FR3137202A1 - Method for detecting a global concavity on a polygonal line - Google Patents

Abstract

The invention relates to a method for detecting a global concavity on a polygonal line (L1), characterized in that it comprises at least a first corner detection step (E1), which executes a line simplification algorithm polygonal line (L1) aimed at limiting the number of points belonging to the polygonal line (L1), the retained points being considered as corners, a corner being considered concave if the angle it forms is reentrant, from the point of view of said image acquisition sensor (16), a second noise detection step (E2) which executes a noise detection algorithm consisting of identifying the corners of the polygonal line (L1) considered to be noise, among the corners identified during the first step (E1), and a third step (E3) of identifying a global concavity, which consists of successively calculating the area of each of the triangles delimited by a corner identified as concave, the previous corner and the next corner of the polygonal line (L1), if the calculated area is greater than a reference area, then said concave identified corner is considered sufficiently concave. Figure for abstract: Figure 1

Description

Method for detecting a global concavity on a polygonal line

The present invention relates to the field of detection of target objects delimited by a polygonal line coming from a cloud of points captured by a three-dimensional data acquisition device, and more particularly to the detection of an overall concavity on a polygonal line.

Also, the present invention is particularly suitable for implementation by a computer which equips a vehicle, in particular an automobile.

It is known to equip a motor vehicle with a driving assistance system commonly called by the English acronym ADAS for “Advanced Driver Assistance System”.

Such an assistance system comprises, in known manner, at least one image acquisition sensor which is mounted on the vehicle and which makes it possible to generate a series of images representing the environment of the vehicle.

Generally, the image acquisition sensor is a three-dimensional data acquisition device such as a Lidar (English acronym for “Light Detection And Ranging”).

The acquired images are used by a computer in order to assist the driver, for example by detecting an “object” (pedestrians, stationary vehicle, objects on the road, etc.), and by calculating for example the time before collision with the detected object.

Lidar is an active optical sensor which emits a laser beam along an optical axis of sight in the direction of a target object.

The reflection of the laser beam from the surface of the target object is detected by a receiving surface arranged in the lidar.

This receiving surface records the time between when the laser pulse is emitted and when it is received by the lidar to calculate the distance between the lidar and the object.

The distance measurements collected by the lidar, associated with positioning information, are transformed into measurements of real three-dimensional points of the target object, each image acquired by the lidar forming a set of three-dimensional points, or cloud of points, which are considered as a sample of a surface.

By three-dimensional point we mean a point which is associated with coordinates in space, in a given three-dimensional coordinate system.

Also, the surface of the detected objects can be represented by a two-dimensional polygonal line which is formed by a series of successive points, which come from the points collected by the lidar.

We speak of a polygon when the polygonal line is closed.

But generally, an object observed by an image acquisition sensor is not visible in its entirety and will therefore be represented by a polygonal line, and not by a closed polygon.

The points that form the polygonal line are connected by consecutive line segments that each connect two consecutive points of the polygonal line.

Two consecutive straight line segments form a concave angle, or a concave corner, if this angle is reentrant, from the point of view of the image acquisition sensor, that is to say oriented towards the interior of the object drawn by the polygonal line.

Conversely, an angle is convex if it is outward, from the point of view of the image acquisition sensor, that is to say oriented towards the outside of the object drawn by the polygonal line.

Also, it is known to detect a concave angle on a polygonal line in order to differentiate two objects drawn by a common polygonal line.

Indeed, it is important to determine whether the polygonal line represents a single object or several objects, particularly in the field of assistance in driving a motor vehicle, so as to interpret the nature of the target objects correctly. .

A concave angle on the polygonal line, that is to say an angle which goes towards the interior of the object, is generally considered as a break between two neighboring objects.

Existing techniques for detecting a concave angle on a polygonal line are local approaches, i.e. the angle is calculated between three consecutive points on the polygonal line.

In other words, with a local approach, the angle is calculated between two consecutive segments of the polygonal line.

A disadvantage of this type of concave angle detection method using a local approach is the risk of errors caused by misinterpretation of a calculated concave angle.

Indeed, a polygonal line can sometimes present a concave angle locally, without this concave angle materializing a break between two neighboring objects.

For example, a polygonal line which represents the outline of a truck may include a locally concave angle which materializes the break between the cab and the trailer of the truck, while the cab and the trailer belong to a single object.

Also, it is common to note that certain points of the polygonal line are parasitic points which form noise and which mix with the so-called useful points of the polygonal line representing the target object.

These parasitic points can form concave angles locally which can lead to a misinterpretation of the geometry of the object represented by the polygonal line.

The present invention aims in particular to resolve the drawback of the aforementioned prior art by proposing a method which makes it possible to detect a global concavity on a polygonal line while ignoring local concavities and noise.

• une première étape de détection de coins, qui exécute un algorithme de simplification de la ligne polygonale visant à limiter le nombre de points appartenant à la ligne polygonale, les points conservés étant considérés comme des coins, un coin étant considéré concave si l’angle qu’il forme est rentrant, du point de vue dudit capteur d’acquisition d’images,
• une deuxième étape de détection du bruit qui exécute un algorithme de détection du bruit consistant à identifier les coins de la ligne polygonale considérés comme étant du bruit, parmi les coins identifiés au cours de la première étape, et
• une troisième étape d’identification d’une concavité globale, qui consiste à calculer successivement l’aire de chacun des triangles délimités par un coin identifié concave, le coin précédent et le coin suivant de la ligne polygonale en faisant abstraction des coins identifiés comme étant du bruit, si l’aire calculée est supérieure à une aire de référence, alors ledit coin identifié concave est considéré suffisamment concave.

This objective, as well as others which will appear on reading the description which follows, is achieved with a method of detecting a global concavity on a polygonal line, said polygonal line being formed by a series of successive points which are connected two by two by a straight line segment and which come from an image captured by an image acquisition sensor, characterized in that it comprises at least:

• a first corner detection step, which executes a polygonal line simplification algorithm aimed at limiting the number of points belonging to the polygonal line, the retained points being considered as corners, a corner being considered concave if the angle that 'it form is re-entrant, from the point of view of said image acquisition sensor,
• a second noise detection step which executes a noise detection algorithm consisting of identifying the corners of the polygonal line considered to be noise, among the corners identified during the first step, and
• a third step of identifying a global concavity, which consists of successively calculating the area of each of the triangles delimited by a corner identified as concave, the previous corner and the following corner of the polygonal line, ignoring the corners identified as being noise, if the calculated area is greater than a reference area, then said concave identified corner is considered sufficiently concave.

Thus, the method according to the invention makes it possible to detect a sufficiently concave corner which is representative of a break between two objects drawn by the polygonal line.

The reference area can be determined according to the degree of concavity that we wish to detect with the present method.

According to other optional characteristics of the invention, taken alone or in combination:

• sélectionner le premier point, formant une première borne, et le dernier point, formant une deuxième borne, de la ligne polygonale,
• déterminer le point distal qui est le point de la ligne polygonale le plus éloigné d’un premier segment de droite de base qui relie la première borne et la deuxième borne,
• calculer la distance entre le point distal et ledit premier segment de droite de base, si ladite distance calculée est supérieure à une distance de référence, alors le point distal est considéré comme étant un coin,

- the simplification algorithm executed during the first step consists of:

• select the first point, forming a first terminal, and the last point, forming a second terminal, of the polygonal line,
• determine the distal point which is the point of the polygonal line furthest from a first basic straight line segment which connects the first terminal and the second terminal,
• calculate the distance between the distal point and said first base line segment, if said calculated distance is greater than a reference distance, then the distal point is considered to be a corner,

and in that the simplification algorithm is called recursively on sub-parts of the polygonal line, if a corner is found. The reference distance can be determined according to the degree of simplification of the polygonal line that we wish to obtain at the end of the first simplification step;

• on mesure la longueur d’un deuxième segment de droite de base reliant le premier point et le dernier point de la ligne polygonale,
• on mesure successivement la longueur d’un troisième segment de droite de base reliant le premier coin et le dernier coin de chacun des triangles qui sont délimités par trois coins successifs de la ligne polygonale,
• si la longueur du deuxième segment de droite de base associé est supérieure à une première longueur de référence prédéterminée, et si la longueur du troisième segment de droite de base est inférieure à une seconde longueur de référence prédéterminée, alors lesdits trois coins successifs associés audit troisième segment de droite de base sont considérés comme étant du bruit. La première longueur de référence et la seconde longueur de référence sont déterminées selon le degré de réduction du bruit souhaité ;

- during the second stage:

• we measure the length of a second base line segment connecting the first point and the last point of the polygonal line,
• we successively measure the length of a third base line segment connecting the first corner and the last corner of each of the triangles which are delimited by three successive corners of the polygonal line,
• if the length of the second associated base line segment is greater than a first predetermined reference length, and if the length of the third base line segment is less than a second predetermined reference length, then said three successive corners associated with said third basic line segment are considered noise. The first reference length and the second reference length are determined according to the desired degree of noise reduction;

- the first step is executed again following the second step, ignoring the corners identified as noise during the second step;

- said reference area used during the third step of identifying a global concavity is determined as a function of the length of a fourth basic straight line segment which connects the first corner and the last corner of the polygonal line;

• une première valeur si la longueur du quatrième segment de droite de base qui relie le premier coin et le dernier coin de la ligne polygonale est supérieure à 8 mètres,
• une deuxième valeur si la longueur dudit quatrième segment de droite de base est comprise entre 8 mètres et 5,5 mètres,
• une troisième valeur si la longueur dudit quatrième segment de droite de base est comprise entre 5,5 mètres et 2,75 mètres,
• une quatrième valeur si la longueur dudit quatrième segment de droite de base est comprise entre 2,75 mètres et 1,5 mètre, et
• une cinquième valeur si la longueur dudit quatrième segment de droite de base est inférieure à 1,5 mètre ;

- said reference area used during the third step of identifying a global concavity is equal to:

• a first value if the length of the fourth basic straight line segment which connects the first corner and the last corner of the polygonal line is greater than 8 meters,
• a second value if the length of said fourth basic straight line segment is between 8 meters and 5.5 meters,
• a third value if the length of said fourth basic straight line segment is between 5.5 meters and 2.75 meters,
• a fourth value if the length of said fourth base line segment is between 2.75 meters and 1.5 meters, and
• a fifth value if the length of said fourth basic straight line segment is less than 1.5 meters;

- said first value of the reference area is equal to 1.5 square meters, said second value of the reference area is equal to 0.75 square meters, said third value of the reference area is equal to 0 .56 square meter, said fourth value of the reference area is equal to 0.4 square meter, said fifth value of the reference area is equal to 0.3 square meter;

The invention also relates to a computer for a vehicle, in particular an automobile, configured to implement the method described above.

Finally, the invention also relates to a vehicle, in particular an automobile, comprising a computer described above and at least one image acquisition sensor.

Other characteristics and advantages of the invention will emerge on reading the description which follows, with reference to the appended figures, which illustrate:

a schematic view of a motor vehicle equipped with a computer which implements the method according to the invention, and a polygonal line connecting several points and drawing a truck and a safety guardrail;

a schematic view of the polygonal line of the comprising a distal point;

a schematic view of the polygonal line of the comprising a plurality of corners detected during the first step of the method;

a schematic view of the polygonal line of the relieved of three corners considered as noise during the second stage of the process;

a flowchart representing the steps of the process according to the invention.

We have schematically represented in the a truck 10 forming a first object to be detected, a safety guardrail 12 forming a second object, and a motor vehicle 14 equipped with an image acquisition sensor 16 and a computer 17.

The image acquisition sensor 16 is for example a remote laser measuring device known by the English acronym LIDAR for “Light Detection And Ranging”.

Without limitation, the image acquisition sensor 16 can be formed by any sensor adapted to generate a cloud of points and to represent objects by a polygonal line.

The truck 10 includes a cockpit 18 and a trailer 20.

In reference to the , a polygonal line L1 draws a portion of the outline of the truck 10 and a portion of the guardrail 12, seen by the image acquisition sensor 16.

For this purpose, the image acquisition sensor 16 captures three-dimensional points in space which are transformed into image points in a two-dimensional reference frame R(Z, , these image points making it possible to form the polygonal line L1.

Thus, the polygonal line L1 includes a series of consecutive points, referenced P1 to P18 at the .

Also, the polygonal line L1 includes a series of consecutive line segments.

We speak of a segment to designate a straight line which connects two successive points of the polygonal line L1, and we speak of successive, or adjacent, segments to designate two segments which have an end point in common and which do not have other point in common.

We speak of an angle to designate an angular sector delimited by two successive segments.

An angle can be concave, that is to say entering towards the interior of the object delimited by the polygonal line L1, such as the angle delimited by the points P5, P6, P7.

Conversely, an angle can be convex, that is to say extending outwards from the object delimited by the polygonal line L1, such as the angle delimited by the points P12, P13, P14.

The invention relates to a method for detecting a global concavity on a polygonal line, such as the polygonal line L1, for the purpose of differentiating adjacent objects.

Indeed, it is expected that an object, in particular an object on a road, is represented by a polygonal line which has convex portions, that is to say portions which have outgoing angles from the point of view of the sensor. image acquisition 16, such as the shape of the truck 10 illustrated in .

Conversely, it is expected that a break, or a separation, between two objects is represented by a portion of the polygonal line which has a concave shape, that is to say a portion which has a re-entrant angle.

We speak of “global” concavity to designate a portion of a polygonal line which is sufficiently concave to represent a break between two objects.

For example, the polygonal line L1 has an overall concavity between the points P5 and P7, a concavity which represents a break between the truck 10 and the safety guardrail 12.

Conversely, a “local” concavity is not interpreted as being a break between two objects.

A local concavity is interpreted as coming from a parasitic point, or as representing a local recess on the contour of one, and only one, object represented by the polygonal line L1.

For example, with reference to the , the angle formed by the points P15 to P17 is a locally concave angle which materializes the break between the cabin 18 and the trailer 20 of the truck 10 drawn by the polygonal line L1, the cabin 18 and the trailer 20 belonging to the same object .

We will understand that the difference between a local concavity and a global concavity is a difference in scale and interpretation.

The method according to the invention aims to distinguish a local concavity from a global concavity.

For this purpose, the method according to the invention successively comprises a first step E1 of detecting corners on the polygonal line L1, a second step E2 of detecting noise and a third step E3 of identifying an overall concavity, illustrated in there .

The first corner detection step E1 executes a simplification algorithm of the polygonal line L1 aimed at limiting the number of points belonging to the polygonal line L1, the retained points being considered as corners.

A corner is an angle that can be concave or convex.

According to a preferred embodiment of the invention, the simplification algorithm is based on an algorithm known as “Ramer-Douglas-Peucker”.

More particularly, the simplification algorithm executed during the first step E1 consists of selecting the first point P1 of the polygonal line L1, forming a first terminal A1, and the last point P18 of the polygonal line L1, forming a second terminal A2.

As we can see in the , the first terminal A1 and the second terminal A2 of the polygonal line L1 are connected by a first base straight segment Db1.

Then, still during the first step E1, we determine the distal point Pdist which is the point of the current polygonal line L1 farthest from the first base line segment Db1, in a direction perpendicular to the first base line segment Db1.

After determining the distal point Pdist, represented by the point P13 according to the example illustrated in , we calculate the distance Dmax between the distal point Pdist and the first base line segment Db1, this distance Dmax corresponding to the height of the triangle having as its base the first base line segment Db1.

If the calculated distance Dmax is greater than a predetermined reference distance, then the distal point Pdist is considered to be a corner.

Conversely, if the distance Dmax is less than the predetermined reference distance, then the points of the polygonal line L1 remain considered as points.

If a corner is detected during the first step E1, then the simplification algorithm is called recursively on a first sub-part and on a second sub-part of the current polygonal line, until the simplification algorithm simplification no longer detects corners.

By current polygonal line we mean the polygonal line, or subpart of a polygonal line, which is concerned by the simplification algorithm at a defined instant, since the polygonal line concerned by the simplification algorithm evolves with the recursion of the first step E1.

It will be understood that the value chosen for the reference distance makes it possible to more or less simplify and “smooth” the polygonal line L1.

According to the example illustrated in , we consider that the distance Dmax is greater than the reference distance, so that the simplification algorithm is called again on a first sub-part of the polygonal line L1 between the first terminal A1 and the distal point Pdist represented by the point P13 which forms a new second terminal, and on a second sub-part between the distal point Pdist represented by the point P13 which forms a new first terminal, and the second terminal A2.

Calculating the distance Dmax amounts to successively calculating the area Δ of each of the triangles formed by the first base line segment Db1 and each of the points of the polygonal line L1, the triangle which has the largest area determining the distal point Pdist.

The area Δ of each triangle can be calculated by the following equation:

with A1 the first terminal of the current polygonal line L1, A2 the second terminal of the current polygonal line L1, x the coordinate along the axis current of one of the n points of the current polygonal line L1.

We represented at the the corners detected during the first step E1, the corners being represented by square pellets, while the points are represented by round pellets.

Note that a corner is considered concave if the angle it forms is re-entrant, from the point of view of the image acquisition sensor 16.

Also, a corner is considered concave if the area Δ of the triangle composed of the current point and the first base line segment Db1 is positive.

Following the first simplification step E1, the method according to the invention executes a second noise detection step E2.

The second step E2 executes a noise detection algorithm consisting of identifying the corners of the polygonal line L1 considered to be noise, among the corners identified during the first step E1.

Note that the corners identified as noise are ignored by the following steps of the process.

During the second noise detection step E2, the length of a second base line segment Db2 connecting the first point and the last point of the polygonal line L1 is measured.

According to the example of embodiment of the invention illustrated in , the second base line segment Db2 joins point P1 and point P18.

Also, we successively measure the length of a third base line segment Db3 connecting the first corner and the last corner of each of the triangles which are delimited by three successive corners of the polygonal line L1.

In other words, during the second step E2, the corners of the polygonal line L1 are covered in groups of three, each group forming a triangle comprising the third base Db3. For the sake of clarity, , only the bases Db3 connecting the corners C10 and C12, on the one hand, and C15 and C17, on the other hand.

If the length of the second base line segment Db2 is greater than a first predetermined reference length, and if the length of the current third base line segment Db3 is less than a second predetermined reference length, then the three successive associated corners to the third current base line segment Db3 are identified as noise.

During the second step E2, all the corners are covered and then each is identified as being noise, or not.

According to the example illustrated in , the three corners C10, C11, C12 which correspond to the three points P10, P11, P12 are considered as noise because the length of the second base line segment Db2 is greater than the first predetermined reference length and the length of the third segment base right Db3 associated is less than the second reference length.

Conversely, the three corners C15, C16, C17 which correspond to the three points P15, P16, P17 are not considered as noise because the length of the third associated base line segment Db3 is greater than the second reference length .

The first reference length and the second reference length are defined in relation to the dimensions of the vehicles likely to be crossed on a road.

According to a preferred embodiment of the invention, the second step E2 of the method aims to apply only to large vehicles greater than 2.75 meters long.

Also, the second reference length can be set at 0.4 meters, for example.

It will be noted that the choice of the first reference length and the second reference length makes it possible to filter more or less the noise present on the polygonal line L1.

Following the second noise detection step E2, the first corner detection step E1 is executed again, then the second occurrence of the first step E1 is followed by the third step E3.

However, without limitation, it is possible to repeat the second noise detection step E2 again following the second occurrence of the first step E1.

The third step E3 of identifying a global concavity is executed by ignoring the corners identified as being noise during the second step E2.

The third step E3 consists of successively calculating the area of each of the triangles delimited by three corners of the polygonal line L1, each of the triangles comprising a first corner identified as concave during the first step E1, a second corner which directly precedes said first corner, and a third corner which directly succeeds said first corner.

If the calculated area is greater than a reference area, then said first concave identified corner which is associated with the calculated area is considered sufficiently concave.

The reference area used during the third step E3 is determined according to the length of a fourth basic straight line segment Db4 which connects the first corner and the last corner of the polygonal line L1.

According to the example of embodiment of the invention illustrated in , the fourth base line segment Db4 joins point P1 and point P18.

According to a preferred embodiment, the reference area used during the third step E3 can take five different values.

A first value if the length of the fourth base line segment Db4 is greater than 8 meters, a second value if the length of the fourth base line segment Db4 is between 8 meters and 5.5 meters, a third value if the length of the fourth base line segment Db4 is between 5.5 meters and 2.75 meters, a fourth value if the length of the fourth base line segment Db4 is between 2.75 meters and 1.5 meters, and a fifth value if the length of the fourth base line segment Db4 is less than 1.5 meters.

So the first value roughly corresponds to the length of a large truck, the second value corresponds to the length of a truck, the third value corresponds to the length of a car or van, the fourth value corresponds to the length of a motorcycle and the fifth value is the length of an object smaller than a motorcycle.

Also, the first value of the reference area is equal to 1.5 square meters, the second value of the reference area is equal to 0.75 square meters, the third value of the reference area is equal to 0.56 square meters, the fourth reference area value is equal to 0.4 square meters, the fifth reference area value is equal to 0.3 square meters.

According to the example of embodiment of the invention illustrated in , we consider that the polygonal line L1 includes two concave corners, namely corner C6 and corner C16.

We also consider that the area of the triangle delimited by the concave corner C6, the directly preceding corner C1 and the directly following corner C7, is greater than the reference area defined as a function of the length of the fourth base line segment Db4 .

Consequently, corner C6 is considered sufficiently concave and therefore as marking a break between two objects, here between truck 10 and guardrail 12.

Conversely, we consider that the area of the triangle delimited by the concave corner C16, the directly preceding corner C15 and the directly following corner C17 is less than the reference area defined according to the length of the fourth line segment basic Db4.

Therefore, corner C16 is not considered sufficiently concave, it does not mark a break between two distinct objects.

Note that the larger the fourth base line segment Db4, the greater the area of the triangle including a concave corner must be for said corner to be considered sufficiently concave.

According to another aspect of the invention, it will be noted that the computer 17 which equips the automobile vehicle 14 is configured to implement the method according to the invention.

Naturally, the invention is described in the above by way of example. It is understood that those skilled in the art are able to carry out different variants of the invention without departing from the scope of the invention.

Claims (8)

Method for detecting a global concavity on a polygonal line (L1), said polygonal line (L1) being formed by a series of successive points which are connected in pairs by a straight line segment and which come from a captured image by an image acquisition sensor (16), characterized in that it comprises at least:

• a first corner detection step (E1), which executes a polygonal line simplification algorithm (L1) aimed at limiting the number of points belonging to the polygonal line (L1), the retained points being considered as corners, a corner being considered concave if the angle it forms is reentrant, from the point of view of said image acquisition sensor (16),
• a second noise detection step (E2) which executes a noise detection algorithm consisting of identifying the corners of the polygonal line (L1) considered to be noise, among the corners identified during the first step (E1), And
• a third step (E3) of identifying a global concavity, which consists of successively calculating the area of each of the triangles delimited by a corner identified as concave, the previous corner and the following corner of the polygonal line (L1) by making abstraction of the corners identified as being noise, if the calculated area is greater than a reference area, then said concave identified corner is considered sufficiently concave.

Method according to claim 1, characterized in that the simplification algorithm executed during the first step (E1), consists of:

• select the first point, forming a first terminal, and the last point, forming a second terminal, of the polygonal line (L1),
• determine the distal point (Pdist) which is the point of the polygonal line (L1) furthest from a first basic straight line segment (Db1) which connects the first terminal and the second terminal,
• calculate the distance between the distal point (Pdist) and said first base line segment (Db1), if said calculated distance is greater than a reference distance, then the distal point (Pdist) is considered to be a corner,

and in that the simplification algorithm is called recursively on sub-parts of the polygonal line (L1), if a corner is found. Method according to any one of the preceding claims, characterized in that during the second step (E2):

• we measure the length of a second basic straight line segment (Db2) connecting the first point and the last point of the polygonal line (L1),
• we successively measure the length of a third base line segment (Db3) connecting the first corner and the last corner of each of the triangles which are delimited by three successive corners of the polygonal line (L1),
• if the length of the associated second base line segment (Db2) is greater than a first predetermined reference length, and if the length of the third base line segment (Db3) is less than a second predetermined reference length, then said three successive corners associated with said third base line segment (Db3) are considered to be noise.

Method according to any one of the preceding claims, characterized in that the first step (E1) is executed again following the second step (E2) ignoring the corners identified as noise during the second step (E2). Method according to any one of the preceding claims, characterized in that said reference area used during the third step (E3) of identifying a global concavity is determined as a function of the length of a fourth line segment base (Db4) which connects the first corner and the last corner of the polygonal line (L1). Method according to claim 5, characterized in that said reference area used during the third step (E3) of identifying a global concavity is equal to:

• a first value if the length of the fourth basic straight line segment (Db4) which connects the first corner and the last corner of the polygonal line (L1) is greater than 8 meters,
• a second value if the length of said fourth basic straight line segment (Db4) is between 8 meters and 5.5 meters,
• a third value if the length of said fourth basic straight line segment (Db4) is between 5.5 meters and 2.75 meters,
• a fourth value if the length of said fourth basic line segment (Db4) is between 2.75 meters and 1.5 meters, and
• a fifth value if the length of said fourth basic straight line segment (Db4) is less than 1.5 meters.

said first value of the reference area is equal to 1.5 square meters, said second value of the reference area is equal to 0.75 square meters, said third value of the reference area is equal to 0, 56 square meter, said fourth value of the reference area is equal to 0.4 square meter, said fifth value of the reference area is equal to 0.3 square meter. Computer (17) for an automobile vehicle (14), configured to implement the method according to any one of the preceding claims. Motor vehicle (14) comprising a computer (17) according to claim 7 and at least one image acquisition sensor (16).