DE102019008369B4 - Method for calibrating a distance measuring sensor of a vehicle - Google Patents

Abstract

Method for calibrating a distance-measuring sensor of a vehicle, characterized in that a transformation parameter value of at least one transformation parameter for a coordinate transformation is determined, with which a measurement position (MP) determined in a sensor coordinate system is transformed into a world coordinate system (WKS), the transformation parameter value of the at least one Transformation parameter is determined by an optimization method in which the measurement positions (MP), which were determined with the sensor at different times during the vehicle's own movement, are evaluated, the optimization method comprising the following steps: a) Transformation of the measurement positions (MP) with a predetermined one Transformation parameter value of the at least one transformation parameter in a coordinate grid (KG) of the world coordinate system (WKS),b) counting of cells (Z) of the coordinate grid (KG) to which at least one of the transformi first measurement positions (TMP),c) repeating steps a) and b) with a respectively modified transformation parameter value of the at least one transformation parameter and determining the transformation parameter value of the at least one transformation parameter, which in step b) leads to a minimum.

Description

The invention relates to a method for calibrating a distance-measuring sensor of a vehicle.

From the prior art, as in the DE 10 2006 000 642 A1 described, a driver assistance system known. The driver assistance system includes a large number of sensors to which calibration parameters are assigned, and storage means for the central storage of the calibration parameters.

From the EP 3 438 777 A1 a method for calibrating a camera and a lidar sensor is known, which is based on the determination of edges from sensor data from the camera and the lidar sensor and on minimizing the mismatch between the edges from the sensor data from the camera and the edges from the sensor data from the lidar sensor.

From the WO 2019/154536 A1 another method for calibrating a lidar sensor with a camera is known, which is based on an extraction of edges from the sensor data of the lidar sensor and the camera and on a determination of similarities with regard to the edges.

From the DE 10 2019 100 497 A1 a method for calibrating a lidar sensor of a vehicle is known, in which an object placed at a predetermined position in the field of view of the lidar sensor is scanned by the lidar sensor and in which a reflectivity of the object is determined from the lidar scans.
The invention is based on the object of specifying a method for calibrating a distance-measuring sensor of a vehicle that is improved over the prior art.

The object is achieved according to the invention by a method for calibrating a distance-measuring sensor of a vehicle with the features of claim 1.

Advantageous configurations of the invention are the subject matter of the dependent claims.

1. a) Transformation der Messpositionen mit einem vorgegebenen Transformationsparameterwert des mindestens einen Transformationsparameters in ein Koordinatengitter des Weltkoordinatensystems,
2. b) Zählen von Zellen des Koordinatengitters, denen zumindest eine der transformierten Messpositionen zugeordnet ist,
3. c) Wiederholen der Schritte a) und b) mit einem jeweils, d. h. bei jeder Wiederholung, modifizierten Transformationsparameterwert des mindestens einen Transformationsparameters und Ermitteln des Transformationsparameterwertes des mindestens einen Transformationsparameters, der in Schritt b) zu einem Minimum führt, d. h. zu einem Minimum einer Anzahl der Zellen des Koordinatengitters, denen zumindest eine der transformierten Messpositionen zugeordnet ist.

In a method for calibrating a distance-measuring sensor of a vehicle, in particular with regard to the vehicle's own movement, a transformation parameter value of at least one transformation parameter for a coordinate transformation is determined, with which a measurement position determined in a sensor coordinate system is transformed into a world coordinate system. The transformation parameter value of the at least one transformation parameter is determined by an optimization method in which measurement positions that were determined with the sensor at different points in time during the vehicle's own movement are evaluated. The optimization process includes the following steps:

1. a) transformation of the measurement positions with a predetermined transformation parameter value of the at least one transformation parameter into a coordinate grid of the world coordinate system,
2. b) counting cells of the coordinate grid to which at least one of the transformed measurement positions is assigned,
3. c) Repeating steps a) and b) with a modified transformation parameter value of the at least one transformation parameter, ie with each repetition, and determining the transformation parameter value of the at least one transformation parameter, which in step b) leads to a minimum, ie to a minimum of a number of the cells of the coordinate grid to which at least one of the transformed measurement positions is assigned.

The basic idea of the method is therefore to represent an environment of the vehicle detected by the sensor with as few cells as possible of the coordinate grid, which is formed from these cells, each with a finite extent. For this purpose, point measurements of the sensor in the world coordinate system, i. H. into a world-fixed coordinate system, which is quantized in cells of a given size.

A respective object point of a respectively detected static object, i.e. a respective world-fixed object point, has unchanging world coordinates and will therefore always appear at the same position in the world coordinate system for measurements at different times during the vehicle’s own movement with a calibrated sensor, i.e. only one for all measurements form the transformed measurement position, and thus occupy only one cell of the coordinate grid. With a decalibrated sensor, however, the same world-fixed object point will appear at different positions in the world coordinate system for measurements at different times during the vehicle's own movement, i.e. different transformed measurement positions are determined at different points in the world coordinate system for the same world-fixed object point of the static object, which means that the same world-fixed object point of the static object, due to the resulting multiple transformed measurement positions, multiple cells of the coordinate grid are occupied. With a large number using the Sensor detected world fixed object points, there is a significant deviation in the number of cells occupied with transformed measurement positions with a calibrated sensor, in which only a single transformed measurement position is determined for the respective world fixed object point, compared to the non-calibrated sensor, in which several for the same world fixed object point transformed measurement positions are determined.

The aim of the optimization process is therefore to reduce the number of required, i. H. cells occupied with at least one transformed measurement position by adapting the transformation parameter value of the at least one transformation parameter, advantageously by adapting a respective transformation parameter value of a plurality of transformation parameters. The number of these required, i. H. cells occupied with at least one transformed measurement position is minimal with optimal calibration of the sensor, since the same transformed measurement position in the world coordinate system is then recorded in all measurements for the respectively recorded world-fixed object point. The number of transformed measurement positions is thus reduced to the number of detected object points, as a result of which the number of transformed measurement positions and thus the number of cells occupied by the transformed measurement positions is minimal.

Using the method according to the invention, calibration is possible without additional reference sensors. In addition, there is no need to find correspondence between measurement positions, as is the case in other possible calibration methods. The method according to the invention works in particular with any surroundings of the vehicle. No preconditions for scenarios to be detected by the sensor have to be met for the method to work. Appropriately, only at least one stationary object needs to be detectable. However, this does not have to be a specific test object, but any stationary object is sufficient, for example a building, a curb, a crash barrier, a delineator post or another structure, for example a piece of furniture in a building or an interior design of the building , for example a production hall in which the vehicle is manufactured.

The method according to the invention also works in scenarios with dynamic objects. Since the object points in dynamic objects are not world-fixed, these dynamic objects generate a different measurement position at different points in time for each object point, since the same object point is at different locations due to the dynamic movement of the object at different points in time, and thus also different transformed positions measurement positions. Since the respective object point of such a dynamic object is therefore not at the same location at different times, there will be no multiple transformed measurement positions for the same object point at the same location resulting from the decalibration of the sensor, which would be reduced to one transformed measurement position by the calibration would. i.e. for dynamic objects, the number of transformed measurement positions and thus also the number of cells occupied by these transformed measurement positions remains the same after the calibration as before the calibration, it does not change during the optimization process and therefore has no influence on this optimization minimizing occupied cells. This minimization of the occupied cells is done exclusively by reducing the transformed measurement positions of the recorded static objects.

The method according to the invention is also robust to distance noise of individual point measurements due to the quantization of the world coordinate system in cells.

Furthermore, no initial calibration of the sensor is necessary using the method according to the invention.

For example, a transformation parameter value is determined for at least one transformation parameter designed as an extrinsic sensor parameter of the sensor and/or for a transformation parameter designed as a time offset between a sensor-internal clock of the sensor and a clock of the vehicle's own motion module. The at least one transformation parameter designed as an extrinsic sensor parameter is, for example, an x-coordinate, a y-coordinate or a z-coordinate of an installation position of the sensor in a vehicle coordinate system or a roll angle, a pitch angle or a yaw angle of the sensor with respect to the vehicle coordinate system.

1. a) Transformation der Messpositionen mit einem vorgegebenen Transformationsparameterwert des jeweiligen Transformationsparameters in das Koordinatengitter des Weltkoordinatensystems,
2. b) Zählen der Zellen des Koordinatengitters, denen zumindest eine der transformierten Messpositionen zugeordnet ist,
3. c) Wiederholen der Schritte a) und b) mit jeweils, d. h. bei jeder Wiederholung, modifizierten Transformationsparameterwerten der Transformationsparameter und Ermitteln der Transformationsparameterwerte der Transformationsparameter, d. h. des Transformationsparameterwertes des jeweiligen Transformationsparameters, die in Schritt b) zu einem Minimum führen, d. h. zu einem Minimum der Anzahl der Zellen des Koordinatengitters, denen zumindest eine der transformierten Messpositionen zugeordnet ist. Dabei kann beispielsweise vorgesehen sein, dass bei jeder Wiederholung der Schritte a) und b) der Transformationsparameterwert nur jeweils eines Transformationsparameters oder der jeweilige Transformationsparameterwert mehrerer oder aller Transformationsparameter gleichzeitig modifiziert wird. Das Verfahren ermöglicht somit vorteilhafterweise eine zeitliche Kalibrierung, durch die Uhrzeitsynchronisation zwischen dem Sensor und dem Eigenbewegungsmodul, d. h. durch die Ermittlung des zeitlichen Versatzes zwischen der sensorinternen Uhr des Sensors und der Uhr des Eigenbewegungsmoduls des Fahrzeugs, und gleichzeitig eine räumliche Kalibrierung, vorteilhafterweise in sechs Freiheitsgraden, d. h. bezüglich der x-Achse, y-Achse und z-Achse des Fahrzeugkoordinatensystems und bezüglich des Rollwinkels, eines Nickwinkels und Gierwinkels zum Fahrzeugkoordinatensystem.

Advantageously, transformation parameter values of several transformation parameters, for example all extrinsic sensor parameters and, for example, the time offset between the sensor's internal clock and the clock of the vehicle's own movement module, are determined for the coordinate transformation, with which a measurement position determined in the sensor coordinate system is transformed into the world coordinate system. i.e. a transformation parameter value is determined for the respective transformation parameter. The transformation parameter values of the transformation parameters are determined by the optimization process, in which measurement positions that were determined with the sensor at different points in time during the vehicle's own movement are evaluated. The optimization process includes the following steps:

1. a) Transformation of the measurement positions with a specified transformation parameter value of the respective transformation parameter into the coordinate grid of the world coordinate system,
2. b) counting the cells of the coordinate grid to which at least one of the transformed measurement positions is assigned,
3. c) Repeating steps a) and b) with modified transformation parameter values of the transformation parameters, ie with each repetition, and determining the transformation parameter values of the transformation parameters, ie the transformation parameter value of the respective transformation parameter, which in step b) lead to a minimum, ie to a minimum the number of cells in the coordinate grid to which at least one of the transformed measurement positions is assigned. It can be provided, for example, that with each repetition of steps a) and b), the transformation parameter value of only one transformation parameter or the respective transformation parameter value of several or all transformation parameters is modified simultaneously. The method thus advantageously enables a time calibration, through the time synchronization between the sensor and the self-motion module, ie by determining the time offset between the sensor-internal clock of the sensor and the clock of the self-motion module of the vehicle, and at the same time a spatial calibration, advantageously in six degrees of freedom , ie with respect to the x-axis, y-axis and z-axis of the vehicle coordinate system and with respect to the roll angle, a pitch angle and yaw angle to the vehicle coordinate system.

For example, a lidar sensor, a radar sensor, an ultrasonic sensor or a stereo camera of the vehicle is calibrated as a sensor, i. H. in particular an environment detection sensor of the vehicle. Such surroundings detection sensors are particularly advantageous for carrying out one or more assistance functions of the vehicle and/or by carrying out an at least partially automated or fully automated or autonomous driving operation of the vehicle. For this it is important that the respective sensor is calibrated in order to deliver exact sensor results and thereby avoid possible malfunctions of the assistance function or the at least partially automated or fully automated or autonomous driving operation of the vehicle.

The vehicle's own movement is determined in particular by means of a vehicle's own movement module. The data on the vehicle's own movement, which is advantageously supplied by this own movement module, is required for the transformation from the sensor coordinate system into the world coordinate system, with this transformation expediently taking place via a vehicle coordinate system as an intermediate step. The transformation from the sensor coordinate system into the vehicle coordinate system takes place in particular by means of the extrinsic sensor parameters and the transformation from the vehicle coordinate system into the world coordinate system by means of the time synchronization between the sensor and the own motion module, i. H. by means of the determined time offset between the sensor's internal clock and the clock of the vehicle's own movement module.

The calibration using this method is carried out, for example, at the end of a production line for the vehicle. This is also referred to as end-of-line calibration, which is possible using the method in a particularly quick and simple manner and without any additional effort. Instead of measuring the vehicle in a complex manner, it is advantageously sufficient to drive the vehicle a few meters. As a result, a process chain can be significantly accelerated. End-of-line calibration is one of the most time-consuming steps in vehicle construction.

Alternatively or additionally, the calibration is carried out using this method, for example during regular driving operation of the vehicle. The method can be used here, for example, to continuously carry out the calibration of the sensor while the vehicle is driving and/or to detect whether the sensor is decalibrated, for example due to a mechanical defect.

Alternatively or additionally, the calibration is carried out using this method, for example during a data determination to create at least one high-resolution digital map. This allows highly accurate object detection and object position detection using the sensor calibrated in this way, thereby improving the accuracy of the high-resolution digital map. The method can be applied to recorded sensor detection sequences in order to obtain a highly precise calibration.

This calibration using recorded sequences of sensor detection can also be used, for example, to serve as a reference for an online calibration. For example, it can be provided that at least one calibration is carried out, in particular as an online calibration, in which measurement positions that were determined with the sensor at different times within a short period of time during the vehicle's own movement are evaluated, and also, in particular as Reference for this online calibration and thus for its verification, at least one calibration is carried out, are evaluated in the measurement positions that were determined with the sensor at different times within a longer period of time during the vehicle's own movement. The longer period of time advantageously includes the points in time of the shorter period of time and additional points in time at which measurement positions were determined with the sensor. In this way, the online calibration can be carried out quickly and in quick succession. A longer period of time and thus also a longer computing time is then available for checking them in order to evaluate the correspondingly higher number of measurement positions.

1. a) Transformation der Messpositionen mit vorgegebenen Transformationsparametern, insbesondere vorgegebenen Transformationsparameterwerten der Transformationsparameter, in ein Koordinatengitter des Weltkoordinatensystems,
2. b) Zählen der Zellen des Koordinatengitters, denen zumindest eine der transformierten Messpositionen zugeordnet ist,
3. c) Wiederholen der Schritte a) und b) mit modifizierten Transformationsparametern, insbesondere modifizierten Transformationsparameterwerten der Transformationsparameter, und Ermitteln der Transformationsparameter, insbesondere deren Transformationsparameterwerte, die in Schritt b) zu einem Minimum führen.

In summary, a method for calibrating the distance-measuring sensor of the vehicle, in particular with regard to the vehicle's own movement, is therefore advantageously carried out, with calibration meaning that transformation parameters, in particular transformation parameter values of the transformation parameters, are determined for the coordinate transformation, with which a measurement position determined in the sensor coordinate system is converted into the World coordinate system is transformed. The transformation parameters, in particular their transformation parameter values, are determined using an optimization method in which measurement positions that were determined with the sensor at different points in time are evaluated. The optimization method is advantageously based on the following steps:

1. a) Transformation of the measurement positions with specified transformation parameters, in particular specified transformation parameter values of the transformation parameters, into a coordinate grid of the world coordinate system,
2. b) counting the cells of the coordinate grid to which at least one of the transformed measurement positions is assigned,
3. c) repeating steps a) and b) with modified transformation parameters, in particular modified transformation parameter values of the transformation parameters, and determining the transformation parameters, in particular their transformation parameter values, which lead to a minimum in step b).

Exemplary embodiments of the invention are explained in more detail below with reference to drawings.

• 1 schematisch mehrere zeitlich versetzte Messungen eines dekalibrierten als Lidarsensor ausgebildeten Sensors, aufakkumuliert in einem Weltkoordinatensystem,
• 2 schematisch eine, eine Anzahl belegter Zellen angebende, Fehlerfunktion in Abhängigkeit von einem als Gierwinkel ausgebildeten Transformationsparameter,
• 3 schematisch eine, eine Anzahl belegter Zellen angebende, Fehlerfunktion in Abhängigkeit von einem als Rollwinkel ausgebildeten Transformationsparameter,
• 4 schematisch die zeitlich versetzten und im Weltkoordinatensystem aufakkumulierten Messungen aus 1 mit kalibriertem als Lidarsensor ausgebildetem Sensor, und
• 5 schematisch eine Ausschnittvergrößerung aus 4.

show:

• 1 Schematic diagram of several temporally offset measurements of a decalibrated sensor designed as a lidar sensor, accumulated in a world coordinate system,
• 2 schematically an error function indicating a number of occupied cells as a function of a transformation parameter designed as a yaw angle,
• 3 schematically an error function indicating a number of occupied cells as a function of a transformation parameter designed as a roll angle,
• 4 schematically shows the measurements that are offset in time and accumulated in the world coordinate system 1 with a calibrated sensor designed as a lidar sensor, and
• 5 schematically shows a detail enlargement 4 .

Corresponding parts are provided with the same reference symbols in all figures.

Based on 1 until 5 a method for calibrating a distance-measuring sensor of a vehicle, in particular with regard to a movement of the vehicle itself, is explained in more detail below. Such a distance-measuring sensor, for example a lidar sensor, ultrasonic sensor, radar sensor or a camera, in particular a stereo camera, is in particular an environment detection sensor of the vehicle.

For an autonomous vehicle in particular, a precise and consistent mapping of the vehicle's surroundings, also referred to as a representation of the surroundings, in a selected coordinate system is a necessary criterion for safe driving. A precise and consistent representation of the environment requires an extrinsic calibration of the sensors, in particular the distance-measuring sensors, of the vehicle to one another or to the vehicle, as well as a temporal calibration, i. H. a synchronization of a sensor-internal clock.

In the method described here, both calibrations, ie the extrinsic and temporal calibration, are advantageously carried out in a single measurement step, ie in a single method process, carried out, advantageously for all distance-measuring sensors of the vehicle. Furthermore, the method is so general and robust that it can be used in almost any vehicle environment without additional measurement and calibration technology. The minimum requirement is, for example, the presence of at least one static object in the area surrounding the vehicle that can be detected by the sensor. This should be the case in practically every conceivable situation in which the vehicle can find itself, in particular when it is moving itself.

Calibrating extrinsically means that an extrinsic transformation T _ext from its sensor coordinate system to a predefined vehicle coordinate system is determined for the respective sensor. The extrinsic transformation text consists of an xyz translation, ie the installation position of the sensor on the vehicle and thus an x-coordinate, y-coordinate and z-coordinate of the installation position of the sensor in the vehicle coordinate system, and an orientation, ie a roll angle R, pitch angle and yaw angle G, also known as yaw, pitch and roll angle together.

The vehicle has a module that provides own motion data in the vehicle coordinate system. i.e. the vehicle's own movement is determined by means of this module of the vehicle, which is also referred to as the vehicle's own movement module, and provides corresponding data of this own movement.

Sensor data of the sensor can thus be transformed into the vehicle coordinate system using the extrinsic transformation text and into a world-fixed coordinate system, hereinafter referred to as world coordinate system WKS, using the own motion data. It must be ensured that, in addition to the extrinsic calibration, the sensor-internal clocks are also adjusted at the same time, i. H. in particular the sensor's internal clock with a clock of the vehicle's own motion module. Without time-aligned clocks, a uniform representation of the environment cannot be created. i.e. it cannot be determined when an object was in which position.

In previous methods, the vehicle is calibrated using reference sensors, which is time-consuming. However, this is very time-consuming, error-prone and not online-enabled, i. H. In particular, it cannot be carried out during regular driving operation of the vehicle. More advanced methods calculate a lidar odometry based on features. Features are calculated for individual points, correspondences between point clouds are sought and the distance between the correspondences is minimized using an error function. The calculated own movement based on the point cloud can now be compared with the own movement of the vehicle and the extrinsic transformation text can be determined from this. Problems with this, however, are that a clear assignment of correspondence points between two point clouds is required. The calculation of suitable correspondence features is also problematic. A significant problem, especially with regard to the simple feasibility and the online capability, is that correspondences to dynamic objects within a detected scene falsify the estimation of the own motion. In addition, the quality of the motion estimation is scene-dependent.

The basic idea for the method described here regarding the advantageous temporal and spatial calibration of a distance-measuring sensor is the assumption that static objects have their position in a world-fixed coordinate system, i. H. in the world coordinate system WCS, do not change. If an object point of such a static object, i. H. a world-fixed object point, measured in the world coordinate system WKS from a moving vehicle from different positions by means of the sensor, then the measured values of the sensor should always deliver the same result for the respective world-fixed object point. However, this requires that the sensor is calibrated relative to the vehicle's own movement, also known as ego-motion. If this is not the case, the sensor will provide different readings for the same fixed object point from different vehicle positions. This applies to all measured/detected world fixed object points.

In the method described here, the world-fixed coordinate system, i. H. the world coordinate system WKS, is quantized into a coordinate grid KG with cells Z of finite extent, for example 10 cm×10 cm×10 cm. For each measurement run with measurements by means of the sensor from different positions, i. H. at different points in time during the vehicle's own movement, those cells Z are counted in which the sensor delivers one or more measurement results. These occupied cells Z are also referred to as marker cells. A number A of these marker cells is therefore a measure of how well the sensor for the vehicle's own movement is calibrated.

If the parameters of the sensor's own movement calibration, ie three translations and three rotations, are now varied independently of one another, the number A of marker cells changes with each parameter change. With a bad calibration, the readings for the same land fixed object point from different positions of the sensor in different cells Z, since due to this poor calibration from different positions of the sensor in the course of the vehicle's own movement for the same fixed object point different in the world coordinate system WKS measuring positions TMP are determined, which are far apart, so that there is a high probability that they are in different cells Z. The number A of marker cells thus increases.

If the calibration is good, the number A of marker cells will decrease, since for the respective world-fixed object point, transformed measurement positions TMP that are close to one another are determined from all positions of the sensor in the course of the vehicle's own movement, which increases the probability that they are in the same cell Z . Accordingly, the optimal calibration is given when the number A of marker cells is minimal. This is particularly the case when the same transformed measurement position TMP is determined from all positions of the sensor in the course of the vehicle's own movement due to the good calibration for the respective world-fixed object point.

The smaller the size of the cells Z of the coordinate grid KG is specified, the better this optimization and thus the calibration can take place, since with very small cells Z the same transformed measurement position TMP or at least transformed measurement positions TMP that are very close to one another are determined for each world-fixed object point must, in order to reduce the number A of cells Z occupied by transformed measurement positions TMP, i. H. of the marker cells. However, as the cells Z become smaller, the calculation effort for this optimization method could also increase, so that the size of the cells Z is adapted, for example, to the required calibration accuracy and the available computing power and computing time.

The method can thus be used to calibrate all distance-measuring sensors, for example stereo cameras, lidar sensors, radar sensors and ultrasonic sensors, without additional tools and in almost any environment. An analogous procedure allows the synchronization of the sensor-internal clock and the clock of the self-motion module.
The procedure is described in detail below. The basic idea for the temporal and spatial calibration of distance-measuring sensors is the assumption that static objects do not change their position in the world coordinate system, ie in the world coordinate system WCS. If the correct extrinsic calibration of the sensor and thus the extrinsic transformation text are known and if the vehicle's own movement data T _ego (t _egomotion ) of the vehicle's own movement are available, then each measurement position MP recorded with the sensor at time t can be converted into a world-fixed coordinate system, ie into the world coordinate system WKS, are transferred: $${\text{TMP=T}}_{\text{ego}}\left({\text{t}}_{\text{egoism}}\right)\times {\text{T}}_{\text{external}}\times \text{MP}$$

If a world-fixed object point in the world coordinate system WKS is measured from different positions of the distance-measuring sensor, the world coordinate system WKS always delivers the same result. If the calibration or the time synchronization between the distance-measuring sensor and the vehicle's own motion module is not correct, a new transformed measuring position TMP is obtained in the world coordinate system WKS for each measuring position MP of the world-fixed object point detected from different sensor positions and thus several different ones for the respective world-fixed object point transformed measurement positions TMP, i. H. with differing world coordinates, in the world coordinate system WCS.

1. 1. Transformiere Messungen, insbesondere Messpositionen MP, von unterschiedlichen Zeitpunkten in das Weltkoordinatensystem WKS.
2. 2. Quantisiere das Weltkoordinatensystem WKS in Zellen Z, beispielsweise 10 cm x 10 cm x 10 cm.
3. 3. Zähle alle mit transformierten Messpositionen TMP belegten Zellen Z.
4. 4. Minimiere die Anzahl A der mit transformierten Messpositionen TMP belegten Zellen Z durch Modifizierung der extrinsischen Sensorparameter und des zeitlichen Versatzes zwischen der sensorinternen Uhr des Sensors und der Uhr des Eigenbewegungsmoduls des Fahrzeugs.

This change in the world coordinates for the same world-fixed object point and the resulting plurality of transformed measurement positions TMP in the world coordinate system WKS for the same world-fixed object point is used in the method described here to calibrate the distance-measuring sensor, in particular to calibrate the vehicle's own motion module. The procedure is as follows:

1. 1. Transform measurements, in particular measurement positions MP, from different points in time into the world coordinate system WKS.
2. 2. Quantize the world coordinate system WKS in cells Z, for example 10 cm x 10 cm x 10 cm.
3. 3. Count all cells Z occupied by transformed measurement positions TMP.
4. 4. Minimize the number A of cells Z occupied by transformed measurement positions TMP by modifying the extrinsic sensor parameters and the time offset between the sensor's internal clock and the clock of the vehicle's own motion module.

The optimum here is that existing objects are represented with as few cells Z as possible, ie several measurements of the same fixed object point by means of the sensor at different times during the vehicle's own movement fall into the same cell Z, they advantageously provide the same transformed measurement position TMP.

1. a) Transformation der Messpositionen MP mit einem vorgegebenen Transformationsparameterwert des mindestens einen Transformationsparameters in das Koordinatengitter KG des Weltkoordinatensystems WKS,
2. b) Zählen der Zellen Z des Koordinatengitters KG, denen zumindest eine der transformierten Messpositionen TMP zugeordnet ist,
3. c) Wiederholen der Schritte a) und b) mit einem jeweils, d. h. bei jeder Wiederholung, modifizierten Transformationsparameterwert des mindestens einen Transformationsparameters und Ermitteln des Transformationsparameterwertes des mindestens einen Transformationsparameters, der in Schritt b) zu einem Minimum führt, d. h. zu einem Minimum der Anzahl A der Zellen Z des Koordinatengitters KG, denen zumindest eine der transformierten Messpositionen TMP zugeordnet ist.

The method described here for calibrating the distance-measuring sensor of the vehicle, in particular with regard to the vehicle's own movement, advantageously results from this method specification, in which a transformation parameter value of at least one transformation parameter for a coordinate transformation is advantageously determined, with which a measurement position MP determined in the sensor coordinate system is converted into the world coordinate system WCS is transformed. The transformation parameter value of the at least one transformation parameter is determined by an optimization method in which measurement positions MP, which were determined with the sensor at different points in time during the vehicle's own movement, are evaluated. The optimization process includes the following steps:

1. a) Transformation of the measurement positions MP with a predetermined transformation parameter value of the at least one transformation parameter in the coordinate grid KG of the world coordinate system WKS,
2. b) counting the cells Z of the coordinate grid KG to which at least one of the transformed measurement positions TMP is assigned,
3. c) Repeating steps a) and b) with a modified transformation parameter value of the at least one transformation parameter, ie with each repetition, and determining the transformation parameter value of the at least one transformation parameter, which leads to a minimum in step b), ie to a minimum number A of the cells Z of the coordinate grid KG, to which at least one of the transformed measurement positions TMP is assigned.

For example, a transformation parameter value is determined for at least one transformation parameter designed as an extrinsic sensor parameter of the sensor and/or for the transformation parameter designed as a time offset between the sensor's internal clock and the clock of the vehicle's own movement module. The at least one transformation parameter designed as an extrinsic sensor parameter is, for example, the x-coordinate, y-coordinate or z-coordinate of the installation position of the sensor in the vehicle coordinate system or the roll angle R, pitch angle or yaw angle G of the sensor with respect to the vehicle coordinate system.

1. a) Transformation der Messpositionen MP mit einem vorgegebenen Transformationsparameterwert des jeweiligen Transformationsparameters in das Koordinatengitter KG des Weltkoordinatensystems WKS,
2. b) Zählen der Zellen Z des Koordinatengitters KG, denen zumindest eine der transformierten Messpositionen TMP zugeordnet ist,
3. c) Wiederholen der Schritte a) und b) mit jeweils, d. h. bei jeder Wiederholung, modifizierten Transformationsparameterwerten der Transformationsparameter und Ermitteln der Transformationsparameterwerte der Transformationsparameter, die in Schritt b) zu einem Minimum führen, d. h. zu einem Minimum der Anzahl A der Zellen Z des Koordinatengitters KG, denen zumindest eine der transformierten Messpositionen TMP zugeordnet ist. Dabei kann beispielsweise vorgesehen sein, dass bei jeder Wiederholung der Schritte a) und b) der Transformationsparameterwert nur jeweils eines Transformationsparameters oder der jeweilige Transformationsparameterwert mehrerer oder aller Transformationsparameter gleichzeitig modifiziert wird.

Expediently, however, not only the transformation parameter value of a single transformation parameter for the coordinate transformation is determined, but in this method for calibrating the distance-measuring sensor of the vehicle, in particular with regard to the vehicle's own movement, advantageously transformation parameter values of several transformation parameters, for example all extrinsic sensor parameters and for example additionally the temporal offset between the sensor-internal clock of the sensor and the clock of the vehicle's own movement module, determined for the coordinate transformation, with which a measurement position MP determined in the sensor coordinate system is transformed into the world coordinate system WKS. i.e. a transformation parameter value is determined for the respective transformation parameter. The transformation parameter values of the transformation parameters are determined by the optimization method, in which measurement positions MP, which were determined with the sensor at different points in time during the vehicle's own movement, are evaluated. The optimization process includes the following steps:

1. a) Transformation of the measurement positions MP with a predetermined transformation parameter value of the respective transformation parameter into the coordinate grid KG of the world coordinate system WKS,
2. b) counting the cells Z of the coordinate grid KG to which at least one of the transformed measurement positions TMP is assigned,
3. c) Repeating steps a) and b) with in each case, ie with each repetition, modified transformation parameter values of the transformation parameters and determining the transformation parameter values of the transformation parameters, which in step b) lead to a minimum, ie to a minimum of the number A of cells Z des Coordinate grid KG, to which at least one of the transformed measurement positions TMP is assigned. It can be provided, for example, that with each repetition of steps a) and b), the transformation parameter value of only one transformation parameter or the respective transformation parameter value of several or all transformation parameters is modified simultaneously.

The advantages of this method are that calibration is possible without additional reference sensors, that no correspondence is needed, and that the method works with any or at least almost any environment of the vehicle. In particular, no preconditions for scenarios have to be met for the method to work.

Furthermore, a temporal and spatial calibration in one process run is possible by means of the method. The method also works in particular in scenarios with dynamic objects, since these dynamic objects, in contrast to the static objects present in the respective scenario, have no influence on the optimization method described.

The method can be used for any distance-measuring sensor, for example for lidar sensors, stereo cameras, radar sensors and ultrasonic sensors.

Due to the quantization of the world coordinate system WKS in cells Z, the method is also robust against distance noise of individual point measurements.

Furthermore, no initial calibration is necessary.

In particular, the method enables a fast end-of-line calibration.

One use of the method is therefore, for example, this end-of-line calibration, also referred to as line decalibration. In this case, the method is used to carry out the calibration of the distance-measuring sensor or the respective distance-measuring sensor of the vehicle at the end of a production line of the vehicle. Instead of measuring the vehicle in a time-consuming process, it is sufficient to drive the vehicle a few meters using the method. This can significantly speed up the process chain. End-of-line calibration is one of the most time-consuming steps in vehicle construction.

A further use of the method is, for example, online calibration and/or decalibration detection. In this case, the calibration using the method is thus advantageously carried out during regular driving operation of the vehicle. The method can be used to continuously calibrate the distance-measuring sensor or the respective distance-measuring sensor of the vehicle while the vehicle is driving and/or to detect whether the distance-measuring sensor or the respective distance-measuring sensor of the vehicle is decalibrated, for example by a mechanical defect.

Another use of the method is, for example, offline calibration for verification and/or HD map generation, i. H. a creation of a high-resolution digital map. In this case, the method can be applied to recorded sequences, for example, in order to obtain a highly precise calibration. This can be used to serve as a reference for an online calibration. Furthermore, this calibration can be used for the construction of high-resolution digital maps.

Based on 1 until 5 an example of the implementation of the method using a distance-measuring sensor of the vehicle designed as a lidar sensor is shown. In the example shown, this sensor is used to capture real data from the area surrounding the vehicle, with dynamic objects also being present in the area and being recorded as well.

1 shows, highly schematized, a point cloud sequence recorded by the sensor, which was transformed into the world coordinate system WCS with the aid of the proper motion data and quantized in cells Z of 10 cm x 10 cm x 10 cm. i.e. According to step a) of the method, the recorded measurement positions MP were converted into the in 1 represented coordinate grid KG of the world coordinate system WKS transformed with the cells Z of 10 cm x 10 cm x 10 cm.

It can be seen that due to the insufficient calibration of the sensor here, all objects smear, i. H. are shown in particular blurred and partially merging into one another. No clearly delimited objects, in particular no clear object edges, are recognizable. The consequence of this is that a very large number of cells Z are required in order to depict the scene. i.e. a large number of cells Z are occupied by transformed measurement positions TMP, so at least one of the transformed measurement positions TMP has been assigned to a large number of cells Z of the coordinate grid KG.

In the following step b) of the method, the cells Z of the coordinate grid KG to which at least one of the transformed measurement positions TMP is assigned are counted. Subsequently, in step c) of the method, steps a) and b) are repeated with respectively modified transformation parameter values of the transformation parameters and the transformation parameter values of the transformation parameters are determined which lead to a minimum in step b), i. H. to a minimum of the number A of cells Z of the coordinate grid KG, to which at least one of the transformed measurement positions TMP is assigned.

In the 2 and 3 is this for changing the extrinsic transformation matrix for this sequence by adjusting the yaw angle G( 2 ) and roll angle R ( 3 ) shown. Are shown on the abscissa axis in 2 the trans formation parameter values for the extrinsic sensor parameter yaw angle G as a transformation parameter and in 3 the transformation parameter values for the extrinsic sensor parameter roll angle R as transformation parameters and the number A of the occupied cells Z of the coordinate grid KG, ie the cells Z of the coordinate grid KG, to which at least one of the transformed measurement positions TMP is assigned, is plotted on the ordinate axis.

This results in 2 an error function indicating the number A of occupied cells Z as a function of the transformation parameter designed as a yaw angle G and in 3 an error function indicating the number A of occupied cells Z as a function of the transformation parameter designed as a roll angle R. i.e. it is the number A of the occupied cells Z of the coordinate grid KG as a function of the yaw angle G ( 2 ) or roll angle R ( 3 ) entered. It can be seen that this leads to an error function with a clear minimum. The optimal transformation parameter value for the respective transformation parameter can thus be determined using the method described here, and the distance-measuring sensor can thus be optimally calibrated.

4 and 5 as a detail of 4 show the result of adapting the transformation parameters, here in particular the extrinsic sensor parameters of the extrinsic matrix, with the optimal angle values of the yaw angle G and roll angle R determined in this way and transferring the point measurements into the world coordinate system WCS. It can be seen that the surroundings of the vehicle detected by means of the distance-measuring sensor are now represented very precisely. Object edges are now sharp, walls of objects result in flat surfaces, freestanding objects are well defined. This shows the functionality of this method in real environmental conditions with dynamic road users.

Furthermore, it becomes clear that the idea of the method, the number A of cells Z of the coordinate grid KG, to which at least one of the transformed measurement positions TMP is assigned, is minimized by modifying the transformation parameter values of the transformation parameters and thereby the optimal transformation parameter value for the respective transformation parameter to determine is a very simple, but at the same time a very good and robust way of defining a minimization function. The comparison of 1 With 4 or their detail section according to 5 , clearly shows that the number A of occupied cells Z, ie cells Z of the coordinate grid KG, to which at least one of the transformed measurement positions TMP is assigned, has been significantly reduced.

Reference List

A

Number

G

yaw angle

KG

coordinate grid

R

roll angle

TMP

transformed measurement position

WCS

world coordinate system

Z

cell

Claims (10)

Method for calibrating a distance-measuring sensor of a vehicle, characterized in that a transformation parameter value of at least one transformation parameter for a coordinate transformation is determined, with which a measurement position (MP) determined in a sensor coordinate system is transformed into a world coordinate system (WKS), the transformation parameter value of the at least one Transformation parameter is determined by an optimization method in which the measurement positions (MP) that were determined with the sensor at different times during the vehicle's own movement are evaluated, the optimization method comprising the following steps: a) Transformation of the measurement positions (MP) with a predetermined one Transformation parameter value of the at least one transformation parameter in a coordinate grid (KG) of the world coordinate system (WKS), b) counting of cells (Z) of the coordinate grid (KG) to which at least one of the transf is assigned to standardized measurement positions (TMP), c) repeating steps a) and b) with a respectively modified transformation parameter value of the at least one transformation parameter and determining the transformation parameter value of the at least one transformation parameter, which leads to a minimum in step b). procedure after claim 1 , characterized in that transformation parameter values of several transformation parameters for the coordinate transformation are determined, with which a measurement position (MP) determined in the sensor coordinate system is transformed into the world coordinate system (WKS), the transformation parameter values of the transformation parameters being determined by the optimization method in which measurement positions (MP) who used the sensor at different times during the prop movement of the vehicle were determined, are evaluated, the optimization method comprising the following steps: a) transformation of the measurement positions (MP) with a predetermined transformation parameter value of the respective transformation parameter in the coordinate grid (KG) of the world coordinate system (WKS), b) counting the cells ( Z) of the coordinate grid (KG), to which at least one of the transformed measurement positions (TMP) is assigned, c) repeating steps a) and b) with respectively modified transformation parameter values of the transformation parameters and determining the transformation parameter values of the transformation parameters, which in step b) become one lead minimum. Method according to one of the preceding claims, characterized in that a lidar sensor, a radar sensor, an ultrasonic sensor or a stereo camera of the vehicle is calibrated as the sensor. Method according to one of the preceding claims, characterized in that the vehicle's own movement is determined by means of a vehicle's own movement module. procedure after claim 4 , characterized in that a transformation parameter value is determined for at least one transformation parameter designed as an extrinsic sensor parameter of the sensor and/or for a transformation parameter designed as a time offset between a sensor-internal clock of the sensor and a clock of the vehicle's own motion module. procedure after claim 5 , characterized in that the at least one transformation parameter designed as an extrinsic sensor parameter is an x-coordinate, a y-coordinate or a z-coordinate of an installation position of the sensor in a vehicle coordinate system or a roll angle (R), a pitch angle or a yaw angle (G) of the Sensor with respect to the vehicle coordinate system. Method according to one of the preceding claims, characterized in that the calibration is carried out at the end of a production line of the vehicle. Method according to one of the preceding claims, characterized in that the calibration is carried out during regular driving operation of the vehicle. Method according to one of the preceding claims, characterized in that the calibration is carried out during a data determination to create at least one high-resolution digital map. Method according to one of the preceding claims, characterized in that at least one calibration is carried out, are evaluated in the measurement positions (MP), which were determined with the sensor at different times within a short period of time during the vehicle's own movement, and at least one calibration is carried out in which the measurement positions (MP) that were determined with the sensor at different points in time within a longer period of time during the vehicle's own movement are evaluated, with the longer period of time including the points in time of the shorter period and additional points in time at which the sensor Measurement positions (MP) were determined includes.

Images