Method for detecting the environment of a vehicle

Abstract

Method for detecting the environment of a vehicle. The invention relates to a method for detecting (1) the environment of a vehicle, comprising: a step of acquiring (2) a point cloud (NP) of the environment of said vehicle (17); a step of filtering (4) the point cloud (NP) to form a filtered point cloud (NF); a step of dividing (6) the filtered point cloud (NF) into distinct zones (Z); a step of determining (8) a point density (D) within each zone (Z) and calculating an average density ( ) of the zones; a selection step (10), in which at least one of the zones (Z) is selected when its point density (D) is greater than the average density ( ); a step of analyzing (14) the at least one selected zone (ZS) to detect the presence of objects. (Figure 1)

Description

Title of the invention: Method for detecting the environment of a vehicle

[0001] The present invention relates to the field of vehicles and, more particularly, to the field of methods for detecting an environment around a vehicle.

[0002] Modern vehicles, especially those with a high degree of autonomy, such as autonomous or semi-autonomous vehicles, must be able to perceive and understand their environment to allow safe driving and compliance with the highway code.

[0003] For this purpose, it is necessary to have precise information on the environment around the vehicle, in particular information on the presence of objects such as road infrastructure or other environmental obstacles, whether these objects are static or dynamic.

[0004] In this context, vehicles can use an environmental sensing method. This type of method often uses a technology that allows the formation of one or more point clouds to model the environment around the vehicle. A technology commonly used in this field is laser remote sensing, known by its English acronym LIDAR.

[0005] The point cloud(s) thus produced allow for a detailed representation of the environment. However, when the environment contains many obstacles, the point cloud(s) may exhibit a high point density, which can pose problems during a subsequent analysis of the point cloud to search for objects in the environment.

[0006] Indeed, too high a point density can lengthen the analysis time, which may lead to late detection of some objects and may then generate an increased risk of accidents.

[0007] In order to solve this problem, some prior art processes seek to eliminate spurious points from the point cloud by applying a neighborhood generation method which evaluates the concentration of points around a given point to determine whether it is part of an object or whether it is a spurious point.

[0008] The objective of the invention described below is to offer an alternative to the existing method by presenting a simpler and faster method to implement. The invention divides the point cloud into distinct zones, these zones then being analyzed in order of priority, according to their point density. The proposed method thus makes it possible to reduce the execution time of the analysis while maintaining sufficient accuracy of said analysis, and this without resorting to neighborhood generation methods.

[0009] The main object of the present invention is therefore a method for detecting the environment of a vehicle, comprising the following steps: - a step of acquiring a point cloud of the environment of said vehicle; - a point cloud filtering step to form a filtered point cloud, said filtering step including at least one elimination of points from the point cloud corresponding to the ground of the environment; - a step of dividing the filtered point cloud into distinct areas; - a step of determining a point density within each zone and calculation of an average density of zones; - a selection step, in which the point density of each zone is compared to the average density of the zones, with at least one of the zones being selected when this point density is greater than the average density; - an analysis step of at least one selected area in order to detect the presence of objects.

[0010] The point cloud acquisition step is a step during which data concerning the vehicle's environment are obtained. This data is then structured as a point cloud, constituting a set of three-dimensional coordinates that allow the vehicle's environment to be modeled. Each point corresponds to a coordinate indicating the position of a surface or part of an object in space.

[0011] The point cloud filtering step makes it possible to obtain, from the point cloud acquired during the acquisition step, the filtered point cloud which no longer contains, or contains very few, points corresponding to the ground of the environment. To do this, the filtering step includes the elimination of points associated with the ground of the environment. This filtering can be carried out by identifying and separating the points corresponding to the ground from the other points, for example by measuring the relative height of the points.

[0012] The filtered point cloud is then divided into distinct zones during the splitting step, for example, into at least two distinct zones, although it is possible to define more. This division into zones allows the filtered point cloud to be organized to facilitate its subsequent processing. By way of non-limiting example, the zones can be formed by dividing the filtered point cloud into circular regions and / or angular sectors arranged around the vehicle.

[0013] The determination step consists of measuring the point density within each zone. In other words, the number of points contained in each zone is counted, thus establishing a point density for each zone.

[0014] During this determination step, the average density of the zones is also determined. This average density is calculated as the sum of the point densities of all the zones, divided by the total number of zones formed during the division step.

[0015] The selection step consists of classifying the zones according to their point density. The zone(s) with a point density higher than the average density are selected, as they have a higher probability of containing an object. Conversely, the zone(s) with a point density lower than the average density are discarded, as they are less likely to contain an object. Thus, this step divides the zones into two groups: a first group comprising the selected zone(s) and a second group comprising the discarded zone(s).

[0016] Finally, the analysis step examines the points contained in at least one selected area to detect the presence of objects. This analysis step may include an evaluation of the concentration of points in the selected area(s), a high concentration of points being interpreted as a likely indication of the presence of an object, while a low concentration suggests the absence of an object.

[0017] It should be noted that the method according to the invention provides that, in the case where several areas are selected, it is possible that not all of them will be analyzed, in particular when constraints, such as the duration of the analysis step, limit said analysis.

[0018] According to an optional feature of the invention, during the splitting step, the filtered point cloud is split into circular regions in order to form said zones.

[0019] More specifically, the filtered point cloud is divided into at least two circular regions. These circular regions may surround the vehicle and may be concentric with each other.

[0020] Alternatively, the filtered point cloud can be divided into irregular regions surrounding the vehicle. By irregular regions, it is appropriate to understand non-circular regions surrounding the vehicle, whether they have an angular or sinusoidal shape.

[0021] According to an optional feature of the invention, during the splitting step, the filtered point cloud is split into a central circular region and at least one peripheral circular region surrounding the central circular region in order to form said zones.

[0022] In other words, the central circular region and at least one peripheral circular region are concentric with each other. The at least one peripheral circular region thus surrounds the central circular region which, for example, may include the vehicle.

[0023] By way of non-limiting example, the diameters of concentric circular regions can be defined by the following mathematical sequence: Dn = 2r + (n — l)r

[0024] where r represents the radius of the central circular region, and n is an index used to differentiate between circular regions. Thus, n = 1 corresponds to the central circular region, and n = 2 to a peripheral circular region directly surrounding the central circular region.

[0025] According to an optional feature of the invention, the method includes a step of processing the central circular region in order to detect the presence of objects, said processing step being prior to the analysis step.

[0026] This processing step is therefore performed before the analysis step and allows the points present in the central circular region to be processed, i.e., analyzed, in order to detect the presence of objects. Thus, regardless of whether the area(s) contained in the central circular region are selected or excluded, the processing step is carried out. This ensures object detection within a perimeter very close to the vehicle, regardless of the number of points present in that perimeter.

[0027] This processing step therefore allows the process of the invention to analyze the points contained in the central circular region, whether they are part of selected or excluded areas, in order to properly examine the immediate environment of the vehicle to reduce the risks associated with driving.

[0028] The analysis step then takes place and consists of examining the selected areas that were not treated during the treatment step.

[0029] According to an optional feature of the invention, during the splitting step, the filtered point cloud is divided into angular sectors surrounding the vehicle in order to form the zones.

[0030] By angular sector, we mean parts of the vehicle environment, which may correspond, for example, to a front or rear part of the vehicle environment.

[0031] By way of example, angular sectors can be defined by straight lines delimiting the different parts of the environment.

[0032] It should be noted that angular sectors and circular regions can be combined to form zones.

[0033] According to an optional feature of the invention, during the splitting step, the filtered point cloud is split to form the zones, into at least:

[0034] - a front angular sector of the vehicle,

[0035] - a rear angular sector of the vehicle,

[0036] - two lateral angular sectors of the vehicle.

[0037] The forward angular sector of the vehicle corresponds to the environment located in front of the vehicle. The area(s) contained within this forward angular sector are particularly relevant to analyze, especially when driving at high speed, in order to react quickly to any obstacle in the vehicle's path.

[0038] The rear angular sector corresponds to the environment located at the rear of the vehicle. The area(s) contained within this rear angular sector are particularly useful to analyze during vehicle maneuvers, for example when reversing or when the vehicle has to stop or decelerate.

[0039] The two lateral angular sectors correspond to the environment located on the sides of the vehicle. The area(s) contained within these lateral angular sectors are particularly useful for analysis during lane changes or for monitoring other traffic lanes.

[0040] According to an optional feature of the invention, an execution time of the processing step and the analysis step is measured, the analysis step being interrupted when this execution time exceeds a maximum level.

[0041] The execution time corresponds to the time taken by the processing step and the analysis step. In other words, it is the time elapsed between the start of the processing step and the present moment, whether during the processing step or the analysis step.

[0042] By way of example, if we are still in the processing stage, the execution time corresponds to the duration elapsed from the beginning of this processing stage to the present moment. If we are in the analysis stage, the execution time corresponds to the duration of the processing stage, plus the duration elapsed from the beginning of the analysis stage to the present moment.

[0043] When the execution time becomes too long, i.e., when it exceeds the maximum level, the analysis step can be interrupted. Indeed, the method according to the invention prevents the execution time from becoming excessive, which could impair the processing of other point clouds and, consequently, lead to problems detecting objects in the environment.

[0044] According to an optional feature of the invention, the interruption of the analysis step is conditioned on a number of areas analyzed.

[0045] It should be noted that the interruption of the analysis step, when an excessively long execution time is observed, depends on the number of areas analyzed during the processing and analysis steps. When the number of areas analyzed is sufficient, i.e., when it exceeds a threshold, the analysis step can be interrupted by complete safety, as an adequate number of zones were analyzed, thus providing a sufficiently accurate view of the vehicle's environment.

[0046] On the other hand, if the number of areas analyzed is insufficient, i.e. below the threshold, the analysis step is not interrupted, because too few areas have been analyzed, which does not allow us to obtain a sufficiently precise idea of ​​the environment.

[0047] According to an optional feature of the invention, the method includes a supplementary analysis step of at least one area discarded during the selection step in order to detect the presence of objects within this discarded area, said supplementary analysis step being carried out when the execution time is less than a minimum level.

[0048] This additional analysis thus makes it possible to examine the points contained in the area(s) excluded during the object search selection step. This additional step is performed only when the execution time is below the minimum level, which means that there is sufficient time remaining to carry out the additional analysis.

[0049] Thus, this additional analysis allows, when time permits, for the analysis to continue on the excluded areas in order to obtain a more complete model of the environment than with only the analysis of the selected areas.

[0050] According to an optional feature of the invention, the minimum level is equal to half of the maximum level.

[0051] The invention also relates to a vehicle environment detection system implementing the method as described in this document.

[0052] The detection system may include different modules, each responsible for one or more steps of the process.

[0053] By way of non-limiting example, the detection system may include, in particular, an acquisition module for acquiring the point cloud. This acquisition module may, in particular, use laser remote sensing (LIDAR) technology.

[0054] The detection system may also include other modules, for example, a filtering module for filtering the point cloud to obtain a filtered point cloud. The system may also include an analysis module, responsible for the division, determination, and analysis stages, as well as the processing and supplementary analysis stages. These modules may be part of an electronic control unit mounted on the vehicle.

[0055] The invention finally relates to a computer program comprising program code instructions for the execution of the steps of the detection process according to the invention, when said program is executed on one or more processors.

[0056] Other features, details and advantages of the invention will become clearer upon reading the following description on the one hand, and the illustrative and non-limiting examples of embodiments given with reference to the accompanying drawings on the other hand, in which:

[0057] [Fig-1] is a schematic representation of a method for detecting the environment of a vehicle.

[0058] [Fig.2] is a schematic representation of zones formed during a step of division of the detection method according to the invention.

[0059] [Fig.3] is a flowchart illustrating stopping conditions of an analysis step of the process according to the invention as well as conditions triggering an activation of a complementary analysis step of the process according to the invention.

[0060] [Fig.4] is a schematic representation of a detection system putting in implements the detection method according to the invention.

[0061] The features and variants of the invention can be combined in various ways, provided they are not incompatible or mutually exclusive. In particular, variants of the invention may be conceived comprising only a selection of the features described below, isolated from the other described features, if this selection of features is sufficient to confer a technical advantage and / or to differentiate the invention from the prior art.

[0062] In the figures, the elements common to several figures retain the same reference.

[0063] Fig. 1 is a schematic representation of a method for detecting the environment of a vehicle.

[0064] This detection method 1 comprises a plurality of steps enabling an analysis of the environment near the vehicle in search of one or more objects. In the described embodiment, these steps are performed successively; however, in other embodiments, they could be performed in a different order.

[0065] The first step of this detection method 1 is an acquisition step 2 of a point cloud NP of the vehicle environment, said acquisition step 2 can be carried out by a detection system as will be described with reference to [Fig.4],

[0066] During this acquisition step 2, the data acquired on the environment are transformed into structured data, more particularly in the form of the NP point cloud.

[0067] The NP point cloud is thus a set of points, that is to say a set of three-dimensional coordinates allowing the geometry of the vehicle's environment. Each point therefore corresponds to a measurement of the position of a surface or object in space, thus allowing the detailed structure of the environment to be captured. This NP point cloud thus makes it possible to analyze volumes, identify shapes, or perform trajectory calculations.

[0068] The second step of this detection method 1 is a filtering step 4 of the point cloud NP acquired during the acquisition step 2. During this filtering step 4, the points of the point cloud NP corresponding to the ground of the environment are filtered so as to obtain a filtered point cloud NF where the points corresponding to the ground are eliminated.

[0069] In this regard, the detection method 1 can examine the points constituting the NP point cloud and categorize them to determine whether they correspond to the ground, before eliminating them if necessary. For example, the system can compare the heights of the points in the NP point cloud and eliminate points with a height that is too low, which would therefore correspond to points resulting from the ground.

[0070] The third step of this detection method 1 is a step of dividing the NF-filtered point cloud into distinct zones Z, which may or may not contain points. To this end, the NF-filtered point cloud can, by way of non-limiting example, be separated into circular regions and angular sectors, thus enabling the formation of said zones Z, as will be detailed in relation to [Fig. 2]. More particularly, the NF-filtered point cloud can be separated into a central circular region and at least one surrounding peripheral circular region, such that the central circular region and the at least one peripheral circular region are concentric with each other.

[0071] It should be noted that, in the schematic representation of [Fig. 1], the NF-filtered point cloud appears to be divided into only two distinct zones Z. This representation is, however, schematic in order to better understand the detection method 1 according to the invention. It should therefore be noted that the NF-filtered point cloud can be divided into more than two distinct zones Z, as will be described in relation to [Fig. 2].

[0072] This delimitation into Z zones allows the analysis of the NF filtered point cloud by priority, focusing on the Z zones that contain the most points and spending less time on Z zones that do not contain or contain few points.

[0073] Once the NF-filtered point cloud has been separated into distinct zones Z, a fourth step of the detection process 1 is a step of determining 8 the point density D of each of these zones Z and calculating an average density D of the zones Z.

[0074] During this determination step 8, the point density D of each zone Z is determined, so that a point density D is obtained for each zone Z formed during the division step 6. Furthermore, during this determination step 8, the The average density D of the Z zones is calculated, which allows us to obtain a single average density value D. This average density D corresponds to the average of the densities at points D determined from the Z zones.

[0075] The fifth step is a selection step 10 of the Z zones. During this selection step 10, at least one of the Z zones is selected based on its density of points D. The objective of this selection step 10 is thus to identify the Z zones with the highest concentration of points, because it is generally these Z zones that present an increased risk of harboring objects in the environment of the vehicle.

[0076] To perform this selection step 10, the data determined and calculated in the previous step are used. More specifically, the point density D of each zone Z is compared to the average density D calculated in the determination step 8. Based on this comparison, the zones Z are either selected or rejected. Thus, the zone(s) Z whose point density D is greater than the average density D are selected, while those with a point density D less than the average density D are rejected.

[0077] In other words, during the selection step 10, the Z zones are divided into two groups: the selected zone(s) ZS, with a point density D greater than the average density D, and the discarded zone(s) ZE, with a point density D less than the average density D.

[0078] The sixth step is a processing step 12 of the central circular region, previously defined and visible in [Fig.2]. This processing step 12 allows the points present in the central circular region to be analyzed in order to detect the presence of objects within said central circular region. Processing step 12 is performed regardless of whether the Z zone(s) within the central circular region were selected or excluded during selection step 10. This is because processing step 12 is systematically performed on the central circular region, as it is the region closest to the vehicle. Therefore, it is beneficial to conduct a precise analysis of the points within this region to detect any potential objects and thus prevent any risk of an accident.

[0079] The seventh step is an analysis step 14 of at least one selected zone. During this analysis step 14, the points present in the selected zone(s) ZS are analyzed to detect the presence of objects. Since several zones Z can be selected, the analysis can cover either all the selected zones ZS or only a portion of the selected zones ZS, as will be specified with reference to [Fig. 3].

[0080] An eighth step of the detection process 1 is a supplementary analysis step 16, during which the points of the excluded zones ZE during Selection step 10 are examined to look for the presence of objects in these remote areas ZE. This additional analysis is optional, which explains its representation as a dotted line on [Fig.1].

[0081] The additional analysis step 16 is carried out under certain conditions, which depend in particular on the execution time of the processing and analysis steps, as well as the number of Z zones processed and / or analyzed during these steps. These conditions will be detailed in the description of [Fig. 3].

[0082] Fig. 2 is a schematic representation of the Z zones formed during the division step 6 of the detection process 1 according to the invention.

[0083] As previously stated, the Z zones are formed during the division step 6, and in this embodiment, are defined by concentric circular regions and angular sectors. More specifically, the Z zones are formed by the central circular region RC and at least one peripheral circular region surrounding it.

[0084] In this embodiment, the Z zones are formed by a plurality of peripheral circular regions surrounding the central circular region RC. In particular, they are formed by a first peripheral circular region 1RP, which is the peripheral circular region closest to the vehicle 17, and a second peripheral circular region 2RP, which is the peripheral circular region furthest from the vehicle 17.

[0085] In other words, the central circular region RC, the first peripheral circular region 1RP, and the second peripheral circular region 2RP form a set of concentric circular regions. The central circular region RC surrounds the vehicle 17, the first peripheral circular region 1RP surrounds the central circular region RC (and therefore the vehicle 17), and the second peripheral circular region 2RP surrounds the first peripheral circular region 1RP, the central circular region RC, and the vehicle 17.

[0086] It should be noted that, in other embodiments, the Z zones could consist solely of circular regions (without necessarily being concentric) or exclusively of angular sectors.

[0087] The diameter of the concentric circular regions can be defined by the following mathematical sequence in this embodiment: - Dn = 2r+(nl)r

[0088] r is a fixed value representing the radius of the central circular region RC. This value is defined according to the size of the environment of the vehicle 17 in which objects are to be detected.

[0089] It should be noted that the value of the radius r may depend in particular on the length of the car, the location of the LiDAR on the vehicle, as well as the Specific characteristics of LiDAR, such as its minimum detection range. This radius value r must ensure the detection of an obstacle located at least 0.5 m in front of and / or behind the vehicle.

[0090] n is an index used to differentiate the central circular region RC from the peripheral circular regions. For example, in this embodiment: - n = 1 corresponds to the central circular region RC, - n = 2 corresponds to the first peripheral circular region 1RP, - n = 3 corresponds to the second peripheral circular region 2RP.

[0091] Thus, by way of example in this embodiment: - the diameter of the central circular region RC is defined by the mathematical formula D = 2r, - the diameter of the first peripheral circular region 1RP is defined by the mathematical formula D = 2r+r, - the diameter of the second peripheral circular region 2RP is defined by the mathematical formula D = 2r+r+r.

[0092] In order to form the Z zones, the filtered point cloud NF is also divided into angular sectors. More specifically, here, the filtered point cloud NF is divided into four angular sectors: a front angular sector SA, a rear angular sector SR, a first lateral angular sector 1SL, and a second lateral angular sector 2SL. The first lateral angular sector 1SL is located to the left of vehicle 17, and the second lateral angular sector 2SL is located to the right of vehicle 17 when it is moving forward.

[0093] Angular sectors can, for example, be delimited by straight lines (d), in particular by two perpendicular and intersecting straight lines (d). These two straight lines (d) intersect at a center C of the vehicle 17, so that the spaces formed between them in the environment of the vehicle 17 constitute the angular sectors.

[0094] In this embodiment, the central circular region RC and the four angular sectors thus form four central zones: a front central zone ZCA, which corresponds to the Z zone of the central circular region RC included in the front angular sector SA; a rear central zone ZCR, which is the Z zone of the central circular region RC included in the rear angular sector SR; and two lateral central zones ZCL, which correspond to the Z zones of the central circular region RC included in the two lateral angular sectors 1SL, 2SL.

[0095] The first peripheral circular region 1RP and the four angular sectors together form four first peripheral zones: a first front peripheral zone 1ZA, which is the zone of the first peripheral circular region 1RP contained within the front angular sector SA; a first rear peripheral zone 1ZR, which is the Z zone of the first peripheral circular region 1RP included in the rear angular sector SR; and the first two lateral peripheral zones 1ZL, which are the Z zones of the first peripheral circular region 1RP included in the two lateral angular sectors 1SL, 2SL.

[0096] The second peripheral circular region 2RP and the four angular sectors in turn form four second peripheral zones: a second front peripheral zone 2ZA, which is the Z zone of the second peripheral circular region 2RP included in the front angular sector SA; a second rear peripheral zone 2ZR, which is the Z zone of the second peripheral circular region 2RP included in the rear angular sector SR; and two second lateral peripheral zones 2ZL, which are the Z zones of the second peripheral circular region 2RP included in the two lateral angular sectors 1SL, 2SL.

[0097] It should be noted that, as previously stated, during processing step 12, the presence of objects in the central circular region RC is investigated by processing the points in this region, and therefore by processing the points included in: the central front zone ZCA, the central rear zone ZCR, and the two central lateral zones ZCL. This processing step 12 is performed independently of the selection or exclusion of these Z zones. In other words, the points in all the Z zones included in the central circular region RC are analyzed before analysis step 14, since these are the areas where the risk of an accident is highest in the presence of objects.

[0098] Regarding the selected zones ZS present in the peripheral regions, at least part of these selected zones ZS present in the peripheral regions is analyzed during the analysis step 14. Indeed, according to certain conditions described in the description of [Fig.3], they may not all be analyzed.

[0099] As regards the ZE outlying areas present in the peripheral regions, they are analyzed during the additional analysis step 16, this analysis step 14 being optional and taking place under certain conditions, also described in the description of [Fig.3].

[0100] Fig. 3 is a flowchart 18 illustrating the stopping conditions of the analysis step 14 as well as the conditions triggering the activation of the complementary analysis step 16.

[0101] In the first block 20 of the flowchart 18, and in parallel with the processing step 12 and the analysis step 14, the execution time T of these two steps is measured, as previously specified. In other words, the combined duration of the processing step 12 and the analysis step 14 is considered as a single continuous period, and the total time thus measured corresponds to said execution time T.

[0102] Furthermore, still within this first block 20, a number N of analyzed and / or processed Z zones is determined. This number N refers to the number of Z zones processed during the processing step 12 and the number of Z zones analyzed during the analysis step 14.

[0103] In other words, when the detection process 1 is still at the processing step 12, the number N corresponds to the number of Z zones processed during said processing step 12. On the other hand, when the detection process 1 is at the analysis step 14, the number N represents the number of Z zones processed during the processing step 12, to which is added the number of Z zones analyzed during the analysis step 14.

[0104] In a second block 22 of the logic diagram 18, the completion of the analysis of the selected zones ZS during the analysis step 14 is then checked. When the analysis of the selected zones ZS is not yet finished, a third block 24 of the logic diagram 18 compares the execution time T to a maximum level Tl of execution time T. If the execution time T is less than this maximum level Tl, this indicates that the duration of the processing step 12 and the analysis step 14 remains reasonable. The analysis step 14, and therefore the analysis of the selected areas ZS, then continues in a fourth block 26 of the flowchart 18. Conversely, if the execution time T exceeds this maximum level Tl, it means that the duration of the processing and analysis steps is too long. It is then necessary to determine whether the analysis step 14 can be interrupted. To do this, a fifth block 28 of the logic diagram 18 conditions the stopping of the analysis step 14 on the number N of zones Z processed and / or analyzed.

[0105] More specifically, within this fifth block 28 of the logic diagram 18, the number N is compared to a threshold NI of the number of zones Z processed and / or analyzed. When the number N exceeds the threshold NI, this indicates that the number of zones Z processed and / or analyzed is sufficient to offer a sufficiently complete view of the environment of vehicle 17. Therefore, the processing step 12 and / or analysis can be stopped in a sixth block 30 of the flowchart 18. On the other hand, when the number N is less than the threshold NI, this means that the areas Z processed and / or analyzed do not allow us to obtain a sufficiently complete view of the environment of vehicle 17. In this case, even when the execution time T exceeds the maximum level Tl, the analysis step 14 continues in a seventh block 32 of the logic diagram 18.

[0106] In the case where, in the second block 22 of the flowchart 18, the completion of the analysis of the selected zones ZS is verified during the analysis step 14, and this analysis is complete, an eighth block 34 of the flowchart 18 compares the execution time T to a minimum execution time level T2. This minimum level T2 can, for example, be equal to half the maximum level TL

[0107] In the case where the execution time T is greater than this minimum level T2, the additional analysis step 16 is not carried out, as indicated by a ninth block 35 of the logic diagram 18.

[0108] On the other hand, if the execution time T is less than this minimum level T2, the additional analysis step 16 of the discarded areas ZE during the selection step 10 is carried out, as indicated by a tenth block 36 of the flowchart 18.

[0109] To summarize this flowchart 18: - When analysis step 14 is not finished and the execution time T is less than the maximum level Tl, analysis step 14 continues. - When analysis step 14 is not completed, the execution time T exceeds the maximum level Tl and the number N is less than the threshold NI, analysis step 14 also continues. - In the event that the analysis step 14 is not completed, the execution time T exceeds the maximum level Tl and the number N exceeds the threshold NI, the analysis step 14 is stopped. - When analysis step 14 is completed, but the execution time T is less than the minimum level T2, a supplementary analysis is carried out on the excluded areas ZE. - In the case where the analysis step 14 is completed and the execution time T exceeds the minimum level T2, the additional analysis is not carried out.

[0110] Fig. 4 is a schematic representation of a detection system 36 implementing the detection method 1 according to the invention. The detection system 36 includes first of all an acquisition module 38. This acquisition module 38 allows the acquisition step 2 to be carried out and thus to form the point cloud NP representing the environment of the vehicle 17. For this, the acquisition module 38 can use laser remote sensing (LiDAR) technology. It then scans its environment using a set of lasers and detectors, sending laser pulses and measuring the delay, also called "time of flight", between the emission of the laser pulses and the detection of these pulses reflected by an object present in the environment. The measurements of the emitted laser pulses that bounce off objects and are then captured by the acquisition module 38 form a data block. This data block is then processed, and using calibration files, it is decoded and converted into the NP point cloud. The detection system 36 may also include a filtration module 40 designed to filter the points of the NP point cloud corresponding to the ground, in order to to produce the NF filtered point cloud, where points corresponding to the ground of the environment are eliminated. To achieve this, the filtering module 40 receives the NP point cloud produced by the acquisition module 38, segments the points corresponding to the ground from those that are not, and then publishes the filtered NF point cloud. For this purpose, it can use, for example, a segmental approach, where the filtered NP point cloud is segmented, each segment is analyzed, and the points corresponding to the ground are removed, thus producing the filtered NF point cloud. The detection system 36 may also include an analysis module 42. This module can be configured to perform the other steps of the detection process 1 according to the invention. Thus, the analysis module 42 can be configured to perform the step 6 of splitting the filtered point cloud NF, then the step 8 of determining the point density D of each zone Z and the average density D of the zones Z, followed by the step 10 of selecting the zones Z, and finally the analysis step 14. The analysis module 42 can also be configured to perform the processing step 12 and the additional analysis step 16, as well as to track the execution time T and the number N of zones Z processed and / or analyzed. Alternatively, all the steps of the detection process 1 could be carried out in a single module. The steps of the detection process 1 could also be distributed across other modules. The example described in [Fig. 4] is therefore a non-limiting example of the invention.

[0111] It should be noted that the invention also relates to a computer program that may include program code instructions for executing the steps of the detection method 1 according to the invention.

[0112] The computer program can in particular be executed on one or more processors.

[0113] As described above, the present invention achieves its objective by providing a method for detecting the environment of a vehicle. This method generates a filtered point cloud and includes a step of dividing this filtered point cloud into distinct zones. The zones are then analyzed in order of priority according to their point density, which makes it possible to target the zone or zones most relevant for object detection. This approach thus enables rapid and efficient detection of objects present in the vehicle's environment and ensures safe driving without resorting to neighborhood generation methods.

[0114] The present invention is not limited to the means and configurations described and illustrated herein and extends also to any equivalent means and configuration as well as to any technically operative combination of such means.

Claims

Demands

1. A method for detecting (1) the environment of a vehicle (17), comprising the following steps: - a step of acquiring (2) a point cloud (NP) of the environment of said vehicle (17); - a step of filtering (4) the point cloud (NP) to form a filtered point cloud (NF), said filtering step (4) comprising at least one elimination of points from the point cloud (NP) corresponding to the ground of the environment; - a step of dividing (6) the filtered point cloud (NF) into distinct zones (Z); - a step of determining (8) a point density (D) within each zone (Z) and calculating an average density (D) of the zones; - a selection step (10), in which the point density (D) of each zone (Z) is compared to the average density (D) of the zones (Z), at least one of the zones (Z) being selected when this point density (D) is greater than the average density (D);- an analysis step (14) of at least one selected zone (SZ) in order to detect the presence of objects.;

2. Detection method (1) according to claim 1, wherein, during the splitting step (6), the filtered point cloud (NF) is split into circular regions (RC, 1RP, 2RP) in order to form said zones (Z).

3. Detection method (1) according to any one of claims 1 or 2, wherein, during the splitting step (6), the filtered point cloud (NF) is split into a central circular region (RC) and at least one peripheral circular region (1RP, 2RP) surrounding the central circular region (RC) in order to form said zones (Z).

4. A detection method (1) according to claim 3, comprising a processing step (12) of the central circular region (RC) in order to detect the presence of objects, said processing step (12) being prior to the analysis step (14).

5. A detection method (1) according to any one of claims 1 to 4, wherein, during the splitting step (6), the cloud of filtered points (NF) is divided into angular sectors (SA, SR, 1SL, 2SL) surrounding the vehicle (17) in order to form the zones (Z).

6. Detection method (1) according to claim 5, wherein, during the splitting step (6), the filtered point cloud (NF) is split to form the zones (Z), into at least: - a front angular sector (SA) of the vehicle (17), - a rear angular sector (SR) of the vehicle (17), - two lateral angular sectors (1SL, 2SL) of the vehicle (17).

7. A detection method (1) according to any one of claims 1 to 6, in combination with claim 4, wherein a run time (T) of the processing step (12) and the analysis step (14) is measured, the analysis step (14) being interrupted when this run time (T) exceeds a maximum level (T1).

8. Detection method (1) according to claim 7, wherein the interruption of the analysis step (14) is conditioned on a number (N) of areas analyzed.

9. A detection method (1) according to any one of claims 7 or 8, comprising a supplementary analysis step (16) of at least one discarded area (ZA) during the selection step (10) in order to detect the presence of objects within that discarded area (ZA), said supplementary analysis step (16) being carried out when the execution time (T) is less than a minimum level (T2).

10. Detection method (1) according to claim 9, wherein the minimum level (T2) is equal to half of the maximum level (T1).

11. A vehicle (17) environment detection system (36) implementing the detection method (1) according to any one of claims 1 to 10.

12. Computer program comprising program code instructions for performing the steps of the detection method (1) according to any one of claims 1 to 11, when said computer program is executed on one or more processors.

Images