JPH081385B2 - Abnormal operation detection device - Google Patents

Description

Description: TECHNICAL FIELD The present invention relates to an abnormal driving detection device, and more particularly to a device that compares an appropriate steering angle with a current steering angle to detect an abnormal driving such as a drowsiness driving.

[Prior Art] Conventionally, various safety devices have been proposed, particularly for drowsy driving, from the viewpoint of pursuing safety when driving a vehicle.

As an example of this type of safety device, there is an alarm device for preventing drowsiness during automatic driving, which is disclosed in JP-A-64-83423 or JP-A-1-122734. FIG. 12 shows a plan view of this conventional device. In this conventional device, the white lines 1 and 2 for displaying the traveling division bands on the road are photographed by a recognition device 5 such as a camera attached to a door mirror 4 of the vehicle 3 and the traveling division bands are displayed by pattern recognition from the obtained image. Recognize the white line and confirm its position. Then, when the position of the white line for displaying the traveling section deviates from the predetermined range, it is determined that the dozing driving is occurring, and an alarm such as a buzzer is given to the driver to restore the normal driving. .

In other words, this device determines that a drowsiness drive is occurring when the vehicle starts to shift to the white line for displaying the lane markings, and warns the driver before the vehicle deviates from the lane markings. To give.

[Problems to be Solved by the Invention] However, the above-mentioned conventional drowsiness driving detection device has some problems. As described above, Japanese Patent Laid-Open No. 64-83
In the dozing driving prevention alarm device disclosed in Japanese Patent Laid-Open No. 423 or JP-A-1-122734, the position of the white line for displaying the driving division zone of the road is confirmed from the image captured by the camera or the like, and the white line of the vehicle is checked. Although it is a method of detecting a deviation and detecting a dozing driving, it is not possible to appropriately give an alarm to the driver according to the running state of the vehicle by just judging the deviation from the white line in this way. is there. For example, when focusing on the vehicle speed of the vehicle, when the vehicle speed is high, the risk of drowsy driving is higher than that when the vehicle speed is low, even if the deviation from the white line is the same. However, this conventional device cannot cope with such a case, and there is a problem that an appropriate and effective alarm cannot be given.

The present invention has been made in view of the problems of the above-described conventional device, and the object thereof is to detect an abnormal driving such as a drowsy driving or a looking aside driving with high accuracy regardless of the running state of the vehicle, and to give an appropriate warning to the driver. An object of the present invention is to provide an abnormal driving detection device for a vehicle, which can provide safe driving.

[Means for Solving the Problems] In order to achieve the above-mentioned object, the abnormal operation detection device of the present invention detects an on-vehicle camera 10 for photographing a forward lane of a vehicle and a traveling speed of the vehicle as shown in FIG. Vehicle speed sensor 12 to detect
A steering angle sensor 14 for detecting a steering angle of the vehicle, and a lateral displacement Δ of the vehicle from the image obtained by the vehicle-mounted camera 10.
l, the yaw angle Δθ, and the vehicle-mounted image processing device 16 for calculating the curve radius R of the road, the vehicle speed v detected by the vehicle speed sensor 12, and the lateral displacement Δl, yaw calculated by the vehicle-mounted image processing device 16. A proper steering amount calculation means 18 for calculating a proper steering amount u ₀ of the vehicle with respect to the traveling path based on the angle Δθ and the curve radius R, and a steering angle u detected by the steering angle sensor 14 and a calculated proper steering amount. It is characterized by comprising an abnormality detecting means 20 for comparing u ₀ to detect an abnormality in the steering angle, and an alarm means 22 for giving an alarm to a vehicle driver based on an abnormality detection signal from the abnormality detecting means 20. There is.

[Operation] The abnormal operation detection device of the present invention has such a configuration, and the vehicle-mounted vehicle has the curve radius R as the road shape in front of the vehicle and the vehicle speed v, lateral displacement Δl, and yaw angle Δθ as the running state of the vehicle. The image obtained by the camera is detected by image processing.

The appropriate steering amount calculation means calculates an appropriate steering amount according to the running shape and the current running state from the detection values thus obtained. That is, the deviation of the vehicle from the proper course at the distance the driver is gazing while the vehicle is traveling (depending on the current vehicle speed v) is obtained from the detected values such as the curve radius R and the lateral displacement Δl, and the deviation is eliminated. In order to achieve this, the amount to be steered is calculated as the proper steering amount.

Then, by comparing the current steering angle from the steering angle sensor 14 with the calculated proper steering angle, how much the current steering angle deviates from the proper steering angle is calculated. When an abnormal driving such as a drowsy driving or a looking aside driving is performed, the driver's ability to recognize the front side is significantly reduced, and the steering angle is largely deviated from the proper steering amount.

Therefore, by comparing the current steering angle with the calculated appropriate steering angle, it is possible to reliably determine whether or not the vehicle is in an abnormal operation, and it is possible to give a warning to the driver to call attention.

[Embodiment] A preferred embodiment of the abnormal operation detection device according to the present invention will be described below with reference to the drawings.

First Embodiment FIG. 2 is a configuration block diagram of the first embodiment of the present invention. The vehicle-mounted camera 10 is attached to a predetermined position of the vehicle, for example, a door mirror or the like, photographs the front running path of the vehicle, and sends the photographed image to the vehicle-mounted image processing device 16. Then, the vehicle-mounted image processing device 16 processes the image sent from the camera 10 to calculate the lateral displacement Δl of the vehicle, the yaw angle Δθ, and the curve radius R of the vehicle front lane. Although various methods are conceivable as methods for calculating these physical quantities, in the present embodiment, a white line for displaying the traffic division zone of the front runway is extracted from the image captured by the camera 10, and the lateral displacement is extracted from the extracted white line. Δl, yaw angle Δθ, and curve radius R are to be calculated. That is, first, the image signal obtained by the vehicle-mounted camera 10 is converted into digital data for each predetermined pixel. Then, a white pixel is detected from the luminance change of each pixel and a white line for displaying a traffic division is extracted.

Next, regarding the extracted white line, the tangent lines of two points having different front distances from the vehicle are obtained, and the crossing angle of these close battles is calculated. Generally, since there is a certain correlation between the tangent intersection angle and the curve radius R, it is possible to calculate the curve radius R of the road ahead of the vehicle from the obtained intersection angle. On the other hand, the lateral displacement Δl of the vehicle can be calculated from the inclination of the image of the extracted white line and the focal length of the camera, and the yaw angle θ can be calculated from the intersection of the extracted white lines. It should be noted that the curve radius R of the front lane and the yaw angle Δθ of the vehicle can be calculated from the image captured by the vehicle-mounted camera 10.
The device for calculating is as follows:
-278269.

By the way, the curve radius R of the road, the lateral displacement Δl of the vehicle and the yaw angle Δθ detected by the in-vehicle image processing device 16 in this manner are sent to the appropriate steering amount calculation means 18. On the other hand, the vehicle speed v detected by the vehicle speed sensor 12 composed of a magnetic resistance element or the like is also input to the appropriate steering amount calculation means 18.

The proper steering amount calculation means 18 stores various parameters.
It has a ROM, a first-in first-out (FIFO) memory that sequentially stores data and outputs the stored data with priority, and a CPU that performs data calculation. By performing the processes described below, the current running state and running path of the vehicle An appropriate steering amount u ₀ according to the shape is calculated.

That is, first, the error Δe is calculated from the distance Δx traveled at every predetermined time T ₀ and the detected curve radius R in the FIFO memory (see FIGS. 3 and 4) and sequentially stored. The number of stored n, the maximum vehicle speed v _max and look-ahead distance L _max described later corresponding to the maximum speed _{v max, n = L max /} (v max · T 0) is determined by ... (1).

Now, as shown in Fig. 5, the vehicle runs on the proper course 10
Consider a case where the vehicle is in a traveling state where the lateral displacement Δl is apart from 0 and the yaw angle Δθ is obtained. The distance that the driver will gaze while driving the vehicle, that is, the deviation of the vehicle from the proper course 100 at the forward gaze distance L ₀ , will exist if it is assumed that there is no lateral displacement Δl at the current position. Error ε _r from course 100, yaw angle Δθ from current vehicle position
Is the sum of the lateral displacement Δl that would exist when 0 was assumed to be zero and the deviation ε ₀ due to the yaw angle Δθ at this current position. Therefore, the proper steering amount u ₀ that must be steered at the current position in order to eliminate the deviation that may occur at the forward gaze distance L ₀ is obtained by multiplying each deviation by a predetermined gain and adding u ₀ = It can be calculated by K ₁ ε _r + K ₂ Δl + K ₃ ε ₀ (2). Where ε _r is the forward gaze distance
Since this is an error from the proper course 100 that may exist when the vehicle advances by L ₀ , the error Δe shown in FIG. 3 or FIG. Determined by However, for m, the maximum m that satisfies the following equation is adopted in order to correspond to the forward gaze distance L ₀ .

Note that the forward gaze distance L ₀ generally has a positive correlation that increases as the vehicle speed v increases, as shown in FIG. 6. Therefore, the forward gaze distance L for the vehicle speed v is stored in the ROM in advance, Forward gaze distance corresponding to the current vehicle speed v ₀
It suffices to read L ₀ from this ROM and perform the above calculation.

Also, it is calculated deviation epsilon ₀ due to the yaw angle [Delta] [theta] of the vehicle by ε _{0 =} L ₀ · Δθ ...... using forward fixed distance L ₀ and the yaw angle [Delta] [theta] corresponding to the current vehicle speed v ₀ (5) . For example, T ₀ = 0.5 seconds, current vehicle speed v ₀
Consider the case where ＝ 30km / h. The vehicle speed v is as shown in FIG. , The detection position every T ₀ = 0.5 seconds is the 9th
As shown in the figure, 5m, 10m, 20m, 30m, 45m, 60m, ………
Becomes Then, the error Δe from the optimum course in the distance Δx advanced every 0.5 seconds is calculated by processing the image of the advanced distance Δx and the vehicle-mounted camera 10 by the vehicle-mounted image processing device 16 to obtain the curve radius R of the track. when calculated from, considering that errors .DELTA.e ₁ is a curve radius R = 100 at a distance [Delta] x = 5 m advanced example when elapsed from the current position 0.5 seconds, 100 - {(100) ² -5 ^{2} 1 / 2} = 0.125 (m). Similarly, the error .DELTA.e is obtained every 0.5 seconds and sequentially stored in the FIFO memory to obtain the data shown in FIG. Then, the front gaze distance L ₀ = 50 m corresponding to the current vehicle speed v ₀ = 30 km / h is read from the ROM, and ε _r is calculated using this front gaze distance L ₀ . Since the maximum m that satisfies the above-mentioned expression (4) is m = 5, ε _r is 0.125 + 0.125 + 0 + 0 + 0.563 = 0.813 (m).

After ε _r , Δl, and ε ₀ are calculated in this way,
The appropriate steering amount u ₀ can be obtained by multiplying these calculated values by predetermined gains K ₁ , K ₂ , and K ₃ and adding them. In the present embodiment, in order to appropriately correct each of the gains K ₁ , K ₂ and K _{3 according} to the driving characteristics of the driver, the driving characteristics of the driver as shown in the block diagram of FIG. A driver model memory 22 that stores the average and variance of the lateral displacement Δl of the vehicle during normal operation, the variance of the yaw angle Δθ, etc. is provided, and the gains K ₁ , K ₂ , and K ₃ should be corrected based on these values. You can do it. For example, when the variance of the yaw angle Δθ of the vehicle during normal driving is large, the appropriate steering amount
By setting the gain K ₃ of the error ε ₀ resulting from the yaw angle Δθ at u ₀ to be small, it is possible to calculate the appropriate steering amount u ₀ more in accordance with the driving characteristics of the driver.

The proper steering amount u ₀ in the current traveling state of the vehicle calculated by the proper steering amount calculation means 18 as described above is sent to the abnormality detection means 20, and the current steering angle detected by the steering angle sensor 14 is detected. Compared with u. When the difference between the proper steering amount u ₀ and the current steering angle u is larger than a predetermined threshold value, it is determined that an abnormality such as a drowsiness driving or a looking aside driving has occurred, and an alarm (not shown) such as a buzzer is issued. An abnormal signal is output to the means.

Note that the driver model memory described above is also used when determining whether the current steering angle u is abnormal by the abnormality detecting means 20.
By appropriately correcting the threshold value for this abnormality detection using the data stored in 22, for example, in the case of relatively rough driving, by setting the threshold value high, it is possible to match the driving characteristics of the driver. It is possible to detect abnormalities.

Second Embodiment FIG. 11 is a configuration block diagram of a second embodiment of the present invention. A feature of the second embodiment is that a neural network is used as the proper steering amount calculation means. This neural network normalizes a vehicle speed v from the vehicle speed sensor 12, a curve radius R from the vehicle-mounted image processing device 16, a lateral displacement Δl, and a yaw angle Δθ.
It is composed of a network unit 18b composed of 18a and formal neurons. As is well known, a formal neuron is a multi-input-1 output non-linear element and outputs a pulse only when the weighted addition of each input exceeds a predetermined threshold value. Then, by appropriately setting the weighting coefficient and the threshold value of each input signal of this type neuron, a logic gate such as an AND gate, an OR gate, and a NOT gate can be configured. Therefore, a neural network formed by combining these formal neurons can perform various arithmetic processes by appropriately changing the combination. In the second embodiment, the proper steering amount u ₀ calculated by the neural network without using the mathematical formula is sent to the abnormality detecting means 20 and the current steering angle sensor 14 outputs the current steering angle sensor 14 as in the first embodiment. The steering angle u is compared with the steering angle u and when it exceeds a predetermined threshold value, it is determined to be abnormal and an abnormal signal is output.

In the present embodiment, when the abnormality detection means 20 detects an abnormality, the difference between the proper steering amount u ₀ and the current steering angle u is compared, and when the difference is equal to or larger than a predetermined threshold value, it is determined that an abnormality has occurred. Instead of making a determination, it is possible to provide a time counter and output an abnormal signal only when the difference is greater than or equal to a threshold value and the duration is greater than or equal to a predetermined time.

Further, when the alarm generated in the present embodiment is canceled by the driver, the abnormality detecting means 20 determines that this is a false alarm and feeds it back to the neural network, and the weighting coefficient and threshold of each type neuron forming the neural network. By correcting the value, it becomes possible to have a learning function that gradually matches the driving characteristics of the driver.

[Effects of the Invention] As described above, according to the abnormal driving detection device of the present invention, abnormal driving such as drowsiness driving or looking aside can be detected with high accuracy and reliability, and safe driving is possible. Has the effect of

[Brief description of drawings]

1 is a configuration block diagram of an abnormal operation detection device according to the present invention, FIG. 2 is a configuration block diagram of a first embodiment of the present invention, FIG. 3 is an explanatory diagram of an error Δe in the same embodiment, FIG. Is an explanatory view of the FIFO memory of the same embodiment, FIG. 5 is an explanatory view of the proper steering amount in the same embodiment, FIG. 6 is a graph showing the relationship between the vehicle speed v and the forward gaze distance L of the same embodiment, 7 is a graph showing a change in vehicle speed v in the same embodiment, FIG. 8 is a graph showing a relationship between the vehicle speed v and a forward gaze distance L in the same embodiment, and FIG. 9 is every 0.5 seconds in the same embodiment. FIG. 10 is an explanatory view of the FIFO memory in the same embodiment, FIG. 11 is a configuration block diagram of a second embodiment according to the present invention, and FIG. 12 is a configuration block diagram of a conventional device. is there. 10 …… In-vehicle camera 12 …… Vehicle speed sensor 14 …… Steering angle sensor 16 …… In-vehicle image processing device 18 …… Proper steering amount calculation means 20 …… Abnormality detection means

Claims (1)

[Claims] 1. A vehicle-mounted camera for photographing a front lane of a vehicle, a vehicle speed sensor for detecting a traveling speed of the vehicle, a steering angle sensor for detecting a steering angle of the vehicle, and a vehicle image from an image obtained by the vehicle-mounted camera. An in-vehicle image processing device that calculates a lateral displacement, a yaw angle, and a curve radius of a track, and a vehicle speed detected by the vehicle speed sensor, and a lateral displacement, a yaw angle, and a curve radius calculated by the in-vehicle image processing device. An appropriate steering amount calculation means for calculating an appropriate steering amount of the vehicle with respect to the traveling path, and an abnormality detection for detecting an abnormality in the steering angle by comparing the steering angle detected by the steering angle sensor with the calculated appropriate steering angle And an alarm means for giving an alarm to a vehicle driver based on an abnormality detection signal from the abnormality detection means, which prevents a drowsy driving or a look-aside driving. .