Lidar sensor data processing

Abstract

The present disclosure relates to a device for processing data from a LIDAR sensor mounted on a vehicle, the data processing device making it possible to determine whether a group of pixels of an array of pixels acquired by the LIDAR sensor corresponds to a particle cloud, in particular from the distribution of normal vectors associated with said pixels.

Description

[Technical field]

[0001] The present invention relates to a device and method for processing LIDAR sensor data. [Background technology]

[0002] A LIDAR (Light Detection and Ranging) sensor is a sensor that emits light waves and, based on the reflection of these waves, determines a matrix that maps the LIDAR sensor's environment.

[0003] The environmental matrix of the lidar sensor may include airborne particles, particularly from exhaust gases, if the lidar sensor is mounted on a vehicle.

[0004] In connection with a lidar sensor placed on a vehicle traveling on a road network, there are a large number of airborne particles, in particular due to exhaust gases from the various vehicles, which can interfere with a correct interpretation of the lidar sensor's environmental matrix.

[0005] The solution proposed in US8818609 involves determining whether a zone of points acquired by a lidar sensor corresponds to exhaust gas or to a solid object based on the density and height profiles of these points. More specifically, the density and height profiles are compared to previously determined profiles and a classifier is used to determine whether it is an exhaust gas or a solid object based on the comparison.

[0006] However, this solution has limitations insofar as solid objects may also be contained in the exhaust gases, and in this case it proves difficult to distinguish between the exhaust gases and the solid objects.

[0007] In this sense, there is therefore scope for improving the accuracy in detecting airborne particles and in distinguishing objects within these particles. Summary of the Invention [Problem to be solved by the invention]

[0008] A first object of the present disclosure therefore involves proposing an alternative device and method adapted to detect whether a group of points (pixels) acquired by a lidar sensor adapted to be mounted on a vehicle corresponds completely or partially to a cloud of particles.

[0009] Another object of the present disclosure includes enabling the device and method to detect, within a group of pixels previously identified as belonging wholly or partially to a particle group, whether a pixel corresponds to a solid element other than a particle. [Means for solving the problem]

[0010] In this regard, the present disclosure provides a data processing device for a lidar sensor adapted to be mounted on a vehicle, the device comprising: - acquiring a matrix of pixels acquired by the lidar sensor, each pixel of the matrix being associated with a light intensity and a position in three-dimensional space; - identifying at least one group of adjacent pixels of the matrix of pixels; For at least one identified group of adjacent pixels, determining a plurality of normal vectors associated with a pixel in a group of adjacent pixels; * Identifying whether a group of adjacent pixels belongs fully or partially to a particle group based on a substantially uniform distribution of a plurality of normal vectors associated with pixels in the group of adjacent pixels. A data processing device is described, comprising a computer configured to:

[0011] Optionally, the computer is also configured to determine, for at least one particular pixel of the group of adjacent pixels identified as belonging wholly or partially to the particle group, whether the particular pixel belongs to a solid element other than the particle group.

[0012] Optionally, the computer is configured to determine that a particular pixel belongs to a solid element if an intensity associated with the pixel is greater than a determined intensity threshold associated with the solid element.

[0013] Optionally, the computer is further configured to determine an average distance of a group of adjacent pixels to the LIDAR sensor. A group of adjacent pixels may be a candidate identified as belonging fully or partially to the particle cloud if the average distance of pixels of the group of adjacent pixels to the LIDAR sensor is less than a predetermined distance threshold.

[0014] Optionally, the computer is further configured to determine an average intensity of the group of adjacent pixels and an average distance of the group of adjacent pixels to the lidar sensor, where the group of adjacent pixels may be a candidate for being identified as belonging fully or partially to the particle group if the average intensity of the pixels in the group of adjacent pixels is less than the determined intensity threshold, with the intensity threshold determined based on the average distance of the group of pixels.

[0015] Optionally, the computer is further configured to determine an azimuth angle for each pixel in the group of adjacent pixels. A group of adjacent pixels may be a candidate identified as belonging fully or partially to a particle group if a distribution of azimuth angles of pixels in the group of adjacent pixels substantially corresponds to a predetermined model.

[0016] Optionally, the computer is configured to identify a group of adjacent pixels as corresponding to a plurality of pixels, each pixel being a distance less than a predetermined adjacency threshold from another pixel of the group of pixels.

[0017] According to another aspect, the present disclosure describes a vehicle fitted with a data processing device according to any of the options described above, and the computer of the data processing device is configured to identify a particle cloud comprising exhaust gas particles.

[0018] According to another aspect, the present disclosure provides a method for processing data from a lidar sensor mounted on a vehicle, the method comprising: - obtaining a matrix of pixels acquired by the LIDAR sensor, each pixel of the matrix being associated with a light intensity and a position in three-dimensional space; - identifying at least one group of adjacent pixels of a matrix of pixels; For at least one identified group of adjacent pixels, determining a plurality of normal vectors associated with a pixel in a group of adjacent pixels; determining whether a group of adjacent pixels belongs completely or partially to a particle group based on a substantially uniform distribution of a plurality of normal vectors associated with pixels in the group of adjacent pixels; A method is described, comprising:

[0019] Optionally, the method further comprises: determining, for at least one particular pixel of the group of adjacent pixels, whether the particular pixel belongs to a solid element other than a particle group.

[0020] The present disclosure also describes a computer program product that includes instructions for performing any one of the methods described herein when implemented by a computer, and a non-transitory computer-readable storage medium having stored thereon code instructions for performing any one of the methods described herein.

[0021] Further features, details and advantages will become apparent upon reading the following detailed description and examining the accompanying drawings. [Brief description of the drawings]

[0022] [Figure 1a] 1 illustrates an example of a data processing device. [Figure 1b] 2 illustrates another example of a data processing device. [Diagram 2] 1 shows an example of a lidar sensor mounted on a vehicle. [Diagram 3] 2 shows an example of various blocks configured to be controlled by a computer of a data processing device. [Figure 4] An example of a group of adjacent pixels and an example of several subsets of pixels are shown. [Figure 5a] 1 illustrates an example of a uniform distribution of pixel normal vector directions among a group of pixels. [Figure 5b] 1 illustrates an example of a non-uniform distribution of pixel normal vector directions among a group of pixels. [Figure 6a] 4 shows an example of a distribution of azimuth angles of pixels among a group of pixels. [Figure 6b] 4 illustrates another example of a distribution of azimuth angles of pixels among a group of pixels. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0023] The present disclosure describes a data processing device for a lidar sensor, comprising a computer configured to perform various operations, which are used in particular to identify whether a group of pixels obtained from the lidar sensor and describing the lidar's environment belongs fully or partially to a particle group. Optionally, it is also possible to determine whether a particular pixel of a group of adjacent pixels identified as belonging fully or partially to a particle group belongs to a solid element other than the particle group.

[0024] These possibilities afforded by the processing device are particularly advantageous, especially in the context of autonomous vehicle traffic. Indeed, the LIDAR sensor mounted on the vehicle allows the environment around the vehicle to be mapped. This mapping can include information related in particular to particles or solid elements. In particular, the particles on the LIDAR matrix can correspond, for example, to the exhaust gases of the vehicle, in particular the vehicle on which the LIDAR is mounted. With regard to the solid elements, the solid elements on the LIDAR matrix can represent obstacles, for example other vehicles around the vehicle on which the LIDAR is mounted.

[0025] However, it should be understood that from the perspective of the lidar, the acquired information merely represents a collection of pixels, whether they are particles or solid elements other than particles, but with respect to the processing functions that use the information received from the lidar, and in particular the navigation functions of an autonomous vehicle, the particles should not be interpreted as obstacles, since unlike solid elements, they do not pose any danger to the occupants and the vehicle.

[0026] The processing device described in the present disclosure is thus suitable for implementing a solution that can identify groups of pixels that may correspond wholly or partially to groups of particles, and optionally distinguish from among these groups pixels that correspond to solid elements other than particles.

[0027] In this case, the described solution is a pre-processing solution in the sense that more thorough processing may be performed following the solution to ensure that the solid elements detected by the processing device do not represent "false positives". Thus, in the described solution, groups of pixels that are specifically suspected to represent solid elements may be carefully differentiated so that they can be directly processed using such more thorough processing. This maintains the ability to detect solid elements that may compromise the physical integrity of occupants and damage the vehicle.

[0028] Reference is now made to FIGS. 1 a and 1 b, which show an example of a data processing device 1 for a lidar sensor 2.

[0029] The data processing device 1 may comprise a computer 3 adapted to execute code instructions enabling the data processing device to control several data processing blocks associated with the lidar sensor 2.

[0030] The code instructions may for example be stored in a memory 31 accessible to the computer 3. The computer 3 may for example be a processor, controller or microcontroller. The computer 3 is thus connected to the memory 31 so as to be able to use the information contained in the memory 31.

[0031] The memory 31 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 storage means suitable for reading out code instructions. The memory may for example comprise optical, electronic or even magnetic storage means.

[0032] In Fig. 1a, an example of a data processing device 1 comprises a lidar sensor 2, a computer 3, and a memory 31. In particular, in the illustrated example, the lidar sensor 2 comprises a computer 3 and a memory 31. Thus, a non-limiting example of a data processing device may correspond to a lidar sensor 2 comprising a computer 3 and a memory 31. In this embodiment, the lidar sensor 2 may be physically connected to the memory 31 of the computer 3 such that data acquired by the lidar sensor following emission of a light beam are transmitted to the memory 31 by a wired connection to be stored. Thus, data processing contained in the memory 31 and controlled by the computer 3 based on the data described later by the present disclosure may be performed locally on the lidar sensor.

[0033] Another example of a data processing device 1 is shown in Fig. 1b. In this example, the data processing device comprises a lidar sensor 2, which may be connected to a computer 3 having a memory 31 via an electrical communication channel. The lidar sensor 2 and the computer 3 having the memory 31 may thus be remote, and data acquired by the lidar sensor 2 is transmitted to the computer 3 by the electrical communication channel to be stored in the memory 31. The computer 3 and the memory 31 may for example form part of a remote server. The electrical communication channel may for example be 3G, 4G, 5G, optical, electromagnetic etc.

[0034] In this case, if the lidar sensor 2 is mounted on the vehicle 10, advantageously at the rear of the vehicle 10, the data can be processed by a computer 3 located directly on the vehicle, as shown in FIG. 2. In this figure, the arrow D represents the direction of travel of the vehicle 10 and the arrow O represents the optical axis of the lidar sensor. In this configuration, the computer can also be understood as the lidar sensor, as in the example shown in FIG. 1a, or it can simply correspond to another computer present on the vehicle 10, as shown in FIG. 2. The computer 3 can for example be an electronic control unit of the vehicle 10. As an alternative embodiment, the lidar sensor 2 is mounted on the vehicle 10 and the computer 3 is located on a server that is remote from the vehicle 10. The data is therefore processed on the remote server.

[0035] In the following, with reference to FIG. 3, a method adapted to be executed by the computer 3 of the data processing device 1 will be described.

[0036] As shown in block 110, the computer 3 is configured to acquire a matrix of pixels p acquired by the LIDAR sensor 2. The matrix is ​​determined by the LIDAR 2 through the emission of a light beam, which is reflected and received by said LIDAR 2, allowing the formation of a matrix of pixels p. Each pixel p of the matrix is ​​assigned an intensity I corresponding to the intensity of light received by the pixel and a position (x, y, z) in three-dimensional space, where x and y are the Cartesian coordinates of the pixel p and z is the depth coordinate. The x, y and z coordinates are calculated based on the position of the pixel p on the receiver matrix of the LIDAR 2 and based on the relative distance of the pixel p with respect to the LIDAR sensor. The distance of the pixel p with respect to the LIDAR sensor is determined as a function of the time from the emission of the light beam to the reception of the beam by the LIDAR 2. The matrix of pixels is then stored in the memory 31 of the computer 3 by using an appropriate communication channel depending on the data processing device used.

[0037] As shown in block 120, the computer 3 is configured to identify at least one group of adjacent pixels of the matrix of pixels M. If the distance between two pixels of the matrix of pixels is less than a predetermined neighbor threshold, these two pixels may correspond to adjacent pixels. In one example, the computer may be configured to identify a group of adjacent pixels as corresponding to a number of pixels, each pixel being at a Euclidean distance of less than a predetermined neighbor threshold of another pixel of the group of pixels. The distance may be calculated based on the coordinates of the pixels in three-dimensional space. In this block, this includes determining one or more groups of candidate pixels that may belong completely or partially to the particle group.

[0038] When distance is referred to herein, it may include Euclidean distance, which is determined based on the coordinates of a pixel in three-dimensional space.

[0039] The data processing device computer is also configured to perform the operations of blocks 130 to 150 and various related examples thereof described below for at least one group of adjacent pixels identified at the end of block 120, and advantageously for all groups of pixels. These blocks are used to determine whether the identified group of adjacent pixels belongs wholly or partially to a particle group, and optionally whether pixels of this group of adjacent pixels belong to solid elements other than a particle group.

[0040] Thus, the processing of the data described in these blocks is applied to each group of pixels identified at the completion of block 120. However, for clarity, although these blocks are described in terms of groups of adjacent pixels, it is understood that they may be applied individually to each of the identified groups of pixels.

[0041] Thus, and as indicated by block 130, the computer 3 is configured to determine a number of normal vectors associated with pixels in the group of contiguous pixels. This step may include determining a normal vector associated with each pixel in the group of contiguous pixels.

[0042] A vector normal to a pixel is understood to be a vector normal to a local surface passing through the pixel under consideration of the group of neighboring pixels. This local surface may be approximated by an average plane passing through the pixel of the group of neighboring pixels.

[0043] The normal vector associated with a particular pixel of the group of neighboring pixels may be determined, for example, based on a subset of pixels that includes the particular pixel. In this regard, the computer may also be configured to determine, for each particular pixel of the group of neighboring pixels, a subset of pixels associated with the particular pixel.

[0044] The computer may be configured to determine the subset of pixels that includes the particular pixel based on, for example, a distance between the pixels in the group of adjacent pixels and the particular pixel.

[0045] In one embodiment, the subset of pixels may include a number of pixels that are located at a distance less than a predetermined subset distance threshold of a particular pixel, and in this regard, the particular pixel may be substantially a central pixel of the subset of pixels.

[0046] In another embodiment, the subset of pixels may include a predetermined number n of pixels, where the pixels of the subset correspond, for example, to the n-1 pixels closest to a particular pixel.

[0047] An example of a subset of pixels for determining a normal vector for a particular pixel in a group of neighboring pixels is shown in Figure 4. In particular, Figure 4 shows a group G of neighboring pixels that includes a number of pixels p.

[0048] Examples of subsets of pixels shown in FIG. 4 are subsets S1, S2, S3, S4, and S5, where the particular pixels respectively associated with each of the subsets are ps1, ps2, ps3, ps4, and ps5.

[0049] The radii of the circles representing subsets S1, S2, and S3 represent an example of a predetermined subset distance threshold. In this case, although Fig. 4 is depicted in two dimensions, it should be understood that the principle can also be applied in three dimensions, since pixels relate to three-dimensional coordinates and the depicted circles may in reality be spheres. Subsets S4 and S5 represent subsets that include a predetermined number n of pixels (seven pixels in the two-dimensional depicted example) corresponding to the n-1-th nearest pixels of a particular pixel, as far as subsets S4 and S5 are concerned.

[0050] In this case, based on the subset of pixels associated with the particular pixel, the computer may be configured to determine an average plane associated with the particular pixel, which may correspond to a plane that minimizes the distance between all pixels of the subset of pixels associated with the particular pixel, and may be determined based on the three-dimensional coordinates of each pixel of the subset of pixels.

[0051] In this embodiment, the computer may be configured to determine that a normal vector associated with a particular pixel of the group of contiguous pixels corresponds to a vector perpendicular to an average plane associated with the particular pixel. The normal vectors associated with the pixels of the group of contiguous pixels obtained at the end of block 160 may be determined, for example, using the respective average planes of each pixel of the group of pixels.

[0052] In an example, the computer 3 may be configured to calculate an average distance of the group of adjacent pixels to the LIDAR sensor. In particular, the computer may be configured to determine the average distance of the group of adjacent pixels to the LIDAR sensor based on the three-dimensional coordinates of each pixel of the group of related adjacent pixels. The average distance of the group of adjacent pixels to the LIDAR sensor may be determined, for example, based on the average of the individual Euclidean distances of each pixel of the group to the LIDAR sensor. The average distance may then be used to refine a specific diagnosis of whether the group of adjacent pixels belongs to a particle group or not. In this case, the group of adjacent pixels cannot form all or part of a particle group based on a specific distance, because at this distance, the refinement of the particles does not allow the particles to reflect the light emitted by the LIDAR sensor sufficiently for said sensor to detect it.

[0053] In an example, the computer 3 may be configured to determine an average intensity of the group of pixels. The average intensity may correspond to the sum of the intensities of each pixel in the group of adjacent pixels divided by the number of pixels in the group of adjacent pixels. The average intensity may then be used to refine the diagnosis for identifying whether the group of pixels belongs to a particle group or not. In fact, particles are so small that they slightly reflect the light beam of the lidar sensor, and the average intensity of the group of adjacent pixels is used to determine whether this group of adjacent pixels may belong completely or partially to a particle group.

[0054] In an example, the computer 3 may be configured to determine an azimuth angle for each pixel of a group of adjacent pixels relative to the LIDAR sensor. The azimuth angle of a particular pixel is defined as the angle between a direction on the horizontal plane corresponding to the optical axis of the LIDAR sensor and a direction of a line projected onto the horizontal plane that connects the LIDAR and passes through the particular pixel. In an example, the azimuth angle associated with a pixel may be expressed as radians and may range from -π / 2 to π / 2. The azimuth angle for the pixel is determined based on a position (x, y, z) in three-dimensional space. The distribution of azimuth angles associated with the pixel of the group of pixels may then be used to refine a diagnosis for identifying whether the group of pixels belongs to a particle group or not.

[0055] As shown in block 140, the computer 3 is configured to determine whether the group of adjacent pixels belongs completely or partially to the particle group. This determination is made based on the distribution of normal vectors associated with pixels of the group of adjacent pixels. The distribution of normal vectors associated with pixels of the group of adjacent pixels advantageously includes all pixels of the group of adjacent pixels. The distribution of normal vectors advantageously relates to the direction (orientation) of the normal vectors of the group of adjacent pixels.

[0056] The distribution of the directions of the normal vectors of the group of adjacent pixels extends between a first direction D1 and a last direction Dn, corresponding to the directions of the normal vectors associated with the first pixel and the second pixel of the group of adjacent pixels, respectively. This distribution includes several ranges of directions that intersect with the distribution between the first direction D1 and the last direction Dn of the multiple ranges of directions, as shown in Figures 5a and 5b. In the example, each range of directions allows the same deviation of the direction between its two limits.

[0057] In one embodiment, a group of adjacent pixels is identified as belonging fully or partially to a particle group if the distribution of directions of a plurality of normal vectors associated with pixels in the group of adjacent pixels is substantially uniform.

[0058] A substantially uniform distribution of the directions of the normal vectors associated with pixels of a group of adjacent pixels corresponds to a substantially smooth, substantially flat distribution of probabilities, in other words a distribution in which no significant bias in the probability of finding a pixel in a first range of directions is observed compared to a second range of directions. In other words, in the distribution of a given group of adjacent pixels, a substantially equal number of pixels of the group are included in each range of directions of the normal vectors of the distribution. In the present application, pixels included in the range of directions of the normal vectors refer to pixels whose directions of the normal vectors are between the two directions that form the range of directions, i.e. the direction Dn-1 and the direction Dn. Thus, a completely uniform distribution of directions of the normal vectors means that, if a pixel is selected randomly among the pixels of the group of adjacent pixels used in the distribution, this pixel has exactly the same probability of belonging to each of the ranges of directions represented in said distribution.

[0059] In this case, the inventors cleverly use the fact that a particle swarm, when dispersed in an open environment, reaches its maximum entropy state almost instantly, which corresponds, from a microscopic point of view, to finding equal probability per particle in all the different possible states associated with the particle swarm. This therefore means that for each pixel of a group of adjacent pixels, and in terms of the direction of the normal vector, the pixels of the group of adjacent pixels have an equal distribution of probability in each range of directions of a plurality of ranges of directions. Thus, if the distribution of directions of the normal vector of a group of adjacent pixels is substantially uniform, this group of pixels is identified as belonging completely or partially to the particle swarm.

[0060] An example of uniform distribution of vector directions is shown in Fig. 5a, and an example of non-uniform distribution of vector directions is shown in Fig. 5b. The abscissa axis represents various ranges D1 / D2 to Dn-1 / Dn of the directions of the normal vectors of the group of adjacent pixels, and the ordinate axis represents the probability of occurrence P of a pixel. The range of directions of the distribution is therefore related to the probability of occurrence of a pixel in this range. In this case, the height P(D1 / D2) of the rectangle shown in the figure corresponds to the probability of selecting a pixel associated with a normal vector whose direction lies between directions D1 and D2 from among the pixels of the group of adjacent pixels.

[0061] The probability of occurrence associated with a particular range of orientations is equal to the number of pixels in the group of adjacent pixels used for the distribution and associated with a normal vector whose orientation is between the limits of the particular range, divided by the total number of pixels in the group of adjacent pixels used for the distribution. For the probability of occurrence P(D1 / D2), it is equal to the number of pixels in the group of adjacent pixels used for the distribution and associated with a normal vector whose orientation is between the orientations D1 and D2, divided by the total number of pixels in the group of adjacent pixels used for the distribution.

[0062] In the example of Figure 5a, the probability of selecting a pixel from among the pixels used in the distribution in each of the range of directions is equal. This is a perfectly uniform distribution of normal vector directions for a group of adjacent pixels. This case is a theoretical case that shows what a uniform distribution of normal vector directions would look like, and what a theoretical distribution of normal vectors for particles in a particle swarm would look like.

[0063] Conversely, Fig. 5b shows the difference in the probability of finding a pixel in various ranges of distribution directions where the difference in probability between the ranges is large. Thus, Fig. 6b shows a non-uniform distribution of the normal vector directions of a group of adjacent pixels.

[0064] Since the normal vector is calculated based on the pixel's three-dimensional coordinates, there are an infinite number of possible directions (possible states) of the normal vector of a group of adjacent pixels, so the direction of the normal vector is shown by a range between two directions in Figures 5a and 5b. In this sense, the distribution of each direction of the normal vector of a group of pixels without defining a priori ranges may not be usable. In fact, it is possible to find different directions for each pixel (since there are an infinite number of directions) and to obtain a uniform distribution with a probability that each direction is 1 divided by the total number of pixels in the group of adjacent pixels, while the distribution of pixels over several ranges of directions will bring completely different results.

[0065] In an example, a distribution of vector directions is substantially uniform if the variance of pixel occurrence probabilities in various ranges of normal vector directions for a group of pixels is less than a predefined directional variance probability threshold.

[0066] In an example, the uniform distribution of the normal vectors of a group of pixels may be evaluated based on a directional uniformity score. The directional uniformity score may be determined according to the different occurrence probabilities of the pixels over the range and according to the number of pixels considered in the distribution. In one example, if the directional uniformity score is greater than a directional uniformity predefined threshold, the directional distribution of the normal vectors of the group of adjacent pixels is considered to be uniform.

[0067] Thus, the processing device described in the present disclosure, and in particular the computer, is configured to identify whether a group of adjacent pixels belongs fully or partially to a particle group based on the distribution of directions of normal vectors associated with pixels in the group of adjacent pixels.

[0068] Below, options are described to aid in the diagnosis of identifying groups of pixels that belong completely or partially to a particle cloud, which may be used, among other things, to exclude groups of adjacent pixels before computing the distribution of normal vectors, e.g. to prioritize computer resources to another group of adjacent pixels.

[0069] Thus, in an example, the computer 3 may also be configured to determine that a group of adjacent pixels is a candidate for being identified as belonging fully or partially to a particle group if the average distance of the pixels of the group of adjacent pixels to the LIDAR sensor is less than a predetermined distance threshold. The term "candidate" should be understood to mean that a group of adjacent pixels is protected for calculating the distribution of normal vectors to determine whether this group of pixels belongs to a particle group or not.

[0070] As mentioned above, a group of pixels cannot form all or part of a particle cloud based on a certain distance because at this distance, the fineness of the particles does not allow the particles to sufficiently reflect the light emitted by the lidar sensor. In this regard, if the average distance of pixels in a group of adjacent pixels is greater than a predetermined distance threshold, this group of adjacent pixels may contain solid elements other than particle clouds and can be directly processed by other more thorough processing functions so that they can be detected. This ensures the safety of the user and the vehicle, especially when the data processing device is mounted on the vehicle.

[0071] In an example, the computer 3 may also be configured to determine that a group of adjacent pixels is a candidate to be identified as belonging fully or partially to a particle group if the average intensity of the group of pixels is below a predetermined intensity threshold, which may be determined as a function of the average distance of the particle group.

[0072] In fact, and as mentioned above, because the particles are so small and contain so little matter, the light emitted by the lidar sensor 2 is only slightly reflected on the particles, and therefore the average intensity of a group of pixels that may correspond to a group of particles is low and decreases with distance.

[0073] Thus, the predefined intensity threshold may be related to the distance, and the determined intensity threshold selected for comparison may, for example, correspond to the predefined intensity threshold whose distance is closest to the average distance of the particle group. It is also possible to determine the intensity threshold related to the average distance of the particle group by linear interpolation based on the predefined intensity threshold. In fact, the intensity of a pixel decreases in the order of the square of the distance. The intensity threshold may depend, in particular, on the lidar sensor. Thus, tests on a test bench may be performed on the lidar sensor in order to predetermine various values ​​of the intensity threshold as a function of distance.

[0074] In this case, an average intensity of the group of adjacent pixels greater than the determined intensity threshold means that the group of adjacent pixels may belong completely or partially to a solid element other than a particle group. In fact, solid objects reflect more of the light emitted by the lidar sensor, which results in an increase in the intensity associated with the various pixels. In this case, the group of adjacent pixels is no longer a candidate to be identified as belonging completely or partially to a particle group, but can be directly processed by other, more thorough, processing functions that allow to detect solid elements from the group of pixels.

[0075] In an example, the computer 3 may also be configured to determine that a group of adjacent pixels is a candidate to be identified as belonging to a particle group, fully or partially, based on a distribution of azimuth angles associated with pixels of the group of adjacent pixels, the distribution of azimuth angles of the group of pixels advantageously including all pixels of the group of adjacent pixels.

[0076] In an example, a group of adjacent pixels is a candidate that is identified as belonging, fully or partially, to a particle group if the distribution of azimuth angles of the group of adjacent pixels substantially corresponds to a predefined model.

[0077] In the example, the given model includes minima at both ends of the distribution of azimuth angles of the group of adjacent pixels and a maximum substantially in the middle of the distribution of azimuth angles of the group of adjacent pixels. In other words, if the distribution of azimuth angles of the group of adjacent pixels extends between a minimum angle corresponding to -π / 2 and a maximum angle corresponding to π / 2, the maximum occurrence probability of the pixel should be observed substantially at an angle corresponding to 0, while two minimum occurrence probabilities should be observed at -π / 2 and π / 2, respectively. This example is particularly shown in Fig. 6a.

[0078] In an example, the predetermined model includes a maximum probability of occurrence of a pixel in the middle of a range of azimuth angles of the pixels in the distributed group of contiguous pixels, and a gradual decrease towards both ends of the range of the intermediate probability of occurrence, in an example, reaching respective minimum probabilities of occurrence at both ends of the range of azimuth angles of the pixels in the distributed group of contiguous pixels.

[0079] An example of a distribution of azimuth angles of groups of adjacent pixels that are candidates for being identified as belonging completely or partially to a particle group is shown in Figure 6a, which distribution substantially corresponds to the predefined model.

[0080] Conversely, to illustrate the contrast between the distribution of azimuth angles for which a group of adjacent pixels may be candidates for being identified as wholly or partially belonging to a particle group and the distribution of azimuth angles for which a group of adjacent pixels may not be candidates, an example of a distribution that does not substantially correspond to the predefined model is shown in Figure 6b.

[0081] In these illustrated examples, the distribution includes pixels whose azimuth angles extend between a minimum angle corresponding to -π / 2 and a maximum angle corresponding to π / 2. However, it should be noted that these are merely examples and the angular range of azimuth angles covered by a group of adjacent distributed pixels may be smaller than that illustrated. In this case, the predefined model is adapted depending on the angular range covered by the group of adjacent pixels.

[0082] In an example, a distribution of azimuth angles associated with pixels of a group of neighboring pixels may be evaluated based on a matching score to a predefined model. The matching score may be determined based on a comparison between the predefined model and the distribution of azimuth angles of the group of neighboring pixels. In one example, if the matching score is greater than a predefined matching threshold, the group of neighboring pixels with the completed distribution of azimuth angles is a candidate to be identified as belonging fully or partially to a particle group.

[0083] Thus, at the end of block 140, the data processing device is used to identify whether a group of pixels acquired based on the matrix of pixels of the LIDAR sensor belongs completely or partially to a particle group.

[0084] This identification facilitates the interpretation of the matrix of pixels of the lidar sensor by the lidar sensor mounted on the vehicle, in particular in relation to driving assistance functions, and more precisely in relation to functions for making the vehicle navigate autonomously.

[0085] In fact, particles such as exhaust gases should not be taken into account by the vehicle's driving functions unless they are dangerous for the occupants or the vehicle. In this sense, particles detected by the lidar sensor can be troublesome, since they are considered dangerous objects by other functions that use the lidar sensor matrix. In this sense, identifying such particles ultimately makes it possible to improve the fluidity of driving and reduce the risk to the vehicle occupants and the vehicle itself.

[0086] Optionally, and as indicated by the dashed line in block 150, the computer may also be configured to determine, for at least one particular pixel of the group of contiguous pixels identified as belonging wholly or partially to the particle group, whether the particular pixel belongs to a solid element other than the particle group, which block is advantageously completed for all pixels of the group of contiguous pixels identified as belonging wholly or partially to the particle group.

[0087] Several additional examples for determining whether a particular pixel belongs to a solid element other than the particle group can be implemented. Each of the examples can be implemented independently of or in addition to other examples to confirm or not identify a particular pixel as belonging to a solid element other than the particle group. In these examples, the computer can be configured to determine that a pixel in a group of adjacent pixels that has not been determined to correspond to a solid element other than the particle group corresponds to a particle of the particle group.

[0088] Thus, in an example, the computer may be configured to determine that a particular pixel of a group of adjacent pixels belongs to a solid element other than a particle group based on an intensity associated with that pixel.

[0089] In these examples, the computer knows that a particular pixel is - belongs to the particle group if the intensity associated with the pixel is below a determined intensity threshold associated with a solid element, or belongs to a solid element if the intensity associated with the pixel is greater than the determined intensity associated with the solid element It may be configured to determine that:

[0090] The intensity thresholds corresponding to the solid elements may be determined according to the distance of the particular pixel to the lidar sensor. In this sense, the determined intensity thresholds associated with the solid elements may be related to distance since the intensity decreases with distance. In an example, the determined intensity thresholds associated with the solid elements are determined to be those whose distances are closest to the distance of the particular pixel to the lidar. In another example, the determined intensity thresholds associated with the solid elements are determined by linear interpolation based on predefined intensity thresholds associated with the solid elements, which are also related to the respective distances.

[0091] As mentioned above, solid elements other than particles reflect the light emitted by the lidar more than particles, so it is possible to distinguish between pixels corresponding to solid elements and pixels corresponding to particles in a group of adjacent pixels.

[0092] In an example, the computer may be configured to determine whether a particular pixel in the group of adjacent pixels belongs wholly or partially to a solid element other than a particle group based on a distribution of azimuth angles associated with the pixel in the group of adjacent pixels.

[0093] In these examples, the computer may be configured to determine that a particular pixel belongs to a solid element other than a particle group if the azimuth angle associated with the particular pixel falls within a range of azimuth angles of a distribution that deviates from a predetermined model.

[0094] In fact, the given model of distribution corresponds to a theoretical representation model of the azimuth angles of the distribution of the particles of the particle swarm. In this regard, pixels having azimuth angles that belong to a zone that deviates from this model may therefore belong to solid elements other than particles.

[0095] Thus, the computer may be configured to determine that a particular pixel belongs to a solid element if the azimuth angle associated with that particular pixel is within a range of azimuth angles of a distribution that deviates from a predetermined model by at least a predetermined azimuth angle deviation threshold.

[0096] It should be appreciated that this determination may be combined with a determination based on the intensity of the pixel, in particular to determine that a pixel in a group of adjacent pixels belongs to a solid element other than a particle group. Thus, a pixel having an intensity greater than the solid element threshold but not falling within the azimuth angle range of the distribution that deviates from the predetermined model may not be identified as belonging to a solid element. Conversely, a pixel having an azimuth angle within the azimuth angle range of the distribution that deviates from the predetermined model but whose intensity is less than the solid element threshold may not be identified as belonging to a solid element.

[0097] In an example, the computer may be configured to determine whether a particular pixel of a group of adjacent pixels belongs to a solid element other than a particle group based on the distribution of normal vectors performed at the end of block 140.

[0098] In these examples, the computer may be configured to determine that a particular pixel belongs to a solid element other than a particle group if the direction of the normal vector associated with the particular pixel belongs to a range of directions having a greater probability of occurrence of the pixel than other ranges of the distribution.

[0099] Indeed, initially, groups of adjacent pixels were determined to belong completely or partially to a particle swarm based on a substantially uniform distribution of the vector normal vector directions. It is now possible to determine to what extent the distribution of directions deviates from the theoretical distribution of a particle swarm (perfectly uniform), thereby finding pixels belonging to solid elements that perturb the distribution so that it is no longer perfectly uniform.

[0100] Thus, the computer may be configured to determine that a particular pixel belongs to a solid element other than a particle if the direction of the normal vector associated with the particular pixel belongs to a range of directions having a probability of occurrence of the pixel greater than another range of a distribution of at least one predetermined directional deviation threshold.

[0101] In an alternative or additional example, the computer may also be configured to determine that a particular pixel belongs to a solid element other than a particle if the direction of a normal vector associated with the particular pixel falls within a range of directions whose probability of occurrence is greater than a particular probability of occurrence of the pixel.

[0102] A particular probability of occurrence of a pixel may be determined, for example, based on an average of the pixel's probabilities of occurrence in the range of the distribution and based on a comparison with this average.

[0103] As with the previous examples, determining that a particular pixel belongs to a solid element based on the direction of the normal vector may be combined in any order with determinations based on the distribution and / or intensity of the azimuth angles.

[0104] Thus, at the end of block 150, in addition to having determined which groups of pixels in the matrix of pixels belong wholly or partially to particle groups, the data processing device can determine, within each group of pixels, which pixels belong to particle groups and which pixels belong to solid elements other than particle groups.

[0105] In this regard, even if a solid object may be combined with a particle cloud, the data processing device can identify, down to the nearest pixel, which information forms part of the particle cloud and which information forms part of another solid object.

[0106] It will be appreciated that in the context of a lidar sensor mounted on a vehicle, being able to distinguish which pixels belong to particles and which pixels belong to solid objects other than particles can, for example, improve the efficiency and performance capabilities of driving assistance functions. Indeed, the data processing device thus described allows for identifying information acquired by the lidar sensor that is unrelated to the movement of the vehicle (i.e. particles) without combining this unrelated information with other relevant information (solid objects other than particles), even if these two types of information are combined in the matrix of pixels of the lidar sensor.

[0107] In this sense, the computer 3 of the processing device may also be configured to select pixels of the matrix of pixels identified as belonging to solid objects and / or to remove pixels identified as belonging to particle groups, for example to transmit only information relevant to the journey to functions that support said journey, or simply to filter out irrelevant information in order to speed up subsequent processing of the matrix of pixels acquired by the lidar sensor and used in other functions.

[0108] The present disclosure also proposes a method for processing data of a lidar sensor mounted on a vehicle. The data processing method can be controlled, for example, by the computer 3 described above. In this case, the steps performed by the method are the same as the blocks and examples described above, configured to be controlled by the computer of the data processing device. Thus, both the data processing device and the method described in the present application make it possible to identify pixels corresponding to suspended particles in the matrix of pixels of the lidar sensor. They also make it possible, in embodiments, to distinguish pixels corresponding to solid objects other than particles in groups of pixels that correspond completely or partially to particle groups. These features are particularly advantageous in the context of a lidar 2 mounted on a vehicle, in particular for identifying particle groups due to exhaust gases, and for distinguishing solid objects from particle groups as being another vehicle, when groups and objects are combined in the matrix of pixels.

Claims

1. A data processing device (1) for a lidar sensor (2) adapted to be mounted on a vehicle (10), said device comprising: obtaining (110) a matrix of pixels (p) acquired by the lidar sensor (2), each pixel (p) of the matrix being associated with a light intensity (I) and a position (x, y, z) in three-dimensional space; identifying (120) at least one group of adjacent pixels in said matrix of pixels; For the identified at least one group of adjacent pixels, determining (130) a plurality of normal vectors associated with pixels in said group of adjacent pixels; determining whether the group of adjacent pixels belongs wholly or partially to a particle group based on a substantially uniform distribution of the normal vectors associated with the pixels in the group of adjacent pixels (140); A data processing device (1) comprising a computer (3) configured to:

2. 2. The device (1) according to claim 1, characterized in that the computer (3) is also configured to determine (150) for at least one particular pixel of the group of adjacent pixels identified as belonging completely or partially to a particle group whether the particular pixel belongs to a solid element other than the particle group.

3. 3. The device (1) according to claim 2, characterized in that the computer (3) is configured to determine that the particular pixel belongs to a solid element if the intensity associated with the pixel is greater than an intensity threshold associated with the determined solid element.

4. 2. The device (1) of claim 1, wherein the computer (3) is further configured to determine an average distance of the group of adjacent pixels to the LIDAR sensor (2), and wherein a group of adjacent pixels is identified as a candidate to belong, fully or partially, to the particle group if the average distance of the pixels of the group of adjacent pixels to the LIDAR sensor (2) is less than a predetermined distance threshold.

5. the computer (3) is further configured to determine an average intensity of the group of adjacent pixels and an average distance of the group of adjacent pixels relative to the lidar sensor (2); 2. The device (1) of claim 1, wherein a group of adjacent pixels is a candidate for being identified as belonging wholly or partially to the particle group if the average intensity of the pixels of the group of adjacent pixels is less than a determined intensity threshold, the intensity threshold being determined based on the average distance of the group of pixels.

6. 2. The device (1) of claim 1, wherein the computer is further configured to determine an azimuth angle for each pixel in the group of adjacent pixels, and wherein a group of adjacent pixels is identified as a candidate for belonging to the particle group, wholly or partially, if a distribution of azimuth angles of pixels in the group of adjacent pixels substantially corresponds to a predetermined model.

7. 2. The device (1) according to claim 1, characterized in that the computer (3) is configured to identify groups of adjacent pixels as corresponding to a plurality of pixels, each pixel being at a distance of less than a predetermined adjacency threshold from another pixel of the group of pixels.

8. A vehicle (10), characterized in that the vehicle (10) is fitted with a data processing device (1) according to any one of claims 1 to 7, a lidar sensor (2) is mounted on a rear part of the vehicle, and the computer (3) of the data processing device (1) is configured to identify particle groups including exhaust gas particles.

9. A method for processing data from a lidar sensor (2) mounted on a vehicle (10), comprising: acquiring (110) a matrix of pixels (p) acquired by the lidar sensor (2), each pixel (p) of the matrix being associated with a light intensity (I) and a position (x, y, z) in three-dimensional space; Identifying (120) at least one group of adjacent pixels in said matrix of pixels; For the identified at least one group of adjacent pixels, determining (130) a plurality of normal vectors associated with pixels in said group of adjacent pixels; determining whether the group of adjacent pixels belongs wholly or partially to a particle group based on a substantially uniform distribution of the normal vectors associated with the pixels in the group of adjacent pixels (140); A method comprising:

10. 10. The method of claim 9, further comprising determining (150) for at least one particular pixel of the group of adjacent pixels identified as belonging wholly or partially to a particle group whether the particular pixel belongs to a solid element other than the particle group.

11. A computer program product comprising instructions for performing any one of the methods according to claims 9 or 10 when implemented by a computer.

12. A non-transitory computer readable storage medium storing code instructions for performing the method of claim 9 or 10.

Images