Method for processing data acquired from a radar sensor

Abstract

Examples are presented as a method for processing data acquired by a radar sensor, enabling, in particular, the determination or refinement of coordinates of digital objects obtained from a two-dimensional occupancy grid. Figure from the abstract: [Fig. 2]

Description

Title of the invention: Method for processing data acquired from a radar sensor Technical field

[0001] The present disclosure relates to the field of processing data acquired from a radar sensor and in particular when this radar sensor is on board a vehicle. Prior art

[0002] A radar sensor is a sensor which allows, by the emission of electromagnetic waves, the acquisition of a cloud of points representing the environment around the radar sensor.

[0003] However, the intrinsic technology of the radar sensor, based on the Doppler effect, is such that static objects in the environment of the radar sensor are difficult to locate by the radar sensor. In other words, a point cloud acquired by a radar sensor during an acquisition includes few points relating to static objects.

[0004] However, many applications require the precise detection of static objects surrounding the radar sensor, particularly when this radar is embedded in a device moving autonomously based on information acquired by the radar sensor.

[0005] A first type of radar sensor used in these applications thus allows the acquisition of a point cloud in which each point of the point cloud is associated with two values ​​corresponding to the abscissa (x) and ordinate (y) coordinates of the space and to a value expressing a radar equivalent surface (also known under the English term “radar cross section”).

[0006] Based on this first type of radar sensor, algorithms using successive point clouds acquired by the radar sensor have been developed in order to construct a two-dimensional ground occupation grid (abscissa x and ordinate y) of the environment of the radar sensor. The two-dimensional occupation grid comprises boxes and each box is associated with a probability of finding an obstacle, and abscissa and ordinate intervals define the position and the surface of the box in the two-dimensional occupation grid. Thus, the probability of finding an obstacle on a box of the occupation grid is updated each time a new point cloud is acquired by the radar sensor. In this way, the objects surrounding the radar sensor, and in particular static objects, are better identified since their detection is based on several acquisitions of the radar sensor.

[0007] Radar sensors of a second type then emerged and now allow the acquisition of point clouds in which each point of a point cloud is associated with three values ​​corresponding to the coordinates of space (x, y, z), where z is an elevation coordinate relative to the ground, as well as a radar equivalent surface value.

[0008] However, the algorithms developed to construct a three-dimensional occupancy grid (x, y, z) from successive acquisitions of second-type radar sensors are expensive in terms of computation time. Consequently, these algorithms cannot be implemented in all applications. In particular, the computation time required to construct the three-dimensional occupancy grid (x, y, z) is not always adapted to the constraints imposed by certain embedded systems that must react in real time, particularly with regard to vehicles in the context of driving assistance functions.

[0009] Furthermore, the surfaces of the boxes of the two-dimensional ground occupation grid do not always represent very precisely the surface actually occupied by an obstacle in the environment of the radar sensor. In this case, the boxes have a relatively large surface, for example of the order of one meter by 0.5 meters. Consequently, the direct use of the data from the two-dimensional ground occupation grid to position obstacles in the environment of the radar sensor is sometimes not sufficient for all the applications for which this ground occupation grid is used, in particular when a position of the obstacle must be determined more precisely. A reduction in the size of the boxes is not suitable since this reduction would increase the processing time so that this type of method would no longer be suitable for real time.Furthermore, since the points representing static objects are few in number since they are rarely detected by the radar sensor, a high-resolution occupancy grid would not be relevant for representing them.

[0010] The invention improves this situation. Summary

[0011] In this regard, a method is proposed for processing data acquired by a radar sensor, the radar sensor being adapted to acquire a point cloud representing its environment in which each point of the point cloud is associated with three values ​​corresponding to three space coordinates and to a value expressing a radar equivalent surface, the method comprising: - obtaining a point cloud acquired by the radar sensor; - obtain, from the point cloud, a two-dimensional land use grid; - detect, on the two-dimensional occupancy grid, a list of 2D digital objects, each 2D digital object being associated with a two-dimensional zone delimited by abscissa and ordinate boundaries; then for at least one 2D digital object from the 2D digital object list, - determining a value of a coordinate to be associated with said processed 2D digital object, the determination comprising: * the selection of a set of points from the point cloud whose ground coordinates belong to the area associated with the processed 2D digital object; * storing the radar equivalent surface values ​​of the selected points in intervals of an elevation grid extending between a minimum elevation value and a maximum elevation value and comprising several successive elevation intervals delimited by limits from the elevation coordinates of the selected points; - determine the coordinate value to be associated with the processed 2D digital object from this elevation grid to obtain a 3D digital object.

[0012] Optionally, the coordinate to be associated with the processed 2D digital object is an elevation value, and determining an elevation value comprises determining a first elevation value corresponding to the elevation value of the selected point associated with the maximum radar equivalent surface value of the elevation grid.

[0013] Optionally, the coordinate to be associated with the processed 2D digital object is an elevation value, and the determination of the elevation value comprises the determination of a pair of elevation extrema, which comprises a minimum elevation value and a maximum elevation value.

[0014] Optionally, determining an elevation extrema pair comprises: selecting a specific elevation interval of the elevation grid associated with the elevation extrema pair, the specific elevation interval comprising the elevation interval storing the maximum radar equivalent surface value and possibly comprising other elevation intervals of the elevation grid, each elevation interval included in the specific elevation interval comprising at least one radar equivalent surface value whose deviation from the maximum radar equivalent surface value is less than a predetermined SER threshold; and - determination of the elevation extrema pair from the specific elevation interval of the elevation grid.

[0015] Optionally, the minimum elevation value corresponds to the minimum elevation value among the elevation values ​​of the points having an equivalent surface value belonging to the specific elevation interval whose deviation from the maximum radar equivalent surface value is less than the predetermined SER threshold; and the maximum elevation value corresponds to the maximum elevation value among the elevation values ​​of the points having an equivalent surface value belonging to the specific elevation interval whose deviation from the maximum radar equivalent surface value is less than the predetermined SER threshold.

[0016] Optionally, the minimum elevation value corresponds to the lower elevation bound of the specific elevation interval; and the maximum elevation value corresponds to the upper elevation bound of the specific elevation interval.

[0017] Optionally, the method comprises, after determining a first pair of elevation extrema: - the removal of the radar equivalent surface (SER) values ​​from the elevation grid belonging to the specific interval; then - the implementation of a new step of determining a pair of elevation extrema to determine a second pair of elevation extrema.

[0018] Optionally, the method further comprises at least one of: - modifying the lower abscissa bound of the processed 2D digital object from the elevation grid, so that said lower abscissa bound corresponds to the minimum abscissa value among the abscissa values ​​associated with the selected points having a stored radar equivalent surface value in the specific elevation interval; - modifying the upper abscissa bound of the processed 2D digital object from the elevation grid, so that said upper abscissa bound corresponds to the maximum abscissa value among the abscissa values ​​associated with the selected points having a stored radar equivalent surface value in the specific elevation interval; - modifying the lower ordinate bound of the processed 2D digital object from the elevation grid, so that said lower ordinate bound corresponds to the minimum ordinate value among the ordinate values ​​associated with the selected points having a stored radar equivalent surface value in the specific elevation interval; or - modifying the upper ordinate bound of the processed 2D digital object from the elevation grid, so that said upper ordinate bound corresponds to the maximum ordinate value among the ordinate values ​​associated with the selected points having a stored radar equivalent surface value in the specific elevation interval.

[0019] Optionally, for each elevation interval of the elevation grid, a single radar equivalent surface value is stored and corresponds to the highest radar equivalent surface value associated with the selected points whose elevation value belongs to the elevation interval concerned.

[0020] Optionally, the method being implemented iteratively and further comprising the tracking of at least one digital object, the method further comprising, for the digital object processed, and as a function of a distance criterion: - the association of the processed digital object with a digital object belonging to a previously determined list of digital objects, the coordinates of the digital object in the previously determined list being updated from the coordinates of the processed digital object with which it is associated; or - creation of a new digital object in the list of digital objects from the coordinates of the processed digital object.

[0021] Optionally, the minimum and maximum elevation values ​​of the elevation grid are determined from the sensor's field of view.

[0022] The application also relates to a vehicle comprising a radar sensor adapted to acquire a point cloud representing its environment in which each point of the point cloud is associated with three values ​​corresponding to the three coordinates of space and to a value expressing a radar equivalent surface, in which the vehicle also comprises a computer adapted to control any one of the methods presented by the present disclosure.

[0023] The application further relates to a computer program product comprising instructions for implementing any of the methods presented by the present disclosure when this program is executed by a processor.

[0024] Finally, the application relates to a non-transitory recording medium readable by a computer on which is recorded a program for implementing any of the methods presented by the present disclosure when this program is executed by a processor.

[0025] The method presented by the present disclosure requires little resource from the computer so that it can be executed in real time so as to determine a precise elevation value for a two-dimensional digital object from the acquisitions of the radar sensor and to redetermine a value for one or more abscissa and / or ordinate terminals of a two-dimensional digital object that are more precise than those determined from the two-dimensional ground occupation grid.

[0026] This method, when implemented iteratively, is particularly suitable for defining the dimensions of static objects in the environment of the radar sensor even when these static elements are poorly detected by the radar sensor. Indeed, even if a static element is not always detected in the ground occupation grid obtained from an acquisition of the radar sensor, it can be detected by other acquisitions so that the dimensions of the static element, once represented by the coordinates of a digital object in the list of digital objects, will be able to be updated from the different acquisitions of the radar sensor each time this object is detected on the ground occupation grid. Brief description of the drawings

[0027] Other characteristics, details and advantages will appear on reading the detailed description below, and on analyzing the attached drawings, in which:

[0028] [Fig-1] [Fig. 1] represents an exemplary system for implementing any of the data processing methods described in the present application.

[0029] [Fig.2] [Fig.2] represents an example of a data processing method acquired by a radar sensor 1.

[0030] [Fig.3] [Fig.3] represents an example of a two-dimensional area of ​​an object 2D digital and point cloud points.

[0031] [Fig.4] [Fig.4] represents an example of an elevation grid. Description of the embodiments

[0032] An example of a system for implementing any of the data processing methods described below is now described with reference to [Fig. 1]. The system comprises a computer 11 adapted to control the implementation of the data processing methods, the data being acquired by a radar sensor 10. The computer has access to a memory 12. The memory 12 can in particular store the data acquired by the radar sensor 10 and the code instructions executed by the computer so as to be able to implement any of the data processing methods described below.

[0033] The computer 11 may for example be of the processor, microprocessor, microcontroller, FPGA, etc. type. The memory 12 may for example comprise a ROM (Read-Only Memory), a RAM (Random Access Memory), an EEPROM (Electrically Erasable Programmable Read-Only Memory) or any other type of suitable storage means. The memory may for example comprise optical, electronic or magnetic storage means.

[0034] The radar sensor 10 is adapted to acquire a point cloud representing its environment. In this case, each point of the point cloud is associated with three values ​​corresponding to three spatial coordinates (x, y, z) and with a value expressing a radar cross section SER. The radar cross section SER is also known by the English term “radar cross section” (RCS). The radar cross section SER can in particular be expressed in decibels (dB).

[0035] Each point acquired by the radar is therefore associated with an abscissa value x, an ordinate value y, an elevation value z and a radar equivalent surface value SER.

[0036] The radar sensor 10 can be embedded in a vehicle 1, as illustrated by [Fig.l], so as to acquire point clouds of the environment around the vehicle 1, in particular in the context of driving assistance functions and in particular autonomous driving. In these exemplary embodiments, the computer 11 making it possible to control the implementation of any one of the data processing methods can also be embedded in the vehicle 1.

[0037] More broadly, the radar sensor 10 can be embedded in a mobile system, in particular a robot vacuum cleaner or a drone, in order to map the environment of this system.

[0038] With reference to [Fig.2], an example of a method 100 for processing data acquired by a radar sensor 10 is described.

[0039] The data processing method presented with reference to [Fig.2] will make it possible to determine a value of a coordinate to be associated with a digital object, for example an elevation coordinate value (z value). A digital object comprises coordinate values ​​defining abscissa (xmin, xmax) and ordinate (ymin, ymax) limits of this object and possibly one or more elevation values ​​defining an elevation of this digital object. The data processing method presented makes it possible in this case to determine a coordinate value to be associated with a specific digital object from an acquisition of a radar sensor 10.

[0040] As illustrated in block 110, the method 100 comprises obtaining a point cloud acquired by the radar sensor 10. It is therefore a point cloud in which each point is associated with the four coordinates detailed above.

[0041] As illustrated in block 120, the method 100 comprises obtaining, from the point cloud, a two-dimensional land use grid. The two-dimensional land use grid comprises a plurality of boxes forming the grid in which each box is associated with a probability of finding an obstacle. Each box is associated with abscissa and ordinate intervals which define the position and the surface area of ​​the box in the two-dimensional land use grid. Advantageously, the two-dimensional land use grid is determined from the static points of the point cloud. A static point designates a point of the point cloud which belongs to a static object relative to the radar sensor 10. The algorithms for determining, from a cloud of points obtained by a radar sensor, which are the static points of the cloud and which are the moving points of the cloud are known to those skilled in the art.

[0042] The algorithms for obtaining a two-dimensional land use grid from a point cloud acquired by a sensor, whether a radar sensor or another type of sensor, are known to those skilled in the art. In particular, a non-limiting example of an algorithm for obtaining a two-dimensional land cover grid is the Bayesian recursive estimation algorithm. The present application is not limited to this example and other algorithms for obtaining such a grid are also applicable.

[0043] As illustrated in block 130, the method 100 comprises detecting, on the two-dimensional occupancy grid, a list of 2D digital objects. A 2D digital object is a digital object associated with a two-dimensional area (x, y) delimited by abscissa (xmin, xmax) and ordinate (ymin, ymax) boundaries. A 2D digital object is therefore an object detected on the ground occupation grid, delimited by abscissa and ordinate boundaries, which expresses a surface area of ​​this occupancy grid.

[0044] In examples, the list of 2D digital objects detected on the occupancy grid comprises all or part of static objects relative to the radar sensor 10. In particular, when the two-dimensional ground occupation grid is determined from the static points of the point cloud, the detection of the list of digital objects necessarily only comprises static objects. In this case, the processing detailed below is particularly advantageous for determining a coordinate of a static object relative to the radar sensor 10 insofar as this type of object is difficult to detect by the radar sensor 10.

[0045] A 2D digital object may, for example, correspond to one or more boxes of the two-dimensional land use grid having a probability of finding an obstacle greater than a predetermined obstacle threshold. In other words, the detection, on the two-dimensional land use grid, of a list of 2D digital objects may comprise a selection of the boxes of the land use grid for which the probability of finding an obstacle is sufficiently significant to seek to determine a coordinate value, for example an elevation value, to be associated with this box. As explained below, the method may also make it possible, as an alternative or in addition to the determination of the elevation value, to redetermine a lower (xmin) or upper (xmax) abscissa bound and a lower (ymin) or upper (ymax) ordinate bound to be associated with the 2D digital object.

[0046] The association of an elevation value with a 2D digital object therefore makes it possible to obtain a 3D digital object since the 2D digital object now comprises a component translating an elevation of this object in space. A 3D digital object therefore comprises a 2D digital object associated with a two-dimensional zone delimited by abscissa (xmin, xmax) and ordinate (ymin, ymax) boundaries expressing a ground surface of the digital object and also comprises at least one elevation value defining an elevation of this digital object.

[0047] Once the 2D digital objects in the list of 2D digital objects have been detected at the end of block 130, the method 100 comprises additional processing for at least one 2D digital object in the list. Advantageously, the processing is carried out for each 2D digital object in the list of 2D digital objects.

[0048] Thus, for at least one 2D digital object from the list of 2D digital objects, the method 100 comprises, as illustrated by block 140, determining a value of a coordinate to be associated with the processed 2D digital object. The determined coordinate may be relative to: - an elevation of the processed 2D digital object, - a lower bound on the abscissa (xmin) of the processed 2D digital object, - an upper bound on the abscissa (xmax) of the processed 2D digital object, - a lower bound on the ordinate (ymin) of the processed 2D digital object, and / or - a lower bound on the ordinate (ymax) of the processed 2D digital object.

[0049] As illustrated by block 141, the determination of a coordinate value for the processed 2D digital object comprises the selection of a set of points from the point cloud whose ground coordinates belong to the two-dimensional zone (xmin, xmax, ymin, ymax) associated with the processed 2D digital object. It is therefore a matter of determining which points of the point cloud acquired by the radar sensor 10 are included in the zone associated with the 2D digital object, that is to say included between the abscissa (xmin, xmax) and ordinate (ymin, ymax) limits of the processed 2D digital object. An example of several points P of the point cloud and of a 2D digital object (O2D) considered is notably represented in [Fig.3]. As shown in [Fig.3], points PI to P5 belong to the zone associated with the digital object O2D so that they are selected at the end of block 141. Conversely, points P6 and P7 do not belong to this zone so that they are not selected.In examples in which the occupancy grid comprises only static objects, the selection of a set of points from the point cloud whose ground coordinates belong to the two-dimensional area is carried out on the static points of the point cloud. In other words, it is a matter of selecting, among the static points of the point cloud, those which have ground coordinates belonging to the two-dimensional area (xmin, xmax, ymin, ymax) associated with the processed 2D digital object.

[0050] As illustrated by block 142, determining a coordinate value for the 2D digital object comprises storing the radar equivalent surface SER values ​​of the selected points in elevation intervals of an elevation grid G. The elevation grid G ​​extends between a minimum elevation value Gzmin and a maximum elevation value Gzmax. In this case, the elevation grid comprises several successive elevation intervals IE delimited by elevation bounds and the storage of the radar equivalent surface SER values ​​is carried out from the z elevation coordinates of selected points. The elevation interval of the elevation grid storing the radar equivalent surface SER value of a selected point is the one for which the z elevation value associated with this point is between its elevation bounds.

[0051] An example of an elevation grid G ​​is notably represented in [Fig.4] and stores the radar equivalent surface SER values ​​of points PI to P5, each of these values ​​being identified by the reference SERPn where the index n is an integer and refers to the point P with which it is associated. The same notation will be applied for the other coordinates (xpn, ypn and zpn) in the rest of the document. The elevation intervals of the example in [Fig.4] are referenced from IE1 to IE5.

[0052] In examples, the minimum elevation values ​​Gzmin and maximum Gzmax of the elevation grid G ​​are determined from the field of view of the sensor. In these examples, the intervals of the elevation grid are constant between the different iterations of the method. The field of view of the sensor designates the elevation interval in which the radar sensor can detect an object, for example between -2m and 8m. In this example, the minimum elevation values ​​Gzmin and maximum Gzmax of the elevation grid G ​​correspond to -2m and 8m respectively.

[0053] In examples, for each elevation interval IE of the elevation grid, a single radar equivalent surface area value SER is stored. This equivalent surface area value may, for example, correspond to the highest radar equivalent surface area value SER associated with the selected points whose elevation value z belongs to the elevation interval IE concerned. In this case, the radar equivalent surface area value SER of a point reflects a level of reliability of the measurement of the point obtained by the radar sensor 10. The higher this value, the less likely the point obtained is to be due to measurement noise or to a non-significant element that is not dangerous for the radar sensor 10 or for the system that incorporates this sensor 10.Storing only the highest radar equivalent surface SER value per elevation interval therefore makes it possible to store only the radar equivalent surface SER values ​​which are associated with the points considered to be the most reliable acquired by the radar sensor 10.

[0054] As illustrated by block 143, the method 100 also comprises determining an elevation value to associate with the 2D digital object from this elevation grid to obtain a 3D digital object.

[0055] In examples, determining an elevation value includes determining a first elevation value corresponding to the elevation value z of the selected point associated with the maximum radar equivalent surface SERmax value of the elevation grid. As explained above, the radar equivalent surface SER value of a point reflects a level of reliability of the measurement of the point obtained by the radar sensor 10. Thus, by choosing the elevation value in z of the point which is associated with the maximum equivalent surface value SERmax of the elevation grid, the method determines an elevation value to be associated with the 2D digital object for which the risks of confusion due to a measurement error of the radar sensor 10 are very low. Furthermore, the method requires little computing time and can thus be implemented in real time by the computer 11.

[0056] The clever use of an algorithm making it possible to obtain a two-dimensional land use grid combined with the construction of an elevation grid from the radar equivalent surface SER values ​​of the points acquired by the radar sensor 10 and belonging to the two-dimensional zone of a 2D digital object detected in the land use grid therefore makes it possible to determine an elevation value to be associated with the 2D digital object which is reliable in terms of measurement and the determination of which is fast enough to be carried out in real time.

[0057] In other examples, the determination of an elevation value comprises the determination of a pair of elevation extrema (zmin, zmax) from the radar equivalent surface SER values ​​of the elevation grid. The pair of elevation extrema (zmin, zmax) comprises a minimum elevation value (zmin) and a maximum elevation value (zmax) to be associated with the 2D digital object. This therefore involves determining elevation limits (zmin and zmax) for the 2D digital object and associating them with it to obtain a 3D digital object comprising the two dimensions of the ground of the 2D digital object delimited by limits (xmin, xmax, ymin, ymax) and the elevation also delimited by limits (zmin, zmax). In this example, a 3D digital object therefore corresponds to a volume indicating an obstacle in the environment of the radar sensor 10, and for example in the environment of a vehicle 1 carrying this radar sensor 10.This embodiment therefore makes it possible to determine characteristics defining an obstacle on a three-dimensional occupancy grid (x, y, z), in particular of an obstacle corresponding to a static element in the environment of the radar sensor 10, by implementing a method suitable for being implemented in real time.

[0058] In these examples, the determination of a pair of elevation extrema (zmin, zmax) comprises: - the selection of a specific elevation interval IES of the elevation grid associated with the elevation extrema pair; and - determination of the elevation extrema pair from the specific elevation interval IES of the elevation grid.

[0059] The specific elevation interval IES includes the elevation interval IE storing the maximum radar equivalent surface value SERmax and may include other elevation intervals of the elevation grid. Each elevation interval included in the specific elevation interval IES comprises at least one radar equivalent surface value SER whose deviation from the maximum radar equivalent surface value SERmax is less than a predetermined SER threshold. The SER threshold may for example be between 1 and 10 dB and advantageously between 2 and 5 dB. In one embodiment, the SER threshold is equal to 3 dB.

[0060] As explained previously, a high radar equivalent surface SER value reflects a certain level of reliability relating to the consistency of the point acquired by the radar sensor such that when several radar equivalent surface SER values ​​are close to the maximum value SERmax, it can also be considered that the points associated with these SER values ​​are reliable and are not due to errors in the acquisition of the radar sensor 10. Consequently, a specific elevation interval IES which includes the elevation interval comprising the SERmax value and which may include other elevation intervals adjacent to this interval, provided that these intervals have a SER value close to that of the SERmax value, represents a reliable and rapid elevation interval to determine for a computer 11. This interval can therefore be used to determine the pair of elevation extrema to be associated with the 2D digital object.

[0061] In first examples, the minimum elevation value zmin of the pair of elevation extrema to be associated with the 2D object may correspond to the lower elevation bound of the specific elevation interval IES.

[0062] In these first examples, the maximum elevation value zmax of the pair of elevation extrema to be associated with the 2D object may correspond to the upper elevation limit of the specific elevation interval IES.

[0063] In second examples, the minimum elevation value zmin may correspond to the minimum elevation value among the elevation values ​​z of the points having an equivalent surface value SER belonging to the specific elevation interval IES whose deviation from the maximum radar equivalent surface value SERmax is less than the predetermined SER threshold.

[0064] In these second examples, the maximum elevation value zmax may correspond to the maximum elevation value among the elevation values ​​z of the points having an equivalent surface value SER belonging to the specific elevation interval IES whose deviation from the maximum radar equivalent surface value SERmax is less than the predetermined SER threshold.

[0065] The second examples therefore make it possible to determine a pair of elevation extrema to be associated with the reliable 2D digital object, precise since it is based on the elevation values ​​acquired by the radar sensor 10, and adapted to be implemented in real time by a computer. The first examples are less precise than the second examples but faster to implement since it is not necessary to go search, in memory, each of the z elevation values ​​of the points having an equivalent surface value SER stored in the elevation grid and compare them.

[0066] It is also possible to determine the zmin value using the first examples and the zmax value using the second examples or vice versa.

[0067] Furthermore, in an embodiment in which the method 100 determines the minimum zmin and maximum zmax elevation values, the method may further comprise: - the determination of an average elevation value zmoy of the 3D digital object corresponding to the sum of the minimum zmin and maximum zmax elevation values ​​divided by two with: zmoy = (zmin + zmax) / 2; - the determination of a height hz of the 3D digital object corresponding to the difference between the maximum elevation value zmax and the minimum elevation value zmin with: hz = zmax - zmin; and / or - the determination of an extent value ze of the 3D digital object corresponding to the square of half the height hz with: ze = (0.5 hz)2.

[0068] In one embodiment, the method 100 comprises, after determining a first pair of elevation extrema, removing the radar equivalent surface SER values ​​from the elevation grid belonging to the specific interval IES and then implementing a new step of determining a pair of elevation extrema to determine a second pair of elevation extrema. This embodiment makes it possible to determine, on the same ground surface, two elevation intervals defined by the two pairs of elevation extrema which may correspond to two different obstacles. On the two-dimensional ground occupation grid, it is by definition impossible to discriminate between two obstacles located at the same place at ground level but which are not located at the same elevation level.This embodiment makes it possible to do this by removing the radar equivalent surface SER values ​​from the elevation grid belonging to the specific elevation interval so that when a second step is implemented, the maximum radar equivalent surface SERmax value is no longer the same, the previous one having been removed. Consequently, a new specific elevation interval IES can be determined from this new maximum radar equivalent surface SERmax value and the remaining radar equivalent surface SER values ​​from the elevation grid that have not been removed. Thus, from this new specific elevation interval IES can thus be determined a second pair of elevation extrema. When this embodiment is implemented, . the determined 3D digital object is therefore associated with the two-dimensional area on the ground delimited by abscissa (xmin, xmax) and ordinate (ymin, ymax) boundaries and it is also delimited in elevation by two intervals (zminl, zmaxl and zmin2, zmax2). There are in reality two different obstacles but which can be represented computer-wise by the same 3D digital object. This scenario makes it possible in particular to represent a bridge located at a first elevation and a vehicle positioned below the bridge at a second elevation.

[0069] In embodiments comprising determining a specific elevation interval, the method may comprise modifying the lower abscissa bound xmin of the processed 2D digital object from the elevation grid, so that said lower abscissa bound corresponds to the minimum abscissa value among the abscissa values ​​associated with the selected points having a radar equivalent surface value stored in the specific elevation interval IES.

[0070] These embodiments make it possible to redetermine a new abscissa value corresponding to the lower abscissa limit xmin of the two-dimensional area of ​​the 2D digital object from the points of the point cloud acquired by the radar sensor 10. In this case, the previous abscissa value of the lower abscissa limit of the two-dimensional area of ​​the 2D digital object comes from the detection of this 2D digital object on the two-dimensional land use grid. However, a box of the two-dimensional land use grid, defined by an abscissa interval (xmin, xmax) and by an ordinate interval (ymin, ymax) and used to describe an obstacle represented by a 2D digital object, can occupy a larger surface area than the actual surface area of ​​the obstacle that it represents. Consequently, an obstacle can be detected closer than it actually is to the radar sensor.In this sense, updating the abscissa value defining the lower abscissa limit xmin in which the obstacle is included using precise data from the point cloud makes it possible to redefine an abscissa position closer to reality than that initially determined by the algorithm for obtaining the ground occupation grid.

[0071] In embodiments comprising determining a specific elevation interval, the method may comprise modifying the upper abscissa bound xmax of the processed 2D digital object from the elevation grid, so that said upper abscissa bound corresponds to the maximum abscissa value among the abscissa values ​​associated with the selected points having a radar equivalent surface value stored in the specific elevation interval IES.

[0072] In embodiments comprising determining a specific elevation interval, the method may further comprise modifying the lower ordinate bound ymin of the processed 2D digital object from the elevation grid, such that said lower ordinate bound corresponds to the minimum ordinate value among the ordinate values ​​associated with the selected points having a stored radar equivalent surface value in the specific elevation interval.

[0073] In embodiments comprising determining a specific elevation interval, the method may further comprise modifying the upper ordinate bound ymax of the processed 2D digital object from the elevation grid, so that said upper ordinate bound corresponds to the maximum ordinate value among the ordinate values ​​associated with the selected points having a stored radar equivalent surface value in the specific elevation interval.

[0074] In the same way as for the abscissa value xmin defining the lower abscissa bound in which the obstacle defined by the 2D digital object is included, it is also possible to redetermine an abscissa value for the upper abscissa bound xmax, an ordinate value for the lower ordinate bound ymin of the digital object and an ordinate value for the upper ordinate bound ymax in two dimensions using the same principle. Consequently, the new bound values ​​(xmin, xmax, ymin, ymax) are more precise than those obtained from the ground occupation grid.

[0075] Thus, in embodiments, the method makes it possible both to determine an elevation value to be associated with the 2D digital object in real time so as to determine a 3D digital object but also to update one or more abscissa (xmin, xmax) and / or ordinate (ymin, ymax) boundaries of the 2D digital object detected on the elevation grid from the data from the points of the point cloud acquired by the radar sensor 10. In this case, this updating of the 2D digital objects detected from the occupancy grid can be implemented in parallel with the determination of the elevation value once the specific elevation interval has been determined.

[0076] The embodiment having a storage of a single radar equivalent surface SER value in each elevation interval of the elevation grid for which the stored value corresponds to the highest radar equivalent surface SER value among the selected points belonging to this elevation interval is therefore all the more advantageous in the embodiments redetermining one or more abscissa (xmin, xmax) and / or ordinate (ymin, ymax) limits of the 2D digital object. Indeed, the abscissa or ordinate value from which the corresponding limit of the 2D digital object is modified is therefore necessarily associated with a point having a radar equivalent surface SER close to the surface maximum radar equivalent SERmax since this is how the specific elevation interval is determined. This is therefore data associated with points considered reliable in the point cloud acquired by the radar sensor 10.

[0077] The method 100 presented by the present disclosure therefore makes it possible, from known algorithms making it possible to obtain a two-dimensional ground occupation grid, to determine an elevation value for the 2D digital objects detected on the two-dimensional occupation grid and to redetermine one or more abscissa (xmin, xmax) and ordinate (ymin, ymax) boundaries of these 2D digital objects in a more precise manner than that obtained from the grid. Furthermore, the method 100 can be carried out in real time by the computer.

[0078] In this case, the present application does not exclude the fact that once the specific elevation interval IES has been determined, the method may comprise the determination of one or more limits (xmin, xmax) on the abscissa and / or on the ordinate (ymin, ymax) of the 2D digital object instead of determining the elevation value to be associated with the digital object. The 2D digital object will therefore not become a 3D digital object since no elevation value will be associated with it.

[0079] The block 140 for determining a value of a coordinate to be associated with a 2D digital object detected on the ground occupation grid can therefore determine a value of a coordinate of one or more abscissa terminals (xmin, xmax) and / or a value of a coordinate of one or more ordinate terminals (ymin, ymax) rather than an elevation value.

[0080] In other words, the block 143 for determining the elevation value can be replaced by a block for determining a terminal on the abscissa or ordinate of the 2D digital object processed by the block 140.

[0081] This alternative block is represented by block 153 in the method 100 of [Fig.l]. The method 100 for processing sensor data acquired by the radar can therefore comprise: - the selection of a specific elevation interval of the elevation grid from the radar equivalent surface SER values ​​stored in this grid; then - a modification of a terminal in abscissa (xmin, xmax) or in ordinate (ymin, ymax) of the 2D digital object processed from the specific elevation interval IES of the elevation grid.

[0082] Examples for obtaining a specific elevation interval have been presented above with reference to obtaining an elevation value. Similarly, examples for determining an abscissa (xmin, xmax) or ordinate (ymin, ymax) boundary of the 2D digital object from the specific elevation interval have been presented above.

[0083] The method presented by the present application therefore makes it possible to determine an elevation value or, as an alternative or in addition to the elevation value, one or more abscissa or ordinate limits of the 2D digital object.

[0084] In one embodiment, the method 100 is implemented iteratively and further comprises tracking at least one previously determined 2D or 3D digital object. As explained previously, a 2D digital object becomes a 3D digital object when an elevation value is associated with it, for example when elevation limits zmin and zmax are associated with it. As such, when the method 100 does not include determining an elevation value to be associated with a 2D digital object but determining an abscissa or ordinate value of a lower limit of this 2D digital object, it will involve tracking a 2D digital object over several iterations. Conversely, when an elevation value is associated with a 2D digital object which will therefore become a 3D digital object, it will involve tracking this 3D digital object over several iterations.

[0085] In this case, the radar sensor 10 acquires successive point clouds to map its environment. Consequently, the data associated with a digital object (xmin, xmax, ymin, ymax and possibly one or more data in z) created from a first point cloud of a first iteration of the method can be updated from the data of a second point cloud newly acquired during a second iteration, etc.

[0086] Thus, in this embodiment in which the method is implemented iteratively, the method may comprise, as illustrated by block 160, for the digital object processed in block 140 and from a distance criterion: - an association of the processed digital object with a digital object belonging to a previously determined list of digital objects, the coordinates of the digital object in the previously determined list being updated from the coordinates of the processed digital object with which it is associated; or - a creation of a new digital object in the previously determined list of digital objects from the coordinates of the processed digital object.

[0087] In other words, it is a matter of associating a digital object processed in a current iteration of the method with a digital object created at the end of a previous iteration of the method and belonging to a previously determined list or of creating a new digital object in this list according to a distance criterion.

[0088] In this case, and as explained below, in an embodiment in which the radar sensor is embedded in a mobile system, the coordinates of the digital objects belonging to the previously determined list of digital objects can be modified at each iteration as a function of the movement of the radar sensor between the iterations so as to modify the coordinates of these digital objects. to predict their current position relative to the position of the radar sensor during the current iteration of the process.

[0089] In examples, when a distance between the two-dimensional area of ​​the digital object and the two-dimensional area of ​​a digital object in the predetermined list is less than a predetermined distance threshold, the processed digital object is associated with the digital object in the predetermined list. Conversely, in these examples, when the distance between the two-dimensional area of ​​the 2D digital object and the two-dimensional area of ​​each digital object in the predetermined list is greater than a predetermined distance threshold, a new digital object is created in the list of digital objects from the two-dimensional area of ​​the 2D digital object.

[0090] As illustrated in [Fig.l], the block 160 for creating a digital object or associating the processed digital object with a digital object from the previously determined list of digital objects is carried out once the processed digital object has been modified so as to update the list of digital objects by taking into account the modifications made to the processed digital object during the current iteration. By modification, it is understood here its association with one or more determined elevation values ​​and / or the modification of one or more of its abscissa and / or ordinate bounds.

[0091] In this case, the creation of a new digital object in the list of digital objects may correspond to a copy of the processed digital object in the list of digital objects. In other words, the coordinates of the processed digital object are copied into the list of digital objects in order to create a new object in this list.

[0092] In examples, an update of the coordinates of the digital object of the list previously determined from the coordinates of the processed digital object may correspond to an at least partial replacement of the values ​​of the coordinates of the digital object of the list previously determined by the values ​​of the coordinates of the corresponding processed digital object.

[0093] In other examples, the coordinates of the digital object of the previously determined list updated from the coordinates of the processed digital object may be updated from the n digital objects most recently associated with the digital object of the previously determined list of which the processed digital object is a part. In examples, a coordinate of the digital object of the previously determined list may correspond to the average of the corresponding coordinates of the n digital objects most recently associated with this digital object. In examples, the number n may be greater than 3 and may advantageously be greater than 5. It is therefore a question of filtering the coordinates of the objects digital objects from the list previously determined from the processed digital objects associated with them during the different iterations in order to refine the coordinates of the digital objects in the list over time.

[0094] In an embodiment in which the radar sensor 10 is embedded in a mobile system, the two-dimensional area of ​​the processed digital object or the two-dimensional area of ​​each object in the previously determined list of digital objects can be updated from a movement of the radar sensor 10. This makes it possible to find a correspondence between the two-dimensional areas of the digital objects in the list and the two-dimensional area of ​​the processed digital object since the acquisition reference mark of the radar sensor 10 moves with the radar sensor 10 during the acquisition of the point clouds.

[0095] In examples in which the two-dimensional area of ​​each object in the previously determined list of digital objects is updated from a movement of the radar sensor 10 and the digital object to be updated is static, the two-dimensional area of ​​the digital object to be updated is updated from a movement of the radar sensor 10 between the iteration at which the digital object to be updated was created in the list and the current iteration of the method.

[0096] In examples in which the two-dimensional area of ​​each object in the previously determined list of digital objects is updated based on a movement of the radar sensor 10 and the digital object to be updated is mobile, the two-dimensional area of ​​the digital object to be updated is updated based on a relative movement of the radar sensor 10 with respect to this digital object to be updated between the iteration at which the digital object to be updated was created in the list and the current iteration of the method. This embodiment therefore makes it possible to plot digital objects over several iterations and to update the data associated with these digital objects based on the newly determined digital objects from the point clouds acquired as and when by the radar sensor 10.

[0097] The method 100 presented by the present disclosure requires little resource from the computer so that it can be executed in real time so as to determine a precise elevation value for a 2D digital object from the acquisitions of the radar sensor 10 and to redetermine a value for one or more abscissa (xmin, xmax) and / or ordinate (ymin, ymax) terminals of a 2D digital object that are more precise than those determined from the two-dimensional ground occupation grid.

[0098] This method, when implemented iteratively, is particularly suitable for defining the dimensions of static objects in the environment of the radar sensor 10 even though these static elements are poorly detected by the radar sensor 10. Indeed, even if a static element is not always detected in the ground occupation grid obtained from an acquisition of the radar sensor, it can be detected by other acquisitions so that the dimensions of the static element, once represented by the coordinates of a digital object in the list of digital objects, can be updated from the different acquisitions of the radar sensor 10 each time this object is detected on the ground occupation grid.

Claims

Claims

1. Method for processing data acquired by a radar sensor, the radar sensor being adapted to acquire a point cloud representing its environment in which each point of the point cloud is associated with three values corresponding to three space coordinates (x, y, z) and to a value expressing a radar equivalent surface (SER), the method comprising: - obtaining (110) a point cloud acquired by the radar sensor; - obtaining (120), from the point cloud, a two-dimensional land occupation grid; - detecting (130), on the two-dimensional land occupation grid, a list of 2D digital objects, each 2D digital object being associated with a two-dimensional area (x, y) delimited by abscissa and ordinate boundaries (xmin, xmax, ymin, ymax);then for at least one 2D digital object from the list of 2D digital objects, - determining (140) a value of a coordinate to be associated with said processed 2D digital object, the determination comprising: * selecting (141) a set of points from the point cloud whose ground coordinates belong to the area associated with the processed 2D digital object; * storing (142) the radar equivalent surface (SER) values of the selected points in intervals of an elevation grid (G) extending between a minimum elevation value (Gzmin) and a maximum elevation value (Gzmax) and comprising several successive elevation intervals delimited by limits from the elevation coordinates (z) of the selected points; - determining (143,153) the coordinate value to be associated with the processed 2D digital object from this elevation grid to obtain a 3D digital object; the coordinate to be associated with the processed 2D digital object being an elevation value.;

2. A method according to the preceding claim, wherein the coordinate to be associated with the processed 2D digital object is an elevation value, and determining an elevation value comprises determining a first elevation value corresponding to the value elevation (z) of the selected point associated with the maximum radar equivalent surface value (SERmax) of the elevation grid.

3. A method according to claim 1 or 2, wherein the coordinate to be associated with the processed 2D digital object is an elevation value, and determining the elevation value comprises determining a pair of elevation extrema (zmin, zmax), which comprises a minimum elevation value (zmin) and a maximum elevation value (zmax).

4. Method according to the preceding claim, wherein the determination of a pair of elevation extrema (zmin, zmax) comprises: - selecting a specific elevation interval (IES) of the elevation grid associated with the pair of elevation extrema, the specific elevation interval comprising the elevation interval storing the maximum radar equivalent surface value (SERmax) and being able to comprise other elevation intervals of the elevation grid, each elevation interval included in the specific elevation interval comprising at least one radar equivalent surface value (SER) whose deviation from the maximum radar equivalent surface value (SERmax) is less than a predetermined SER threshold; and - determining the pair of elevation extrema (zmin, zmax) from the specific elevation interval (IES) of the elevation grid.

5. Method according to the preceding claim, wherein the minimum elevation value (zmin) corresponds to the minimum elevation value among the elevation values (z) of the points having an equivalent surface area value (SER) belonging to the specific elevation interval (IES) whose deviation from the maximum radar equivalent surface area value (SERmax) is less than the predetermined SER threshold; and the maximum elevation value (zmax) corresponds to the maximum elevation value among the elevation values (z) of the points having an equivalent surface area value (SER) belonging to the specific elevation interval (IES) whose deviation from the maximum radar equivalent surface area value (SERmax) is less than the predetermined SER threshold.

6. The method of claim 4, wherein the minimum elevation value (zmin) corresponds to the lower elevation bound of the specific elevation interval (SEI); and the maximum elevation value (zmax) corresponds to the upper elevation bound of the specific elevation interval (SEI).

7. Method according to any one of claims 4 to 6, wherein the method comprises, after the determination of a first pair of elevation extrema: - the deletion of the radar equivalent surface (SER) values from the elevation grid belonging to the specific interval (IES); then - the implementation of a new step of determining a pair of elevation extrema to determine a second pair of elevation extrema.

8. A method according to any one of claims 4 to 7, wherein the method further comprises at least one of: - modifying the lower abscissa bound (xmin) of the processed 2D digital object from the elevation grid, so that said lower abscissa bound corresponds to the minimum abscissa value among the abscissa values associated with the selected points having a radar equivalent surface value stored in the specific elevation interval; - modifying the upper abscissa bound (xmax) of the processed 2D digital object from the elevation grid, so that said upper abscissa bound corresponds to the maximum abscissa value among the abscissa values associated with the selected points having a radar equivalent surface value stored in the specific elevation interval;- modifying the lower ordinate bound (ymin) of the 2D digital object processed from the elevation grid, so that said lower ordinate bound corresponds to the minimum ordinate value among the ordinate values associated with the selected points having a radar equivalent surface value stored in the specific elevation interval; or - modifying the upper ordinate bound (ymax) of the 2D digital object processed from the elevation grid, so that said upper ordinate bound corresponds to the maximum ordinate value among the ordinate values associated with the selected points having a radar equivalent surface value stored in the specific elevation interval.;

9. A method according to any preceding claim, wherein for each elevation interval of the elevation grid, a single radar cross section (RCS) value is stored and corresponds to the highest radar cross section (RCS) value associated with the selected points whose elevation value falls within the relevant elevation interval.

10. Method according to any one of the preceding claims, the method being implemented iteratively and further comprising tracking at least one digital object, the method further comprising, for the processed digital object, and as a function of a distance criterion: - associating the processed digital object with a digital object belonging to a previously determined list of digital objects, the coordinates of the digital object of the previously determined list being updated from the coordinates of the processed digital object with which it is associated; or - creating a new digital object in the list of digital objects from the coordinates of the processed digital object.

11. A method according to any preceding claim, wherein the minimum (Gzmin) and maximum (Gzmax) elevation values of the elevation grid (G) are determined from the field of view of the sensor.

12. Vehicle comprising a radar sensor adapted to acquire a point cloud representing its environment in which each point of the point cloud is associated with three values corresponding to the three spatial coordinates (x, y, z) and to a value expressing a radar equivalent surface (SER), the vehicle also comprising a computer adapted to control a method according to any one of claims 1 to 11.

Images