FR3064791A1 - DEVICE FOR ASSISTING A VEHICLE DRIVER BY DETERMINING A LEVEL OF INATTENTION BY IMAGE COMPARISON - Google Patents

Abstract

A device (DA) is intended to assist a driver of a vehicle (V) during a taxi phase, the vehicle (V) having a kinematic defined by primary information and comprising acquisition means (MAQ) for acquiring data. primary images of an area in front of it and means of observation (MO) determining the direction of observation of the driver. This device (DA) comprises: - first analysis means (MA1) generating from a primary image a first image representative of an observable area in front of the vehicle (V), taking into account the primary information, - seconds analysis means (MA2) generating from the same primary image a second image representative of a zone observed by the driver in front of the vehicle (V), taking into account the direction of observation, and - processing means ( MT) comparing the first and second images and deducing from this comparison a first level of inattention of the driver.

Description

Holder (s): PEUGEOT CITROEN AUTOMOBILES SA Société anonyme.

Extension request (s)

Agent (s): PEUGEOT CITROEN AUTOMOBILES SA Public limited company.

DEVICE FOR ASSISTING A DRIVER OF A VEHICLE BY DETERMINING A LEVEL OF CAREFULLY BY COMPARING IMAGES.

FR 3 064 791 - A1 _ A device (DA) is intended to assist a driver of a vehicle (V) during a driving phase, this vehicle (V) having a kinematics defined by primary information and comprising means for acquisition (MAQ) to acquire primary images of an area in front of him and observation means (MO) determining the direction of observation of the driver. This device (DA) includes:

first analysis means (MA1) generating from a primary image a first image representative of an observable area in front of the vehicle (V), taking into account the primary information,

second analysis means (MA2) generating from this same primary image a second image representative of an area observed by the driver in front of the vehicle (V), taking account of the direction of observation, and

- processing means (MT) comparing the first and second images and deducing from this comparison a first level of driver inattention.

PRB

MAQ

MA2 MA1

RC ^ mt \ ^plb

i

DEVICE FOR ASSISTING A DRIVER OF A VEHICLE BY DETERMINING A LEVEL OF CAREFUL BY COMPARISON OF IMAGES

The invention relates to vehicles, possibly of the automotive type, and more specifically to devices which assist the drivers of such vehicles.

As those skilled in the art know, many assistive devices are already offered. They can be classified into at least five groups depending on the type of assistance (or help) they provide. Thus, a first group comprises the devices which alert the driver via a man / machine interface when the driving time has exceeded a chosen threshold (for example equal to 2 hours).

A second group includes the devices which are responsible for detecting signs of loss of vigilance on the part of the driver, for example as a function of variations in the angular position of the steering wheel. Note that in what follows the expression "loss of alertness" is synonymous with the word "inattention". In both cases, it is a distraction from the attention of the

0 driver compared to the observation of the lane used and / or driver drowsiness.

A third group includes the devices which are responsible for detecting abnormal swaying of the vehicle on its lane or steering wheel strikes given by the driver when the vehicle is located near a line delimiting its lane. These latter devices need to have images acquired by at least one digital camera in front of the vehicle (for example installed on the windshield).

A fourth group includes the devices that are responsible for detecting signs of distraction and drowsiness on the face of the

0 driver. These latter devices need to observe the orientation of the driver’s head and gaze as well as his eyes, for example using a digital camera.

A fifth group includes devices that are responsible for detecting signs of fatigue by observing the driver’s health, such as his heartbeat. These latter devices require the driver to be equipped with sensor (s).

Such assistive devices are responsible for generating alarms when they detect at least one abnormal parameter. These alarms can be intended for the driver of the vehicle and / or at least one computer on board the vehicle and responsible for controlling at least part of the latter's kinematics (speed, acceleration, trajectory, braking).

îo None of these assistive devices really puts itself in the place of the driver of the vehicle to react according to the actual level of his inattention (for example a driver can deliberately sway).

In addition, the aids offered by these assistance devices have a relatively high error rate (both in overestimating risk and underestimating risk).

The invention therefore aims in particular to improve the situation.

To this end, it notably proposes an assistance device intended to assist a driver of a vehicle during a driving phase on a traffic lane, this vehicle having kinematics defined by

0 primary information and comprising means of acquisition suitable for acquiring primary images of an area in front of the vehicle and means of observation suitable for determining a direction of observation of the driver.

This assistance device is characterized by the fact that it includes:

- first analysis means capable of generating from a primary image a first image representative of an observable area in front of the vehicle, taking into account the primary information,

- second analysis means capable of generating from this same primary image a second image representative of an area observed by the driver in front of the vehicle, taking into account the direction of observation, and

- processing means capable of comparing the first and second images and of deducing from this comparison a first level of driver inattention.

Thanks to this comparison between what is observed in front of the vehicle and what is actually observed by the driver, there is a level of inattention which is very representative of the attention that the driver actually pays to the traffic lane used. .

The assistance device according to the invention may include other characteristics which can be taken separately or in combination, and in particular:

its first analysis means may be suitable for determining in the primary image an observable frontal area of the vehicle with respect to a first predefined reference point, taking account of at least one of the primary information, and its second analysis means may be suitable for determining an optimal vision zone of the driver with respect to a second predefined reference mark, from the direction of observation. In this case, its processing means may be suitable for determining a position of the optimal vision zone with respect to the frontal zone observable in the first predefined benchmark, then deducing from this determined position a second level of driver inattention, then to estimate a residue representative of a difference between the first and second levels of inattention, and to consider valid the first level of inattention when the estimated residue is less than a first selected threshold;

> its first means of analysis can be specific to determining in the primary image lines delimiting the traffic lane, then to construct the observable frontal zone of the vehicle with portions of these lines;

> its processing means may be capable of determining a cause of the residue when the latter is above the first threshold, and of generating an alarm representative of this cause;

> its processing means may be able to estimate the residue also based on data representative of a behavioral profile of the driver (known elsewhere);

- its processing means may be capable of generating a message intended to be broadcast in the vehicle and representative of the first level of inattention when the latter is greater than a second selected threshold;

- its processing means may be capable of generating an alarm or a setpoint intended for at least one vehicle computer involved in controlling the kinematics of the vehicle when the first level of attention is greater than a third selected threshold;

- It can include the observation means, the latter comprising at least one digital camera suitable for observing the driver's head.

The invention also provides a vehicle, possibly of the automobile type, having a kinematics defined by primary information, and comprising acquisition means suitable for acquiring primary images of an area in front of it and a device for assisting the type of the one presented above.

Other characteristics and advantages of the invention will appear on examining the detailed description below, and the attached drawings, in which:

FIG. 1 schematically and functionally illustrates, in a side view, a part of a motor vehicle equipped with an exemplary embodiment of an assistance device according to the invention, and

0 - FIG. 2 schematically illustrates an example of an observable frontal zone of the vehicle of FIG. 1, and two positions, relative to this observable frontal zone, of an example of the zone of optimal vision of the driver of this vehicle, determined by two different moments.

The purpose of the invention is in particular to propose a device 25 for DA assistance intended to assist a driver CD of a vehicle V during a driving phase on a traffic lane.

In what follows, it is considered, by way of nonlimiting example, that vehicle V is of the automobile type. This is for example a car. However, the invention is not limited to this type of vehicle. It concerns everything

0 type of vehicle that can be driven by a driver. Consequently, it relates to land vehicles, river (or maritime) vehicles, and air vehicles.

There is schematically shown in Figure 1 a non-limiting example of vehicle V driven by a driver CD, installed in a passenger compartment H, and equipped with an assistance device DA according to the invention.

As illustrated in FIG. 1, a DA assistance device, according to the invention, comprises at least first MA1 and second MA2 analysis means and MT processing means.

In order for the invention to be implemented, the vehicle V must include MAO acquisition means capable of acquiring primary IP images of an area in front of it (V) and observation means MO îo proper to determine a direction of observation of its driver CD. In addition, the kinematics of the vehicle V must be defined by primary information, such as for example its current speed vector, its current acceleration vector, its current trajectory (X, Y) or its current braking level. All this primary information is available in the i5 V vehicle, for example from at least one computer and via an on-board RC communication network, possibly of the multiplexed type, and allowing the exchange of data and commands between electronic equipment.

MAO acquisition means can, for example, include

0 minus a digital camera, for example installed in a front part of the vehicle V (possibly in the bumper or front bumper, or in the passenger compartment H (possibly in the foot of the central rear view mirror RV which is secured to the windshield PRB The primary IP images acquired by these MAO acquisition means (and representative of the area located in front of the vehicle V) are for example transmitted via the communication network RC.

Note that the MAO acquisition means can possibly be part of the DA assistance device. But it is not compulsory. In fact, they could be part of another device on board the vehicle V, such as for example an obstacle detection device located in a

0 front part of vehicle V.

The observation means MO can, for example, comprise at least one digital camera suitable for observing the face of the driver CD and calculation means suitable for deducing from this observation the direction of observation of the driver CD. This direction of observation can be deduced from the images acquired by the digital camera either directly when it observes at least the eyes of the driver CD, or indirectly from the inclination of the head of the driver CD when it observes only the head in its entirety or the look.

It will be noted that any technique known to those skilled in the art can be implemented, alone or in combination, by the observation means MO to deduce the direction of observation of the driver CD.

Furthermore, the observation direction deduced by the observation means 10 MO is for example transmitted via the communication network

RC.

It will also be noted that the observation means MO are preferably part of the assistance device DA. But it is not compulsory. Indeed, they could be part of another device on board vehicle V, such as for example a device for monitoring the driver's health or defining the driver's profile.

In the exemplary embodiment illustrated without limitation in FIG. 1, the assistance device DA is an electronic device which is connected to the abovementioned communication network RC. It therefore includes a

0 computer comprising the first MA1 and second MA2 analysis means and the MT processing means. But in an alternative embodiment, it could be part of a computer (such as the on-board computer of vehicle V) or another device, such as multimedia equipment, for example. Consequently, the DA assistance device can be produced in the form of software modules (or computer or "software"), or else of a combination of electronic circuits (or "hardware") and software modules.

The first MA1 analysis means are suitable for generating from a primary IP image (provided by the MAQ acquisition means) a

0 first image 11 which is representative of an observable area in front of vehicle V, taking into account primary information (available and defining the kinematics of vehicle V).

It is important to note that each first image 11 can be defined by at least part of the data file defining the corresponding primary IP image, or by a fully constructed data file, possibly in augmented reality, notably from the file. of data defining the corresponding primary IP image. In other words, it can be a real image or a virtual image (the latter alternative is currently preferred).

The second analysis means MA2 are capable of generating from this same primary image IP a second image 12 which is representative of an area observed by the driver CD in front of the vehicle ίο V, taking into account the direction of observation ( provided by the observation means MO). It will be understood that each second image 12 represents the part of a primary image IP which is observable by the driver CD taking into account its direction of observation in progress. Consequently, if the direction of observation is shifted to the left relative to the center of the primary IP image, the second image 12 will not include a right portion of this primary IP image, the missing right portion being all the more important that the shift to the left is important. Conversely, if the direction of observation is shifted to the right relative to the center of the primary IP image, the second image 12 will not include a portion

0 left of this primary IP image, the missing left portion being all the more important as the shift to the right is important.

It is important to note that each second image 12 is defined by at least part of the data file defining the corresponding primary IP image, or else (and as is currently preferred) by a fully constructed data file, possibly in reality increased, in particular from the data file defining the corresponding primary IP image. In other words, it can be a real image (possibly truncated) or a virtual image (the latter alternative is currently preferred).

0 It is also important to note that the first 11 and second I2 images are defined in the same frame of reference of vehicle V, for example located at the base of the windshield in the axis of the driver.

When the observation means MO cannot determine the direction of observation in progress, for example due to the closure of the driver's eyelids (case of drowsiness), the second image 12 can be "black".

The processing means MT are suitable for comparing the first 11 and second I2 images with each other, and for deducing from this comparison a first level of attention NI1 of the driver CD.

The larger the second image I2 will have a missing portion compared to the first image 11, the greater the first level of attention NI1 and therefore the lower the vigilance. For example, the first level îo of attention NI1 may be representative of the percentage of correspondence (or of substantially identical pixels) between the first 11 and second I2 images or the number of identical objects observed (determined by correlation of images).

The first level of inattention NI1 being determined by a comparison between what is observed in front of vehicle V (11) and what is actually observed by the driver CD (I2), it is therefore very representative of the attention which the CD driver actually carries to the traffic lane taken.

Note that any technique known to those skilled in the art can be

0 implemented, alone or in combination, by the processing means MT to compare the first 11 and second I2 images with each other.

In order to determine whether the first level of attention NI1 determined is reliable, it is possible to introduce redundancy in the processing carried out by the DA assistance device. This can be done, for example, as described below.

First of all, the first analysis means MA1 may be suitable for determining in the primary image IP an observable frontal area ZFO of the vehicle V with respect to a first predefined reference point, taking account of at least one of the primary information , and the second analysis means

0 MA2 may be suitable for determining an optimal vision zone ZVO of the driver CD with respect to a second predefined reference mark from the direction of observation.

For example, the first predefined mark can be that which is attached to vehicle V and described above (for example located at the base of the windshield in the driver's axis).

In this exemplary embodiment, dedicated to redundancy, the processing means MT are first of all suitable for determining a position of the optimal vision zone ZVO with respect to the observable frontal zone ZFO in the first predefined reference frame. Then, the processing means MT deduce from this determined position a second level of inattention N12 of the driver. Then, the processing means MT estimate a residue representative of a difference between the first NU and second NI2 levels of inattention, and consider valid the first level of inattention NU when this estimated residue is less than a first threshold S1 chosen. .

It will be understood that the second level of inattention NI2 is information which is redundant with respect to the first level of inattention NU, and therefore the comparison between these first NU and second NI2 levels i5 of inattention is intended to determine the reliability of this first level of NU inattention. If there is a noticeable difference between the first NU and second NI2 levels of inattention, this means that at least one of them is wrong and therefore that the first level of inattention NU cannot, at least temporarily , be taken into account due to a problem.

0 The estimation of the residue can, for example, be carried out by the MT processing means when the noise level measured on a sensor (MAQ acquisition means or MO observation means) is too high compared to a selected threshold. But it could be estimated permanently.

The determination of the position of the optimal vision zone ZVO with respect to the observable frontal zone ZFO in the first predefined reference mark requires a prior change of reference mark of the optimal vision zone ZVO from the second predefined reference point to the first predefined reference point. This change can possibly be achieved by the second means

0 MA2 analysis, otherwise it is carried out by the MT processing means.

For example, the optimal vision zone ZVO can be materialized by a geometric figure. Thus, it could be, for example, a circle (as illustrated without limitation in FIG. 2), a square, a rectangle, ίο a triangle, or a hexagon. In this case, the center of the geometric figure can represent the intersection of the plane with the direction of observation deduced, and the "width" of the geometric figure can represent the uncertainty of calculation on the direction of observation deduced.

Also for example, and as illustrated without limitation in FIG. 2, the first analysis means MA1 may be suitable for determining in the primary image IP lines which delimit the traffic lane, then for constructing the observable frontal area ZFO of the vehicle V with portions of these lines.

îo In the purely illustrative example of FIG. 2, the observable frontal zone ZFO comprises a left part PLG representing a portion of the left line delimiting the traffic lane taken, a right part PLD representing a portion of straight line delimiting this traffic lane , a front part connecting the front ends of the left parts

PLG and right PLD, and a rear part connecting the rear ends of the left PLG and right PLD parts. Such an observable frontal zone ZFO, substantially symmetrical with respect to a median vertical plane, corresponds to a trajectory of the vehicle V substantially linear and centered on a linear portion of the taxiway located in front of this vehicle V. When the portion

0 of taxiway located in front of vehicle V is curvilinear (for example in a bend), the left PLG and right PLD parts of the observable frontal area ZFO are very asymmetrical with respect to the median vertical plane.

In FIG. 2 is schematically illustrated an example of an observable frontal zone ZFO of the vehicle V, and two examples of positions of the optimal vision zone ZVO of the driver CD with respect to this observable frontal zone ZFO, determined at two different times. It is considered here that between these two different moments the observable frontal zone ZFO of vehicle V has not changed shape. Here, the optimal vision zone ZVO on the left is slightly off center in the front left part

0 of the observable frontal area ZFO, which means that the second level of attention NI2 of the CD conductor is higher. The ZVO optimal viewing area on the right is off center in the rear right of the ZFO observable frontal area, which means that the second level of attention CD driver NI2 is medium lower or even low. When the ZVO optimal viewing area is located outside the ZFO observable frontal area, this means that the second level of N12 attention from the CD driver is (very) low, or even zero. The further the ZVO optimal vision area is located outside the ZFO observable frontal area, the greater the second level of attention CD driver NI2.

It will also be noted that any technique known to those skilled in the art can be implemented, alone or in combination, by the processing means MT to determine the residue from the first NI1 and second ίο N12 levels of attention.

It will also be noted that the processing means MT may be suitable for estimating the residue also as a function of data representative of a behavioral profile of the conductor CD which is known elsewhere.

This behavioral profile can be used once the driver 15 CD has been recognized, for example by means of observation MO.

For example, one can have a behavioral profile resulting from a recording of the usual behavior of the CD driver as a function of the day of the week and / or the time and / or the geographic position. In this case, one can, for example, consider that the driver CD is not attentive when the first level of inattention NI1 in progress does not correspond to his usual behavior, including when the residue (representative of the difference between the first NI1 and second NI2 inattention levels) is less than the first threshold S1.

For example, the behavioral profile of the driver can notably indicate whether he is rather distracted or concentrated and / or if he tends to cut corners.

It will also be noted that the processing means MT can also be suitable for determining a cause of the residue when the latter is greater than the first threshold S1, and for generating an alarm representative of

0 this cause. This alarm may be intended for a computer of the vehicle V, for example involved in controlling the kinematics of the latter (V), and / or the driver CD. In the latter alternative, the alarm can be displayed in the form of text and / or an image on at least one screen of vehicle V (for example that of the central instrument panel of the dashboard PLB or of the instrument panel CT dashboard), and / or broadcast by at least one speaker in vehicle V.

The cause can result from a problem with a sensor (for example an oversized noise), or with the MAQ acquisition means. For example, the problems may be: noises within the system (comprising the DA assistance device and the MAQ acquisition means), or a difference between the mathematical model of this system and the real response îo of this system, or a failure of a component of this system.

The mathematical modeling of the system can, for example, be expressed by the following state equations:

X (t / t-1) = F (X (t-1), U (t-1), ΔΧ, v (t-1)), and y (t) = G (X (t / t-1 ), ε (ΐ), ΔΧ, ω (ΐ-1)).

In these equations, X (t / t-1) is the predicted state of the system at time t from the previous instant and the excitation U (t) of the system, ΔΧ represents a mathematical modeling of l uncertainty on the mathematical model of the chosen system, and v (t) is a source of noise on the system.

In the second equation, y (t) represents the predicted observations

0 of the sensors, ω (ΐ-1) is an error representative of the noise on the MAQ acquisition means and ε (ΐ) is an error due to a failure.

We consider here that the different types of problem are decoupled from each other. In this case, the residue r (t) can be defined by the vector equation r (t) = C * e (t) + f _s (t), where C is a transformation matrix, f _s (t) is a function representing the problems of the sensors, and e (t) is a vector presenting the difference between the state predicted by the mathematical model of the system and the state estimated after correction, and linked to the uncertainties of the mathematical model of the system and to the problems of the system. The choice of the transformation matrix C is essential so that the decoupling of

0 problems be carried out effectively.

Preferably, it is required that the residue have a known distribution centered around 0 when the system is operating correctly, and a distribution different from that known in the presence of a problem (for example, an eccentric mean and / or a different variance).

If the mathematical model of the system is perfectly known, and 5 the noises on the system and on the sensors are considered to be Gaussian, the residue r (t) can be the difference between the prediction and the estimation made by a Kalman filter . We can then deduce directly from the residue r (t) an indication of a sensor problem or a problem (abnormal behavior) of the system.

îo For example, a mathematical model of the ARX (“Auto Regressive eXogenous”) system can be used, which is well suited to systems subject to noise. It is recalled that the ARX model is autoregressive taking into account external inputs. It can be defined by the following equation:

i5 y (k) = F (y (k-1), y (k-2), ..., y (kn), u (k-1), u (k-2), ..., u (km)) + s (k).

In this case, the residue can be calculated independently for each type of problem using an equation of the type:

R (k) = Zimi * (Vi - vj) + Σ, η, * (ω, - ω ~), where m, and η are weights defined during experiments.

0 As mentioned above, the analysis of the residue can be carried out by means of hypothesis tests, and comparisons with thresholds. It is intended to enable effective detection of a problem and the identification of the type of problem and its cause.

As a purely illustrative example, if a residual in nominal mode 25 (excluding the problem) of the system at a value rO, then the state of this same residue varies in the presence of a problem, assuming that two problems cannot appear simultaneously. We can then use predefined laws to distinguish the different types of problem from each other. These laws are defined during preliminary analyzes intended to

0 identify each of the possible problems and the consequences of these problems on the residual system.

The determination of the problems and their causes makes it possible to better detect the risk situations and not to take into account certain untimely situations.

It will also be noted that the processing means MT may be able to generate a message intended to be broadcast in the vehicle V and representative of the first level of attention NI1 when the latter (NI1) is greater than a second threshold S2 chosen. Diffusion by at least one speaker of the vehicle V is preferred here to the display on a screen, because in the event of strong inattention of the driver CD, this last (CD) is generally not able to notice the fact that '' an alert message has just been displayed on a screen of vehicle V (unless it is observing this screen).

It will also be noted that the processing means MT may be capable of generating an alarm or a setpoint intended for at least one computer of the vehicle V involved in the control of the kinematics of the vehicle V when the first level of inattention NI1 is greater than a third threshold S3 chosen. This third threshold S3 is preferably greater than the second threshold S2.

In the presence of this last option, the alarm generated by the processing means MT can be used to trigger at least one action on vehicle V, depending on the level of attention of the driver CD. This action can be, for example, a temporary activation of the horn (or "horn") and / or the hazard warning lights (or "warning") from a minimum inattention threshold, braking, gradual reduction of a maximum speed programmed by the CD driver or authorized on the portion of the taxiway used, deactivation of the cruise control selected by the CD driver, or a simulation of lifting the foot of the pedal accelerator, or a ban on the display of certain pages of information, or a variation in the intensity of the sound signal broadcast by the speakers of vehicle V, or a reduction in the temperature of the air distributed in the passenger compartment H, or an inhibition of certain commands from vehicle V, or the prevention of displaying certain menus or submenus, or a configuration in “hard” suspension mode.

It will be noted that the first S1, second S2 and third S3 thresholds can be adjustable over time, for example according to the profile of the driver (known elsewhere).

It will also be noted that in the event of drowsiness, the second image 12 may be a "black" image, as mentioned previously, and therefore an alert is preferably triggered.

Claims (10)

1. Assistance device (DA) to assist a driver of a 5 vehicle (V) during a driving phase on a traffic lane, said vehicle (V) having a kinematics defined by primary information and comprising acquisition means (MAQ) suitable for acquiring primary images of a zone located in front of said vehicle (V) and observation means (MO) capable of determining a direction of observation of said driver, characterized in that it comprises i) first analysis means (MA1) capable of generating from a primary image a first image representative of an observable area in front of said vehicle (V), taking into account said primary information, ii) second analysis means (MA2) suitable for generating a second from this same primary image image i5 representative of an area observed by said driver in front of said vehicle (V), taking into account said direction of observation, and iii) processing means (MT) suitable for comparing said prem first and second images and to deduce from this comparison a first level of inattention of said driver. 2 0 2. Device according to claim 1, characterized in that said first analysis means (MA1) are capable of determining in said primary image an observable frontal area of said vehicle (V) with respect to a first predefined reference point, taking into account at least one of said primary information, in that said second analysis means (MA2) 25 are suitable for determining an optimal vision area of said conductor with respect to a second predefined reference mark, from said direction of observation, and in that said processing means (MT) are suitable for determining a position of said area of optimal vision in relation to said frontal area observable in said first predefined reference frame, then to deduce from this Determined position a second level of inattention of said driver, then estimating a residue representative of a difference between said first and second levels of inattention, and considering valid said first level of inattention when said estimated residue is less than a first selected threshold. 3. Device according to claim 2, characterized in that said first analysis means (MA1) are suitable for determining in said primary image lines delimiting said traffic lane, then for constructing said observable frontal area of the vehicle (V) with portions 5 of these lines. 4. Device according to one of claims 2 and 3, characterized in that said processing means (MT) are capable of determining a cause of said residue when the latter is greater than said first threshold, and of generating an alarm representative of this cause . îo 5. Device according to one of claims 2 to 4, characterized in that said processing means (MT) are suitable for estimating said residue also according to data representative of a behavioral profile of said conductor. 6. Device according to one of claims 1 to 5, characterized in that 15 said processing means (MT) are capable of generating a message intended to be broadcast in said vehicle (V) and representative of said first level of attention when the latter is greater than a second selected threshold. 7. Device according to one of claims 1 to 6, characterized in that 2 0 said processing means (MT) are capable of generating an alarm or a setpoint intended for at least one computer of said vehicle (V) involved in controlling the kinematics of said vehicle (V) when said first level of inattention is higher to a third selected threshold. 8. Device according to one of claims 1 to 7, characterized in that it 25 includes said observation means (MO), the latter (MO) comprising at least one digital camera suitable for observing a head of said conductor. 9. Vehicle (V) having kinematics defined by primary information and comprising acquisition means (MAQ) suitable for acquiring 3 0 primary images of an area located in front of said vehicle (V), characterized in that it further comprises an assistance device (DA) according to one of the preceding claims. 10. Vehicle according to claim 9, characterized in that it is of the automobile type. 1/1