FR3128039A1 - Process for authenticating a shot captured in the passenger compartment of a motor vehicle. - Google Patents

Abstract

Process for authenticating a shot captured in the passenger compartment of a motor vehicle. Method for authenticating a candidate shot (C1, C2, C3, C4) by an image sensor (2) fixed in a passenger compartment (1) of a motor vehicle (100), characterized in that it comprises at least one of the following steps: a step (E2) of recognition, in the candidate shot (C1, C2, C3, C4), of elements of the passenger compartment (11, 12, 13, 14) located at the edge of a face or body of a driver of the motor vehicle (100), a step (E3) of monitoring, by analysis of several images from the sensor (2), the continuity of an identity of 'a driver of the motor vehicle (100) over the total duration of an uninterrupted journey between a moment of boarding of a driver in the motor vehicle (100) and an end moment of a first time interval (DT2, DT5 ) of shots including the candidate shot (C1, C2, C3, C4), Figure for abstract: 6

Description

Process for authenticating a shot captured in the passenger compartment of a motor vehicle.

The invention relates to a method for authenticating a shot captured in the passenger compartment of a motor vehicle. The invention also relates to a device for authenticating a shot captured in the passenger compartment of a motor vehicle. The invention also relates to a computer program implementing the mentioned method. The invention finally relates to a recording medium on which such a program is recorded.

The automotive industry is now developing the use of facial recognition to activate various vehicle features, which may require higher or lower levels of security. For example, facial recognition processing used to validate a bank transaction will implement a higher level of security than facial recognition processing used for personalized driver greeting (including, for example, automatic seat position adjustment driver).

Facial recognition processing can be broken down into three steps:
- an enrollment step corresponding to the capture of a portrait of a candidate person by an image sensor (in particular a camera) in order to construct a candidate facial biometric fingerprint,
- an identification step in which the candidate facial biometric fingerprint is compared with a database of facial biometric fingerprints, the person being identified if one of the fingerprints of the database corresponds to the candidate fingerprint, and
- an authentication step in which it is verified that the candidate person is actually present in front of the camera.

The invention relates to the authentication step. It relates more particularly to the detection of a fraud consisting in presenting a decoy at the input of the facial recognition processing, in order to be falsely identified and to take advantage of the rights of the usurped person. A most common mode of fraud and the easiest to implement consists of presenting photos or videos of the impersonated person facing the sensor. It is therefore essential to guard against this type of fraud.

Depending on the type of sensor used to capture the images of a person (for example an RGB camera or an infrared camera), different technologies can detect the presence of a decoy. For example, infrared camera technology makes it possible to discriminate a decoy from a human face by analyzing the reflections present on the captured image. Alternatively, authentication based on images from an RGB camera will use the detection of eye or face movements, blood pulsation, skin texture, or face geometry.

These authentication solutions require the use of algorithms that are costly in terms of processing time and the purchase of licenses. In addition, since these solutions are essentially developed for the banking or customs sector, the processing offered is generic and poorly adapted to the specificities of facial recognition applied to the automotive sector.

The object of the invention is to provide an image authentication method remedying the above drawbacks and improving the image authentication methods known from the prior art. In particular, the invention makes it possible to carry out a method which is simple and reliable and which allows rapid, reliable and inexpensive authentication in terms of calculation time.

To this end, the invention relates to a method for authenticating a candidate shot by an image sensor fixed in a passenger compartment of a motor vehicle, comprising at least one of the following steps:
- a recognition step, in the candidate shot, of elements of the passenger compartment located at the edge of a face or a body of a driver of the motor vehicle,
- a step of monitoring, by analysis of several images from the sensor, the continuity of an identity of a driver of the motor vehicle over the total duration of an uninterrupted journey between a moment of boarding of a driver in the vehicle automobile and an end time of a first time interval of shots including the candidate shot,
- a step of monitoring the continuity of movements of a driver of the motor vehicle and/or of recognizing movements of a driver of the motor vehicle from among a predefined set of movements compatible with a driving situation, by analyzing the images from the sensor over a second continuous time interval including a first time interval of shots including the candidate shot, and the second time interval starting at the latest at the start of the first time interval, the first time interval being by example initiated by an authentication request sent by an application of a motor vehicle.

The authentication method may comprise a step of receiving an authentication request for a shot, the authentication request coming from an application secured by facial recognition of the motor vehicle.

The authentication method may comprise, preceding the step of receiving a request, a step of initializing a position and a geometry of a set of elements of the passenger compartment from at least a reference shot of the empty passenger compartment captured by the sensor,
and the recognition step may include
- a sub-step of determining an outline of a face and a body of a driver,
- a sub-step of determining a set of so-called probationary zones of the candidate shot, distributed over the contour of a face and a body of a driver, so that each zone of the set of probationary zones includes both a portion of a driver's face and a portion of the passenger compartment, the determination of a probationary zone comprising the determination of its position and its geometry, and
- a sub-step of comparing the content of each zone of the set of probationary zones with the content of a reference zone of the reference shot, the reference zone being at the same position and of the same geometry as the probationary area.

The comparison sub-step can be performed, for each probationary zone, on a context sub-zone corresponding to a portion of the probationary zone excluding a face and/or a body of a driver.

The comparison sub-step may comprise a calculation of a spatial correlation rate between each of the context zones and a corresponding reference zone, and a comparison of each correlation rate with a single minimum,
and the candidate shot may be authenticated if all the correlation rates are above the minimum threshold, and the candidate shot may be rejected if not.

The spatial correlation rates can be calculated by a neural network, in particular by a convolutional neural network.

The method may comprise the step of monitoring the continuity of movements of a driver of the motor vehicle and/or of recognizing movements of a driver of the motor vehicle taking place at a given period, and the delay between the start of the second time interval and the beginning of the first time interval can be equal to six times or about six times the given period.

The motor vehicle may be equipped with a driver drowsiness detection system, the driver drowsiness detection system being capable of receiving shots captured by the sensor, and the step of monitoring the continuity of a driver's identity may include receiving a diagnosis from the driver's drowsiness detection system.

The motor vehicle may be equipped with a system for monitoring the vigilance of a driver, the vigilance monitoring system being able to receive shots captured by the sensor, and the step for monitoring the continuity of movements of A driver may include receiving a diagnosis from a driver's alertness monitoring system.

The invention also relates to a device for authenticating a candidate shot of a passenger compartment of a motor vehicle equipped with an image sensor. The device comprises hardware and/or software elements implementing the method as defined previously, in particular hardware and/or software elements designed to implement the method according to the invention, and/or the device comprising means of implement the method as defined previously.

The invention also relates to a computer program product comprising program code instructions recorded on a computer-readable medium to implement the steps of the method as defined above when said program runs on a computer. The invention also relates to a computer program product downloadable from a communication network and/or recorded on a data carrier readable by a computer and/or executable by a computer, comprising instructions which, when the program is executed by the computer, lead it to implement the method as defined previously.

The invention also relates to a data recording medium, readable by a computer, on which is recorded a computer program comprising program code instructions for implementing the method as defined above. The invention also relates to a computer-readable recording medium comprising instructions which, when executed by a computer, lead the latter to implement the method as defined previously.

The appended drawing shows, by way of example, an embodiment of an image authentication device according to the invention and an embodiment of an image authentication method according to the invention.

There represents a motor vehicle equipped with an authentication device.

There describes an embodiment of a system for identifying a person in a shot.

There represents a shot of the passenger compartment of the motor vehicle equipped with an authentication device.

There represents relative positions of an image sensor and a driver's seat according to different settings of the driver's seat.

There is a flowchart of an embodiment of an image authentication method according to the invention.

There illustrates an example of probationary zones relating to a shot captured in the passenger compartment.

There illustrates the comparison processing in a probationary zone and a reference zone corresponding to the probationary zone.

There illustrates an example of probationary areas placed on a candidate image containing a decoy.

There illustrates a first authentication sequence of a shot of a passenger compartment of a vehicle equipped with a system for detecting drowsiness of a driver and a system for monitoring the vigilance of a driver.

There illustrates a second authentication sequence of a shot of a passenger compartment of a vehicle equipped with a system for detecting drowsiness of a driver and a system for monitoring the vigilance of a driver.

There is an alternative representation of a method for authenticating a shot of a passenger compartment comprising a step for recognizing elements of the passenger compartment.

There is an alternative representation of a method for authenticating a shot of a passenger compartment comprising a step of detecting the continuity of a driver's identity.

There is an alternative representation of a method for authenticating a shot of a passenger compartment comprising a step of detecting the continuity of the movements of a driver.

There illustrates the detection of fraudulent authentication from data processed by a process for monitoring the vigilance of the driver.

An example of a motor vehicle 100 equipped with an embodiment of an image authentication device is described below with reference to the .

The motor vehicle 100 can be a vehicle of any type, for example a passenger vehicle or a utility vehicle, or a public transport vehicle.

Motor vehicle 100 mainly comprises:
- a passenger compartment 1 defining the interior space of the motor vehicle 100,
- an image sensor 2 arranged in the passenger compartment 1,
- a set 3 of driver management systems analyzing shots captured by the image sensor 2, the set 3 comprising a facial recognition system 4, and
- a set 5 of applications secured by facial recognition communicating with the facial recognition system 4.

In one embodiment, application set 5 may, for example, include:
- a personalized welcome application 51 for the driver using the driver's facial recognition and allowing for example to configure vehicle settings according to the preferences of the driver, in particular the adjustment of the position of the driver's seat,
- an application 52 for managing bank transactions validated by the driver's facial recognition, these payments being able to concern, for example, the payment of parking, a toll or fuel,
- A vehicle start application 53, allowing the vehicle to be started automatically by facial recognition of the driver.

In a preferred embodiment, when an application 51, 52, 53 is started automatically or by a user of the vehicle (in particular a driver of the vehicle), the application sends a facial recognition request RRF to the facial recognition system 4.

Preferably, the facial recognition system 4 can provide differentiated processing depending on the application, in order to take account of the level of security required by each of the applications 51, 52, 53.

For example, the vehicle start-up application 53 requires very secure facial recognition processing in order to prevent the theft of the vehicle, while fraudulent use of the facial recognition system associated with the personalized welcome application 51 of the driver will have limited consequences.

In addition, the frequency of use of facial recognition processing varies according to the application 51, 52, 53:
- for the vehicle start-up application 53, facial recognition will only take place each time the vehicle is started,
- for the application 52 for managing banking transactions, facial recognition can be performed several times between two vehicle starts (for example during successive toll payments).

It would thus be acceptable for the user of the vehicle to submit to potentially long facial recognition processing (due to a very advanced level of authentication), knowing that it is carried out only once when the vehicle is started, while a facial recognition processing associated with an application used more frequently and/or having to be executed quickly will have to respond to a compromise between the level of authentication required and the speed of implementation of the authentication.

Furthermore, the use of the applications 51, 52, 53 can take place according to a defined time frame. In one embodiment, it is possible, for example, to condition the use of the user system 52 for managing banking transactions on the fact that the transaction takes place during an uninterrupted journey between the driver getting into the motor vehicle (implying strong security of the facial recognition processing of the driver for starting the vehicle) and facial recognition processing associated with the bank transaction. Thus, the level of security of the facial recognition processing associated with the banking transaction is reinforced by the prerequisite of strong authentication for starting the vehicle.

The facial recognition system 4 described in this document allows the implementation of an intermediate level facial recognition processing, which can for example be used to authenticate the shots captured for the operation of a user system 52 for managing banking transactions. To this end, the facial recognition system 4 comprises an identification subsystem 6 and a calculation unit 7 described later in this document and implementing an authentication method according to the invention.

In addition, the facial recognition system 4 may include alternative authentication systems 8 fitted to the vehicle; the systems 8 can be used during facial recognitions relating to the securing of applications of the set of applications 5, for example by the application 53 of starting the vehicle.

When the facial recognition system 4 receives a facial recognition request RRF, it transmits an identification request RI to the identification subsystem 6 and an authentication request RA to the calculation unit 7 or to the systems of authentication alternatives 8. The identification processing can be executed before the authentication processing. Alternatively, some authentication processing can be executed before the identification processing.

In one embodiment, the operation of the identification subsystem 6 can be described with reference to the . The identification of a person is carried out from a candidate shot.

A candidate shot is a shot captured by sensor 2 which is intended to be subject to facial recognition processing.

In the rest of the document, the term "candidate shot" can be used to refer to a shot that has not yet been authenticated or is in the process of being authenticated. This term can also refer to a shot in which the identity of a driver has not yet been determined.

In a first step Id1, the identification subsystem 6 analyzes the candidate shot to extract candidate biometric markers therefrom. Biometric markers are a unique identifier of a person based on their facial geometry and facial expressions. Candidate biometric markers are biometric markers extracted from a candidate shot captured by the sensor 2 and intended to be compared with a database of biometric markers for the identification of a driver of the motor vehicle 100.

In a second step Id2, the identification subsystem 6 compares the candidate biometric markers with the biometric markers of each person already registered in a database of biometric markers. This comparison makes it possible to determine a match score between the candidate biometric markers and the biometric markers of each person registered in the database.

When the match score between the candidate biometric markers and the biometric markers of a given person is greater than a minimum threshold, in a third step Id3, the identification subsystem 6 determines the identity of the person represented in the candidate shot as that of the given person.

The identification subsystem 6 can be common to all the user applications 51, 52, 53. Alternatively, the facial recognition system 4 could comprise different identification subsystems, each being adapted to the specific needs of one or more user applications 51, 52, 53.

The image sensor 2 can be an RGB type camera or an infrared camera. Other embodiments of the sensor 2 are possible. In the rest of the document, the images or data representative of the images coming from the sensor 2 are called “shots”. In the remainder of the document, the images are assimilated to the data representative of these images.

The shots can be for example an RGB or infrared photo or an RGB or infrared video.

An RGB photo or video (for the acronym "Red Green Blue") is a digital photo or video in which the color of each pixel is defined according to a percentage of blue, green and red.

In one embodiment, sensor 2 captures 30 images per second. Higher frequency sensors such as 60, 120 frames per second can also be considered, although less deployed. When an RRF facial recognition request is sent by an application 51, 52, 53, an RRF snapshot is transmitted to the facial recognition system 4. The snapshot C1, C2, C3, C4 can be a single image captured at the time of transmission of the RRF request. Alternatively, a shot C1, C2, C3, C4 can be a video sequence starting at the time of transmission of the RRF request. A shot could possibly be a video sequence starting before the time of transmission of the RRF request, for example implementing a system for storing a video stream.

Optionally, the set 3 of driver management systems could integrate an automatic driver identity control system issuing RRF facial recognition requests in order to carry out checks, in particular periodic checks, of the driver's identity, and this independently of the activation of applications 51, 52, 53.

The sensor 2 is fixed to a structure or a non-moving element of the passenger compartment 1 (windscreen, central mirror 11, body pillar 12) so that the shots entirely include the face of a driver, whatever whatever their size and whatever their seat adjustment, when the driver is installed on the seat of the driving position.

Sensor 2 is intangible, i.e. its position and settings cannot be modified.

As is particularly illustrated by the , the position and the technical characteristics of the sensor 2 define a framing 101 of the shots. In this example, the framing encompasses not only the driver's face, but also a driver's hand and elements of the passenger compartment. Certain elements of the cabin have a fixed position relative to the sensor, such as the dashboard, the steering wheel 13, or the uprights of the body 12. Other elements have a variable position relative to the sensor 2, such as for example the front 14 and rear seats and the center armrest.

The framing thus defines a set of constant (or invariant) elements which are found in the same position regardless of the view taken from the sensor 2 (dashboard, steering wheel, door pillar). Other variable elements may move from one shot to another but will be present on any shot taken from sensor 2 (driver's seat headrest, seat belt, etc.). There also illustrates a shot of the passenger compartment in which appear the central mirror 11, a door pillar 12, the steering wheel 13, the backrest of the driver's seat 14 of the passenger compartment.

In a preferred embodiment, the set 3 of driver control systems comprises, in addition to the facial recognition system 4:
- a driver drowsiness detection system 9, and/or
- A system 10 for monitoring the vigilance of the driver.

These systems, although intended to provide other functions, can advantageously be used to ensure the implementation of the methods that are the subject of the invention. Alternatively, additional systems can be dedicated to the implementation of the methods that are the subject of the invention.

In the embodiment described, the systems 9 and 10 operate continuously, and collaborate with the calculation unit 7, in particular via a memory 110 shared between the systems of the set of systems 3 for managing the driver.

The drowsiness detection system 9 uses images taken from the sensor 2 to analyze the facial movements of the driver, in particular the movements of his eyes, and to detect signs of fatigue or drowsiness of the driver. The detection of such signs can lead to the emission of a visual or audible alarm intended for the driver of the motor vehicle 100.

The analysis of the driver's facial movements is carried out at one to a T9 period called drowsiness detection. It advantageously includes driver identification processing, in order to adapt the analysis of his facial expressions according to his biometric markers. The system 9 can therefore detect a discontinuity in the identity of the driver, during a presentation of a decoy to authenticate a shot. In this case, the system 9 can record an anomaly diagnosis on the driver's identity in a memory 110 shared between the driver management systems 3.

The drowsiness detection period T9 can, for example, be equal to one second. In an advantageous embodiment, the period T9 can be modified according to the driving phases, and also according to the needs of the authentication system according to the invention. For example, the period T9 can be reduced when the motor vehicle 100 finds itself in a situation liable to generate banking transactions and therefore authentications of shots.

To this end, the drowsiness detection system 9 comprises means for processing a request for modification of the period T9 which can be issued by one of the driver management systems, in particular by the control unit 7 of the system facial recognition 4.

The system 10 for monitoring the vigilance of the driver uses images taken from the sensor 2 to analyze the movements of the driver's face and body. In one embodiment, the movements or postures of the driver are compared to a set E of movements or postures compatible with driving. For example, FIGS. 8 and 9 respectively illustrate on graphs G2 and G4 the detection of the position 71 of the driver's hands on the steering wheel.

In an advantageous embodiment, the system 10 could also detect a set of movements making it possible to suspect authentication fraud, such as for example a movement of presentation of a decoy to the sensor 2 as illustrated in the , at time TC. In this case, the system 10 can record an anomaly diagnosis on the analysis of the movements of the driver in a memory 110 shared between the systems 3 for managing the driver.

The analysis of the movements of the driver's face and body is carried out at a so-called monitoring period T10. The period T10 can, for example, be equal to one second. In an advantageous embodiment, the period T10 can be modified according to the driving phases, and also according to the needs of the authentication system according to the invention. For example, the period T10 can be reduced when motor vehicle 100 is in a situation likely to generate bank transactions and therefore authentications of shots.

To this end, the system 10 for monitoring the vigilance of the driver comprises means for processing a request for modification of the period T10 which can be issued by one of the driver's management systems, in particular by the control unit 7 facial recognition system 4.

The driver vigilance monitoring system 10 is capable of determining the posture of the driver by locating, in the plane of a shot of the passenger compartment, the Cartesian coordinates of characteristic points of the driver's body and head.

In the example illustrated by the , the following characteristic points define a structure of the driver's body and head:
- the point (x1,y1) defines the center of the head,
- the points (x2,y2), (x3,y3), (x4,y4), (x5,y5), (x6,y6), (x7,y7) determine the position of the driver's arms,
- the points (x8,y8), (x9,y9) determine the position of the driver's hands,
- other points not shown on the can determine the position of the driver's trunk.

The characteristic points together form a structure of segments which characterizes a posture of the driver.

The monitoring of the driver's posture is for example a monitoring of the structure of segments, which consists in monitoring the variations in position of the points (xi, yi) forming the structure of segments between a time t and a time t+T10, in particular by applying the following formula:
(∆xi, ∆yi) = (xi(t+T10), yi(t+T10)) - (xi(t), yi(t))

The comparison of the variations (Δxi, Δyi) with a maximum threshold can determine discontinuities in the driver's posture.

The structure of segments also makes it possible to characterize each of the postures listed in the set E of movements or postures compatible with driving. The set E can include for example:
- a first structure of segments ST1 describing a posture P1 corresponding to a normal driving posture,
- a second structure of segments ST2 describing a posture P2 in which the driver speaks with a passenger by turning his head towards him,
- a third structure of segments ST3 describing a posture P3 in which the driver faces the rear seats,
- a fourth structure of segments ST4 describing a posture P4 in which the driver looks at the central screen for a long time,
- a fifth structure of ST5 segments describing a P5 posture in which the driver uses his phone to send SMS, turning his head or eyes downwards,
- a sixth structure of segments ST6 a posture P6 in which the driver reads a book or a newspaper by turning the head or the eyes downwards.

In a case where the driver vigilance monitoring system 10 detects sudden variations in the structure of segments, for example a sudden variation in the points determining the location of the hands, it is advantageous to check that the structure of ST segments captured at the moment of the sudden variation of the structure of segments, is well described in the set E. This verification will make it possible to reinforce a suspicion of fraudulent authentication when the structure of segments ST is not described in the set E.

In one embodiment, the set E could also contain a description of a set of suspicious postures, for example postures characterized by suspicious hand positions, in particular a posture of hands holding a decoy, such as the posture represented in the on the fraud attempt time interval DT5. Thus, a fraud attempt could be detected or confirmed when the driver's posture is among the suspicious postures listed in set E.

Furthermore, the shared memory 110 advantageously includes information relating to persons authorized to drive the vehicle, or persons authorized to use one of the applications secured by facial recognition. The information contained in the shared memory may include the facial biometric fingerprints of the persons authorized to drive the vehicle. These fingerprints will be used when identifying the face of the driver of the vehicle among the persons authorized to drive the vehicle.

In one embodiment, the memory 110 allows the alternative authentication systems 8 to share information with the control unit 7. For example, the memory 110 can contain information concerning the successes or failures of the authentications carried out by each of the alternative authentication systems 8.

The motor vehicle 100 can also comprise a memory 120 shared between the set 3 of driver management systems and the set 5 of applications secured by facial recognition. For example, if the motor vehicle 100 includes a personalized driver welcome user system 51, then the position of the driver's seat can be stored in the shared memory 120.

The position of the driver's seat will in particular make it possible to calculate an angle AP and a distance DP for shooting the face as represented by the . The AP angle and the DP distance may be used during the authentication of a candidate shot to check the consistency of the location of the driver's face in the candidate shot, as well as the relative positions of the elements of the shot. passenger compartment and the driver's face.

The calculation unit 7 comprises a microprocessor 71, a local electronic memory 72 and communication interfaces 73 enabling the microprocessor 71 to communicate with the sensor 2, the set of application systems 5, the identification system 6, the system 9 for drowsiness detection, the system 10 for monitoring the vigilance of the driver and the memories 110, 120.

In one embodiment, the microprocessor 71 makes it possible to execute software comprising the following modules, which collaborate with each other:
- an initialization module 710, which collaborates with the sensor 2 and the local electronic memory 72,
- a module 711 for receiving an authentication request, which collaborates with the set of application systems 5,
- a module 712 for recognizing elements of the passenger compartment, which collaborates with the sensor 2, the shared memory 120 and the local electronic memory 72, and
- a module 713 for monitoring the continuity of the identity of a driver of the motor vehicle, which collaborates with the drowsiness detection system 9 and the shared memory 110,
- a module 714 for monitoring the continuity of the movements of a driver of the motor vehicle, which collaborates with the vigilance monitoring system 10 and the shared memory 110,
- a module 715 for transmitting an authentication response which collaborates with the set of application systems 5.

The motor vehicle 100, in particular the facial recognition system 4, preferably comprises all the hardware and/or software elements configured so as to implement the method defined in the subject of the invention or the method described below.

One mode of execution of the process for authenticating a shot is described below with reference to the . In this mode of execution, the method comprises six steps E0 to E5, the first and second steps E0, E1 following on from the three steps E2, E3 and E4 which are executed in parallel; at the end of the three steps E2, E3 and E4, we move on to a sixth step E5.

In step E0, a position and a geometry of elements 11, 12, 13, 14 of the passenger compartment 1 are initialized from at least one so-called reference shot R1 of the empty passenger compartment 1. Advantageously, this processing is carried out from different shots captured under different luminosity conditions. All the shots analyzed come from image sensor 2.

In one embodiment, different possible positions can be stored for each mobile element of the passenger compartment (for example the headrest of the driver's seat).

Each element 11, 12, 13, 14 of the passenger compartment is associated with a position and a geometry projected in a plane corresponding to the shot plane.

In step E1, an authentication request RA for a candidate shot C1, C2, C3, C4 is received. Depending on the order in which the identification and authentication steps are controlled by the facial recognition system 4, the authentication request RA can come
- from one of the applications secured by facial recognition, for example it can come from the application 52 for managing banking transactions if the authentication is carried out before the identification or simultaneously with the identification,
- of the identification system 6, if the identification is carried out before the authentication.

We then move on to steps E2, E3 and E4 which run simultaneously.

In step E2, more particularly described by FIGS. 6 to 8, it is sought to recognize in the candidate shot C1, C2 elements of the passenger compartment 11, 12, 13, 14 situated at the edge of a face or of a body of a driver of motor vehicle 100.

To this end, in a sub-step E21, the driver's face and body are located in the candidate shot C1, C2 to determine an outline 102 of the driver's face and body.

Then, in a sub-step E22, a set of zones ZP _i , called probationary zones, of the candidate shot C1, C2 distributed over the contour 102 of the driver's face and body is determined, so that each probationary zone includes both a portion of the driver's face and a portion of the passenger compartment. Determining a probationary zone ZP includes determining its position and its geometry.

This step can be executed previously and the zones ZP _i can be predefined by presupposing a location of the driver's head during the shooting.

In the rest of the document, a probationary zone can be denoted ZP(x,y) , x and y designating the coordinates of the probationary zone in the plane of the candidate shot, for example the coordinates of the center of the probationary zone by relation to the center of the candidate shot C1.

Thus, each probationary zone ZP _i is determined by its position (x _i , y _i ) in the plane of the candidate shot C1, and by its geometry G _i . By way of example, the geometry G _i of a probationary zone can be determined as being a rectangle of length L _i and of width l _i .

There illustrates a set of eight probationary zones ZP1 to ZP8 defined for the authentication of the candidate shot C1:
- zone ZP1 comprises a portion of central mirror 11,
- zones ZP3 and ZP4 include an upright of the body 12 corresponding to a portion of the frame of the rear windscreen,
- the ZP5 and ZP7 zones include a portion of the driver's backrest 14,
- zone ZP6 includes a portion of steering wheel 13, and
- zones ZP2 and ZP8 do not include any element of the cockpit.

In a sub-step E23, the content of each probationary zone ZP _i is compared with the content of a reference zone ZR _i of a reference shot R1 of the empty passenger compartment, the reference zone ZR _i being of the same position (x _i , y _i ) and of the same geometry G _i as the probationary zone ZP _i .

For each probationary zone ZPi, the comparison processing excludes the part of the probationary zone ZPi corresponding to the face of the driver. In other words, for each probationary zone, a context zone ZC _i is defined corresponding to the portion of the probationary zone ZP _i representing the passenger compartment, and a comparison processing of the candidate and reference shots is applied only on the set of context zones ZC _i .

For example, the illustrates the comparison of the probationary zone ZP4 with a reference zone ZR4 corresponding to it. The comparison processing excludes the part of the probationary zone ZP4 corresponding to the driver's face. In other words, the comparison processing relates to the context zone ZC4 corresponds to the portion of ZP4 representing the passenger compartment.

There illustrates an example of probationary zones ZP11, ZP12, ZP13, ZP14 and ZP15 placed on a candidate image containing a decoy L1. The detection of probationary zones is imprecise: certain probationary zones ZP13, ZP14 are not located at the edge of the face. In addition, the content of the probationary areas differs from the content of the passenger compartment. For example, zone ZP13 only includes a mirror part 11, zone ZP12 does not include the body pillar 12 and zone ZP13 does not include the backrest of the driver's seat 14.

In one embodiment, the comparison includes
- a calculation of a spatial correlation rate COR _i associated with each of the probationary zones ZP _i will be calculated between the context zone ZC _i of the probationary zone ZP _i and a corresponding reference zone ZR _i , and
a comparison of each correlation rate COR _i with a single minimum CORMIN, the minimum threshold CORMIN possibly being 97% for example.

For example, a correlation rate COR ₄ associated with the probationary zone ZP ₄ will be calculated between the context zone ZC ₄ and the reference zone ZR ₄ .

The spatial correlation rate takes into account the level of similarity between each pixel of a context zone ZC _i and a corresponding pixel of a corresponding reference zone ZR _i . Before comparing the zones, processing can be carried out on the data of one or more of the shots to take account of the fact that the different shots may have been taken under different lighting conditions.

The calculation of each spatial correlation rate COR _i can be performed by a neural network, in particular a convolutional neural network. The training of the neural network comprises a first phase of learning reference zones in shots of the empty passenger compartment then a second phase of learning the recognition of reference zones in candidate shots.

Alternatively, the calculation of the spatial correlation rates can be based on a matrix distance method between the matrix of the image points of the context zones ZC _i of a candidate image and the corresponding image points of a reference shot R1.

In one embodiment, the correlation rate COR _i is a centered normalized cross-correlation ZNCC (with reference to the English acronym “Zero mean Normalized Cross-Correlation”).

According to this calculation method, if we consider a first image i1 in the form of its data vector Ui1 and a second image i2 in the form of its data vector Vi2, then the centered normalized cross-correlation ZNCC is defined by the following formula :

The absolute value of the ZNCC correlation changes between the value 1 when the first and second images are identical and the value 0 when the first and second images are different.

We can consider methods established by the ‘Computer Vision’ domain refining this definition by geometry of the image data.

Each spatial correlation rate COR _i can also be determined by calculation methods based on convolutional neural networks (CNN) dedicated to the correspondence between images, in particular the CHM method (in reference to the English acronym "Convolutional Hough Matching (CHM )») recognized as efficient in the geometric correspondence of images. The CHM method is advantageously used for example in the matching of road content images (motorcycles, pedestrians), the problem of which has similarities with the image comparison needs relating to our invention. The CHM method exploits the correlations within the convolutional neural network in its intermediate layers and calculates at output a CHM coefficient similar to a correlation rate and which can be used as such.

We define a single minimum CORMIN of the correlation rate COR _i between a probationary zone and an associated reference zone, so that:
- a candidate shot C1, C2 will be authenticated if all the correlation rates COR _i are greater than the minimum threshold CORMIN,
- the candidate view C1, C2 will be rejected otherwise.

The minimum threshold CORMIN can for example be set at 97%, or even 98% or 99%. It can also be calibrated to lower values, for example 95% or 90%.

In step E3 which can be executed simultaneously with step E2, the continuity of the identity of the driver is monitored. For this, the driver drowsiness detection system 9 is used.

The system 9 performs throughout the duration of the journey of the motor vehicle 100, a periodic processing of identification of the driver at a period T9, which can for example be equal to one second. When a discontinuity is detected in the identity of the driver, the system 9 issues a diagnosis signaling the identity discontinuity. This diagnosis is advantageously transmitted to the control unit 7, for example via the shared memory 110.

In step E3, it is possible if necessary to calibrate the period T9 of the system 9 according to the needs relating to the authentication processing.

For example, when it is identified that the vehicle is approaching a parking lot payment terminal, the period T9 can be modified to a value T=0.5 seconds to more finely monitor the continuity of the driver's face and detect a possible presentation of a decoy.

Figures 9 and 10 illustrate the progress of step E3.

We can assume, for example, that the vehicle 100 arrives near a toll. At time T0, we anticipate that a bank transaction will be requested. The period T9 of the drowsiness detection system 9 is then reduced to the value T=0.5 seconds. At a given instant TC, a shot C3 of the driver is captured to authenticate the bank transaction.

In step E3, contrary to what is done in step E2, the TC view is not analyzed. To authenticate the TC shot, the diagnosis issued by the system 9 on the continuity of the driver's identity is used, this diagnosis taking into account the C3 shot.

The graph G1 of the illustrates the continuity of the identity of the driver between a first period DT1 preceding the shot C3, a second period DT2 of capturing a shot and a third period DT3 following the period of capturing a shot. In the case illustrated by , the diagnosis issued by the system 9 concerning the continuity of the driver's identity will rule on the absence of an incident.

In a preferred embodiment of step E3, the shot C3 will be considered authentic only if the identity of the driver is continuous over the total duration of an uninterrupted trip between a moment when the driver climbs into the motor vehicle 100 and an end instant of the shooting period DT2.

The G3 graph of the illustrates a discontinuity of the identity of the driver between a fourth period DT4 preceding a capture of a shot C4, a fifth period DT5 of capture of the shot C4. At the instant TC of capture of the shot C4 used to authenticate the payment, the driver conceals his face behind a decoy. The identity of the driver identified by the drowsiness detection system 9 at the instant TC is therefore different from the identity of the driver over the period DT4 preceding the capture of the shot C4. The detection of this discontinuity by the system 9 generates an incident diagnosis in the identification of the driver, the incident diagnosis being advantageously recorded in the shared memory 110.

Thus, in step E3, the authenticity of a shot C3, C4 is determined by a diagnosis provided by the system 9 for detecting drowsiness. This diagnosis relates to the continuity of the identity of the driver between the instant of capture of the shot and a period of time preceding the instant of capture of the shot, the period of time being able to cover the total duration of an uninterrupted path between a moment when the driver climbs into the motor vehicle 100 and the moment when the shot is captured.

In step E3, after a discontinuity has been detected in the identity of the driver, it is possible to command a reduction in the period T9 of the drowsiness detection system in order to better analyze the behavior of the driver in order to confirm or not a attempted fraud.

Conversely, following the successful authentication of a shot, it is possible to command an increase in the period T9 of the drowsiness detection system, in order to reduce the frequency of driver identity checks , and thus reduce the computational load of the system 9.

In a first embodiment of step E4, which can be executed simultaneously with step E2 and/or step E3, the continuity of the movements of the driver is monitored by analyzing the images from the sensor 2. This monitoring is carried out over a second continuous time interval including a first time interval DT2, DT5 of shots including the candidate shot C1, C2, C3, C4, and the second time interval starting at the latest at the start of the first time interval DT2, DT5, the first time interval being for example initiated by an authentication request sent by an application 51, 52, 53.

In one embodiment, the delay between the start of the second time interval and the start of the first time interval is a multiple of the period T10, for example this delay is equal to six times the period T10.

According to a second embodiment of step E4, complementary or alternative to the first embodiment previously described, step E4 comprises recognition of the driver's movements from among the predefined set E of movements compatible with a driving situation, by analyzing the images from the sensor 2 over said second time interval.

In other words, in step E4, to process an authentication request RA sent at a time TReq and relating to the authentication of shots taken over a first time interval DT2, DT5
- a second time interval beginning before the instant TReq and comprising the first time interval DT2, DT5 is considered, and
- over the entire duration of the second time interval, the continuity of the driver's movements and/or the recognition of the driver's movements among the predefined set E is checked.

For this, the system 10 for monitoring the vigilance of a driver is used.

The system 10 performs throughout the duration of the journey of the motor vehicle 100, a periodic processing of analysis of the movements of the driver at a period T10, which can for example be equal to one second. In particular, the analysis of the driver's movements includes a recognition of movements compatible with automobile driving. For example, as illustrated by FIGS. 9 and 10, the system 10 performs recognition of the placement 71 of the driver's hands on the steering wheel of the vehicle. Thus, the detection of an incompatible movement of driving (for example when the driver turns towards the rear passengers while the vehicle is moving), a visual and/or sound alert can be emitted by the system 10 towards the driver.

The analysis of the movements carried out by the system 10 further comprises a detection of the discontinuity of the movements of the driver. In one embodiment, it could advantageously include detection of a set of movements making it possible to suspect authentication fraud, such as for example a movement of presentation of a decoy to the sensor 2 as illustrated in the , at time TC. In this case, the system 10 can store an anomaly diagnosis on the analysis of the movements of the driver in the shared memory 110.

In step E4, it is possible if necessary to calibrate the period T10 of the system 10 according to the needs relating to the authentication processing.

For example, when it is identified that the motor vehicle 100 is approaching a parking lot payment terminal, the period T10 can be modified to a value T=0.5 seconds to monitor the continuity of the driver's movements more precisely. and detecting a possible presentation of a decoy.

Figures 9 and 10 illustrate the progress of step E4.

It can be assumed, for example, that the motor vehicle 100 arrives near a toll. At time T0, we anticipate that a bank transaction will be requested. The period T10 of the vigilance monitoring system 10 is then reduced to the value T=0.5 seconds. At a given instant TC, a shot C3 of the driver is captured to authenticate the bank transaction.

In step E4, as in step E3, the shot C3 is not analyzed. To authenticate the shot C3, the diagnosis transmitted by the system 10 on the continuity of the movements of the driver is used, this diagnosis taking into account the shot C3.

The G2 graph of the illustrates the continuity of the driver's movements between a first period DT1 preceding the shot C3, a second period DT2 of capturing a shot and a third period DT3 following the period of capturing a shot. In the case illustrated by , the diagnosis issued by the system 10 concerning the continuity of the identity of the driver will rule on the absence of an incident.

The G4 graph of the illustrates a discontinuity of the driver's movements between a fourth period DT4 preceding a capture of a shot C4, a fifth period DT5 of capture of the shot C4. At the instant TC of capture of the shot C4 used to authenticate the payment, the driver conceals his face behind a decoy. The position of the hands detected by the alertness monitoring system 10 at the instant TC is therefore different from the position of the driver's hands over the period DT4 preceding the capture of the shot C4. Moreover, this position can be identified among a set of movements allowing authentication fraud to be suspected. The detection of this discontinuity by the system 10 generates an incident diagnosis in the continuity of the driver's movements, the incident diagnosis being advantageously recorded in the shared memory 110.

Thus, in step E4, the authenticity of a shot C3, C4 is determined by a diagnosis provided by the vigilance monitoring system 10. This diagnosis mainly relates to the continuity of the driver's movements between the instant of capture of the shot and a period of time preceding the instant of capture of the shot. This diagnosis can also relate to the recognition of the driver's movements among a set of movements making it possible to suspect authentication fraud.

In step E4, when an anomaly is detected by the vigilance monitoring system 10, it is possible to command a reduction in the period T10 of the vigilance monitoring system in order to better analyze the behavior of the driver.

Conversely, following the successful authentication of a shot, it is possible to command an increase in the period T10, in order to reduce the frequency of verifications of the movements of the driver, and therefore the calculation load of the system. 10.

In a step E5, the result of the authentication of the shot is transmitted to the system having required authentication of a shot, that is to say to an application 51, 52, 53 originally the RA authentication request or the identification system 6.

The authentication result is obtained by aggregating the results of steps E2 to E4. Thus, the result of the three steps E2, E3, E4 is grouped together in a final authentication result delivered in step E5.

In one embodiment, a shot is authenticated only if it has been authenticated in all of the three steps E2 to E4.

In an alternative embodiment, the authentication result comprises a confidence index calculated from the diagnostics of steps E2 to E4. For example, the confidence index can take into account both the spatial correlation rate COR calculated in step E2, and the processing periods T9, T10 of systems 9 and 10.

More generally, it is possible to define logical fraud scores S2 to S4 respectively associated with each of the steps E2 to E4, so that:
- S2=1 if fraud has been detected in step E2, otherwise S2=0, and
- S3=1 if fraud has been detected in step E3, otherwise S3=0, and
- S4=1 if fraud has been detected in step E4, otherwise S4=0.

The aggregation of the results of steps E2 to E4 can then be determined by a so-called aggregation function having as input parameters the fraud scores S2 to S4 and calculating an aggregated fraud score S5 as output. In other words: S5 = f(S2, S3, S4).

For example, the aggregation function can be a logical AND function, the S5 score then being obtained according to the formula:
S5 = S2 AND S3 And S4

In this embodiment, the aggregated fraud score is equal to 1 only if all the scores S2 to S4 are equal to 1. In other words, in step E5 the authentication will be considered to be fraudulent only if a fraud has been detected in each of steps E2 to E4.

Alternatively, the aggregation function can be a weighted aggregation making it possible to calibrate the relative importance of steps E2, E3, and E4 in the detection of fraudulent authentication. The S5 score is then obtained according to the following formula:
S5 = a.S2 + b.S3 + c.S4, where a + b + c = 1 and a, b and c are positive.

Thus, if it is desired to assign a high weight to the score S2 resulting from step E2 and a low weight to the scores S3 and S4 resulting respectively from steps E3 and E4, it is possible, for example, to assign the following values to the coefficients a , b, and c: a=80%, b=10% and c=10%.

The value of the coefficients a, b and c can change during the course of the process according to the invention. For example, the value of the coefficients a, b, and c can take into account the quality of the capture which could influence the relative relevance of the steps E2, E3 and E4.

The aggregated fraud score S5, calculated using one of the methods previously described, is then compared to a minimum threshold beyond which fraudulent authentication is detected. For example, fraud can be detected when the S5 aggregate fraud score is greater than 95%.

There is an alternative representation of a method for authenticating a shot of a passenger compartment comprising a step E2 of recognizing elements of the passenger compartment in the shot:
- In a first step F1, a shot of the passenger compartment is captured.
- Then, in a second step F2, the face of the driver is located.
- Then, in a third step F3, N context zones ZC _i (x _i , y _i ) are determined from N probationary zones ZP _i (x _i , y _i ).
- Then, in a fourth step F4, from a reference shot R1 of the empty passenger compartment recorded in the local memory 72, the N reference zones ZR _i (x _i , y _i ) associated with the N context zones ZC _i (x _i , y _i ).
- Then, in a fifth step F5, the N spatial correlation rates T _i are calculated respectively between each context zone and the associated reference zone.
- Then, in a sixth step F6, a Boolean fraud indicator IF1 is determined by comparing each spatial correlation rate Ti with a minimum threshold CORMIN, the indicator IF1 being initialized to the value 0, and the indicator IF1 being fixed at the value 1 if at least one spatial correlation rate Ti is strictly lower than CORMIN.
- Then, in a seventh step F7, the value of the fraud indicator IF1 is tested: if the indicator IF1 is equal to 1, we go on to a step F8 of rejection of the authentication, otherwise, we go on to a authentication validation step F9.

There is an alternative representation of a method for authenticating a shot of a passenger compartment comprising a step E3 of monitoring the continuity of the identity of a driver of the motor vehicle. The process loops back, at a given period T9, to the following steps:
- At a time t, in a first step H1, a shot of the passenger compartment is captured.
- Then, in a second step H2, an identification processing is carried out to find the face of the driver among a set of persons authorized to drive the vehicle, the facial biometric fingerprints necessary for this processing being contained in the shared memory 110.
- Then, in a third step H3, on the basis of the previous steps, an identifier Id(t) of the driver of the vehicle at time t is determined,
- Then in a fourth step H4, a Boolean fraud indicator IF2 is determined by comparing the identifier Id(t) of the driver at time t with the identifier Id(t-T9) of the driver at time t- T9: the value of the IF2 indicator is set to 0 if Id(t)= Id(t_T9), otherwise the value of the IF2 indicator is set to 1.
- Then, in a fifth step H5, the value of the fraud indicator IF2 is tested: if the indicator IF2 is equal to 1, we continue with a step H7 of rejection of the authentication, and we reduce the period of verification T9 if necessary, otherwise we go on to a step H6 for validating the authentication and we increase the verification period T9 if necessary.

There is an alternative representation of a method for authenticating a shot of a passenger compartment comprising a step E4 of monitoring the continuity of the movements of a driver of the motor vehicle. The process loops back, at a given period T10, to the following steps:
- At a time t, in a first step J1, a shot of the passenger compartment is captured.
- Then, in a second step J2, an identification processing is carried out to find the posture of the driver among a set of reference positions (or postures) compatible with driving the vehicle, the reference positions necessary for this processing being contained in the shared memory 110.
- Then, in a third step J3, on the basis of the previous steps, a position identifier Id(t) of the driver of the vehicle is determined at time t.
- Then, in a fourth step J4, a Boolean fraud indicator IF3 is determined by comparing the position Pos(t) of the driver at time t and the position Pos(T-T10) of the driver at time t-T10 : we set the value of the IF3 indicator to the value 0 if the positions Pos(t) and Pos(t-T10) are in continuity with each other, otherwise we set the value of the IF3 indicator to 1.
- Then, in a fifth step J5, the value of the fraud indicator IF3 is tested: if the indicator IF3 is equal to 1, we move on to a step J7 of rejection of the authentication, and we reduce the period of verification T10 if necessary, otherwise, we go on to a step J6 of validation of the authentication and we increase the verification period J10 if necessary.

Finally, the authentication method according to the invention makes it possible to take advantage of the specific context of facial recognition in the automotive field.

The specificity of the automotive context primarily concerns the invariability of the framework in which the facial recognition processing takes place, in particular the invariability of the position and geometry of certain elements of the passenger compartment.

The invention thus provides a first authentication solution consisting in checking only a few so-called probationary zones in the shot to be authenticated, the probationary zones comprising part of the driver's face and part of the passenger compartment. This first solution has several advantages. First of all, it is difficult to bypass by creating decoys. In addition, it is fast, which is valuable for users when authentication requests are repetitive.

Furthermore, the specificity of the automotive context relates to the presence of at least two driver management systems in any vehicle: a drowsiness detection system and an alertness monitoring system. These two systems will soon be mandatory in all motor vehicles. However, these systems respectively provide periodic diagnostics on the identity and movements of the driver. By exploiting these diagnostics, the invention thus provides a second and a third authentication solution requiring little processing time and reusing processing already implemented in the vehicle.

In addition to the advantage of reducing the processing time compared to an authentication system that would not be specific to the automotive context, the invention makes it possible to avoid the purchase of licenses relating to market algorithms.

In addition, the first authentication solution makes it possible to very quickly reject a fraudulent shot because the authentication step can be carried out before the driver identification step.

The invention also makes it possible to combine the three authentication solutions described above, which reinforces the reliability of the authentication.

Claims (12)

Method for authenticating a candidate shot (C1, C2, C3, C4) by an image sensor (2) fixed in a passenger compartment (1) of a motor vehicle (100), characterized in that it includes at least one of the following steps:

• a step (E2) of recognition, in the candidate shot (C1, C2, C3, C4), of elements of the passenger compartment (11, 12, 13, 14) situated at the edge of a face or a body of a motor vehicle driver (100),
• a step (E3) of monitoring, by analysis of several images from the sensor (2), the continuity of an identity of a driver of the motor vehicle (100) over the total duration of an uninterrupted journey between a moment of boarding of a driver in the motor vehicle (100) and an end time of a first time interval (DT2, DT5) of shots including the candidate shot (C1, C2, C3, C4),
• a step (E4) of monitoring the continuity of movements of a driver of the motor vehicle and/or of recognizing movements of a driver of the motor vehicle from among a predefined set (E) of movements compatible with a driving situation, by analysis of the images from the sensor (2) over a second continuous time interval including a first time interval (DT2, DT5) of shots including the candidate shot (C1, C2, C3, C4), and the second time interval beginning at the latest at the start of the first time interval, the first time interval being for example initiated by an authentication request sent by an application of a motor vehicle.

Authentication method according to the preceding claim, characterized in that it comprises a step (E1) of receiving an authentication request (Req) for a shot (C1, C2, C3, C4), the authentication request (Req) coming from an application secured by facial recognition (51, 52, 53) of the motor vehicle. Authentication method according to one of the preceding claims, characterized in that it comprises, preceding the step (E1) of receiving a request, a step of initializing (E0) a position and a geometry of a set of elements (11, 12, 13) of the passenger compartment (1) from at least one reference shot (R1) of the empty passenger compartment captured by the sensor (2),
and in that the recognition step (E2) comprises
- a sub-step of (E21) for determining an outline (102) of a driver's face and body,
- a sub-step (E22) of determining a set of so-called probationary zones (ZP _i ) of the candidate shot (C1), distributed over the contour (102) of a face and a body of a driver, so that each zone (ZP _i ) of the set of probationary zones comprises both a portion of a driver's face and a portion of the passenger compartment, the determination of a probationary zone comprising the determination of its position and geometry, and
- a comparison sub-step (E23) of the content of each zone (ZP _i ) of the set of probationary zones with the content of a reference zone (ZR _i ) of the reference shot (R1), the reference zone (ZR _i ) being at the same position and with the same geometry as the probationary zone (ZP _i ). Authentication method according to the preceding claim, characterized in that the comparison sub-step (E23) is performed, for each probationary zone (ZP _i ), on a context sub-zone (ZC _i ) corresponding to a portion of the probationary zone (ZP _i ) excluding a face and/or a body of a driver. Authentication method according to the preceding claim, characterized in that the comparison sub-step (E23) comprises a calculation of a spatial correlation rate (COR _i ) between each of the context zones (ZC _i ) and a corresponding reference (ZR _i ), and a comparison of each correlation rate (COR _i ) with a single minimum (CORMIN),
in that the candidate shot (C1) is authenticated if all the correlation rates (COR _i ) are greater than the minimum threshold (CORMIN) and in that the candidate shot (C1) is rejected in the opposite case. Authentication method according to the preceding claim, characterized in that the spatial correlation rates (COR _i ) are calculated by a neural network, in particular by a convolutional neural network. Authentication method according to one of the preceding claims, characterized in that it comprises the step of monitoring the continuity of movements of a driver of the motor vehicle and/or of recognizing movements of a driver of the motor vehicle taking place at a given period (T10) and in that the delay between the start of the second time interval and the start of the first time interval is equal to six times or about six times the given period (T10). Authentication method according to one of the preceding claims, the motor vehicle (100) being equipped with a system (9) for detecting the drowsiness of a driver, the system (9) for detecting the drowsiness of a driver receiving shots (C1, C2, C3, C4) captured by the sensor (2), characterized in that the step (E3) of monitoring the continuity of an identity of a driver comprises receiving a diagnosis resulting from the system (9) for detecting the drowsiness of a driver. Authentication method according to one of the preceding claims, the motor vehicle (100) being equipped with a system (10) for monitoring the vigilance of a driver, the system (10) for monitoring the vigilance receiving sockets of views captured (C1, C2, C3, C4) by the sensor (2), characterized in that the step (E4) of monitoring the continuity of movements of a driver comprises receiving a diagnosis from the system (10) monitoring the vigilance of a driver. Device (20) for authenticating a candidate shot of a passenger compartment (1) of a motor vehicle (100) equipped with an image sensor (2), the device comprising elements (6, 7 , 8, 9, 10, 71, 72, 73, 110, 710, 711, 712, 713, 714, 715) hardware and/or software implementing the method according to one of claims 1 to 9, in particular elements hardware (6, 7, 8, 9, 10, 71, 72, 73, 110) and/or software designed to implement the method according to one of the preceding claims. A computer program product comprising program code instructions recorded on a computer-readable medium for carrying out the steps of the method according to any one of claims 1 to 9 when said program runs on a computer. Data storage medium, readable by a computer, on which is recorded a computer program comprising program code instructions for implementing the method according to one of Claims 1 to 9.