JP7320187B2 - Driver posture detector - Google Patents

Description

The present invention relates to a driver posture detection device provided in a vehicle such as an automobile.

2. Description of the Related Art In recent years, in the driving of vehicles such as automobiles, there have been a series of traffic accidents due to sudden changes in the driver's physical condition during driving. According to statistics, about 10% of automobile accidents in Japan (30% in the case of commercial vehicles) are caused by changes in the driver's physical condition. In the analysis of traffic accidents caused by such sudden changes in the driver's physical condition (endogenous diseases, etc.), the driver's state immediately before the accident was affected by the collapse of the head posture due to loss of muscle tone, and the inability to face forward while driving. collapse has been found, and it is speculated that these are signs that an abnormality has occurred in the driver.

Detection of the driver's posture is essential for determining such driver anomalies, but with recent advances in the functionality of imaging means and the speeding up of numerical processing units, advanced image processing functions can be used at low cost. Therefore, by performing the recognition processing of the captured image of the driver in real time, it is possible to accurately recognize the driver state from the driver image, and based on this, highly accurate judgment of the abnormal state of the driver is performed. It is becoming possible to do so. Therefore, in recent years, various techniques have been proposed for detecting the driver's posture using a camera or sensor installed in the vehicle and detecting an abnormality of the driver from the driver's posture.

For example, in Patent Document 1 (Japanese Patent Application Laid-Open No. 2017-77817), a reference range of the head position of the driver of the vehicle is set, and in the image of the driver acquired by the image acquisition unit, the head position is specified. An invention has been proposed in which it is determined that there is an abnormal state when it does not fall within the reference range of . In addition, in Patent Document 2 (Japanese Patent Application Laid-Open No. 2017-206173), in the image of the driver captured by the imaging unit, when the angle of the forward leaning posture of the driver of the vehicle exceeds a predetermined threshold angle An invention for judging the state has been proposed.

JP 2017-77817 A JP 2017-206173 A

However, the in-vehicle camera for capturing the image of the driver used in such technology is limited in its imaging range due to the limited mounting location in the vehicle. For this reason, when the driver makes a large change in posture, the driver's head, which is necessary for determining the posture, is out of the shooting range of the in-vehicle camera.

FIG. 5 shows an example of detection results obtained by a conventional driver posture detection device. In the figure, the upper graph shows the detected vertical angle of the driver's head (face), and the lower graph shows the front-rear position of the driver's head (face). As can be seen from the figure, each detection result has missing parts (broken parts in the graph). If such a missing measurement occurs in the detection of the driver's state, real-time abnormality determination cannot be performed during that time, and the accuracy of abnormality determination decreases.

SUMMARY OF THE INVENTION It is an object of the present invention to provide a driver posture detection apparatus capable of appropriately estimating the driver's head state even when the driver's head state detection is missing. is.

In order to achieve the above object, the present invention adopts the following solutions. That is, as described in claim 1, in a driver posture detection device provided in a vehicle, a photographing unit for photographing an image of the driver, and based on the image photographed by the photographing unit, the position of the driver's head is detected. and/or a head state calculation unit that calculates a measured value of the driver's head state, which is the posture, and if the measured value of the driver's head state cannot be calculated by the head state calculation unit, the driver's head state is measured. The calculated value of the measured value of the driver's head state immediately before the value becomes uncalculable is the assumed head state, which is the predicted value at the time when the predetermined predicted recovery time has elapsed from the time when the driver's head state became uncalculable. value, and using the calculated value of the driver's head state before the head state of the driver becomes uncalculable and the assumed head state value, during the period when the calculation is impossible by the head state calculation unit a head state estimating unit for estimating a driver's head state, wherein the head state estimating unit calculates two measured values of the driver's head state immediately before the driver's head state becomes uncalculable. and the assumed head state value, the driver's head state is estimated.

According to the above-described solution method, even during a period (missing time) in which the calculation (detection) of the measured value of the driver's head state by the driver's head state calculation unit is impossible, the driver's head state estimation unit detects the driver's head state can be appropriately estimated, processing using the driver's head state (for example, abnormality determination) can be continuously executed in real time. Further, in estimating the driver's head state, the estimation is performed on the assumption that the driver's head state returns to the assumed head state value after the return prediction time elapses. Estimates can be obtained. Further, interpolation for obtaining the estimated value of the head state can be performed with high accuracy.

Preferred modes based on the above-described solution method are as described in claims 2 and subsequent claims. That is, as the time during which the driver's head state cannot be calculated increases, the predicted return time is extended (corresponding to claim 2 ). In this case, even if the missing measurement time of the driver's head state is long, the appropriate driver's head state can be estimated.

An abnormality determination section is provided for determining an abnormality in the driver's posture using the calculated value by the driver's head state calculator or the estimated value by the driver's head state estimator (corresponding to claim 3 ). In this case, even if the driver's head condition is not measured, the abnormality can be determined based on the estimated value of the driver's head condition. can be judged appropriately.

According to the present invention, it is possible to appropriately estimate the driver's head state even during a period when the driver's head state cannot be detected, so it is possible to accurately detect a driver's abnormality in real time based on the driver's posture.

1 is a block configuration diagram showing an example of a control system of the present invention; FIG. FIG. 4 is a diagram for explaining an outline of driver posture detection according to the present invention; 4 is a flowchart showing an example of control of the present invention; The figure which shows the measurement result in the driver posture detection to which this invention is applied. The figure which shows the detection result in the conventional driver posture detection.

BEST MODE FOR CARRYING OUT THE INVENTION An embodiment of the present invention will be described below with reference to the accompanying drawings.

FIG. 1 is a block diagram showing an example of a control system of the driver posture detection device of the present invention. The driver posture detection device includes an in-vehicle camera 1 and a controller U as shown. The in-vehicle camera 1 is a photographing means capable of photographing the driver in the vehicle interior, photographs the state of the driver including the head (face) (for example, the driver's upper body), and transmits the photographed image to the controller U.

The controller U is a control device configured using a microcomputer, and includes a head condition calculator 11, a head condition estimator 12, and an abnormality determiner 13 as processing means (programs). .

The head state calculation unit 11 analyzes the image information of the driver acquired from the in-vehicle camera 1 to obtain the measured value of the driver's head state (specifically, the position and posture (angle) of the driver's head). calculate. Specific driver head states include, for example, the vertical angle of the driver's head (face), the front-back position of the driver's head (face), and the like (see FIG. 4). Calculation of the measured value of the driver's head condition by the head condition calculator 11 is repeated at predetermined calculation intervals.

The head condition estimating unit 12 detects the driver's head condition when the head condition calculating unit 11 cannot calculate the driver's head condition (for example, when the driver's head is out of the photographable range of the in-vehicle camera 1). When the driver's head cannot be recognized in the image information from the driver's head), an estimated value of the driver's head state is calculated. The details of the calculation of the estimated value by the head state estimator 12 will be described later with reference to FIG.

The abnormality determination unit 13 detects an abnormality in the driver's posture based on the calculated value of the driver's head state calculated by the head state calculation unit 12 or the estimated value of the driver's head state estimated by the head state estimation unit 13. It is determined whether or not has occurred. Specifically, when the calculated value or estimated value of the driver's head condition exceeds a predetermined threshold value (abnormality judgment value) and continues beyond a predetermined abnormality judgment time, the driver's posture is abnormal (physical condition It is determined that a change in posture, etc.) has occurred.

Next, details of estimation of the driver's head state by the head state estimation unit 12 will be described with reference to FIG. FIG. 2 is a graph plotting calculated values (measured values) or estimated values of the driver's head state at each calculation time or estimation time. In the graph of FIG. 2, points (plots) C1 to C4 indicate the calculation of the driver's head state by the head state calculator 11, and the calculation (detection) at the point C4 is the last to calculate the driver's head state. is disabled.

When the driver's head state becomes incalculable in this way, the driver state estimating unit 12 first calculates the calculated value immediately before the driver's head state becomes incalculable (that is, the detected value of the driver's head state at point C4). The assumed head state value (detection resumption predicted value) is set to X, and after a predetermined recovery prediction time T has elapsed from the measurement immediately before the driver's head state becomes unrecoverable (point C4), the driver's head state Assume that the hypothetical head state value X is restored. That is, it is assumed (predicted) that the state of the driver's head (the position and posture of the head) suddenly changes for some reason, but returns to the original state after a predetermined return prediction time T elapses. In FIG. 2, a return prediction point R1 indicates a point on the graph where the return is assumed.

Subsequently, using the two calculation points C3 and C4 immediately before the driver's head state cannot be calculated and the return prediction point R1, the value of the driver's head state in the section from point C4 to point R1 is calculated. , interpolate using the second-order least-squares method. Then, an estimated value of the driver's head state after a predetermined detection interval t has elapsed from the point C4 is obtained as a value at the estimated point E1 on the interpolated curve V1.

If the state in which the driver's head state cannot be calculated continues, the initial return prediction time T is lengthened by the duration of the state in which the calculation is not possible. That is, when the next detection timing arrives while the driver's head state cannot be calculated, the predicted return time T is lengthened by the detection interval t, and the extended predicted return times (T+t, T+2t, T+3t) are , . . . ) Assuming that a return is made at the assumed head state value X at the predicted return points (R2, R3, R4, . Repeat interpolation. The figure shows a case where five estimation points E1 to E5 (interpolated curves V1 to V5) are determined using return prediction points R1 to R5 to determine an estimated value of the driver's head state. When the posture of the driver returns and the calculation of the driver's head condition becomes possible again, the calculation of the driver's head condition is resumed as indicated by calculation points C5 to C7.

Thus, according to this example, the three points necessary for determining the parameters of the second-order least-squares method are the two detection points (C3, C4) immediately before the driver's head state becomes undetectable and the return prediction point ( Since the estimated value of the driver's head state is obtained by using the three points R1 to R5), highly accurate interpolation can be performed and an appropriate estimated value can be obtained. In addition, since extrapolation is performed using upwardly convex curves (V1 to V5), the extrapolated values will not diverge even if the state of the drive head cannot be detected continuously (for example, Extrapolation that does not use the return prediction point does not always yield a convex curve, and the extrapolation value diverges.) As a preprocessing for improving the accuracy of such estimation calculation, a more favorable result can be obtained by performing the first-order least-squares method on the effective value interval.

Next, an example of control in the present invention will be described according to the flowchart of FIG. In step S1, a driver image is acquired from the in-vehicle camera 1. FIG. In subsequent step S2, it is determined whether or not the driver's head condition can be calculated from the acquired driver image. If the driver's head condition can be calculated, the process proceeds to step S3, where the driver's head condition is calculated, and in the following step S7, an abnormality determination is made based on the calculated value of the driver's head condition, and one cycle of processing ends. .

If it is determined in step S2 that the driver's head state cannot be calculated, the process proceeds to step S4 to set the predicted recovery time. Note that if the incalculable state continues, the predicted recovery time is extended by the duration.

In step S5, the calculated value immediately before the driver's head state becomes incalculable is set as the assumed head state value. In step S6, an estimated value of the driver's head state is calculated by interpolation using the two calculated values immediately before the driver's head state becomes incalculable and the assumed head state value. An abnormality determination is made based on the estimated value of the part state, and one round of processing ends.

FIG. 4 shows an example of the driver's head state detection result using the present invention. In the figure, the upper graph shows the detected vertical angle of the driver's head (face), and the lower graph shows the front-back position of the driver's head (face). As shown, in each detection result, the missing part in the conventional detection result (FIG. 5) is properly compensated (without divergence) by the estimated value (indicated by the dashed line).

Although the embodiments of the present invention have been described above, the present invention is not limited to the above-described embodiments, and appropriate modifications can be made within the scope of the claims. For example, in the above embodiment, the interpolation is performed using the quadratic least squares method, but the interpolation method is not limited to the least squares method, and other methods may be used.

INDUSTRIAL APPLICABILITY The present invention can be used for driver abnormality detection based on the driver posture in a vehicle such as an automobile.

U controller 1 camera body 11 driver head state calculation unit 12 driver head state estimation unit 13 abnormality determination unit C1 to C7 driver head state calculation (detection) points R1 to R5 return prediction points E1 to E5 driver head state Estimated point T Estimated return time

Claims (3)

In a driver posture detection device provided in a vehicle,
a photographing unit for photographing an image of the driver;
a head state calculation unit that calculates a measured value of the driver's head state, which is the position and/or posture of the driver's head, based on the image captured by the imaging unit;
When the measured value of the driver's head state cannot be calculated by the head state calculation unit, the calculated value of the measured value of the driver's head state immediately before the measured value of the driver's head state becomes incapable of being calculated is calculated as the driver's head state. It is set to an assumed head state value that is a predicted value at the time when a predetermined recovery prediction time has elapsed from the time when the measured value of the state becomes uncalculable, and is set to the assumed head state value before the driver's head state becomes uncalculable. a head state estimating unit for estimating the driver's head state during a period in which the head state calculating unit cannot calculate, using the calculated value of the driver's head state and the assumed head state value ;
The head state estimating unit calculates the driver's head state by interpolation using the calculated values of the measured values of the driver's head state for two times immediately before the driver's head state becomes uncalculable and the assumed head state value. A driver posture detection device for estimating a part state . In the driver posture detection device according to claim 1 ,
A driver posture detection device that extends the predicted return time as the time during which the driver's head state cannot be calculated increases. In the driver posture detection device according to claim 1 or claim 2 ,
A driver posture detection device comprising an abnormality determination unit that determines an abnormality in the driver's posture using a value calculated by the driver head state calculation unit or an estimated value by the driver head state estimation unit.

Images