Vehicle tracking by tracking lateral objects

Abstract

The invention relates to a method of tracking a neighboring vehicle (VV1) in the environment of a host vehicle (VP), said method being implemented by a driving assistance device (2) of the host vehicle (VP), said driving assistance device being able to acquire images within a field of view (FOV) at acquisition times (TO, T1), said method comprising: - obtaining (400), at an initial acquisition time (TO), a first set of parameters associated with a virtual object associated with the neighboring vehicle (VV1); - detecting (410), from a first image acquired at a first acquisition time (T1) following the initial acquisition time (TO), a lateral object (OBJ) associated with a second set of parameters; - tracking the neighboring vehicle (VV1) by updating (430) the first set of parameters with the second set of parameters, and, if a second criterion is met (440), tracking the neighboring vehicle by estimating (450) a length associated with the virtual object (OV).

Description

Technical Field

[0001] This invention relates to the field of driver assistance systems, and more particularly to the field of automatic detection of vehicles in regions of interest, especially in lateral regions. Background Technology

[0002] The rise of Intelligent Transportation Systems (ITS) has spurred the development of numerous systems installed in vehicles, particularly highway transport vehicles. These in-vehicle systems include, in particular, driver assistance systems or autonomous driving systems. Specifically, object detection and / or tracking, including vehicles, play a crucial role in issues such as traffic flow, road safety, and road infrastructure management (e.g., variable message signs or automatic speed cameras).

[0003] For example, in the context of systems installed in autonomous or semi-autonomous vehicles operating in road traffic (e.g., on highways), reliable and continuous tracking of objects around the vehicle in question is particularly important in order to, for example, avoid collisions between the vehicle in question and other vehicles. Specifically, in lane change assist and automatic lane change systems, reliable real-time tracking of obstacles, such as the median strip or other vehicles in the lane adjacent to the vehicle in question, is required. In fact, if the autonomous vehicle in question fails to detect, and in particular fails to track, other vehicles in the adjacent lane in real time and therefore swerves suddenly, a collision between the two vehicles is highly likely.

[0004] In this context, most existing vehicle detection and tracking methods are based on classification algorithms that identify and track vehicles from image sequences or video streams acquired by vision sensors (typically cameras). Such vision sensors are usually located at the front and / or rear of the vehicle in question to acquire images and / or image sequences of the field of view in front of and / or behind the vehicle. The classification algorithm can then identify one or more vehicles adjacent to the vehicle in question based on frontal recognition of nearby vehicles (i.e., by identifying and classifying vehicles detected in the field of view according to their frontal or rearal position). Furthermore, existing classifiers can track detected vehicles by maintaining frontal detection of the vehicles.

[0005] However, such tracking methods using vehicle classifiers prove ineffective when the front or rear of adjacent vehicles has not yet been detected in the field of view, even if these adjacent vehicles are partially visible, for example, via a partial lateral plane (or side). Therefore, if the vehicle in question is in a highway lane and an adjacent vehicle is positioned at the same level (or slightly behind) it in the adjacent lane, a camera placed behind the vehicle in question will capture the rear lateral portion of the adjacent vehicle. In such a case, the camera placed behind the vehicle in question cannot see the front of the adjacent vehicle, for example, until the speed difference between the vehicle in question and the adjacent vehicle is large enough for the vehicle in question to completely overtake the adjacent vehicle. Before such a complete overtake occurs, the adjacent vehicle is therefore essentially at the same level as the vehicle in question, and the vehicle in question changing lanes into the adjacent lane will result in a collision with the adjacent vehicle. Therefore, existing classification methods cannot detect such critical situations.

[0006] Furthermore, most existing vehicle tracking methods rely on predefined dimensions associated with the detected vehicle type. These dimensions are typically not updated, so even if vehicle tracking is implemented (e.g., through a classifier), such tracking cannot reliably determine the vehicle's size and therefore cannot accurately track the spatial extent occupied by the detected vehicle within the host vehicle's environment.

[0007] Therefore, it is necessary to ensure the safety of driver assistance and / or autonomous driving system decisions, especially in the context of lane change decisions. Specifically, it is necessary to detect, track, and understand such surrounding vehicles early and reliably, even before the front or rear of vehicles around the vehicle in question can be seen within the field of view of the vehicle's vision sensors—and therefore before such surrounding vehicles can be routinely tracked by existing classifiers. Summary of the Invention

[0008] This invention improves upon this situation.

[0009] This invention proposes a method for tracking at least one neighboring vehicle present in the environment of a master vehicle, wherein the master vehicle and the neighboring vehicle are motor vehicles, the method is implemented by a device configured to provide driving assistance functions for the master vehicle, the device being connected to at least one camera (camera), the at least one camera being mounted in the master vehicle and capable of acquiring images of the scene surrounding the master vehicle at an acquisition time according to at least one field of view, the method comprising the following steps: - During the initial acquisition time, a first set of parameters associated with a virtual object is acquired, the virtual object being associated with the detection of neighboring vehicles in the environment. - Detect objects having a vertical plane extending laterally relative to the master vehicle from at least one first image acquired after the initial acquisition time, the objects being associated with a second set of parameters. If the first criterion is met, the adjacent vehicles are tracked between the initial acquisition time and the first acquisition time by updating the first parameter set with the second parameter set, the update including estimating the length associated with the virtual object if the second criterion is met.

[0010] Therefore, the proposed method advantageously allows for real-time tracking of adjacent vehicles. In particular, such tracking is possible even when adjacent vehicles are no longer fully visible in the camera's field of view, and the detected vehicle portion corresponds only to the lateral portion of the vehicle (e.g., the front of the adjacent vehicle is no longer in the field of view). Furthermore, the method allows for updating parameters associated with the tracked vehicle, specifically the length associated with the tracked vehicle. The method can estimate the actual size of adjacent vehicles, which provides higher tracking accuracy, thereby improving road traffic safety and assistance.

[0011] The initial acquisition time can refer to the moment when the camera's field of view captures the initial state of the environment, and particularly of neighboring vehicles relative to the main vehicle. At such an initial acquisition time, an initial image can be captured where neighboring vehicles are fully contained within the field of view (e.g., the front of a neighboring vehicle is visible). The first acquisition time following the initial acquisition time can refer to an acquisition time after the initial acquisition time during which movement of neighboring vehicles and the main vehicle alters the state captured by the field of view. Specifically, a neighboring vehicle may move forward relative to the main vehicle, causing its front to no longer be visible in the field of view. For example, the field of view may only include the lateral portion of the neighboring vehicle. At the first acquisition time, the method detects (lateral) objects, which may or may not correspond to neighboring vehicles.

[0012] The first and second parameter sets can refer to one or more parameters that represent objects detected in the environment. Elements of these parameter sets may include, for example, position (or a set of positions), velocity, and acceleration. These parameter sets can be associated with three-dimensional coordinates. In particular, these parameter sets can also be associated with a set of pixels (or two-dimensional coordinates) in an acquired image.

[0013] Virtual objects can be used to refer to the modeling of neighboring vehicles detected in the environment. Such modeling can be represented in two dimensions on an image, or it can be represented as a set of locations in the environment, representing the occupancy of neighboring vehicles estimated by the method.

[0014] In another aspect, the present invention proposes an apparatus configured to provide driving assistance functions for a primary vehicle by tracking neighboring vehicles present in the environment of the primary vehicle. The apparatus is connected to at least one camera mounted in the primary vehicle and capable of acquiring images of the scene surrounding the primary vehicle at an acquisition time according to at least one field of view. The apparatus is configured to track the neighboring vehicles by implementing the proposed method.

[0015] In another aspect, the present invention proposes a computer-readable non-transitory recording medium having a program recorded thereon for implementing the proposed method when the program is executed by a processor.

[0016] The features described in the following paragraphs may be implemented optionally, independently of each other, or in combination of each other: In one implementation, the first criterion is satisfied if the difference between at least the first element in the first parameter set and the second element in the second parameter set does not exceed a predetermined first threshold.

[0017] Therefore, the method allows us to test whether an object detected at the first acquisition time could belong to (or correspond to) a virtual object, and thus belong to a detected neighboring vehicle. If the first criterion is met, the first object and the virtual object are considered to be associated and correspond to a neighboring vehicle being tracked.

[0018] The gap can be used to refer to offset, difference, or mathematical distance (e.g., Mahalanobis distance), which allows us to quantify the magnitude difference between two elements.

[0019] The first and second elements can refer to two homogeneous, comparable quantities that characterize the virtual object and the detected object, respectively. Such a first / second element could be, for example, position or velocity. In other words: In one implementation, the first parameter set includes a first location associated with a virtual object, the second parameter set includes a second location associated with a detected object, and a first criterion is satisfied at least when the difference between the first location and the second location does not exceed a predetermined first location threshold; or In one implementation, the first parameter set includes a first speed associated with a virtual object, and the second parameter set includes a second speed associated with a detected object, wherein the first criterion is satisfied if the difference between the first speed and the second speed does not exceed a predetermined first speed threshold.

[0020] In one implementation, a first pixel set and a second pixel set are associated with a virtual object and an object detected on the acquired first image, respectively, and a first criterion is met if the difference between the first pixel set and the second pixel set does not exceed a predetermined first threshold.

[0021] Therefore, the method advantageously allows testing whether virtual objects and detected objects can be correlated on an image. In particular, the method takes into account a range of uncertainties (e.g., related to the imprecision of obtaining virtual objects and / or detecting objects): even if virtual objects and detected objects do not overlap, they can still be correlated (i.e., a predetermined first threshold can be greater than 0).

[0022] The difference between sets of pixels can be called the offset or mathematical distance between two nearest pixels, each belonging to a separate set. If the intersection of two sets of pixels is not zero, then such a difference is considered zero.

[0023] In one implementation, the first parameter set includes a first position associated with a virtual object, and the second parameter set includes a second position associated with a detected object, wherein the second criterion is satisfied if the difference between the first position and the second position does not exceed a predetermined second position threshold.

[0024] Therefore, the method advantageously allows for the determination of criteria for updating parameters associated with the tracked neighboring vehicles. In fact, the relevance of such updates, particularly the updates to the estimated lengths of neighboring vehicles, may depend on the situation of virtual objects in the field of view, such as the occupancy of these virtual objects within the field of view.

[0025] In one implementation, the first set of pixels is associated with a virtual object on the acquired first image, and the second criterion is satisfied if at least one of the following conditions is met: - The size of the first pixel set does not exceed a predetermined second threshold. - The ratio between the size of the first pixel set and the number of pixels occupied by adjacent vehicles does not exceed a predetermined ratio when the front of the adjacent vehicles is visible in the field of view.

[0026] Therefore, the method advantageously allows determining the relevance of an update to the length associated with a virtual object based on the number of pixels the virtual object occupies in the acquired image and / or the percentage of the virtual object's size contained in the image. In fact, if the virtual object is still "overly" contained in the image, its length estimate is not considered relevant.

[0027] For example, when the front of an adjacent vehicle is visible in the field of view, the number of pixels occupied by that adjacent vehicle can be obtained, for example, at the initial acquisition time, by quantizing the number of pixels occupied by the virtual object at that moment.

[0028] In one implementation, a first parameter set includes a first position associated with a virtual object and a second parameter set includes a second position associated with a detected object. The first parameter set includes a first size associated with the virtual object, and the length estimation includes a value that expands one of the first sizes relative to the difference between the first and second positions.

[0029] Therefore, the method advantageously proposes to update the length of the virtual object using parameters of the detected object (which is considered to belong to an adjacent vehicle at this stage). Specifically, the expansion of the first dimension can correspond to the expansion of the estimated length associated with the tracked adjacent vehicle. Furthermore, the method advantageously allows the estimated length of the virtual object to be updated even when the detected object does not precisely overlap with the virtual object.

[0030] In one implementation, the first position and the second position correspond, respectively, to a set of coordinates in a predefined coordinate system that are associated with the portion of the environment occupied by the virtual object and the detected object.

[0031] Therefore, the method allows us to compare and quantify the differences between virtual objects and objects detected in the environment (such as objects in an image) when they correspond to a set of locations (e.g., the location set of optical flow). In other words, in the environment, virtual objects and detected objects can each correspond to a set of locations reflecting the space occupied by each object. In an image, these sets of coordinates (or locations) can correspond to the set of pixels occupied by each object captured in the field of view on the image.

[0032] In one implementation, tracking of adjacent vehicles is provided for the multiple acquisition times by repeating detection, update, and estimation steps for multiple objects detected in multiple images acquired at multiple acquisition times.

[0033] Therefore, the proposed method can track adjacent vehicles using multiple acquired images. Furthermore, the method also allows testing the correlation between virtual objects and several objects detected in the images (such detected objects may correspond to different sections of adjacent vehicles, or to other elements / objects existing in the environment that are different from the tracked vehicle).

[0034] In one implementation, tracking of a neighboring vehicle ends at the final acquisition time when the probability of its presence in the field of view associated with the neighboring vehicle is less than a predetermined threshold. The probability of its presence depends at least on the difference between the initial acquisition time and the final acquisition time, as well as a first parameter set and a second parameter set.

[0035] Therefore, the proposed method advantageously optimizes vehicle tracking resources by limiting the tracking time for each detected vehicle. For example, starting from the initial acquisition time of the first detection of a given vehicle and an estimate of its speed, the method can estimate the evolution of such vehicles in the environment (based on assumptions about the movement of the master vehicle and the tracked vehicles) to estimate a certain duration beyond which the tracked vehicle will no longer be in the field of view. Thus, the method is capable of advantageously tracking multiple objects in the environment by optimizing the use and storage of computational resources required during association and / or creation of detected objects. Such an implementation is particularly advantageous when tracking multiple vehicles and newly detected objects may be associated with each of these tracked vehicles. Attached Figure Description

[0036] Other features, details, and advantages will become apparent upon reading the following detailed description and analyzing the accompanying drawings, in which: [ Figure 1 ] Figure 1 A schematic diagram of the main vehicle according to one implementation scheme is shown.

[0037] [ Figure 2 ] Figure 2 An overhead view of the scene surrounding the main vehicle according to one implementation scheme is shown.

[0038] [ Figure 3 ] Figure 3 A schematic diagram of a driving assistance device according to one embodiment is shown.

[0039] [ Figure 4 ] Figure 4 The steps of a vehicle tracking method according to one implementation scheme are shown.

[0040] [ Figure 5 ] Figure 5 An overhead view of the scene surrounding the main vehicle according to one implementation scheme is shown.

[0041] [ Figure 6 ] Figure 6 An overhead view of the scene surrounding the main vehicle according to one implementation scheme is shown.

[0042] [ Figure 7 ] Figure 7 An overhead view of the scene surrounding the main vehicle according to one implementation scheme is shown. Detailed Implementation

[0043] refer to Figure 1 . Figure 1This schematically represents the lead vehicle (VP). The lead vehicle (VP) can be a motor vehicle. There are no restrictions on the type of vehicle to which the lead vehicle (VP) belongs. The lead vehicle (VP) can be, for example, a private vehicle, a multi-purpose vehicle, an industrial vehicle, and can correspond to, for example, a car, a van, a two-wheeler, a truck, or even a bus. The lead vehicle (VP) can also be a towed or pulling vehicle, such as a trailer, semi-trailer, or motorhome. The dimensions of the lead vehicle (VP) can be, for example, between 2 meters and 20 meters in length, between 0.5 meters and 5 meters in width, and between 1 meter and 5 meters in height.

[0044] The main vehicle VP is equipped with at least one onboard system that provides various functions or applications for the main vehicle VP. These functions may, for example, correspond to speed regulation, power steering, automatic airbag deployment, and automatic headlight adjustment.

[0045] The main vehicle (VP) is specifically equipped with an onboard system that allows the detection of objects around the VP, such as as part of driver assistance functions of the type of obstacle detection, lane change assist, or automatic lane change. For this purpose, the onboard system of the VP includes a driver assistance device 2. This device 2 is capable of providing driver assistance functions based on multiple snapshots / images (or images) of the VP's environmental ENV. For this purpose, the device 2 is specifically connected to one or more visual sensors, such as an image capture device or camera 1 (in the remainder of the description, the visual sensor will be considered as camera 1). The device 2 is also capable of transmitting or exchanging data, for example, related to the provided driver assistance functions. For this purpose, the device 2 can be connected to a communication interface 50. In one specific embodiment, the interface 50 can be integrated into the device 2. For example, such a communication interface 50 can be a human-machine interface. The interface 50 can include a display screen, a touchscreen, a dashboard, and / or a speaker. The data transmitted from the device 2 to the interface 50 can, for example, correspond to assistance information instructing the driver of the VP whether they may change lanes.

[0046] Now for reference Figure 2 . Figure 2 The environment (or scenario) where the main vehicle VP is located is shown in the ENV. Figure 2 This is an overhead view of the ENV in this scene.

[0047] The environment (ENV) can be defined in a three-dimensional reference frame (X,Y,Z), which is called the "world coordinate system". Figure 1 and 2 As shown, the origin of such a world coordinate system (X,Y,Z) is predefined and fixed in the environment ENV.

[0048] The main vehicle (VP) is considered to be moving within the scene ENV. Such a scene ENV, for example, corresponds to a road or highway consisting of multiple traffic lanes. These traffic lanes may, in particular, be parallel to each other, such as... Figure 2 As shown by the vertical dashed line in the image. For example, Figure 2 Three traffic lanes are shown, with the main vehicle VP located in the middle lane. The main vehicle VP is considered to move along the principal direction X of the reference frame (X,Y,Z), which is called the longitudinal direction. This movement occurs in... Figure 1 and Figure 2 The image is schematically illustrated by an arrow attached to the main vehicle VP. The main vehicle VP can also move along the Y direction in the reference frame (X,Y,Z)—referred to as the lateral direction. For example, such movement along the Y direction, called lateral movement (or displacement), can occur when the main vehicle VP changes traffic lanes. Displacement of the main vehicle VP along the Z direction (i.e., the height direction) is considered nonexistent or negligible. Therefore, the main vehicle VP is considered to remain on the ground and thus has a constant height along the Z direction, which corresponds to a predefined dimension of the main vehicle VP. In one embodiment, such height along the Z direction of the main vehicle VP can vary on the order of centimeters or decimeters, and the height variation along the Z direction is related, for example, to undulations or roughness present in the shock absorbers in the main vehicle VP and / or in the environmental ENV (particularly on the traffic lane of the main vehicle VP). In the following description, such height along the Z direction of the main vehicle VP is considered known.

[0049] The scene ENV surrounding the main vehicle VP also includes other element Vs and vehicles VV1, VV2, and VV3. These vehicles VV1, VV2, and VV3 are adjacent vehicles of the main vehicle VP. Like the main vehicle VP, these adjacent vehicles VV1, VV2, and VV3 are motorized vehicles moving within the scene ENV. These adjacent vehicles VV1, VV2, and VV3 may specifically have similar or different dimensions and movement characteristics (e.g., in terms of speed or acceleration) than the main vehicle VP. For example, Figure 2 An object can represent a primary vehicle (VP) moving in a road lane and adjacent vehicles (VV1, VV2, VV3) traveling in a road lane adjacent to the lane used by the primary vehicle (VP). An object (OV) is a neighboring element of the primary vehicle (VP) and can refer to any element different from the neighboring vehicles (VV1, VV2, VV3). For example, a neighboring element (OV) can correspond to an obstacle in the scene, such as a median separating two traffic lanes, a sign, or even a bird flying in the scene (ENV).

[0050] like Figure 2 As shown, the main vehicle VP is believed to be equipped with a camera 1 positioned at the rear of the main vehicle VP. In another embodiment ( Figure 2(Not shown in the image), camera 1 can be located at the front of the main vehicle VP, or several cameras 1 can be located simultaneously at the front and rear of the main vehicle VP. In the context of this description, camera 1 is considered to be located at the rear of the main vehicle VP, and the field of view (FOV) is the rear field of view of the main vehicle VP. Camera 1 allows for the capture of images (or videos) of the scene ENV of the main vehicle VP within the FOV. Such a FOV depends particularly on the type of camera 1 used and the positioning of camera 1 within (or on) the main vehicle VP. Reference Figure 2 The field of view (FOV) covers a portion of the scene's enveloping volume (ENV), allowing the capture of a limited portion of the scene's ENV based on the FOV. Therefore, in Figure 2 In the diagram, the field of view (FOV) covers a portion of the lane where the main vehicle (VP) is located, as well as corresponding portions of adjacent traffic lanes. Specifically, in Figure 2 Considering time T1, the field of view (FOV) covers a portion of the scene's enveloping volume (ENV) such that adjacent vehicle VV2 is completely located within this portion, and adjacent vehicle VV1 is partially located within this portion (the gray area not shaded). As an example, such as... Figure 2 As shown, the left rear end of adjacent vehicle VV1 (e.g., including the left rear wheel of adjacent vehicle VV2) belongs to the field of view (FOV). However, the marker point W, for example, located in the middle of the front of the vehicle, is not within the FOV. Figure 2 Within the field of view (FOV). For comparison, the same adjacent vehicle VV1 is shown at an initial time T0 prior to the time under consideration: such adjacent vehicle VV1 is within... Figure 2 The image is shown in dashed lines. At the initial time T0, the adjacent vehicle VV1 is fully visible within the field of view (FOV) of camera 1. Specifically, at such an initial time T0, for example, at the marker point W located in the middle of the front of the adjacent vehicle VV1... n Visible within the field of view (FOV).

[0051] At the time under consideration—which is after the initial time T0—a portion of the adjacent vehicle VV1 (specifically including marker point W) is no longer visible in the field of view (FOV) of the master vehicle VP. For example, Figure 2 The situation of the adjacent vehicle VV1 shown between time T0 and time T1 can correspond to the situation where the main vehicle VP is overtaken from the right by the adjacent vehicle VV1.

[0052] exist Figure 2 In this context, the adjacent vehicle VV3 is not visible within the field of view (FOV) of the primary vehicle VP. Figure 2 The shaded areas represent the parts of adjacent vehicles VV1 and VV3 that are not visible in the field of view (FOV). Adjacent objects OV (such as birds) are visible in the FOV.

[0053] In such Figure 2In the context of the scene ENV represented by the coordinate system (X,Y,Z), it is assumed that, for example, the main vehicle VP has a known main velocity. Movement. For ease of subsequent description, this main velocity... It is considered to have a constant value V VP The adjacent vehicle VV1 is considered to have an adjacent speed relative to the driver assistance device 2 of the main vehicle VP that is unknown. Movement. In the following implementation scheme, the speeds of adjacent vehicles can be considered... The value V VP Greater than the main speed The value V VP For example, consider the case where an adjacent vehicle VV1 overtakes the main vehicle VP in the adjacent lane. Under this assumption, and assuming the adjacent speeds of the adjacent vehicles VV1... It remains substantially constant over the considered time interval, just like the main speed of the main vehicle VP. Similarly, the distance difference in the longitudinal X direction between the main vehicle VP and the adjacent vehicle VV1 will change over time, such that, for example, the adjacent vehicle VV1 will approach the main vehicle VP when overtaking it (in the longitudinal X direction), and then move away from the main vehicle VP once the overtaking is complete (in the longitudinal X direction). Therefore, at the considered initial time T0 (or initial acquisition time T0), the adjacent vehicle VV1 can be fully visible in the rear field of view (FOV) of the main vehicle VP. When the adjacent vehicle VV1 performs an overtaking maneuver, the adjacent vehicle VV1 can gradually "move out" of the rear field of view (FOV) until it is completely out of the FOV (as with, for example, adjacent vehicle VV3). For example, in Figure 5 and Figure 6-7 The spaces between them represent the evolution of adjacent vehicles VV1 within the field of view (FOV).

[0054] Figure 5 This represents the image captured by camera 1 at the initial acquisition time T0. At this initial acquisition time T0, the adjacent vehicle VV1 is completely within the field of view (FOV). Specifically, the front of the adjacent vehicle VV1 is visible within the FOV.

[0055] Figure 6 and Figure 7 This represents an image acquired by camera 1 at a first acquisition time T1 after the initial acquisition time T0, according to two different implementation schemes. At this first acquisition time T1, the neighboring vehicle VV1 has advanced relative to the main vehicle VP within the environment ENV, causing the neighboring vehicle VV1 to gradually reach the level of the main vehicle VP. Specifically, the front face of the neighboring vehicle VV1 is no longer visible in the field of view (FOV). At the first acquisition time T1, only the front face of the neighboring vehicle VV1 is visible. Figure 6 and Figure 7The lateral portion shown includes, for example, the left rear wheel of the adjacent vehicle VV1. In fact, Figure 6 and Figure 7 Each of the frames shown in the diagram defines the shooting range and corresponds to the field of view (FOV). Therefore, elements outside the frames (which will be described later) are shown for illustrative purposes and are not visible from the perspective of the FOV.

[0056] In the context of driver assistance functions provided by device 2 (intended to, for example, assist the primary vehicle VP in lane-changing operations), a process is needed for tracking vehicles detected in the environmental ENV of the primary vehicle VP, such that if the primary vehicle VP is at risk of colliding with one of the surrounding vehicles, the primary vehicle will not move into an adjacent traffic lane. Existing techniques for detecting and tracking vehicles around the primary vehicle VP rely heavily on the use of classifiers, such as those based on… Convolutional Neural Networks (CNN) K-Nearest Neighbors Algorithm (KNN) or Support Vector Machine (SVM). Such classifiers, capable of detecting and tracking vehicles around the main vehicle (VP), rely particularly on the detection and classification of frontal views (front or rear views) of vehicles surrounding the VP. The use of such classifiers may specifically include a learning phase based on multiple images representing frontal views of various vehicle types. Therefore, referencing... Figure 2 This refers to the frontal view of the adjacent vehicle VV2 within the field of view (FOV) of camera 1, which allows for the detection and tracking of adjacent vehicle VV2 based on existing classification techniques. This also applies to... Figure 5 The adjacent vehicle VV1 in the image at the initial acquisition time T0.

[0057] However, by Figure 2 Time T1 and Figure 6 and Figure 7 The positions of adjacent vehicles VV1 shown correspond to the following situation: due to Figure 6 and Figure 7 There is no visible front view of the adjacent vehicle VV1 in the image. Figure 2 (The marker point W located on the front of the adjacent vehicle VV1 is no longer visible in the FOV of camera 1 during these stages). Existing classifiers cannot effectively track adjacent vehicle VV1, or at least cannot do so without incurring significant cost and / or computation time. These situations correspond to critical situations: during which, without lane change assist information indicating that an adjacent vehicle VV1 has been detected near such a location, if the primary vehicle VP changes position and moves into the lane of the adjacent vehicle VV1, the primary vehicle VP may collide with the adjacent vehicle VV1.

[0058] Therefore, a method for tracking adjacent vehicles VV1 is proposed, and through... Figure 4Describe the process to track neighboring vehicles VV1, especially... Figure 2 , Figure 6 and Figure 7 The stage of the situation shown. Such a tracking method can be specifically described as a method that continues tracking even when a vehicle leaves the field of view (FOV), that is, even if the front face (front side) of the vehicle is beyond the FOV (and therefore no longer visible in the FOV), the initially detected vehicle can still be tracked. This method of tracking adjacent vehicles VV1 is... Figure 3 The driving assistance system shown is implemented.

[0059] Now for reference Figure 3 . Figure 3 This diagram illustrates the in-vehicle system of the main vehicle VP. Specifically, such an in-vehicle system corresponds to the driver assistance system of the main vehicle VP. When the main vehicle VP is, for example... Figure 2 When moving in the lane shown, the driver assistance system of the main vehicle VP allows, in particular, the lane change assist or automatic lane change function of the main vehicle VP.

[0060] The system first includes a driving assistance device 2. The device 2 itself may include a unit 20 for detecting objects on the acquired images, a unit 30 for comparing the objects detected by the unit 20 with existing virtual objects, and a unit 40 for updating parameters associated with existing virtual objects to track elements in the environment ENV associated with existing virtual objects.

[0061] On one hand, the driving assistance device 2 is also connected to the visual sensor of the imaging device or camera 1. The device 2 may also include an input unit ( Figure 3 (Not shown in the image) to allow device 2 to receive data streams from camera 1 substantially in real time. Depending on the field of view (FOV) of camera 1, such a data stream corresponds to a discrete or continuous sequence of images (or videos) of the scene's envelopment (ENV). Each received image is associated with the image acquisition time of camera 1. Each image may also include a timestamp.

[0062] On the other hand, the driver assistance device 2 can be connected to the communication interface 70. Such an interface 70 can correspond to a human-machine interface integrated into the onboard system of the main vehicle VP. Such an interface 70 can also be integrated into the device 2. The communication interface 70 can also be a remote interface. The communication interface 70 can include a display screen, touchscreen, dashboard, or speaker, which allows the transmission of instructions related to driver assistance of the main vehicle VP, for example, via visual, tactile, and / or auditory information. In particular, the communication interface allows the transmission of the detection status of vehicles adjacent to the main vehicle VP or instructions related to the possibility of lane changes by the main vehicle VP. Such information transmitted by the interface 70 can, for example, correspond to an aerial visual representation of the position of the main vehicle VP and the positions of adjacent elements in the field of view (FOV) in real time (e.g., as shown in the image). Figure 2 (as shown), audio stimuli warning of collision risks, or virtual objects that are highlighted and updated in real time and superimposed on the image stream from camera 1, for example Figure 5 , Figure 6 and Figure 7 As shown.

[0063] Units 20, 30, and 40 of device 2 may each include processing circuitry, which includes at least one processor (21, 31, 41) and a storage unit (22, 32, 42) to implement one or more steps of a method for tracking adjacent vehicles VV1, which will... Figure 4 As described in the text. Specifically, each processing unit 20, 30, and 40 of device 2 can rely on data processed and / or acquired by other units to implement one or more steps of the method for tracking adjacent vehicles VV1, such as... Figure 3 As shown by the arrow in the image.

[0064] Now for reference Figure 4 . Figure 4 A series of steps for implementing a method to track previously detected neighboring vehicles VV1 are shown. For example... Figure 3 As shown, such a tracking method can be implemented through a system including the driver assistance device 2 of the main vehicle VP.

[0065] Throughout the method described below, apparatus 2 receives from camera 1 multiple images (or videos) associated with each image acquisition time. Such images can be received continuously, for example, via a video stream. In such cases, apparatus 2 can discretize the received video stream to obtain a set of discrete, timestamped images associated with the corresponding acquisition times.

[0066] In step 400, the adjacent vehicle VV1 has been detected in the initial image acquired at the initial acquisition time T0. For example, Figure 5An initial image is shown. At this initial acquisition time T0, the neighboring vehicle VV1 is fully visible in the field of view (FOV). Specifically, the front face of the neighboring vehicle VV1 is visible in the FOV. The neighboring vehicle VV1 can then be detected by conventional methods, such as by a vehicle classifier based on identifying different types of vehicle front faces. In another embodiment, the neighboring vehicle VV1 can be detected by other methods, such as methods based on optical flow detection. In step 400, the neighboring vehicle VV1 is then detected on the initial image, and a first position associated with the neighboring vehicle VV1 can be determined in the image reference frame (y,z) and / or the world reference frame (X,Y,Z). In one implementation, the first location associated with the adjacent vehicle VV1 This can correspond to a set of coordinates associated with a first portion of the environmental ENV occupied by the adjacent vehicle VV1. This first set of coordinates can be represented in three dimensions according to the world coordinate system (X,Y,Z), such that the first position reflects the occupied space of the environmental ENV occupied by the detected adjacent vehicle VV1. Equivalently, this first set of coordinates can also be represented in two dimensions according to the image reference system (y,z), such that the first position reflects the occupied space of the initial image by the adjacent vehicle VV1 detected on the initial image. The first position is defined between the image reference system (y,z) and the world coordinate system (X,Y,Z) image. The transformation can be obtained from the pinhole model and assumptions (e.g., the planar world assumption). Alternatively, the first position associated with the adjacent vehicle VV1. This can correspond to the first marker point W detected on the adjacent vehicle VV1. n The location, for example, on the front of the adjacent vehicle VV1. Such a marker point W n It can be represented using the world coordinate system (X,Y,Z), or equivalently, it can be a (virtual) marker point w represented in the image reference system (y,z). n (For example, corresponding to, for example, a pixel). For example, Figure 5 This indicates the marker point w. n .

[0067] More generally, in step 400, the detected neighboring vehicle VV1 can be associated with a first parameter set, which may include one or more of the following parameters: - First Position , - First speed It reflects the speed of adjacent vehicle VV1 detected in the environmental ENV. - First Size This includes, for example, a first value of length, width, and / or height associated with a neighboring vehicle VV1 detected in the environmental ENV. - First collision time associated with the detected adjacent vehicle VV1 , - First lateral distance This reflects, for example, the lateral offset that separates the detected neighboring vehicle VV1 from the main vehicle VP (i.e., the so-called lateral Y in the world coordinate system (X,Y,Z) or y in the image reference system (y,z).

[0068] For example, such parameters can be determined or estimated using one or more classifiers based on one or more initial images acquired in association with the initial acquisition time T0. Such parameters can also be estimated based on measurements taken from the acquired initial images.

[0069] Based on this first set of parameters, the neighboring vehicle VV1 detected in step 400 can be compared with the virtual object. Related. For example, such virtual objects. This can correspond to a set of pixels (and thus to two-dimensional coordinates in the image reference frame (y,z)), which can be superimposed on the space occupied by adjacent vehicles VV1 in the initial image. Specifically, virtual objects... It can also correspond to a three-dimensional virtual object in the world coordinate system (X,Y,Z), thus allowing modeling of the space occupied by adjacent vehicles VV1 in the environment ENV. Such virtual objects... The feature is one or more kinematic parameters, which correspond to a first parameter set and allow the detected neighboring vehicle VV1 to be defined in time (here, the initial acquisition time T0) and space (here, the environment ENV). Therefore, at the initial acquisition time, the neighboring vehicle VV1 is detected and detected via a virtual object. Modeling (or, equivalently, modeling using the first set of parameters).

[0070] In step 410, object OBJ is detected on at least one of the acquired images (referred to as the first image). References represent two possible variations of the detected object OBJ. Figure 6 and Figure 7The detected object OBJ is represented by the first acquisition time T1. In one embodiment, the detected object OBJ is a so-called lateral object in the acquired image because the detected object OBJ1 belongs to the lateral plane of the image, such a lateral plane being, for example, parallel to the plane (X,Z) in the world coordinate system (X,Y,Z). This lateral object OBJ can also be detected when the front of an adjacent vehicle VV1 is not visible in the field of view (FOV). Specifically, at the stage of step 410, the detected object OBJ corresponds to a non-inertial body that has not been a priori assimilated into the previously detected adjacent vehicle VV1. In other words, at the stage of step 410, the detected object OBJ corresponds to a moving body in the environment ENV (and therefore in the world coordinate system (X,Y,Z)) and thus exhibits at least one velocity, referred to as a second velocity. (For example, so-called longitudinal velocity, which corresponds to displacement along the X-axis in the world coordinate system (X,Y,Z); so-called lateral velocity, which corresponds to displacement along the Y-axis; and / or rotational velocity).

[0071] In step 410, an object OBJ can be detected on an image by processing one or more consecutively acquired images (e.g., Figure 6 Images or Figure 7 (Image in the image). In one implementation, the object OBJ can be detected by, for example, a classifier capable of detecting lateral objects OBJ on the acquired image, such as a classifier for detecting lateral portions of a vehicle, like a vehicle wheel detector. As a variation, the object OBJ can be detected by performing a homography transformation on at least two consecutively acquired images to identify one or more related sets or optical flows with relatively uniform velocity on the image within a defined time interval. According to such a variation, the detection of said object OBJ1 then includes determining the displacement of pixels at a substantially common velocity from one image to the next, i.e., optical flow. Each optical flow can be determined by point matching between multiple consecutive images, for example by the Lucas-Kanade method. The determined optical flow can be segmented, in particular, into one or more segments having common points or vanishing lines and belonging to lateral planes within the field of view (FOV). In other words, the acquired image can be segmented into one or more segments, each segment defining a set of optical flows (and therefore a set of pixels with substantially the same velocity and describing the same movement within a given time interval).

[0072] Therefore, in step 410, the detected object OBJ is a certain set of pixels, which can be, for example, instances, contours, polygons, or even "..." detected and classified by a classifier through segmentation of objects in the acquired image. boundary box"Formed. As a variant, the detected object OBJ is a certain set of pixels formed by optical flow or segments containing several optical flows. In particular, the set of pixels that forms the detected object OBJ can be a set of coplanar pixels parallel to the horizontal plane in the world coordinate system (X,Y,Z).

[0073] Specifically, the detected object OBJ can be associated with a second parameter set, which may include, for example, one or more of the following parameters: - Second position , - Second speed , - Second size This includes, for example, a second value associated with the length, width, and / or height of the object OBJ detected in the environment ENV. - Second collision time associated with the detected object OBJ , - Second lateral distance This reflects, for example, the lateral offset that separates the detected object OBJ from the main vehicle VP (i.e., the so-called lateral Y in the world coordinate system (X,Y,Z) or y in the image reference system (y,z).

[0074] and virtual objects The first associated position Similarly, the second location associated with the detected object OBJ This can correspond to a set of coordinates (in the world coordinate system (X,Y,Z) and / or the image reference system (y,z)) representing a portion of the environment ENV occupied by the detected object OBJ, or, as a variation, to the position of a reference point belonging to the detected object OBJ. For example, such a reference point could correspond to the pixel closest to the ground.

[0075] For example, the second set of parameters can be determined or estimated by one or more classifiers based on the first image acquired in association with the first acquisition time T1. Such parameters can also be estimated based on measurements performed on the first image.

[0076] Therefore, in step 420, the device, on the one hand, obtains a virtual object associated with a first set of parameters representing the initially detected neighboring vehicle VV1. Furthermore, on the other hand, the detected object OBJ associated with the second parameter set is obtained.

[0077] In step 420, a consistency (or reasonableness) test is performed to determine the detected object OBJ and virtual object. Does it correspond to the same element in the environment ENV, i.e., adjacent vehicle VV1? For this purpose, conformance testing may include testing against a first standard. For example, such a first standard could correspond to a virtual object... The first associated position and the second location associated with the detected object OBJ A comparison between them. If the first position Second position If the offset between the two positions is less than a predetermined first position threshold, then the first criterion is met. Specifically, if the first position... Second position These correspond to virtual objects respectively. marker point Given the coordinates of the first position and the coordinates of the reference point belonging to the detected object OBJ, the first position threshold can be defined at the first position. Second position On the horizontal coordinate (i.e., along the X-axis), respectively denoted as and In other words, a consistency test can be passed if the following conditions are met: Mathematical Formula 1

[0078] in: - It is the first position The coordinates along the X-axis in the world coordinate system (X,Y,Z). - It is the second position. The coordinates along the X-axis in the world coordinate system (X,Y,Z). - It is the predetermined first position threshold.

[0079] As a variant, if the first position Second position Corresponding to the coordinate set (i.e., reflecting the virtual object) If the offset between two nearest points (or pixels) (which belong to two different coordinate sets) is less than the first position threshold, then the first criterion can be satisfied.

[0080] As a variant, the first criterion can also be divided by position in the parameter set. , Other parameters besides these are relevant. In one implementation, if the first speed... Second speed If the difference between them does not exceed the predetermined first speed threshold, then the first criterion can be met.

[0081] As a variation, the first criterion can be satisfied if the Mahalanobis distance determined between the first parameter set and the second parameter set does not exceed a predetermined first threshold.

[0082] The first criterion can also be related to one or more of the comparisons mentioned above.

[0083] If, at the end of step 420, the detected object OBJ is not determined to belong to the pre-existing virtual object OV. n (In other words, if the first criterion is not met), then step 421 can be implemented. Such step 421 may in particular include creating a virtual object different from the one associated with the adjacent vehicle VV1. A new virtual object is then created. In other words, if the first criterion is not met, the detected object OBJ is determined to reflect the movement of a mobile body in the environment that is different from the adjacent vehicle VV1. Such a mobile body could, for example, correspond to another vehicle different from the adjacent vehicle VV1, an obstacle in the environment ENV, or a bird, etc. A new virtual object can then be created based on the second set of parameters. In another embodiment, if the detected object OBJ is not determined to belong to a pre-existing virtual object OV, a new virtual object can be created. n If so, the detected object OBJ can be ignored, and steps 400 and 410 can be repeated until the detected object OBJ is identified as belonging to a pre-existing virtual object OV. n (In other words, the method can attempt to track a specific vehicle VV1).

[0084] On the other hand, if at the end of step 420 it is determined that the detected object OBJ belongs to a pre-existing virtual object. (In other words, if the first criterion is met), then proceed to step 430.

[0085] In step 430, the pre-existing virtual object is processed. Update in order to obtain updated virtual objects. To this end, a second set of parameters can be fed into the data estimator, for example, into a Kalman filter of the extended Kalman filter (EKF) type. Such a data estimator can also obtain a first set of parameters (e.g., supplied by one or more classifiers of the system). Based on the parameters provided to the data estimator, apparatus 2 can obtain an updated set of parameters in step 430. This updated set of parameters can be estimated, in particular, by the Kalman filter on the one hand, based on the provided second set of parameters, and on the other hand, based on the estimation predictions made by the Kalman filter based on at least the provided first set of parameters.

[0086] Therefore, in step 430, device 2 can obtain an updated parameter set. Specifically, such an updated parameter set may include at least: - Updated location It is at least based on the first position Second position Confirmed, and - Update speed It is at least based on the first velocity Second speed Sure.

[0087] Based on these updated parameters, virtual objects associated with adjacent vehicle VV1 can be... Update to the latest virtual object This allows tracking of neighboring vehicles VV1 moving within the environment ENV between times T0 and T1. Specifically, the updated marker points... Can be updated virtual objects Related. For example, Figure 6 and Figure 7 This updated marker point is shown. .

[0088] In most existing tracking methods, the size associated with the tracked neighboring vehicle VV1 (i.e., the first size) is... ) corresponds to a pre-existing size (e.g., the size of a typical truck), so that it can be determined based on the marked points (e.g., The space occupied by the detected vehicle is estimated using the virtual object and its predetermined dimensions. This is done when updating the virtual object. At that time, this size will be preserved, and the updated virtual object will be updated. The space occupied will be moved accordingly.

[0089] In the context of this invention, updating the dimensions associated with adjacent vehicle VV1 is also considered, in order to allow, in particular, estimation of the actual length of the detected adjacent vehicle VV1.

[0090] Therefore, in step 440, a length estimation test is performed to determine whether to perform a length estimation test on a pre-existing virtual object. The length is updated. In step 440, the second standard is tested.

[0091] For example, a second criterion could include estimating the size of a portion of an adjacent vehicle VV1 that remains visible within the field of view (FOV). For this purpose, virtual objects could be estimated. The space occupied in the acquired image (represented by the number of pixels occupied in the image). In one implementation, the virtual object can be estimated. The horizontal space occupied in the acquired image (represented by the number of pixels in the image). If a virtual object... If the number of pixels occupied in the image does not exceed a predetermined pixel threshold, such as 100 pixels, then the second criterion is met. As a variation, if the virtual object... In the images under discussion (e.g.) Figure 6 or Figure 7 The number of pixels occupied by the virtual object In the initial image ( Figure 5 If the ratio between the number of pixels occupied by a given element does not exceed a predetermined percentage of pixels, such as 10%, then the second criterion is met. More generally, other variations can be determined if the ratio is smaller than the entire virtual object. If a given percentage is visible in the image, then the second criterion is met.

[0092] In fact, if an adjacent vehicle VV1 is completely or almost completely within the field of view (FOV), then updating the length of the adjacent vehicle VV1 can be considered overly inaccurate. The FOV does indeed affect the viewpoint relative to virtual objects. The associated lengths produce deviations.

[0093] In one implementation, if with virtual objects The first associated position and the second location associated with the detected object OBJ If the offset δ between them does not exceed a predetermined second position threshold, then the second criterion is satisfied. Similar to the first criterion described in step 420 according to an embodiment (e.g., the embodiment of mathematical formula 1), if the virtual object... If the spaces occupied by the detected object OBJ substantially overlap, then the second criterion is satisfied. This predetermined second position threshold can be similar to or different from the first position threshold. In particular, the existence of the second position threshold means that even virtual objects... If there is no overlap with the detected object OBJ, the second criterion can still be met. (Reference) Figure 6 It illustrates an implementation scheme in which virtual objects The offset between the detected object OBJ and the object itself is considered zero, i.e., a virtual object. Overlapping with the detected object OBJ, such as Figure 6 The shaded area in the image represents this. (See reference.) Figure 7 It illustrates an implementation scheme in which virtual objects The offset between the detected object OBJ and the virtual object is not zero. Non-overlapping with detected object OBJ: Virtual object There is a non-zero offset δ between the detected object OBJ and the object itself, such as... Figure 7 As shown. Specifically, the tolerance of offset δ being less than a predetermined second position threshold means that if the detected object OBJ is related to the position as represented by... Virtual objects whose horizontal components extend virtually. If the values ​​(corresponding to the second position threshold) overlap, then the second criterion is considered to be met. Figure 7 The Chinese side considers that such conditions have been met. In one implementation, based on one or two of the above comparisons, the second criterion can be considered to be met.

[0094] If the second criterion is not met at the end of step 440, then step 451 is performed. In step 451, the updated dimensions associated with the adjacent vehicle VV1 are determined. Reserved, and corresponding to .

[0095] If the second criterion is met at the end of step 440, then step 450, estimating the updated length of the adjacent vehicle VV1, is performed. In step 450, the length associated with the adjacent vehicle VV1 and the initial size is determined. Different update sizes Specifically, the updated dimensions are determined in step 450. The updated horizontal component is denoted as Therefore, a second dimension associated with the detected object OBJ is considered. Specifically, consider the distance between the virtual object and the detected object OBJ. The horizontal coordinate associated with the farthest pixel. This horizontal coordinate then corresponds to the updated horizontal coordinate of the end pixel of the updated virtual object. In other words, horizontal objects. Extend laterally until it encompasses the detected object OBJ in the lateral direction. The value of this extension corresponds to the compensation for the first position. Second position The offset between them (based on the horizontal component) makes the updated virtual object Extend until it occupies the space occupied by the detected object OBJ in the horizontal direction. Figure 6 and Figure 7 Each of these demonstrates such an extension in order to obtain updated virtual objects. Specifically, the virtual objects that are determined and updated. Related updated size This allows for the estimation of the length of adjacent vehicle VV1.

[0096] Therefore, it is possible to track neighboring vehicles VV1 between time T0 and T1, while also allowing the estimation of the length of neighboring vehicles VV1.

Claims

1. A method for tracking at least one neighboring vehicle (VV1) present in an environment (ENV) of a master vehicle (VP), wherein the master vehicle (VP) and the neighboring vehicle (VV1) are motor vehicles, the method being implemented by a device (2) configured to provide driving assistance functions for the master vehicle (VP), the device (2) being connected to at least one camera (1), the at least one camera being mounted in the master vehicle (VP) and capable of acquiring images of the scene surrounding the master vehicle (VP) according to at least one field of view (FOV) at acquisition times (T0, T1). The method includes the following steps: - At the initial acquisition time (T0), (400) and the virtual object ( The first set of parameters associated with the virtual object ( This is associated with the detection of the adjacent vehicle (VV1) in the said environment (ENV). - Detect (410) an object (OBJ) having a vertical plane extending laterally relative to the master vehicle (VP) from at least one first image acquired after the initial acquisition time (T0) and at a first acquisition time (T1), the object (OBJ) being associated with a second parameter set. - If the first criterion (420) is met, the neighboring vehicle (VV1) is tracked between the initial acquisition time (T0) and the first acquisition time (T1) by updating (430) the first parameter set with the second parameter set, the update (430) including estimating (450) the length associated with the virtual object (OV) if the second criterion (440) is met.

2. The method according to claim 1, wherein, The first criterion is satisfied if the difference between at least one first element in the first parameter set and the second element in the second parameter set does not exceed a predetermined first threshold.

3. The method according to any one of the preceding claims, wherein the first parameter set includes parameters related to the virtual object ( The first position associated with it () The second parameter set includes a second location associated with the detected object (OBJ). ),in, If the first position ( ) and the second position ( If the difference between the two positions does not exceed a predetermined first position threshold, then the first criterion is satisfied.

4. The method according to any one of the preceding claims, wherein the first parameter set includes parameters related to the virtual object ( The first velocity associated with it () The second set of parameters includes a second velocity associated with the detected object (OBJ). ),in, At least if the first velocity ( ) and the second speed ( If the difference between the two speeds does not exceed a predetermined first speed threshold, then the first criterion is satisfied.

5. The method according to any one of the preceding claims, wherein the first pixel set and the second pixel set are respectively associated with the virtual object ( The image is associated with the object (OBJ) detected on the first image acquired, wherein... If the difference between the first pixel set and the second pixel set does not exceed a predetermined first threshold, then the first criterion is met.

6. The method according to any one of the preceding claims, wherein the first parameter set includes parameters related to the virtual object ( The first position associated with it () The second parameter set includes a second location associated with the detected object (OBJ). ),in, At least if the first position ( ) and the second position ( If the difference between the two positions does not exceed a predetermined second position threshold, then the second criterion is satisfied.

7. The method according to any one of the preceding claims, wherein the first pixel set and a virtual object on the acquired first image ( Related to, among which, The second criterion is satisfied if at least one of the following conditions is met: - The size of the first pixel set does not exceed a predetermined second threshold. - The ratio between the size of the first pixel set and the number of pixels occupied by the neighboring vehicle (VV1) does not exceed a predetermined ratio when the front of the neighboring vehicle (VV1) is visible in the field of view (FOV).

8. The method according to any one of the preceding claims, wherein the first parameter set includes parameters related to the virtual object ( The first position associated with it () The second parameter set includes a second location associated with the detected object (OBJ). The first parameter set includes parameters related to the virtual object ( The first dimension associated with it () ),in, The estimation of the length includes taking the first dimension ( ) dimensions ( ) Extend with the first position ( ) and the second position ( The values ​​related to the differences between ).

9. The method according to any one of the preceding claims, wherein the first parameter set includes parameters related to the virtual object ( The first position associated with it () The second parameter set includes a second location associated with the detected object (OBJ). ),in, The first position ( ) and the second position ( In the predefined coordinate system, they correspond respectively to the virtual object ( The coordinate set associated with the portion of the environment (ENV) occupied by the detected object (OBJ).

10. The method according to any one of the preceding claims, wherein, By repeating the detection step (410), update step (430), and estimation step (450) on multiple objects detected in multiple images acquired at multiple acquisition times, tracking of the adjacent vehicle (VV1) is provided for the multiple acquisition times.

11. The method according to any one of the preceding claims, wherein, When the probability of the neighboring vehicle (VV1) being present in the field of view (FOV) is less than a predetermined threshold, the tracking of the neighboring vehicle (VV1) ends at the final acquisition time, the probability of being present depending at least on the difference between the initial acquisition time (T0) and the final acquisition time, as well as the first parameter set and the second parameter set.

12. An apparatus (2) configured to provide driving assistance to a primary vehicle (VP) by tracking neighboring vehicles (VV1) present in the environment (ENV) of the primary vehicle (VP), the apparatus (2) being connected to at least one camera (1), the at least one camera being mounted in the primary vehicle (VP) and capable of acquiring images of the scene surrounding the primary vehicle (VP) according to at least one field of view (FOV) at acquisition times (T0, T1), the apparatus (2) being configured to track the neighboring vehicles (VV1) by means of a processor of the apparatus (2) implementing the method according to any one of claims 1 to 11.

13. A computer-readable non-transitory recording medium having a program recorded thereon, the program being used to implement the method according to any one of claims 1 to 11 when executed by a processor.

Images