JP3646501B2 - Vehicle dangerous driving judgment device - Google Patents

Description

[0001]
BACKGROUND OF THE INVENTION
The present invention relates to a vehicle dangerous driving determination device, and more particularly to a device that determines a dangerous driving when the running stability of the vehicle is lowered.
[0002]
In recent years, safety awareness has increased in society, and safety devices that prevent traffic accidents have been demanded. Dangerous driving such as snoozing, fatigue, and looking away is one of the main causes of serious accidents. . If such dangerous driving can be detected, accidents can be prevented, and many research institutions and automobile companies are working on these dangerous driving determination devices.
[0003]
[Prior art]
Conventional dangerous driving determination devices include (1) directly determining the driver's dangerous state, and (2) indirectly determining the driver's dangerous state by determining the state in which the driving stability of the vehicle has decreased. It is known that it is estimated automatically.
[0004]
The former {circle around (1)} is, for example, as described in Japanese Patent Laid-Open No. 1-250221, a doze driving determination device or a fatigue determination device, and conventionally known determination methods use physiological information such as a driver's brain waves and blinks. It is what is used.
[0005]
The latter (2) is, for example, a dozing detection device disclosed in Japanese Patent Laid-Open No. 52-25336. In this dozing detection device, when the arousal level is low, the steering wheel operation of the driver is reduced and becomes messy, and minute correction steering is not performed and the vehicle flutters left and right. That is, the deviation frequency of the steering wheel steering angle with respect to a predetermined detection angle (running stability = meandering degree) is detected and compared with a reference frequency (threshold value) expected at awakening to determine whether normal driving or dozing driving. A traveling stability decrease determining means for determining is used.
[0006]
In any method, the detection accuracy of the snoozing driving is important for the dozing detection sensitivity of the sensor used and robustness against noise such as disturbance.
[0007]
[Problems to be solved by the invention]
However, in the case of the above method (2), if the traveling stability reduction determining means is mounted on an actual vehicle as it is as a dangerous driving determination device for a vehicle, the vehicle traveling stability reduced state and the driver's dangerous state are It occurs in scenes that do not necessarily correspond one-on-one.
In other words, even though the driver's condition is normal, the actual vehicle has problems with sound, vibration, and heat, and the weather, road surface conditions, traffic conditions, driver characteristics, etc. are all erroneous as disturbances. It becomes a factor of detection.
[0008]
If the number of false detections becomes too great, the reliability of the device will be significantly reduced, and there is a risk that the meaning of the alarm will not be meaningful in a critical situation. Therefore, reducing the number of false alarms is essential for improving the accuracy of the dangerous driving determination device.
[0009]
One way to reduce false alarms is to raise the judgment threshold of the vehicle as a dangerous driving judgment device to make it difficult to issue an alarm, but this will cause delays and leaks in the case of a real dangerous driving. There's a problem. On the contrary, if the judgment threshold value of the dangerous driving judgment device of the vehicle is lowered to make it easy to issue an alarm and the judgment of the dangerous driving can be made earlier, the false alarm will also increase this time.
[0010]
Therefore, techniques for obtaining such a determination threshold are disclosed in, for example, Japanese Patent Application Laid-Open Nos. 60-157927 and 6-107029. In these conventional techniques, a frequency component related to a predetermined dangerous driving is extracted from a signal such as a yaw rate or a steering wheel rotation angle indicating a vehicle behavior state continuously detected at the start of driving of the vehicle, and an absolute value thereof. The threshold value is determined by a value obtained by integrating the values (shaded portions in FIG. 8A) or an average value thereof.
[0011]
However, in this case, as shown in FIG. 1A, the integration value is performed at regular intervals, and each integration range is independent and there is no overlapping portion, so that there is a large variation for each integration value. There is.
[0012]
Furthermore, since the frequency component of the extracted vehicle behavior detection signal is fixed and the threshold value is single according to this, it is not possible to consider the frequency component due to individual differences in doze driving. That is, among the extracted vehicle behavior detection signals, there are individual differences in the frequency at which it is possible to detect that the driver is in a drowsy driving state. There is a problem of erroneous detection that the sign does not appear in the component, and conversely, the sign appears in the extracted frequency component even though the component is not asleep.
[0013]
Therefore, in view of the above-described problems, the present invention obtains the driving stability of the vehicle based on the vehicle behavior detection signal and compares the driving stability with a threshold value to determine the dangerous driving of the vehicle. It is an object of the present invention to eliminate variations in integral values depending on the integral range for obtaining the running stability and to eliminate as much as possible false detections caused by individual differences in sleep driving.
[0014]
[Means for Solving the Problems]
To achieve the above object, the vehicle dangerous driving determination device according to the present invention further includes a filter for extracting a plurality of frequency components related to the dangerous driving of the output signal of the vehicle behavior detecting means, and the driving stability reduction determination Means obtains an average value of the movement integral value within the learning section as the running stability from the data for each frequency component extracted from the filter, determines a plurality of temporary threshold values based on each average value, and sets each temporary threshold value; Is stored as a true threshold value only when it is larger than the maximum movement integral value in the corresponding learning section.
[0015]
That is, in the present invention, movement integration is performed at regular time intervals as shown in FIG. 8 (2) so that the integration ranges partially overlap. Then, an average value of the movement integral values is obtained, and thus a threshold value of the running stability is determined based on the average value. Therefore, since the variation of the integral value is reduced, a more appropriate value can be obtained as the threshold value.
[0016]
However, in this case, as described above, if the threshold value is one, it is not always possible to set a threshold value that suits individual differences among various drivers.
Therefore, in the present invention, a filter is further provided, and a plurality of frequency components related to the dangerous driving of the output signal of the vehicle behavior detecting means are extracted from the filter.
[0017]
The traveling stability decrease determination means determines each threshold value as described above using data for each frequency component extracted from the filter, and sets this threshold value as a temporary threshold value. There are a plurality of temporary thresholds.
[0018]
Then, the maximum movement integral value in each learning section for which these temporary threshold values are obtained is taken out, stored as a true threshold value only when the temporary threshold value is larger than the maximum movement integral value, and deleted when it is smaller.
[0019]
Since the true threshold value obtained in this way is obtained from a plurality of frequency components in the vehicle behavior detection signal even if the drivers are different, it is possible to prevent the driver from erroneously detecting the drowsy driving state.
[0020]
In the present invention, instead of the above threshold value, a standard deviation can be further obtained from the average value of the moving integral values, and the threshold value can be determined based on the standard deviation.
[0021]
That is, when the driver wakes up, for example, when the steering wheel is operated, there are people who hardly move the steering wheel using the play of the steering wheel, and conversely, people who drive while constantly finely adjusting the steering wheel, In both cases, although the width of the change of the movement integral value is different, the average value is almost the same, so the threshold values are equal.
[0022]
Therefore, in the case of the latter driver, the running stability may exceed the threshold value in spite of normal driving, which is troublesome and impairs the reliability of the device.
[0023]
In order to avoid this, in the present invention, after obtaining the average value of the moving integral values, the standard deviation is further obtained. If the threshold value is determined based on the standard deviation, even if the movement integral value of the running stability fluctuates, it is not immediately affected.
[0024]
In the present invention, vehicle speed detection means is further provided, and only when it can be determined from the output signal of the vehicle speed detection means that the vehicle speed has been stable for a predetermined time or more and the vehicle speed has been stable, Means may determine the dangerous driving.
[0025]
Further, the vehicle is further provided with a steady operation detecting means, and the traveling stability lowering judging means judges the dangerous driving only when the steady operation detecting means detects the steady driving state of the vehicle and can judge that the driving state is stable. You may go.
[0026]
Further, a filter that extracts a frequency component related to dangerous driving of the output signal of the vehicle behavior detecting means may be provided, and the driving stability decrease determining means may determine the driving stability from the data extracted from the filter. .
[0027]
Further, the vehicle behavior detection means may be any one of a vehicle yaw direction angular velocity detection means, a steering wheel rotation angle detection means, a lateral acceleration detection means, and a travel locus measurement means.
Further, the above dangerous driving includes a snoozing driving.
[0028]
DETAILED DESCRIPTION OF THE INVENTION
FIG. 1 shows an embodiment of a vehicle dangerous driving determination apparatus according to the present invention. In this embodiment, a snooze driving detection device is used as the dangerous driving determination device. In the figure, reference numeral 1 denotes a steering angle sensor as vehicle behavior detecting means for detecting the behavior of the vehicle. In addition to the steering angle sensor, a vehicle yaw direction angular velocity sensor, a vehicle lateral acceleration sensor, or a travel locus measuring means is provided. In this embodiment, a steering angle sensor is used as a representative example.
[0029]
Further, 2 is a vehicle speed sensor, 3 is a winker operation monitoring sensor for monitoring the winker operation by the driver, and 4 is a brake operation monitoring sensor for monitoring the brake operation by the driver. Reference numeral 5 denotes a signal processing unit (ECU) as a running stability lowering determination unit that inputs the output signals of these sensors 1 to 4 and performs the signal processing shown in FIGS. 5 is an alarm device that issues an alarm in response to the output signal 5.
[0030]
The operation of the embodiment of FIG. 1 will be described below with reference to flowcharts of programs stored and executed in the signal processing unit 5 shown in FIGS.
[0031]
First, these flowcharts are routines of a fixed period that are started when the engine is started. FIG. 2 is a flow for obtaining a plurality of thresholds, FIG. 3 is a learning flow for one threshold, and FIG. It is a dozing driving determination flow.
[0032]
Multiple provisional threshold determination flow (Fig. 2)
In this embodiment, two threshold values are obtained. First, the temporary threshold value Ax is calculated (step S21), and then the temporary threshold value Ay is calculated (step S22).
[0033]
These steps S21 and S22 are subroutines, the details of which are shown in FIG. 3, and will be described below.
[0034]
Threshold learning flow (Figure 3)
First, the calculation of the temporary threshold Ax will be described. Based on the output signal of the vehicle speed sensor 2, the signal processing unit 5 determines whether or not a state of a predetermined vehicle speed or more has continued for a predetermined time (step S1). This is a step for checking whether or not the vehicle speed is stable, and it is because an accurate threshold value cannot be obtained until the vehicle speed is stabilized after starting.
[0035]
Next, even if the vehicle speed is stable, an accurate threshold value cannot be obtained until the driving state is stabilized, so it is checked whether or not the vehicle is in steady operation (step S2). Specifically, the operation statuses of the blinker operation monitoring sensor 3 and the brake operation monitoring sensor 4 are monitored, and it is determined that a predetermined time is not a steady operation after the operation. Further, an operation monitoring sensor such as a headlight may be further added. For example, when driving in a lane or traveling in a tunnel, the driving state changes compared to normal driving. If a threshold value is set at this time, it is impossible to make an accurate doze driving determination.
[0036]
Next, a timer t and a variable n described later are reset (step S3).
Thereafter, the signal processing unit 5 inputs an output signal of the steering angle sensor 1 (step S4). The waveform of this output signal is shown in FIG. The horizontal axis represents time, and the vertical axis represents the steering angle.
[0037]
The steering angle input in this way is passed through a band pass filter (BPF), and smoothing processing is performed to obtain a desired frequency component signal (step S5). This is because, in this embodiment, the frequency component X for detecting the drowsy driving is obtained. The waveform at this time is shown in FIG.
[0038]
In this way, a moving integration process is performed on the data extracted from the bandpass filter (step S6). The moving integration process itself is a well-known technique as disclosed in Japanese Patent Laid-Open No. 8-255690. As shown in FIG. 8 which has already been described, the range of the normal integration in FIG. 1A is performed without overlapping every fixed section, but in the case of the moving integration shown in FIG. Has an integral range that overlaps for the same amount of time.
[0039]
The movement integral value An thus obtained is stored in a memory (not shown) built in the signal processing unit 5 (step S7).
Then, it is determined whether or not the timer t has exceeded a certain learning time T (step S8). Since the learning time T is naturally not initially exceeded, the timer t is incremented by “1” and the variable n is also set to “1”. "" Is incremented (step S10), and the process returns to step S4.
[0040]
In this way, when the processing of step S4 to step S7 is executed by the number of times of movement integration shown in FIG. 8 (2), for example, when the timer t exceeds the learning time T, the temporary threshold Ax is calculated (step S9). ).
[0041]
Here, the provisional threshold Ax can be considered in the following two ways.
(1) First, as a first method, the movement integral values An stored in step S7 are summed for the learning time T and the average value is obtained. Then, a provisional threshold Ax is obtained by multiplying this average value by a certain coefficient (for example, 1.5).
[0042]
That is, as shown in FIG. 6 (1), an average value Aav indicated by a solid line can be obtained through the learning section T, and by multiplying the average value Aav by a coefficient 1.5, a dotted line indicates A temporary threshold value Ax is obtained.
[0043]
(2) The second method is the same as the above method (1) until the average value is calculated by summing the movement integral values An obtained in step S7, but based on this average value. Further, the standard deviation S is obtained, the standard deviation S is multiplied by an arbitrary coefficient (for example, “3”), and a value obtained by adding the above average value Aav to the provisional threshold value.
[0044]
The reason for determining the temporary threshold in this way is as follows.
In the case of the above method (1), as shown in the figure (1), when the movement integral value has a small fluctuation as shown by the waveform (1), or when the fluctuation is large as shown by the waveform (2). Also, the average value Aav shows the same value.
[0045]
As a result, the temporary threshold value Ax exceeds the temporary threshold value Ax in the movement integral waveform {circle around (2)} as shown in the figure, so that the signal processing unit 5 drives the alarm device 6 to generate an alarm.
[0046]
However, the movement integral waveforms (1) and (2) are substantially normal for the driver. However, due to the characteristics of the driver, in the case of the waveform (1), steering of the steering wheel is very small. This is a case of a person, and the waveform (2) only shows the case of a large person.
Therefore, in spite of being normal, in the case of waveform {circle around (2)}, a false alarm (false detection) occurs, which is not only troublesome but also lacks the reliability of the apparatus.
[0047]
Therefore, as shown in the above method (2), if the standard deviation S is further obtained from the average value Aav, and then the standard deviation S is multiplied by a certain multiple and added to the average value Aav, (2) in FIG. As shown in FIG. 5, the waveform (1) does not reach the temporary threshold value Ax1 of the waveform (1), and no false alarm is generated. In the case of the waveform (2), the waveform (2) does not reach the provisional threshold Ax2, and no false alarm is generated.
[0048]
Thus, in the case of the above method (2), by considering only the average value, it is possible to avoid a state in which the movement threshold value is the same even if the movement integration value varies. It is possible to eliminate the possibility of false alarms when there is a large variation in.
[0049]
Although the above learning flow has been described for the temporary threshold Ax, the temporary threshold Ay can be obtained in the same manner.
[0050]
True threshold decision flow (Figure 2)
After obtaining the temporary threshold values Ax and Ay in this manner, the signal processing unit 5 first extracts the maximum value Anx from the movement integral value An stored in step S7 in the learning step in which the temporary threshold value Ax is calculated. And compared with the temporary threshold Ax (step S23).
[0051]
As a result, if Ax> Anx, the temporary threshold value Ax is stored as the true threshold value Ax in the frequency component X (step S24). If Ax> Anx is not satisfied, the temporary threshold value Ax is deleted (step S25). This is because the temporary threshold value Ax that is equal to or less than the maximum value of the movement integral value An is likely to cause erroneous detection as shown by the relationship between the waveform (2) and the threshold value Ax shown in FIG. This is because if the temporary threshold value Ax is larger than the maximum value of the movement integral value An, there is little possibility of causing erroneous detection.
[0052]
Similarly, the temporary threshold value Ay is also compared with the maximum value Any in step S26. If Ay> Any, the temporary threshold value Ay is stored as the true threshold value Ay (step S27). Otherwise, this temporary threshold value is stored. Ay is deleted, and the process returns to step S21 (step S28).
After step S27, the program proceeds to the dozing determination flow (step S29). This subroutine proceeds to step S11 in FIG.
[0053]
FIG. 7 shows an example in which the threshold determination flow in FIG. 2 is not adopted as a temporary threshold. That is, in this example, one temporary threshold value is created, and in the case of FIGS. 7 (1) to (3), the maximum value of the movement integral value in the learning section T is equal to or less than the temporary threshold value. Is the true threshold, but in the example of FIG. 4 (4), the maximum value of the movement integral value in the learning section T is larger than the temporary threshold, and there is a risk of false detection, so it is deleted. The following doze driving determination flow is not delivered.
[0054]
Doze driving determination flow (steps S11 to S16: FIG. 4)
After the threshold calculation is completed as described above, the dozing operation determination flow is executed.
That is, as in steps S4 to S7, the steering angle signal is input (step S11), the smoothing process (step S12), the movement integration process (step S13), and the calculation process (step S13) of the movement integration value Bn. S14) is executed.
[0055]
Then, the movement integral value Bn obtained in this way is compared with the true threshold value Ax (or Ay) stored in step S24 (or S27) (step S15). In other words, the processing is performed on the true threshold values Ax and Ay obtained in steps S24 and S27, and when more threshold values are actually obtained, the processing is executed accordingly.
[0056]
As a result, when Bn> Ax (Ay) is not satisfied, the process returns to step S11 to perform the next movement integration calculation. When Bn> Ax (Ay) is satisfied, the signal processing unit 5 drives the alarm device 6 to output an alarm output. (Step S16). Thereafter, the process returns to step S11, and the alarm routine is continuously executed.
[0057]
In the above embodiment, the two frequency components X and Y have been described. However, the frequency component is not limited to this and can be applied to various components.
[0058]
When the travel locus measuring means for obtaining the meandering deviation amount from the travel locus of the vehicle is used as the vehicle behavior detecting means, the image input means cited as the travel locus measuring means, as is well known, on the road on which the vehicle travels. The image including the lane is taken in, and the taken image is subjected to image processing by a signal processing means (not shown) to specify the position of the lane.
[0059]
Then, if the relative position deviation between the vehicle center and the lane center is obtained in the signal processing unit 5, the traveling locus of the vehicle can be known.
The running stability is obtained from the integral value of the absolute value of the running locus thus obtained.
[0060]
Further, as the travel locus measuring means, a magnetic signal input means can be used in addition to the image input means. Examples of a device that emits a magnetic signal include a magnetic coil that is installed continuously on a road in advance.
[0061]
In this case, if the signal of the magnetic coil is detected by the magnetic signal input means of the vehicle and the relative position between the vehicle and the magnetic coil is continuously obtained as is well known by the signal processing means, the traveling locus of the vehicle can be obtained. it can.
[0062]
Also, radio signal input means can be used instead of image input means. GPS satellites are examples of devices that emit radio signals. The signal processing means obtains the position of the vehicle in the same manner as the position detection is well known in the navigation system. The subsequent operation principle is the same as that when the image input means is used.
[0063]
【The invention's effect】
As described above, according to the dangerous driving determination device for a vehicle according to the present invention, a plurality of frequency components related to the dangerous driving of the vehicle behavior detection signal are extracted from the filter, and the data for each frequency component is moved and integrated within the learning interval. To determine the average value, determine a plurality of temporary threshold values based on each average value, and use only as a true threshold value when each temporary threshold value is greater than the maximum moving integral value in the corresponding learning interval. Because it is configured, it is possible to eliminate variation when integrating or averaging the vehicle behavior detection signal data, making it possible to make more accurate dangerous driving judgments and setting an accurate threshold regardless of individual differences among drivers. Can be set.
[Brief description of the drawings]
FIG. 1 is a block diagram showing an embodiment of a vehicle dangerous driving determination device according to the present invention.
FIG. 2 is a flowchart (No. 1) of a control program executed by a signal processing unit as a running stability decrease determining unit used in the dangerous driving determining apparatus for a vehicle according to the present invention.
FIG. 3 is a flowchart (No. 2) of a control program executed by a signal processing unit as a traveling stability decrease determining unit used in the dangerous driving determining apparatus for a vehicle according to the present invention.
FIG. 4 is a flowchart (No. 3) of a control program executed by a signal processing unit serving as a traveling stability decrease determining unit used in the dangerous driving determination device for a vehicle according to the present invention.
FIG. 5 is a waveform diagram before and after passing an actual steering angle signal obtained through the vehicle dangerous driving determination device according to the present invention through a band-pass filter;
FIG. 6 is a waveform diagram for explaining how to obtain a threshold value in the dangerous driving determination apparatus for a vehicle according to the present invention.
FIG. 7 is a waveform diagram for explaining a relationship between a plurality of temporary threshold values and the maximum movement integral value in the learning section in the dangerous driving determination device for a vehicle according to the present invention.
FIG. 8 is a diagram illustrating the principle of movement integration.
[Explanation of symbols]
DESCRIPTION OF SYMBOLS 1 Steering angle sensor 2 Vehicle speed sensor 3 Winker operation monitoring sensor 4 Brake operation monitoring sensor 5 Signal processing part (ECU)
6 In the alarm device diagrams, the same reference numerals indicate the same or corresponding parts.

Claims (7)

The vehicle risk is determined by the travel stability lowering determination means based on the detection signal of the vehicle behavior detection means for detecting the behavior of the vehicle, and determining the dangerous driving by comparing the travel stability with a threshold value. In the driving determination device,
The vehicle behavior detecting means further includes a filter for extracting a plurality of frequency components related to the dangerous driving, and the driving stability reduction determining means uses the driving stability from the data for each frequency component extracted from the filter. The average value of the movement integral value within the learning interval is determined, and a plurality of temporary threshold values are determined based on each average value, and only true when each temporary threshold value is greater than the maximum movement integral value in the corresponding learning interval. A dangerous driving determination device for a vehicle, characterized in that it is stored as a threshold. In claim 1,
A vehicle dangerous driving determination device characterized in that a standard deviation is further obtained from the average value and a threshold value determined based on the standard deviation is used instead of the threshold value based on the average value. In claim 1 or 2,
The vehicle further comprises a vehicle speed detecting means, and the running stability decrease determining means determines the dangerous driving only when a vehicle speed of a predetermined value or more continues from the output signal of the vehicle speed detecting means for a predetermined time or more. Dangerous driving judgment device. In claim 3,
Further, the vehicle is provided with a steady operation detection means, and the vehicle stability reduction judgment means judges the dangerous driving only when the steady operation detection means detects the steady operation state of the vehicle. apparatus. In any one of Claims 1 thru | or 4,
The vehicle behavior detecting means further comprises a filter for extracting a frequency component related to the dangerous driving, and the traveling stability reduction determining means obtains the traveling stability from the data extracted from the filter. A vehicle dangerous driving judgment device. In any one of Claims 1 thru | or 5,
A dangerous driving determination device for a vehicle, wherein the vehicle behavior detection means is any one of a vehicle yaw direction angular velocity detection means, a steering wheel rotation angle detection means, a lateral acceleration detection means, and a travel locus measurement means. In any one of Claims 1 thru | or 6.
A dangerous driving determination device for a vehicle, wherein the dangerous driving is a drowsy driving.

Images