METHOD FOR THE DYNAMIC ESTIMATING OF THE PITCH AND ROLL MOTION OF A MOTOR VEHICLE USING AT LEAST ONE IMAGE CAPTURE SENSOR

Description

[Technical domain]

[0001] This patent application relates to the field of driver assistance systems using a video camera-type image acquisition sensor, and more particularly to the optimization of the acquisition of information relating to the environment of a motor vehicle by a process of dynamic estimation of the pitch and roll of a motor vehicle by means of at least one image acquisition sensor which is mounted on the vehicle. [Etat de la technique antérieure]

[0002] Advanced Driver Assistance Systems (ADAS) are becoming increasingly common in vehicles. They offer a growing number of features designed to improve the comfort and safety of vehicle users. Ultimately, the production of fully autonomous vehicles is a possibility.

[0003] Many of the features offered by such systems rely on real-time analysis of the immediate environment of the vehicle in which they are installed. Whether it's assisting with emergency braking or improving a car's trajectory control, constantly taking into account the characteristics of the vehicle's environment is necessary to ensure the functionality of driver assistance systems.

[0004] In this context, one way to acquire information about a vehicle's immediate surroundings is to use image acquisition and analysis tools. A vehicle can be equipped with one or more video cameras designed to observe its environment at all times, in one or more directions. The images collected by the cameras are then processed by the driver assistance system, which can use them to potentially trigger a driver assistance action, such as activating an audible and / or visual alert, or even triggering braking or a change of trajectory.

[0005] This type of operation assumes that the images collected by the camera(s) allow for the reliable deduction of the desired relevant information about the vehicle's environment, and with the lowest possible computational cost. .This notably implies that the acquired images are truly representative of the observed reality and free from any degradation that could distort the driver assistance system's perception of the environment. It also implies, for certain functionalities, that the driver assistance system must be able to correctly locate the various objects present in the three-dimensional space in which the vehicle is moving, based on the images acquired by the camera.

[0006] This is the case, for example, in the classic situation where the driver assistance system uses images acquired by a camera mounted on the front windshield of a vehicle, near its rearview mirror. This use presupposes, for certain functionalities, knowing the relationship between the spatial coordinates of a point in space and the corresponding point in the image acquired by the camera, which in turn requires knowing the spatial relationship between the camera and the vehicle. To ensure that these relationships are correctly understood, it is common practice to calibrate the camera, more or less regularly. This calibration aims to determine parameters of the camera, known as intrinsic and extrinsic parameters.

[0007] Intrinsic parameters are linked to the camera's specific technical characteristics, such as its focal length, the position of its optical center, or its distortion model. Calibration allows for the estimation of these parameters and is therefore performed at the factory at the end of the camera's production line.

[0008] The extrinsic parameters, on the other hand, correspond to the positioning and orientation of the camera relative to a world frame of reference, assumed to be parallel to the road, subject to any potential tilt of the vehicle relative to the road on which it is traveling. These parameters include the three translations and three rotations necessary to go from the coordinates of a point, expressed in a frame of reference defined above and called the "world frame of reference," to the coordinates of that point expressed in a frame of reference linked to the camera, called the "camera frame of reference." Calibrating these parameters (six in total) then allows the camera's positioning and orientation relative to the road to be estimated.Furthermore, since these parameters are likely to vary more frequently than intrinsic parameters, for example in the event of a change in vehicle load, the calibration of extrinsic parameters can be carried out relatively regularly in order to continuously adjust their values ​​to the observed situation.

[0009] The concept of "calibration" often refers to a variable that varies little and determines an "absolute" angle between the camera and the world reference frame. This is an average, or a "nominal" value, that does not take into account camera oscillations. A distinction is made between end-of-line vehicle calibration, which uses a production target, and "online" calibration, which relies on the camera's optical flow.

[0010] Finally, the purpose of camera calibration is to facilitate the use of images acquired by the camera in order to improve the spatial localization of the various objects observed.

[0011] Despite this, no matter how regularly the calibration can be performed, it does not allow tracking of the dynamic pitch and roll that generate significant dynamic changes in camera orientation.

[0012] Camera pitch is defined as an angular movement of the camera around a transverse axis perpendicular to the longitudinal trajectory of the motor vehicle, and roll is defined as an angular movement of the camera around a longitudinal axis parallel to the longitudinal trajectory of the motor vehicle; their value is estimated using the optical flow of the camera.

[0013] The concept of dynamic pitch and roll represents the same physical quantity as calibration, that is, the angle of the camera relative to a reference frame, generally the "world" reference frame.

[0014] However, dynamic pitch and roll do not represent an average or nominal values ​​but an instantaneous value that oscillates around the calibration value.

[0015] Typically, in situations such as going over a speed bump, during significant acceleration or deceleration, or when experiencing a sudden change in gradient, the instantaneous pitch value may temporarily deviate significantly from its calibrated value. Similarly, when the vehicle travels over a speed bump or pothole "asymmetrically"—that is, when only the wheels on one side of the vehicle cross the speed bump or pothole—or in the case of a sharp turn, the instantaneous roll value may temporarily deviate significantly from its calibrated value.

[0016] Accurate knowledge of the pitch and roll angles has a crucial impact on the spatial localization of objects observed by a vehicle's camera. Therefore, even a temporary miscalculation of these angles can affect the precise determination of the position of objects surrounding the vehicle, thus limiting or even disabling certain driver assistance system functionalities that rely on this information.

[0017] To estimate the instantaneous pitch, which we will simply refer to as pitch in the following, and the instantaneous roll, which we will simply refer to as roll in the following, of a camera mounted in a vehicle in a dynamic way, that is to say in order to appreciate the rapid variations linked to punctual changes in the driving environment, we already know the use of the simultaneous localization and mapping method or SLAM from the English "Simultaneous Localization And Mapping".

[0018] This method consists, firstly, of simultaneously creating a map of a location and locating oneself within it. It can also be used to estimate the camera's pitch and roll, notably by tracking certain prominent objects in the image and their movement from one frame to the next. However, it is highly susceptible to detection noise and is relatively limited in terms of the responsiveness at which the pitch and roll values ​​can be effectively updated, especially when a robust estimation is the objective.

[0019] Another known method involves processing images acquired by a vehicle's onboard camera to identify, in each image, the position of the road markings on which the vehicle is traveling, and to deduce, in real time, one or more camera tilt angles. This method can dynamically estimate pitch but requires traveling on a road with parallel and recognizable markings.

[0020] Furthermore, in the case of the two methods discussed so far, the estimation of pitch and roll values ​​is only effective when these values ​​remain sufficiently close to the expected values. In other words, as soon as the amplitude of the pitch variation is very large, as is the case, for example, when crossing a speed bump or during strong acceleration, or as soon as the amplitude of the roll variation is very large, as is the case, for example, when crossing a speed cushion or a pothole, the two methods mentioned above may fail to keep up with the changes in the camera angle.

[0021] To overcome the drawbacks of the two methods mentioned, a process for dynamically estimating the pitch and roll of a motor vehicle is known and is described and represented in document EP3579191B1.

[0022] The process described in document EP3579191B1 includes a first step of calculating an extrinsic tilt angle of a camera estimated by integrating the relative motion of the camera at a current time t, and a second step of estimating an extrinsic tilt angle of the camera with divergence correction.

[0023] The term "angle of inclination" refers to both a pitch angle and a roll angle, that is to say the angle between the camera frame of reference and the world frame of reference.

[0024] The first calculation step determines the angle of inclination using the following equation: θ int ( t ) = θ ( t - 1) - θ rel ( t ) with θ ( t - 1) the absolute extrinsic angle of inclination of the camera at a previous instant t-1, and θ rel ( t) the relative tilt angle of the camera at the current time t which expresses the tilt angle of the sensor between said previous time t-1 and the current time t.

[0025] The second step estimates the camera's tilt angle using the following equation: θ ( t ) = θ int ( t ) + DAC(t) * ( θ calib ( t ) - θ int ( t )) with θ ( t ) the absolute extrinsic angle of inclination of the camera, at the current time t, θ int ( t ) the extrinsic tilt angle of the camera estimated by integrating the relative motion of the camera during the first calculation step above, at the current time t, DAC ( t ) an anti-divergence coefficient between zero and one, at the current time t, and θ calib ( t) the extrinsic tilt angle of calibration at the current time t, which is known at each instant and which corresponds to the nominal value of the tilt angle of the camera in the absence of dynamics of the motor vehicle.

[0026] The anti-divergence coefficient is also known by the English name "Drift Avoiding Coefficient" for the acronym DAC.

[0027] Advantageously, the process described in document EP3579191B1 takes into account the movement of the camera and the multitude of road scenarios without diverging, by using the anti-divergence coefficient.

[0028] However, the anti-divergence coefficient depends only on camera movement, so when inter-frame movement is high, the anti-divergence coefficient tends towards zero and the estimated tilt angle θ ( t) approaches the extrinsic tilt angle of the camera estimated by integrating the relative motion of the camera θ int ( t ).

[0029] Conversely, when the inter-image movement is small, the anti-divergence coefficient tends towards its maximum value and the estimated tilt angle θ ( t ) approaches its reference value θ calib ( t ) .

[0030] However, it is observed that the motor vehicle can exhibit a prolonged tilt under certain conditions.

[0031] For example, in the case of a motor vehicle such as a truck equipped with a suspended cab, the cab can remain inclined at a constant angle for several seconds, due to the inertial force exerted on the cab.

[0032] Thus, in the case of emergency braking or in the case of a long turn, the cabin of the motor vehicle and the camera mounted on the cabin can maintain a pitch angle and a roll angle respectively, for several seconds.

[0033] Therefore, under certain conditions, the method described in document EP3579191B1 does not allow for reliable estimation of roll and pitch angles, which can lead to loss of the tracked object or an erroneous distance estimate to the tracked object.

[0034] US patent 6292759 B1 (SCHIFFMANN JAN KONRIED [US]) (2001-09-18) discloses a vehicle attitude angle estimator that minimizes errors that may be present in automotive-grade sensors. The attitude angle estimator uses an attitude angular velocity sensor to detect the attitude angular velocity, and an accelerometer to detect the vehicle's lateral or longitudinal acceleration. A controller determines an estimate of the attitude angle and updates this estimate based on the attitude angular velocity. The controller further provides a mixing coefficient and updates the attitude angle estimate to generate an estimate of the vehicle's current attitude angle based on the acceleration-based attitude angle and the mixing coefficient, and produces a vehicle angle estimation signal, such as a roll angle estimate and / or a pitch angle estimate.

[0035] US patent 6678631 B2 (DELPHI TECH INC [US]) (2004-01-13) discloses an attitude angle estimator (e.g., roll). The attitude angle estimator uses an attitude angular velocity sensor to detect a vehicle's attitude angular velocity, a first accelerometer to detect the vehicle's vertical acceleration, and a second accelerometer to detect the vehicle's horizontal (e.g., lateral) acceleration. A controller determines an acceleration-based attitude angle based on the detected vertical and horizontal accelerations, and further determines an attitude angle estimate based on the angular attitude rate and the acceleration-based attitude angle. [Exposé de l'invention]

[0036] The present invention aims in particular to resolve the aforementioned drawbacks of the prior art.

[0037] This objective, as well as others which will appear upon reading the description that follows, is achieved with a method for dynamically estimating the pitch and roll of a motor vehicle according to claim 1.

[0038] Depending on other optional features of the process according to the invention, taken alone or in combination: the anti-divergence coefficient DAC ( t ) depends in part on the movement of the sensor between the current time t and a previous time; the second estimation step includes the calculation of the anti-divergence coefficient DAC ( t ) by the following equation: DAC t = min DAC rel t , DAC a → t with : DAC rel ( t ) a relative anti-divergence coefficient which depends on the movement of the sensor and which is obtained by the following equation: DAC rel t = DAC maxrel ∗ 1 − min ∫ t − Nrel t θ rel x θAMR dx , 1 with DAC maxrel a predetermined maximum value of the relative anti-divergence coefficient DAC rel ( t ), Nrela predetermined number corresponding to the chosen value of the number of previous moments used to determine the relative anti-divergence coefficient DAC rel ( t ) , θ rel ( t ) the relative tilt angle of the sensor at the current time t, θAMR a predetermined empirical maximum value of the relative tilt angle of the sensor at the current time t θ rel ( t ), DAC to ( t ) an anti-divergence coefficient of acceleration which depends on the acceleration experienced by the sensor, and which is obtained by the following equation: DAC a → t = DAC max a → ∗ 1 − min ∫ t − Na t a → x a → AMR dx , 1 with DAC maxa a predetermined maximum value of the acceleration anti-divergence coefficient DAC to ( t ) , Na the predetermined number corresponding to the chosen value of the number of previous instants used to determine the anti-divergence coefficient of acceleration DAC to ( t ) , a ( t) the acceleration of the image acquisition sensor at time t, aAMR a predetermined empirical maximum value of the acceleration a ( t ) of the image acquisition sensor; the absolute extrinsic tilt angle θ ( t The sensor's roll angle corresponds to an angular oscillation of the sensor around a longitudinal axis that is generally parallel to the longitudinal trajectory of the motor vehicle; the acceleration a ( t ) of the image acquisition sensor is a transverse centrifugal acceleration along a direction generally perpendicular to the trajectory of the motor vehicle; the absolute extrinsic tilt angle θ ( t The sensor's pitch angle corresponds to an angular oscillation of the sensor around a transverse axis that is generally perpendicular to the longitudinal trajectory of the motor vehicle; the accelerationa ( t ) of the image acquisition sensor is a longitudinal acceleration along the longitudinal trajectory of the motor vehicle; the motor vehicle is equipped with a driver's cab which is mounted to oscillate in roll and pitch relative to the chassis of said vehicle and which carries the image acquisition sensor.

[0039] The present invention also relates to a device for dynamically estimating the pitch and roll of a motor vehicle by means of at least one image acquisition sensor mounted on said vehicle, intended to implement the method described above, the device comprising at least one image acquisition sensor mounted on said vehicle and a processing unit for determining the absolute extrinsic tilt angle. θ ( t ) of the sensor. [Description des dessins]

[0040] Other features and advantages of the invention will become apparent from the following description, with reference to the attached figures, which illustrate: [ Fig. 1 ] : a schematic view of a motor vehicle equipped with a suspended cabin and a video camera that does not have a dynamic tilt angle; ] Fig. 2 ] : a schematic view of the motor vehicle of the figure 1 in braking position in which the camera has a pitch angle; [ Fig. 3 ] : an example of how camera orientation changes relative to the pitch angle; [ Fig. 4 ] : a schematic view of the motor vehicle of the figure 1 in a canted position in which the camera has a roll angle at time t-1; [ Fig. 5 ] : a schematic view similar to that of the figure 4 in which the camera has a roll angle at a current instant t; [ Fig. 6] : an example of the evolution of the camera orientation relative to the roll angle.

[0041] In the documents of this patent application, the terminology longitudinal, vertical and transverse shall be adopted without limitation with reference to the L, V, T trihedron shown in the figures, considering that the motor vehicle extends longitudinally and moves longitudinally forward.

[0042] Across all these figures, identical or similar elements are identified by identical or similar reference symbols across all figures. [Description of production methods]

[0043] We represented at the figure 1 a motor vehicle 10 of the truck type which carries an image acquisition sensor consisting of a video camera 12.

[0044] The video camera 12 is mounted on the front windshield of a driver's cab 13, for example near the rearview mirror, so as to monitor the immediate surroundings of the motor vehicle 10, as illustrated in the figure 1 through the field of view 14 of the video camera 12.

[0045] The driver's cab 13 is mounted to move in roll and pitch relative to the vehicle chassis, for example by means of a suspension device.

[0046] The digital images acquired by the video camera 12 are transmitted to a processing unit 16 belonging to a driver assistance system, for example to enable the driver assistance system to detect objects and make an active intervention decision in the driving of the motor vehicle 10.

[0047] Also, the driving assistance system includes a dynamic estimation device 18 of the pitch and roll of the motor vehicle 10 which is adapted to implement a dynamic estimation method of the pitch and roll of the motor vehicle 10 according to the invention.

[0048] Roll is defined as the dynamically estimated absolute extrinsic tilt angle θ(t) of the video camera 12, which corresponds to the tilt angle of the video camera 12 around a longitudinal axis A that is globally parallel to the longitudinal trajectory of the motor vehicle 10, as can be seen in the figures 4 And 5 .

[0049] In other words, roll corresponds to an inclination of the structure of the motor vehicle 10, and therefore of the video camera 12, to the left or to the right in reference to the trajectory of movement of the motor vehicle 10 in forward motion.

[0050] Also, pitch is understood to mean an absolute extrinsic tilt angle θ(t) of the video camera 12, estimated dynamically, which corresponds to a tilt angle of the video camera 12 around a transverse axis B that is globally perpendicular to the longitudinal trajectory of the motor vehicle 10, as can be seen in the figure 2 .

[0051] In the following description, the absolute extrinsic tilt angle θ(t) will designate mutatis mutandis the pitch angle or the roll angle of the video camera 12, particularly in the mathematical formulas described below.

[0052] We represented at figures 1 to 5 a first three-dimensional frame RC {x C , y C , z C} which is called camera frame and which is linked to the video camera 12, and a second three-dimensional frame RM {x M , y M , z M} which is called world frame.

[0053] The RM world reference frame is linked to the front running gear 20 of the motor vehicle 10 and is projected onto the road.

[0054] In general, pitch and roll angles are angles of rotation between axes of the camera frame RC and axes of the world frame RM.

[0055] It should be noted that the orientation of the world frame RM is linked to the slope on which the motor vehicle 10 is moving. For example, if the vehicle is moving on a sufficiently long slope, the world frame is oriented along this slope.

[0056] As it appears more clearly on the figure 3 the absolute extrinsic angle of inclination θ ( t ) represents a pitch angle estimated at a time t defined as the angle formed, at that time t, between the z C axis of the camera RC frame and the x M axis of the world RM frame.

[0057] Similarly, with reference to figures 4 to 6 the absolute extrinsic angle of inclination θ ( t) represents a roll angle estimated at a time t defined as the angle formed, at that time t, between the x C axis of the RC camera frame and the y M axis of the RM world frame.

[0058] Those skilled in the art will appreciate that the pitch and roll represented herein and whose estimates are the subject of the invention are extrinsic pitch and roll, in the sense defined above in relation more generally to the extrinsic parameters of the video camera 12.

[0059] The potential variation of the intrinsic parameters, as defined above, is not the subject of the estimation described below. Indeed, these parameters are calibrated and, in general, not subject to significant variations over short time scales. They can therefore be considered stable with regard to the dynamic phenomena that the estimation method according to the invention aims to track, which are related to driving events such as passing over a speed bump or a dip in the road surface, i.e., over a sudden convex or concave change in elevation, respectively, in the case of pitch, or over a speed cushion, a sharp turn, or a pothole in the case of roll, on a road surface.

[0060] The method for dynamically estimating the pitch and roll of the motor vehicle 10 according to the invention comprises a first step of calculating an extrinsic tilt angle of the video camera 12 estimated by integrating the relative motion θ int ( t ) inter-image of camera 12 at a current instant t, by the following equation: θ int t = θ t − 1 − θ rel t with θ ( t - 1) the absolute extrinsic tilt angle of the video camera 12 at a previous instant t-1, and θ rel ( t ) the relative tilt angle of the video camera 12 at the current time t which expresses the tilt angle of the video camera 12 between the previous time t-1 and the current time t.

[0061] By previous instant, we mean for example the instant of the previous image acquisition by the video camera 12 in a sequence of successive acquisitions, or the instant of the acquisition of an image separated from a given number of acquisitions, number chosen for example in advance, in said sequence, or even the instant of the acquisition of a previous image acquired a certain given time before instant t.

[0062] Examples of changes in the orientation of video camera 12 are given to figures 3 And 6 .

[0063] The method according to the invention includes a second step of estimating an absolute extrinsic tilt angle of the video camera 12, which is carried out following the first step, by the following equation: θ t = θ int t + DAC t ∗ θ calib t − θ int t with θ ( t ) the absolute extrinsic tilt angle of the video camera 12, at the current time t, θ int ( t) the extrinsic tilt angle of video camera 12 estimated by integrating the relative motion of the inter-image camera during the first calculation step above, at the current time t, DAC ( t ) an anti-divergence coefficient between zero and one, at the current time t, and θ calib ( t ) the extrinsic tilt angle of calibration at the current time t, which is known at each instant and which corresponds to the nominal value of the tilt angle of the video camera 12 in the absence of dynamics of the motor vehicle 10.

[0064] The value of the extrinsic calibration tilt angle θ calib ( t ) as a function of time can be obtained by any known extrinsic calibration method, such as, for example, the SLAM method or white line tracking.

[0065] The extrinsic calibration tilt angle θ calib ( t) is a reference angle, it is also possible to use the angle of the video camera 12 decided during the design of the vehicle, or the angle estimated during calibration at the end of the vehicle production line.

[0066] The anti-divergence coefficient DAC ( t ) avoids, as its name suggests, the divergence in the estimation of the angle of inclination θ ( t ) .

[0067] Indeed, the angle of inclination θ ( t The estimated value may diverge from the true value due to the accumulation over time of errors in estimating the tilt angle estimated by integrating the relative motion of the camera.12 θ int ( t ).

[0068] In the absence of the anti-divergence coefficient DAC ( t ) , the potential errors in estimating the relative angle of inclination θ rel ( tThe video camera's 12 signals would be summed over time, which could lead to errors in estimating the tilt angle. θ ( t ) estimated and, in extreme cases, significant discrepancies between the angle of inclination θ ( t ) estimated and the actual angle of inclination.

[0069] Depending on the value assigned to the anti-divergence coefficient DAC ( t ), the calculation of the angle of inclination θ ( t ) is more or less restricted by the extrinsic calibration tilt angle θ calib ( t ) .

[0070] In order to estimate the angle of inclination θ ( t ) of the video camera 12 can take into account the multitude of road scenarios without diverging, the anti-divergence coefficient DAC ( t ) is calculated, during the second stage of process estimation, by the following equation: DAC t = min DAC rel t , DAC a → t with DAC rel ( t ) a relative anti-divergence coefficient which depends on the movement of the video camera 12 and which is obtained by the following equation: DAC rel t = DAC maxrel ∗ 1 − min ∫ t − Nrel t θ rel x θAMR dx , 1 with DAC maxrel a predetermined maximum value of the anti-divergence coefficient DAC rel ( t ), Nrel a predetermined number corresponding to the chosen value of the number of previous moments used to determine the relative anti-divergence coefficient DAC rel ( t ), θ rel ( t ) the relative tilt angle of the video camera 12 at the current time t, θAMR a predetermined empirical maximum value of the relative tilt angle of the video camera 12 at the current time t θ rel ( t ) .

[0071] The anti-divergence coefficient of acceleration DAC to ( t) is an anti-divergence coefficient which depends on the acceleration experienced by the video camera 12 and which is obtained by the following equation: DAC a → t = DAC max a → ∗ 1 − min ∫ t − Na t a → x a → AMR dx , 1 with DAC maxa a predetermined maximum value of the acceleration anti-divergence coefficient DAC to ( t ), Na the predetermined number corresponding to the chosen value of the number of previous instants used to determine the anti-divergence coefficient of acceleration DAC to ( t ), a ( t ) the acceleration of the video camera 12 at time t, and aAMR a predetermined empirical maximum value of the acceleration a ( t ) of the video camera 12.

[0072] Note that the parameters DAC maxrel , Nrel, θAMR , DAC maxa , Na And aAMR can be modified to adjust the behavior of the algorithm that estimates the absolute extrinsic tilt angle of the video camera 12 θ ( t) .

[0073] Advantageously, the anti-divergence coefficient DAC ( t ) depends both on the movement of the video camera 12 between the current instant t and a previous instant with the relative anti-divergence coefficient DAC rel ( t ) , and also the acceleration experienced by the video camera 12 during the movement of the motor vehicle 10 with the anti-divergence coefficient of acceleration DAC to ( t ) .

[0074] Indeed, when the inter-image movement is high, such as when passing over a pothole or a Berlin cushion, the relative anti-divergence coefficient DAC rel ( t ) tends towards zero and the angle of inclination θ ( t The estimated angle of inclination of video camera 12 approaches the extrinsic angle of inclination of video camera 12 estimated by integrating the relative motion of camera 12. θ int ( t ).

[0075] Conversely, when the inter-image movement is small, for example during a long emergency braking maneuver or when going around a roundabout, the relative anti-divergence coefficient DAC rel ( t ) tends towards the predetermined maximum value DAC maxrel of the acceleration anti-divergence coefficient DAC to ( t ) and the estimation of the angle of inclination θ ( t ) approaches its reference value, which is the extrinsic calibration tilt angle θ calib ( t ) according to the example described here.

[0076] The anti-divergence coefficient of acceleration DAC to ( t ) does not depend on the movement of the video camera 12 but on the acceleration a ( t ) suffered by video camera 12.

[0077] It should be noted that the acceleration a ( t) of the video camera 12 is a transverse centrifugal acceleration along a direction generally perpendicular to the trajectory of the motor vehicle 10 when the absolute extrinsic angle of inclination θ ( t ) of the video camera 12 designates a roll angle.

[0078] Centrifugal acceleration a ( t The force exerted on the video camera 12, and therefore on the cabin of the motor vehicle 10, can be estimated using a set of sensors designed for this purpose. In the case of the small-angle approximation, the centrifugal acceleration a ( t ) is determined by the following equation: a → t = ω t ∗ v t with ω the angular velocity of the cabin approximated to the angular velocity of the chassis of the motor vehicle 10, and v the speed of the cabin approximated to the speed of the chassis of the motor vehicle 10.

[0079] Conversely, acceleration a ( t) of the video camera 12 is a longitudinal acceleration along the longitudinal trajectory of the motor vehicle 10 when the absolute extrinsic angle of inclination θ ( t ) of the video camera 12 designates a pitch angle.

[0080] Making the anti-divergence coefficient dependent on acceleration DAC to ( t ) of the acceleration a ( t ) allows us to take into consideration scenarios in which the cabin, or more generally the support of the video camera 12, is tilted for a non-negligible time by an inertial force.

[0081] For example, if motor vehicle 10 enters a roundabout at low speed, the acceleration a ( t )) remains low and the anti-divergence coefficient of acceleration DAC to ( t ) takes a value that tends towards the predetermined maximum value DAC maxaof the anti-divergence coefficient of acceleration, and the angle of inclination θ ( t ) approaches the value of the extrinsic calibration tilt angle θ calib ( t ).

[0082] On the other hand, if motor vehicle 10 enters a roundabout dynamically, or if motor vehicle 10 initiates emergency braking, the acceleration a ( t ) is high and the anti-divergence coefficient of acceleration DAC to ( t ) takes a value that tends towards zero, and the angle of inclination θ ( t ) will not undergo filtering that brings it closer to the value of the extrinsic calibration tilt angle θ calib ( t ).

[0083] Thus, the method according to the invention makes it possible to estimate the absolute extrinsic angle of inclination. θ ( t) of the video camera 12, whether it is a roll angle or a pitch angle, stably in phases of low acceleration and reactively in phases of high acceleration or deceleration such as passing over a speed bump, a roundabout or a short section of inclined road.

[0084] Naturally, the invention described above is by way of example. It is understood that a person skilled in the art is capable of carrying out different embodiments of the invention without departing from its scope.

[0085] For example, a different combination of the acceleration anti-divergence coefficient DAC to ( t ) and the relative anti-divergence coefficient DAC rel ( t ) can be considered.

[0086] Similarly, but not limited to, the relative anti-divergence coefficient DAC rel ( t ), or the anti-divergence coefficient DAC ( t), can be determined to a constant value between zero and one.

Claims

1. A computer-implemented method of dynamically estimating the pitch and roll of a motor vehicle (10) by means of at least one image-acquiring sensor (12) that is located on board said motor vehicle (10), which method comprises at least: - a first step of computing at least an extrinsic tilt angle of the sensor (12) estimated by integrating the relative inter-image movement of the sensor (12) θint(t) at a current time t, using the following equation: θ int t = θ t − 1 − θ rel t with: i. θ(t - 1) the absolute extrinsic tilt angle of the sensor (12) at a previous time t-1, ii. θrel(t) the relative tilt angle of the sensor (12) at the current time t, which expresses the tilt angle of the sensor (12) between said previous time t-1 and the current time t, - a second step of estimating at least an absolute extrinsic tilt angle of said sensor (12), using the following equation: θ t = θ int t + DAC t * θ calib t − θ int t with: i. θ(t) the absolute extrinsic tilt angle of the sensor (12), at the current time t, ii. θint(t) the extrinsic tilt angle of the sensor (12) estimated by integrating the relative movement of the sensor (12) in the course of the preceding first computing step, at the current time t, iii. DAC(t) a drift-avoidance coefficient, comprised between zero and one, at the current time t, iv. θcalib(t) the calibration extrinsic tilt angle at the current time t, which is known at all times and which corresponds to the nominal value of the tilt angle of the sensor (12) when the motor vehicle (10) is not moving, where the drift-avoidance coefficient DAC(t) depends partly on the acceleration experienced by the image-acquiring sensor (12) in the course of the movement of the motor vehicle (10).

2. The method as claimed in claim 1, characterized in that the drift-avoidance coefficient DAC(t) depends partly on the movement of the sensor (12) between the current time t and a previous time.

3. The method as claimed in either one of the preceding claims, characterized in that the second estimating step comprises computing the drift-avoidance coefficient DAC(t) using the following equation: DAC t = min DAC rel t , DAC a → t with: - DACrel(t) a relative drift-avoidance coefficient that depends on the movement of the sensor (12) and that is obtained using the following equation: DAC rel t = DAC maxrel ∗ 1 − min ∫ t − Nrel t θ rel x θAMR dx , 1 with DACmaxrel a predetermined maximum value of the relative drift-avoidance coefficient DACrel(t), Nrel a predetermined number corresponding to the selected value of the number of previous times used to determine the relative drift-avoidance coefficient DACrel(t), θrel(t) the relative tilt angle of the sensor (12) at the current time t, and θAMR a predetermined empirical maximum value of the relative tilt angle of the sensor (12) at the current time t θrel(t), DACa(t) an acceleration drift-avoidance coefficient that depends on the acceleration experienced by the sensor (12), and that is obtained using the following equation: DAC a → t = DAC max a → ∗ 1 − min ∫ t − Na t a → x a → AMR dx , 1 with DACmaxa a predetermined maximum value of the acceleration drift-avoidance coefficient DA Ca(t), Na the predetermined number corresponding to the selected value of the number of previous times used to determine the acceleration drift-avoidance coefficient DACa(t),a(t) the acceleration of the image-acquiring sensor (12) at the time t, and aAMR a predetermined empirical maximum value of the acceleration a (t) of the image-acquiring sensor (12).

4. The method as claimed in any one of the preceding claims, characterized in that the absolute extrinsic tilt angle θ(t) of the sensor is a roll angle corresponding to an angular oscillatory movement of the sensor (12) about a longitudinal axis (A) that on the whole is parallel to the longitudinal path of the motor vehicle (10).

5. The method as claimed in claims 3 and 4, characterized in that the acceleration a(t) of the image-acquiring sensor (12) is a transverse centrifugal acceleration in a direction that on the whole is perpendicular to the path of the motor vehicle (10).

6. The method as claimed in any one of claims 1 to 3, characterized in that the absolute extrinsic tilt angle θ(t) of the sensor is a pitch angle corresponding to an angular oscillatory movement of the sensor (12) about a transverse axis (B) that on the whole is perpendicular to the longitudinal path of the motor vehicle (10).

7. The method as claimed in claims 3 and 6, characterized in that the acceleration a(t) of the image-acquiring sensor (12) is a longitudinal acceleration along the longitudinal path of the motor vehicle (10).

8. The method as claimed in any one of the preceding claims, characterized in that the motor vehicle (10) is equipped with a driver's cab (13) that is mounted so as to be able to oscillate rollwise and pitchwise with respect to the chassis of said vehicle (10) and that bears the image-acquiring sensor (12).

9. A device (18) for dynamically estimating the pitch and roll of a motor vehicle (10) by means of at least one image-acquiring sensor (12) that is located on board said vehicle, said device being intended to implement the method as claimed in any one of the preceding claims, the device including at least one image-acquiring sensor (12) that is mounted on said vehicle (10) and a processing unit (16) for determining the absolute extrinsic tilt angle θ(t) of the sensor (12).