Method for calibrating onboard cameras and enriching a calibration database shared by several vehicles

Abstract

Method for calibrating on-board cameras and enriching a calibration database shared by several vehicles.A camera calibration method (2) comprising: - prior to calibrating a camera (2) of a vehicle (1), selecting at least one suitable area for calibration based on: - the location of the vehicle (1) and / or a route entered into the navigation system (7) and - information regarding the possibility of self-calibrating the camera (2) based on the environment of the vehicle (1) as observed by the camera (2), - acquiring images necessary for the self-calibration of the camera when the vehicle (1) is present in the area for calibration (2), - enriching and / or updating the database at least when the result of the self-calibration is one of success and one of failure, at least with information regarding the result of the self-calibration in said area. Figure for the abstract: Fig. 6.

Description

Title of the invention: Method for calibrating on-board cameras and enriching a calibration database shared by several vehicles technical field

[0001] The present invention relates to autonomous vehicles or vehicles offering a possibility of assisted driving, equipped with cameras which are arranged to observe the road and the environment of the vehicle and provide the necessary data to the on-board computer which manages the driving and / or assists the driver.

[0002] The invention relates more particularly to a method of calibrating cameras on board the vehicle.

[0003] The term “vehicle” means land vehicles, aircraft and boats.

[0004] The term “camera” encompasses digital image acquisition devices operating in the visible or non-visible spectrum, of any resolution and technology. Previous technique

[0005] Cameras are sensors widely used in vehicles for Advanced Driver Assistance Systems (ADAS), parking, infotainment, and multimedia applications in general. In order to provide a correct image, each camera requires calibration of various intrinsic and extrinsic parameters.

[0006] The intrinsic parameters are internal to the camera and include in particular the focal length, the image magnification factors or "aspect ratio", the distortion parameter ("skew factor") and the coordinates of the projection of the optical center of the camera onto the image plane ("principal point").

[0007] The extrinsic parameters define the positioning of a reference frame linked to the camera with respect to another given reference frame, and include in particular the rotation and translation components to go from one reference frame to the other.

[0008] A study relating to camera self-calibration was published by Olivier Faugeras, QT Luong and Stephen J. Maybank in European Conference on Computer Vision, 19 May 1992

[0009] A conventional initial calibration process is generally implemented in the factory at the end of production. This calibration can conventionally be carried out using targets as described in publications US 9 978 146 B2 and CN 107 133 988 B.

[0010] It is also known to perform calibration by driving the vehicle equipped with the camera to be calibrated along a route intended for this purpose, as described in US 11,265,514 B2.

[0011] It is often necessary to perform calibration during the vehicle's lifetime Automatic or auto-calibration of cameras. This method consists of determining various camera parameters directly from several uncalibrated images and unstructured scenes, notably without test patterns or other special objects, unlike the classic initial calibration process defined above.

[0012] US publications 9 201 425 Bl, JP 6458439 B2, WO 2019 / 233330, and US 110 554 951 B2 describe various methods of self-calibrating motor vehicle cameras.

[0013] US patent 10,970,878 B2 discloses a method for automatically calibrating the cameras of a motor vehicle based on the use of a reference map. A calibration model is generated from the comparison between certain location features identified on the one hand in images taken by the vehicle's cameras and on the other hand as recorded on the reference map, knowing the vehicle's location.

[0014] This solution is not entirely satisfactory since it requires the use of a relatively large high-definition map (or "HDmap"). Here, "high-definition map" refers to a map that includes significantly more precise information than a traditional map, such as the number of road lanes, road markings, and the types and positions of traffic signs. Furthermore, when this map is stored on a remote server, this method may require the transfer of a relatively large volume of data, which is a constraint on the bandwidth of the associated communication system. Summary of the invention

[0015] The invention aims to facilitate the self-calibration of vehicle cameras and achieves this, according to one of its aspects, through a method of calibrating on-board cameras and enriching a calibration database shared by several vehicles, each equipped with at least one on-board camera to be calibrated, a navigation system enabling vehicle location, and a map in memory and / or access to this map, the latter listing areas accessible to the vehicle and associated location characteristics likely to be useful for calibrating the on-board camera, a method comprising: - Prior to calibrating a vehicle camera, the selection of at least one suitable area for calibration from: - the location of the vehicle and / or a route for it entered into the navigation system and - information regarding the possibility of the camera self-calibrating based on the vehicle's environment as observed by the camera, this information being provided by the calibration database and / or derived from the card,

[0016] - the acquisition of images necessary for the camera's self-calibration when the The vehicle is present in the area designated for calibration.

[0017] - the enrichment and / or updating of the database at least when the The result of the autocalibration is one of success and one of failure, at least with information relating to the result of the autocalibration in said candidate area.

[0018] By "accessible area," it should be understood that the area is accessible by the vehicle, and can in particular be reached or traversed by it. In particular, the set of accessible areas of a given terrestrial region forms a convex subregion.

[0019] A "suitable calibration zone" is defined as an area with specific location characteristics observable by the camera, allowing for successful camera self-calibration. An area is considered more suitable for calibration the higher the probability of successful calibration.

[0020] Thanks to the invention, the data streams required for self-calibration can remain relatively small.

[0021] The use of the shared database according to the invention leads to an efficient and relatively inexpensive self-calibration method. The number of images to be processed in order to calibrate the camera can be reduced due to the quality of the specific location features present in the area, as well as the associated algorithmic complexity.

[0022] The method according to the invention can also make it possible to plan the execution of the different associated tasks in time in order to reduce the computational load related to self-calibration and the possible additional costs related to the amount of computer memory required.

[0023] Enriching the database allows us to obtain, over time, increasingly precise, reliable and exhaustive data useful for self-calibration, and thus to further improve calibration performance.

[0024] The invention makes it possible to efficiently track and record information representative of the ability of a given area to allow correct calibration of the camera.

[0025] The invention can in particular allow calibration performance to be comparable to that which would be obtained in the factory.

[0026] In addition, replacing factory calibration with self-calibration reduces the time and costs of vehicle production in the factory and reduces the need for maintenance.

[0027] Furthermore, the invention does not require the use of high-definition mapping, and thus a saving can be achieved in terms of the overall cost of the system.

[0028] The need for camera calibration can be previously identified from the detection of a defect or distortion in at least one image provided by it, and / or from a command by a vehicle user or an operator responsible for its maintenance.

[0029] The aforementioned candidate area may already be listed in the database, the latter then including a corresponding label and / or a score representing the self-calibration success rate associated with the area.

[0030] In this way, the continuous updating of a zone's score by the various vehicles that enter it provides a precise and reliable dynamic indicator of the zone's capacity for successful self-calibration. It is thus possible to update the score based on the results of the different vehicles' self-calibration attempts, notably by incrementing or decrementing it, particularly by a predefined value, especially a constant one, for each successful or unsuccessful attempt, or by a predefined value determined by a law that makes the increment value dependent on the number of evaluations already performed by the different vehicles.

[0031] The score update can be performed, in particular, based on a number of consecutive positive or negative evaluations of the candidate area, for example, when several consecutive evaluations are either positive or negative. In this way, it is possible to dynamically amplify the trend of the evaluations of said area and to converge more quickly towards a score representative of its actual performance.

[0032] The score can be updated in real time or with a delay. For example, the vehicle can store information regarding the success or failure of the self-calibration attempt and subsequently enrich the database with the result, for example, when conditions for a lower-cost data transfer are met. Alternatively, the score can be updated at approximately the same time as the success or failure of the self-calibration attempt is detected.

[0033] The selected candidate area may be the one with the highest score.

[0034] The score value in the database can be initialized before returning vehicle experience, based on a probability of successful calibration relative to said usable area, calculated from the map.

[0035] The candidate area may not be listed in the database, the selection of the candidate area being carried out from a probability of success of the self-calibration obtained from the map or calculated from it.

[0036] The selected candidate area may be the one with the highest probability of success.

[0037] The probability of successful calibration can thus be calculated from location characteristics suitable for calibration identified on the map, in particular the presence of linear road segments, multi-lane road segments, segments roads near urban areas or city centers, traffic signs, road markings, peripheral infrastructure elements such as bus stops, telephone booths or fire hydrants, and / or the slope value of a road segment.

[0038] Alternatively, the probability of successful calibration is already provided in the card and extracted from it without calculation.

[0039] The probability of successful calibration can be refined using dynamic information, such as available parking spaces with markings at a given time (for example, outside of store opening hours, the parking lot is likely to be sparsely occupied, and vice versa), and / or real-time geolocated traffic information, as heavy traffic can reduce the visibility of the markings. This improves the accuracy of the calibration probability estimate at a given time.

[0040] The map and / or the probability of successful calibration can also be updated, where appropriate, based on new location characteristics suitable for calibration identified on images provided by at least one camera of a vehicle moving in said area.

[0041] When calibration is successful, the vehicle can communicate a success notification to a database update server for the area where calibration was performed.

[0042] When calibration fails, the vehicle can communicate a failure notification for that area to a database update server.

[0043] The database can list at least one label identifying areas recognized as suitable for calibration, or vice versa.

[0044] When calibration fails and the score associated with said zone falls below a predefined threshold, a failure label can be recorded and / or updated in the database.

[0045] The presence of a label in the database makes it possible to effectively distinguish between areas suitable for calibration and areas not suitable for calibration. It is then possible, for example, to filter the data relating to areas during the search phase so as to consider only those areas suitable for calibration.

[0046] When reliable means of communication exist between the vehicle and a remote server through which the database is accessed, it is not necessary to duplicate all or part of the database in the vehicle's memory. However, to facilitate database access and reduce access times, the database may, if necessary, be duplicated at least partially in the vehicle's memory, and the enrichment of said database may be carried out later, for example by a merge operation.

[0047] The map can be low or medium resolution.

[0048] The invention also relates to a vehicle equipped with at least one on-board camera, a navigation system enabling the location of the vehicle and a map in memory and / or access to this map, the latter listing areas accessible by the vehicle and associated location characteristics that may be useful for calibrating the on-board camera, the vehicle having access to a calibration database shared by several vehicles, access to the database being via a means of communication, the vehicle further comprising a computer program product including code instructions which, when executed on a computer system of the vehicle, allow the implementation of the method according to the invention as defined above.

[0049] Such a computer program product can make it possible to implement the process with all or part of the process characteristics defined above. Brief description of the drawings

[0050] The invention will be better understood upon reading the detailed description that follows, the non-limiting examples of embodiments thereof, and upon examination of the accompanying drawing, on which:

[0051] [Fig-1] Fig. 1 illustrates, schematically and partially, an example equipment for a vehicle adapted to implement the process according to the invention.

[0052] [Fig.2] Fig.2 illustrates, schematically and partially, an example architecture of a system adapted to implement the process according to the invention.

[0053] [Fig.3] Fig.3 is a schematic representation of an example of data relating to accessible areas as they could be listed in the database.

[0054] [Fig.4] Fig.4 is a schematic representation of an example of data relating to accessible areas as they might be listed on the map.

[0055] [Fig. 5] Fig. 5 is a block diagram illustrating steps in an example of calibration method according to the invention.

[0056] [Fig. 6] Fig. 6 illustrates steps in an example calibration process and enrichment of the shared database from an identification of the area from the map.

[0057] [Fig.7] Fig.7 illustrates steps in an example of a self-calibration process and enrichment of the calibration database from an identification of listed areas.

[0058] [Fig.8] Fig.8 is a schematic representation of an example display on a user interface. Detailed description Vehicle

[0059] Figure [1] illustrates an example of a motor vehicle in top view. having equipment adapted to the implementation of the process according to the invention.

[0060] Vehicle 1 can be thermal, for example of petrol, diesel, gas, hydrogen, or hybrid type, or even electric.

[0061] The vehicle 1 may include various sensors, in particular at least one camera 2.

[0062] The camera 2 can be mounted on the roof, on a side, at the front, at the rear, outside or inside the vehicle 1, including behind the windshield.

[0063] The vehicle 1 may optionally include, as illustrated, at least one radar 3 enabling precise measurement of the distances of objects present in the vicinity of the vehicle 1.

[0064] In the example considered, vehicle 1 has side radars 3 left and right, a front radar 4.

[0065] Vehicle 1 is also equipped with a computer 5, which receives data from the camera 2 and from any radars 3,4.

[0066] The computer 5 includes one or more processors executing one or more programs enabling the implementation of the method according to the invention.

[0067] In particular, the computer 5 can control the various tasks necessary to carry out the self-calibration of the camera 2 at least.

[0068] The computer 5 may be composed of all or part of an embedded computer system comprising, for example, in addition to the aforementioned processor(s), at least one RAM memory, at least one ROM memory used to save the applications of the vehicle 1, one or more optional digital-to-analog converters as well as one or more input / output interfaces used to communicate with the various sensors.

[0069] The computer 5 can access an on-board memory 6 storing a digital map allowing it to provide information relating to the environment of the vehicle 1. The map can in particular allow the identification of the different areas accessible 33 by the vehicle.

[0070] The map can be of variable resolution, of road, topographic, orthophotographic, satellite, nautical, aeronautical type, or even of hybrid type by combining several types of views.

[0071] The map may include two-dimensional or even three-dimensional data, and in particular the locations (longitude and latitude) of practicable paths in a given space.

[0072] The map may in particular include additional information relating to certain specific areas or locations, such as the presence and / or form of infrastructure, characteristic places in the environment (for example, historical monuments or rivers).

[0073] The vehicle 1 further includes a navigation and location system 7, preferably with a GNSS receiver, for example of the GPS type, enabling in particular the driver to enter a route for the vehicle 1 and / or the latter to know its position in real time. Self-calibration system

[0074] Figure 2 shows a self-calibration system 10 for implementing the invention, comprising a set of means on board the vehicle and a set of remote means.

[0075] By "remote means", it is understood that these means are not physically present on board the vehicle 1, and that access to them is via a means of communication 13 of any type, for example, a 4G or 5G network or other.

[0076] The calculator 5 can be connected to any type of interface 31 allowing information to be presented to the user of the vehicle 1.

[0077] The latter can use the camera 2 to capture a number of images or videos of the environment and can use the localization system 7 to collect location data associated with each image or video captured.

[0078] As illustrated in [Fig.2], the set of remote means may include a remote server, in particular “cloud” 16.

[0079] The shared calibration database can be stored on this server 16.

[0080] Server 16 is accessible by a set of several vehicles 1 according to the invention, the connection of each of the vehicles to the server 16 can be carried out via a secure protocol.

[0081] The database includes at least one organized collection of structured data relating to different usable areas.

[0082] The database can be hierarchical, network, object-oriented, relational, non-relational or NoSQL.

[0083] The database can be controlled by any database management system.

[0084] The data contained in the database can be encrypted where appropriate.

[0085] Different types of access to the database can be defined, including administrator access or restricted access, such as only the ability to view and read data, or only the ability to modify or add data. Database

[0086] Fig. 3 illustrates an example of a sample of 20 data relating to several practicable areas listed in the database.

[0087] Each row of data in the database relates to a usable area 33, and includes For example, as illustrated, its location 21, a score 22 and the number of evaluations 23.

[0088] The location 21 of an area can be defined by a first latitude coordinate and a second longitude coordinate.

[0089] Alternatively, the location 21 of an area can be defined by the convex hull of a plurality of points, each point being defined by a first latitude coordinate and a second longitude coordinate.

[0090] Alternatively, the location 21 of an area can be defined by a circle of radius R and the coordinates of the center of this circle.

[0091] The 22 score of a zone is a measure representing the actual success rate of self-calibration in that zone. This 22 score is a positive number between 0 and a strictly positive number V, for example between 0 and 1. The higher the 22 score, and therefore here closer to 1, the more successful the self-calibrations have been in practice in the zone under consideration.

[0092] A score of 22 not entered in the database means that the data is not available, for example that it has not yet been initialized after one or more auto-calibration attempts.

[0093] The score 22 is modified according to the result of the self-calibration attempts, in particular increased after a successful self-calibration, and decreased otherwise.

[0094] The score update 22 is for example to be carried out by a constant or variable increment, and / or can take as a parameter the number of previous evaluations 23, and / or the results of the most recent self-calibration attempts.

[0095] The number of evaluations 23 corresponds to the number of updates to the score 22 associated by the different vehicles 1. Thus, after a vehicle 1 performs a self-calibration attempt in a given area, it sends the server 16 corresponding information in the form of a failure or success notification, the associated score 22 is updated, and the number of evaluations 23 is incremented.

[0096] In the example of [Fig.3], the first line of data describes an area particularly suitable for calibration, its score 22 being close to 1, and the number of evaluations 23 is relatively large.

[0097] The third line also describes an area particularly well suited to calibration, the score 22 being close to 1, but less certainly in view of the relatively small number of evaluations 23.

[0098] Each data line may also include a label indicating the suitability of the zone for calibration, for example OK or NOK, respectively indicating that the associated zone is suitable for calibration, or not suitable.

[0099] Labels can be useful for providing summary information on capacity of a given area to allow in practice a successful self-calibration.

[0100] The initialization of the label of a zone can be carried out following a first attempt at self-calibration by a vehicle that has gone to that zone.

[0101] For example, after a first successful attempt, the label can be initialized to OK.

[0102] Similarly, after a first failed attempt, the label can be initialized to NOK. Map data

[0103] Fig. 4 illustrates an example of a sample of data 24 relating to several practicable areas 33 taken from the map.

[0104] Each line of data from the map is for example related to a usable area 33, and includes for example, as illustrated, its location 21, a probability of success 25, as well as a list of location characteristics 26 or “features”.

[0105] The probability of success 25 is, for example, a positive number between 0 and a strictly positive number V, for example between 0 and 1

[0106] The higher the probability of success 25, and therefore closer to 1, the more the area considered theoretically gives rise to successful self-calibrations.

[0107] The probability of success 25 of an area can be calculated a priori, even before a vehicle goes to the area and makes an attempt at self-calibration, from the location information 26 provided by the map.

[0108] Alternatively, the probability of success 25 is calculated a posteriori, after the self-calibration attempt has taken place.

[0109] The probability of success 25 is established from the number and nature of the different location characteristics suitable for calibration 26 present in the area. Some characteristics 26 may be more suitable than others for leading to successful calibration.

[0110] The characteristics 26 of an area can be directly extracted from the map by the calculator 5.

[0111] On [Fig.4], the fourth and fifth lines correspond to two areas where no self-calibration attempt has yet been made by a vehicle 1. The fourth area seems, however, particularly suitable for calibration in view of the probability of success 25 established from the location characteristics 26 present, unlike the fifth area. Calibration process

[0112] An example of a camera 2 calibration method will now be described with regard to [Fig.5].

[0113] Vehicle 1 is considered to have at least one camera 2 to be recalibrated and launches In step 100, a search mode for a suitable area for calibration.

[0114] The location of vehicle 1 is recorded by the computer 5 using the location system 7.

[0115] The computer 5 accesses the shared database via the communication means 13 in order to extract the coordinates 21 of at least one area close to the current position of the vehicle 1 or its route, suitable for calibration. An area can be considered close if it is located at a distance less than a predefined value, for example 5 km.

[0116] To do this, it can send a request to server 16 to select the nearest area with an OK success label along vehicle 1's route. Card identification process

[0117] If no area suitable for calibration is returned after consulting the shared database in step 50, a card identification process is initiated in step 101.

[0118] This card identification process 101 illustrated in [Fig.6] allows the identification of an area suitable for calibration from the knowledge of a probability of success 25 associated with this area, obtained from the card.

[0119] At step 103 the calculator 5 accesses the map and calculates at least one probability of success 25 relating to an area whose coordinates it knows 21.

[0120] This calculation can be carried out by an algorithm which defines for a given area the probability of success 25 from its location characteristics 26 identified on the map.

[0121] For example, an area with a large number of place features 26 guaranteeing the presence of many vertical and horizontal lines, and / or parallel lines and / or regular spacing, will lead to a higher probability of success 25 than an area with place features containing, for example, only a few such lines or other geometric patterns of interest for calibration.

[0122] The probability calculation law can be determined empirically by making trials and / or from a mathematical model.

[0123] In this way, the probabilities can be calculated for a predefined number of zones, located on the route or within a given radius around vehicle 1 during the search phase.

[0124] The selection of the zone can then be carried out in step 104 by the calculator 5. This selection can consist, for example, of taking the one with the best probability of success 25, or with the best ratio of probability of success 25 to distance to travel to reach it.

[0125] A filter can be applied where appropriate when selecting the area to take into account the route of vehicle 1 entered into the navigation system 7, and for example exclude areas whose access by vehicle 1 would cause a detour of a distance greater than a predefined value.

[0126] When vehicle 1 goes to the selected area, the computer 5 executes in step 105 a calibration algorithm which generates in a way known per se a calibration model from the images taken by the camera 2. The images taken by the latter advantageously include at least part of the location characteristics 26 of the area, identified a priori on the map.

[0127] If the self-calibration attempt is successful, a success notification is sent by vehicle 1 to server 16 via communication means 13 for updating and / or enriching the database.

[0128] If the auto-calibration attempt fails, a failure notification is communicated and search mode 100 is initiated again.

[0129] Finally, the card identification process ends with a step 106 of sharing the result and associated data.

[0130] Step 106 may include recording in the shared database the coordinates 21 of the area.

[0131] The score 22 can be initialized where appropriate by taking as value the probability of success 25 calculated by vehicle 1 during this process. Identification process by listed areas

[0132] Returning to [Fig.5], we see that if at least one area close to vehicle 1 and suitable for calibration is listed in the database and received by vehicle 1 in return for querying server 16 at step 50, an identification process by listed areas 102 as illustrated in [Fig.7] is initiated.

[0133] Such a process allows the identification of an area suitable for calibration from the labels and / or scores 22, and / or the numbers of evaluations 23 associated.

[0134] When identification by listed zones 102 is initiated, a zone is selected as a target by the computer 5 in step 107. The selection of a zone can be carried out by a selection algorithm that determines an optimal zone from a set of given zones, including locations 21, scores 22, and associated numbers of evaluations 23. For example, the selection algorithm is configured to select from a set of zones the one with the best score 22, or with the best performance ratio, the performance ratio of a given zone being defined, for example, by a function taking as parameters its score 22, its number of evaluations 23, and its relative distance to the vehicle 1.The set of zones considered by the selection algorithm can be filtered taking into account the route of vehicle 1 entered into the navigation system 7, in particular by excluding zones whose access by vehicle 1 would entail a detour of a distance greater than a predefined value.

[0135] When vehicle 1 then moves to the area selected in step 105, the computer 5 executes a calibration algorithm that generates a calibration model at starting from the images taken by camera 2, as described above.

[0136] The result of the self-calibration attempt is evaluated in step 60.

[0137] If the self-calibration attempt is successful, a success notification is communicated by vehicle 1 to server 16 via communication means 13 at step 110.

[0138] The score 22 associated with the zone can then be increased by server 16.

[0139] Similarly, if the self-calibration attempt fails, a notification The failure is communicated by vehicle 1 to server 16 at step 111.

[0140] The score 22 associated with the zone can be decreased by server 16.

[0141] If the score 22 reaches a predefined minimum threshold after the update, a label NOK is also registered for the area.

[0142] In the event of a failed attempt at self-calibration, the candidate area search process can be restarted, and we resume at step 100, searching for a new candidate area. Display of areas suitable for calibration

[0143] The self-calibration process can take place without notifying the vehicle user 1.

[0144] Alternatively, the user may be notified, and may agree to go to the selected area, or even drive vehicle 1 into that area.

[0145] In all cases, the method may include displaying on a plan the location 21 of the area or areas suitable for calibration.

[0146] Fig. 8 illustrates an example of displaying 30 on a screen 31 the location of different areas suitable for calibration 35, listed in the database and / or identified from the map.

[0147] A marker 32 materializes the position of the vehicle 1 within the practicable areas 33 and non-practicable areas 34.

[0148] Where appropriate, the suitable areas 35 are displayed with specific colours depending, for example, on the chances of successfully calibrating them. Variants

[0149] Of course, the invention is not limited to the examples just described.

[0150] In one variant, the database is duplicated at least partially in the memory 6 of vehicle 1, and the enrichment of said database is carried out asynchronously, for example by a merge operation. The database can be duplicated periodically from the version stored on server 16.

[0151] In one variant, the card is stored on the server 16; access to the card is then via the means of communication 13.

[0152] In one variant, at least a probability of success 25 is already specified on the map, and this is then extracted without calculation.

[0153] It is possible, if desired, to reassess the probability of success 25 of the calibration, to take into account a difference between the location characteristics 26 actually observed in the area and those listed on the map.

[0154] The probabilities of success 25 can be calculated in the background, outside of an activation of the map area identification process as described above.

[0155] At least part of the calculations relating to the generation of the calibration model can be carried out on the server 16. The results can then be transmitted to the computer 5 via the communication means 13.

[0156] The types of place characteristics 26 identified on the map can be predefined. Conversely, their number can increase as new types of relevant place characteristics 26 become identifiable.

[0157] Although the invention is primarily applicable to the calibration of cameras 2 of motor vehicles, it extends to other types of vehicles such as ships and aircraft. In the latter case, the self-calibration is carried out in areas sufficiently close to the ground, or even on the ground, for example during driving or parking.

Claims

Demands

1. A method for calibrating on-board cameras (2) and enriching a calibration database shared by several vehicles (1), each equipped with at least one on-board camera (2) to be calibrated, a navigation system (7) enabling the location of the vehicle (1), and a map in memory (6) and / or access to this map, the latter listing areas (33) accessible to the vehicle (1) and associated location characteristics (26) that may be useful for calibrating the on-board camera (2), a method comprising: - prior to calibrating a camera (2) of a vehicle (1),the selection of at least one suitable area (33) for calibration based on: - the location (21) of the vehicle (1) and / or a route entered into the navigation system (7) and - information regarding the possibility of self-calibration of the camera (2) based on the environment of the vehicle (1) as observed by the camera (2), this information being provided by the calibration database and / or from the map, - the acquisition of images necessary for the self-calibration of the camera when the vehicle (1) is present in the area for calibration (2), - the enrichment and / or updating of the database at least when the result of the self-calibration is one of success and one of failure, at least with information relating to the result of the self-calibration in said area.

2. A method according to claim 1, wherein the need for calibration of the camera (2) is previously identified from the detection of a defect or distortion in at least one image provided by it, and / or from a command from a user of the vehicle (1) or an operator responsible for its maintenance.

3. A method according to any one of claims 1 and 2, wherein the candidate area is already listed in the database, the latter then comprising a corresponding label and / or a score (22) representative of the self-calibration success rate associated with the area.

4. A method according to claim 3 wherein the score (22) is updated based on the results of the self-calibration attempts of the different vehicles (1), in particular by incrementing or decrementing it, by particular of a predefined value, in particular constant, at each successful or unsuccessful attempt, or of a predefined value given by a law making the increment value dependent on the number of evaluations (23) already carried out by the different vehicles (1).

5. A method according to any one of claims 3 and 4, wherein the score update (22) is performed based on a number of consecutively positive or negative evaluations (23) of the candidate area.

6. A method according to any one of the three preceding claims, wherein the selected candidate area is the one having the highest score (22).

7. A method according to any one of claims 1 and 2, wherein the candidate area is not listed in the database, the selection of the candidate area being carried out based on a probability of success (25) of the self-calibration (25) obtained from the map or calculated from it.

8. Cl. Method according to claim 7, wherein the selected candidate area is the one having the highest probability (25) of success.

9. A method according to any one of claims 7 and 8, wherein the probability of success (25) is calculated from location characteristics suitable for calibration (26) identified on the map, including the presence of linear road segments, multi-lane road segments, road segments near an urban area or city center, traffic signs, road markings, peripheral infrastructure elements such as bus stops, telephone booths or fire hydrants, and / or a slope value of a road segment.

10. Method according to claims 7 and 8, wherein the probability of success (25) of calibration is already provided in the card and extracted from it without calculation.

11. A method according to any one of claims 7 to 10, wherein the probability of success (25) is refined from dynamic information, including available parking spaces with ground markings at a given time and / or real-time geolocated traffic information.

12. A method according to any one of the preceding claims, wherein when calibration is successful, the vehicle communicates to a database update server (16) a success notification for the area where calibration was performed.

13. Vehicle (1) equipped with at least one on-board camera (2), a system navigation (7) enabling the location of the vehicle (1) and a map in memory (6) and / or access to this map, the latter listing areas accessible to the vehicle (1) and associated location characteristics (26) that may be useful for calibrating the on-board camera (2), the vehicle (1) having access to a calibration database shared by several vehicles (1), access to the database being via a means of communication (13), the vehicle further comprising a computer program product configured to: - prior to calibrating a camera (2) of a vehicle (1), select at least one usable area (33) suitable for calibration, starting from at least: • the location (21) of the vehicle (1) and / or a route thereof (1) entered into the navigation system (7) and • at least one piece of information relating to the possibility of self-calibration of the camera (2) from the environment of the vehicle (1) as observed by the camera (2), this information being provided by the calibration database and / or from the map, - when the vehicle (1) is present in the candidate area for calibration, proceed with the acquisition of images necessary to camera self-calibration (2), - enrich or update the database at least when the result of the autocalibration is one of a success and one of a failure, at least with information relating to the result of the autocalibration in said candidate area.

Images