Image processing method in a motor vehicle

Abstract

The invention relates to a method for authenticating or identifying a person using a camera (10), comprising the steps of generating a sequence of images by the camera (10), and for each image of said image sequence, of detecting (E1) at least one person in said image, of defining (E2) in the image at least one area called "region of interest" for each detected person or group of detected persons, and for each identified region of interest, of rectifying (E3) the region of interest, of detecting (E4) the face of at least one person in the rectified region of interest, of generating (E5) a 3D model representing each detected face, and of rotating (E6) each generated 3D model so that said face is visible from the front in the acquisition plane of the camera, an authentication or identification (E7) of the at least one person from the at least one face having undergone the rotation. Figure for the abbreviation: Figure 1

Description

Title of the invention: Image processing method in a motor vehicle. Technical field

[0001] The present invention relates to the field of authentication and identification and more particularly concerns a method for processing images generated by a camera in order to identify or authenticate a person, in particular for use in a motor vehicle. Previous technique

[0002] Nowadays, it is known to use on-board cameras to assist the driver of a vehicle, for example during reversing phases with a camera placed in the rear of the vehicle (trunk lid or rear bumper) or to allow a vehicle user access to the passenger compartment. In the latter case, the camera is mounted at the driver's side door.

[0003] In order to provide a wide view while reducing the size of onboard cameras, it is common to use panoramic cameras, known as "fisheye" cameras, in vehicles. These cameras offer a 360° view without blind spots, particularly for authenticating or identifying a vehicle user to grant them access to the passenger compartment. Authentication involves determining whether the face of a person appearing in the images is a real face or a fake one, for example, printed on a medium, while identification involves comparing the face of a person appearing in the images with images of faces stored in a memory area to determine the person's identity.

[0004] With this type of panoramic camera and in the absence of specific processing, 360° images appear distorted. It is known to preprocess them before analyzing them to detect a face for the purpose of authenticating or identifying it.

[0005] In existing solutions, this preprocessing consists of a rectification of the images generated by the panoramic camera, which are then processed to detect a face.

[0006] However, it appears that defects such as occlusion and / or distortion may occur with these existing processes, which may lead to a failure of authentication or identification.

[0007] Similarly, with existing methods, varying lighting conditions can reduce the accuracy or sharpness of images, which can again lead to authentication or identification failure and / or increased vulnerability to attacks identity theft, for example using 3D masks, presents a risk to vehicle security.

[0008] Finally, existing processes may reduce the performance of authentication or identification due to insufficient robustness to deal with the variability of pose, appearance and expression of faces, which also presents a disadvantage in terms of security.

[0009] A simple, reliable and effective solution that would at least partially remedy these drawbacks would therefore be advantageous. Description of the invention

[0010] To this end, the invention first relates to a method for authenticating or identifying a person using a camera configured to generate two-dimensional images in an acquisition plane, said method comprising the steps of:

[0011] - generation of an image sequence by the camera,

[0012] - for each image in said image sequence:

[0013] - detection of at least one person in said image,

[0014] - definition in the image of at least one area called "region of interest" for each person detected or group of people detected,

[0015] - for each identified region of interest:

[0016] - rectification of the region of interest,

[0017] - detection of the face of at least one person in the rectified region of interest,

[0018] - generation, also called reconstruction, of a 3D model representing each Face detected

[0019] - rotation of each generated 3D model so that said face is visible from the front in the camera acquisition plan,

[0020] - authentication or identification of at least one person from at least a face that has undergone rotation.

[0021] The method according to the invention facilitates the authentication or identification of an authorized user based on a 3D facial model created from a rectified region of interest in the image and rotated as if the user were facing the camera, thereby improving the quality of the authentication or identification. In particular, the method according to the invention provides a better input image for facial recognition algorithms, due to the increased pixel input resulting from the rectification and construction of 3D facial models; it mitigates the impact of distortion on the accuracy of identification algorithms; it increases the security of the access system by preventing identity theft attacks; and it improves performance. global facial identification algorithms in driver assistance systems where system efficiency is crucial.

[0022] Preferably, the method is implemented by an electronic control unit of a motor vehicle, with at least one camera mounted in said vehicle. Alternatively, the method can be implemented by any image processing system outside a vehicle, for a camera mounted inside or outside a vehicle, for example on a security gate or an ATM.

[0023] Preferably, the method includes, among other things, the steps of tracking at least one person detected in a previous image into a subsequent image of the generated image sequence, implementing the method on the subsequent image, and merging the 3D facial model of the tracked person generated from the subsequent image with the 3D facial model of the tracked person generated in the previous image. Tracking each person, and therefore each face, makes it possible to create a succession of 3D models of each face for merging and further improving authentication and / or identification.

[0024] Advantageously, the generation of the 3D model is carried out using a 3D candid face mask type process, which allows simple and detailed modeling.

[0025] In one embodiment, the method includes, following the detection of the face of at least one person in the rectified region of interest, an attempt to authenticate or identify the face, the generation of a model representing each detected face and subsequent steps being carried out only if the authentication or identification has failed.

[0026] Preferably, the camera is a panoramic camera, allowing 360° "fish-eye" type images.

[0027] The invention also relates to a computer program product characterized in that it comprises a set of program code instructions which, when executed by one or more processors, configure the processor(s) to implement a process as described above.

[0028] The invention also relates to an electronic control unit for a motor vehicle, said electronic control unit being configured to receive a sequence of images from a vehicle camera and, for each image in said image sequence, to:

[0029] - detect at least one person in said image,

[0030] - define in the image at least one area called "region of interest" for each person detected or group of people detected,

[0031] - for each identified region of interest:

[0032] - rectify the region of interest,

[0033] - detect the face of at least one person in the rectified region of interest,

[0034] - generate a model representing each detected face,

[0035] - rotate each generated model so that said face is visible facing the camera in its acquisition plane,

[0036] - to authenticate or identify at least one person from at least one face having undergone rotation.

[0037] Preferably, the electronic control unit is further configured to:

[0038] - follow at least one person detected in a previous image in an image next in the sequence of generated images,

[0039] - implement the process on the following image,

[0040] - merge the face model of the tracked person generated from the image next with the face model of the person being tracked generated in the previous image.

[0041] Advantageously, the electronic control unit is configured to generate the model using a 3D candid face mask type process.

[0042] In one embodiment, the electronic control unit is configured to, following the detection of the face of at least one person in the rectified region of interest, perform an attempt at face authentication or identification, the generation of a model representing each detected face and subsequent steps being carried out only if authentication or identification has failed.

[0043] The invention also relates to a motor vehicle comprising an electronic control unit as described above and at least one camera configured to generate a sequence of images and transmit it to said electronic control unit.

[0044] Preferably, at least one camera is a panoramic camera. Brief description of the drawings

[0045] Other features and advantages of the invention will become apparent from the following description. This description is purely illustrative and should be read in conjunction with the accompanying drawings, in which:

[0046] [Fig-1] Fig.1 schematically illustrates one embodiment of the system according to the invention.

[0047] [Fig.2] Fig.2 schematically illustrates one embodiment of the process according to the invention. Description of the implementation methods

[0048] Fig. 1 illustrates an example of a motor vehicle according to the invention.

[0049] The vehicle 1 includes a camera 10 and an electronic control unit 20.

[0050] The camera 10 is a panoramic camera configured to generate a sequence of images II, 12, ..., In (with reference to [Fig.2]) in two dimensions in a so-called "acquisition" plane and to transmit said sequence of images to the electronic control unit 20.

[0051] The electronic control unit 20 is configured to receive a sequence of images from the camera 10.

[0052] The electronic control unit 20 is configured, for each image of said image sequence, to detect at least one person in said image, to define in the image at least one area called "region of interest" for each person detected or group of people detected.

[0053] The electronic control unit 20 is configured, for each identified region of interest, to rectify said region of interest, to detect the face of at least one person in the rectified region of interest, to generate a model representing each detected face, to perform a rotation of each generated model so that said face is visible from the front in the camera's acquisition plane, and to authenticate or identify the at least one person from the at least one face that has undergone the rotation.

[0054] The electronic control unit 20 is configured to track at least one person detected in a previous image in a subsequent image of the generated image sequence, to implement the process on the subsequent image, and to merge the face model of the tracked person generated from the subsequent image with the face model of the tracked person generated in the previous image.

[0055] The electronic control unit 20 is configured to generate the model produced using a 3D candid face mask type process.

[0056] The electronic control unit is configured to, following the detection of the face of at least one person in the rectified region of interest, perform an attempt at face authentication or identification, the generation (also called reconstruction) of a 3D model representing each detected face and subsequent steps being carried out only if authentication or identification has failed.

[0057] In material terms, the electronic control unit includes a processor capable of implementing a set of instructions to perform these different functions.

[0058] Example of implementation of the invention

[0059] An example of implementation of the process will now be described with reference to the [Fig.2],

[0060] To implement the method, the camera 10 generates a sequence of images and transmits it to the electronic control unit 20. For each image in the sequence of images, the electronic control unit 20 will perform a plurality of processing steps.

[0061] First, the electronic control unit 20 detects one or more persons in the first image II in a step E1, then defines an area called the "region of interest" for each detected person or group of detected persons in a step E2. A group of persons corresponds to several persons located in the same area of ​​the image.

[0062] For each identified region of interest, the electronic control unit 20 rectifies the area of ​​the first image II corresponding to said region of interest in a step E3. Rectification is the process of applying geometric corrections to an image, in this case to a portion of an image, based on information extracted from a source image (or raw image) according to the chosen geometry model. It is then said that the raw image is rectified, and the synthesized image is called the rectified image. Such a process is known per se and will not be described in further detail here.

[0063] The electronic control unit 20 then detects the face of at least one person in the rectified region of interest in a step E4.

[0064] Once detection has been carried out, the electronic control unit 20 can proceed to attempt to authenticate or identify the detected face(s).

[0065] If authentication or identification fails, the electronic control unit 20 generates (or reconstructs) a 3D model representing each face detected in a step E5. This generation of the 3D model is preferably carried out using a 3D candid face mask type process, known per se, or any other suitable method.

[0066] The electronic control unit 20 then performs a rotation of each 3D model generated in a step E6 so that said face is visible from the front in the acquisition plane of the camera 10.

[0067] The electronic control unit 20 makes an attempt to authenticate or identify at least one person from at least one modeled face which has undergone rotation in a step E7.

[0068] In the absence of authentication or identification of a person in the first image II, the electronic control unit 20 carries out, in the second image 12 of the sequence of generated images, a tracking of the person or persons detected in the first image, in a manner known per se, in a step E8 and then implements the steps E1 to E6 of the process described above on the second image 12 as well as the steps E61 and E62 described below.

[0069] In a step E61, the electronic control unit 20 checks whether face tracking detected in step E4 has been carried out through step E8.

[0070] If so, the electronic control unit 20 performs the fusion of the 3D face model of the tracked person generated from the second image 12 and the 3D face model of the tracked person generated in the first image II.

[0071] The electronic control unit 20 repeats the process on each image of the image sequence by merging the successive models of each face until an authorized user of the vehicle 1 is authenticated or identified in a step E7 and preferably activates a function of the vehicle such as, for example, unlocking the doors.

[0072] Identification can be achieved by comparing the 3D model of a face with an image of the face of an authorized user or a 3D model of the face of an authorized user stored in a memory area of ​​the electronic control unit 20.

[0073] In the event that no person is authenticated or identified, no function of vehicle 1 is activated.

[0074] The invention makes it possible to rectify only the region or regions of interest of the image in order to allow easy use of a 3D face model which is turned to face the camera and allow authentication or identification.

Claims

Demands

1. A method for authenticating or identifying a person using a camera (10) configured to generate two-dimensional images in an acquisition plane, said method comprising the steps of: - generating a sequence of images by the camera (10), - for each image in said image sequence: - detecting (E1) at least one person in said image, - defining (E2) in the image at least one area called a "region of interest" for each detected person or group of detected persons, - for each identified region of interest: - rectifying (E3) the region of interest, - detecting (E4) the face of at least one person in the rectified region of interest, - generating (E5) a 3D model representing each detected face, - rotating (E6) each generated 3D model so that said face is visible from the front in the camera's acquisition plane,- authentication or identification (E7) of at least one person based on at least one face that has undergone rotation.

2. A method according to claim 1, comprising inter alia the steps of: - tracking at least one person detected in a previous image in a subsequent image of the generated image sequence, - implementing the method on the subsequent image, - merging the 3D face model of the tracked person generated from the subsequent image with the 3D face model of the tracked person generated in the previous image.

3. A method according to any one of the preceding claims, wherein the generation of the 3D model is carried out using a 3D candid face mask type method.

4. A method according to any one of the preceding claims, comprising, following the detection of the face of at least one person in the rectified region of interest, an attempt to authenticate or identify the face, the generation of a 3D model representing each detected face and subsequent steps being carried out only if authentication or identification has failed.

5. A method according to any one of the preceding claims, wherein the camera (10) is a panoramic camera

6. Product computer program characterized in that it comprises a set of program code instructions which, when executed by one or more processors, configure the processor(s) to implement a method according to any one of the preceding claims.

7. Electronic control unit (10) for a motor vehicle (1), said electronic control unit (20) being configured to receive a sequence of images from a camera (10) of the vehicle (1) and, for each image of said image sequence, to: - detect at least one person in said image, - define in the image at least one area called "region of interest" for each detected person or group of detected persons, - for each identified region of interest: - rectify the region of interest, - detect the face of at least one person in the rectified region of interest, - generate a 3D model representing each detected face, - perform a rotation of each generated 3D model so that said face is visible from the front in the acquisition plane of the camera, - authenticate or identify the at least one person from the at least one face that has undergone the rotation.

8. Electronic control unit (20) according to claim 7, further configured to: - track at least one person detected in a previous image in a subsequent image of the generated image sequence, - implement the method on the subsequent image, - merge the 3D face model of the tracked person generated from the subsequent image with the 3D face model of the tracked person generated in the previous image.

9. Electronic control unit (20) according to any one of claims 7 or 8, configured to generate the 3D model made using a 3D candid face mask type process. 10

10. Motor vehicle (1) comprising an electronic control unit (20) according to any one of claims 7 to 9 and at least one camera (10) configured to generate a sequence of images and transmit it to said electronic control unit (20).

Images