the smoothed value of the time derivative of the output value calculated by the time derivative smoothed value calculating means, and the determined change tendency of the input value The smoothing filter device according to claim 1, wherein the addition ratio is determined based on a relationship with time differentiation of the input value.   The addition ratio determining unit calculates a difference between a value obtained by amplifying the smoothed value of the time derivative of the output value calculated by the time derivative smoothed value calculating unit by a predetermined coefficient and the time derivative of the input value, The smoothing filter device according to claim 1, wherein the addition ratio is determined based on the difference.   4. The predetermined coefficient is set so that a value obtained by amplifying a smoothed value of a time derivative of the output value by the predetermined coefficient falls within an amplitude of a time derivative of the input value. The smoothing filter device described in 1.   The predetermined coefficient is a positive value, and a value obtained by amplifying the smoothed value of the time derivative of the output value by the predetermined coefficient becomes a value near the end of the amplitude of the time derivative of the input value. The smoothing filter device according to claim 3, wherein the smoothing filter device is set.   The said addition ratio determination means determines the said addition ratio weighted to the said input value so that it may become a larger value, so that the absolute value of the said difference is small. The smoothing filter device according to item.   The addition ratio determination means includes a table in which an addition ratio is assigned in advance for each absolute value range of the difference divided into a plurality of ranges, and determines the addition ratio based on the table. The smoothing filter device according to claim 6.   The smoothing filter device according to claim 6, wherein the addition ratio determining unit includes a monotone decreasing function for the absolute value of the difference, and determines the adding ratio based on the monotone decreasing function.   The said addition ratio determination means determines the said addition ratio to the preset fixed value, while the frequency | count of input of the said input value is less than predetermined number of times, The any one of Claims 1-8 characterized by the above-mentioned. The smoothing filter device according to one item.   The time differential smoothed value calculating means weights and adds the time differential of the output value and the smoothed value of the time differential of the output value before the predetermined period by a predetermined ratio, and the output value of the current cycling period The smoothing filter device according to any one of claims 1 to 9, wherein a smoothing value of a time derivative of is calculated.   The smoothing according to any one of claims 1 to 10, wherein the physical quantity data is data of a change rate of the other physical quantity calculated by differentiating the other physical quantity with respect to time. Filter device. The physical quantity data is data of the change rate of the other physical quantity obtained by differentiating the other physical quantity with time,
11. The smoothing filter device according to claim 3, wherein the predetermined coefficient is varied so as to be smaller as the value of the other physical quantity is smaller. A speed detection device that detects the speed of an object using the smoothing filter device according to any one of claims 1 to 10,
The physical quantity data is speed data input from a measuring unit that measures the speed of the object, and outputs the smoothed speed. A speed detection device that detects the speed of an object using the smoothing filter device according to claim 11 or 12,
The other physical quantity is a distance input from a measuring unit that measures the distance of the object, and the physical quantity data is data of the speed of the object obtained by time differentiation of the distance, and the smoothed data A speed detection device that outputs a speed.   The speed detection apparatus according to claim 13 or 14, wherein the speed is a vehicle speed of the host vehicle.   The speed detection apparatus according to claim 13 or 14, wherein the speed is a relative speed between the host vehicle and a preceding vehicle.   The time differential smoothed value calculating means weights and adds the time differential of the output value and the smoothed value of the time differential of the output value before the predetermined period by a predetermined ratio to smooth the time differential of the output value When calculating the value, the larger the distance between the host vehicle and the preceding vehicle, the larger the predetermined ratio weighted to the smoothed value of the time derivative of the output value before the predetermined period. The speed detection device according to claim 16. Means for detecting that the vehicle is traveling on a highway,
The time differential smoothing value calculation means, when the means detects that the host vehicle is traveling on a highway, the time differentiation of the output value and the time differentiation of the output value before the predetermined period. When calculating the smoothed value of the time derivative of the output value by weighting the smoothed value of the output value at a predetermined ratio and weighting the smoothed value of the time derivative of the output value before the predetermined period. The speed detection apparatus according to any one of claims 15 to 17, wherein the ratio is increased.

Images