EP1038189A1

EP1038189A1 - Method for predicting the existence of a curve in a road portion

Info

Publication number: EP1038189A1
Application number: EP98962797A
Authority: EP
Inventors: Magnus Kamel; Maria ÖGREN
Original assignee: CelsiusTech Electronics AB
Current assignee: CelsiusTech Electronics AB
Priority date: 1997-12-10
Filing date: 1998-12-10
Publication date: 2000-09-27
Also published as: AU1795499A; WO1999030183A1; SE511013C2; JP2001526397A; SE9704612D0; SE9704612L

Abstract

The present invention relates to a method and a device for predicting the existence of a curve in a road portion in front of a vehicle in connection with the utilisation of a radar arranged on the vehicle which, by transmitting a radar signal and recording received reflections thereof, provides information indicating the position of objects located in an area in front of the vehicle. According to the invention, values are derived which represent directions of hypothetical lines between said objects and from said directions are derived the corresponding values each of which defines the position of a respective one of said lines between said objects. Subsequently, it is determined whether at least a subset of said values of directions and corresponding positions fulfils a functional relationship between direction and position, whereby the existence of said curve is predicted on the basis thereof.

Description

METHOD FOR PREDICTING THE EXISTENCE OF A CURVE IN A ROAD PORTION

Field of the Invention

The present invention relates to a method for predicting the existence of a curve in a road portion in front of a vehicle in connection with the utilisation of a radar arranged on the vehicle which, by transmitting a radar signal and recording received reflections thereof, provides information indicating the position of objects located in an area in front of the vehicle. The invention also relates to a device for carrying out the method.

Background of the Invention

Vehicle-mounted radar systems of the type described above, for example for cars, buses, lorries and the like, are currently being developed. The purpose of such sys- terns is to assist the driver of the vehicle by providing functions such as primarily cruise control and collision warning, but in future also automatic breaking or other types of avoidance manoeuvring in the case of an imminent collision hazard, based on the detection of vehicles, stationary obstacles, and similar objects in front of the vehicle .

In such radar systems, it is often desirable, on the basis of the detected objects located in an area in front of the vehicle, to provide a prediction of the direction of the path in front of the vehicle, i.e. to determine whether the road is most likely to be straight or curved, and in the latter case the magnitude of an existing curve. In addition to cars travelling in front of the vehicle, such objects may include road signs, guard rails and the like which in this connection are used to predict the direction of the path.

One method for determining the curvature of the road is known from the European patent specification EP 464 821, in which a functional approximation is uti- lised so that a hypothetical functional relationship describing the hypothetical curvature of a road can be made to correspond to the greatest possible extent with the position of the objects appearing in the image.

Objects of the Invention

A problem associated with, for example, the above- mentioned prior art type of approximation is that reflections originating from objects which do not define the direction of the path will influence the approximation of the functional relationship, i.e. every object which does not delimit or define the path contributes to disturbing the appearance of the function and thus causes a false prediction of the direction of the path. In addition, if it can be assumed that the objects originate from several lines or curves, approximation of several functions is required. In this connection, it will be appreciated that the processor capacity required in the system increases with the number of lines that it must be capable of determining.

Consequently, one object of the invention is to reduce or obviate the problems associated with the influence of irrelevant objects when predicting the direction of the path in front of the vehicle. A further object of the invention is to simplify processing when several lines or curves can be expected to be found when predicting the direction of the path in front of the vehicle.

Yet another object of the present invention is to provide a solution which reduces the calculation capacity required in a radar system of the kind mentioned above.

Summary of the Invention

The above-mentioned objects are achieved by means of the invention as defined in the appended claims.

According to a first aspect of the invention, there is provided a method of the kind mentioned by way of introduction, characterised by the steps of: deriving values representing directions of hypothetical lines between said objects and to said directions associating derived corresponding values each of which defines the position of a respective one of said lines between said objects; and determining whether at least a subset of said values of directions and corresponding positions fulfils a functional relationship between direction and position and predicting the existence of said curvature on the basis thereof.

Accordingly, rather than starting from the actual position of the objects, the invention is based on the idea of starting from directions indicated by the objects, more specifically directions of hypothetical lines between the objects, and of basing the prediction of the direction of the path upon these directions. Since the directions relate to hypothetical lines between hypothetical objects, no prior analysis is required in order to determine these directions. Rather, the directions may be derived directly from the information regarding the positions of each combination of two objects, which saves processor capacity.

The derivation of directions of hypothetical lines between the objects in the radar image is similar to a kind of spatial differentiation of the positional information. In principle, this means that the information about the horizontal position of the objects is differentiated away. However, this has the advantage that directional values of lines between objects located at the same distance from the vehicle but in different positions laterally, e.g. relating to reflections from guard rails on both sides of a road, will reinforce each other in the subsequent analysis. When the known curvature approximation based directly on the positions of the objects is used, separate curvature approximations must be carried out for separate potential curvatures in the image. According to the invention, curves on both sides of the road will together provide directions which jointly indicate the direction of the path without requiring separate processing at least initially.

According to a preferred embodiment of the inven- tion, the Hough transform is utilised for determining, on the basis of the number pairs obtained from said values of directions and the corresponding positions, whether a functional relationship between direction and position is fulfilled by at least a subset of said number pairs, and, if so, for determining the parameters of the probable appearance of this functional relationship. In this connection, it should be noted that the utilisation of the Hough transform per se in connection with image analysis for detecting lines therein is well known to the person skilled in image processing. It should be noted that the Hough transformation according to the invention is carried out on the basis of a direction/position space and not on the basis of the original spatial space.

If, for example, the path in front of the vehicle can be likened to an arc, the relationship between direction and distance will be essentially linear. In this case, the Hough transform is used for finding such a linear relationship between direction and distance for at least a subset of the number pairs obtained from the above-mentioned values of directions and corresponding positions .

More specifically, said Hough transform preferably comprises the steps of transforming number pairs, obtained from said values of directions and corresponding posi- tions, into respective curves in a parameter space, whose dimensions correspond to the parameters included in said functional relationship of direction as a function of position for a hypothetical range, the parameter combinations obtained from said curves indicating possible appearances of said functional relationship; and determining one or more intersecting points between the curves in said parameter space and, on the basis of these, work- ing out parameter combinations indicating the specific appearances of said functional relationship, corresponding to specific appearances of curvatures present in the road portion. It will be appreciated that an intersecting point in the parameter space defines the parameters of the functional relationship between direction and position, and that, in turn, the relationship between direction and position defines a curvature of the road. Since each number pair gives rise to a separate curve in the parameter space, it will be appreciated that a set of number pairs will give rise to a set of curves in the parameter space, and consequently a plurality of intersecting points will exist between these curves. How- ever, this does not mean that all intersecting points correspond to relevant functional relationships. Only in cases where a plurality of intersecting points are concentrated in a small area in the parameter space does this lead to the conclusion that there exists a func- tional relationship common to a plurality of the number pairs, and only then is a path prediction deemed to exist. Naturally, many false intersecting points will appear, but there is little probability that they will combine to create a significant false maximum. It is preferred that the Hough transformation be based on rectilinear relationships expressed as first degree polynomials and that, consequently, said parameter space be two-dimensional. However, the transform can also be used with respect to multidimensional or non-linear relationships. For example, the above-mentioned curves in the parameter space may be straight lines as well as sinusoidal relationships or higher order polynomials.

The analysis of the curves in said parameter space produced by means of the Hough transform is advantageous- ly carried out by incrementing or otherwise changing the element values in a matrix having the same dimensions as the parameter space. In order to determine relevant intersecting points in the parameter space when it is represented by a matrix, an initial mean value formation, the purpose of which is to even out the amplitudes of said matrix, is advantageously employed, followed by a determination of local maxima of the matrix. By choosing a suitable resolution of the matrix, i.e. the number of elements in the matrix, of the representation of the width of the curves (calculated as the number of matrix elements) , and of the size of the window utilised in the mean value formation, one ensures that relevant local maxima are derived from the matrix.

At first, all elements of the matrix are set to zero. Subsequently, the elements of the matrix which correspond to curves produced by means of the Hough transform of the respective number pairs are updated. This means that each number pair provides an incrementation of all elements on a surface of a dimension lower than the dimension of the matrix, i.e. a curve in the case of a two-dimensional matrix. A major advantage of utilising the Hough transform compared to, for example, least squares approximation is that number pairs which do not represent a curvature of the road do not influence the prediction of the direction of the path and it is not necessary to determine how many curves one is looking for until after the actual transformation to the parameter space has been achieved, or to determine which number pairs belong to which curve.

Reference is made to the book "Digital Image Processing" by Rafael C. Gonzales and Richard E. Woods, published by the Addison-Wesley Publishing Company, for a more detailed description of Hough transformation in image processing.

Although using the Hough transform is the preferred alternative for analysing the directions and positions derived according to the invention, it will be appreciated that other methods are applicable for this purpose. According to an alternative embodiment, least squares approximation of a functional relationship between direction and position for a hypothetical road portion to the number pairs obtained from said values of directions and corresponding positions is utilised, after which the pro- bable appearance of a curvature present in the road portion is determined on the basis of said least squares approximated functional relationship.

Advantageously, in addition to determining directions of lines between objects in the radar image, direc- tions of moving objects are also derived by deriving, with respect to an object which on the basis of radar information gathered on two separate occasions is determined to be a moving object, a value of a direction of the object in the form of the direction of motion of the object, and a corresponding value representing the position of the moving object. The number pairs used in the subsequent analysis, preferably carried out using the Hough transform, can thus relate to lines between objects in the image as well as directions of motion of moving objects. It will be appreciated that the direction of motion of the vehicle in question can also be used as a basis for a number pair corresponding thereto.

To the extent that moving objects are treated differently than static objects, the derivation of direc- tions of lines between different objects is preferably limited to lines between static objects, since it is unlikely that a line between a static object and a moving object represents the direction of the path.

Moreover, the analysis according to the invention is preferably limited to include only those derived directional values which deviate less than a predetermined limit value from the direction of travel of the vehicle. This is based on the assumption that directions extending more or less transversely of the direction of travel of the vehicle are unlikely to represent the road upon which the vehicle is travelling. Consequently, the analysis will be both quicker and simpler if such obviously false directional values are simply ignored.

In addition, the analysis according to the invention is preferably limited to include only those directional values which deviate less than a predetermined limit value from a previously predicted direction of the path in the corresponding position. This is based on the insight that the direction of the road in a portion far away from the vehicle in question can be significantly different from the direction of travel of the vehicle, and that a classification of the relevance of an obtained directional value should preferably be related to a predicted direction of the path in the position in question, to the extent that such a prediction has been achieved in a previous phase.

Since, for example, the direction of motion of a vehicle travelling in front of the vehicle in question most likely corresponds to the direction of the path, while the direction of an individual line between two optional static objects does not have an obvious connection to the direction of the path, it is possible to achieve a more reliable analysis by weighting the corresponding number pairs differently depending on a classification of the object or objects from which said number pair derives. For example, a number pair corresponding to the position and direction of motion of a moving object is preferably more heavily weighted than a number pair corresponding to the direction and the position of an individual line between two objects. Yet another alter- native is to weight number pairs of moving objects differently depending on the speed of the objects.

In order further to simplify the processing, it is sufficient to form lines between certain objects only rather than between all objects found. For example, it is less likely that a line between two objects located far from each other actually corresponds to the direction of the path. Since objects which delimit the road, such as guard rails, lamp posts and the like, usually appear at relatively regular and short intervals, it is more likely that a line between objects located such a short distance from each other actually represent the direction of the path. Accordingly, in a preferred embodiment, only lines between objects located less than a predetermined distance, e.g. 10 metres, from each other are processed.

In connection with continuous processing of radar information which is recorded at successive points in time, it is suitable to temporarily save values from previous processing occasions, for example by saving the element values in said matrix. On subsequent processing occasions, it is then sufficient to derive new directional and positional values with respect to lines to/ from new objects in relation to the previous information, whereupon the new values are used for updating the matrix. When utilising such updating, it is important to take into consideration the altered distance resulting from the movement of the vehicle in question. The processing of the above-mentioned derived values may advantageously be divided into partial processing operations with respect to different subareas of the area in front of the vehicle. Such subareas can be defined either in relation to the vehicle or in relation to the environment. The choice of definition of the subareas will thus affect the choice of how previously derived number pairs or calculated matrices are saved and updated.

Each direction obtained, represented by the respec- tive line or direction of motion, shall be taken into consideration with respect to the position for which the direction can be considered to be relevant. Consequently, according to the invention, a corresponding positional value is derived for each directional value obtained. These positions may be expressed in several different ways. The preferred alternative is for said positions to be expressed as the distance thereto from the vehicle in question. According to another alternative, said positions are expressed as distances thereto from a chosen reference point, which may be defined either in relation to the vehicle or in relation to the environment. For example, one of the objects may be chosen as a reference point .

Said positional values may, for example, be expressed as the straight distance between the vehicle and the line/object, as the distance between the vehicle and the line/object projected onto an axis extending along the direction of travel of the vehicle in question, or as the arc distance to the line/object taking into account the angle thereto, seen from the vehicle in question. If the processing is performed in subareas, as discussed above, positional information may also be obtained from the knowledge of which subarea is relevant to a specific line/object. Similarly, directions are preferably indicated as spatial directions or angles in a system of coordinates which is defined with respect to the vehicle, but other ways of defining direction may, of course, also be used, for example in relation to a ground-based system of coordinates. Moreover, it will be appreciated that all directional values can preferably be related in some way to the direction of travel of the vehicle. When determining, for example, the distance to a direction of a line, the distance to a point half-way between two objects from which the line derives is advantageously chosen, since it is likely that the direction of the line best corresponds to the direction of the path at a distance half-way between the objects. However, it should be noted that the invention is, of course, not restricted to this choice and that other bases for determining the distance between the vehicle in question and the line between two objects can be used. For example, the distance from the vehicle to the closer of the two objects can be chosen. According to further embodiments of the invention, the initial derivation of number pairs is repeated with respect to the actual number pairs themselves. By viewing the number pairs as points in a two-dimensional image, the directional analysis can be carried out yet another time with respect to lines between these points (number pairs) prior to carrying out the actual determination of a functional relationship. In this case, the result can be likened to a second derivative of the original posi- tional information.

It will be appreciated that the invention is advantageously implemented using conventional microprocessor technology and that, according to a second aspect, the invention thus relates to a device, such as a micropro- cessor, for carrying out the steps and measures recited in the appended claims and discussed above.

Further features and advantages of the invention will appear from the following description of preferred embodiments thereof as well as from the appended claims.

Brief Description of the Drawings

Embodiments of the invention will now be described by way of example with reference to the accompanying drawings, in which: Fig. 1 schematically shows a representation of the positions of objects in relation to a vehicle;

Fig. 2 schematically shows a geometric division of a road into straight lines and curves of a predefined kind; Figs 3a and 3b schematically show the first and the second derivative of the direction of the road as a function of the range in Fig. 2;

Fig. 4 schematically shows choices of lines and distances with respect to the objects in Fig. 1 in accordance with an embodiment of the invention; Fig. 5 schematically shows a diagram of directions as a function of distance from the radar image in Fig. 4; Figs 6a and 6b show an example of the utilisation of the Hough transform to detect lines in an x-y plane;

Figs 7a and 8a schematically show examples of the utilisation of the Hough transform to detect lines in the diagram shown in Fig. 5; and

Fig. 8 schematically shows the division of the processing of the objects in Fig. 1 into different areas according to an embodiment of the invention.

Detailed Description of Preferred Embodiments

For the sake of simplicity, Fig. 1 shows a schematic representation of the positions of objects detected by a vehicle-mounted radar in an area located in front of the vehicle. The radar provides successively renewed posi- tional information, e.g. ten times per second, by transmitting a scanning radar signal and recording received reflections thereof from objects located in the area in front of the vehicle. Although the description with reference to the Figures herein is based on the application of the invention to a given amount of information, the invention can in fact advantageously be used with respect to information obtained successively, which when put together provides better decision support data.

In Fig. 1, the radar is mounted on the vehicle in question, which by way of illustration is schematically shown at 20 in the Figure in order to indicate, by way of illustration, the location of the vehicle 20 in relation to the other objects.

As can be seen in Fig. 1, a number of objects 31, 32, 33 are located in front of the vehicle. One of the objects is a vehicle 31 travelling in a direction opposite to that of the vehicle in question. Another object is a vehicle 32 which is travelling in the same direction as the vehicle in question and which is located in a curve in the road. The other objects 33 are static objects in the environment, such as trees, lamp posts, guard rails, road signs, and the like. By comparing, according to the prior art, information from successively effected scans, for example, or by analysing the frequency content of the individual reflections, moving objects are distinguished from static ones. As can be seen from Fig. 1, it is possible on the basis of the location of the static objects 33 as well as the location the moving objects 31, 32 and the direction of motion to predict the direction of the path in front of the vehicle, schematically shown by dashed lines 40 in the Figure.

Figs 2, 3, and 3b illustrate a geometric division of a road into straight lines and curves of a predefined kind. Most roads, particularly those made for speeds over 60 km/h, consist of segments or portions which with a high degree of accuracy, can be approximated by straight lines, circles, and clothoids. (It should be noted that a straight line can actually be viewed as a circle with an infinite radius of curvature.)

In the case of a straight line, the direction of the line is constant (as in road portions I and V in Fig. 2) . In this case, the derivative of the direction Dir of the path, with respect to the distance travelled (Range) , is zero and the second derivative of the direction of the path, with respect to the distance travelled, is also zero, as shown at I and V in Figs 3a and 3b.

In the case of a circle, the curvature of the line is constant (as in road portion III in Fig. 2) . In this case, the derivative of the direction Dir of the path with respect to the distance travelled (Range) is con- stant, and the second derivative of the direction of the path with respect to the distance travelled is consequently zero, as shown at III in Figs 3a and 3b.

A clothoid is defined by the fact that the radius of curvature R of the line is proportional (or alternatively inversely proportional) to the arc length. Consequently, the derivative of the direction of the path as a function of the distance travelled is proportional to the distance travelled, and the second derivative of the direction of the path with respect to the distance travelled is constant, as shown at II and IV in Figs 3a and 3b.

It will be appreciated that by looking for straight lines in a diagram of the type shown in Figs 3a or 3b, it is possible to establish the existence of curvatures of the road, something which will be discussed in more detail below, inter alia with reference to Figs 7a and 7b. Fig. 4 shows a schematic selection of lines and positions (distances) of the objects in Fig. 1 according to an embodiment of the invention. According to the invention, the direction of hypothetical lines between the static objects 33 is calculated. In this embodiment, processing of lines from all objects to all objects does not take place, rather only lines between each static object and the static objects located within a certain distance therefrom, e.g. 10 metres, are calculated. This choice is based on the assumption that it is less likely that lines between objects located far away from each other represent the actual direction of the path. In the basic case, two objects are connected by one line only, i.e. no line is counted twice.

Each line represents a direction. This direction may be represented in several ways. For example, as the quotient Δy/Δx, shown at 53, or as the angle α, shown at 54. It should be noted that these two directional indications are relatable to the direction of travel of the vehicle 20 (Δy/Δx = ∞ or α = 90°) . It should also be noted that the directions and positions in Fig. 4 are indicated relative to a coordinate system which is defined in relation to the vehicle 20. However, it will be appreciated that these values could just as well be calculated on the basis of a ground-based system of coordi- nates. In the case of moving objects 31, 32 similar indications of direction denote the direction of motion of the object itself in relation to the direction of travel of the vehicle 20, as indicated by arrows adjacent to the vehicles 31 and 32. Normally, this means that it is necessary to compare the position of the moving objects at two different points in time.

According to a further alternative, all derived directions deviating more than a set limit value, e.g. 50 degrees, from the direction of travel of the vehicle are quite simply ignored, since it is unlikely that such directions represent the direction of the path in front of the vehicle. A limit value for a deviation in relation to a predicted direction of the path, based on a previous prediction carried out with a certain degree of reliability, could also be used.

Each direction obtained, represented by a respective line or direction of motion, is relevant with respect to its position. Consequently, for each directional value obtained, a corresponding positional value is derived, which is indicated in Fig. 4 with the vehicle as the point of reference. This distance value can be expressed in several ways. For example, as the shortest distance between the vehicle 20 and the line/object, as indicated by dashed lines at 50, or as the distance between the vehicle and the line/object projected onto an axis extending along the direction of travel of the vehicle, as indicated by dashed lines at 51, or as the arc distance to the line/object taking into the account the angle thereto, seen from the vehicle, as indicated by dashed lines at 52. With respect to lines found, a point in the middle of the respective line is advantageously chosen when determining the position as described above, since the direction of the respective line most likely best represents the direction of the path at a distance halfway between the objects. If the directions and distances derived from Fig. 4 as described above are drawn as points in a diagram showing direction as a function of distance, a relationship similar to the one shown in Fig. 5 is obtained. Since this diagram can be said to represent a spatial differentiation of the original positional information, in accordance with the discussion with reference to Figs 2, 3a, and 3b above, the existence of an essentially circular curve in the spatial space will be represented by an essentially linear relationship between the points in the direction/distance diagram, as indicated by a dashed line in Fig. 5. By determining the parameters of this straight line, on the basis of the points obtained, it is easy to calculate the parameters of the corresponding curve in the road.

A preferred method for deriving by means of the Hough transform the parameters of the straight line indicated between points in Fig. 5 will now be described with reference to Figs 6a, 6b, 7a and 7b. Suppose that we wish to determine the straight line between two points (or number pairs) (xi, yi) and (x₂, y₂) in an image with axes x and y, as shown in Fig. 6a. In this case, the line sought in the x-y plane can be described by means of- the equation y = ax + b, where a and b are parameters which must be determined in order for the line to be unambiguously defined. Since the line sought intersects the point (x_lf y_x) , it follows that yi = axi + b. This relationship can also be expressed as b = yi - axi, which can be seen as a straight line in a plane which has the parameters a and b as its axes, a so- called parameter space (which in this case has two dimensions), as shown in Fig. 6b. This straight line in the parameter space represents all combinations of a and b resulting in lines in the x-y plane which intersect the point (xi, yi) .

Since the line sought also intersects the point (x₂, y₂) it also follows that y₂ = ax₂ + b. Correspondingly, this can be expressed as b = y₂ - ax₂, which can also be seen as a straight line in said parameter space. Each point in the x-y plane in Fig. 6a can thus be represented by a line in the parameter space in Fig. 6b. The line intersecting both the point (xi, yi) and the point (x₂, y₂) has parameters a and b which shall be found in both corresponding lines in the parameter space, i.e. which correspond to the intersecting point between the lines in the parameter space, as indicated by dashed lines in Fig. 6b. By determining the intersecting point in the parameter space in Fig. 6b, the line in the x-y plane in Fig. 6a is unambiguously determined.

A description corresponding to the one given above with reference to Figs 6a and 6b will now be provided on the basis of a diagram showing direction as a function of distance in Fig. 7a, which corresponds, by and large, to the diagram described with reference to Fig. 5. For each of the six points in Fig. 7a, a corresponding line in the parameter space a-b is determined. In practice, this line is represented by weighted elements in a parameter matrix. In the case of five of these lines a probable intersecting point can be determined, which gives a corresponding line in the direction/distance diagram. Since the lines do not intersect at exactly the same point, a probable common intersecting point is derived by calculating local maxima in said matrix subsequent to a preliminary mean value formation or equalisation thereof. The maximum thus determined in the parameter space can subsequently be converted into a line in the direction/ distance space, which in turn can be converted into a curve in the original spatial space.

It should be noted that the sixth line, corresponding to the divergent point in the direction/distance diagram, intersects the other lines in Fig. 7b at points located too far apart from the other intersecting points, subsequent to said mean value formation and calculation of maxima, for it to be assumed to correspond to a prob- able line in the direction/distance plane. Consequently, the divergent point in Fig. 7a will not influence the determination of the values of the parameters of the functional relationship, unlike the case where a line in the direction/distance space is determined by means of, for example, least squares approximation on the basis of all the points.

It will be appreciated that carrying out the invention does not involve actually drawing the obtained values of directions and positions (distances) in a diagram of the kind shown in Figs 5 and 7a. Rather, the corresponding curves can be formed in the parameter space matrix (corresponding to Fig. 7b) starting directly from the directions (number pairs) obtained in the radar image in Fig. 1. Correspondingly, it will be appreciated that the positional information need not be represented in the form of an image of the kind shown in Fig. 1 in order to carry out the steps of the invention. Rather, the processing can be carried out on the basis of raw signal data with no intermediate image analysis.

Fig. 8 schematically shows the division of an area in front of the vehicle into subareas according to an embodiment of the invention. In Fig. 8, area A relates to a first portion closest to the vehicle and area B relates to an adjacent portion located farther away from the vehicle. For example, in the phase schematically illustrated in Fig. 8, there is no substantial line in area A, while a curvature will probably be predicted in area B. As a result of this division the processing described with reference to the above Figures is advantageously carried out separately for separate areas. In this connection, the Hough transform described above is preferable carried out in separate matrices for the separate areas A and B. Areas A and B can be defined either with respect to the vehicle 20 or with respect to the environment. In the latter case, areas A and B will move towards the vehicle as the vehicle moves forward in the environment, and new areas will successively replace areas A and B.

With reference to the above illustrations it will be appreciated that all possible objects, representing all existing radar reflections, need not be taken into consideration in the processing. Rather, the processing may advantageously be limited to such reflections as can be determined with some reliability to correspond to actual objects.

It will be appreciated that the above description of preferred embodiments is provided by way of example only, and that modifications and combinations thereof can be carried out within the scope of the invention, which is defined in the appended claims.

Claims

1. A method for predicting the existence of a curve in a road portion in front of a vehicle in connection with the utilisation of a radar arranged on the vehicle which, by transmitting a radar signal and recording received reflections thereof, provides information indicating the position of objects located in an area in front of the vehicle , c h a r a c t e r i s e d by the steps of : deriving values representing directions of hypothetical lines between said objects and to said directions associating derived corresponding values each of which defines the position of a respective one of said lines between said objects; and determining whether at least a subset of said values of directions and corresponding positions fulfils a functional relationship between direction and position and predicting the presence of said curve on the basis thereof.

2. A method according to claim 1, wherein the step of determining whether at least a subset of said values of directions and corresponding positions fulfils a functional relationship between direction and position comprises Hough transformation of the number pairs obtained from said values of directions and corresponding positions .

3. A method according to claim 1, wherein the step of determining whether at least a subset of said values of directions and corresponding positions fulfils a functional relationship between direction and position com- prises Hough transformation of the number pairs obtained from said values of directions and corresponding positions into curves in a parameter space, determining one or more intersecting points between these curves in said parameter space, and on the basis of these intersecting points determining the values of parameters of said functional relationship.

4. A method according to claim 1, 2 or 3, wherein the step of determining whether at least a subset of said values of directions and corresponding positions fulfils a functional relationship between direction and position comprises: transforming number pairs, obtained from said values of directions and corresponding positions, into respective curves in a parameter space, whose dimensions correspond to the parameters included in said functional rela- tionship of direction as a function of position for a hypothetical range, the parameter combinations obtained from said curves indicating possible appearances of said functional relationship; and determining one or more intersecting points between the curves in said parameter space and on the basis of these working out parameter combinations indicating specific appearances of said functional relationship, corresponding to specific appearances of curves present in the road portion.

5. A method according to claim 4, wherein the representation of said curves in said parameter space is achieved by changing element values in a matrix having the same dimensions as the parameter space.

6. A method according to claim 5, wherein said determination of one or more intersecting points comprises determination of maxima in said matrix.

7. A method according to claim 4, 5, or 6, wherein said functional relationship as well as said curves are obtained from first degree polynomials.

8. A method according to claim 1, wherein the step of determining whether at least a subset of said values of directions and corresponding positions fulfils a functional relationship between direction and position com- prises: least squares approximation of a functional relationship between direction and position for a hypothetical road portion to the number pairs obtained from said values of directions and corresponding positions; and determining the probable appearance of a curve in the road portion on the basis of said least squares approximated functional relationship.

9. A method according to any one of the preceding claims, comprising processing only derived values of determined directions deviating less than a predetermined limit value from the direction of travel of the vehicle.

10. A method according to any one of the preceding claims, comprising processing only derived values of determined directions deviating less than a predetermined limit value from a predicted direction of the path in the position corresponding to the respective determined direction on the basis of a previous prediction of the direction of the path.

11. A method according to any one of the preceding claims, comprising weighting the presence of said number pairs depending upon a classification of the object or objects from which said number pairs derive.

12. A method according to any one of the preceding claims, wherein the step of deriving values comprises the measure of deriving with respect to at least one of the objects a value representing a direction of a line between that object and another object.

13. A method according to claim 12, wherein said measure comprises deriving for said at least one of the objects a plurality of values representing directions of lines between that object and a plurality of the other objects.

14. A method according to claim 12 or 13, comprising carrying out said measure for a plurality of the objects.

15. A method according to claim 12, 13 or 14, comprising carrying out said measure with respect to such objects as have been added to objects which have already been processed in a previous stage.

16. A method according to any one of the preceding claims, wherein said values of positions of said lines are represented as distances from a reference point to said lines.

17. A method according to any one of the preceding claims, wherein said values of positions of said lines are represented as distances from the vehicle to said lines .

18. A method according to claim 17, wherein said distances from the vehicle to the respective lines are chosen to be the distances from the vehicle to respective points on the respective lines.

19. A method according to claim 18, wherein the respective point is chosen halfway between the objects between which the corresponding line extends.

20. A method according to claim 18, wherein said distance is approximated as the shortest distance between the vehicle and a point on the respective line.

21. A method according to any one of the preceding claims, comprising deriving only directions of lines between objects located within a predetermined distance from each other.

22. A method according to any one of the preceding claims, wherein said values representing directions are related to the direction of travel of the vehicle.

23. A method according to any one of the preceding claims, wherein said directions are obtained with respect to a system of coordinates which is defined with respect to the vehicle.

24. A method according to any one of the preceding claims, wherein said directions are obtained in relation to a ground-based system of coordinates.

25. A method according to any one of the preceding claims, wherein said step is carried out in separate processing operations for separate subareas of said area.

26. A method according to any one of the preceding claims, wherein the step of deriving values comprises deriving, with respect to an object which on the basis of information obtained at separate points in time is determined to be a moving object, a value of a direction of the object in the form of the direction of motion of the object.

27. A device for carrying out the method according to any one of the preceding claims.