Determining the orientation of an object

Abstract

A method for determining the reliability of a hypothesis about the orientation of an object near a motor vehicle is described. The method is implemented by a computer using a cluster of object detection points collected by a radar or lidar system mounted on the vehicle and at least one hypothesis about the object's orientation. The method comprises: determining the object's outline from the cluster of detection points and the hypothesis about the object's orientation; determining at least one critical zone of the object from the relative positions of the object and the vehicle; and calculating the reliability of the hypothesis based on the position of the detection points relative to the critical zone. Abstract: Figure 3b

Description

Title of the invention: Determining the orientation of an object technical field

[0001] The present disclosure falls within the field of connected vehicles, and relates in particular to a method for determining the reliability of a hypothesis on the orientation of an object near a motor vehicle from a cluster of object detection points collected by a radar or lidar on board the vehicle, at least one hypothesis on the orientation of the object and a predictive curve of the vehicle's trajectory. Previous technique

[0002] Advanced driver assistance systems (ADAS) use sensors to detect and track objects around a motor vehicle. These sensors, such as radar and lidar, generate point clouds that represent the detected objects. Applications include obstacle detection, parking assistance, and collision avoidance. The sensors provide essential data for assessing the position and movement of surrounding objects, thus contributing to the safety and efficiency of autonomous vehicles. ADAS systems aim to provide early warnings to drivers, automate certain driving functions, and facilitate navigation in complex environments.

[0003] For these applications, the accuracy of object detection near vehicles is a critical issue. Systems must be able to distinguish the shape of different types of objects and accurately determine their orientation to prevent inappropriate operation of driver assistance systems, such as activating emergency braking when there is no obstacle. The ability to correctly estimate object orientation is essential to ensure an appropriate vehicle response to driving situations.

[0004] Although ADAS systems already use sensors to detect and track objects around a vehicle, the precise determination of the orientation of these objects still needs improvement. Indeed, the possible orientations of an object from a simple point cloud are numerous, which complicates the estimation of the object's exact position in space. Current methods often rely on several assumptions based on the object's velocity vector or the distribution of detection points. However, these methods do not always guarantee an accurate estimate, particularly for slow-moving or stationary objects.

[0005] There is therefore a need for solutions enabling the estimation of the reliability of a hypothesis on the orientation of detected objects. Summary

[0006] This disclosure improves the situation.

[0007] In particular, one purpose of the present disclosure is to enable a more reliable determination of the orientation of an object detected by a sensor on board a vehicle.

[0008] A method is proposed for determining the reliability of a hypothesis about the orientation of an object near a motor vehicle, the method being implemented by a computer using a cluster of object detection points collected by a radar or lidar system mounted in the vehicle and at least one hypothesis about the object's orientation, the method comprising: - Determining the object's outline based on the cluster of detection points and the assumption about the object's orientation, - A determination of at least one critical area of ​​the object based on the relative positions of the object and the vehicle, - A calculation of the reliability of the hypothesis based on the position of the detection points relative to the critical zone.

[0009] In embodiments, the determination of the critical area of ​​the object is implemented from a predictive curve of the vehicle's trajectory.

[0010] In embodiments, the critical zone is a surface of the object consisting of a set of points included in the contour of the object located at a distance less than a predetermined critical distance from the predicted curve of the vehicle's trajectory.

[0011] In embodiments, the critical area includes at least one corner of the object closest to the predicted curve of the vehicle's trajectory.

[0012] In embodiments, the critical zone comprises the two nearest corners of the predicted trajectory curve of the vehicle.

[0013] In embodiments, the reliability of the hypothesis is calculated as the density of detection points in the critical area.

[0014] In embodiments, the reliability of the hypothesis is expressed p = id - threshold) / seiûL where p is the reliability of the hypothesis, is the distance between said corner and the nearest detection point of said corner, and threshold is a predetermined constant.

[0015] In embodiments, the object comprises at least two critical zones and the calculation of the reliability of the hypothesis is carried out from the position of detection points relative to each critical zone.

[0016] In embodiments, the process further includes a hypothesis validation step in which the reliability of the hypothesis is greater than a determined threshold.

[0017] According to another object, a method for determining the orientation of an object near a motor vehicle is described, the method being implemented by a computer using a cluster of object detection points collected by a radar or lidar, the method comprising: - A determination of at least one hypothesis regarding the object's orientation, - A determination of the hypothesis's reliability according to the preceding description, - A comparison of the reliability of the hypothesis at a given threshold and the assignment of a confidence marker to the hypothesis, or: • When the determined reliability is greater than the threshold, a validation of the hypothesis, • When the determined reliability is below the threshold, a rejection or a modification of the hypothesis.

[0018] In embodiments, the method for determining the orientation of an object further includes adapting the movement of the vehicle according to the determined orientation.

[0019] According to another object, a computer program product is described comprising code instructions for the implementation of a process described above, when this program is executed by a computer.

[0020] According to another object, a device for determining the orientation of objects relative to a vehicle is described, comprising at least one computer, a memory, a radar or lidar mounted in a vehicle, the device being configured to implement a method according to the preceding description.

[0021] According to another object, a vehicle is described, comprising a device for determining the orientation of objects according to the preceding description. Technical effects

[0022] The method improves the estimation of the orientation of an object detected by radar or lidar by allowing the reliability of each hypothesis about this orientation to be evaluated. Furthermore, calculating the reliability of a hypothesis based on the position of the detection points relative to a critical area makes it possible to focus the analysis on the parts of the object most likely to have an impact or degrade driving functions (for example, by triggering emergency braking, degrading adaptive cruise control, or interfering with the vehicle's trajectory).

[0023] The method makes it possible to validate or reject a hypothesis about the orientation of an object and ensures that only the most precise orientations are used by the systems driver assistance systems help prevent incorrect driving behavior. This improves the vehicle's response to driving situations, contributing to a better user experience and increased safety.

[0024] The dynamic adaptation of vehicle movement, based on the assumptions validated according to the process, makes it possible to improve traffic flow by minimizing unnecessary emergency braking and ensuring smoother driving. Brief description of the drawings

[0025] Other features, details and advantages will become apparent from reading the detailed description below and from analyzing the accompanying drawings, in which: Fig. 1

[0026] [Fig.1] represents a vehicle adapted for implementing a method for determining the orientation of an object according to a particular embodiment or a method for determining the reliability of a hypothesis on the orientation of an object according to another embodiment. Fig. 2

[0027] [Fig. 2] is a flowchart illustrating the main steps of a process for determining the orientation of an object according to a particular embodiment. Fig. 3

[0028] [Fig.3a] shows a vehicle on a road network and an object according to a first hypothesis. Fig. 3b

[0029] [Fig.3b] shows the vehicle of [Fig.3a] on a road network and the object according to a second hypothesis. Description of the implementation methods

[0030] Figure 1 represents a vehicle 100 traveling on a road network. The vehicle 100 includes an on-board object detection sensor 128 configured to collect information about the vehicle's environment. For example, the object detection sensor 128 could be a radar or a lidar.

[0031] The vehicle 100 is equipped with a device 120 for determining the orientation of objects relative to the vehicle. The device 120 includes at least one memory 124, comprising code instructions that can be executed by a computer. The device 120 further includes at least one computer 126 that can be configured to implement a method from data acquired by the sensor 128, for example the method described below with reference to [Fig. 2].

[0032] In one embodiment, the memory 124 and the computer 126 can be mounted in the vehicle 100. According to an alternative, only the sensor 128 can be mounted, the memory 124 and the computer 126 being able to be remote, for example within a server. In this case, vehicle 100 may also include means of communication enabling it to access the remote server, for example via a telecommunications network.

[0033] A method 200 for determining the orientation of an object in the vicinity of a vehicle, in other words sufficiently close to the vehicle to be detected by a sensor (for example a radar or lidar) on board the vehicle, implemented by the device 120, will now be described with reference to [Fig.2].

[0034] During a step 201, the device 120 emits a signal via the sensor 128 and receives a return signal reflected on an object near the vehicle 100. The return signal is interpreted by the computer 126 as a cluster of object detection points.

[0035] From the cluster of detection points, the computer 126 can determine at least one hypothesis about the orientation of the object.

[0036] The assumption about the object's orientation is an estimate of the direction or angle of a detected object with respect to an axis of a given reference frame, for example, a reference frame attached to the vehicle 100, and in particular the angle formed between the detected object and an axis parallel to the longitudinal direction of the vehicle. This assumption can be derived from the point cluster, for example, by analyzing the spatial distribution of the points.

[0037] By having the cluster of detection points and at least one hypothesis on the orientation of the object, the computer 126 can determine the reliability of this hypothesis by a method 210, comprising steps 211 to 213.

[0038] Step 211 consists of determining an object contour from the cluster of detection points and the assumption about the object's orientation. The contour may have a predetermined shape. In one particular embodiment, the object contour is rectangular. Alternatively, the object contour could, for example, also be an ellipse, or a rectangle with rounded corners.

[0039] During a step 212, the computer 126 determines at least one critical area of ​​the object from the relative positions of the object and the vehicle.

[0040] A critical zone is a part of the detected object that is particularly important for assessing its potential interaction with the vehicle. In one embodiment, the determination of the object's critical zone is implemented using a predictive curve of the vehicle's trajectory. The critical zone can be determined based on the proximity of points included in the contour associated with the object to the vehicle's predicted trajectory.

[0041] The predictive trajectory curve of vehicle 100 designates an estimate of the future trajectory that the vehicle is likely to follow. This curve can be calculated in real time by another device of vehicle 100 from data such as the current speed, the direction of the vehicle (derived, for example, from the angle steering wheel) and driving intentions (for example a predefined route or settings relating to the preferred type of road).

[0042] According to one embodiment, the critical zone is a surface of the object consisting of a set of points included in the contour associated with the object located at a distance less than a predetermined critical distance from the predicted curve of the trajectory of the vehicle 100.

[0043] According to one embodiment, the critical zone comprises at least one corner of the object closest to the predicted trajectory curve of vehicle 100, for example, the two corners closest to the predicted trajectory curve of vehicle 100. In another case, a relevant side of the object can be determined based on the object's speed and / or that of vehicle 100, allowing, for example, consideration of only one lateral or longitudinal side of the object. The critical zone could then comprise two corners belonging to this relevant side. In this embodiment, the shape and size of the critical zone can be predetermined. The size can be predetermined as an absolute value or as a proportion of the surface area formed by the object's contour. Furthermore, in the case of a rectangular contour with rounded edges, the corners can be defined as the rounded area connecting two adjacent sides.In the case of an elliptical contour, the corners can be defined as the areas of the ellipse located at the ends of the major and minor axes.

[0044] Step 213 of the process 210 corresponds to calculating the reliability of the hypothesis based on the position of the detection points relative to the critical zone. In particular, the reliability calculation of a hypothesis can be implemented using a number of detection points included within the critical zone. The reliability of a hypothesis can be represented by a numerical value, which can allow the reliability value to be compared to a threshold. Thus, according to one embodiment, the process 210 may include a hypothesis validation step in which the reliability of the hypothesis, determined in step 213, is greater than a specified threshold.

[0045] An alternative may be to compare the hypotheses with each other according to their respective reliability values. In some cases, for example if a threshold value is not available, determining the orientation of the object may consist of selecting the hypothesis with the highest reliability.

[0046] The reliability of the hypothesis can be calculated as the density of detection points in the critical zone, i.e., for example, the number of detection points in said zone divided by the area of ​​said zone.

[0047] Alternatively, the reliability of the hypothesis can be expressed as p = (d - threshold) / threshold, where p is the reliability of the hypothesis, d is the distance between said corner and the nearest detection point of said corner, and threshold is a predetermined constant, threshold reflecting the fact that an area of ​​the object without a detection point exceeding a threshold distance on the trajectory of vehicle 100 could be critical for the functions of vehicle 100. For example, threshold may be 50 cm.

[0048] When an object has at least two critical zones, the reliability of the hypothesis is calculated based on the position of detection points relative to each critical zone. For example, a reliability value can be calculated using one of the methods described above for each of the two critical zones. The reliability of the hypothesis can then be expressed as the average of the two calculated values. Alternatively, the reliability of the hypothesis can be taken as the highest reliability value calculated among the critical zones.

[0049] This calculation method can be generalized when there are more than two critical zones, by calculating the average or by selecting the maximum value of the reliability values ​​obtained from each of the critical zones.

[0050] Step 220 consists of comparing the reliability of the hypothesis to a predetermined threshold. When the determined reliability is greater than the threshold, the hypothesis is validated; when the determined reliability is less than the threshold, the hypothesis is rejected or modified. In other words, when the hypothesis is validated—that is, when the reliability of the hypothesis is greater than the predetermined threshold—the hypothesis concerning the orientation of the object is retained and can be used by the driver assistance systems. For example, the vehicle's movement can be adjusted according to the orientation of the object.

[0051] In some embodiments, a second threshold is defined, lower than the determined threshold or first threshold. If the reliability of the hypothesis is lower than the second threshold, the hypothesis is rejected and the computer 126 executes step 201 of process 200 again to determine another hypothesis. If the reliability of the hypothesis is between the second threshold and the first threshold, the hypothesis is modified, for example by slightly changing the orientation or outline of the object. The computer then resumes determining the reliability of the new hypothesis from step 211.

[0052] In the event that only one hypothesis is identified, for example, when the same unsatisfactory hypothesis is determined and rejected at each iteration of process 200, said unsatisfactory hypothesis can nevertheless be retained by being accompanied by a confidence marker. The confidence marker can indicate that the reliability is low and can be used both by driver assistance systems and brought to the attention of the vehicle user.

[0053] A computer program product may include code instructions for implementing the process 200 or 210, when this program is executed by a computer, for example the computer 126 of the device 120.

[0054] Figures 3a and 3b represent a motor vehicle 300, which may be similar to vehicle 100, travelling on a road. The vehicle 300 may be equipped with an object detection sensor, for example sensor 128, enabling it to detect objects in a field of vision 302.

[0055] Several detection points 312 representative of an object 310 are captured by the object detection sensor. The vehicle 300 is a connected vehicle equipped with a device for determining the orientation of objects relative to the vehicle, for example the device 120 of the vehicle 100, adapted to implement the method 200. Furthermore, the predicted trajectory curve 304 of the vehicle 300 is known and can be used by the object orientation determination device.

[0056] By applying step 201, two hypotheses Ha and Hb on the orientation of the object are determined, consisting of an orientation along an axis Xa and an orientation along an axis Xb respectively.

[0057] During step 211, contours 306a and 306b of the object are determined respectively for the orientation hypotheses along Xa and Xb.

[0058] During step 212, critical zones are identified for each hypothesis, for example critical zones 308a in [Fig.3a] for hypothesis Ha and 308b in [Fig.3b] for hypothesis Hb. In this example, critical zones 308a and 308b represent all the surfaces of the object made up of points located at a distance less than a predetermined critical distance, for example 1 m, from the predicted curve 304 of the trajectory of the vehicle 300.

[0059] During step 213, the reliability of each hypothesis Ha, Hb is calculated. There are three detection points 312 in area 308a and no detection points 312 in area 308b. For example, if each critical area 308a, 308b has an area of ​​1 m², the reliability of hypothesis Xa would be 2 detections / m² and the reliability of hypothesis Xa would be J Xb would be "_ £ _ n detections / m2. h ~ i " u

[0060] If the value of the determined threshold was 2 detections / m2, during step 220, the hypothesis Ha would be validated (because 3 > 2) and the hypothesis Hb would be rejected (because 0 < 2).

[0061] Thus, of the assumptions determined in step 201, only assumption Ha is validated. It is therefore this assumption that can be used by the vehicle's driver assistance systems, for example, to adapt the vehicle's movement. Without the use of method 200, assumption Hb could have been retained and could have caused an emergency stop of the vehicle 300, even if the actual object 310 did not intersect the curve 304 of the vehicle's predicted trajectory.

[0062] According to an alternative, the determination of the critical zone of the object can be implemented from the relative positions of the object and the vehicle without taking into account a predicted trajectory of the vehicle. For example, if an object is placed at the To the right of the vehicle in the direction of travel, a critical zone can be defined as a set of points within the object's boundary located to the left of the object and covering a predetermined area. This predetermined area can, for example, be defined beforehand or depend on the size of the object or the distance between the vehicle and the object.

Claims

Demands

1. A method for determining the reliability of a hypothesis about the orientation of an object near a motor vehicle, the method being implemented by a computer from a cluster of detection points of the object collected by a radar or lidar mounted in the vehicle and at least one hypothesis about the orientation of the object, the method comprising: - A determination (211) of a contour of the object from the cluster of detection points and the hypothesis about the orientation of the object, - A determination (212) of at least one critical zone of the object from the relative positions of the object and the vehicle, - A calculation (213) of the reliability of the hypothesis from the position of the detection points relative to the critical zone.

2. A method according to claim 1, wherein the determination of the critical area of ​​the object is implemented from a predictive curve of the vehicle's trajectory.

3. A method according to claim 2, wherein the critical zone is a surface of the object consisting of a set of points included in the contour of the object located at a distance less than a predetermined critical distance from the predicted curve of the vehicle's trajectory.

4. A method according to any one of claims 2 or 3, wherein the critical area includes at least one corner of the object nearest to the predicted curve of the vehicle's trajectory.

5. Method according to claim 4, wherein the critical area comprises the two nearest corners of the predicted curve of the vehicle's trajectory.

6. A method according to any one of the preceding claims, wherein the reliability of the assumption is calculated as the density of detection points in the critical area.

7. A method according to any one of claims 1 to 5, wherein the reliability of the hypothesis is expressed as p - (d-threshold) / threshold, where p is the reliability of the hypothesis, d is the distance between said corner and the nearest detection point of said corner, and threshold is a predetermined constant.

8. A method according to any one of the preceding claims, wherein the object comprises at least two critical zones and the calculation of the reliability of the hypothesis is carried out from the position of detection points relative to each critical zone.

9. A method according to any one of the preceding claims, further comprising a hypothesis validation step in which the reliability of the hypothesis is greater than a determined threshold.

10. A method for determining the orientation of an object near a motor vehicle, the method being implemented by a computer from a cluster of object detection points collected by a radar or lidar, the method comprising: - A determination (201) of at least one hypothesis on the orientation of the object, - A determination (210) of the reliability of the hypothesis according to one of claims 1 to 7, - A comparison (220) of the reliability of the hypothesis to a determined threshold and the assignment of a confidence marker to the hypothesis, or: • When the determined reliability is greater than the threshold, a validation of the hypothesis, • When the determined reliability is less than the threshold, a rejection or a modification of the hypothesis.

11. A method according to claim 10, further comprising an adaptation of the movement of the vehicle according to the determined orientation.

12. Product computer program comprising code instructions for implementing the method according to any one of the preceding claims, when this program is executed by a computer.

13. Device for determining the orientation of objects relative to a vehicle, comprising at least one computer, a memory, a radar or lidar mounted in a vehicle, the device being configured to implement the method according to any one of claims 1 to 11.

14. Vehicle, comprising a device for determining the orientation of objects according to the preceding claim.

Images