Processing of data acquired by LiDAR sensors

Abstract

Examples set forth a method of processing data acquired by a LiDAR sensor, a computer configured to perform the processing, a vehicle carrying a computer so configured, a computer program product, and a computer readable non-transitory storage medium.

Description

Technical Field

[0001] The present disclosure relates to the field of data processing. Background Art

[0002] The processing of data acquired by sensors is increasingly used in real time to assess situations and make decisions about them.

[0003] For example, there are methods for tracking in real time features of an object identified on a sequence of images acquired by a camera in order to estimate the relative movement of the object with respect to the camera, which methods may in particular enable the movement of the object to be predicted.

[0004] These methods are extremely useful in particular in the context of driver assistance functions or automated driving functions, since they enable real-time tracking of motor vehicles in the surroundings of the vehicle on which the camera is mounted and possible anticipation of the trajectories of these vehicles.

[0005] However, the method of tracking features of an object can be improved. Summary of the invention

[0006] In this regard, a computer-implemented method for processing data acquired by a LiDAR sensor is proposed, the method comprising:

[0007] - using the acquisition of the LiDAR sensor to obtain a first point matrix and a second point matrix, the points of the first matrix and the points of the second matrix respectively represent the environment of the LiDAR sensor at the first moment and the second moment; each point of these point matrices is associated with a coordinate in the three-dimensional space and an intensity value;

[0008] - Projecting these point matrices onto a two-dimensional projection plane to obtain an intensity image and a depth image for each of these matrices;

[0009] For each of the first and second point matrices:

[0010] * determining at least one feature of an element of the environment of the LiDAR sensor on the matrix of points using the three-dimensional coordinates and the intensity values ​​of the points of the matrix of points, the feature being thus associated with the points of the matrix; then

[0011] For at least one determined feature on each matrix:

[0012] * determining a first neighboring point window in the intensity image using the two-dimensional coordinates of the points of the matrix and the intensity values ​​of the points in the intensity image, the first neighboring point window being associated with the feature and including the points associated with the feature;

[0013] *determine a second neighboring point window associated with the feature, the neighboring points of the second neighboring point window corresponding to the neighboring points of the first neighboring point window whose distance from the point representing the feature in the three-dimensional space or the depth image is less than a predetermined threshold; then

[0014] - compare the second neighboring dot window of the first dot matrix with the second neighboring dot window of the second dot matrix; then

[0015] - using the comparison to associate features associated with a second neighboring window of dots of the first matrix of dots with corresponding features associated with a second neighboring window of dots of the second matrix of dots.

[0016] Optionally, the method may further comprise: determining a displacement of a feature of an element of an environment belonging to the LiDAR sensor between the first moment and the second moment using two second neighboring point windows associated with corresponding associated features.

[0017] Optionally, determining a displacement of a feature of an element of the environment belonging to the LiDAR sensor between the first moment and the second moment may comprise:

[0018] - determining in the intensity image a two-dimensional displacement of the feature between a first moment in time and a second moment in time using the two-dimensional coordinates and intensity values ​​of points of two second point windows in the intensity image associated with the feature; and

[0019] - determining a three-dimensional displacement of the feature between the first and second moments in time using the two-dimensional displacement in the intensity image and using the two-dimensional coordinates and depth values ​​of points of two second point windows associated with the feature in the depth image.

[0020] Optionally, the method may further include: for a second adjacent point window associated with the feature,

[0021] - determining a descriptor of the second neighboring point window, the descriptor corresponding to a feature value of the neighboring point window determined using the coordinates and intensity values ​​of the points of the second point window in the intensity image; and

[0022] Features associated with the second neighboring point window of the first matrix may be associated with corresponding features associated with the second neighboring point window of the second matrix using a distance between the descriptors associated with the second neighboring point window of the first matrix and the descriptors associated with the second neighboring point window of the second matrix.

[0023] Optionally, a binary robust independent basic feature method can be used to implement the following operations: determining a descriptor of a second neighboring point window; and associating features related to the second neighboring point window of the first point matrix with corresponding features related to the second neighboring point window of the second point matrix using the distance between the descriptor of the second neighboring point window of the first point matrix and the descriptor of the second neighboring point window of the second point matrix.

[0024] Optionally, the Lucas-Kanade method can be used to implement the following operations: determining a descriptor of the second neighboring point window; and associating a feature associated with the second neighboring point window of the first point matrix with a corresponding feature associated with the second neighboring point window of the second point matrix. In this option, the displacement of the feature of the element between the first moment and the second moment can be determined based on minimizing the distance between the descriptors of the two neighboring point windows associated with the corresponding associated features.

[0025] The application also relates to a computer configured to implement any of the data processing methods set forth in the present disclosure, and to a vehicle carrying a computer having one of these configurations.

[0026] The present application also relates to a computer program product comprising instructions for implementing any of the methods set forth in the present disclosure when the program is executed by a processor.

[0027] Finally, the present application relates to a computer-readable non-transitory storage medium on which a program is stored, which, when executed by a processor, is used to implement any of the methods set forth in the present disclosure.

[0028] Thus, the method according to the present disclosure enables the use of information acquired with the LiDAR sensor to track features of an element of the environment of the LiDAR sensor between two or possibly more point matrices to track features of an element of the environment of the LiDAR sensor over time. Thus, in applications where the LiDAR sensor is carried on a motor vehicle, the tracked features may, for example, belong to another motor vehicle, enabling tracking of motor vehicles travelling close to the vehicle carrying the LiDAR sensor. In particular, the method enables tracking of features over time, including in the event of occlusion. Thus, optionally, the method may in particular enable determination of a relative displacement of the feature over time. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] Other features, details and advantages will become apparent from a reading of the following detailed description and an analysis of the accompanying drawings, in which:

[0030] [ Figure 1 ] schematically represents an example data processing device for implementing a method for processing data acquired by a LiDAR sensor.

[0031] [ Figure 2 ] schematically represents an example vehicle including a data processing device and a LiDAR sensor.

[0032] [ Figure 3] schematically represents an example of a method for processing data acquired by a LiDAR sensor.

[0033] [ Figure 4 ] schematically shows an example of two dot matrices illustrating an occlusion situation. DETAILED DESCRIPTION

[0034] refer to Figure 1 A method for implementing the processing of data acquired by the LiDAR sensor 10 is described, in particular with reference to Figure 3 An example data processing device 1 of an example data processing method is described.

[0035] The data processing device 1 may be designed to be carried on the vehicle 2 .

[0036] The data processing device 1 comprises a computer 11 and a memory 12. The device is configured to process data acquired by a light detection and ranging (LiDAR) sensor 10.

[0037] The memory 12 may store code instructions that are executed by the computer 11 and used to control the LiDAR sensor 10 to acquire data and process the data. Therefore, the computer 11 is able to access information stored in the memory. The memory 12 may also be designed to store data acquired by the LiDAR sensor 10.

[0038] For example, the memory 12 may be a read-only memory (ROM), a random access memory (RAM), an electrically erasable programmable read-only memory (EEPROM) or any other suitable storage device. For example, the memory may include optical storage devices, electrical storage devices or even magnetic storage devices.

[0039] The data processing device 1 may be in a motor vehicle, such as Figure 2 This figure also shows a LiDAR sensor 10, which enables acquisition of data processed by a data processing device.

[0040] The light detection and ranging (LiDAR) sensor 10 is a sensor that emits light waves and determines, from the reflections of these light waves, a matrix of dots representing the environment of the LiDAR sensor.

[0041] The LiDAR sensor 10 is designed to acquire a matrix of points. Each point is associated with a three-dimensional coordinate (x, y, z) representing the environment of the LiDAR sensor 10 and an intensity value. With respect to the three-dimensional coordinates, the x-coordinate of a point is the horizontal coordinate of the point relative to the sensor 10. The y-coordinate of a point is the vertical coordinate of the point relative to the sensor 10. The z-coordinate of a point is the depth coordinate of the point relative to the LiDAR sensor 10. Figure 2The example shown shows a top view of the outline of the vehicle 2 and the LiDAR sensor 10, as well as an abscissa axis X and a depth axis Z of the LiDAR sensor 10. A ordinate axis Y of the sensor 10 is perpendicular to the axis X and the axis Z. Figure 2 In the example shown, the LiDAR sensor 10 is mounted on the front of the vehicle and is oriented to emit a light beam in the direction of movement of the vehicle.

[0042] Each point acquired by the LiDAR sensor 10 is also associated with an intensity value. The intensity value is the intensity of the light beam received by the LiDAR sensor after it is reflected on the surface.

[0043] In a first example, the LiDAR sensor 10 according to the present disclosure may be a scanning LiDAR sensor. This is a LiDAR sensor that acquires a matrix of points representing its environment, in which each point of the matrix is ​​associated with a different moment in time. More specifically, each point of the matrix of points is acquired by emitting a different light beam, so that there is a time difference between each point of the matrix of points representing the environment of the LiDAR.

[0044] In a second example, the LiDAR sensor 10 according to the present disclosure may be a flash LiDAR sensor. Unlike a scanning LiDAR sensor, a flash LiDAR sensor acquires a dot matrix representing its environment by emitting a single beam with a wide cross section, so that each dot of the dot matrix is ​​acquired at the same time.

[0045] Reference below Figure 3 An exemplary method 100 for processing data acquired by a LiDAR sensor 10 is described. The method may be implemented, for example, by a computer 11 of a data processing device 1 .

[0046] It should be noted that Figure 3 This is merely an illustration of an example of method 100, using blocks to represent various operations that are optionally included in the method and described below in the document. Therefore, this illustration does not reflect any order between the operations unless such order is specified in the present disclosure. In other words, reference Figure 3 The operations described are not necessarily performed sequentially, but can be performed in parallel. Figure 3 The order of execution of the operations shown in the figure is different from that shown in the figure. In addition, it is not necessary to repeat a given operation twice after each operation is executed once. The frequency of execution of each operation is specific to itself and is not necessarily related to the execution of other operations.

[0047] like Figure 3As shown in , method 100 includes an operation 110 of obtaining a first point matrix and a second point matrix. These point matrices are obtained from the acquisition of the LiDAR sensor 10. The points of the first point matrix and the points of the second point matrix represent the environment of the LiDAR sensor at the first moment and the second moment, respectively. As explained above, each point of the point matrix is ​​associated with a three-dimensional coordinate and an intensity value.

[0048] In a first example where the LiDAR sensor 10 is a scanning LiDAR sensor, obtaining the first matrix of points and the second matrix of points may include correcting the three-dimensional coordinates of the points in each of these matrices to compensate for the time lag between the corresponding points in each matrix. In fact, in an example where the scanning LiDAR 10 moves during the acquisition of the points of the matrix, for example when it is mounted on a motor vehicle, the time lag between the acquisition of each point of the matrix will result in a lag in the three-dimensional coordinates of the point, which must be corrected so that all points in the matrix are considered to be acquired at the same time. Methods for correcting such lags are known to those skilled in the art. In particular, a method for correcting such lags is described, for example, in the document "High-Precision Motion Compensation for LiDAR based on LiDAR Odometry" by QIN et al.

[0049] Furthermore, it will be appreciated that in the second example where the LiDAR sensor 10 is a flash LiDAR sensor, such correction is not necessary, since the corresponding points in each of the matrices are acquired simultaneously. Indeed, acquiring two matrices of points at two different moments in time by the flash LiDAR sensor directly provides two matrices of points associated with the two different moments in time.

[0050] like Figure 3 As shown, method 100 then includes operation 120: projecting the two point matrices onto a two-dimensional projection plane to obtain an intensity image and a depth image for each of the matrices. The intensity image is an image comprising a plurality of points in a two-dimensional space, each point being associated with an intensity value. The depth image is an image comprising a plurality of points in a two-dimensional space, in which case each point is associated with a depth value. When the points in the matrices are projected into a two-dimensional space while retaining the intensity values ​​of the intensity image and the depth values ​​of the depth image, these points may also be referred to as pixels of the intensity image and pixels of the depth image.

[0051] like Figure 3As shown in , the method 100 then includes an operation 130 performed on each of the first matrix and the second matrix: determining at least one feature of an element of the environment of the LiDAR sensor on the matrix of points so as to associate the feature with the points in the matrix. The at least one feature is determined using the three-dimensional coordinates and the intensity values ​​of the points.

[0052] The term "feature" in this disclosure is understood to have the meaning used in the field of image processing, in which case the image is the matrix of points under consideration. In this field, the French term "caractéristique" generally corresponds to the English term "feature". In this case, in the field of image processing, the term "feature" can refer to a visual attribute or unique property of an image, such as contour, texture, pattern, color, shape, angle, corner point, etc.

[0053] In some examples, operation 130 of determining at least one feature of an element of the environment of the LiDAR sensor on the considered point matrix may include determining a contour, texture, pattern, color, shape, angle, or corner point of an element of the environment of the LiDAR sensor.

[0054] In some examples, operation 130 of determining at least one characteristic of an element of the environment belonging to the LiDAR sensor on the matrix of points may include an angle of the vehicle.

[0055] The determined feature on the point matrix is ​​associated with a point of the point matrix under consideration. In the example where the determined feature is included in several points of the point matrix, the point of the point matrix associated with the feature is selected from the points of the subset including the determined feature. In the example where the determined feature is included in a single point of the point matrix, the point of the point matrix associated with the feature corresponds to the point including the feature.

[0056] In some examples, intensity gradients are used to determine characteristics of elements of the environment of the LiDAR sensor 10 over a matrix of points.

[0057] like Figure 3 As shown in , the method 100 then comprises an operation 140 performed for at least one determined feature on each matrix. Advantageously, the operation 140 is performed during the operation 130 for each feature determined on the first matrix and on the second matrix.

[0058] Operation 140 is performed in the intensity image of the matrix. It involves determining a first neighboring point window, which is related to the considered feature of the matrix and includes the points in the intensity image associated with the feature. The first neighboring point window is determined using the two-dimensional coordinates of the points of the matrix and the intensity values ​​of the points in the intensity image. In this case, when the feature is determined on the matrix during operation 130 and the feature is associated with a given point of the matrix, it is entirely possible to track this point projected onto the intensity image and / or the depth image. This enables the determination of a first neighboring point window in the intensity image, which includes the point associated with the feature. As the name implies, the neighboring point window includes a plurality of points located adjacent to each other in the two-dimensional space of the image, and the window is determined on the two-dimensional space of the image.

[0059] In some examples, the first neighboring point window associated with a particular feature may include a point associated with the particular feature and neighboring points of the point, i.e., a point whose distance from the point associated with the particular feature is below a predetermined distance threshold. In these examples, the point associated with the particular feature may be a center point of the first neighboring point window.

[0060] like Figure 3 As shown in , method 100 then includes operation 150: determining a second neighboring point window associated with the feature. The neighboring points of the second neighboring point window are neighboring points of the first neighboring point window whose distance from the point representing the feature in the three-dimensional space or the depth image is below a predetermined threshold. This operation involves removing points from the first neighboring point window that are too far away from the point associated with the feature in the three-dimensional space or in the depth image to determine the second neighboring point window. In fact, since these points are too far away from the point representing the feature in the space associated with the depth value, they may potentially belong to elements other than the elements that include the determined feature in the space. As described below, method 100 enables tracking the displacement of the feature between two moments based on a second point window associated with the feature at two moments, that is, the first moment is represented by the first matrix and the second moment is represented by the second matrix.

[0061] like Figure 3 , method 100 then comprises operation 160: comparing a second neighboring point window belonging to the first point matrix (at the first moment) with a second neighboring point window belonging to the second point matrix (at the second moment). The second point windows are compared in the intensity image. Advantageously, each second neighboring point window of the first point matrix is ​​compared with each second neighboring point window of the second point matrix, respectively.

[0062] In some examples, the operation 160 of comparing the second neighboring point window belonging to the first point matrix with the second neighboring point window belonging to the second point matrix includes: determining the distance between the neighboring point windows to be compared and comparing the distance between the neighboring point windows to be compared. In particular, the distance used during this operation can be the Hamming distance.

[0063] like Figure 3 As shown in , method 100 includes operation 170: using the comparison to associate features associated with the second neighboring point window of the first matrix of points with corresponding features associated with the second neighboring point window of the second matrix of points. Therefore, during operation 170, a correspondence is established between features of the first matrix of points and features of the second matrix of points using the corresponding second neighboring point windows of the points.

[0064] In some examples, features associated with a second neighboring point window of the first point matrix are associated with features associated with a second neighboring point window of the second point matrix if the distance between corresponding point windows of the points is below a predetermined threshold.

[0065] Thus, the method 100 according to the present disclosure makes it possible to use the information acquired with the LiDAR sensor 10 to track features of elements of the environment of the LiDAR sensor between two point matrices. Although the method is explained for two point matrices, it can be implemented on more than two point matrices, in particular in the case where these point matrices are acquired successively, so as to track features of elements of the environment of the LiDAR sensor 10 over time. Thus, in applications where the LiDAR sensor 10 is carried on a motor vehicle, the tracked features can, for example, belong to another motor vehicle, making it possible to track motor vehicles traveling close to the vehicle carrying the LiDAR sensor 10, for example to estimate the speed of these motor vehicles or their respective trajectories. Of course, many other applications can be envisaged and the present disclosure is not limited to automotive applications.

[0066] In the present disclosure, features are tracked using information acquired by the LiDAR sensor 10. In particular, the advantage of the LiDAR sensor 10 compared to a camera is that points with coordinates in a three-dimensional space are acquired. This means that feature tracking based on three-dimensional coordinates is more accurate than feature tracking based on two-dimensional coordinates.

[0067] For example, using the method 100 for acquiring data from the LiDAR sensor 10 also enables tracking of features in occlusion situations. An occlusion situation is a situation in which a tracked feature belonging to a first element of the sensor's environment is partially hidden or obscured on a point matrix by a second element of the sensor's environment. In these situations, defining a neighboring point window associated with a feature of the first element of the sensor's environment in the point matrix without taking into account any depth information of these points may make it difficult to track the feature, because such a window may include points belonging to a second element different from the first element of the sensor's environment, which second element may move relative to the first element. This affects the association of corresponding features of the two point matrices based on the comparison of the neighboring point windows associated with these features, because the neighboring point window of the first point matrix representing the sensor's environment at time t1 may include points belonging to the second element, whereas if the second element has moved relative to the first element between time t1 and time t2, the neighboring point window of the second matrix representing the sensor's environment at time t2 will no longer include any such points. The method according to the present disclosure enables coping with this situation in particular by considering that the points determined in the second neighboring point window are points close to the feature in the three-dimensional space or depth image (i.e., by taking into account the depth information in this way). This reduces the probability that some of the points constituting the second neighboring point window belong to elements other than the element including the feature over time.

[0068] In particular, Figure 4An example of an occlusion situation is schematically shown in order to facilitate understanding of this situation. This figure shows a first point matrix M1 representing the environment of the camera at time t1 and a second point matrix M2 representing the environment of the camera at time t2. On the first point matrix M1, a first neighboring point window f1 associated with feature C is schematically shown, and the first neighboring point window includes points belonging to the first element E1 and the second element E2, which partially hides the first element E1. The first neighboring point window f1 is associated with feature C belonging to element E1. On the second point matrix M2, a second neighboring point window f2 is schematically shown, which is also associated with feature C of element E1, and the second neighboring point window includes points belonging to the first element E1, but no longer includes points belonging to the second element E2, because the element E2 has moved relative to the first element E1 between acquiring the first matrix M1 and acquiring the second matrix M2. It will be understood here that if the neighboring points are considered in a two-dimensional coordinate space, then comparing the first neighboring point window f1 with the second neighboring point window f2 does not allow to determine that the feature C associated with the window f1 and the window f2 is indeed the corresponding feature C of the element E1, because the points of the neighboring point windows being compared are relatively different. The method 100 according to the present disclosure precisely enables to avoid this situation, because the second neighboring point window being compared during operation 160 should no longer or almost no longer include points belonging to elements other than the element including the tracked feature, because the depth information is used to select the points of the second point window.

[0069] Other operations may optionally be incorporated into method 100 and are set forth in the remainder of this disclosure. Unless otherwise indicated in this disclosure, these operations may be incorporated into method 100 in combination with each other.

[0070] In some examples, prior to the comparison operation 160, the method 100 may further include an operation 155 of determining a descriptor corresponding to a feature value of the second neighboring point window for each of the second neighboring point windows compared during operation 160. Examples of calculating descriptors of the considered neighboring point windows are particularly described in the document "BRIEF: Binary Robust Independent Elementary Features" by CALONDER et al. The BRIEF document particularly describes a method called "Binary Robust Independent Elementary Features Method" (BRIEF Method).

[0071] In some examples, a descriptor of the second window of neighboring points may be determined using the two-dimensional coordinates and intensity values ​​of points of the second window of neighboring points in the intensity image.

[0072] In the example of determining descriptors for the neighboring point windows compared during operation 160, operation 170 of associating features associated with the second neighboring point window of the first point matrix with corresponding features associated with the second neighboring point window of the second point matrix can be performed using the distance between the descriptor associated with the second neighboring point window of the first matrix and the descriptor associated with the second neighboring point window of the second matrix (e.g., whether the distance between the descriptor associated with the second neighboring point window of the first matrix and the descriptor associated with the second neighboring point window of the second matrix is ​​below a predetermined distance threshold). For example, the distance between the descriptors mentioned herein can be a Hamming distance.

[0073] In these examples, the binary robust independent elementary features method or the Lucas-Kanade method can be used to implement: operation 155 of determining descriptors of neighboring point windows; and operation 170 of associating features associated with the neighboring point windows of the first point matrix with corresponding features associated with the neighboring point windows of the second point matrix using distances between the descriptors of the neighboring point windows of the first point matrix and the descriptors of the neighboring point windows of the second point matrix.

[0074] In some examples, method 100 may include operation 180: using two neighboring point windows associated with the associated features to determine the displacement of the features belonging to the element between the first moment and the second moment. In particular, by comparing the position of the second neighboring point window on the second point matrix with its position on the first point matrix and knowing the moment associated with each of the first matrix and the second matrix, the displacement of the neighboring point window between the first point matrix and the second point matrix can be determined, the displacement corresponding to the displacement of the feature between the first moment and the second moment.

[0075] In some examples, the operation 180 of determining a displacement of a feature of an element of an environment belonging to the LiDAR sensor between a first time instant and a second time instant includes:

[0076] - determining in the intensity image a two-dimensional displacement of the feature between a first moment in time and a second moment in time using the two-dimensional coordinates and intensity values ​​of points of two second point windows in the intensity image associated with the feature; and

[0077] - determining a three-dimensional displacement of the feature between the first and second moments in time using the two-dimensional displacement in the intensity image and using the two-dimensional coordinates and depth values ​​of points of two second point windows associated with the feature in the depth image.

[0078] In the example, this includes:

[0079] - an operation of determining 155 descriptors and associating 170 using distances between the descriptors, implemented using the Lucas-Kanade method; and

[0080] - an operation 160 of determining the displacement of a feature belonging to the element between a first moment and a second moment using the three-dimensional coordinates of two adjacent point windows associated with corresponding associated features,

[0081] The displacement of a feature of an element between a first time instant and a second time instant may be determined based on minimization of a distance between descriptors of two adjacent point windows associated with corresponding associated features.

[0082] These examples make it possible to determine the displacement of features between two matrices at a scale smaller than the scale of the points of the matrix of points obtained from the acquisition of the LiDAR sensor 10 , so that the determined displacement is obtained in an extremely precise manner.

[0083] Thus, the method 100 according to the present disclosure enables the use of information acquired using the LiDAR sensor 10 to track a feature of an element of the environment of the LiDAR sensor between two point matrices and optionally determine the displacement of the feature between the two matrices. The method 100 is described for two point matrices, but the method can be implemented on more than two point matrices in order to track a feature of an element of the environment of the LiDAR sensor 10 over time and optionally determine the displacement of the feature over time in an extremely precise manner.

Claims

1. A method (100) for processing data acquired by a LiDAR sensor (10) implemented by a computer (11), the method (100) comprising: - using the acquisition of the LiDAR sensor to obtain (110) a first point matrix and a second point matrix, the points of the first matrix and the points of the second matrix respectively representing the environment of the LiDAR sensor (10) at a first moment and a second moment; each point of these point matrices is associated with a coordinate in three-dimensional space and an intensity value; - projecting (120) these point matrices onto a two-dimensional projection plane to obtain an intensity image and a depth image for each of these matrices; For each of the first point matrix and the second point matrix: * determining (130) at least one feature of an element belonging to the environment of the LiDAR sensor (10) on the matrix of points using the three-dimensional coordinates and the intensity values ​​of the points of the matrix of points, whereby the feature is associated with the points of the matrix; then For at least one determined feature on each matrix: * determining (140) a first neighboring point window in the intensity image using the two-dimensional coordinates of the points of the matrix and the intensity values ​​of the points in the intensity image, the first neighboring point window being associated with the feature and comprising the points associated with the feature; * determining (150) a second neighboring point window associated with the feature, the neighboring points of the second neighboring point window corresponding to the neighboring points of the first neighboring point window whose distance from the point representing the feature in the three-dimensional space or the depth image is less than a predetermined threshold; then - comparing the second neighboring dot window of the first dot matrix with the second neighboring dot window of the second dot matrix (160); then - using the comparison to associate (170) features associated with a second neighboring window of dots of the first matrix of dots with corresponding features associated with a second neighboring window of dots of the second matrix of dots.

2. The method according to the preceding claim, further comprising: - determining (180) a displacement of a feature of an element of the environment of the LiDAR sensor (10) between the first moment and the second moment using two second neighboring point windows associated with corresponding associated features.

3. The method according to the preceding claim, wherein: Determining the displacement of a feature of an element of the environment of the LiDAR sensor (10) between the first moment and the second moment comprises: - determining in the intensity image a two-dimensional displacement of the feature between the first moment and the second moment using the two-dimensional coordinates and intensity values ​​of points of two second point windows associated with the feature in the intensity image; and - determining a three-dimensional displacement of the feature between the first moment and the second moment using the two-dimensional displacement in the intensity image and the two-dimensional coordinates and depth values ​​of points of two second point windows associated with the feature in the depth image.

4. The method according to any one of the preceding claims, further comprising: for a second neighboring point window associated with the feature, determining (155) a descriptor for the second neighboring point window, the descriptor corresponding to a value of the feature of the neighboring point window determined using coordinates and intensity values ​​of points of the second point window in the intensity image; and In which, the distance between the descriptor associated with the second neighboring point window of the first matrix and the descriptor associated with the second neighboring point window of the second matrix is ​​used to associate the features related to the second neighboring point window of the first point matrix with the corresponding features related to the second neighboring point window of the second point matrix.

5. The method according to the preceding claim, wherein: The binary robust independent basic feature method is used to implement: determining a descriptor of a second neighboring point window; and associating features associated with the second neighboring point window of the first point matrix with corresponding features associated with the second neighboring point window of the second point matrix using a distance between the descriptor of the second neighboring point window of the first point matrix and the descriptor of the second neighboring point window of the second point matrix.

6. The method according to claim 4 in combination with claim 2, wherein: The Lucas-Kanade method is used to implement: determining a descriptor of a second neighboring point window; and associating features associated with the second neighboring point window of the first matrix of points with corresponding features associated with the second neighboring point window of the second matrix of points; and in, The displacement of the feature of the element between the first moment and the second moment is determined based on minimization of the distance between the descriptors of two adjacent point windows associated with the corresponding associated feature.

7. A computer program product comprising instructions for implementing any of the methods of claims 1 to 6 when the program is executed by a processor. 8 . A computer-readable non-transitory storage medium having stored thereon a program for implementing any one of the methods of claims 1 to 6.

9. A computer (11) configured to implement the method according to any one of claims 1 to 6.

10. A motor vehicle (2) comprising a computer (11) as claimed in the preceding claim.

Images