METHOD FOR CLASSIFYING OBJECTS

Abstract

A method for classifying objects into object classes on the basis of information of an ultrasonic sensor of a vehicle, including receiving multiple detections of a vehicles's ultrasonic sensor. Items of position information and directional information are assigned to each detection. The position information indicates a reflection location where an ultrasonic sensor's signal was reflected and the directional information indicates a direction along which the ultrasonic signal propagates between the reflection location and ultrasonic sensor. The method includes forming detection clusters based on the received detections, with one cluster including multiple detections; calculating statistical distribution information of the position information and the directional information of the detections assigned to the respective cluster; and classifying an object into an object class based on the statistical distribution information of the position information and the directional information of the clusters.

Claims

1. A method for classifying objects into object classes on the basis of information of at least one ultrasonic sensor of a vehicle, comprising: a) receiving multiple detections of at least one ultrasonic sensor of a vehicle, wherein an item of position information and an item of directional information are assigned to each detection, wherein the position information indicates a reflection location at which an ultrasonic signal of the at least one ultrasonic sensor was reflected and wherein the directional information indicates a direction along which the ultrasonic signal propagates between the reflection location and the at least one ultrasonic sensor; b) forming clusters of detections on the basis of the received detections, wherein one cluster comprises multiple detections; c) calculating information on a statistical distribution of the position information and information on a statistical distribution of the directional information of the detections assigned to the respective cluster; and d) classifying an object into an object class on the basis of the information on the statistical distribution of the position information and on the statistical distribution of the directional information of the clusters.

2. The method according to claim 1, herein the position information comprises at least one first coordinate and at least one second coordinate, wherein the information on the statistical distribution of the position information comprises information which is based on a covariance matrix of the first and second coordinates of the position information.

3. The method according to claim 2, wherein the information on the statistical distribution of the position information comprises eigenvalues of the covariance matrix of the first and second coordinates of the position information.

4. The method according to claim 3, wherein the information on the statistical distribution of the position information comprises a ratio of the eigenvalues of the covariance matrix of the first and second coordinates of the position information.

5. The method according to claim 1, wherein the information on the statistical distribution of the directional information of the detections comprises a variance of the directional information.

6. The method according to claim 1, wherein the information on the statistical distribution of the directional information of the detections comprises a derivative over time of the directional information.

7. The method according to claim 1, wherein the information on the statistical distribution of the directional information of the detections comprises a derivative over time of directional information filtered by a filter function.

8. The method according to claim 1, wherein the classification is conducted at least on the basis of a first threshold value and a second threshold value, wherein the first threshold value indicates a threshold value for the information on the statistical distribution of the position information and the second threshold value indicates a threshold value for the information on the statistical distribution of the directional information of the detections.

9. The method according to claim 8, wherein the first and second threshold values are determined by training data which have label information on the respective object classes.

10. The method according to claim 1, further comprising carrying out the classification by a decision tree or a random forest.

11. The method according to claim 1, further comprising using a neural network for the classification, wherein the neural network being trained by training data which have label information on the respective object classes.

12. The method according to further comprising carrying out a classification into the object classes vehicle and not a vehicle.

13. The method according to claim 1, further comprising selectively estimating a height of the object depending on a result of the classification of the object.

14. A system for classifying objects into object classes on the basis of information of at least one ultrasonic sensor of a vehicle, wherein the system has a computing unit which is configured to execute a method comprising: receiving multiple detections of at least one ultrasonic sensor of a vehicle, wherein an item of position information and an item of directional information are assigned to each detection, wherein the position information indicates a reflection location at which an ultrasonic signal of the at least one ultrasonic sensor was reflected and wherein the directional information indicates a direction along which the at least one ultrasonic signal propagates between the reflection location and the at least one ultrasonic sensor; forming clusters of detections on the basis of the received detections, wherein one cluster comprises multiple detections; calculating information on a statistical distribution of the position information and information on the statistical distribution of the directional information of the detections assigned to the respective cluster; and classifying the object into an object class on the basis of the information on the statistical distribution of the position information and on the statistical distribution of the directional information of the clusters.

15. A vehicle comprising a system according to claim 14.

Description

BRIEF DESCRIPTION OF THE DRAWINGS

[0033] The present disclosure is explained in greater detail below on the basis of figures with reference to exemplary embodiments, wherein:

[0034] FIG. 1 shows, by way of example and schematically, a vehicle having ultrasonic sensor technology including multiple ultrasonic sensors distributed around the vehicle on the circumference and a computer unit for evaluating the information provided by the ultrasonic sensor technology;

[0035] FIG. 2 shows, by way of example, a diagram which shows the position information of detections for a longitudinal parking system between two vehicles;

[0036] FIG. 3 shows, by way of example, a diagram which shows the position information and partially the directional information of detections for the longitudinal parking system according to FIG. 2;

[0037] FIG. 4 shows, by way of example, multiple clusters which were determined on the basis of the detections shown in FIG. 2 and FIG. 3;

[0038] FIG. 5 shows, for example, a decision tree for classifying objects into object classes; and

[0039] FIG. 6 shows, by way of example, a block diagram which illustrates the steps of the method for classifying objects into object classes.

DETAILED DESCRIPTION

[0040] FIG. 1 shows, by way of example and merely schematically, a vehicle 1. The vehicle 1 has a plurality of ultrasonic sensors 2 by means of which the surroundings are captured.

[0041] The ultrasonic sensors 2 are coupled to at least one computing unit 3, by means of which the method described below for classifying an object O into an object class is carried out.

[0042] FIG. 2 shows, by way of example, a diagram in which a plurality of detections is entered which are captured by the ultrasonic sensors 2 of the vehicle 1 and assigned to multiple objects O. The detections refer, for example, to a longitudinal parking system in which a longitudinal parking space is formed between two vehicles. The linearly arranged detections between the two vehicles refers, for example, to a curb which laterally delimits the parking space.

[0043] A detection is represented by a point. The reflection location at which the detection occurred due to a reflection by a surrounding object is indicated by position information. The position information can include at least two coordinates, by means of which the reflection location is defined in a horizontal plane. These are in particular x and y coordinates which are plotted on the respective diagram axes in the diagram of FIG. 2. In other words, the position information is indicated in Cartesian coordinates. Alternatively, it is possible to indicate the position information in a cylindrical coordinate system or a spherical coordinate system.

[0044] The statistical distribution of the position information is enlisted for classifying an object.

[0045] If available, the position information can also contain information regarding the height of the object region at which the reflection occurred. In other words, the position information can indicate the reflection location in the three-dimensional space. The information on the third dimension (i.e., the height) can also be enlisted to classify the object.

[0046] An item of directional information is also determined in each case for the detections. The directional information indicates from which direction the detection was received. The directional information can, for example, be a spanned angle in the horizontal plane, which indicates the direction of a connecting line which connects the reflection location to the sensor position of the ultrasonic sensor which emitted the reflected ultrasonic signal and/or which received the reflected ultrasonic signal. The angle can, for example, be measured relative to an axis of coordinates, for example relative to the x axis. For example, the directional information can therefore include the angle which the connecting line, which connects the reflection location to the sensor position of the ultrasonic sensor, encloses with the x axis.

[0047] FIG. 3 shows, by way of example, a diagram similar to that of FIG. 2, wherein an item of directional information is additionally assigned to each detection which is assigned to the left vehicle. The directional information is in each case signified by a line. The line indicates the direction in which the detection was determined. In other words, the line indicates the direction in which the ultrasonic sensor 2 was situated when emitting and/or receiving the ultrasonic signal. It should be noted that, due to the high propagation speed and the associated short transit time of the ultrasonic signal between emission and reception, it can be approximately assumed that the ultrasonic sensor 2 is situated in the same location when sending an ultrasonic signal and receiving the reflected components of this ultrasonic signal.

[0048] In order to assign multiple detections to one object, at least one cluster C is formed from the captured detections. Known clustering algorithms can be used to form a cluster, for example, density-based clustering methods, partitioning clustering methods, etc. In particular, a K-means clustering algorithm or a DBSCAN algorithm can be used to form the clusters.

[0049] FIG. 4 illustrates multiple clusters C which were determined on the basis of the detections depicted in FIGS. 2 and 3. The ellipses depicted in FIG. 4 each represent the covariance matrix of the x, y coordinates of those detections assigned to the respective cluster C. The clusters C refer to locally connected groups of detections and may indicate how the respective group of detections is locally arranged in a horizontal plane (i.e., the x-y plane).

[0050] A threshold value may be predefined for the number of detections. The number of the detections which should form a cluster must exceed this threshold value. This can prevent a cluster from already being formed due to fewer detections.

[0051] Following the cluster formation, information on the statistical distribution of the position information and information on the statistical distribution of the directional information of the detections can be determined for the respective clusters.

[0052] In particular, the mean value of the position information is calculated for the respective clusters C. This mean value therefore indicates the center of the cluster. It can be determined, for example, by averaging the x and y coordinates of the detections of a cluster. In addition, the covariance to the x and y coordinates of the detections assigned to the respective cluster can be calculated. In other words, the two-dimensional Gaussian distribution of the x and y coordinates of the detections is calculated.

[0053] The information on the statistical distribution of the directional information of the detections of a cluster can include the mean value and the variance of the directional information of the detections of a cluster. In this case, the mean value and the variance can either be calculated directly or the directional information can be filtered prior to the calculation of the mean value and the variance, for example, a smoothing filtering or a filtering by means of which statistical outliers are filtered out.

[0054] Further information or variables which can be enlisted for classifying the objects can be calculated from the information on the statistical distribution of the position information and the information on the statistical distribution of the directional information.

[0055] For example, the determinant of the covariance matrix of the first and second coordinates of the position information can be calculated. This is larger in the case of multidimensionally shaped objects than in the case of linear objects. Therefore, the determinant of the covariance matrix can serve to distinguish between multidimensionally shaped objects and linear objects.

[0056] Furthermore, the eigenvalues of the covariance matrix of the first and second coordinates of the position information can be calculated. In particular, this produces a first and a second eigenvalue. The eigenvalues indicate the extension of the cluster C along its main axis and its auxiliary axis running perpendicularly thereto. In particular, the eigenvalues indicate the length of the main and auxiliary axis of an ellipse which can be used to reproduce the location and alignment of the cluster. These ellipses assigned to clusters C are depicted in FIG. 4.

[0057] Furthermore, the quotient of the first and second eigenvalues of the covariance matrix of the first and second coordinates of the position information can be calculated. The ratio of the eigenvalues provides an indication of whether it is a linear object or not, since the ratio of the eigenvalues differs considerably for linear objects and two-dimensionally shaped objects.

[0058] The variance of the directional information of the detections of a cluster likewise provides an indication of whether it is a linear object or a two-dimensionally shaped object. Thus, for example, the variance of the directional information is very small in the case of linear objects, and very high in the case of round objects such as posts or similar. In the case of vehicles, the variance of the directional information lies in the middle between the variance of the directional information of linear and round objects, since vehicles have both straight vehicle regions and diffusely reflecting regions, for example mirrors, door handles, etc.

[0059] In addition, the derivative over time of statistical properties of the directional information of the detections or of the filtered directional information of the detections can be determined, i.e., for example, the change in the mean value, the variance, etc.

[0060] The information on the statistical distribution of the position information and the information on the statistical distribution of the directional information may be determined iteratively, and indeed in such a way that the formation of clusters C is updated when one or more detections are received. As a result, updated clusters are obtained. The information on the statistical distribution of the position information and the information on the statistical distribution of the directional information of the detections assigned to a cluster are likewise updated following the updating of a cluster, i.e., recalculated on the basis of the detections newly added to the cluster.

[0061] The information on the statistical distribution of the position information and the information on the statistical distribution of the directional information of the detections can subsequently be enlisted to form decision rules, wherein the classification of the objects into object classes is conducted on the basis of the decision rules.

[0062] Training data can be used to form the decision rules. The training data have information on objects and label information assigned to the objects. The label information indicates which object class the respective object should be assigned to. The decision rules can in particular be threshold values. The height of the threshold values can be fixed on the basis of the training data.

[0063] The threshold values can be assigned to specific information on the statistical distribution of the position information or specific information on the statistical distribution of the directional information. The threshold values can in particular indicate that an item of information below the threshold value indicates a classification into a first object class and an item of information above the threshold value indicates a classification into a second object class.

[0064] A decision tree or a random forest can be deployed for classifying the objects.

[0065] The structure and the decision rules of the decision tree or of the random forest can be fixed by the training data.

[0066] FIG. 5 shows, by way of example, a decision tree. The decision tree serves, by way of example, to classify the objects into the classes curb, wall and vehicle. In particular, the decision tree is used for classification into the object classes vehicle and not a vehicle.

[0067] The decision rules of the decision tree refer to information on the statistical distribution of the position information and information on the statistical distribution of the directional information. The quotient of the first and second eigenvalues of the covariance matrix of the first and second coordinates of the position information, i.e., the ratio of the two eigenvalues of the covariance matrix, is used as information on the statistical distribution of the position information. The variance of the directional information is used as information on the statistical distribution of the directional information. On the basis of the information on the statistical distribution of the position information, the information on the statistical distribution of the directional information, and the threshold values which are assigned to this information, the detections of the respective cluster C can be assigned to an object class. Therefore, it can in particular be decided whether the detections of the recognized cluster C refer to the object class vehicle or not.

[0068] According to the decision tree of FIG. 5, if the variance of the directional information is greater than 0.00665 and the ratio of the eigenvalues is greater than 0.0012, it can be assumed with a probability of 99% that the detections of the cluster refer to a vehicle.

[0069] Following the classification of the objects or the clusters assigned to objects into object classes, a height estimation algorithm can be selectively performed on the basis of the information of the ultrasonic sensor technology. In particular, those detections which are assigned to clusters of the object class vehicle can be excluded from the height estimation in order to avoid incorrect estimations.

[0070] FIG. 6 shows, in a schematic representation, the steps of a method according to the invention for classifying objects into object classes by means of the ultrasonic sensor technology of a vehicle.

[0071] Initially, multiple detections of at least one ultrasonic sensor of a vehicle are received (S10). An item of position information and an item of directional information are assigned to each detection. The position information indicates the reflection location at which an ultrasonic signal of the at least one ultrasonic sensor was reflected. The directional information indicates the direction along which the ultrasonic signal propagates between the reflection location and the at least one ultrasonic sensor.

[0072] Clusters of detections are subsequently formed on the basis of the received detections. One cluster includes multiple detections (S11).

[0073] Information on the statistical distribution of the position information and information on the statistical distribution of the directional information of the detections assigned to the respective cluster are subsequently calculated for the clusters (S12).

[0074] Finally, a cluster which is assigned to one object is classified into an object class, and indeed on the basis of the information on the statistical distribution of the position information and on the statistical distribution of the directional information (S13).

[0075] The invention has been described above using exemplary embodiments. It goes without saying that numerous changes as well as modifications are possible without leaving the scope of protection defined by the claims.

LIST OF REFERENCE NUMERALS

[0076] 1 Vehicle [0077] 2 Ultrasonic sensor [0078] 3 Computing unit [0079] C Cluster [0080] E Decision tree [0081] O Object [0082] s1 First threshold value [0083] s2 Second threshold value