Method and identification system for locating an object

Abstract

The invention relates to a method (1) for locating an object. In this process, position data (PD1, PD2-PDn) from at least two sensor units (S1, S2-Sn) are provided and filtered separately. The filtered position data (PD1, PD2-PDn) of all sensor units (S1, S2-Sn) are fused into fused position data (FPD). Velocity data and / or direction data are derived from the fused position data (FPD) and filtered. The invention also relates to an identification system for carrying out the method (1).

Description

[0001] Mercedes-Benz Group AG

[0002] Methods for locating an object and

[0003] an identification system for this

[0004] The invention relates to a method for locating an object from position data from at least two sensor units. The invention also relates to an identification system for executing the method.

[0005] Precise localization of an object or user is particularly important in safety-critical applications, such as vehicle access control. The position data used to locate the object or user can originate from various sensor sources. However, each sensor source has its own cycle time and noise level. Therefore, localization of the object or user based on fused position data from disparate sensor sources can only be inaccurate. For example, differing cycle times of position data from various sensor sources can lead to unreliable calculations of the user's speed. Furthermore, existing filtering methods, which are designed for linear data, cannot determine the direction of the object or user.

[0006] DE 102014211 176 A1 discloses a method for correcting measurement data from a sensor base system. The sensor base system and at least one other sensor system acquire measurement data. Using the measurement data from the sensor system, error values ​​of the measurement data from the sensor base system are determined. The error values ​​are corrected by applying corrections.

[0007] DE 102019208872 A1 discloses a method for error assessment in position determination. In this method, first and second position values ​​are recorded synchronously using different measurement methods. From the measured values, first and second trajectories are generated and shifted along a displacement vector such that the difference vectors between position values ​​are minimized. The error in the position determination is then determined based on one or more magnitudes of the minimized difference vectors.

[0008] The object of the invention is therefore to provide an improved or at least alternative embodiment of a method of the generic type, in which the described disadvantages are overcome. The object of the invention is also to provide an identification system for carrying out the method.

[0009] This problem is solved according to the invention by the subject matter of the independent claims. Advantageous embodiments are the subject matter of the dependent claims.

[0010] The method according to the invention is designed for locating an object, person, or user. Localization can, in particular, include determining the object's position, speed, and / or direction. The method involves carrying out measures a), b), c), d), and / or e). Measures a), b), and c) are carried out in the order specified herein. After measure c), measures d) and / or e) can then be carried out. The order of measures d) and e) is arbitrary. Measures d) and e) can also be carried out simultaneously.

[0011] In measure a), position data comprising multiple position values ​​representing the object's position over time are provided by at least two sensor units. The position data can be provided by the sensor unit in the form of a UWB sensor (Ultra-Wideband) and / or a BLE sensor (Bluetooth Low Energy) and / or a surround camera and / or a radar sensor (RADAR: Radio Detection and Ranging). In measure b), the position data from each sensor unit is then filtered separately. The position data can be filtered, for example, using a median filter and / or a low-pass filter. In measure c), the filtered position data from all sensor units are fused or combined to create fusion position data. The filtered position data from all sensor units can be fused to create the fusion position data, in particular, using a Kalman filter.After step c), the fusion position data can be made available to read the object's position. In step d), velocity data is derived and filtered from the fusion position data. The velocity data comprises several velocity values ​​representing the object's velocity over time. After step d), the velocity data can be made available to read the object's velocity. In step e), direction data is derived and filtered from the fusion position data. The direction data comprises several direction values ​​representing the object's direction over time. After step e), the direction data can be made available to read the object's direction.

[0012] In the method according to the invention, the position data of the sensor units are filtered separately, thereby solving the problem of inconsistent data. In particular, outliers and noise can be removed from the position data of the sensor units by filtering, thereby increasing the reliability and accuracy of the position data of the sensor units. As a result, the position of the object can be reliably determined from the fusion position data according to measure c). Furthermore, in measures d) and e), the velocity data and / or the direction data can then be reliably extracted from the fusion position data, thereby reliably determining the velocity and / or the direction of the object. The velocity data extracted or derived from the fusion position data and / or the direction data extracted or derived from the fusion position data can be used to determine the position of the object.Derived direction data can be used to reduce the false acceptance and rejection rates in an identification system designed to execute the procedure. This can improve the overall performance of the identification system. In total, the procedure can improve the accuracy of object localization and optimize data continuity for determining the object's velocity and direction.

[0013] The object's speed is a crucial parameter for the identification system. In particular, the object's speed can be used to assess its potential future intentions. This can be mitigated by using the average time difference, rather than the instantaneous time difference, when calculating the speed, thus compensating for irregularities in the cycle times and enabling a reliable and robust speed assessment.

[0014] In one possible embodiment of the method, steps i), ii), iii), and iv) of measure d) can be carried out sequentially in the order mentioned. In step i), x-velocity data and y-velocity data can be derived from the fusion position data. The x-velocity data comprises several x-velocity values ​​representing the object's x-velocity as a function of time, and the y-velocity data comprises several y-velocity values ​​representing the object's y-velocity as a function of time. The object's x-velocity and y-velocity values ​​can, for example, be calculated for each position value of the fusion position data in a manner known to those skilled in the art. In step ii), the x-velocity and y-velocity data can each be filtered or smoothed separately using a polynomial smoothing filter.The noise in the x-velocity and y-velocity data can be reduced by using a polynomial smoothing filter. For example, a Savitzky-Golay filter can be used as the polynomial smoothing filter for each of the x-velocity and y-velocity data. In step iii), the filtered x-velocity and y-velocity data can be combined to obtain the velocity data. In step iv), the velocity data can then be filtered or smoothed using a smoothing filter. A Kaiser window method can be used as the smoothing filter for filtering or smoothing the velocity data. The Kaiser window method reduces peaks in the velocity's time course.As described above, the speed data can then be provided to read the speed of the object.

[0015] The object's direction is a crucial parameter for the identification system. In particular, the object's direction can be used to assess its further intentions.

[0016] In one possible embodiment of the method, steps i), ii), iii), iv), v), vi), and vii) in measure f) can be carried out sequentially in the order mentioned. In step i), x-velocity and y-velocity data can be derived from the fusion position data. The x-velocity data comprises several x-velocity values ​​representing the object's x-velocity as a function of time, and the y-velocity data comprises several y-velocity values ​​representing the object's y-velocity as a function of time. The object's x-velocity and y-velocity values ​​can, for example, be calculated for each position value of the fusion position data in a manner known to those skilled in the art. In step ii), the direction values ​​of the direction data in degrees can be calculated from the x-speed values ​​of the x-speed data and the associated y-speed values ​​of the y-speed data.The direction values ​​of the direction data calculated in step ii) can, for example, lie in a range between 180° and -180°. After step ii) and before step iii), the calculated direction values ​​of the direction data can then be converted from the range between 180° and -180° to a range between 0° and 360°. In step iii), each direction value of the direction data can be converted from degrees to radians. This ensures mathematical consistency and ease of calculation during further processing of the direction data. In step iv), a complex direction value can then be calculated from each direction value of the direction data. A cosine of the direction value can be assigned to a real part of the complex direction value, and a sine of the direction value to an imaginary part of the complex direction value.In step v), the real parts and imaginary parts of the complex direction values ​​of the direction data can be filtered separately using Savitzky-Golay filters. This allows successive subsets of the real parts and imaginary parts of the direction values ​​to be fitted with a polynomial of a specific degree, thus reducing noise without distortion. In step vi), a real direction value can be calculated from each complex direction value of the direction data. The real direction value can be calculated as the arctangent of the ratio of the imaginary and real parts of the complex direction value. This provides the real direction value in radians. In step vii), the real direction values ​​of the direction data can then be converted from radians to degrees.This allows the direction data to be converted back into its original format. As described above, the direction data can then be made available for reading the direction of the object. The invention also relates to an identification system for locating an object, a person, or a user, wherein the identification system is designed to execute the method described above. The identification system may include software and / or hardware necessary for executing the method.

[0017] Further important features and advantages of the invention will become apparent from the dependent claims, the drawings and the associated description of the figures based on the drawings.

[0018] It is understood that the features mentioned above and those to be explained below can be used not only in the combination specified, but also in other combinations or on their own, without leaving the scope of the present invention.

[0019] Preferred embodiments of the invention are shown in the drawings and are explained in more detail in the following description, wherein identical reference numerals refer to identical or similar or functionally identical components.

[0020] They show, schematically, each one

[0021] Fig. 1 shows a schematic sequence of a method according to the invention with measures a), b), c), d) and e);

[0022] Fig. 2 shows a schematic sequence of measure d) of the method according to the invention;

[0023] Fig. 3 shows a schematic partial sequence of measure e) of the method according to the invention;

[0024] Fig. 4 shows a spatial diagram of a reversal of an object;

[0025] Fig. 5 shows a direction-time diagram of the reversal of the object from Fig. 4. Fig. 1 shows a schematic sequence of a method 1 according to the invention for locating P of an object, person, or user. In method 1, measures a), b), c), d), and e) are carried out. Measures a), b), and c) are carried out sequentially, and after measure c), measure d) and / or e) can be carried out.

[0026] In step a), position data PD1, PD2 to PDn are provided by various sensor units S1, S2 to Sn. The position data PD1, PD2 to PDn comprise several position values ​​representing the object's position P over time. The sensor units S1, S2 to Sn can be a UWB sensor and / or a BLE sensor and / or a surround camera and / or a radar sensor. In step b), the position data PD1, PD2 to PDn from each sensor unit S1, S2 to Sn are then filtered separately using a median filter and / or a low-pass filter. In step c), the filtered position data PD1, PD2 to PDn are fused into fusion position data FPD using a Kalman filter. After step c), the fusion position data FPD is available for reading the object's position P.

[0027] In measure d), step i), velocity data is derived from the Fusion Position Data (FPD) and filtered in steps ii) to iv). The velocity data comprises several velocity values ​​representing the object's velocity v over time. After measure d), the velocity data is made available for reading the object's velocity v. In measure e), steps i) and ii), direction data is derived from the Fusion Position Data (FPD) and filtered in steps iii) to vii). The direction data comprises several direction values ​​representing the object's direction R over time. After measure e), the direction data is made available for reading the object's direction R.

[0028] Fig. 2 shows a schematic sequence of step d) of the method 1 according to the invention. In step d), the velocity data are derived from the fusion position data FPD and filtered. In step i), x-velocity and y-velocity data are derived from the fusion position data FPD. This can be done in a manner known to those skilled in the art. The x-velocity data comprises several x-velocity values ​​v representing the object's x-velocity as a function of time. x and the y-velocity data, several y-velocity values ​​representing the object's y-velocity in a time-dependent manner. yIn step ii), the x-velocity and y-velocity data are smoothed or filtered separately using a polynomial smoothing filter, such as a Savitzky-Golay filter. In step iii), the smoothed x-velocity and y-velocity data are combined or merged to form the velocity data. The x-velocity values ​​v are then used to calculate the velocity. x and the corresponding y-velocity values ​​v y The velocity values ​​v were calculated in a manner known to a person skilled in the art.

[0029] In step iv), the velocity data is then smoothed or filtered using a Kaiser window method. After step d), the velocity data is ready to read the velocity v of the object.

[0030] Fig. 3 shows a schematic partial sequence of step e) of the inventive method 1. In step e), the direction data are derived and filtered from the fusion position data FPD. In step i), x-velocity data and y-velocity data are derived from the fusion position data. This can be done in a manner known to those skilled in the art. The x-velocity data includes an x-velocity value v for each position value of the fusion position data FPD. x The object's y-velocity data includes a y-velocity value v for each position value of the FPD fusion position data. y of the object. In step ii), v is calculated from the x-velocity values. x the x-velocity data and the associated y-velocity values ​​v yThe direction values ​​of the y-velocity data are calculated in degrees. The calculated direction values ​​lie in a range between 180° and -180° and are converted to a range between 0° and 360°. In step iii), each direction value is converted from degrees to radians. In step iv), a complex direction value is then calculated from each direction value. A cosine of the direction value is assigned to a real part of the complex direction value, and a sine of the direction value is assigned to an imaginary part. In step v), the real parts and imaginary parts of the complex direction values ​​are then filtered separately using Savitzky-Golay filters. In step vi), a real direction value is calculated from each complex direction value.The real direction value can be calculated as the arctangent of the ratio of the imaginary and real parts of the complex direction value. This provides the real direction value in radians. In step vii), the real direction values ​​are then converted from radians to degrees. This provides the direction data in its original format. The direction data is then available for reading the object's direction.

[0031] Fig. 4 shows a spatial diagram of an object's about-turn. The space is spanned here in an xy-plane. The object or user moves from left to right and then makes a clockwise about-turn. A dashed arrow indicates the object's or user's trajectory, and a dotted circle marks an area of ​​interest.

[0032] Fig. 5 shows a direction-time diagram of the turnaround of the object or user from Fig. 4 in the area of ​​interest indicated in Fig. 4. R denotes direction and t denotes time. R1 and dashed lines indicate the direction of the object or user obtained from direction data using a conventional method with linear filters. R2 and solid lines indicate the direction of the object or user obtained from processed direction data using the method 1 according to the invention. R3 and dotted lines indicate raw direction data and direction data not processed by any method, respectively.

[0033] Conventional methods apply outlier filters to linear data. These outlier filters cannot completely eliminate discontinuities caused by a so-called wrap-around effect at 360 degrees. This often leads to inaccuracies and misinterpretations of the direction data. For example, the direction profile R1 of the object or user obtained using the conventional method suggests an abrupt counterclockwise turn (10° -> 40° -> 160° -> 350°).

[0034] In the inventive method 1, a so-called circular filter or circular filter method is used when processing the direction data. This allows outliers to be removed from the direction data and the direction data to be smoothed. The direction profile R2 of the object or user reflects a gentle clockwise turn due to the application of the circular filter, which is closer to the actual trajectory of the object or user in Fig. 4.

Claims

Mercedes-Benz Group AG Patent claims 1. Method (1) for locating an object, a) wherein position data (PD1, PD2-PDn), comprising several position values ​​representing the position (P) of the object in a time-dependent manner, are provided by at least two sensor units (S1, S2-Sn), b) wherein the position data (PD1, PD2-PDn) of each sensor unit (S1, S2-Sn) are filtered separately, c) wherein the filtered position data (PD1, PD2-PDn) of all sensor units (S1, S2-Sn) are fused to form fusion position data (FPD), d) wherein velocity data, comprising multiple velocity values ​​representing a time-dependent velocity (v) of the object, are derived and filtered from the fusion position data (FPD), and / or e) wherein direction data, comprising multiple direction values ​​representing a time-dependent direction (R) of the object, are derived and filtered from the fusion position data (FPD).

2. Method (1) according to claim 1, characterized by that after measure c) the fusion position data (FPD) are made available to read the position (P) of the object, and / or that after measure d) the speed data are provided to read out the speed (v) of the object, and / or that after measure e) the direction data are provided to read out the direction (R) of the object.

3. Method (1) according to claim 1 or 2, characterized by that in measure a) the position data (PD1, PD2-PDn) are provided by the sensor unit (S1, S2-Sn) in the form of a UWB sensor and / or a BLE sensor and / or a surround camera and / or a radar sensor.

4. Method (1) according to any one of the preceding claims, characterized by that in measure b) the position data (PD1, PD2-PDn) of each sensor unit (S1, S2-Sn) are filtered using a median filter and / or a low-pass filter.

5. Method (1) according to any of the preceding claims, characterized by that in measure c) the filtered position data (PD1, PD2-PDn) of all sensor units (S1, S2-Sn) are fused to the fusion position data (FPD) using a Kalman filter.

6. Method (1) according to any of the preceding claims, characterized by the fact that in measure d) i) from the fusion position data (FPD) x-velocity data, which contains multiple x-velocity values ​​(v) representing the object's x-velocity over time. x ) include, and y-velocity data, which includes multiple y-velocity values ​​(v) representing the object's y-velocity over time. x ) include, be derived, ii) the x-speed data and the y-speed data are each filtered separately using a polynomial smoothing filter, iii) the filtered x-speed data and the filtered y-speed data are combined to form the speed data, and iv) the speed data are filtered using a smoothing filter.

7. Method (1) according to claim 6, characterized by that a Savitzky-Golay filter is used as the polynomial smoothing filter for filtering the x-speed data and the y-speed data, and / or that a Kaiser window method is used as the smoothing filter for filtering the speed data.

8. Method (1) according to any of the preceding claims, characterized by the fact that in measure f): i) from the fusion position data (FPD) x-velocity data, which contains multiple x-velocity values ​​(v) representing the object's x-velocity over time. x ) include, and y-velocity data, which includes multiple y-velocity values ​​(v) representing the object's y-velocity over time. x ) include, be derived, ii) from the x-velocity values ​​(v x ) the x-velocity data and the associated y-velocity values ​​(v y ) the direction values ​​of the direction data are calculated in degrees from the y-velocity data, iii) each direction value of the direction data is converted from degrees to radians, iv) a complex direction value is calculated from each direction value of the direction data, where a cosine of the direction value is assigned to a real part of the complex direction value and a sine of the direction value is assigned to an imaginary part of the complex direction value, v) the real parts of the complex direction values ​​of the direction data and the imaginary parts of the complex direction values ​​of the direction data are filtered separately from each other using a Savitzky-Golay filter, vi) a real direction value of the direction data is calculated from each complex direction value of the direction data, where the real direction value is calculated as an arctangent of a ratio of the imaginary part and the real part of the complex direction value, vii) the real direction values ​​of the direction data are converted from radians to degrees.

9. Method (1) according to claim 8, characterized by that the direction values ​​of the direction data calculated in measure i) are in a range between 180° and -180°, and that after measure i) and before measure ii) the calculated direction values ​​of the direction data are transformed from the range between 180° and -180° into a range between 0° and 360°.

10. Identification system for locating an object, wherein the identification system is designed to perform the method (1) according to any of the preceding claims.

Images