DE102022212898A1 - Sensor device and method for detecting objects on a roadway of a vehicle - Google Patents

Abstract

The invention relates to a sensor device (20) for detecting objects (5) on a roadway (2) of a vehicle (1), comprising: an input interface (21) for receiving a sensor signal from a lidar sensor (3), the sensor signal comprising information on reflections of laser beams (4) from the lidar sensor (3) on illuminated objects (5) in an environment of the vehicle (1); an accumulation unit (22) for storing positions of the reflections as a data point cloud (7) and for accumulating subsets of the reflections to form lidar rings; a computing unit (23) for adapting parameters of a local roadway model in a near area of the vehicle (1) and a global roadway model in a far area of the vehicle (1) using adjacent lidar rings and for generating a roadway plane (12) based thereon; a filter unit (26) for filtering data points (6) of the data point cloud (7) with a minimum distance (10) to the road surface (12) in sections along the road surface (12), wherein each filtered section has a filter convolution value (13); a detection unit (24) for detecting objects (5) on the road surface (2) of the vehicle (1) based on the filter convolution value (13); and an output unit (25) for outputting an output signal with information about the detected objects (5). The invention further relates to a method and a system (19) for detecting objects (5) on a road surface (2) of a vehicle (1).

Description

The present invention relates to a sensor device for detecting objects on a roadway of a vehicle. The present invention further relates to a method and a system for detecting objects on a roadway of a vehicle and to a computer program product.

Modern vehicles (cars, vans, trucks, motorcycles, etc.) have a large number of sensors that provide the driver with information and control individual functions of the vehicle partially or fully automatically. Sensors record the vehicle's surroundings and other road users. Based on the recorded data, a model of the vehicle's environment can be created and changes in this vehicle environment can be responded to.

An important sensor principle for detecting the environment is lidar technology (light detection and ranging). A lidar sensor is based on the transmission of light signals and the detection of the reflected light. A distance to the location of the reflection can be calculated using a time-of-flight measurement. It is also possible to determine a relative speed. Both unmodulated pulses and frequency-modulated signals (chirps) can be used here (frequency-modulated continuous wave lidar, FMCW lidar). A target can be detected by evaluating the received reflections. With regard to the technical implementation of the lidar sensor, a distinction is made between scanning and non-scanning systems. A scanning system is usually based on micromirrors and scanning the environment with a light spot, whereby a coaxial system is referred to when the transmitted and received light pulses are deflected by the same micromirror. In non-scanning systems, several transmitting and receiving elements are arranged statically next to each other (especially a so-called focal plane array arrangement). The determined distances and their positions are often stored in the form of a so-called data point cloud, which allows further analyses and conclusions to be drawn about objects in the vehicle's surroundings.

A relevant function of a lidar sensor in the vehicle environment is the detection of obstacles on the road, such as lost cargo, tires or injured people. Injured people and tires in particular can take up a comparatively small area visible to the lidar sensor. When used on a motorway or expressway, such obstacles should be detected in a range of 100 to 300 m. To achieve this reliably, a high spatial resolution is necessary. For example, a resolution of 0.05° in the horizontal direction and 0.025° in the vertical direction may be necessary to ensure that an object is hit by several scanning points.

In order to ensure reliable detection of objects, the current technology often works based on models of the road. Deviations in the measurement data from these models are detected and, based on this, potentially disruptive or dangerous objects are identified.

A challenge with previous evaluation approaches is that reflective laser beams emitted at an unfavorable angle are no longer reflected back to the lidar sensor and are therefore not detectable by the lidar sensor. This means that, depending on the vehicle's surroundings, smaller and larger gaps can arise in the data point cloud, on the basis of which the vehicle's surroundings cannot always and everywhere be reliably determined. Another challenge is distinguishing between an object lying on the road and a property of the road itself. If, for example, the gradient of the road changes in an area in front of the vehicle, this can lead to ambiguous situations when detecting obstacles.

Based on this, the object of the invention is to provide an approach for reliably detecting objects in the environment of a vehicle. In particular, compared to previous approaches, an improved distinction between objects on the road and road properties, such as a change in gradient, should be made possible.

1. a) einer Eingangsschnittstelle zum Empfangen eines Sensorsignals eines Lidar-Sensors, wobei das Sensorsignal Informationen zu Reflektionen von Laserstrahlen des Lidar-Sensors an angestrahlten Objekten in einer Umgebung des Fahrzeugs umfasst;
2. b) einer Akkumulationseinheit zum Abspeichern von Positionen der Reflektionen als Datenpunktwolke und zum Akkumulieren von Teilmengen der Reflektionen zu Lidar-Ringen;
3. c) einer Recheneinheit zum Anpassen von Parametern eines lokalen Fahrbahnmodells in einem Nahbereich des Fahrzeugs und eines globalen Fahrbahnmodells in einem Fernbereich des Fahrzeugs anhand benachbarter Lidar-Ringe und zum darauf basierenden Erzeugen einer Fahrbahnebene;
4. d) einer Filtereinheit zum Filtern von Datenpunkten der Datenpunktwolke mit einem Mindestabstand zur Fahrbahnebene abschnittsweise entlang der Fahrbahnebene, wobei jeweils ein gefilterter Abschnitt einen Filterfaltungswert aufweist;
5. e) einer Detektionseinheit zum Detektieren von Objekten auf der Fahrbahn des Fahrzeugs basierend auf dem Filterfaltungswert; und
6. f) einer Ausgabeeinheit zum Ausgeben eines Ausgabesignals mit Informationen zu den detektierten Objekten.

The object is achieved according to the invention by a sensor device for detecting objects on a roadway of a vehicle, comprising:

1. a) an input interface for receiving a sensor signal from a lidar sensor, the sensor signal comprising information on reflections of laser beams from the lidar sensor on illuminated objects in an environment of the vehicle;
2. b) an accumulation unit for storing positions of the reflections as a data point cloud and for accumulating subsets of the reflections into lidar rings;
3. c) a computing unit for adapting parameters of a local road model in a near area of the vehicle and a global road model in a far area of the vehicle based on neighboring Lidar rings and for generating a road surface plane based on them;
4. d) a filter unit for filtering data points of the data point cloud with a minimum distance to the road surface in sections along the road surface, whereby each filtered section has a filter convolution value;
5. e) a detection unit for detecting objects on the roadway of the vehicle based on the filter convolution value; and
6. f) an output unit for outputting an output signal with information about the detected objects.

1. a) Empfangen eines Sensorsignals eines Lidar-Sensors, wobei das Sensorsignal Informationen zu Reflektionen von Laserstrahlen des Lidar-Sensors an angestrahlten Objekten in einer Umgebung des Fahrzeugs umfasst;
2. b) Abspeichern von Positionen der Reflektionen als Datenpunktwolke;
3. c) Akkumulieren von Teilmengen der Reflektionen zu Lidar-Ringen;
4. d) Anpassen von Parametern eines lokalen Fahrbahnmodells in einem Nahbereich des Fahrzeugs und eines globalen Fahrbahnmodells in einem Fernbereich des Fahrzeugs anhand benachbarter Lidar-Ringe zu einer basierend auf dem angepassten lokalen Fahrbahnmodell sowie globalen Fahrbahnmodell erzeugten Fahrbahnebene;
5. e) Filtern von Datenpunkten der Datenpunktwolke mit einem Mindestabstand zur Fahrbahnebene abschnittsweise entlang der Fahrbahnebene, wobei jeweils ein gefilterter Abschnitt einen Filterfaltungswert aufweist;
6. f) Detektieren von Objekten auf der Fahrbahn des Fahrzeugs basierend auf dem Filterfaltungswert; und
7. g) Ausgeben eines Ausgabesignals mit Informationen zu den detektierten Objekten.

Furthermore, the object is achieved according to the invention by a method designed according to the device with the following steps:

1. a) receiving a sensor signal from a lidar sensor, the sensor signal comprising information on reflections of laser beams from the lidar sensor on illuminated objects in an environment of the vehicle;
2. b) Saving positions of reflections as a data point cloud;
3. c) Accumulating subsets of reflections into lidar rings;
4. (d) adapting parameters of a local road model in a near area of the vehicle and of a global road model in a far area of the vehicle using adjacent lidar rings to form a road plane generated based on the adapted local road model and global road model;
5. e) filtering data points of the data point cloud with a minimum distance to the road surface in sections along the road surface, whereby each filtered section has a filter convolution value;
6. f) detecting objects on the vehicle’s path based on the filter convolution value; and
7. g) Outputting an output signal with information about the detected objects.

In addition, one aspect of the invention relates to a system for detecting objects on a roadway with a sensor device as described above and a lidar sensor for detecting objects in an environment of the vehicle.

The invention also relates to a computer program product with program code for carrying out the steps of the method according to the invention when the program code is executed on a computer.

It is understood that the features mentioned above and those to be explained below can be used not only in the combination specified in each case, but also in other combinations or on their own, without departing from the scope of the present invention. In particular, the method, the system and the computer program product can be designed in accordance with the embodiments described for the sensor device in the dependent claims.

Lidar sensors usually have sources of error and noise, which are the reason why reflections of objects on the road, reflections of the road itself and falsified reflections cannot always be reliably distinguished. In this respect, the invention provides for a virtual road surface to be generated based on a local and a global road model. With the help of this road surface and predetermined parameters, it can be determined whether subsets of the data point cloud belong to the road or can be assigned to a relevant object on the road.

The inventive use of a local road model in the near area of the vehicle and a global road model in the far area of the vehicle increases the robustness in the detection of relevant objects on the road. The global road model detects reflections that are further apart from one another in a more distant area of the vehicle's surroundings. The local road model detects reflections from objects that are a short distance from the vehicle with a higher resolution. The inventive combination of both road models enables relevant objects to be reliably detected from a distance from the road surface of approx. 0.2 m up to a distance of at least approx. 40 m and to be detected as being at a distance from the road surface.

The road surface level can be determined based on an equation, for example a plane equation, by means of a regression analysis. The use of a robust regression algorithm, for example a median-weighted RANSAC algorithm (random sample consensus), and a multiple repetition of the regression analysis with newly recorded reflections enables a reliable determination of the road surface level that is not distorted by a few outliers. Practice has shown that the regression analysis takes at least about twenty times, ideally at least about thirty times.

According to the invention, not only distances between individual or many data points of the data point cloud can be used to detect objects on the road. If data points of the data point cloud are projected onto a filter plane, which can correspond in particular to a road plane, from a minimum distance to the road plane (for example 2 cm, 4 cm or more), the filter unit can filter these data points along the road plane using a two-dimensional filter. A filter kernel of the two-dimensional filter can be circular and have a radius of 80 cm, for example. An advantage of this configuration is that, for example, a small child lying on the road can also be detected and distinguished from a (smaller) pile of leaves. Alternatively, filter kernels with smaller or larger radii are conceivable, with rectangular, triangular or elliptical filter kernels also being conceivable.

Projecting the data points onto the filter plane and then filtering them with a two-dimensional filter makes use of the fact that volume information above the road plane can be mapped using a filter convolution value. This makes it easier to distinguish, for example, a bulky person lying on the road from a pile of leaves or a cardboard box on the road. For example, an object can be detected if a filter convolution value - which corresponds to a volume value - is above a convolution threshold along the vehicle's roadway.

Optionally, it is possible to omit projecting the data points onto the filter plane and to directly use a three-dimensional filter with, for example, a spherical or polyhedral filter kernel.

In particular, the approach according to the invention makes it possible to avoid unnecessary emergency braking or speed reductions due to negligible objects on the road.

In order to ensure a quick reaction when objects are detected, an advantageous embodiment of the invention can provide that the output signal is sent to a control unit for controlling the vehicle taking the detected objects into account, in particular for triggering an emergency braking maneuver. The control unit can, for example, directly influence a braking process of the vehicle or initiate a steering maneuver to avoid the object on the road.

In order to compensate for uneven road profiles, according to an advantageous embodiment of the invention, the accumulation unit can be designed to project a subset of the data point cloud onto the road surface of the vehicle using a road map. The road map can, for example, correspond to a map of the surroundings received from an Internet server or stored on a corresponding data carrier. An object on the road is often indistinguishable from a road section with an incline. In this context, reliability can be further increased by additionally using a road map. If the road profile for the current and subsequent road section is known using an environmental or road map, the subsets of the data point cloud on the road map can be projected onto the road surface so that the incline of the road section can be compensated and objects on the road can be recognized based on their distance from the road surface.

According to an advantageous embodiment, it can be provided that the computing unit is designed to adapt the parameters of the local road model based on a straight line equation along the lidar rings. Optionally, according to the method according to the invention, it can be provided that when adapting the parameters of the local road model, a straight line equation based on a straight line along the lidar rings is used.

The local road model records the road using the lidar rings in the immediate vicinity of the vehicle, so that a straight line equation can be adapted to each individual lidar ring using the regression analysis according to the invention. Deviations from the individual straight line equation indicate an elevation of the road, which can indicate relevant objects on the road that are to be detected.

In practice, the following basic straight line equation has proven to be useful: $$\begin{array}{ccc}z& \ne & 0=b\cdot y+c+z_0\\ -z_0& =& b\cdot y+c\end{array}$$

With z and y as coordinates in the direction of a z- or y-axis with respect to the road plane, z ₀ as the position in the direction of the z-axis at the beginning of the lidar ring and b and c as regression parameters meters that can be adjusted using regression analysis so that the road surface has a minimal, in particular median-weighted error to the data point cloud. This is because real measured values belonging to a lidar ring regularly deviate from the starting position z ₀ during a measurement run due to tolerances and measurement errors of the lidar sensor. According to this exemplary design, it is assumed that the deviation is taken into account linearly.

Insofar as it is advantageous according to the invention to carry out N repetitions of the regression algorithm with newly recorded reflections and a new data point cloud so that the regression parameters b and c result in an improved local road model of the lidar rings, the following N-dimensional matrix equation can be advantageous: $$\begin{array}{ccc}\left[-{\mathrm{<figure-callout\; id="1"\; label="z"\; filenames="DE102022212898A1\_0005.png,DE102022212898A1\_0006.png"\; state="\{\{state\}\}">z</figure-callout>}}_1,\dots ,-z_N\right]& =& \left[b,c\right]\left[\begin{array}{ccc}{\mathrm{<figure-callout\; id="1"\; label="y"\; filenames="DE102022212898A1\_0005.png,DE102022212898A1\_0006.png"\; state="\{\{state\}\}">y</figure-callout>}}_1& \dots & {\mathrm{<figure-callout\; id="1"\; label="y\; N"\; filenames="DE102022212898A1\_0005.png,DE102022212898A1\_0006.png"\; state="\{\{state\}\}">y</figure-callout>}}_N\\ 1& \dots & 1\end{array}\right]\\ Z& =& \left[b,c\right]\cdot Y\\ \left[b,c\right]& =& Z\cdot Y^T{\left(Y\cdot Y^T\right)}^{-1}\end{array}$$

Alternatively, according to an embodiment of the sensor device according to the invention, it can be provided that the computing unit is designed to adapt the parameters of the local road model based on a quadratic road equation over the respective lidar rings. Optionally, it can be provided for the method according to the invention that a quadratic road equation is used over the respective lidar rings when adapting the parameters of the local road model. Accordingly, the following road equation is used for each lidar ring, so that horizontal curvatures of the road along the lidar ring can also be detected by the local road model: $$\begin{array}{ccc}z& \ne & 0=a\cdot y+b\cdot y\cdot y+c+z_0\\ -z_0& =& a\cdot y+b\cdot {\mathrm{<figure-callout\; id="2"\; label="y"\; filenames="DE102022212898A1\_0005.png,DE102022212898A1\_0006.png"\; state="\{\{state\}\}">y</figure-callout>}}^2+c\end{array}$$

According to a particularly advantageous embodiment of the invention, it can be provided that the computing unit is designed to adapt the parameters of the global road model based on a plane equation over adjacent lidar rings. According to an advantageous embodiment of the method according to the invention, it can optionally be provided that a plane equation based on a plane over two lidar rings is used when adapting the parameters of the global road model. Alternatively, the method according to the invention can be adapted, wherein a plane equation based on a plane over three lidar rings is used when adapting the parameters of the global road model.

Erfindungsgemäß ist es entsprechend der Geradengleichung vorteilhaft, N Wiederholungen des Regressionsalgorithmus mit neu erfassten Reflektionen und einer neuen Datenpunktwolke durchzuführen, sodass die Regressionsparameter a, b und c ein verbessertes globales Fahrbahnmodell basierend auf den Lidar-Ringen ergeben. If two or more neighboring lidar rings are included in the adaptation of the global road model, the parameters of the plane equation can be adapted to the included lidar rings to the road plane. $$\begin{array}{ccc}z& \ne & 0=a\cdot x+b\cdot y+c+z_0\\ -z_0& =& a\cdot x+b\cdot y+c\end{array}$$

According to the invention, it is advantageous, according to the straight line equation, to carry out N repetitions of the regression algorithm with newly recorded reflections and a new data point cloud, so that the regression parameters a, b and c result in an improved global road model based on the lidar rings.

According to an advantageous embodiment of the invention, it can be provided that the filter unit for filtering all data points of the data point cloud with a maximum distance to the road surface is designed in sections along the road surface. Optionally, a maximum distance can be used as an additional criterion. Reflections and data points of the data point cloud that are above a maximum distance to the road surface are not included in the filtering of the data point cloud. In this respect, they are considered irrelevant or implausible. Advantageously, data points of the data point cloud can be ignored from a predefined maximum distance to the road surface of, for example, 0.2 m to 2 m. This saves computing effort and computing time and improves efficiency. Collisions with massive and voluminous objects on the road surface are avoided.

In order to avoid unnecessary braking processes, according to an advantageous embodiment of the invention, it can be provided that the detection unit is designed to detect the objects based on a distance-dependent folding threshold value. Often, despite the detection of an object assessed as voluminous or relevant with a large filter folding value, no reaction (e.g. braking) is necessary because the object is still sufficiently far away. To reflect this, the folding threshold value can be distance-dependent, i.e. depend on the distance to the detected object. For example, a linear dependence of the folding threshold value on the distance can be used.

According to a particularly advantageous embodiment of the invention, it can be provided that the detection unit is designed to segment the road surface of the vehicle in the surrounding map and to accumulate the reflections in each segment. Optionally, it can be provided that a grid field is used when segmenting the road surface of the vehicle in the surrounding map and to accumulate the reflections in each segment. Accordingly, it can be advantageous to segment the road surface according to a grid field so that all data points of the data point cloud located in a grid can be accumulated into a single segment value. This can save computing time and make it easier to recognize contours of objects on the road. In practice, a grid field size of 10 cm to 30 cm has proven to be sufficient.

Typically, the sensor device communicates with at least one lidar sensor, which can be mounted on a roof of the vehicle, for example. The lidar sensor emits laser beams along an annular section of a circumferential line around the vehicle and receives the laser beams reflected from surfaces.

A lidar ring is understood here in particular to be the sum of all reflections that correspond to a row of a line-based scanning lidar sensor. The reflections belonging to a row form a lidar ring, so to speak. Neighboring lidar rings correspond in particular to reflections from directly adjacent rows. A road model is understood here to be a mathematical model for mapping the road, in particular in coordinates of the data point cloud. Both models have corresponding model parameters. A local road model is understood to be a mapping of the course of the road in a close area, for example at a distance of a few meters. A global road model is understood to be a mapping of the course of the road in a more distant (far) area. The areas do not have to be directly adjacent to one another and there can also be overlaps. A data point cloud includes data points that represent the positions of reflections.

• 1 eine schematische Darstellung eines erfindungsgemäßen Systems;
• 2 eine schematische Darstellung einer erfindungsgemäßen Vorrichtung;
• 3 eine von einem Laserstrahl eines Lidar-Sensors erfasste Person;
• 4 eine von zwei Lidar-Sensoren erfasste Person auf einer Fahrbahn;
• 5 ein zweidimensionales Diagramm eines mit Lidar-Sensoren erfassten Umgebungsbereichs eines Fahrzeugs; und
• 6 eine schematische Darstellung eines erfindungsgemäßen Verfahrens.

The invention is described below by way of example using schematic representations. They show:

• 1 a schematic representation of a system according to the invention;
• 2 a schematic representation of a device according to the invention;
• 3 a person detected by a laser beam from a lidar sensor;
• 4 a person on a road detected by two lidar sensors;
• 5 a two-dimensional diagram of an area surrounding a vehicle captured by lidar sensors; and
• 6 a schematic representation of a method according to the invention.

In 1 a vehicle 1 with a system 19 for detecting an object 5 in an environment of the vehicle 1 is shown schematically. The illustration corresponds to a side sectional view of the situation. The object 5 can be, for example, a car tire, a vehicle part lying on the roadway 2 or a person. The system 19 comprises a sensor device 20 and a lidar sensor 3. In the exemplary embodiment shown, the system 19 is arranged inside the vehicle 1 or integrated into the vehicle. The lidar sensor 3 is, for example, mounted in the area of a roof of the vehicle 1. The sensor device 20 is, for example, integrated into a control unit or into the lidar sensor 3 itself. It goes without saying that it is also possible for the lidar sensor 3 and the sensor device 20 to be designed separately and, for example, integrated into a mobile device. In the following, an x-axis or an x-direction is understood to mean a direction along a vehicle's longitudinal axis in a vehicle-fixed coordinate system, a y-axis or a y-direction is understood to mean a direction along a vehicle's transverse axis, and a z-direction or a z-axis is understood to mean a direction along a vehicle's vertical axis. However, the axes can also be understood as axes in another coordinate system that is used for the sensor data.

In 2 The sensor device 20 according to the invention is shown schematically. The sensor device 20 comprises an input interface 21, an accumulation unit 22, a computing unit 23, a detection unit 24, a filter unit 26 and an output unit 25. The sensor device 20 can be integrated, for example, in a lidar measuring device or in a vehicle control unit or a driver assistance system. The units and interfaces can be partially or completely implemented in hardware and/or software.

A sensor signal is received via the input interface 21, for example via a corresponding bus system. This sensor signal is evaluated and processed in the accumulation unit 22, the computing unit 23, the detection unit 24 and the filter unit 26. An output signal is output via the output unit 25, whereby the output signal is used on the one hand for informing ization of a vehicle driver or for direct further processing in a control unit to control a driving function in an autonomous or semi-autonomous vehicle, in particular to initiate an (emergency) braking process.

In 3 a vehicle 1 is shown on a roadway 2, the vehicle 1 having a lidar sensor 3 with which several laser beams 4 are sent in the direction of an object 5 (in the example shown, a person). Reflections from a head and a body of the person and the roadway 2 are received by means of a detector and displayed in a data point cloud 7 having data points 6, each data point 6 of the data point cloud 7 comprising the coordinates assigned to a reflection. A data point 6 can in particular correspond to a coordinate specification. In this respect, a sensor device according to the invention on the vehicle 1 can detect a distance 8 of the data points 6 on the person that are spaced apart from the roadway 2 as belonging to a relevant object 5 on the roadway 2.

Advantageously, a local road model of the road 2 for a short range is used to detect relevant objects 5 in order to be able to monitor the road 2 in this short range in as much detail as possible, so that, for example, even smaller objects 5 on the road 2 can be detected. For more extensive monitoring of the road 2, even at greater distances, a global road model can be used that maps a long-distance area of the vehicle 1. In comparison to the local road model, the global road model can, for example, be based on detecting a more distant environmental area with lower resolution. By combining both road models, more distant or larger objects 5 can be detected early on based on the global road model, and smaller objects 5 in the close range of the vehicle 1 can also be reliably detected based on the local road model.

In 4 an object 5 (standing person) located in front of a vehicle 1 is shown on a lane 2 of the vehicle 1. Two lidar sensors 3 emit laser beams 4, the reflections of which from the person standing on the lane 2 can be received again by the lidar sensors 3. After an accumulation of the data points 6, for example to form lidar rings, it can be seen that all or many data points 6 or lidar rings are not significantly spaced from one another in the x-direction.

3 and 4 show a distance window 9 with a minimum distance 10 and a maximum distance 11 to the road surface 12. Data points 6 of the data point cloud 7 within this distance window 9 are used for section-by-section filtering along the road surface 12. If a filter folding value generated by the filtering exceeds a folding threshold value, it can be assumed that there is an object 5 (in the example shown, a person) on the road surface 2. In this respect, the folding threshold value can be designed to be distance-dependent - for example, linearly dependent on the distance - based on a distance of the object 5 to the vehicle 1 or to the lidar sensor 3. A lower filter folding value of a person lying on the road surface 2 near the vehicle 1 can be sufficient to initiate emergency braking or to issue a warning, whereas a higher filter folding value of a person walking next to the vehicle 1 can only cause a reduction in speed.

In 5 a two-dimensional diagram 14 of an area surrounding the vehicle recorded with a lidar sensor on a vehicle is shown. The y-axis 15 shows a position of the reflections in relation to the vehicle's transverse axis. The x-axis 16 shows a distance to the vehicle in the direction of the vehicle's longitudinal axis. Two clusters 17 are shown in diagram 14, with the smaller cluster on the right representing a pile of leaves in autumn and the larger cluster on the left representing a person lying on the road. In this respect, it is immediately apparent that two-dimensional filtering of the clusters 17 along the road plane produces a different filter convolution value 13 - represented by the circles of different sizes around the clusters 17 - so that the person can be clearly distinguished from the pile of leaves, since different numbers of data points 6 of the clusters 17 lead to different filter convolution values 13.

In 6 A method according to the invention for detecting objects on a roadway is shown schematically. The method comprises steps of receiving S10 a sensor signal, storing S12 positions of the reflections as a data point cloud, accumulating S14 subsets of the reflections to form lidar rings, adjusting S16 parameters of a local road model and a global road model, filtering S17 data points, detecting S18 objects and outputting S20 an output signal. The method can in particular be implemented in software that is executed on a lidar sensor or in a vehicle control unit.

The invention has been fully described and explained with reference to the drawings and the description. The description and explanation are to be understood as exemplary and not restrictive. The invention is not limited to the disclosed embodiments. Other embodiments or variations will become apparent to those skilled in the art upon use of the present invention and upon careful analysis of the drawings, the disclosure and the following claims.

In the patent claims, the words "comprising" and "having" do not exclude the presence of further elements or steps. The undefined article "a" or "an" does not exclude the presence of a plurality. A single element or unit can perform the functions of several of the units mentioned in the patent claims. An element, unit, interface, device and system can be partially or completely implemented in hardware and/or software. The mere mention of some measures in several different dependent patent claims should not be understood to mean that a combination of these measures cannot also be used advantageously. Reference signs in the patent claims are not to be understood as limiting.

Reference symbols

1

vehicle

2

roadway

3

Lidar sensor

4

laser beam

5

Person/relevant object

6

Data point

7

Data point cloud

8th

Distance to a roadway/level

9

Distance window

10

Minimum distance

11

Maximum distance

12

Road level

13

Filter convolution value

14

diagram

15

y-axis

16

X axis

17

Cluster

19

system

20

Sensor device

21

Input interface

22

Accumulation unit

23

Computing unit

24

Detection unit

25

Output unit

26

Filter unit

Claims (12)

Sensor device (20) for detecting objects (5) on a roadway (2) of a vehicle (1) with: a) an input interface (21) for receiving a sensor signal from a lidar sensor (3), the sensor signal comprising information on reflections of laser beams (4) of the lidar sensor (3) on illuminated objects (5) in an environment of the vehicle (1); b) an accumulation unit (22) for storing positions of the reflections as a data point cloud (7) and for accumulating subsets of the reflections to form lidar rings; c) a computing unit (23) for adapting parameters of a local roadway model in a near area of the vehicle (1) and a global roadway model in a far area of the vehicle (1) using neighboring lidar rings and for generating a roadway plane (12) based thereon; d) a filter unit (26) for filtering data points (6) of the data point cloud (7) with a minimum distance (10) to the road surface (12) in sections along the road surface (12), wherein each filtered section has a filter convolution value (13); e) a detection unit (24) for detecting objects (5) on the road surface (2) of the vehicle (1) based on the filter convolution value (13); and f) an output unit (25) for outputting an output signal with information about the detected objects (5). Sensor device (20) according to Claim 1 , wherein the output unit (25) is designed to output the output signal to a control unit for controlling the vehicle (1) taking into account the detected objects (5), in particular for triggering an emergency braking. Sensor device (20) according to one of the preceding claims, wherein the accumulation unit (22) is designed to project a subset of the reflections of the data point cloud (7) based on a road map into the road plane (12) of the vehicle (1). Sensor device (20) according to one of the preceding claims, wherein the computing unit (23) is designed to adapt the parameters of the local road model based on a straight line equation along the lidar rings. Sensor device (20) according to one of the Claims 1 until 3 , wherein the computing unit (23) is designed to adapt the parameters of the local road model based on a quadratic road equation via the respective lidar rings. Sensor device (20) according to one of the preceding claims, wherein the computing unit (23) is designed to adapt the parameters of the global road model based on a plane equation over two lidar rings. Sensor device (20) according to one of the preceding claims, wherein the filter unit (26) is designed to filter all data points (6) of the data point cloud (7) with a maximum distance (11) to the roadway plane (12) in sections along the roadway plane (12). Sensor device (20) according to one of the preceding claims, wherein the detection unit (24) is designed to detect the objects (5) based on a distance-dependent convolution threshold value. Sensor device (20) according to one of the preceding claims, wherein the detection unit (24) is designed to segment the road plane (12) of the vehicle (1) in the environment map and to accumulate the reflections in each segment. Method for detecting objects (5) on a roadway (2) of a vehicle (1), the method comprising the following method steps: a) receiving (S10) a sensor signal from a lidar sensor (3), the sensor signal comprising information on reflections of laser beams (4) of the lidar sensor (3) on illuminated objects (5) in an environment of the vehicle (1); b) storing (S12) positions of the reflections as a data point cloud (7); c) accumulating (S14) subsets of the reflections to form lidar rings; d) adapting (S16) parameters of a local roadway model in a near area of the vehicle (1) and a global roadway model in a far area of the vehicle (1) using neighboring lidar rings to form a roadway plane (12) generated based on the adapted local roadway model and global roadway model; e) filtering (S17) data points (6) of the data point cloud (7) with a minimum distance (10) to the road surface (12) in sections along the road surface (12), wherein each filtered section has a filter convolution value (13); f) detecting (S18) objects (5) on the road surface (2) of the vehicle (1) based on the filter convolution value (13); and g) outputting (S20) an output signal with information about the detected objects (5). Computer program product with program code for carrying out the steps of the method according to Claim 10 when the program code is executed on a computer. System (19) for detecting objects (5) on a roadway (2) of a vehicle (1) with a sensor device (20) according to one of the Claims 1 until 9 and a lidar sensor (3) for detecting objects (5) in an environment of the vehicle (1).

Images