Method and device for determining depth using a monoscopic vision system learned by supervision of a stereoscopic vision system

Abstract

A method or device implementing a method for determining the depth of a pixel by a depth prediction model learned in a learning phase comprising receiving (31) first and second images acquired by two cameras of a vision system, cropping (32) these images to generate a first and second cropped image comprising pixels corresponding to objects in the scene seen by the two cameras, predicting (34) first depths associated with the pixels of the first cropped image by another depth prediction model, and learning the depth prediction model by minimizing a difference between the first and second depths associated with the pixels of the first cropped image predicted by the depth prediction model. Figure for the abstract: Figure 3

Description

Title of the invention: Method and device for determining depth using a monoscopic vision system learned by supervision of a stereoscopic vision system. Technical field

[0001] The present invention relates to methods and devices for determining depth using a vision system mounted in a vehicle, for example, in a motor vehicle. The present invention also relates to a method and device for measuring the distance between an object and a vehicle equipped with a vision system. Technological background

[0002] Many modern vehicles are equipped with Advanced Driver-Assistance Systems (ADAS). Such ADAS systems are passive and active safety systems designed to eliminate human error in driving all types of vehicles. ADAS systems use advanced technologies to assist the driver while driving and thus improve performance. ADAS systems use a combination of sensor technologies to perceive the environment around a vehicle and then provide information to the driver or act on certain vehicle systems.

[0003] There are several levels of ADAS, such as reversing cameras and blind spot sensors, lane departure warning systems, adaptive cruise control or automatic parking systems.

[0004] The AD AS systems embedded in a vehicle are powered by data obtained one or more onboard sensors, such as cameras. These cameras make it possible to detect and locate other road users or potential obstacles around a vehicle in order to, for example: • to adapt the vehicle's lighting according to the presence of other road users; • to automatically regulate the vehicle's speed; • to act on the braking system in case of risk of impact with an object.

[0005] Training a depth prediction model associated with one of these cameras is tedious and requires a large amount of good-quality training data. The training data needed for the self-monitoring of a depth prediction model associated with a monoscopic vision system, i.e., the images acquired by a single camera, must not include pixels corresponding to dynamic objects in a three-dimensional scene. The observed data must also be acquired by the moving camera. Therefore, the training data requires prior processing to enable self-monitoring of the training of the depth prediction model associated with the monoscopic vision system. It is also preferable to use annotated data for such training.

[0006] Moreover, the known methods of learning a depth prediction model associated with a monoscopic vision system require working on all or a large part of the images acquired by the camera of the monoscopic vision system, but such learning requires a long learning time. Summary of the present invention

[0007] One object of the present invention is to solve at least one of the problems of the technological background described above.

[0008] Another object of the present invention is to improve the quality of data from the processing of an image acquired by a vision system, in particular by a depth prediction model implemented by a neural network making it possible to determine a depth associated with a pixel corresponding to an object located in the field of vision of a camera.

[0009] Another object of the present invention is to improve road safety, in particular by improving the reliability of AD AS systems powered by data obtained from a vision system.

[0010] According to a first aspect, the present invention relates to a method for determining the depth of a pixel of an image acquired by a first camera of a vision system comprising the first and a second camera arranged so as to each acquire an image of a three-dimensional scene from a different point of view, the depth being determined by a depth prediction model associated with the first camera, called a monoscopic prediction model, implemented by a convolutional neural network, the method being implemented by at least one processor, and being characterized in that the monoscopic prediction model is learned in a learning phase comprising the following steps: - reception of representative data of a first image and a second image acquired by respectively the first camera and the second camera at the same time instant of acquisition; - cropping of the first and second images to generate a first cropped image and a second cropped image, the first cropped image including pixels from the first image corresponding to objects in the three-dimensional scene seen by both the first and second cameras, and the second cropped image including pixels from the second image corresponding to said objects of the three-dimensional scene seen by both the first and second cameras, the first and second cropped images having the same definition; - training a depth prediction model associated with the first and second cameras, called a stereoscopic prediction model, from the first and second cropped images; - prediction of initial depths associated with the pixels of the first cropped image by the stereoscopic prediction model from the first and second cropped images; and - learning the monoscopic prediction model on a part of the first image by minimizing a result of a comparison of the first depths to second depths predicted by the monoscopic prediction model from the first cropped image and associated with the pixels of the first cropped image.

[0011] Such a learning process becomes more efficient by resizing the images to process only pixels located within a field of view common to both cameras. The monoscopic prediction model thus learned makes it possible to accurately predict the depth of a pixel in an image acquired by one of the cameras of the vision system, even when that pixel corresponds to an object in the three-dimensional scene seen only by one of the cameras. The depth associated with this pixel also corresponds to a distance separating the vehicle carrying the camera from an object in the three-dimensional scene associated with that pixel.

[0012] Furthermore, the monoscopic prediction model enabling the prediction of depth associated with a pixel of an image from a single camera is learned under the supervision of annotated data, this annotated data corresponding to data generated by the vision system itself and the annotation being obtained via the stereoscopic prediction model associated with the stereoscopic vision system comprising several cameras.

[0013] According to a variant of the method, learning is achieved by minimizing the following function representing a recalibration error: ^s-^m ~ 11 ( P ) ( P ) 11 j With : - a loss of alignment of the first depth relative to the second depth; - p a pixel of the first cropped image; - Dtm the second predicted depth for pixel p; and - Dts the first predicted depth for pixel p.

[0014] According to a further variant of the method, the first image includes a distortion symmetrical with respect to its vertical median axis, the part of the first image then being half of the first image.

[0015] According to another embodiment of the method, the stereoscopic prediction model learning step further comprises the following steps: - prediction of depths associated with the pixels of the first cropped image by the stereoscopic prediction model from the first and second cropped images; - generation of a third image from the first cropped image and the first depths; - Determining a reconstruction error by comparing the second cropped image and the third image; and - learning the stereoscopic prediction model by minimizing the reconstruction error.

[0016] According to yet another variant of the method, the cropping of the first and second images comprises cropping the first and second images according to coordinates of at least one image pixel determined by a transformation function applied to a pixel of the first image, called the first pixel, a determined depth being associated with the first pixel and / or by an inverse transformation function applied to a pixel of the second image, called the second pixel, the determined depth being associated with the second pixel, the transformation function determining coordinates in the second image of a pixel corresponding to the first pixel according to the coordinates of the first pixel in the first image and the depth associated with the first pixel,and the inverse transformation function determining the coordinates in the first image of a pixel corresponding to the second pixel as a function of the coordinates of the second pixel in the second image and the depth associated with the second pixel.

[0017] According to yet another variant of the process, the transformation function and the inverse transformation function are defined by mathematical equations.

[0018] According to a second aspect, the present invention relates to a device for learning a depth prediction model associated with a vision system embedded in a vehicle, the device comprising a memory associated with at least one processor configured for the implementation of the steps of the process according to the first aspect of the present invention.

[0019] According to a third aspect, the present invention relates to a vehicle, for example of the automobile type, comprising a device as described above according to the second aspect of the present invention.

[0020] According to a fourth aspect, the present invention relates to a computer program which includes instructions adapted for carrying out the steps of the process according to the first aspect of the present invention, in particular when the computer program is executed by at least one processor.

[0021] Such a computer program may use any programming language and be in the form of source code, object code, or an intermediate form between source code and object code, such as in a partially compiled form, or in any other desirable form.

[0022] According to a fifth aspect, the present invention relates to a computer-readable recording medium on which is recorded a computer program comprising instructions for carrying out the steps of the process according to the first aspect of the present invention.

[0023] On the one hand, the recording medium can be any entity or device capable of storing the program. For example, the medium can include a storage means, such as a ROM, a CD-ROM or a microelectronic circuit-type ROM, or a magnetic recording means or a hard disk drive.

[0024] On the other hand, this recording medium can also be a transmissible medium such as an electrical or optical signal, such a signal being able to be transmitted via an electrical or optical cable, by conventional or radio frequency, by self-directing laser beam, or by other means. The computer program according to the present invention can, in particular, be downloaded from an Internet-type network.

[0025] Alternatively, the recording medium may be an integrated circuit in which the computer program is incorporated, the integrated circuit being adapted to execute or to be used in the execution of the process in question. Brief description of the figures

[0026] Other features and advantages of the present invention will become apparent from the description of the specific and non-limiting embodiments of the present invention below, with reference to the attached Figures 1 to 6, in which:

[0027] [Fig-1] schematically illustrates a vision system equipping a vehicle, according to a a particular and non-limiting example of a realization of the present invention;

[0028] [Fig.2] illustrates a flowchart of the different steps of a process for determining the depth of a pixel of an image by a monoscopic prediction model associated with a vision system embedded in the vehicle of [Fig.1], according to a particular and non-limiting embodiment of the present invention;

[0029] [Fig.3] illustrates a flowchart of the different stages of a learning process for the monoscopic prediction model used in the process of [Fig.2], according to a particular and non-limiting example of the present invention;

[0030] [Fig.4] schematically illustrates images received and / or generated during the learning process of [Fig.3], according to a particular and non-limiting example of the present invention;

[0031] [Fig. 5] schematically illustrates a device configured to determine the depth of a pixel of an image acquired by a vision system embedded in the vehicle of [Fig. 1], according to a particular and non-limiting embodiment of the present invention; and

[0032] [Fig.6] illustrates a flowchart of the different stages of a learning process for the stereoscopic prediction model used in the learning process of [Fig.3], according to a particular and non-limiting embodiment of the present invention. Description of examples of achievements

[0033] A method and device for determining the depth of a pixel of an image by a neural network associated with a vision system embedded in a vehicle will now be described in what follows with joint reference to Figures 1 to 6. The same elements are identified with the same reference signs throughout the description that follows.

[0034] The terms "first," "second" (or "firsts," "seconds"), etc., are used in this document by arbitrary convention to allow for the identification and distinction of different elements (such as operations, means, etc.) implemented in the embodiments described below. Such elements may be distinct or correspond to a single element, depending on the embodiment.

[0035] According to a particular and non-limiting example of an embodiment of the present invention, a method for determining the depth of a pixel of an image by a depth prediction model associated with a camera, called a monoscopic prediction model, implemented by a convolutional neural network associated with a vision system comprising several cameras is implemented by at least one processor.

[0036] Indeed, the monoscopic prediction model is learned in a learning phase comprising the reception of a first image and a second image acquired by two cameras of the vision system and the cropping of the first and second images to generate a first and second cropped images comprising pixels corresponding to objects of the three-dimensional scene seen by the two cameras.

[0037] Initial depths associated with the pixels of the first cropped image are then predicted by a depth prediction model associated with the two cameras, called the stereoscopic prediction model, from the two cropped images and after learning the stereoscopic prediction model.

[0038] The monoscopic prediction model is then learned on half of the first image by minimizing a result of a comparison of the first depths and second depths predicted by the monoscopic prediction model from the first cropped image and associated with the pixels of the first cropped image.

[0039] Fig. 1 schematically illustrates a vision system equipping a vehicle, according to a particular and non-limiting embodiment of the present invention.

[0040] Such an environment 1 corresponds, for example, to a road environment consisting of a network of roads accessible to the vehicle 10.

[0041] In this example, vehicle 10 corresponds to a vehicle with an internal combustion engine, an electric motor(s), or a hybrid vehicle with an internal combustion engine and one or more electric motors. Vehicle 10 thus corresponds, for example, to a land vehicle such as a car, a truck, a bus, or a motorcycle. Finally, vehicle 10 corresponds to an autonomous or non-autonomous vehicle, that is to say, a vehicle operating according to a predetermined level of autonomy or under the total supervision of the driver.

[0042] The vehicle 10 advantageously comprises at least two on-board cameras, a first camera 11 and a second camera 12, configured to acquire images of a three-dimensional scene unfolding in the environment of the vehicle 10 from distinct observation positions. The first camera 11 and the second camera 12 form a stereoscopic vision system when used together, as illustrated in [Fig. 1]. The first camera 11 forms a monoscopic vision system when used alone, and similarly, the second camera 12 forms another monoscopic vision system when used alone. The present invention, however, extends to any vision system comprising at least two cameras, for example, 2, 3, or 5 cameras.

[0043] The intrinsic parameters of the first camera 11 characterize the transformation which associates, for an image point, hereafter called "point", its three-dimensional coordinates in the reference frame of the first camera 11 with the pixel coordinates in an image acquired by the first camera 11. These parameters do not change if the first camera 11 is moved. The intrinsic parameters of the first camera 11 include in particular a first focal length fl associated with the first camera 11.

[0044] The intrinsic parameters of the second camera 12 characterize, for their part, the transformation which associates, for an image point, its three-dimensional coordinates in the reference frame of the second camera 12 with the pixel coordinates in an image acquired by the second camera 12. These parameters do not change if the second camera 12 is moved. The intrinsic parameters of the second camera 12 include in particular a second focal length f2 associated with the second camera 12.

[0045] Distortions, which are due to imperfections in the optical system such as defects in the shape and positioning of the camera lenses, will deflect the light beams and thus induce a positioning error for the projected point relative to an ideal model. It is then possible to complete the camera model by introducing the three distortions that generate the most significant effects, namely radial, decentering, and prismatic distortions, induced by defects in curvature, lens parallelism, and coaxiality of the optical axes.

[0046] According to a particular embodiment, the first camera 11 and / or the second camera 12 is of the "wide-angle" type, a wide-angle camera being, for example, equipped with a lens designed to acquire a representative image of a three-dimensional scene seen over a wider field of view than a standard camera, also sometimes called a panoramic lens. In other words, a wide-angle lens makes it possible to capture a larger portion of the three-dimensional scene unfolding in front of or around the camera, which is particularly useful in situations where it is necessary to include more objects in the frame of the image acquired by that camera. The angle α of the field of view of the first camera 11 is, for example, equal to 120°, 145°, 180°, or 360°, whereas a standard camera offers, for example, a field of view open at an angle of 45° or less.Such a first camera 11 corresponds, for example, to a camera equipped with mirrors or a "fisheye" camera. Wide-angle lenses have a shorter focal length compared to standard lenses, making them suitable for capturing images of landscapes, architecture, road intersections, or any other subject requiring a wide perspective. Wide-angle cameras are, for example, used to capture immersive and dynamic images with an extended depth of field.

[0047] These two cameras 11, 12 are arranged so that each acquires an image of a scene from a different viewpoint. The first viewpoint is, for example, located on or in the left-hand rearview mirror of the vehicle 10 or at the top of the windshield of the vehicle 10. The second viewpoint is, for example, located on or in the right-hand rearview mirror of the vehicle 10 or at the top of the windshield of the vehicle 10. If both cameras are located at the top of the windshield of the vehicle, they are then placed at a certain distance. In this example, the first camera 11 is located at at the top of the windshield of vehicle 10, the second camera 12 is located in the right-hand rearview mirror of vehicle 10.

[0048] A first marker is associated with the first camera 11: - the direction of the x-axis is defined as horizontal and normal to the optical axis of the first camera 11. The distance B separating the optical center of the first camera 11 from the projection of the optical center of the second camera 12 onto the horizontal plane passing through the optical center of the first camera 11 is called the reference basis (in English "baseline"); - the direction of the y-axis is defined as vertical and normal to the optical axis of the first camera 11; - The direction of the z-axis is defined as orthogonal to the directions of the x and y axes. The three axes x, y and z thus form an orthonormal coordinate system.

[0049] The extrinsic parameters related to the position of cameras 11, 12 are the following parameters: - three translations in the x, y, and z directions: Tx, Ty, and Tz, constituting the translation vector T; and - three rotations in the x, y and z directions: 0x, 0y and 0z.

[0050] An extrinsic matrix of the vision system then includes the extrinsic parameters previously defined.

[0051] Note that the first and second cameras 11,12 are arranged so that the optical axis Cl of the first camera 11 is in a first plane parallel to a second plane comprising the optical axis C2, these two planes being separated by a distance H along the vertical direction z. Thus, the angle 0 of rotation is defined between the two optical axes and includes only the 0z component.

[0052] The extrinsic parameters are determined, for example, during a calibration phase of the stereoscopic vision system comprising the first camera 11 and the second camera 12.

[0053] A key constraint of stereoscopic vision systems used in automobiles is, for example, the large distance between the two cameras. Indeed, to cover a measurement range of 200 meters, the reference base must be 60 cm for cameras commonly used in this field.

[0054] The two cameras 11, 12 acquire images of a scene located in front of the vehicle 10, the first camera 11 alone covering a first acquisition field 13, the second camera 12 alone covering a second acquisition field 14, and both cameras 11, 12 together covering a third acquisition field 15. The first and third acquisition fields 13, 15 thus allow a monoscopic view of the scene by the first camera 11, the second and third acquisition fields 14, 15 allow a monoscopic view of the scene by the second camera 12, and the The third acquisition field 15 allows a stereoscopic view of the scene by the stereoscopic vision system composed of the two cameras 11, 12.

[0055] An obstacle 18 is placed in the acquisition field of the cameras, for example in the third acquisition field 15. The presence of the obstacle 18 defines an occlusion field for the stereoscopic vision system composed here of the three fields 16, 17 and 19.

[0056] Among these three fields, field 16 is visible from the second camera 12. The part of the scene present in this field 16 is therefore observable using the monoscopic vision system comprising the second camera 12.

[0057] The field 17 is visible from the first camera 11. The part of the scene present in this field 17 is therefore observable using the monoscopic vision system comprising the first camera 11.

[0058] Finally, field 19 is not visible to any of the cameras. The part of the scene present in this field 19 is therefore not observable.

[0059] According to one particular embodiment, the field of view of the second camera 12 covers at least half of the field of view of the first camera 11.

[0060] It is evident that it is possible to use such a stereoscopic vision system to take images of scenes located on the sides or behind the vehicle 10 by equipping it with cameras placed and oriented differently.

[0061] The images acquired by the cameras 11, 12 at a time instant of acquisition are presented in the form of data representing pixels characterized by: - ​​coordinates in each image; and - data relating to the colors and brightness of the objects of the observed scene in the form, for example, of RGB colorimetric coordinates (from the English "Red Green Blue", in French "Rouge Vert Bleu") or HSL (Tone, Saturation, Luminosity).

[0062] According to a particular embodiment, an image acquired by the first camera 11 and / or by the second camera 12 includes a distortion equal to 0.5%, 0.8% or greater than 1%. The measurement of such distortion corresponds to the determination of a ratio between: - the maximum spacing of a pixel in the image of a straight line in the first three-dimensional scene whose image is a line touching the longest edge of the first image, either at the center of the image edge or at the corners of the image edge, and - the length of this edge.

[0063] In the present invention, in the event of distortion, the distortion in an image acquired by the first camera 11 is considered symmetrical with respect to at least one axis of an image acquired by this first camera 11; for example, the distortion is symmetrical with respect to a vertical median axis of an image acquired by the first camera 11. According to other variants, the distortion is symmetrical with respect to a horizontal median axis of the first camera 11 or even symmetrical with respect to both the horizontal median axis and with respect to the vertical median axis of an image acquired by the first camera 11.

[0064] In the world of photography, distortion is commonly considered to be: • negligible if it is less than 0.3%, • not very sensitive if it is between 0.3% or 0.4%, • sensitive if it is between 0.5% and 0.6%, • very sensitive if it is between 0.7% and 0.9%, and • bothersome if it is greater than or equal to 1% or more.

[0065] Barrel distortion is characterized by a positive percentage, while crescent distortion is characterized by a negative percentage.

[0066] Each pixel of the acquired image represents an object in the three-dimensional scene present in the camera's field of view. Indeed, a pixel of the acquired image is the smallest visible unit and corresponds to a point of light resulting from the emission or reflection of light by a physical object present in the three-dimensional scene. When light strikes this object, photons are emitted or reflected, which are captured by a photosensitive sensor in the camera after passing through its lens. This sensor divides the three-dimensional scene into a grid of pixels. Each pixel records the light intensity at a specific location, thus capturing visual details. The combination of millions of pixels creates an image that faithfully represents the physical object observed by the camera. An image point described above is therefore a point on the surface of an object in the three-dimensional scene.

[0067] The images acquired by cameras 11, 12 represent views of the same scene taken from different viewpoints, the camera positions being distinct. This scene includes, for example: - buildings; - road infrastructure; - other stationary users, for example a parked vehicle; and / or - other mobile users, for example another vehicle, a cyclist or a moving pedestrian.

[0068] According to a particular embodiment, a field of view of the first camera 11 covers at least half of a field of view of the second camera 12, and a field of view of the second camera 12 covers at least half of a field of view of the first camera 11. In other words, more than half of the pixels of an image acquired by the first camera 11 correspond to an object of The three-dimensional scene seen by the second camera 12, pixels of an image acquired by the second camera 12 also corresponding to this object of the three-dimensional scene. Similarly, more than half of the pixels of an image acquired by the second camera 12 correspond to an object of the three-dimensional scene seen by the first camera 11, pixels of an image acquired by the first camera 11 also corresponding to this object of the three-dimensional scene.

[0069] The images acquired by the first camera 11 and by the second camera 12 are sent to a computer of a device equipping the vehicle 10 or stored in a memory of a device accessible to a computer of a device equipping the vehicle 10.

[0070] A method for determining depth by a vision system on board the vehicle 10 is advantageously implemented by the vehicle 10, i.e. by a processor, a computer or a combination of computers of the on-board system of the vehicle 10, for example by the computer or computers in charge of the vision system of the vehicle 10.

[0071] Figure 2 illustrates a flowchart of the different steps of a method 2 for determining the depth of a pixel in an image by a monoscopic prediction model implemented by a convolutional neural network associated with a vision system embedded in a vehicle, for example in vehicle 10 of Figure 1, according to a particular and non-limiting embodiment of the present invention. Method 2 is implemented, for example, by a device of the vision system embedded in vehicle 10 or by device 5 of Figure 5.

[0072] In a step 21, data representative of an image acquired by the first camera 11 are received.

[0073] In a step 22, depths associated with a set of pixels of the received image are predicted by the monoscopic prediction model from the received image.

[0074] Each determined depth then corresponds to a distance separating the vehicle 10 or a part of the vehicle 10 from an object in the three-dimensional scene to which a pixel is associated, the determination of a depth of a pixel then corresponding to a measurement of a distance separating an object from the vehicle carrying the vision system.

[0075] If the ADAS uses these depths or distances as input data to determine the distance between a part of the vehicle 10, for example the front bumper, and another road user, the ADAS is then able to determine this distance precisely. For example, if the ADAS's function is to activate the braking system of the vehicle 10 in the event of a risk of collision with another road user, and the distance separating the vehicle 10 from that same road user is road narrows sharply, so ADAS is able to detect this sudden approach and act on the vehicle's braking system 10 to avoid a possible accident.

[0076] Figure 3 illustrates a flowchart of the different stages of a method for learning a depth prediction model associated with a camera, for example the monoscopic prediction model used in a method for determining the depth of a pixel of an image, for example in method 2 of Figure 2, according to a particular and non-limiting embodiment of the present invention.

[0077] The learning process 3 is for example implemented by the device on board the vehicle 10 implementing the method of determining a depth by vision system on board a vehicle or by the device 5 of the [Fig.5].

[0078] In a step 31, representative data of a first image 41 and a second image 42 are received, the first image 41 being acquired by the first camera 11 at an acquisition time instant and the second image 42 being acquired by the second camera 12 at the same acquisition time instant.

[0079] According to a particular embodiment, the first image and second image are of the same definition, that is to say they have the same number of pixels, have the same number of pixels according to their height and the same number of pixels according to their width.

[0080] According to another particular embodiment, the first and second images are not of the same resolution. An additional step then consists of resizing or cropping them to obtain a first and second image of the same resolution.

[0081] As shown with regard to [Fig.1], according to one particular embodiment, the first camera 11 acquires images without distortion, the first camera 11 comprising a pinhole lens, or according to another example, the first camera 11 acquires images having symmetrical distortion with respect to at least one axis of symmetry in the image acquired by the first camera 11, here the first image 4L. Thus, the first image 4L comprises at least two parts arranged on either side of at least one axis of symmetry.

[0082] Fig. 4 presents an example of a particular embodiment in which the axis of symmetry As divides the first image 41 into two equal parts, a left part 41G and a right part 41D.

[0083] In a step 32, the first and second images are cropped to generate a first cropped image 41' and a second cropped image 42'. Thus, the first cropped image includes pixels from the first image corresponding to objects of the three-dimensional scene seen by both the first camera 11 and the second camera 12, in other words, pixels corresponding to objects present in the field of view common to both cameras, the acquisition field 15 shown in [Fig. 1]. Similarly, the second cropped image includes pixels from the second image corresponding to objects in the three-dimensional scene seen by both the first and second cameras.

[0084] It should be noted that the first and second cropped images necessarily have the same definition.

[0085] It should also be noted that, according to the example illustrated in [Fig.4], the first cropped image 41' covers a large part of the right side 41D.

[0086] The detection and reframing of the first and second images 41, 42 is carried out, for example, according to any method known to a person skilled in the art.

[0087] According to a particular embodiment, step 32 consists of cropping the first and second images according to the coordinates of at least one image pixel determined by a transformation function applied to a pixel of the first image 41, called the first pixel, a determined depth being associated with the first pixel, and / or by an inverse transformation function applied to a pixel of the second image 42, called the second pixel, the determined depth being associated with the second pixel. The transformation function thus determines the coordinates in the second image of a pixel corresponding to the first pixel as a function of the coordinates of the first pixel in the first image and the depth associated with the first pixel, here the determined depth.Similarly, the inverse transformation function determines the coordinates in the first image 41 of a pixel corresponding to the second pixel based on the coordinates of the second pixel in the second image 42 and the depth associated with the second pixel, here also the determined depth. The determined depth corresponds, for example, to an average depth within a depth prediction range. For example, if the depth of objects in observed three-dimensional scenes is between 2 and 100 m, then the determined depth is 51 m. According to other examples, the determined depth corresponds to a median value of predicted depths for different observed three-dimensional scenes or to an average value of predicted depths for the reference pixel in subsequent observations.

[0088] Figure 4 illustrates an example of determining the cropping limits of the first and second images, in particular through the definition of the abscissas and ordinates of these limits.

[0089] According to the example illustrated in [Fig. 4], an abscissa in the first image 41 of a first target pixel 410a is determined by applying the inverse function of the transformation function, hereafter referred to as the transformation function inversely, to a first reference pixel 420a belonging to a lateral edge 421v of the second image 42 as a function of the determined depth. Similarly, an ordinate in the second image 42 of a second target pixel 420b is determined by applying the transformation function to a second reference pixel 410b belonging to a horizontal edge 41Ih of the first image 41 as a function of the determined depth.

[0090] The transformation function determines, for example, for a pixel of the first image 41 to which a depth is associated: • spatial coordinates of an intermediate point in the three-dimensional scene by a first reprojection model based on the coordinates of the pixel in the first image 41 and the depth associated with the pixel, and • the coordinates of a pixel in the second image 42 by a first projection model as a function of the spatial coordinates of the intermediate point.

[0091] The intermediate transformation function determines, for example, for a pixel of the second image 42 to which a depth is associated: • spatial coordinates of an intermediate point in the three-dimensional scene by a second reprojection model based on the coordinates of the pixel in the second image 42 and the depth associated with the pixel, and • coordinates of a pixel in the first image 41 by a second projection model as a function of the spatial coordinates of the intermediate point.

[0092] The transformation function and the inverse transformation function are, for example, defined by mathematical equations such as:

[0093] [Math.l] p^^K^P^P^, D( P^)]) And [Math.l]

[0094] With: • P\ the coordinates of a pixel in the first image 41, • P2 the coordinates of a pixel in the second image 42, •77 a function to convert from homogeneous coordinates to pixel coordinates by removing one dimension from a vector, • K is a direction prediction model associated with camera 11, • K' a direction prediction model associated with the second 12, • T and p1 of matrices including extrinsic parameters, • 0 a projection function in the three-dimensional scene of a pixel Pi as a function of its coordinates in the first image 41, and of the depth ) which associated with it is, here, the determined depth, and • 0' a projection function in the three-dimensional scene of a pixel ^2 in

[0095]

[0096]

[0097] function of its coordinates in the second image 42, and of the depth j Who associated with it is, here, the determined depth. A calibration method is known to those skilled in the art, and is presented for example in the document "Single View Point Omnidirectional Camera Calibration from Planar Grids", written by Christopher Mei and Patrick Rives published in April 2007. Thus, the intrinsic parameters of the first and second cameras 11, 12 are determined and can be used to determine the coordinates of a pixel of an image acquired by the first camera 11 and / or by the second camera 12 from the relative position of a point in the three-dimensional scene with respect to the observation position of the first camera 11 and / or the second camera 12, the pixel being associated with this point. According to a particular embodiment, an analytical correspondence between a pixel of the first image 41 and a pixel of the second image 42 is determined by the following function: [Math.2] With : —'—f— +Zjys 1 n 0+B J t J Z^x^^ > \ r +^jyCOSul / Z^rOsinS „1 ] ( ---------+Zn!cos01 + Cy • xr an abscissa of the pixel of the second image 42, • the ordinate of the pixel in the second image 42, • xi an abscissa of the pixel of the first image 41, • J / an ordinate of the pixel of the first image 41, • d an angle between an optical axis of the first camera 11 and an optical axis of the second camera 12, • H is a height separating the optical axis of the first camera 11 and the optical axis of the second camera 12, • ft a focal length associated with the first camera 11, • fr a focal length associated with the second camera 12, and * Z{w a depth associated with the pixel of the first image 41.

[0098] According to this particular embodiment, an abscissa in the first image 41 of a first target pixel 410a corresponding to a first reference pixel 420a belonging to a lateral edge 421v of the second image 42 is determined analytically, the first target pixel 410a being included in a target set of pixels associated with the first reference pixel 420a. Similarly, an ordinate in the second image 42 of a second target pixel 420b corresponding to a second reference pixel 410b belonging to a horizontal edge 41Ih of the first image 41 is also determined analytically, the second target pixel 420b being included in a target set of pixels associated with the second reference pixel 410b.These analytical determinations are a function of a determined depth and geometric parameters, the geometric parameters including intrinsic parameters of the first and second cameras 11, 12 and extrinsic parameters of the vision system, these extrinsic parameters being representative of the relative positions of the first and second cameras 11, 12.

[0099] The particular embodiment illustrated in [Fig. 4] corresponds to a specific relative position of the first and second cameras 11, 12, the first camera 11 being positioned to the left and below the second camera 12. The lateral edge 421v of the second image 42 then corresponds to the left edge of the second image 42 and the horizontal edge 41Ih of the first image 41 then corresponds to the upper edge, or top edge, of the first image 4L. Coordinates of the first and second target pixels 410a, 420b are, for example, determined analytically by the following functions when the frames associated with the images are defined so as to place the origin of each frame in the upper left of each image:

[0100] [Math.3] in cos(9 And [Math.3] j +C' With : [Math.3] •^410« the x-coordinate of the first target pixel 410a, [Math.3] the ordinate of the second target pixel 420b, [Math.3] -»411 the abscissa of the horizontal edge 41 Ih of the first image 41, [Math.3] ri an abscissa of an optical center of the first image 41, [Math.3] an ordinate of an optical center of the first image 41, [Math.3] ci an abscissa of an optical center of the second image 42, [Math.3] ey an ordinate of an optical center of the second image 42, [Math.3] *1 the x-coordinate of a pixel in the first image 41, [Math.3] H the height separating the optical axis of the first camera 11 and the optical axis of the second camera 12, • [Math.3] HAS a focal length associated with the first camera 11, [Math.3] fr a focal length associated with the second camera 12, and [Math.3] y / the determined depth.

[0101] According to another particular embodiment, when the first camera 11 is to the left above the second camera 12, the lateral edge 421v of the second image 42 corresponds to the left edge of the second image, and the horizontal edge 41Ih of the first image 41 corresponds to the bottom edge of the first image. The coordinates of the first and second target pixels 410a, 420b are, for example, determined analytically by the following functions when the frames associated with the images are defined so as to place the origin of each frame in the upper left corner of each image:

[0102] [Math.4] frsin&+frl^-Ztwcxcos0 X410a. ~ i iei . / , Â + Cx ZjJ—s mcos#] And [Math.4] dVM n _ _____i__2______l__ > ^r I i \ 5 U---7---+Zwcos# \ l / With : [Math.4] the x-coordinate of the first target pixel 410a, • [Math.4] v Am the abscissa of the horizontal edge 41 Ih of the first image 41, [Math.4] d an abscissa of an optical center of the first image 41, [Math.4] cy an ordinate of an optical center of the first image 41, [Math.4] an abscissa of an optical center of the second image 42, [Math.4] an ordinate of an optical center of the second image 42, [Math.4] the x-coordinate of a pixel in the first image 41, [Math.4] H the height separating the optical axis of the first camera 11 and the optical axis of the second camera 12, • [Math.4] HAS a focal length associated with the first camera 11, [Math.4] fr a focal length associated with the second camera 12, and [Math.4] the determined depth.

[0103] The first resized image 41' is generated by cropping the first image 41, and the second resized image 42' is generated by cropping the second image 42. The cropping is a function of the x-coordinate of the first target pixel 410a and the y-coordinate of the second target pixel 420b. The first resized image 41' and the second resized image 42' have the same resolution. The first and The second resized images, 41' and 42', have a width determined from the x-coordinate x4i0a of the first target pixel 410a. Thus, the width L4n of the first resized image 41' is equal to the width L42i of the second resized image 42', and is equal to the difference between the width L4i of the first image 41 (the latter being equal to the width L42 of the second image 42) and the x-coordinate x4i0a of the first target pixel 410a. We therefore obtain: L41 = L42 and L411 = L421 = L4rx4i0a.

[0104] Similarly, the first and second resized images 41', 42' have a height determined from the y420b coordinate of the second target pixel 420b. Thus, the height H4n of the first resized image 41' is equal to the height H42 of the second resized image 42' and is equal to the difference between the height H4i of the first image 41, the latter being equal to the height H42 of the second image 42, and the y420b coordinate of the second target pixel 420b. We thus obtain: H41 = H42 and H421 = H411 = H42-y420b.

[0105] The first resized image 41' is generated from the first image 41, with pixels of the first resized image 41' corresponding to pixels of the first image 41, of which: • an abscissa is located between the abscissa x4i0a of the first target pixel 410a and an abscissa of the right edge 41 Iv of the first image 41, and • an ordinate is between the ordinate y4B equal to the height H4n and an ordinate of the upper edge 41 Ih of the first image 41.

[0106] The pixels of the first image 41 retained to generate the first resized image 41' then correspond to objects of the three-dimensional scene observable by both the first camera 11 and the second camera 12, that is to say, objects of the three-dimensional scene located in the common field of view of the first and second cameras 11, 12, the third acquisition field 15. Indeed, some areas of the first image 41 include pixels corresponding to objects located outside the field of view of the second camera 12 and therefore do not have a corresponding pixel in the second image 42. Thus, the first resized image 41' includes only pixels each having a corresponding pixel in the second image 42 in the absence of occlusion.

[0107] Similarly, the second resized image 42' is generated from the second image 42, with pixels in the second resized image 42' corresponding to pixels in the second image 42, of which: • an abscissa is located between an abscissa x42[ equal to the width L42[ and the abscissa of the left edge 421v of the second image 42, and • an ordinate is between the ordinate y420b of the second target pixel 420b and the ordinate of the lower edge 42Ih of the second image 42.

[0108] The pixels of the second image 42 retained to generate the second resized image 42' then correspond to objects of the three-dimensional scene observable by both the first camera 11 and the second camera 12, that is to say, objects of the three-dimensional scene located in the common field of view of the first and second cameras 11, 12, the third acquisition field 15. Indeed, some areas of the second image 42 include pixels corresponding to objects located outside the field of view of the first camera 11 and therefore do not have a corresponding pixel in the first image 41. Thus, the second resized image 42' includes only pixels each having a corresponding pixel in the first image 41 in the absence of occlusion.

[0109] The use of a predetermined depth and cropping the first and second images 41, 42 horizontally and vertically is not intended to remove every pixel from one image that does not have a corresponding pixel in the other image, but rather to retain a large proportion of pixels corresponding to objects in the three-dimensional scene observed by the two cameras 11, 12. For example, the pixels in one image that have a corresponding pixel in the other image represent more than 90 or 95% of the pixels in a resized image. Any error induced by the presence of pixels in one image without a corresponding pixel in the other image is then rendered negligible in the subsequent steps.

[0110] In a step 33, a depth prediction model associated with the first and second cameras 11, 12, called a stereoscopic prediction model, is learned from the first and second cropped images 41', 42'.

[0111] Such a stereoscopic prediction model implemented by a convolutional neural network is known to those skilled in the art and is presented for example in the paper "Unifying Flow, Stereo and Depth Estimation" written by Haofei Xu, Jing Zhang, Jianfei Cai, Hamid Rezatofighi, Fisher Yu, Dacheng Tao and Andreas Geiger, published in July 2023.

[0112] The stereoscopic prediction model is, for example, learned in another learning phase comprising steps 331, 332, 333 and 334 shown below opposite [Fig.6].

[0113] In a step 34, first depths associated with the pixels of the first cropped image 41' are predicted from the first and second cropped images 41', 42' by the stereoscopic prediction model associated with a vision system comprising several cameras and using images acquired simultaneously by these cameras.

[0114] Thus, the depths are predicted for each pixel of a resized image 41', 42', the definition of a depth map is then equal to the definition of each resized image and avoids any approximation or interpolation to determine depths associated with each pixel of a resized image 41', 42', thus saving computation time.

[0115] The depth associated with each pixel of the first resized image 41' or with each pixel of the second resized image 42' is determined by comparing the positions of the pixels in the first and second resized images 41', 42'.

[0116] In a step 35, the monoscopic prediction model is learned on one half of the first image, the right part 41D according to the example illustrated in [Fig.4], by minimizing a result of a comparison of the first depths to second depths associated with the pixels of the first cropped image 41'.

[0117] The second depths associated with the pixels of the first cropped image 41' are predicted from the first cropped image 41' by the monoscopic prediction model. Such a monoscopic prediction model is known to those skilled in the art and is, for example, presented in the paper "Digging Into Self-Supervised Monocular Depth Estimation" by Clément Godard, Oisin Mac Aodha, Michael Firman and Gabriel Brostow published in August 2019. This monoscopic prediction model makes it possible to predict the depth associated with a pixel for a monoscopic vision system, i.e., one comprising only one camera, here the first camera 11.

[0118] According to a particular embodiment, the result of the comparison is defined by the following function: [Math.5] / \ 11 ~ ^tsP^ ^2 /

[0119] With: - a loss of alignment of the first depth relative to the second depth; - p a pixel of the first cropped image; - Dtm the second predicted depth for the pixelp; and - Dts the first predicted depth for pixel p.

[0120] According to another particular embodiment, the result of the comparison is defined by the following function: [Math.6] ^s-^m ~ ( P ) ^ts ( P l)

[0121] With: - L^m a loss of recalibration of the first depth relative to the second depth; - p a pixel of the first cropped image; - Dtm the second predicted depth for pixelp; and - Dts the first predicted depth for pixelp.

[0122] The training of the monoscopic prediction model consists of adjusting the input parameters of the convolutional neural network in order to minimize the previously calculated registration error. It should be noted that the training step comprises several iterations, for which second depths are predicted at each iteration, with the second depths becoming increasingly accurate.

[0123] The monocular prediction model is then able to predict a depth for any pixel of an image acquired by the first camera 11, in particular the first image 41, via the use of symmetry. The axis of symmetry As in the acquired image is the axis of symmetry of the distortion in any image acquired by the first camera 11. According to the example illustrated in [Fig. 4], training is performed on the right part 41D of the first image 4L. During prediction, the left part 41G of the first image 41 is to be flipped by symmetry with respect to the axis of symmetry As and is an input to the monocular prediction model.After the prediction, the resulting depth map is to be reversed by symmetry with respect to the symmetry axis As to be consistent with the original image, that is to say that the depths determined for the symmetry pixels of the left part 41G are associated with the original pixels of the left part 41G of the first image 4L The combination of the depth map defined for the right part 41D and the depth map determined for the left part 41G thus forms the depth map for the entirety of the first image 4L.

[0124] The resizing steps of the received images make it possible in particular to optimize computation times, the learning process is thus faster to implement than other known learning processes, reducing in particular the time to search for corresponding pixels and not requiring depth interpolation steps, the depths being predicted for each pixel of each of the resized images.

[0125] Furthermore, the training of the monoscopic prediction model is performed on a portion of an image acquired by the first camera and then extended, once trained, to the entire image acquired by the first camera. The partial learning process is thus faster to implement and allows supervision by the stereoscopic prediction model even when the pixels of an image acquired by the The first camera does not all correspond to objects positioned in the field of vision common to both cameras.

[0126] According to other variants, the distortion in an image acquired by the first camera is symmetrical with respect to several axes, for example with respect to the vertical median axis and the horizontal median axis of such an image. Partial learning is then performed on a quarter of the image and then extended via two symmetries to the entire image, thus further reducing the training time of the monoscopic prediction model.

[0127] Thus, the monoscopic prediction model is learned from the depths predicted by the stereoscopic prediction model. In other words, the monoscopic vision system is supervised by the stereoscopic vision system in the parts of an image corresponding to pixels located in the field of view of several cameras.

[0128] The annotated data, here the second depths predicted by the depth prediction model associated with the stereoscopic vision system, are data obtained directly from the cameras on board the vehicle. They are therefore particularly relevant because they are representative of the different road environments in which the vehicle travels, the objects in the three-dimensional scenes traversed being representative of real objects present in these road environments.

[0129] To this must be added that the data being acquired by the vision system on board the vehicle, these are numerous and easy to obtain.

[0130] Figure 6 illustrates a flowchart of the different stages of a method for learning a depth prediction model associated with several cameras, for example the stereoscopic prediction model used in the learning method of Figure 3, according to a particular and non-limiting embodiment of the present invention.

[0131] The learning process 3' of the stereoscopic prediction model is notably carried out from two cropped images, for example from the first and second cropped images 41', 42' previously generated and constitutes another learning phase comprising the following steps.

[0132] In a step 331, depths associated with the pixels of the first cropped image 41' are predicted by the stereoscopic prediction model from the first and second cropped images 41', 42'.

[0133] During a first iteration, these depths are unreliable, but become more and more accurate after further iterations of the steps of the learning process 3'.

[0134] In a step 332, a third image 43 is generated from the first resized image 41' by applying the transformation function to the pixels of the first image according to the depths associated with these pixels.

[0135] Generating an image from an image acquired by a camera of the vision system and resized consists of reprojecting a pixel of the resized image into the three-dimensional scene as a point, and then projecting this point into the image plane of another camera of the vision system, so as to obtain an image corresponding to a view of the three-dimensional scene from the viewpoint of the other camera. The image plane of a camera corresponds to a plane defined in the camera's frame of reference, normal to the camera's optical axis and located at the camera's first focal length. Thus, the third image 43 generated from the first resized image 41' is comparable to the second resized image 42'. Since objects can obscure other objects in the scene, the generated images are not identical to acquired images.Furthermore, depth prediction, like the models used to generate the images, is not error-free. Therefore, comparing a generated image to an acquired and resized image allows for an evaluation of the accuracy of the different models used.

[0136] In a step 333, a reconstruction error is determined by comparing the second cropped image 42' and the third image 43.

[0137] According to a first particular embodiment, the reconstruction error is determined from a first error associated with the pixels of the second cropped image 42'. Each first error is a photometric error as presented in the document "Digging Into Self-Supervised Monocular Depth Estimation" and is determined by the following function: [Math.7] ^i(p) = EJ (l-«) ■ \I(P)-I(p) 1 +«• (l-jSSIMÇKp^ïip) ) ) ]

[0138] With: • the first reconstruction error associated with a pixel defined by its coordinates in the second cropped image 42', • l(p) a pixel value p in the first cropped image 42', * l(p) a pixel value p in the third image 43, • SSIM is a function that takes into account a local structure, and • has a weighting factor depending in particular on the type of road environment in which the vehicle travels 10.

[0139] According to a second particular embodiment, the reconstruction error is determined from a second error associated with the pixels of the third image 43 determined by the following function:

[0140] [Math.8] ^smooth ( ( P ) ' ®) = W(p)

[0141] With: * ^smooth(D(p),W» a second error for a pixel p of the third image 43, • D(p) is a depth associated with a pixel p, • VF is a parameter matrix, • 0 is the order of a smoothing gradient, • an L1 norm of the second-order depth gradients is calculated with W = 1, et° = 2, • x and are the dimensions of the third image, • fi is a hyperparameter dependent on the road environment in which vehicle 10 is traveling, and • It^p^ is a pixelp value in the third image 43.

[0142] This second function is generally used to deal with discontinuity at the boundary of objects.

[0143] The reconstruction error is thus defined, for example, from the first and second errors previously defined and is determined by the following function: [Math.9] L' = Yi^m.(L[(p),L2(p) )

[0144] With: • The reconstruction error, • L^p) the first error for a pixelp of the second cropped image 42', and • L2(p) the second reconstruction error for a pixelp of the third image 43 corresponding to the pixel p of the second cropped image 42'.

[0145] Note that according to variant embodiments, weighting coefficient factors are added in the preceding function in order to favour the first or second error.

[0146] In a step 334, the depth prediction model is learned by minimizing the previously defined reconstruction error.

[0147] Learning the stereoscopic prediction model involves adjusting the input parameters of the convolutional neural network to minimize the previously calculated reconstruction error. It should be noted that learning the stereoscopic prediction model involves several iterations, in which depths are predicted at each iteration, with the depths becoming increasingly accurate. Learning ends when a target accuracy level is reached, for example, when more than 95% of the predicted depths have an error of less than 25%.

[0148] Thus, the stereoscopic prediction model used for predicting the depth of a pixel in an image acquired by the first camera 11 or the second camera 12 is made more reliable through this learning process, ensuring a high level of reliability for the annotated data used in the learning process 3 of the monoscopic prediction model. Furthermore, the data used for this learning are obtained from the stereoscopic vision system itself; they therefore correspond to data that perfectly represent the use of the vision system mounted on the vehicle 10.

[0149] The steps of resizing the received images and determining the target sets of pixels make it possible in particular to optimize computation times, the learning process is thus faster to implement than other known learning processes, reducing in particular the time to search for corresponding pixels and not requiring depth interpolation steps, the depths being predicted for each pixel of each of the resized images.

[0150] Figure 5 schematically illustrates a device 5 configured for learning a depth prediction model associated with a vision system embedded in a vehicle 10 and / or for predicting a depth associated with a pixel of an image acquired by the vision system embedded in a vehicle 10, according to a particular and non-limiting embodiment of the present invention. The device 5 corresponds, for example, to a device embedded in the first vehicle 10, for example, a computer associated with the stereoscopic vision system.

[0151] Device 5 is, for example, configured to carry out the operations described opposite Figures 1 and 4 and / or the steps described opposite Figures 2 and 3. Examples of such a device 5 include, but are not limited to, embedded electronic equipment such as a vehicle's on-board computer, an electronic control unit such as an ECU (Electronic Control Unit), a smartphone, a tablet, or a laptop computer. The elements of device 5, individually or in combination, may be integrated into a single integrated circuit, into several integrated circuits, and / or into discrete components. Device 5 may be implemented in the form of electronic circuits or software (or computer) modules, or a combination of electronic circuits and software modules.

[0152] The device 5 comprises one (or more) processor(s) 50 configured to execute instructions for carrying out the steps of the process and / or for executing instructions from the software embedded in the device 5. The processor 50 may include integrated memory, an input / output interface, and various circuits known to those skilled in the art. The device 5 further comprises at least one memory 51 corresponding, for example, to volatile and / or non-volatile memory and / or includes a memory storage device which may include volatile and / or non-volatile memory, such as EEPROM, ROM, PROM, RAM, DRAM, SRAM, flash, magnetic or optical disk.

[0153] The computer code of the embedded software(s) including the instructions to be loaded and executed by the processor is for example stored on memory 51.

[0154] According to various particular and non-limiting embodiments, the device 5 is coupled in communication with other similar devices or systems (for example other computers) and / or with communication devices, for example a TCU (Telematic Control Unit), for example via a communication bus or through dedicated input / output ports.

[0155] According to a particular and non-limiting embodiment, the device 5 includes a block 52 of interface elements for communicating with external devices. The interface elements of the block 52 include one or more of the following interfaces: - radio frequency RF interface, for example of the Wi-Fi® type (according to IEEE 802.11), for example in the 2.4 or 5 GHz frequency bands, or of the Bluetooth® type (according to IEEE 802.15.1), in the 2.4 GHz frequency band, or of the Sigfox type using UBN (Ultra Narrow Band) radio technology, or LoRa in the 868 MHz frequency band, LTE (Long-Term Evolution), LTE-Advanced; - USB interface (from the English "Universal Serial Bus" or "Universal Serial Bus" in French); HD MI interface (from the English "High Definition Multimedia Interface", or "High Definition Multimedia Interface" in French); - LIN interface (from the English "Local Interconnect Network", or in French "Réseau interconnecté local").

[0156] According to another particular and non-limiting embodiment, the device 5 includes a communication interface 53 which enables communication with other devices (such as other computers in the embedded system) via a communication channel 530. The communication interface 53 corresponds, for example, to a transmitter configured to transmit and receive information and / or data via the communication channel 530. The communication interface 53 corresponds, for example, to a wired CAN (Controller Area Network) or CAN FD (Controller Area Network Flexible Data-Rate) type network. flexible data rate”), FlexRay (standardized by ISO 17458) or Ethernet (standardized by ISO / IEC 802-3).

[0157] According to a particular and non-limiting embodiment, the device 5 can provide output signals to one or more external devices, such as a display screen 540, touch or not, one or more speakers 550 and / or other peripherals 560 via the output interfaces 54, 55, 56 respectively. According to a variant, one or more of the external devices is integrated into the device 5.

[0158] Of course, the present invention is not limited to the embodiments described above but extends to a method for measuring the distance between an object and a vehicle equipped with a vision system, which would include secondary steps without falling outside the scope of the present invention. The same would apply to a device configured for implementing such a method.

[0159] The present invention also relates to a vehicle, for example an automobile or more generally an autonomous land-powered vehicle, comprising the device 5 of [Fig.5].

Claims

1. Demands Method for determining the depth of a pixel of an image acquired by a first camera (11) of a vision system comprising the first and a second camera arranged so as to each acquire an image of a three-dimensional scene from a different viewpoint, said depth being determined by a depth prediction model associated with the first camera (11), called a monoscopic prediction model, implemented by a convolutional neural network, said method being implemented by at least one processor, and being characterized in that the monoscopic prediction model is learned in a learning phase comprising the following steps: - reception (31) of data representative of a first image (41) and a second image (42) acquired respectively by the first camera (11) and the second camera (12) at the same time instant of acquisition; - cropping (32) of the first and second images to generate a first cropped image (41') and a second cropped image (42'), the first cropped image comprising pixels of the first image corresponding to objects of the three-dimensional scene seen by both the first and second camera and the second cropped image comprising pixels of the second image corresponding to said objects of the three-dimensional scene seen by both the first and second camera, the first and second cropped images having the same definition; - training (33) of a depth prediction model associated with the first and second cameras (11, 12), called stereoscopic prediction model, from the first and second cropped images (41', 42'); - prediction (34) of first depths associated with the pixels of the first cropped image from the first and second cropped images by the stereoscopic prediction model; and - learning (35) of the monoscopic prediction model on a part of the first image (41) by minimizing a result of a comparison of the first depths to second depths predicted by the monoscopic prediction model from the first cropped image (41') and associated with the pixels of the first cropped image (41').

2. Method according to claim 1, wherein said learning (35) is obtained by minimizing the following function representing a registration error: ^s-^m — 11 ^tm (P) ~ D ( P^ 11 1 "2 / With: - Lg^m a registration loss of the first depth with respect to the second depth; - p a pixel of the first cropped image; - Dtm the second depth predicted for pixel p; and - Dts the first depth predicted for pixel p.

3. A method according to claim 1 or 2, wherein the first image (41) comprises a symmetrical distortion with respect to a vertical median axis, the portion of the first image (41) then being a half of the first image (41) and said symmetrical extension being made with respect to said vertical median axis of the first image (41).

4. A method according to any one of claims 1 to 3, wherein the learning step (33) of the stereoscopic prediction model further comprises the steps of: - predicting (331) depths associated with the pixels of the first cropped image by the stereoscopic prediction model from the first and second cropped images; - generating (332) a third image (43) from the first cropped image (41') and the first depths; - determining (333) a reconstruction error by comparing the second cropped image (42') and third image (43); and - learning (334) the stereoscopic prediction model by minimizing the reconstruction error.

5. A method according to any one of claims 1 to 4, wherein the cropping (32) of the first and second images comprises cropping the first and second images according to coordinates of at least one image pixel determined by a transformation function applied to a pixel of the first image (41), referred to as the first pixel, a determined depth being associated with the first pixel and / or by an inverse transformation function applied to a pixel of the second image (42), called second pixel, the determined depth being associated with the second pixel, the transformation function determining coordinates in the second image of a pixel corresponding to the first pixel as a function of the coordinates of the first pixel in the first image and the depth associated with the first pixel, and the inverse transformation function determining coordinates in the first image of a pixel corresponding to the second pixel as a function of the coordinates of the second pixel in the second image and the depth associated with the second pixel.

6. A method according to claim 4 or 5, wherein the transformation function and the inverse transformation function are defined by mathematical equations.

7. A computer program comprising instructions for carrying out the method according to any one of the preceding claims, when such instructions are executed by a processor.

8. Device (5) for determining depth by means of a vision system mounted in a vehicle (10), said device (5) comprising a memory (51) associated with at least one processor (50) configured for carrying out the steps of the method according to any one of claims 1 to 6.

9. Vehicle (10) comprising the device (4) according to claim 8.

Images