Checking the compliance of a light beam in a vehicle

Abstract

The invention relates to a method for evaluating the compliance of a light beam of a motor vehicle light module, the method comprising a step of storing (203) a model in the motor vehicle. The model is capable of determining an item of information on the compliance of a light beam on the basis of an image received as input to the model. The method comprising: - checking (221) that a light module of the motor vehicle has been activated; - obtaining (223) a first image on the basis of items of data from an image acquisition device of the vehicle; - applying (225) the model to the first image in order to determine a first item of information on the compliance of a light beam; - evaluating (226) the compliance of a light beam of the at least one light module according to the at least one first item of information on the compliance of a light beam.

Description

Description Title of the invention: Verification of conformity of lighting beam in a vehicle

[0001] The present invention relates to the field of evaluating the conformity of a lighting beam projected by a lighting module for a motor vehicle. It relates in particular to a method and a device for evaluating the lighting beam of a lighting module for a motor vehicle.

[0002] It is particularly advantageous for enabling correction of the aiming of a lighting module for a motor vehicle or correction of the distribution of light intensity in the projected light beam.

[0003] A motor vehicle includes lighting modules at the front of the vehicle, on the right and left, in order to perform several lighting functions, such as the “low beam” function, also called LB for “Low Beam” in English, and the “high beam” function, also called HB for “High Beam” in English.

[0004] . A lighting module may be capable of performing both of the above-mentioned functions from a single light source. Alternatively, several separate modules are provided: a first pair of modules is dedicated to the LB function while another pair of modules is dedicated to the HB function.

[0005] . Aiming means the direction and shape of the light beam projected by a lighting module for a motor vehicle. Such a shape and direction may be subject to regulatory constraints, in particular to allow good visibility of the scene in front of the vehicle by the driver, without dazzling other road users.

[0006] . In addition, regulatory constraints may define the distribution of the projected light intensity within the beam. In particular, it may be required that the ground projections of the left lighting module for a given function have a symmetrical shape, with a homogeneous distribution of the light intensity within each lighting beam.

[0007] The aiming of a lighting module can be observed, or characterized, by its projection on the ground. In particular, the aiming of two lighting modules of the same pair can be evaluated by their respective projections in a reference zone. It is thus made possible for a specialist, during a technical inspection of the vehicle, to check whether the aiming of each lighting module complies with regulatory constraints, or whether the aims of lighting modules of the same pair are indeed symmetrical. In addition, it is possible for such a specialist to check the correct distribution of the light intensity within each beam of a lighting module, and between several beams of a pair of lighting modules.

[0008] . However, such a technical inspection of the vehicle is infrequent, and an undue change in the aiming or distribution of the light intensity of a lighting module may not be detected for several months or even years. In the meantime, such an undue change in aiming may lead to reduced visibility of the scene by the driver and / or dazzling of other road users, which may cause accidents.

[0009] . Thus, the absence of regular assessment of the conformity of the characteristics of a light beam of a lighting module in relation to predetermined characteristics, for example regulatory or technical, causes safety problems for the passengers of a vehicle and for other road users.

[0010] There is therefore a need to assess the conformity of a light beam of a lighting module for a motor vehicle, on a regular basis, without having to resort to a check by a specialist.

[0011] . To this end, a first aspect of the invention relates to a method for evaluating the conformity of a lighting beam of at least one lighting module of a motor vehicle, the method comprising a prior step of storing a model in a memory of the motor vehicle, the model being capable of determining at least one piece of lighting beam conformity information from at least one image received as input to the model, the method further comprising the following common steps: - verification of activation of at least one lighting module of the motor vehicle; - obtaining at least a first image from data from an image acquisition device of the vehicle, said image acquisition device being arranged so as to obtain data representative of a scene facing the vehicle; - applying the model to said at least one first image to determine at least one first lighting beam conformity information; - evaluation of a conformity of a lighting beam of said at least one lighting module as a function of said at least one first conformity information.

[0012] . Thus, the invention makes it possible to evaluate the conformity of a lighting beam of a motor vehicle lighting module in an automated manner, from a model stored in the vehicle, and by applying the model to a first image obtained from data from an image acquisition device. Such an image acquisition device being generally integrated into most recent vehicles, the invention does not require an additional sensor dedicated to the invention, and can therefore be implemented at lower cost.

[0013] . According to embodiments, the method may further comprise a prior step of obtaining the model by machine learning from a training data set.

[0014] . Thus, the model can be derived from machine learning, which allows for a reliable prediction of the first compliance information.

[0015] . Additionally, the model may be obtained by supervised learning, and the training data set may comprise associations between at least one training image and at least one piece of lighting beam conformity information.

[0016] . Thus, by constituting a training database comprising both training images and respectively associated compliance information, an accurate compliance information prediction model can be obtained, stored in the vehicle, and then used in the current phase to evaluate the compliance of lighting beams.

[0017] . According to embodiments, the model may be capable of determining a binary compliance category from among a compliant category and a non-compliant category.

[0018] . Thus, the model can be a simple classifier with two possible outputs, which reduces the costs associated with training the model, as well as the runtime when applying the model.

[0019] . Alternatively, the model may be capable of determining one type of non-conformity from among at least two distinct types of non-conformity.

[0020] . Thus, richer compliance information than binary information can be obtained by differentiating between several types of non-compliance. Such differentiation makes it possible to improve the evaluation provided at the output of the process, and makes it easier for an operator to implement corrective action in the event of non-compliance of a lighting beam.

[0021] . In addition, when applying the model to said at least one first image, a first type of non-conformity can be determined, the evaluation can include an indication of a correction to be made to the lighting module, said correction depending on the first type of non-conformity determined.

[0022] . Thus, the implementation of corrective action by an operator can be facilitated in the event of non-conformity of a lighting beam.

[0023] . According to one embodiment, the image acquisition device may be a camera capable of obtaining color or grayscale images.

[0024] . Thus, it is made possible to take advantage of a camera present in most motor vehicles, used for example for a function of assisting in steering the lateral or longitudinal trajectory of the vehicle, or for a function of assisting in parking the vehicle. Thus, the costs associated with the implementation of the method according to the invention are reduced.

[0025] In addition, the model may be capable of determining at least one piece of lighting beam conformity information from at least one color or grayscale image received as input, and said at least one first image may be in color or grayscale and may come from the image acquisition device.

[0026] . Thus, the camera images can be used directly as input to the model, which speeds up the implementation of the method according to the invention.

[0027] . Alternatively, the model may be a first model capable of determining at least one piece of lighting beam conformity information from at least one luminance image, the method may further comprise a prior step of storing a second model in said motor vehicle, said second model being capable of determining at least one luminance value of a pixel of a luminance image from of at least one value of a pixel of a color or grayscale image. Said at least one first image obtained may be a first luminance image. Obtaining said at least one first luminance image may comprise applying the second model to pixels of at least one first color or grayscale image obtained by the image acquisition device. Said first model may be applied to said at least one first luminance image to determine said at least one piece of lighting beam conformity information.

[0028] . The prediction of conformity information of a lighting beam is improved by taking into account a luminance image as input to the model. Indeed, a luminance image allows a better contrast and thus a better identification of both the geometric characteristics of a lighting beam but also the distribution of the luminous intensity within the beam. What is more, the luminance image is determined from a second model, which avoids the use of a luminance camera which is both expensive and slow in its processing times.

[0029] . According to another variant, the image acquisition device is an infrared camera capable of obtaining infrared images, the model may be capable of determining at least one piece of lighting beam conformity information from at least one infrared image received as input.

[0030] . Taking into account an infrared image as input to the model allows a better prediction of the lighting beam conformity information compared to using a grayscale or color image.

[0031] . According to embodiments, said at least one activated lighting module may comprise a right low beam module and a left low beam module of the motor vehicle.

[0032] . Thus, the conformity of dipped headlights can be assessed, which makes it possible to detect non-conformity which would be likely in particular to dazzle other users of the road on which the vehicle is traveling, and / or to poorly illuminate the scene in front of the vehicle, which deteriorates the visibility of the driver of the vehicle.

[0033] . According to embodiments, the first model may be capable of determining at least one piece of lighting beam conformity information from a series of images received as input, said at least one first image may be a series of first images, and the model may be applied to the first series of images to obtain said at least one piece of first information.

[0034] . Taking into account a series of images as input to the model allows for better accuracy in predicting the lighting beam conformity information.

[0035] A second aspect of the invention relates to a computer program comprising instructions for implementing the method according to the first aspect of the invention, when these instructions are executed by a processor.

[0036] A third aspect of the invention relates to a vehicle device comprising: - a memory storing a first model capable of determining at least one piece of lighting beam conformity information from at least one image received as input to the model, or an interface for accessing said first model; - a processor configured to: verify activation of at least one lighting module of the vehicle; obtain at least one first image from data from an image acquisition device of the vehicle, said image acquisition device being arranged so as to obtain data representative of a scene facing the vehicle; apply the model to said at least one first image to determine at least one first piece of lighting beam conformity information evaluate a conformity of a lighting beam of said at least one lighting module based on said at least one first piece of conformity information

[0037] . Other characteristics and advantages of the invention will appear on examining the detailed description below, and the appended drawings in which:

[0038] . [Fig 1] illustrates a motor vehicle according to embodiments of the invention;

[0039] . [Fig 2] illustrates the steps of a method according to embodiments of the invention;

[0040] . [Fig 3a] illustrates lighting beams of lighting modules of a motor vehicle according to embodiments of the invention, in a case of conformity of the lighting beams;

[0041] . [Fig 3b] illustrates lighting beams of lighting modules of a motor vehicle according to embodiments of the invention, in a case of non-conformity of the lighting beams;

[0042] . [Fig 4] illustrates a device for calibrating a lighting module for a motor vehicle according to embodiments of the invention;

[0043] The description focuses on the characteristics that distinguish the method and device from those known in the state of the art.

[0044] . Figure 1 illustrates a vehicle 100 according to embodiments of the invention.

[0045] The vehicle 100 comprises a device 101 according to embodiments of the invention, the device 101 being capable of evaluating the conformity of a lighting beam of at least one lighting module of the vehicle 100. Note that the conformity is evaluated with respect to a set of predefined standards, which may be regulatory standards applying to all vehicles circulating in a given territory, and / or technical standards issued by the vehicle manufacturer 100.

[0046] The vehicle 100 in fact comprises a pair of lighting devices 103 comprising at least one lighting module capable of performing at least one lighting function of the vehicle 100.

[0047] . In the example of Figure 1, the vehicle 100 is shown in profile so that only one lighting device 103, such as the left lighting device, is visible in the figure. It will be understood, however, that the vehicle 100 further comprises a lighting device located to the right of the vehicle, arranged symmetrically to the lighting device 103, relative to a front-rear axis of the vehicle.

[0048] . In the example of Figure 1, the lighting device 103 comprises a first lighting module 104 dedicated to a first lighting function and a second lighting module 105 dedicated to a second lighting function distinct from the first lighting function. Alternatively, a single lighting module is capable of implementing the first and second lighting functions.

[0049] The first lighting function can be a dipped beam function, called LB, for Low Beam in English, while the second lighting function can be a main beam function, called HB, for High Beam in English.

[0050] . The vehicle 100 further comprises a control module 102 of the lighting device 103, capable of activating / deactivating the first function and / or the second function of the vehicle. When the light sources of the lighting modules 104 and 105 are of the matrix type with a plurality of individually controllable light elements, the control module 102 can further transmit pixelated lighting photometry, from which the lighting device 103 is capable of selectively activating / deactivating light elements of the source of one or other of the lighting modules.

[0051] . The control module 102 may be dedicated to controlling the lighting devices or may also provide control of other functions of the vehicle. The control module 102 may in particular be a central control module of the vehicle 100.

[0052] The vehicle 100 according to the invention may further comprise an image acquisition device 106.

[0053] The image acquisition device 106 is capable of acquiring data representative of the scene of the vehicle located in front of the vehicle 100, that is to say in front of the vehicle 100 in a direction of travel of the vehicle 100. In particular, the image acquisition device 106 may be located at the front of the vehicle, for example at the top of the windshield of the vehicle 100 or at the bumper of the vehicle 100.

[0054] The image acquisition device 106 may be an RGB type camera. RGB type cameras, for “red-green-blue”, in English, are capable of acquiring data which are color images, which are capable of being processed in real time, and are thus capable of measuring light parameters relating to the environment of the vehicle.

[0055] . Thus, the data acquired by the image acquisition device 106 may be color images, that is to say that each image comprises pixels arranged in a matrix, and that each pixel is coded so as to correspond to one color among strictly more than two colors. For example, each pixel may be coded in red-green-blue, or RGB for “Red Green Blue”, that is to say that the set of colors is represented by three coordinates respectively red, green and blue. Each pixel may thus be coded on several bits, for example on a byte coding 256 colors.

[0056] . Alternatively, the data acquired by the image acquisition device 106 may be grayscale images, i.e. each image comprises pixels, arranged in a matrix, each pixel being coded on strictly more than two levels, with at least one grayscale level intermediate between white and black. Each pixel can thus be coded on several bits, for example on a byte coding 256 gray levels.

[0057] . As a further variant, the image acquisition device 106 may be an infrared camera capable of acquiring data which are infrared images, comprising pixels, arranged in a matrix, each pixel being associated with an infrared value.

[0058] The image acquisition device 106 may be capable of acquiring a fixed image regularly, at a given frequency, for example every second. However, preferably in order to allow the use of the data from the image acquisition device 106 for real-time functions of the vehicle, the image acquisition device 106 may be capable of obtaining images constituting video frames at a frequency of several frames per second, or fps, for “frame per second” in English, for example at a frequency greater than 10 fps, in particular equal to 30 fps.

[0059] . The vehicle 100 may comprise a memory 107, capable of storing data, such as the first model and / or the second model described in the following. Alternatively, the memory 107 may be integrated into the control module 102 or into the device 101.

[0060] . The vehicle 100 may comprise a human-machine interface 108, or HMI 108. No restriction is attached to the HMI 103, which comprises any interface element capable of receiving a command from the driver or a passenger of the vehicle, whether this command relates to the piloting of the vehicle or to the control of interior or exterior equipment, and / or to transmit information to the driver and / or to a passenger of the vehicle. For this purpose, the HMI 103 may comprise a screen, such as a touch screen, a set of one or more buttons, a speaker, a microphone, a dashboard capable of displaying one and / or more luminous pictograms of predefined shapes, a steering wheel vibration system, etc.

[0061] . Figure 2 is a diagram illustrating the steps of a method for verifying the conformity of a lighting beam of at least one lighting module, according to embodiments of the invention.

[0062] The method may comprise a preliminary phase 200 and a current phase 220. The preliminary phase comprises steps 201 to 203 described below, and may further comprise steps 211 to 213 in a particular embodiment.

[0063] . In a step 201, a first training database is formed, the first training database comprising first training data. The first training data comprises first training images obtained by cameras installed in respective motor vehicles, in a position close to that of the camera 106 described previously, that is to say: - if the camera 106 is arranged in an upper part of the vehicle windshield, the training images are acquired by cameras arranged in respective upper parts of motor vehicle windshields; - if the camera 106 is arranged in a rocker panel of the vehicle at the front, in particular to perform a vehicle parking assistance function, the training images are acquired by cameras arranged in respective rocker panels of motor vehicles.

[0064] . Training images are representative of a scene facing a given vehicle when at least one lighting module of the given vehicle is turned on.

[0065] . The first training images can be: - grayscale images; - color images; - infrared images; or - luminance images, also called luminance maps.

[0066] There are no restrictions on how the initial training data is obtained. It can be collected from a fleet of vehicles in circulation, for example during technical inspections.

[0067] In one embodiment, the first training database comprises time series of training images.

[0068] . Each first training image or each time series of training images may be associated with an illumination beam conformity information, in the training database. The first training data therefore further comprises the conformity information of each first training image or time series of training images.

[0069] Each piece of conformity information associated with one or more training images is called reference conformity information, since it is not a prediction but true information, unlike the conformity information that is obtained (or predicted) at the output of the model described below.

[0070] . The conformity information is representative of the conformity or non-conformity of at least one of the light beams present in the first associated training image (or in the associated series of training images). The conformity information may indicate the conformity or non-conformity of all the projected lighting beams, for example of the two lighting beams projected by dipped beam lighting modules.

[0071] . Thus, the conformance information is used to label the first training images (or sets of training images) of the first database.

[0072] . There is no restriction on the standard to which the conformity relates: it can be a technical standard of the manufacturer or a regulatory standard.

[0073] . Furthermore, compliance may be relative: - the shape and / or orientation of the lighting beam(s), the conformity of which can be assessed from the projection on the ground; and / or - to the distribution of light intensity in a lighting beam and / or in the two lighting beams.

[0074] . According to one embodiment, the compliance information may be a binary compliance category, that is, the compliance information may take two values: a “compliant” category and a “non-compliant” category.

[0075] . Alternatively, the compliance information may take at least three values: a “compliant” category, a “non-compliant, first type of non-compliance” category, and a “non-compliant, second type of non-compliance” category. More generally, the compliance information may identify N types of non-compliance, where N is an integer greater than or equal to 2. The type of non-compliance may: - indicate a level of non-compliance, among several levels, for example “slight”, “medium” and “serious”; and / or - describe the non-conformity. The type of non-conformity may indicate, for example, that a right or left beam is too high, too low, too far to the right, or too far to the left. In addition, the type may indicate a value of a deviation from a non-conformity situation, the value being a distance or an angular value, for example.

[0076] . In a step 202, a first model is trained by machine learning from the first training data of the training database obtained in step 202. The model thus trained is capable of: - receive at least one first image as input. As indicated previously, depending on the first training images used (in grayscale, color, luminance or infrared), the model is able to receive a first image of the same type as the first training images; - determine, or predict, at least one piece of conformity information from the first image received as input.

[0077] . Such a model can be obtained by supervised learning, that is to say that the reference conformity information associated with the first training images is compared with the predicted conformity information output by the model. The model thus minimizes its errors in determining said at least one conformity information.

[0078] The principle of supervised learning is well known and is not described further in this application.

[0079] The first model thus trained is capable of predicting at least one piece of conformity information based on a first image received as input, or based on a first series of images received as input.

[0080] Step 202 can be implemented by a training module not shown in FIG. 1, which can be external to the vehicle 100, and which is capable of accessing the first training database.

[0081] . In a step 203, the first model is stored in the vehicle 100, so as to be accessible by the device 101. The first model can for example be stored in the memory 107 of the vehicle, or in an internal memory of the device 101.

[0082] . There is no restriction on the type of the first model, which can be a support vector machine (SVM) classifier, or a convolutional neural network, or any other model that can be built by machine learning and capable of receiving an image as input.

[0083] . When the first model is trained to receive grayscale or color images, according to embodiments, the input of the first model may be capable of directly receiving the images from the image acquisition device 106, which is then a camera in grayscale or color. Such a camera is available in most current vehicles, particularly for a lateral / longitudinal steering assistance function of the vehicle (camera at the top of the windshield) or for a parking assistance function (camera in the underbody of the vehicle). The costs associated with these embodiments are thus reduced.

[0084] . According to another embodiment, the first model is trained to receive infrared images. In this case, the input of the first model may be capable of directly receiving the images from the image acquisition device 106, which is then an infrared camera. The use of infrared images allows better accuracy in the prediction of said at least one piece of conformity information, due to better contrast compared to grayscale or color images.

[0085] . According to yet another embodiment, the first model is trained to receive luminance images.

[0086] Luminance is a quantity that designates the luminous flux coming from an illuminated surface and which is reflected in the eye, which is expressed in candela per square meter, or cd / m 2 Luminance thus translates a visual sensation of brightness of a surface.

[0087] More precisely, luminance is the power of visible light passing through or being emitted by a surface element in a given direction, per unit area and per unit solid angle.

[0088] In outdoor lighting, the measurement of luminance is normalized according to the distance from the surface, the position and height of the observer, the angle of observation, etc.

[0089] . Thus, the luminance map is distinct from a grayscale image, in which the grayscale of a pixel illustrates the brightness perceived by a camera sensor at a given location, but which does not correspond to luminance as defined above.

[0090] . Predicting compliance information based on a luminance image allows for greater accuracy in prediction, particularly when the scene in front of the vehicle is lit by street lighting or interior lighting (parking lot), which makes it difficult to distinguish ambient lighting from that coming from the lighting beam of the vehicle's lighting module.

[0091] . In this embodiment, the input of the first model may be capable of directly receiving the images from the image acquisition device 106, which is then a luminance camera. However, such a luminance camera is expensive and also requires significant processing times to obtain a luminance image. It is therefore incompatible with real-time applications.

[0092] . Preferably according to this embodiment of the invention, a second model is trained to predict a luminance image from a grayscale or color image received as input to the model. The second model can thus be placed in interception between the image acquisition device 106, and the first model implemented by the device 101, and in this case, the image acquisition device 106 can be a color or grayscale camera, which has the advantages presented above.

[0093] . The second model can be obtained according to steps 211 to 213 described below.

[0094] . At a step 211, a second training database can be created.

[0095] The second training database may comprise second training data comprising second grayscale or color training images from one camera, or from several separate cameras. In particular, the second training images of the second training database may come from different cameras, in particular from cameras of different types, two different types having at least one characteristic technique that differs or has different manufacturers. The creation of a second database with second training images from cameras of different types allows the training of a generic model, which can then be integrated into any vehicle, regardless of the image acquisition device 106 (capable of acquiring grayscale or color images) of the vehicle 100.

[0096] Alternatively, all the images come from cameras of the same type as the image acquisition device 106, in which case the second model trained from the second training data is specific to the image acquisition device 106 of the vehicle 100.

[0097] . Preferably, the second training data of the second base constituted in step 211 comprise associations between the aforementioned second training images and respective training luminance maps. Each training luminance map, or image, associated with a second training image can be acquired by a luminance camera, placed next to the camera acquiring the second training image, so as to acquire visual data relating to the same scene as that of the camera.

[0098] . At a step 212, a second model may be trained based on the second training data of the second base trained at step 211. No restriction is attached to the second model resulting from the training of step 212, which may be: - a curve or function matching a value of a pixel in a grayscale or color image with a luminance value of a corresponding pixel in a luminance map. Such a curve can be obtained by a curve fitting analysis technique, or "curve fitting" in English, from the associations stored in the second training database. Indeed, from these associations in the second training database, each association associating a second training image with a training luminance map, pixel associations can be determined between a given pixel in the training image and a corresponding pixel in the training luminance map. These pixel associations feed into the curve fitting analysis to obtain a model in the form of a curve; - a support vector machine, or SVM model, for "Support Vector Machine" in English, capable of determining, by classification, a luminance value of a pixel of a luminance map, from a value of a corresponding pixel in a grayscale or color image. An SVM model can be constructed by supervised learning from the associations of the second training database. Indeed, from these associations of the second training database, each association associating a training image with a training luminance map, pixel associations can be determined between a given pixel of the second training image and a corresponding pixel in the training luminance map. These associations between pixels feed the supervised learning of the SVM model.The advantage of an SVM model is that it can be trained from a second training database with little data; - a neural network, for example a convolutional neural network, capable of determining an entire luminance map from a grayscale or color image received as input. The neural network can be obtained by machine learning from the second training data of the second training database constituted in step 211. In the case of supervised learning, the neural network can be configured by modifying parameters of certain neurons, as a function of a measured difference between the luminance map determined, or predicted, by the neural network when a second given training image is submitted to it, and the training luminance map associated in the second training database with the second given training image; - any other model resulting from machine learning or a given analysis technique, and capable of determining a luminance value from of a pixel value in grayscale or color, or capable of determining an entire luminance map from a grayscale or color image received as input

[0099] Step 212 can be implemented by a training module not shown in FIG. 1, which can be external to the vehicle 100, and which is capable of accessing the second training database.

[0100] . At the end of the training of step 212, which is an analysis or machine learning as detailed previously, a second model according to the invention is obtained, the model being capable of determining a luminance value from a value of a pixel in grayscale or in color, or capable of determining an entire luminance map from a grayscale or color image received as input.

[0101] . In a step 213, the second model resulting from the training of step 212 is stored in the motor vehicle 100, either in a memory of the device 101, or in a memory external to the device 101 but accessible to the device 101.

[0102] The method according to the invention further comprises a current phase 220, comprising steps 221 to 227 described below.

[0103] . In a step 221, the device 101 verifies that at least one lighting module of the vehicle 100 is switched on. In particular, according to the invention, the device 101 can verify that the right and left dipped beam lighting modules are switched on. Such a verification can consist of checking the activation state of the dipped beam lighting modules by interrogating the control module 102, or can be based on an identification of the lighting beams in the data received from the image acquisition device received in step 222 (in which case, the verification of the activation of the lighting module of step 221 is implemented after step 222).

[0104] . In the case where said at least one lighting module of the vehicle 100 is not lit, the device 101 can request the activation of said at least one lighting module, in particular from the control module 102. As a variant, the device regularly interrogates, at a given frequency for example, the activation state of said at least one lighting model, and the following steps of the current phase 220 are only implemented once said at least one lighting module is lit.

[0105] . In a step 222, the device 101 receives data from the image acquisition device, the data comprising one or more images from the image acquisition device 106.

[0106] . In a step 223, the device 101 obtains at least a first image from the data received in step 222 from the image acquisition device 106.

[0107] . Said at least one first image obtained in step 223 may be a single first image or a series of several first images.

[0108] . When the data from the image acquisition device 106 comprises a single image, this image constitutes the first single image obtained in step 223. When the data from the image acquisition device 106 comprises several images acquired successively, the first single image can be: - an image selected from among the several images acquired successively by the image acquisition device 106; - an image determined from the several images acquired successively. The image thus determined can be a stabilized or averaged image, which makes it possible to obtain a sharper image than each of the images acquired by the image acquisition device 106.

[0109] . Alternatively, said at least one first image is a series of first images, in the case where the first model is capable of receiving a series of images as input. In this case, the series of first images may be the series of images acquired by the image acquisition device 106.

[0110] Each first image 106 may be, depending on the embodiment corresponding to the first model obtained during the preliminary phase 200, a grayscale image, a color image, an infrared image or a luminance image.

[0111] . In the case where the first model is capable of receiving a luminance image as input, the method may comprise a step 224, implemented by the device 101 between steps 222 and 223, of applying the second model stored in step 213, to the data from the image acquisition device 106, in order to obtain said at least one first luminance image (first single luminance image or series of luminance images).

[0112] To obtain a series of first luminance images, the device 101 applies the second model sequentially to a series of images (in grayscale or in color) from the image acquisition device 106.

[0113] . In a step 225, which follows step 223 of obtaining said at least one first image, the device 101 applies the first model stored in the vehicle 100 to said at least one first image to obtain, by prediction, at least one piece of conformity information.

[0114] As indicated above, the conformity information can be binary or can take strictly more than two values, in order to indicate, in the event of non-conformity, a type of non-conformity.

[0115] . In a step 226, the device 101 evaluates the conformity of a lighting beam of at least one lighting module as a function of said at least one first piece of conformity information. In particular, the evaluation indicates whether the lighting beam(s) from the lighting modules are or are not compliant with a technical and / or regulatory standard.

[0116] . The evaluation of step 226 can be obtained from: - of at least one first conformity information from step 225, when the first model is applied only once to said first picture; - several pieces of compliance information from at least two iterations of steps 222 to 225 described previously. At least one piece of first compliance information is obtained at each iteration, and the first pieces of compliance information are taken into account to evaluate the compliance of the lighting beam(s), which improves the precision associated with the compliance assessment.

[0117] In a step 227, the device 101 transmits the conformity assessment obtained in step 226 to another entity of the vehicle 100. Such a transmission may only be implemented when the assessment indicates that at least one lighting beam is not compliant.

[0118] . Thus, the assessment can be transmitted to the driver or to a third party, which makes it possible to implement a corrective action in the event of non-compliance. The assessment can also indicate a corrective action, or correction, to be implemented, the correction being determined by the device 101 from said at least one first piece of compliance information.

[0119] . The assessment can for example be transmitted: - to the HMI 108 in order to indicate to the driver that a non-conformity of at least one lighting beam is detected, or in order to indicate that the lighting beams are compliant; - to memory 107 for storage and later consultation; - to a wireless communication module for transmitting the evaluation to a remote server, accessible via a telecommunications network.

[0120] . Figure 3a illustrates lighting beams 301.1 and 301.2 of lighting modules 103.1 and 103.2 of the motor vehicle 100 according to embodiments of the invention, in a case of conformity of the lighting beams.

[0121] . Figure 3a illustrates the projection on the ground of lighting beams projected by lighting modules 103.1 and 103.2, which are dipped beam lighting modules, right and left, for illustrative purposes. The projection of the illumination beams 301.1 and 301.2 is presented from an aerial view above the road onto which the illumination beams 301.1 and 301.2 are projected. Such a view is distinct from the perspective of the image acquisition device 106 arranged in the vehicle. However, the aerial view is presented to facilitate the understanding of a given standard for illustrative purposes.

[0122] . A standard against which the conformity of the lighting beams 301 .1 and 301 .2 may be that the lighting beams 301 .1 and 301.2 do not intersect within a reference zone 302, which may be an interval of predefined distances from the front of the vehicle. More specifically, the standard may require that the two beams intersect at the end of the reference zone 302.

[0123] . Thus, in the example shown in Figure 1, the lighting beams 301.1 and 301.2 comply with the standard described above. In the case where the first model is trained to identify compliance information with respect to this standard only, then a training image obtained for lighting beams whose projection is similar to that of Figure 3a, can be associated with compliance information indicating that the lighting beams are compliant.

[0124] The standard defined from the reference area is given for illustrative purposes. Other standards may be assessed, in addition or as a variant, to characterize compliance.

[0125] . Figure 3b illustrates lighting beams 311.1 and 311.2 of the lighting modules 103.1 and 103.2 of the vehicle 100, according to embodiments of the invention, in a case of non-conformity of the lighting beams with respect to the standard described previously.

[0126] . In the case of Figure 3b, the lighting beams 311.1 and 311.2 intersect in the reference area, so they do not comply with the standard described above.

[0127] . In the case where the first model is trained to identify compliance information with respect to this standard only, then a training image obtained for lighting beams whose projection is similar to that of Figure 3b, can be associated with compliance information indicating that the lighting beams are non-compliant.

[0128] . As described above, the compliance information may be more specific than simply indicating non-compliance: it may further indicate a type of non-compliance. In the example of Figure 3b, the compliance information may indicate that the non-compliance is due to a beam crossing in the reference area 302. More specifically, the compliance information may indicate that the right beam is in too low a position.

[0129] . Figure 4 shows the structure of the device 101 according to embodiments of the invention.

[0130] . The device 101 comprises a processor 401 configured to communicate unidirectionally or bidirectionally, via one or more buses or via a direct wired connection, with a memory 402 such as a memory of the “Random Access Memory” type, RAM, or a memory of the “Read Only Memory” type, ROM, or any other type of memory (Flash, EEPROM, etc.). Alternatively, the memory 402 comprises several memories of the aforementioned types.

[0131] The memory 402 is capable of storing, permanently or temporarily, at least some of the data used and / or resulting from the implementation of the steps of the method according to the invention illustrated with reference to figure 2.

[0132] In particular, the memory 402 may be capable of storing the first model, during step 203, and optionally the second model during step 213.

[0133] . Furthermore, said at least one first image obtained in step 223 can be stored, at least temporarily, in the memory 402.

[0134] The processor 401 is capable of executing instructions, stored in the memory 402, for the implementation of steps 221 to 227 of the current phase 220, described with reference to FIG. 2. Alternatively, the processor 401 can be replaced by a microcontroller designed and configured to carry out steps 221 to 227 of the current phase 220, described with reference to FIG. 2.

[0135] The device 101 comprises a first interface 403 capable of receiving data from the image acquisition device 106.

[0136] The device 101 comprises a second interface 404 capable of transmitting the evaluation to an entity external to the device 101, during the step 227 described previously.

[0137] The device 101 comprises a third interface 405 capable of accessing the first model and / or the second model, when the first model and / or the second model is stored in the vehicle 100 outside the device 101.

[0138] The present invention is not limited to the embodiments described above as examples; it extends to other variants.

Claims

Claims

1. Method for evaluating the conformity of a lighting beam (301.1; 301.2; 311.1; 311.2) of at least one lighting module (105; 104; 104.1; 104.2) of a motor vehicle (100), the method comprising a prior step of storing (203) a model in a memory (107; 402) of the motor vehicle, the model being capable of determining at least one piece of lighting beam conformity information from at least one image received as input to the model, the method further comprising the following common steps: - verification (221) of an activation of at least one lighting module of the motor vehicle; - obtaining (223) at least one first image from data from an image acquisition device (106) of the vehicle, said image acquisition device being arranged so as to obtain data representative of a scene facing the vehicle; - application (225) of the model to said at least one first image to determine at least one first lighting beam conformity information; - evaluation (226) of a conformity of a lighting beam of said at least one lighting module as a function of said at least one first conformity information.

2. Method according to claim 1, further comprising a prior step of obtaining (202) the model by machine learning from a training data set.

3. The method of claim 2, wherein the model is obtained (202) by supervised learning, and wherein the training data set comprises associations between at least one training image and at least one piece of illumination beam conformity information.

4. Method according to one of the preceding claims, in which the model is capable of determining a binary conformity category from among a conforming category and a non-conforming category.

5. Method according to one of claims 1 to 3, in which the model is capable of determining a type of non-conformity from among at least two distinct types of non-conformity.

6. Method according to claim 5, wherein, upon applying (225) the model to said at least one first image, a first type of non-conformity is determined, wherein the evaluation comprises an indication of a correction to be made to the lighting module, said correction depending on the first type of non-conformity determined.

7. Method according to one of the preceding claims, in which the image acquisition device (106) is a camera capable of obtaining color or grayscale images.

8. Method according to claim 7, in which the model is capable of determining at least one piece of lighting beam conformity information from at least one color or grayscale image received as input, in which said at least one first image obtained is in color or grayscale and comes from the image acquisition device (106).

9. Method according to claim 7, wherein the model is a first model capable of determining at least one lighting beam conformity information from at least one luminance image, wherein the method further comprises a prior step of storing (213) a second model in said motor vehicle (100), said second model being capable of determining at least one luminance value of a pixel of a luminance image from at least one value of a pixel of a color or grayscale image; wherein said at least one first image obtained is a first luminance image; wherein obtaining said at least one first luminance image comprises applying (224) the second model to pixels of at least one first color or grayscale image obtained by the image acquisition device (106); said first model being applied to said at least one first luminance image to determine said at least one piece of lighting beam conformity information.

10. Method according to one of claims 1 to 6, in which the image acquisition device (106) is an infrared camera capable of obtaining infrared images, in which the model is capable of determining at least one piece of lighting beam conformity information from at least one infrared image received as input,

11. Method according to one of the preceding claims, wherein said at least one activated lighting module (104; 105; 104.1; 104.2) comprises a right dipped beam module (104.2) and a left dipped beam module (104.1) of the motor vehicle (100).

12. Method according to one of the preceding claims, in which the first model is capable of determining at least one piece of lighting beam conformity information from a series of images received as input, in which said at least one first image is a series of first images, and in which the model is applied (225) to the first series of images to obtain said at least one piece of first conformity information.

13. Computer program comprising instructions for implementing the method according to one of the preceding claims, when these instructions are executed by a processor (401).

14. Device (101) for vehicle (100) comprising: - a memory (402) storing a model capable of determining at least one piece of lighting beam conformity information from at least one image received as input to the model, or an access interface (405) to said model; - a processor (401) configured to: verify an activation of at least one lighting module (104; 105; 104.1; 104.2) of the vehicle; obtaining at least one first image from data from an image acquisition device (106) of the vehicle, said image acquisition device being arranged so as to obtain data representative of a scene facing the vehicle; applying the model to said at least one first image to determine at least one first lighting beam conformity information; evaluating a conformity of a lighting beam of said at least one lighting module as a function of said at least one first conformity information.

Images