Method and apparatus for calibrating vehicle sensors

Abstract

The invention relates to a method for calibrating a vehicle sensor (5) of a motor vehicle, the method comprising the following steps: determining (S1) sensor data by means of the vehicle sensor (5) at a plurality of measuring times, wherein the motor vehicle moves relative to an object in the surroundings of the motor vehicle; calculating (S2) an object position of the object based on the determined sensor data; calculating (S3) a Hough transform based on the calculated object position; determining (S4) an orientation of the vehicle sensor (5) relative to a travel axis (A1) of the motor vehicle based on the calculated Hough transform; and calibrating (S5) the vehicle sensor (5) based on the determined orientation of the vehicle sensor (5) relative to the travel axis (A1) of the motor vehicle.

Description

Technical Field

[0001] The present invention relates to a method for calibrating a vehicle sensor of a motor vehicle. The present invention also relates to a corresponding device for calibrating a vehicle sensor of a motor vehicle. Background Art

[0002] Driver assistance systems access the sensor data of the vehicle's sensors. In order for driver assistance systems to be usable right from the factory, the vehicle's sensors must be calibrated. For calibration, the vehicle can be moved on the chassis (Fahrwerkstand) towards a suitable target while the driving axis is measured. However, static measurement of the driving axis and calibration based on this can be time-consuming and expensive. In a production environment in particular, determining the orientation of the sensor relative to the driving axis of the vehicle can be problematic because short parking times are expected. Therefore, the cost of static methods increases linearly with each additional sensor, which can be difficult for multi-sensor systems consisting of many sensors. In addition, tracking (tracing) objects is very expensive. An "object" is understood to be any point to which coordinates can be assigned based on sensor data - for example, radar reflections, etc.

[0003] In the case of dynamic calibration of sensors by trilateration on fixed objects, it can be difficult to monitor the positions of the objects relative to each other to ensure that the objects are not moving.

[0004] DE 10 2013 209 494 A1 discloses a method for determining the dejustment of a radar sensor, in which a measured angle is compared with an angle calculated from a distance measurement.

[0005] Furthermore, EP 1 947 473 A2 discloses a method for calibrating a distance sensor and a measuring section in which road markings are arranged on the roadway. Summary of the Invention

[0006] The present invention provides a method for calibrating a vehicle sensor of a motor vehicle and a device for calibrating a vehicle sensor of a motor vehicle.

[0007] According to a first aspect, the present invention relates to a method for calibrating a vehicle sensor of a motor vehicle. The vehicle sensor acquires sensor data at a plurality of measurement times, wherein the motor vehicle moves relative to an object in its surroundings. The object position is calculated based on the acquired sensor data. A Hough transform is calculated based on the calculated object position. The orientation of the vehicle sensor relative to the vehicle's axis of travel is determined based on the calculated Hough transform. The vehicle sensor is calibrated based on the determined orientation of the vehicle sensor relative to the vehicle's axis of travel.

[0008] According to a second aspect, the present invention relates to a device for calibrating vehicle sensors of a motor vehicle, comprising an interface, a computing device, and a calibration device. The interface receives sensor data that has been ascertained by the vehicle sensors at multiple measurement points in time while the vehicle is moving relative to an object in its surroundings. The computing device calculates the object position based on the ascertained sensor data. The computing device calculates a Hough transform based on the calculated object position. The computing device also calculates the orientation of the vehicle sensor relative to the vehicle's axis of travel based on the calculated Hough transform. The calibration device calibrates the vehicle sensor based on the ascertained orientation of the vehicle sensor relative to the vehicle's axis of travel.

[0009] Advantages of the present invention

[0010] The present invention enables dynamic calibration, meaning that the calibration process can be performed while driving, without downtime or adjustments to the axis of travel. Thus, for example, a drive from a factory to a parking lot or to a transporter can be used for calibration. The present invention can thus be used to optimize time flows within a factory.

[0011] The calibration of the vehicle sensor according to the present invention can also be performed without measuring (radial) velocity, yaw rate or other signals. Object tracking is also not required. Moreover, the method is more accurate than static methods, especially in the elevation direction.

[0012] Finally, the method is easier to implement. Although the results are good, the method does not rely on a fixed object, which reduces costs because the object does not have to be continuously monitored with respect to its installation position.

[0013] Furthermore, it is possible to correct for static angular deviations caused by surfaces in front of the radar (e.g., a cover) through which the radar emits its radiation. Furthermore, it is possible to correct for the constant magnitude of refraction effects caused by the front panel.

[0014] According to another embodiment of the method for calibrating a vehicle sensor, the vehicle sensor comprises at least one of a radar sensor, a lidar sensor, an ultrasonic sensor, an infrared sensor, and a camera sensor. The present invention can be used for all environment scanning sensors without requiring additional external information.

[0015] According to another embodiment of the method for calibrating a vehicle sensor, the speed and yaw rate of the vehicle are also determined in order to perform post-filtering of the orientation of the vehicle sensor relative to the driving axis of the motor vehicle, i.e. with respect to the orientation error angle in the azimuth and elevation directions.

[0016] According to a further embodiment of the method for calibrating a vehicle sensor, the calculation of the Hough transform includes generating a sinusoidal diagram.

[0017] According to another embodiment of the method for calibrating vehicle sensors, object positions are weighted when generating the sinusoidal diagram. Weighting can lead to better results because quality characteristics of the object positions associated with the objects can be taken into account. For weighting, device information of the vehicle sensors, in particular the sensor range or the angular or distance dependence of the quality of the sensor data ascertained by the vehicle sensors, can be taken into account. The more reliable this data, the higher the weighting of the corresponding information.

[0018] According to another embodiment of the method for calibrating a vehicle sensor, a main direction of the object position is determined in Hough space based on a sinusoidal diagram by an intensity maximum estimation method (for example, by kernel density estimation), wherein determining the orientation of the vehicle sensor relative to the axis of travel of the motor vehicle includes calculating an angle between the main direction and the axis of travel of the motor vehicle.

[0019] According to a further embodiment of the method for calibrating a vehicle sensor, an azimuth angle and an elevation angle are calculated, which extend between the respective main direction and the axis of travel of the motor vehicle.

[0020] According to another embodiment of the method for calibrating a vehicle sensor, a sinusoidal diagram includes intensity values ​​ascertained from the sensor data, wherein a histogram is calculated by applying a threshold function to the intensity values. The histogram depends on the azimuth and / or elevation angle relative to the vehicle's axis of travel. The orientation of the vehicle sensor relative to the vehicle's axis of travel is ascertained based on the calculated histogram.

[0021] According to a further embodiment of the method for calibrating a vehicle sensor, the orientation of the vehicle sensor relative to the axis of travel of the motor vehicle is ascertained by applying a filter function to the histogram.

[0022] According to another embodiment of the method for calibrating a vehicle sensor, a main direction of the object position is determined by applying a filter function to the histogram. Determining the orientation of the vehicle sensor relative to the axis of travel of the motor vehicle includes calculating the angle between the main direction and the axis of travel of the motor vehicle.

[0023] According to another embodiment of the method for calibrating vehicle sensors, the Hough transform is calculated in spherical coordinates. This calculation is preferably performed without previously calculating a grid. Using spherical coordinates has the advantage that the position of the object is usually already present in spherical coordinates, allowing it to be used without transforming to a Cartesian coordinate system. This improves performance. The representation in spherical coordinates allows for higher accuracy, comparable to that of the Radon transform. However, in resource-constrained situations, this representation offers significantly higher performance because, in contrast to the Cartesian representation of the Hough transform, the object position is not reduced to the resolution of the grid size in Cartesian space, but can be described continuously.

[0024] According to a further embodiment of the method for calibrating a vehicle sensor, an error in the orientation of the vehicle sensor relative to the axis of travel of the motor vehicle can be calculated by means of an error propagation method.

[0025] According to another embodiment of the method for calibrating a vehicle sensor, the calibration of the vehicle sensor can be used continuously during driving. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] The accompanying drawings show:

[0027] Figure 1 A schematic block diagram shows a device for calibrating a vehicle sensor of a motor vehicle according to one embodiment of the present invention;

[0028] Figure 2 showing a grid with object positions for determining azimuthal orientation;

[0029] Figure 3 showing a grid with object positions for determining elevation orientation;

[0030] Figure 4 An exemplary sinusoidal diagram is shown;

[0031] Figure 5 An exemplary histogram is shown;

[0032] Figure 6 A flow chart of a method for calibrating a vehicle sensor of a motor vehicle according to one specific embodiment of the present invention is shown. DETAILED DESCRIPTION

[0033] Figure 1A schematic block diagram shows a device 1 for calibrating a vehicle sensor 5 of a motor vehicle. Device 1 includes an interface 2 coupled to the vehicle sensor 5 to receive sensor data from the vehicle sensor. The sensor data is generated based on measurements taken by the vehicle sensor 5 at multiple measurement times. For example, the vehicle sensor 5 may perform measurements at predetermined time intervals. The vehicle sensor 5 may be an environmental detection sensor, such as a radar sensor, a lidar sensor, an ultrasonic sensor, an infrared sensor, or a camera sensor. During the measurement, the vehicle moves relative to an object in its surroundings. Preferably, this involves a linear relative motion at a constant speed. For example, the object is fixedly positioned, while the vehicle moves in a straight line on a flat surface at a constant speed. In this case, the driving axis corresponds to the longitudinal axis of the vehicle. However, in principle, the vehicle's acceleration or steering motion can also be read by the sensor and taken into account in the calculation.

[0034] Device 1 also includes a computing device 3 that is coupled to interface 2 and further evaluates the received sensor data. Computing device 3 can generate a grid with a predetermined grid size in the vehicle coordinate system of the motor vehicle. In another embodiment, for example, when performing calculations in spherical coordinates, the calculation of the grid can be omitted.

[0035] Based on the sensor data, the computing device 3 calculates the object position, which indicates the position of the object relative to the motor vehicle.

[0036] The calculation device 3 calculates the object's position. The object's position is detected at different times due to the vehicle's movement relative to the object, resulting in a "flow" of the object. When using a grid, each object is assigned to the grid multiple times, as the corresponding position of the object is determined for each measurement time. Object positions can also be referred to as "locations." Consequently, each object is assigned multiple locations.

[0037] The calculation device 3 calculates a Hough transformation for each partial measurement time period (eg, based on a grid with object positions for this partial measurement time period), ie, transforms the object positions into a dual space or Hough space. This generates a sinogram.

[0038] The main directions are determined by Hough transformation in the Cartesian coordinate system according to the following formula:

[0039] r_n=x·cos(θ_n)+y·sin(θ_n),

[0040] Here, r_n represents the ordinate of the Hough sine plot point, and θ_n represents the abscissa of the Hough sine plot point. Furthermore, x and y represent the Cartesian coordinates of the object's position.

[0041] When generating the sinogram, computing device 3 can weight the object positions classified in the grid. Weighting factors can be calculated based on the quality characteristics of the respective object positions. In particular, the weighting factors can be determined based on the sensor characteristics of vehicle sensor 5 and can depend, for example, on the distance of the object position from the position of vehicle sensor 5.

[0042] The Hough algorithm modified by weighting can have the following form:

[0043]

[0044]

[0045] Alternatively, the calculation device 3 may determine the main directions of the object positions classified into the grid from the sinogram by kernel density estimation.

[0046] Instead of calculating in Cartesian coordinates, the calculation device 3 can also calculate the Hough transform in spherical coordinates, that is, according to the following formula:

[0047] r_n(ξ_n)=d_R·sin(ξ_R-ξ_n),

[0048] Where ξ_R represents and θ_R. The ordinate and abscissa of the Hough map (r_n, ξ_n) can be calculated from the radial distance d_R and the angle θ_R (azimuth or elevation) of the object position. Based on the calculated Hough map coordinates (r_n, ξ_n), the weights of the associated object positions (d_R, θ_R) are added to the Hough sinusoidal map. Quantization or partitioning of r_n and ξ_n can be performed with any degree of precision, which determines the resolution of the sinusoidal map.

[0049] The sinogram includes intensity values, to which computing device 3 applies a threshold function to calculate a histogram. The histogram depends on the possible azimuth or elevation angles of vehicle sensor 5 relative to the vehicle's axis of travel. Computing device 3 applies a filter function to the histogram to determine the main direction of the object positions classified in the grid. The actual azimuth or elevation angle of vehicle sensor 5 corresponds to the angle between the main direction and the vehicle's axis of travel.

[0050] The device 1 further comprises a calibration device 4, which is coupled to the computing device 3 and calibrates the vehicle sensor 5 based on the orientation of the vehicle sensor 5 ascertained by the computing device 3. For example, the actual orientation of the vehicle sensor 5 can be taken into account in all measured values ​​of the vehicle sensor 5. Provision can also be made to calculate an orientation error—that is, a deviation of the calculated orientation of the vehicle sensor 5 from a desired orientation.

[0051] For ease of explanation, Figure 2A grid with object positions for determining azimuth angles is shown. Each marked measurement point in the grid corresponds to a measurement signal, which is assigned by computing device 3 based on sensor data to the spatial region corresponding to the grid point in the vehicle's surroundings. For each object, the measurement points or object positions lie along mutually parallel main directions. As an example, the main direction A2 of one of the objects is marked. The object positions along this main direction A2 correspond to measurements of the object at different measurement times. The vehicle's travel axis A1 is also marked. Figure 2 This corresponds to a view from above the vehicle, so that the azimuth angle β_AZ lies between the main direction A2 and the axis of travel A1 .

[0052] Figure 3 A grid with object positions for determining elevation orientation is shown. Thus, Figure 3 Corresponding to the side view, the elevation angle β_EL is therefore between the main direction A3 and the axis of travel A1 .

[0053] Figure 4 The sinogram obtained by Hough transformation for a single object is shown. This generates multiple essentially sinusoidal curves that intersect at the point B of maximum intensity. For multiple objects, additional parallel, shifted curves are generated that also intersect at the point of maximum intensity for each object. The sinogram shows the projection position P as a function of the angle α (e.g., elevation or azimuth). Angles α with higher intensities correspond to a greater number of measurement points lying on the associated straight line.

[0054] Figure 5 An exemplary histogram is shown, which is calculated using a threshold function for the intensity values. For each angle α, the intensities at all projection positions P (i.e., r_n) are added together, with only points whose intensities exceed a predetermined threshold being considered. The corresponding histogram yields the maximum value A of the accumulated intensity for a specific angle α. This angle corresponds to the corresponding angle between the main directions A2, A3 and the driving axis A1 of the motor vehicle, i.e., the actual elevation angle or azimuth angle, for example. The accumulated intensity or total intensity corresponds to the total weight of the associated angle α and can be calculated using a filter function. The angle can be calculated using angle estimation methods (in particular, Kalman filters, averaging methods, etc.).

[0055] Figure 6 A flow chart of a method for calibrating vehicle sensor 5 of a motor vehicle according to one specific embodiment of the present invention is shown.

[0056] In a first method step S1 , vehicle sensors 5 transmit sensor data acquired at a plurality of measuring times during a movement of the motor vehicle relative to an object in the motor vehicle's surroundings. The motor vehicle and the object may in particular move linearly relative to one another at a constant speed.

[0057] In a second method step S2, the object positions are assigned to a grid in the vehicle coordinate system of the motor vehicle based on the ascertained sensor data. Each object is assigned multiple times, i.e., the corresponding coordinates ("locations") are entered into the grid for each measurement time. For other calculations, such as those using spherical coordinates, the generation of a grid can be omitted.

[0058] In a third method step S3, a Hough transform is calculated from the grid with the object positions. For this purpose, a sinogram can be calculated, wherein the object positions can be weighted. The Hough transform can be calculated in Cartesian coordinates or in spherical coordinates.

[0059] In a fourth method step S4, the orientation of vehicle sensor 5 relative to the vehicle's axis of travel A1 is calculated based on the calculated Hough transform. For this purpose, a kernel density estimation can be performed to determine the main direction. Alternatively, a histogram can be calculated by applying a threshold function to the sinusoidal intensity values, which histogram depends on the azimuth and / or elevation angle relative to the vehicle's axis of travel. The orientation of vehicle sensor 5 relative to the vehicle's axis of travel is determined based on the calculated histogram. A filter function can be applied to the histogram.

[0060] In a fifth method step S5 , vehicle sensor 5 is calibrated as a function of the ascertained orientation of vehicle sensor 5 relative to driving axis A1 of the motor vehicle.

Claims

1. A method for calibrating a vehicle sensor (5) of a motor vehicle, the method comprising the following steps: Sensor data are determined (S1) by the vehicle sensor (5) at a plurality of measuring times, wherein: The motor vehicle moves relative to objects in the surroundings of the motor vehicle; Calculating (S2) the object position of the object based on the determined sensor data; calculating (S3) a Hough transform based on the calculated object position; determining (S4) the orientation of the vehicle sensor (5) relative to the driving axis (A1) of the motor vehicle based on the calculated Hough transform, wherein calculating the Hough transform includes: generating a sinusoidal graph, wherein the sinusoidal graph includes intensity values ​​determined based on the sensor data, wherein a histogram is calculated for the intensity values ​​without the orientation and yaw rate of the driving axis of the motor vehicle, wherein a threshold function is applied to the intensity values, the histogram being dependent on possible azimuth angles and / or elevation angles relative to the driving axis (A1) of the motor vehicle, wherein the orientation of the vehicle sensor (5) relative to the driving axis (A1) of the motor vehicle is determined based on the calculated histogram; The vehicle sensor (5) is calibrated (S5) as a function of the ascertained orientation of the vehicle sensor (5) relative to the axis of travel (A1) of the motor vehicle.

2. The method according to claim 1, wherein The motor vehicle and objects in the surroundings of the motor vehicle execute a substantially linear relative movement at a constant speed.

3. A method according to any one of the preceding claims, wherein Weighting is performed on the object positions when generating the sinogram.

4. The method according to any one of claims 1 to 3, wherein Based on the sinusoidal diagram, the main direction (A2, A3) of the object position is determined by determining the intensity maximum, wherein determining the orientation of the vehicle sensor (5) relative to the driving axis (A1) of the motor vehicle includes calculating the angle between the main direction (A2, A3) and the driving axis (A1) of the motor vehicle.

5. The method according to any one of claims 1 to 4, wherein The orientation of the vehicle sensor (5) relative to the driving axis (A1) of the motor vehicle is determined by applying a filter function to the histogram.

6. The method according to claim 5, wherein: The main direction (A2, A3) of the object position is determined by applying the filter function to the histogram, wherein determining the orientation of the vehicle sensor (5) relative to the driving axis (A1) of the motor vehicle includes calculating the angle between the main direction (A2, A3) and the driving axis (A1) of the motor vehicle.

7. A method according to any one of the preceding claims, wherein: The calculation of the Hough transform is performed in spherical coordinates.

8. A device (1) for calibrating a vehicle sensor (5) of a motor vehicle, the device being configured to carry out the method according to any one of claims 1 to 7, the device comprising: An interface (2) for receiving sensor data, the sensor data being determined by the vehicle sensor (5) at a plurality of measurement times, wherein: The motor vehicle moves relative to objects in the surroundings of the motor vehicle; A calculation device (3) configured to: calculate an object position of the object based on the ascertained sensor data; calculate a Hough transform based on the calculated object position; and calculate an orientation of the vehicle sensor (5) relative to a travel axis (A1) of the motor vehicle based on the calculated Hough transform; A calibration device (4) is designed to calibrate the vehicle sensor (5) as a function of the ascertained orientation of the vehicle sensor (5) relative to a travel axis (A1) of the motor vehicle.

Images