Method for evaluating software to control an adaptive light projector

Abstract

Method for evaluating software for controlling an adaptive headlight. Method for evaluating software (33) for controlling an adaptive headlight (2) for a first motor vehicle (1), the first vehicle comprising said adaptive headlight and a camera (4) arranged to observe a scene in front of the first vehicle, the evaluation method comprising a first step (E1) of providing a video recording (6) and the control software (33) to be evaluated, the video recording comprising a predefined duration and representing a scene in front of the first vehicle in which the second vehicle appears. Figure for the abstract: Figure 4

Description

Title of the invention: Method for evaluating software for controlling an adaptive light projector. Technical field of the invention

[0001] The invention relates to a method for evaluating software for controlling an adaptive headlight for a motor vehicle. The invention also relates to a device for evaluating an adaptive headlight, the evaluation device comprising software and hardware means adapted to implement such an evaluation method. Prior art

[0002] Motor vehicles are equipped with headlights designed to illuminate a scene in front of the vehicle. At night, when no other motorist is present in the scene in front of the vehicle, and when the vehicle is traveling outside built-up areas, the headlights are generally used in the so-called "high beam" configuration. In this configuration, the headlights emit a powerful beam of light, allowing for very effective illumination of the scene. Conversely, when another motorist is present in the scene in front of the vehicle or when the vehicle is traveling in a built-up area, the headlights must be used in the so-called "low beam" configuration. In this configuration, the headlights emit a less powerful beam of light and / or have characteristics of shape and position designed to avoid dazzling other motorists or road users.

[0003] Some motor vehicles are equipped with so-called adaptive headlights. Such headlights are configured to automatically adapt the type of light beam they emit based on the detection of the presence or absence of at least one other driver. This automatic detection is generally performed by an electronic control unit installed in the vehicle, the electronic control unit being coupled to a camera configured to observe the scene in front of the vehicle. In particular, the electronic control unit includes control software. The control software is configured to process a video signal from the camera and to perform high-frequency detection algorithms designed to effectively detect the presence or absence of other drivers.

[0004] The detection algorithms of the control software are complex and rely primarily on identifying the headlights of other motor vehicles. The detection algorithms must lead to effective detection of other Vehicles in highly varied situations, including diverse weather conditions and / or when the scene ahead is obstructed by obstacles or luminous objects that do not correspond to vehicle headlights. In particular, road signs and / or painted road markings generally have reflective properties: when illuminated, these objects become very bright and can easily be mistaken for the headlights of other vehicles. In all these situations, detection algorithms must effectively detect the presence of other drivers so as not to dazzle them, and must also effectively detect the absence of other drivers to allow for comfortable illumination of the scene.

[0005] Detection algorithms, front-facing cameras, and vehicle headlights are constantly evolving. These developments require validation to ensure that the detection algorithm functions correctly. To evaluate a detection algorithm, tests are generally conducted on vehicles at night in various situations, and it is verified that these situations are correctly interpreted by the algorithm. These tests require significant human resources: not only to drive the vehicle being tested, but also to generate different scenarios.

[0006] In addition to verifying that the detection algorithm correctly detects the presence of a second vehicle, it is useful to know the distance between the vehicle being tested and the second vehicle at the time of this detection. Indeed, to comply with automotive regulations, adaptive headlights must automatically switch from high beam to low beam as soon as the distance between two vehicles is less than or equal to a given threshold. During validation testing, measuring this distance is also very complex, as it requires precise knowledge of the relative positions between the vehicle being tested and the second vehicle at the moment the presence of the second vehicle is detected by the vehicle being tested.

[0007] There is therefore a need to evaluate more simply the performance of a software for controlling an adaptive light projector. Presentation of the invention

[0008] The object of the invention is to provide a method for evaluating software for controlling an adaptive light projector, remedying the above drawbacks and improving the evaluation methods known in the prior art.

[0009] More specifically, a first object of the invention is a method for evaluating a software for controlling an adaptive light projector that is simple and quick to implement. Summary of the invention

[0010] The invention relates to a method for evaluating software for controlling an adaptive headlight for a first motor vehicle, the first vehicle comprising said adaptive headlight and a camera arranged to observe a scene in front of the first vehicle, the control software to be evaluated comprising: - a detection algorithm adapted to verify the presence of at least a second vehicle in front of the first vehicle by processing images provided by said camera, the detection algorithm being capable of calculating an initial verification result, and - a module for adapting a beam of light emitted by the adaptive projector based on said first verification result, the evaluation method comprising: - a first step of providing a video recording and the piloting software to be evaluated, the video recording comprising a predefined duration and representing a scene in front of the first vehicle on which the second vehicle appears, then - a second step of executing the piloting software to be evaluated, the detection algorithm being carried out by substituting the images provided by the camera with images from the video recording, then - a third stage of evaluating the control software based on the first verification result obtained at the end of the second stage, during the execution of the control software to be evaluated.

[0011] The third step may include: - a first sub-step of automatic verification of the presence of the second vehicle in the video recording by processing the images of the video recording, the first sub-step leading to a second verification result, then - a second sub-step of comparing the first verification result with the second verification result.

[0012] The first sub-step may include: - a sub-step of identifying a first image of said video recording at the moment when the first vehicle detects the presence of the second vehicle during the execution of the piloting software to be evaluated, and - a sub-step of identifying the second vehicle in the first image.

[0013] The third step may include a substep of estimating the distance separating the first vehicle from the second vehicle at a time when the first vehicle detects the presence of the second vehicle during the execution of the piloting software to be evaluated.

[0014] Said substep of estimating the distance separating the first vehicle from the second vehicle may include an estimation of a height of said second vehicle on said first image, then a calculation of the distance separating the first vehicle from the second vehicle as a function of the height of said second vehicle on said first image.

[0015] Said substep of identifying the second vehicle in the first image may include: - a step of identifying the second vehicle on a second image of the video recording, the second image being later than the first image in the chronological order of the video recording, then - the identification of the second vehicle on the first image of the video recording, by reverse tracking of the second vehicle on at least part of the images separating the second image from the first image in the video recording.

[0016] The step of identifying the second vehicle on the second image of the video recording may include - the calculation of a given R ratio using the following formula: R = intersection (Ax, By) / union (Ax, By) Or : Ax designates a first area of ​​the second image, the first area being determined by means of a first algorithm, in particular a shape detection algorithm, B designates a second area of ​​the second image, the second area being determined by means of a second algorithm, in particular a light intensity detection algorithm, then - the comparison of the R ratio with a predefined threshold.

[0017] The second image can be selected from a subset of the images in the video recording by calculating a score for each of the images in said subset.

[0018] Said score can be calculated for a given image using the following formula: S = max (intersection (Ax, By) / union (Ax, By)) where: Ax designates a first area of ​​the given image, the first area being determined by means of a first algorithm, in particular a shape detection algorithm, By designates a second area of ​​the given image, the second area being determined by means of a second algorithm, in particular a light intensity detection algorithm.

[0019] The invention also relates to a device for evaluating control software intended to control an adaptive headlight for a motor vehicle, the evaluation device comprising software and hardware means adapted to implement the evaluation method as defined above. Figures

[0020] These objects, features and advantages of the present invention will be described in detail in the following description of a particular embodiment, given by way of non-limiting example, with reference to the accompanying figures, among which:

[0021] Fig. 1 is a schematic view of a motor vehicle according to one embodiment of the invention.

[0022] Fig. 2 is a schematic view of an evaluation device for control software for an adaptive light projector.

[0023] Fig. 3 is a synoptic diagram of a method for evaluating software for controlling an adaptive light projector according to an embodiment of the invention.

[0024] Fig. 4 is a first image of a scene in front of the vehicle, the first image being from a camera mounted in the vehicle.

[0025] Fig. 5 is a second image of a scene in front of the vehicle, the second image being from the camera mounted in the vehicle.

[0026] Fig. 6 is a schematic view of an image from the camera mounted in the vehicle.

[0027] Fig. 7 is a nomogram illustrating a relationship between the height of an object detected on an image, the image being from a camera mounted in the vehicle, and a distance separating said object from said camera. Detailed description

[0028] Figure 1 schematically illustrates a first motor vehicle 1. The first vehicle 1 can be, for example, a passenger car, a commercial vehicle, a truck, or even a bus. The first vehicle 1 comprises an adaptive headlight 2, an electronic control unit 3, and at least one camera 4 (or any other type of sensor, such as a radar or LiDAR sensor, capable of providing a representation of a scene in the form of an image). The adaptive headlight 2 and the camera 4 are connected to the electronic control unit 3. The adaptive headlight 2 is configured to emit two types of light beams. A first type of light beam, called the "high beam," is a powerful beam that very effectively illuminates the scene in front of the vehicle 1. The second type of light beam, called the "low beam," is less powerful than the "high beam" and illuminates the scene in front of the vehicle 1 without dazzling other drivers or road users. The camera 4 is capable of observing the scene in front of the vehicle. Specifically, the camera 4 is capable of providing images of the scene to the electronic control unit 3 at a given frequency. The electronic control unit 3 includes a memory 31 and a microprocessor 32. The memory 31 contains control software 33 adapted to control, that is, to drive, the type of light beam emitted by the adaptive headlight 2 based on a video signal received by the camera 4. The microprocessor 32 is capable of executing the control software 33.

[0029] In particular, the control software 33 includes: - a detection algorithm 35 adapted to verify the presence of at least a second vehicle in front of the first vehicle 1 by processing the images provided by the camera 4, the detection algorithm being capable of calculating a first verification result, and - an adaptation module 36 of a lighting beam emitted by the adaptive light projector 2 according to said first verification result. The detection algorithm 35 and the adaptation module 36 contain instructions, in the form of computer code, the execution of which by the microprocessor 32 leads to the implementation of a method for controlling the adaptive light projector 2.

[0030] According to one embodiment, said first result may be binary information simply indicating the presence or absence of any vehicle in front of the first vehicle 1. When the first result indicates the presence of a second vehicle in front of the first vehicle, the adaptation module 36 issues a command to the adaptive headlight 2 so that the latter produces a low beam. When the first result indicates the absence of any vehicle in front of the first vehicle, the adaptation module 36 issues a command to the adaptive headlight 2 so that the latter produces a high beam. In addition, the first result may also include information relating to the distance between the first vehicle and the detected vehicle.Switching from "high beam" to "low beam" lighting can only be controlled if the distance between the first vehicle and the detected vehicle is less than or equal to a given threshold.

[0031] The object of the invention is to evaluate simply and quickly the performance of the control software 33 integrated, or intended to be integrated, into the memory 31 of the electronic control unit 3. To this end, the invention consists of a method for evaluating the control software 33 in the laboratory, and not by means of vehicle tests. [Fig.2] illustrates an evaluation device 5 of the control software 33. The evaluation device 5 can be, for example, a computer equipped with a memory 51 and a microprocessor 52. The memory 51 includes instructions, in the form of computer code, the execution of which by the microprocessor 52 leads to the implementation of said evaluation method.

[0032] According to one embodiment, the evaluation device 5 is configured to directly test the control software 33, which thus constitutes input data for the evaluation method that will be described. Alternatively, the evaluation device 5 can be configured to test the electronic control unit 3, with the control software 33 being loaded into the memory 31 of the electronic control unit 3. The evaluation device 5 can also be configured to observe a command issued by the adaptation module 36. Alternatively, the electronic control unit 3 can be supplied with the adaptive light projector 2, and this assembly can be mounted on a test bench. In this case, the evaluation device 5 can be configured to observe a lighting switching of the adaptive light projector 2. An evaluation device 5 based solely on the supply of the control software 33 is simpler to implement.An evaluation device 5, based on the provision of the electronic control unit 3 with the control software 33 loaded into memory 31, or even based on the provision of the electronic control unit 3 and the adaptive headlight 2 connected to the electronic control unit 3, allows for a more representative evaluation of real-world operating conditions. In all cases, no vehicle is required, nor even a camera 4. Instead of a camera, one or more video recordings 6 are used to evaluate the control software 33. Thus, the evaluation device 5 receives as input the command issued from the adaptation module 36 to the adaptive headlight 2, which is essentially the output of the control software 33, and at least one video recording 6. The evaluation device 5 provides as output an evaluation result 7 of the control software 33.Advantageously, the evaluation device 5 includes a screen for displaying the evaluation result 7. The evaluation result 7 provides a performance level of the piloting software 33 and / or provides an indication of scenarios that the piloting software 33 fails to interpret correctly. In particular, the evaluation result 7 can indicate whether the detection algorithm 35 correctly detects a second vehicle appearing on the video recording 6.

[0033] The evaluation method is now described in more detail in relation to the synoptic diagram illustrated in [Fig. 3]. According to the embodiment presented, the evaluation method mainly comprises three steps E1, E2, E3.

[0034] In a first step E1, a video recording 6 is provided, the video recording comprising a predefined duration and representing a scene in front of the first The video recording 6 may be in the form of a computer file, for example, an "avi" file or equivalent. The video recording may be from a database containing numerous video recordings. The video recording may have been previously made using a camera of the same type as camera 4 or even a different type. The video recording 6 may be broken down into a series of successive images, or frames, each image corresponding to a given moment. Advantageously, the video recording 6 includes a timestamp for each image.

[0035] The use of a video recording avoids the need for a test vehicle to obtain a video signal for evaluating the control software 33. In particular, the video recording may have been made at night, and the evaluation method using the recording 6 can be implemented during the day. Advantageously, the same video recording 6 can be reused to test different control software. Another advantage of using a video recording is that it can be played back in fast motion without compromising the representativeness of the evaluation method. Thus, using a video recording reduces the time required to validate the control software 33 and / or allows for testing a greater number of configurations within the same period.

[0036] Next, in a second step E2, the control software 33 is executed by feeding it a video signal from the video recording 6. For this purpose, the evaluation device 5 may advantageously include a conversion module 8 capable of producing a video signal compatible with the control software 33. The control software 33 may be executed using an emulator 9 adapted for this purpose. Alternatively, the control software 33 may be executed using the electronic control unit 3. During the execution of the control software 33, the detection algorithm 35 calculates said first verification result.

[0037] Next, in a third step E3, the control software is evaluated according to the first verification result obtained during the execution of the control software 33. In practice, four different situations can then occur: IF) the video recording actually contains the representation of a second vehicle and this has been correctly detected by the detection algorithm 35. S2) the video recording does indeed contain the representation of a second vehicle but this one was not detected by the detection algorithm 35. S3) the video recording does not contain any representation of a second vehicle, however the detection algorithm 35 wrongly detected the presence of a second vehicle. S4) the video recording does not contain any representation of a second vehicle, and the detection algorithm 35 has correctly detected the absence of any vehicle. Of course, the SI and S4 situations mentioned above lead to a positive evaluation of the control software 33, while the S2 and S3 situations lead to a negative evaluation of the control software 33.

[0038] To determine whether video recording 6 contains the representation of a second vehicle or not, several strategies are possible.

[0039] According to a first strategy, which is very simple to implement, the video recording 6 can be viewed beforehand by an operator to determine whether or not it contains the representation of a second vehicle. The operator then manually enters a second result associated with the video recording 6. This second result can be a simple binary value indicating whether or not the video recording 6 contains the representation of a second vehicle. The second result can then be compared to the first result calculated by the detection algorithm 35 to determine the corresponding situation among the four situations S1, S2, S3, and S4 described previously. This method is simple to implement but relatively tedious: if one wishes to reproduce the validation method with numerous video recordings, the prior viewing of many hours of recording can prove prohibitive.

[0040] Thus, according to a second particularly advantageous strategy, the third step E3 comprises a first substep E31 of automatic verification of the presence of the second vehicle in the video recording by processing the images of the video recording, the first substep leading to a second verification result. The third step E3 then comprises a second substep E32 of comparison of the first verification result with the second verification result. The first substep E31 is therefore carried out by automatic processing of the images of the video recording.

[0041] To automatically verify the presence of the second vehicle on the video recording 6, the following procedure can be applied. In a first substep E311, a first image 11 of the video recording 6 is identified at the instant when the first vehicle detects the presence of the second vehicle during the execution of the control software 33. This first substep can be carried out simply by identifying the image of the video recording whose moment of appearance best corresponds to the instant when the control software commands the switching of the adaptive lighting device. The timestamp of the video recording can be used for this purpose.

[0042] Next, in a second substep E312, the second vehicle V2 is identified on the first image il. According to a first embodiment, the second vehicle V2 can be detected directly by processing the first image il. However, this method has a drawback: the second vehicle V2 is often barely visible in the first image, and detection based on the processing of the first image would introduce uncertainty into the second result. Therefore, to accurately assess the performance of the control software 33, it is particularly important that the automatic verification of the presence of the second vehicle V2 in the first image be highly reliable, and in any case, more reliable than the detection algorithm 35 integrated into the control software 33.

[0043] Advantageously, the automatic verification of the second vehicle V2 on the video recording 6 during the second substep E312 can be performed a posteriori, that is, this verification can be carried out by using the entire video recording 6. This automatic verification can, in particular, use images from the video recording on which the second vehicle V2 is very close to the first vehicle VI and clearly appears. A posteriori verification of the presence of the second vehicle V2 makes it possible to obtain a second, highly reliable verification result.

[0044] Thus, according to an advantageous embodiment, the second substep E312 comprises a first substep E3121 for identifying the second vehicle V2 on a second image i2 of the video recording, the second image i2 being chronologically subsequent to the first image i1 (according to the chronology of the recording). In particular, assuming that the second vehicle V2 is traveling in the opposite direction to the first vehicle VI, the second image i2 may be an image corresponding to a moment when the first vehicle V1 crosses paths with the second vehicle V2. Assuming that the second vehicle V2 is traveling in the same direction as the first vehicle VI, the second image i2 may be an image corresponding to a moment when the first vehicle VI has stabilized its safe distance from the second vehicle V2.In any case, identifying the second V2 vehicle in the second image i2 is easy to do since the second V2 vehicle appears in relatively close-up in this image.

[0045] Next, in a second substep E3122, the second vehicle V2 is identified on the first frame i1 of the video recording by reverse tracking of the second vehicle V2 over at least a portion of the frames separating the second frame i2 from the first frame i1. In other words, the video recording is played back in reverse chronological order from the second frame i2 to the first frame i1, and the movement of the second vehicle V2 is tracked over all the frames separating the second frame i2 from the first frame i1, or at least over a portion of the frames separating the second frame i2 from the first frame i1. This reverse tracking is simple to implement because the movement of the second vehicle V2 between two consecutive frames or two frames close in time is relatively small. We can thus identify the position of the second vehicle V2 on the first image by cleverly reading the video recording 6 in reverse.

[0046] To illustrate these explanations, an example of the first image 11 is shown in [Fig. 4]. This image 11 corresponds to the frame in the video recording 6 at the moment when the control software 33 detects the second vehicle V2 and commands the adaptive headlight system to switch to low beam. In this image 11, the second vehicle V2 is very far away and could easily be mistaken for a reflective sign or any other luminous object present in the scene. It is therefore difficult to state with certainty, based solely on the analysis of the first image 11, that the detection of the second vehicle V2 by the control software 33 was performed correctly. Next, an example of the second image 12, subsequent to the first image 11 in the video recording 6, is shown in [Fig. 5]. In the second image 12, the two headlights of the second vehicle V2 are clearly identifiable.Identifying the second vehicle V2 in the second image V2 is much simpler and more reliable. By performing a reverse tracking of the second vehicle V2 between the second image i2 and the first image il, we can confirm the presence of the second vehicle V2 in the first image il.

[0047] An embodiment of the identification of the second vehicle V2 on the second image i2 during the first substep E3121 is now described with reference to [Fig. 6]. According to this embodiment, two different algorithms are applied to the second image i2. In the example shown, the first algorithm leads to the detection of a first set of rectangular objects A1, A2, A2, A3, A4, A5, identified by solid lines. The second algorithm leads to the detection of a second set of rectangular objects B1, B2, B3 identified by dashed lines. Each rectangular object A1, A2, A2, A3, A4, A5, B1, B2, B3 can be identified by the coordinates of all or some of its corners on the image i2 and / or by the coordinates of at least one of its points combined with the values ​​of its height and width.Next, for each pair Ax, By consisting of an object from the first set of rectangular objects Al, A2, A2, A3, A4, A5 and an object from the second set of rectangular objects Bl, B2, B2, we calculate the ratio R with the following formula: R = intersection (Ax, By) / union (Ax, By). . The function "intersection (Ax, By)" calculates the area of ​​the surface belonging to both object Ax and object By. The function "union (Ax, By)" calculates the area of ​​the surface formed by the union of objects Ax and By. The ratio R is thus automatically zero for any pair of objects Ax, By with no shared surface. According to the example in [Fig. 6], the ratio R is maximal for the pair formed by objects A1 and B1. When the ratio R is greater than or equal to a given threshold, we can conclude that the second vehicle V2 is present in the second image i2.

[0048] The first algorithm could, for example, be an object detection algorithm. Such an algorithm is particularly effective for detecting the second vehicle at close range. The second algorithm could be a light intensity detection algorithm. The second algorithm could, for example, identify objects whose light intensity is greater than or equal to a given threshold. Such an algorithm is particularly effective for detecting the second vehicle V2 at a distance. The first and / or second algorithms could be developed using artificial intelligence techniques.

[0049] As a note, to select the second image i2 from the set of images in the video recording 6, substep E3121, which has just been described, can be performed for each of the images subsequent to the first image i1 or for a subset of images subsequent to the first image i1. For each of these images, a score S can be calculated as the maximum of the ratios R for that image. The second image can then be selected as the one with the highest score S.

[0050] Furthermore, it is possible that the control software 33 does detect the presence of the second vehicle in the video recording, but that this detection occurs too late relative to the desired performance. A detection of the second vehicle that is too late would lead to a delayed switching of the adaptive headlight 2, and therefore to momentary glare for the driver of the second vehicle.

[0051] Advantageously, the third step E3 also includes a third substep E33 of estimating the distance separating the first vehicle VI from the second vehicle V2 at a time when the first vehicle VI detects the presence of the second vehicle V2 during the execution of the piloting software 33 to be evaluated, and then a fourth substep E34 of comparing this distance with a predefined distance, for example a distance imposed by motoring legislation.In particular, automotive legislation mandates the use of dipped headlights when the second vehicle V2 is travelling in the opposite direction to the first vehicle VI and is less than 400m from the first vehicle VI, and when the second vehicle V2 is travelling in the same direction as the first vehicle and is less than 200m from the first vehicle VL. Thus, if the distance estimated during sub-step E33 is greater than or equal to 400m or respectively to 200m depending on the case considered, the driving software 33 will be evaluated positively.

[0052] The distance estimated in substep E33 is a virtual distance since the evaluation method is carried out without recourse to automotive tests. According to one embodiment, the third substep E33 includes an estimation substep E331 The height H of the second vehicle V2 in the first image (il) is determined. This height H can be estimated relative to a horizontal reference line (identified by a dashed line in [Fig. 4]). The height H can be expressed, for example, as a number of pixels and / or as a proportion of the total height of the first image (il). Then, in substep E332, the distance between the first vehicle V1 and the second vehicle V2 is calculated as a function of the height H in the first image (il). A nomogram such as the one shown in [Fig. 7] can be used for this purpose: the x-axis represents the height of the second vehicle V2 in the first image (il), expressed in pixels. The y-axis represents the distance, in meters, between the first vehicle (VI) and the second vehicle V2. The higher the second vehicle V2 appears in the first image (il), the greater the distance between the first vehicle (VI) and the second vehicle (V2).Of course, such a nomogram is more accurate if the video recording 6 was made on a perfectly horizontal road. Therefore, video recordings made on horizontal roads should preferably be used. A numerical function x=f(y) that best characterizes the dashed curve Cl in [Fig.7] can be used to quickly calculate the distance between the first vehicle and the second vehicle. The function x=f(y) can be formulated as follows: . .6 (y — k) b

[0053] Where a, b, k and H are parameters that can be determined by learning and / or artificial intelligence techniques called curve fitting.

[0054] The distance thus estimated can then be compared to the thresholds previously mentioned, in particular the thresholds of 200m and 400m previously mentioned, in order to confirm the performance of the control software 33.

[0055] Finally, thanks to the invention, we have a method for evaluating a control software 33 for an adaptive headlight that is dematerialized, that is to say, whose repetition requires no testing on a motor vehicle. This evaluation method makes it possible to quickly simulate a very large number of scenarios and requires only a computer with sufficient computing power. The method makes it possible to identify scenarios for which the control software fails to perform appropriate detection. The control software can thus be more easily validated and / or improved if necessary.

Claims

Demands

1. Method for evaluating a control software (33) for an adaptive headlight (2) for a first motor vehicle (1), the first vehicle comprising said adaptive headlight and a camera (4) arranged to observe a scene in front of the first vehicle, the control software (33) to be evaluated comprising: - a detection algorithm (35) adapted to verify the presence of at least a second vehicle (V2) in front of the first vehicle by processing images provided by said camera (4), the detection algorithm being capable of calculating a first verification result, and - an adaptation module (36) for a lighting beam emitted by the adaptive headlight as a function of said first verification result, the evaluation method comprising: - a first step (E1) of providing a video recording (6) and the control software (33) to be evaluated,The video recording, comprising a predefined duration and depicting a scene in front of the first vehicle on which the second vehicle appears, then - a second step (E2) of execution of the piloting software to be evaluated, the detection algorithm being carried out by substituting the images provided by the camera with images from the video recording, then - a third step (E3) of evaluation of the piloting software based on the first verification result obtained at the end of the second step (E2), during the execution of the piloting software to be evaluated.

2. Evaluation method according to the preceding claim, characterized in that the third step (E3) comprises: - a first substep (E31) of automatic verification of the presence of the second vehicle (V2) in the video recording by processing the images of the video recording, the first substep leading to a second verification result, then - a second substep (E32) of comparison of the first verification result with the second verification result.

3. Evaluation method according to the preceding claim, characterized in that the first substep (E31) comprises: - a substep (E311) of identifying a first image (11) of said video recording at the moment when the first vehicle (1) detects the presence of the second vehicle (V2) during the execution of the driving software to be evaluated, and - a substep (E312) of identifying the second vehicle on the first image.

4. Evaluation method according to any one of the preceding claims, characterized in that the third step (E3) includes a substep (E33) of estimating the distance separating the first vehicle (1) from the second vehicle (V2) at a time when the first vehicle detects the presence of the second vehicle during the execution of the piloting software (33) to be evaluated.

5. Evaluation method according to claim 3 and according to claim 4, characterized in that said substep (E33) of estimating the distance separating the first vehicle from the second vehicle comprises an estimation (E331) of a height (H) of said second vehicle on said first image (il), then a calculation (E332) of the distance separating the first vehicle from the second vehicle as a function of the height of said second vehicle on said first image.

6. Evaluation method according to claim 3 or 5, characterized in that said substep of identifying the second vehicle (V2) on the first image (il) comprises: - a step of identifying the second vehicle on a second image (i2) of the video recording (6), the second image being later than the first image (il) in the chronological order of the video recording, then - the identification of the second vehicle (V2) on the first image (il) of the video recording, by reverse tracking of the second vehicle on at least a part of the images separating the second image from the first image in the video recording.

7. Evaluation method according to the preceding claim, characterized in that the step of identifying the second vehicle on the second image (i2) of the video recording (6) comprises - the calculation of a ratio R given by the following formula: R = intersection (Ax, By) / union (Ax, By) where: Ax designates a first area of ​​the second image, the first area being determined by means of a first algorithm, in particular a shape detection algorithm, By designates a second area of ​​the second image, the second area being determined by means of a second algorithm, in particular a light intensity detection algorithm, then - the comparison of the ratio R with a predefined threshold.

8. Evaluation method according to claim 6 or 7, characterized in that the second image (i2) is selected from a subset of the images of the video recording (6) by calculating a score for each of the images of said subset.

9. Evaluation method of the preceding claim, characterized in that said score is calculated for a given image by means of the following formula: S = max (intersection (Ax, By) / union (Ax, By)) where: Ax designates a first area of ​​the given image, the first area being determined by means of a first algorithm, in particular a shape detection algorithm, By designates a second area of ​​the given image, the second area being determined by means of a second algorithm, in particular a light intensity detection algorithm.

10. Evaluation device (5) of a control software (33) intended to control an adaptive light projector for a motor vehicle, the evaluation device comprising software and hardware means (51, 52) adapted to implement the evaluation method according to one of the preceding claims.

Images