METHOD FOR DETECTING LONG OR ARTICULATED VEHICLES

Abstract

A method for observing the environment of a primary vehicle is described. The method is implemented by a device comprising a radar or lidar-type object presence sensor mounted on the vehicle, and a processing unit communicating with the object presence sensor. The method includes: acquiring data relating to the vehicle's environment by the object presence sensor; detecting a plurality of objects (O1, O2, O3) from the acquired data, each object being associated with a set of characteristics; tracking a plurality of detected objects to predict a future state of the tracked objects; and creating a chain comprising a succession of tracked objects associated with one another, including: associating a set of tracked objects forming the chain, and determining characteristics associated with the chain of objects from the characteristics of the objects forming said chain. Abstract Figure: Figure 6a

Description

Title of the invention: METHOD FOR DETECTING LONG OR ARTICULATED VEHICLES Technical field

[0001] The present disclosure relates to the observation of the environment of vehicles for the implementation of driving assistance functions. It applies in particular to the detection and tracking of long or articulated vehicles. Prior art

[0002] It is known to equip a vehicle with a lidar or radar type object presence sensor to obtain information concerning the environment of the vehicle. In particular, lidar or radar sensors conventionally implement methods for observing the environment of a vehicle comprising the acquisition of a point cloud by the sensor, the identification of a set of objects, for example other vehicles, from a processing of this point cloud, and the tracking of the identified objects, in order to be able to track and anticipate the trajectory of the objects surrounding the vehicle on which the sensor is mounted.

[0003] With regard to the representation and tracking of long vehicles, such as trucks, buses, including in particular articulated buses, vehicles comprising a tractor and a trailer, or even trams, etc., two approaches have been proposed.

[0004] A first approach consists of creating a single object describing the entire long vehicle, and tracking this single object. However, this approach suffers from significant drawbacks, including on the one hand a lack of precision, because the point clouds obtained by radar or lidar type sensors are often distributed heterogeneously, which results in a geometry of the detected object which is approximate, this approximation being all the more important as the object is large. In addition, the kinematic tracking of such an object depends significantly on the position of its center of rotation.However, as shown in Figures 1a and 1b, which represent possible trajectories Tb T2, T3 of the same object O, for identical initial kinematic conditions but with a different position of the center of rotation, the approximation mentioned above is also applicable to the position of the center of rotation of the object, which can result in a very large margin of error for long vehicles, and therefore a great inaccuracy in the prediction of the trajectories (resulting in [Fig.lb] in three hypothetical positions Ob O2 and O3 for the object).

[0005] The other approach is to create a set of small objects that are tracked independently of each other. This approach solves the problems of inaccuracy described above, but having to track a plurality of objects corresponding to a single vehicle represents a significant cost in terms of memory and processing capacity for this vehicle. In addition, data concerning objects detected and tracked by lidar or radar type sensors are usually used by driver assistance applications, and representing a single object by a set of small objects can on the one hand lead to saturation of the processing capacity of the application, and on the other hand generate a risk of confusion for users. Summary

[0006] The present disclosure improves the situation.

[0007] A method for observing the environment of a main vehicle is proposed, the method being implemented by a device comprising a radar or lidar type object presence sensor mounted on the main vehicle, and a processing unit in communication with the object presence sensor, the method comprising: - the acquisition of data relating to the environment of the main vehicle by the object presence sensor, - the detection of a plurality of objects from the acquired data, each object being associated with a set of characteristics, - tracking a plurality of detected objects to predict a future state of the tracked objects, - the creation of a chain comprising a succession of tracked objects associated with each other, including: • the association of a set of tracked objects forming the chain, and • the determination of characteristics associated with the chain of objects, from characteristics of the objects forming said chain.

[0008] In embodiments, the method further comprises using the characteristics of the object chain in implementing at least one driving assistance application of the main vehicle.

[0009] In embodiments, each object of a chain is associated by a front end, respectively rear end, with the rear end, respectively front end, of another object of the chain.

[0010] In embodiments, associating a set of tracked objects forming a chain comprises: - the identification of at least one first tracked object of which at least one end, called free, is not associated with another object, - the association of a second tracked object with the first, if the pair formed by the first and second tracked objects satisfies at least the following conditions: • the second tracked object has a free end which can be associated with the free end of the first object, • the two objects are spatially neighbors, and • the two objects have equal kinematic characteristics within a given tolerance.

[0011] In embodiments, the association of a second tracked object with the first is implemented if the pair formed by the first and the second object further verifies a condition on the trajectory of the objects of the pair, such that the second object is located in a region of interest located at the free end of the first object and whose position is determined as a function of the yaw angular velocity of the first object.

[0012] In embodiments, the region of interest is a geometric shape whose center is located on the circle of curvature defined by the yaw rate of the first object.

[0013] In embodiments, identifying at least one first tracked object comprising a free end comprises a preliminary step of selecting a set of candidate objects from among the set of tracked objects, the selected candidate objects being moving tracked objects.

[0014] In embodiments, determining characteristics associated with the chain includes determining kinematic characteristics and dimensions of the chain.

[0015] In embodiments, the method comprises identifying a main object among the set of objects in the chain, and the kinematic characteristics of the chain are those of the main object.

[0016] In embodiments, the primary object is the object in the chain closest to the object presence sensor.

[0017] In embodiments, determining the dimensions of the chain comprises determining an area of ​​space encompassing all of the grouped objects, the dimensions of the chain corresponding to the dimensions of said area.

[0018] In embodiments, determining the area of ​​space encompassing all of the grouped objects comprises: - the calculation of a first rectangular area of ​​space grouping together all the grouped objects, said first area having a main dimension in a direction parallel to the direction of the object closest to the sensor among the grouped objects, - the calculation of at least one second rectangular zone of space grouping together all of the grouped objects, said second zone having a main dimension in a direction different from the first zone, - the allocation of a score to each zone based on a confidence criterion and a criterion of area of ​​the zone and, - the selection of the zone of the space from among the first and said at least second zone according to the score attributed to each zone.

[0019] In embodiments, the method comprises: - the calculation of a second rectangular area of ​​space grouping together all the grouped objects, said second area having a main dimension parallel to an axis connecting a front end of the first object in the chain to a rear end of the last object in the chain, - the calculation of a third rectangular area of ​​space grouping together all the grouped objects, said third area having a main dimension parallel to an estimated orientation of the main vehicle if it were at the location of the object in the chain closest to the sensor.

[0020] A device is also described comprising a radar or lidar type object presence sensor and a processing unit in communication with the object presence sensor, the device being configured for implementing the method according to the preceding description.

[0021] According to another object, a vehicle is described, comprising a device according to the preceding description, the object presence sensor being on board the vehicle and oriented so as to acquire data on the presence of objects in the environment of the vehicle.

[0022] It is also proposed to observe the environment of a main vehicle, the method being implemented by a device comprising a radar or lidar type object presence sensor mounted on the main vehicle, and a processing unit in communication with the object presence sensor, the method comprising: - the acquisition of data relating to the environment of the main vehicle by the object presence sensor, - the detection of a plurality of objects from the acquired data, each object being associated with a set of characteristics, - tracking a plurality of detected objects to predict a future state of the tracked objects, - creating a chain comprising a succession of tracked objects associated with each other, comprising associating a set of tracked objects forming the chain, wherein the association of a second tracked object with the first is implemented if the pair formed by the first and the second object satisfies a condition on the trajectory of the objects of the pair, such that the second object is located in a region of interest located at the free end of the first object and the position of which is determined as a function of the yaw angular velocity of the first object. In embodiments, the region of interest is a geometric shape whose center is located on the circle of curvature defined by the yaw rate of the first object.

[0023] In embodiments, the method further comprises determining characteristics associated with the chain of objects, from characteristics of the objects forming said chain, and using the characteristics of the chain of objects in the implementation of at least one driving assistance application of the main vehicle.

[0024] In embodiments, each object of a chain is associated by a front end, respectively rear end, with the rear end, respectively front end, of another object of the chain.

[0025] In embodiments, the association of a set of tracked objects forming the chain further comprises: the identification of at least one first tracked object of which at least one end, called free, is not associated with another object, - the association of a second tracked object with the first, if the pair formed by the first and second tracked objects satisfies at least the following conditions: • the second tracked object has a free end that can be associated with the free end of the first object, • the two objects are spatially neighbors, and - the two objects have equal kinematic characteristics within a given tolerance

[0026] The proposed method for observing the environment makes it possible to detect long or articulated vehicles and to take these vehicles into account in driving assistance applications in a precise manner, while reducing the amount of memory or computing resources required. Indeed, long or articulated vehicles are represented as a chain of objects, each object being tracked, which, due to the reduced size of each object, ensures tracking accuracy. In addition, the chain of objects is also associated with specific characteristics, determined from the tracked objects that compose it, so that it can be used as a single object, in particular for driving assistance applications using acquisitions from a radar or lidar type sensor. Brief description of the drawings

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

[0028] [Fig. la] [Fig. la] illustrates the uncertainty of the position of a center of rotation of a long vehicle.

[0029] [Fig. 1b] [Fig. 1b] illustrates the impact of the uncertainty of the position of the center of rotation of a long vehicle on the prediction of a trajectory of this vehicle.

[0030] [Fig.2] [Fig.2] represents the main stages of a method of observing the environment of a vehicle according to one embodiment.

[0031] [Fig.3] [Fig.3] represents the main steps of a process of associating tracked objects to form a chain.

[0032] [Fig.4] [Fig.4] represents the main steps of a characterization process of a chain of objects.

[0033] [Fig.5a] [Fig.5a] represents an example of a region of interest located at the end of an object whose yaw angle is negative.

[0034] [Fig.5b] [Fig.5b] represents a second example of a region of interest located at the end of an object whose yaw angle is negative and greater than that of [Fig.5a]

[0035] [Fig.5c] [Fig.5c] represents a third example of a region of interest located at the end of an object whose yaw angle is positive.

[0036] [Fig.6a] [Fig.6a] represents a first example of an area grouping together a set of objects forming a chain of objects.

[0037] [Fig.6b] [Fig.6b] represents a second example of an area grouping together a set of objects forming a chain of objects.

[0038] [Fig.6c] [Fig.6c] represents a third example of an area grouping together a set of objects forming a chain of objects.

[0039] [Fig.7] [Fig.7] represents an example of a vehicle carrying a device observation suitable for implementing the process. Description of the embodiments

[0040] A method for observing the environment of a main vehicle V will now be described. With reference to [Fig.7], this method is implemented by a device 1 comprising a sensor 10 for the presence of objects O of the radar or lidar type, and a processing unit 11. The processing unit may comprise one or more computers 12, for example of the processor, microprocessor, microcontroller, FPGA, etc. type, and one or more memories 13, storing in particular code instructions that can be executed by the computer for the implementation of the method described below. The memory 13 may in particular be a read-only memory, for example ROM (Read-Only Memory), EEPROM (Electrically Erasable Programmable Read-Only Memory); the memory may also comprise optical, electronic or magnetic storage means.

[0041] The object sensor 10 may be a radar sensor. In this case, the sensor is adapted to detect a point cloud comprising for each point at least three data: speed, distance from the sensor and azimuth, the distance and the azimuth being able to be converted into location in two dimensions (X,Y) in a reference frame associated with the main vehicle. Some radar sensors can also acquire additional elevation data. In this case, the distance, azimuth and elevation data can be converted into a three-dimensional location (X, Y, Z) in a reference frame associated with the main vehicle. In the case of a radar, the object sensor also acquires for each point a radar equivalent surface (SER) data, also known by the English acronym RCS for Radar Cross Section.

[0042] Alternatively, the object sensor 10 may be a lidar sensor. In this case, the sensor is adapted to detect a point cloud comprising for each point a three-dimensional position (x,y,z) where x and y are the positions of a pixel p in a matrix of receivers of the lidar sensor, and z is a depth coordinate, i.e. distance from the lidar. These coordinates may be converted into three-dimensional coordinates (X, Y, Z) in a frame of reference associated with the main vehicle. The data acquired by the lidar sensor also comprise for each point of the point cloud, a light intensity datum I reflected by the point.

[0043] The object presence sensor 10, radar or lidar, is on board the main vehicle V, being positioned so as to be able to acquire a point cloud representative of the environment of the main vehicle. For example, the radar or lidar sensor can be mounted on the vehicle near a front or rear end of the main vehicle, being oriented respectively towards the front or the rear of the vehicle.

[0044] The processing unit 11 may be co-located, i.e. also be embedded in the main vehicle, to locally process the data acquired by the sensor 10. In embodiments, the processing unit 11 may be integrated with the object presence sensor 10. Alternatively, the processing unit 11 may be located remotely and the sensor 10 may comprise a remote communication interface for transmitting acquired data for the implementation of the remote processing by the processing unit. For example, certain steps of point cloud acquisition, detection and tracking of objects may be implemented by a processing unit integrated with the sensor 10, and other steps, such as for example the chain creation step, may be implemented by a remote processing unit. According to another example, all of the processing described below may be carried out by the sensor 10.

[0045] With reference to [Fig.2], the method for observing the environment of the main vehicle comprises a step 100 of acquiring data relating to the environment of the main vehicle by the object presence sensor. This step is implemented for example by acquiring a point cloud by the radar or lidar sensor. This step is advantageously repeated at a given acquisition frequency, in order to be able to track the objects detected using the sensor.

[0046] During a step 200, a plurality of objects is detected from the acquired data, for example by grouping points of the point cloud. The detected objects are objects present in the environment of the vehicle. They may therefore be other vehicles, moving or parked, pedestrians, elements of street furniture, etc. Each detected object is associated with characteristics, including for example kinematic characteristics: position relative to the main vehicle, speed, acceleration, and dimension characteristics. The dimension characteristics may be those of an area of ​​the space surrounding the detected object. At least some SER data of the points making up an object, in the case of a radar sensor, or the intensity data, in the case of a lidar sensor, are also stored by being associated with the detected object.For example, an object may be associated with a SER or intensity value corresponding to the maximum value observed among the set of points belonging to the object. The step 200 of detecting objects from a radar or lidar point cloud is known to those skilled in the art. It may for example be implemented by a DBSCAN type algorithm (acronym for Density-Based Spatial Clustering of Applications with Noise) or by grouping points on the basis of statistical distances such as the Mahalanobis distance.

[0047] Once these objects have been detected, they are subject to tracking 300 making it possible to predict their trajectory relative to the vehicle. This tracking can for example be implemented by a Kalman filter (or EKF for extended Kalman Filter) which makes it possible to predict the future kinematic characteristics of each object from its current kinematic characteristics. The tracking of an object comprises the recording in the memory 13 of an identifier of the tracked object, its dimensions, and its kinematic characteristics. In the case where the tracking is implemented by a Kalman filter, the recorded kinematic characteristics are those which are updated at the end of the prediction of the filter at each iteration thereof.

[0048] The method also comprises a step 400 of creating at least one chain from the tracked objects, the chain comprising a succession of tracked objects associated with each other by their ends. A chain created during this step therefore comprises a succession of tracked objects, each object comprising a front end and a rear end (previously determined and identified as such from the coordinates of the points forming corners of the object and the kinematic properties of the object), and each object in the chain being connected to another object by its front end and / or by its rear end, each object being able to be connected at a given end to only one other object. The chain created during this step thus makes it possible to represent a long object, for example a long and / or articulated vehicle, as a succession of several objects. A long or articulated vehicle may be a truck, for example example a semi-trailer, a bus, especially an articulated bus, a tram, a vehicle towing a trailer, etc.

[0049] Indeed, long and / or articulated vehicles are not detected as a single object by a radar or lidar sensor, but are subject to several independent detections. In general, the maximum dimensions of a detected object are constrained to approximately 4 to 5 m, which is sufficient to represent an individual car but not a longer or articulated vehicle. The construction of a chain makes it possible to associate these objects to create a single object for downstream driver assistance applications, while the constituent objects of the chain are tracked individually (for example by the Kalmann filter) for better precision.

[0050] The method 400 for creating a chain of objects comprises a step 410 of associating a set of tracked objects forming the chain, and a step 420 of determining a set of characteristics specific to the chain, from the characteristics of the objects constituting the chain.

[0051] With reference to [Fig.3], a method 410 of associating a set of tracked objects to form a chain is shown in more detail.

[0052] In embodiments, step 410 of associating a set of objects may comprise a preliminary step 411 of preselecting a set of candidate objects for chaining, i.e. objects eligible to belong to a chain of objects.

[0053] This step may comprise selecting, from among the set of tracked objects, a subset of objects satisfying one or more conditions. According to a first condition, the selected objects are moving objects. In one embodiment, this selection is carried out by verifying that the speed of an object is not zero.

[0054] A second condition may relate to the intensity or SER value associated with the tracked objects, to ensure that the selected object may be a vehicle. In one embodiment, the selected objects are objects whose intensity or SER value is greater than a predetermined threshold. For example, the selected objects may be objects whose SER value is greater than a threshold between -10dB and 10dB, for example -5dB.

[0055] Once a subset of tracked objects has been selected, if applicable, the association of a set of objects comprises a step 412 of identifying at least one object of which at least one end has not already been associated with another object. For this object, the method then comprises the identification 413 of a second object which can be associated with the first.

[0056] The second object to be associated with the first is determined so as to satisfy a set of conditions. First, the second object must comprise at least one free end, that is to say one which has not already been associated with another object.

[0057] In embodiments, the set of conditions further comprises at least at least one similarity condition on the kinematic properties of the two objects. In one embodiment, the similarity condition on the kinematic properties is met if the two objects have equal speed within a predetermined speed tolerance, and equal acceleration within a predetermined acceleration tolerance. In embodiments, a similarity condition on the kinematic properties may also include a similarity condition on the orientations of the two objects. This condition may for example be verified by determining an angle formed between the orientation of the first object, the orientation being provided by the direction of the object along its main dimension, and the orientation of the second object, and verifying that this angle is equal within a predetermined tolerance.Given that the chain allows for the representation of articulated objects, and to take into account the possibly variable orientation between two successive objects in the chain, the tolerance on the relative orientations of the two objects can be quite high; for example, between 10 and 45° in absolute value, for example between 15 and 30°. A tolerance equal to 30° means that the similarity condition on the orientations of the two objects is respected if the orientation of the second object is equal to that of the first object ± 30°.

[0058] In embodiments, the set of conditions also includes a condition of proximity of the two objects, i.e. similarity of positions of the two objects. This condition can be verified by establishing an area centered on each object, for example circular, and of predetermined dimensions, and by determining whether the areas centered on the two objects have at least one point of intersection.

[0059] In embodiments, the set of conditions that the second object must satisfy in order to be able to be associated with the first object further comprises a condition relating to the compatibility between the relative positions of the two objects and their potential chaining. As indicated previously, the identification of a second object that can be associated with the first comprises a selection of an object of which at least one end is free, just like the first object. The compatibility of the two objects taking into account their potential chaining therefore corresponds to the possibility of associating the two objects taking into account on the one hand their free ends, and on the other hand their relative position. If the two objects both have the same free end, whether it is the front or rear end, these two objects cannot be associated.Furthermore, two objects each having at least one free end that is respectively different can only be associated if these free ends are sufficiently close to each other, that is to say that the front end of the first object is located at a distance from the rear end of the second object less than a determined distance threshold, or vice versa with the rear end of the first object and the front end of the second object.

[0060] In embodiments, the identification 413 of the second object to be associated with the first comprises the verification of several of the conditions presented above, namely at least the following conditions: - The second object has a free end, for example the front end, which can be associated with the free end of the first object, for example the rear end, - The second object has kinematic properties similar to those of the first object, within a predetermined tolerance, and - The second object is spatially close to the first object.

[0061] Furthermore, the set of conditions that the second object must satisfy may include a condition of trajectory similarity with the first vehicle. As indicated above, the chaining must make it possible to represent articulated objects, which do not necessarily have the same orientation at all points. To verify that an object follows another object, and can therefore be chained to it, even if it does not have the same orientation, it is determined whether the second object is located in a zone of interest located at the free end of the first object, the position of the zone of interest being determined as a function of the yaw angular velocity of the first object.

[0062] More precisely, with reference to Figures 5a to 5c, the area of ​​interest ZI is positioned so that its center CI is located on a circle C formed by the instantaneous yaw angular velocity of the object. This circle C represents the extension of the instantaneous trajectory of the first object 01, and centering the area of ​​interest ZI on the circle C makes it possible to search for the second object in the extension of the trajectory of the first object 01. Indeed, if the two objects belong to an articulated vehicle, the different parts of this articulated vehicle can form an angle between them, so that the second object is not located on a rectilinear axis corresponding to the instantaneous direction of the leading vehicle. On the other hand, the different objects have the same trajectory, and it is therefore possible to find an object associated with the first object by propagating its trajectory forwards or backwards.

[0063] By way of illustration, [Fig.5a] represents a first example of an area of ​​interest ZI with a relatively low (compared to [Fig.5b]) and negative yaw angular velocity. [Fig.5b] represents a second example of an area of ​​interest ZI with a relatively high and negative yaw angular velocity. [Fig.5c] represents a third example of an area of ​​interest ZI with a relatively low and positive angular velocity. It can be seen in these three figures that the area of ​​interest ZI, in this case elliptical in shape, follows the trajectory of the first object.

[0064] The shape of interest can be of any shape but it preferably has a main dimension; it can be for example a rectangle or an ellipse.

[0065] Determining a second object 413 that can be associated with the first object can be implemented systematically by verifying, for each tracked object other than the first object, all of the conditions described above. When all the conditions are met, the object verifying all the conditions is associated with the first object. Alternatively, these conditions can be verified so as to progressively select subsets of objects satisfying the successive conditions. For example, one can first identify a subset of objects spatially neighboring the first object, then verify, among the objects in this subset, those which have kinematic properties close to the first object, etc.

[0066] Once a second vehicle has been identified to be associated with the first object, the association of the two objects is recorded 414 in the memory 13. To do this, the memory further stores a list of object strings with an identifier associated with each string, and the recorded information regarding a tracked object is updated to add: - The identifier of the chain to which it belongs, - An identifier of the object to which it is associated, - An indication of the end of the object by which it is associated with another object.

[0067] According to an additional condition, each chain cannot exceed a determined length. Consequently, in the case where the first object that was associated with a second object in step 410 already belonged to a chain of objects, by being associated with another object by one end, an additional condition can be verified during step 412, that is to say before determining a second object to be associated with it, according to which the chain to which the object belongs must have a length less than a determined threshold. For this, the length of the chain is calculated as the sum of the lengths of the objects that compose it, to which is added the length of the object to be added. If the length obtained is less than the determined length, step 413 of determining a second object to be associated with the first is implemented.

[0068] With reference to [Fig. 4], the step 420 of determining the characteristics associated with the chain formed by associating the objects one after the other is shown schematically. In one embodiment, this step may comprise a preliminary step of determining a main object 421 of the chain. The main object of the chain may be the object of the chain closest to the object presence sensor 10. Alternatively, the main object of the chain is the object for which the trajectory predicted for the object has a maximum confidence level among all the objects of the chain. For example, in the case where a Kalman filter is used, the Kalman filter generates a confidence metric—for example, a covariance matrix completing the state vector—for all the estimated kinematic signals. This confidence metric may be used as a confidence level for se- Select the main object of the chain.

[0069] The step of determining the characteristics of the chain may further comprise the determination of kinematic characteristics 422 of the chain, and dimensional characteristics 423 of the chain.

[0070] Concerning the kinematic characteristics, they can be those of one of the objects making up the chain. In one embodiment, the kinematic characteristics are those of the main object of the chain.

[0071] Determining the dimensional characteristics 423 of the chain comprises determining a zone of space encompassing all of the objects in the chain, the dimensions of the chain corresponding to the dimensions of said zone.

[0072] In one embodiment, shown schematically in [Fig.6a], the zone Z of the space encompassing all of the objects 01, 02, 03 of the chain is a parallelepiped zone, for example rectangular, whose main dimension extends according to the orientation of the main object of the chain, for example the object of the chain closest to the object sensor (here 01).

[0073] In one embodiment, shown schematically in [Fig.6b], the zone Z of the space encompassing all of the objects 01, 02, 03 of the chain is a parallelepiped zone, for example rectangular, whose main dimension extends parallel to an axis A connecting the front end of the object located at the head of the chain to the rear end of the object located at the tail of the chain. By "head of the chain" is meant the first object of the chain, the front end of which is free and the rear end is connected to another object (for example here the object 01), and by "tail of the chain" is meant the last object of the chain, the rear end of which is free and the front end is linked to another object (for example here the object 03). In one example, the axis may connect a point halfway across the front end of the object at the head of the chain to a point halfway across the back end of the object at the tail of the chain.

[0074] In another embodiment, shown schematically in [Fig.6c], the zone Z of the space encompassing all the objects 01, 02, 03 of the chain is a parallelepiped zone, for example rectangular, whose main dimension is parallel to an estimated orientation B of the main vehicle V if it were at the location of the main object of the chain. The estimated orientation B of the main vehicle V is determined by calculating a time difference t between the current position of the main vehicle V and the position of the object of the chain, this time difference being equal to the speed of the main vehicle multiplied by the distance between the main vehicle and the main object of the chain. This time difference is then multiplied by the yaw angular speed to obtain the estimated orientation of the main vehicle.

[0075] In one embodiment, the area of ​​space encompassing all of the objects of the chain is determined by calculating the area of ​​at least two zones among the three examples of zones described above, assigning a score to each zone based on the area of ​​each zone, and retaining the zone having the best score. In one embodiment, the score assigned to each zone may also depend on a confidence criterion on the zone. For example, an area defined from the orientation of the object closest to the sensor may have a higher confidence than other areas because the object closest to the sensor is generally detected and tracked with better precision.

[0076] In addition, the characteristics associated with the chain may include an identifier of the first object in the chain, i.e., the object located at the head of the chain, in the direction of movement of the objects constituting the chain. The characteristics associated with the chain may also include other characteristics derived from the characteristics of the objects constituting the chain, such as for example: - a SER or intensity value associated with the chain, - a number of points detected by the sensor and associated with the chain, cor corresponding to the sum of the number of points associated with each object in the chain, - the coordinates of the centroids of all the objects in the chain, or an average of these coordinates, - a probability of existence, which may be equal to the maximum probability observed among all the objects in the chain, - for each object in the chain, the source of the object detection, whether it is sensor 10 or possibly another component, - a timestamp of the objects in the chain, corresponding to the most recent moment when an object was detected by the sensor. For the entire chain, the most recent timestamp among all the objects in the chain can be selected. It is observed that depending on the nature of the data, the characteristics of the chain can be those of the main object of the chain, or be an average, a minimum or a maximum of the values ​​of the characteristics of the objects of the chain.

[0077] The characteristics may also include an indication of the length of time the chain has existed, i.e., the number of iterations of the process for which the chain has existed.

[0078] At the end of step 420, the chain formed by association of objects therefore constitutes an object distinct from the objects which constitute it, this object having a set of specific characteristics. The “chain” object and its characteristics can then be used in other algorithms, and in particular in processing implemented for driving assistance functionalities 500. Thus, a driving assistance algorithm whose input data is a set of objects detected in the environment of the main vehicle, each detected object being associated with characteristics including in particular kinematic characteristics and dimensions, can receive as input a chain of objects representative of a long vehicle, each object of which is also tracked, for example by a Kalman filter, which allows good precision in determining the dimensions and kinematic positions of the chain.

[0079] In embodiments, the above method is iterated at a given frequency which is the frequency of execution of the object detection algorithms by the object presence sensor 10. During successive iterations, the chains formed at a previous iteration are retained and step 400 of creating a chain may comprise a preliminary step of verifying that the objects composing the previously created chains still verify all or part of the different conditions for chaining, mentioned in step 410. The verifications may be carried out for each chaining of two consecutive objects, and comprise for example: - A verification of the same state of stationarity or mobility of two chained objects, that is to say that the conservation of the chain requires that two consecutive objects are either both mobile or both stationary, - One or more similarity condition(s) on the kinematic properties of two chained objects, in particular speed and acceleration, - The proximity condition on the positions of two chained objects, - The condition of similarity of trajectory of two consecutive chained objects.

[0080] If these conditions are not met for two consecutive objects, the chaining of these objects is removed. In the case where a chain is formed only of these two objects, then the chain is removed. In the case where the two objects belong to a longer chain, if these two objects are located at one end of the chain, the two objects are separated and the end of the remaining chain is updated. If the two objects are not located at any end of a chain, removing the chaining of the objects amounts to dividing the chain into two chains. The properties of the objects and the chains are updated. In particular, the objects whose chain is removed are no longer associated, for the ends concerned, with another object and these ends become free.

[0081] In one embodiment, the chaining between two objects is not directly deleted if the conditions mentioned above are not met, but each chaining between two objects can be associated with a score. This score can be initialized to a first value when the objects are chained, and be increased if, during the verification, the conditions are met, or be decreased if they are not, or not all of them. The deletion of the chaining between two objects takes place when the score reaches a too low a level previously determined. Parameters such as the initial value, the increment or decrement level, and the low threshold for removing the chaining can be determined based on the level of robustness or flexibility that one wishes to give to the chaining.

Claims

Claims

1. A method for observing the environment of a main vehicle, the method being implemented by a device (1) comprising an object presence sensor (10) of radar or lidar type mounted on the main vehicle (V), and a processing unit (11) in communication with the object presence sensor, the method comprising: - the acquisition (100) of data relating to the environment of the main vehicle by the object presence sensor, - the detection (200) of a plurality of objects from the acquired data, each object being associated with a set of characteristics, - the tracking (300) of a plurality of detected objects to predict a future state of the tracked objects, - the creation (400) of a chain comprising a succession of tracked objects associated with each other, comprising: • the association (410) of a set of tracked objects forming the chain,and • determining (420) characteristics associated with the chain of objects, from characteristics of the objects forming said chain.,

2. The method of claim 1, further comprising using the characteristics of the object chain in implementing at least one application (500) for assisting the driving of the main vehicle.

3. Method according to claim 1 or 2, in which each object of a chain is associated by a front end, respectively rear end, with the rear end, respectively front end, of another object of the chain.

4. Method according to any one of the preceding claims, in which the association (410) of a set of tracked objects forming a chain comprises: - the identification (412) of at least one first tracked object of which at least one end, called free, is not associated with another object, - the association (413) of a second tracked object with the first, if the pair formed by the first and the second tracked objects satisfies at least the following conditions: • the second tracked object has a free end that can be associated with the free end of the first object, • the two objects are spatially neighbors, and • the two objects have equal kinematic characteristics within a given tolerance.

5. Method according to the preceding claim, in which the association (413) of a second tracked object with the first is implemented if the pair formed by the first and the second object further verifies a condition on the trajectory of the objects of the pair, such that the second object is located in a region of interest located at the free end of the first object and the position of which is determined as a function of the angular yaw velocity of the first object.

6. Method according to the preceding claim, in which a region of interest is a geometric shape whose center is located on the circle of curvature defined by the yaw angular velocity of the first object.

7. Method according to any one of claims 4 to 6, wherein the identification (412) of at least one first tracked object comprising a free end comprises a preliminary step of selecting (411) a set of candidate objects from among the set of tracked objects, the selected candidate objects being moving tracked objects.

8. A method according to any preceding claim, wherein determining characteristics (420) associated with the chain comprises determining kinematic characteristics (422) and dimensions (423) of the chain.

9. Method according to the preceding claim, comprising the identification (421) of a main object among the set of objects of the chain, and the kinematic characteristics of the chain are those of the main object.

10. The method of claim 9, wherein the primary object is the object in the chain closest to the object presence sensor (10).

11. A method according to any one of claims 8 to 10, wherein determining the dimensions of the chain (423) comprises determining an area of ​​space encompassing all of the grouped objects, the dimensions of the chain corresponding to the dimensions of said area.

12.

13.

14. Method according to the preceding claim, in which the determination of the area of ​​space encompassing all of the grouped objects comprises: - the calculation of a first rectangular area of ​​space grouping together all the grouped objects, said first area having a main dimension in a direction parallel to the direction of the object closest to the sensor among the grouped objects, - the calculation of at least one second rectangular zone of space grouping together all of the grouped objects, said second zone having a main dimension in a direction different from the first zone, - the allocation of a score to each zone based on a confidence criterion and a criterion of area of ​​the zone and, - the selection of the zone of the space from among the first and said at least second zone according to the score attributed to each zone. The method of claim 12, comprising: - the calculation of a second rectangular area of ​​space grouping together all the grouped objects, said second area having a main dimension parallel to an axis connecting a front end of the first object in the chain to a rear end of the last object in the chain, - the calculation of a third rectangular area of ​​space grouping together all the grouped objects, said third area having a main dimension parallel to an estimated orientation of the main vehicle if it were at the location of the object in the chain closest to the sensor. Device (1) comprising an object presence sensor (10) of radar or lidar type and a processing unit (11) in communication with the object presence sensor, the device being configured for implementing the method according to any one of the preceding claims.

15. Vehicle (V), comprising a device (1) according to the preceding claim, the object presence sensor (10) being on board the vehicle (V) and oriented so as to acquire data on the presence of objects in the environment of the vehicle (V).

Images