Method and device for deleting at least one landmark position of a landmark in a radar map

Abstract

The solution presented herein relates to a method (200) for deleting at least one landmark position (102) of a landmark (105) in a radar map (110). The method comprises at least a reading step and a deleting step. In the reading step, at least one road position (125) of at least one road section of a road (130) imaged in the radar map (110) is read. In the deleting step, the landmark position (102) is deleted (210) from the radar map (110) if the landmark position (102) is at least a predetermined distance (135) from the road position (125).

Description

Technical Field

[0001] The invention relates to a device and a method for deleting at least one landmark position of a landmark in a radar map. The invention also relates to a computer program. Background Art

[0002] Modern driver assistance systems (ADAS) and highly automated vehicle systems for UAD (urban automated driving) increasingly require detailed knowledge of the vehicle's surroundings and situational awareness. This requires precise positioning. Devices are known for extracting landmarks from aerial photography of roads. These landmarks are stored in a positioning map and subsequently used for vehicle positioning. Furthermore, radar maps provided by satellite can be used to optimize the positioning map. Summary of the Invention

[0003] Based on this background, a method for deleting at least one landmark position of a landmark in a radar map is proposed with the solution presented here, further a device for deleting at least one landmark position of a landmark in a radar map is proposed, and finally a corresponding computer program is proposed.

[0004] The advantage achieved by the described approach is that landmark positions displayed in the radar map that are not necessary for vehicle positioning can be deleted, ie eliminated from the radar map, so that a radar map can be provided that only includes changes in landmark positions located in the road area.

[0005] A method for deleting at least one landmark position of a landmark in a radar map is proposed, wherein the method comprises at least the following steps:

[0006] reading at least one road position of at least one section of the road imaged in the radar map; and

[0007] When a landmark position is at least a predetermined distance away from a road position, the landmark position is deleted from the radar map. The road position can be extracted from a conventional navigation map, for example.

[0008] Landmarks are understood to be stationary objects in a road area that can be used for vehicle positioning and / or navigation, such as infrastructure elements such as traffic signs or streetlight masts or traffic sign bridges. These landmarks useful for vehicle positioning and / or navigation are hereinafter referred to as pillars, for example. However, landmarks are also understood to be man-made structures such as trees or other structures that are not easily perceived by vehicles traveling on the road and are therefore not used for vehicle positioning and / or navigation.

[0009] The radar map can be a satellite-provided radar map that contains or images highly accurate landmark locations. The method described herein can provide a variable radar map that advantageously only images landmark locations up to a predetermined distance from the road, as landmarks further away are less important for vehicle positioning. The predetermined distance can be, for example, 5 to 10 meters.

[0010] When a modified radar map is used, for example, in a method for optimizing a surroundings map read by a vehicle system with onboard sensor systems, this can facilitate or shorten the associated calibration between the currently modified radar map and the surroundings map. Surroundings maps generated using conventional vehicle sensor signals from a vehicle reading unit or sensor system in the vehicle are often too inaccurate to be used as localization maps. Such surroundings maps can be optimized using radar maps by a system according to the approach presented herein and can then be used as highly accurate localization maps. This approach can limit the need for expensive mapping sensor systems or mapping vehicle systems. By combining this with an already highly accurate surroundings map based on measurements from mapping sensors, the method and / or system presented herein offers further possibilities for advantageously influencing the mapping results.

[0011] Because roads or road routes are often difficult to identify on radar maps, the method presented herein advantageously includes, according to one embodiment, an identification step, performed before the reading step, in which a road location is identified by matching the radar map with a topological road map that contains or images at least one other road location of a road segment. A matching method can be understood as a method that overlays or compares the two maps, particularly by highlighting or extracting consistent features. In this matching method, the radar map and the topological road map, which is difficult to identify the road, can be overlaid, for example. In the identification step, if another road location imaged on the topological road map during the matching method corresponds to a road location on the radar map within a tolerance range, then that road location is identified, particularly if that road location is displayed in a darker color. The tolerance range can be understood, for example, as two to five times the road width. Thus, the road route in the topological road map can be quickly and easily used to identify the road in the radar map. Asphalt roads are typically displayed in a dark color, such as black, on radar maps, so the dark areas on the radar map can be used to verify the road location displayed by the topological road map.

[0012] To distinguish different landmarks that are not removed by the method (because they are located sufficiently close to the road), the method advantageously includes a classification step in which at least one landmark imaged at a landmark location or another landmark imaged at another landmark location on the radar map is classified using a classifier. Classification (also understood as grading) can, for example, further remove landmark locations that, despite being located near the road, do not belong to the expected category (e.g., not pillars) and / or can only classify landmarks into different categories. Therefore, the classification step can be used to classify or grade landmarks using a classifier (or classifier) that has been trained by creating a laser point cloud map comprising at least one laser point cloud. The laser point cloud map can be, for example, a Velodyne laser point cloud map provided by a laser scanner. In the classification step, if the landmark location or another landmark location is within a certain tolerance range in the laser point cloud map within a laser point cloud region representing a pillar or within another laser point cloud location representing a pillar, then a further matching method between the radar map and the laser point cloud map can be used to classify the landmark located at the landmark location or another landmark location, particularly where the landmark or landmark can be classified as a pillar. As mentioned above, a pillar can be understood as a vehicle-related infrastructure, such as a street light, a traffic light, and / or a traffic sign. A pillar can therefore be inferred using a representative laser point cloud in the area of this landmark location or another landmark location.

[0013] The method may further include a providing step, wherein if the landmark is classified as unclassifiable in the classification step, at least the location of the landmark or the location of the other landmark is provided for use in the vehicle's positioning map. The location of the landmark classified as unclassifiable in the providing step may also be directly stored in the vehicle's positioning map. This step may be used, for example, by the vehicle itself, another associated vehicle, or a backend server for subsequent classification of the landmark.

[0014] Furthermore, the method advantageously includes an elimination step in which, if a landmark or another landmark located at a landmark position is classified as unclassifiable in the classification step, said landmark position or said another landmark position is eliminated from the radar map. Thus, only the relevant pillars remain in the radar map.

[0015] The method can be implemented in the control unit, for example, in the form of software or hardware, or in a hybrid form of software and hardware.

[0016] In addition, the solution presented here proposes a device that is configured to execute, control or implement the steps of the variant of the method presented here in a corresponding device. The purpose on which the solution is based can also be achieved quickly and effectively through the implementation variant of the solution in the form of a device.

[0017] To this end, the device may include at least one computing unit for processing signals or data, at least one memory unit for storing signals or data, at least one interface to a sensor or actuator for reading sensor signals from a sensor or outputting data signals or control signals to an actuator, and / or at least one communication interface for reading or outputting data embedded in a communication protocol. The computing unit may be, for example, a signal processor, a microcontroller, etc., while the memory unit may be a flash memory, an electrically programmable read-only memory (EPROM), or a magnetic storage unit, etc. The communication interface may be configured for wireless and / or wired reading or outputting, wherein a communication interface capable of reading or outputting wired data may, for example, electrically or optically read such data from a corresponding data transmission line or output such data to a corresponding data transmission line.

[0018] Here, a device can be understood as an electrical device that processes sensor signals and outputs control signals and / or data signals accordingly. The device may have an interface that can be configured using hardware and / or software. In a hardware configuration, the interface can, for example, be part of a so-called application-specific integrated circuit (ASIC) system that contains the various functions of the device. However, it is also possible for the interface to be its own integrated circuit or to be at least partially composed of discrete components. In a software configuration, the interface can be, for example, a software module that exists alongside other software modules on a microcontroller.

[0019] In one advantageous embodiment, the device controls a deletion signal for deleting at least one landmark location from the radar map. For this purpose, the device may, for example, access sensor signals, such as a road position signal having at least one road position. The control is performed via actuators, such as a reader for reading the road position signal and a deletion device for outputting the deletion signal.

[0020] Also advantageous is a computer program product or a computer program having a program code, which can be stored on a machine-readable carrier or storage medium (such as a semiconductor memory, hard disk memory or optical memory) and is used to execute, implement and / or control the steps of the method according to any of the preceding embodiments, in particular when the program product or the program is run on a computer or device. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] An exemplary embodiment of the solution presented here is shown in the drawings and explained in more detail in the following description.

[0022] Figure 1 A schematic diagram of an apparatus for deleting at least one landmark position of a landmark in a radar map is shown according to an embodiment;

[0023] Figure 2According to one embodiment, a flow chart of a method for deleting at least one landmark position of a landmark in a radar map is shown;

[0024] Figure 3 A schematic diagram showing a method for matching a radar map with a topological road map according to an embodiment is shown;

[0025] Figure 4 A schematic diagram illustrating application of a classifier on a radar map according to an embodiment;

[0026] Figure 5 A schematic diagram of a radar map is shown; and

[0027] Figure 6 A schematic diagram showing a topological road map. DETAILED DESCRIPTION

[0028] In the following description of advantageous exemplary embodiments of the present solution, identical or similar reference numerals are used for similarly acting elements that are shown in the various figures, wherein a repeated description of these elements is omitted.

[0029] Figure 1 According to one exemplary embodiment, a schematic diagram of a device 100 for deleting at least one landmark position 102 of a landmark 105 in a radar map 110 is shown.

[0030] According to the present embodiment, apparatus 100 includes a reading device 115 and a deleting device 120. Reading device 115 is configured to read at least one road position 125 of at least one section of road 130, which is imaged in radar map 110. Deleting device 120 is configured to delete at least landmark position 102 of landmark 105 in radar map 110 when landmark position 102 is at least a predetermined distance 135 from road position 125.

[0031] According to this embodiment, landmark 105 is a tree. According to this embodiment, deletion device 120 deletes landmark position 102 of the tree in radar map 110 because landmark position 102 of the tree is farther from road position 125 of road 130 than predetermined distance 135 .

[0032] According to the present embodiment, radar map 110 is provided via satellite 140 and includes at least another landmark position 142 configured as another landmark 145 of a traffic sign. According to the present embodiment, deletion device 120 does not delete another landmark position 142 of the traffic sign from radar map 110 because another landmark position 142 of the traffic sign is less than predetermined distance 135 from road position 125. According to the present embodiment, the radar map changed by deleting landmark position 102 is configured for reading by vehicle 150.

[0033] The features of the device 100 are further described in more detail below:

[0034] A highly accurate and optimized positioning map is an important component for sufficiently accurate and stable positioning of a highly automated vehicle system for vehicle 150. The cost-effective creation and updating of such a highly accurate positioning map constitutes a major obstacle to profitable market introduction. To cost-optimize the creation of such a highly accurate positioning map, measurements of classified landmarks, for example, base points of pillars, in this case in the form of landmark positions 142, can be used by radar satellites, here satellites 140.

[0035] However, a significant challenge lies in performing the classification so that the satellite data can be used to create a highly accurate positioning map. The device 100 presented here uses radar measurements from satellites 140 in the form of a radar map 110 with landmark locations 125; 142, and according to another embodiment, a topological road map as input signals, thereby generating a modified radar map with classified, highly accurate pillar locations from the satellite measurements. This modified radar map can then be used to cost-optimize the generation of a highly accurate positioning map.

[0036] Unlike some known satellite systems, which create radar maps with highly accurate base points, preferably metal pillars, the apparatus 100 described herein can provide a modified radar map 110 by removing landmark locations 102 at least at a predetermined distance 135. Advantageously, the radar map 110 is no longer associated with man-made objects and can thus be directly used for optimizing the global positioning map. The subsequent selection of base points for pillars (such as streetlights, traffic lights, and / or traffic signs) can be automated, resulting in less time and lower costs.

[0037] Unlike some known systems that extract lane markings from aerial photography and use them for positioning, and in addition use standard image processing methods to detect clearly visible lane markings, the device 100 described herein achieves navigation results based on low noise content in radar map 110 and clearly distinguishable differences between pillars (e.g., base points) and other landmarks (e.g., man-made objects). Device 100 automatically classifies pillars in radar map 110. After classification, landmarks that do not represent pillar base points can be removed from radar map 100, and the positioning map can be optimized.

[0038] The purpose of the device 100 is to classify high-precision grid points in a given radar map of the satellite 140. This classification is the basic prerequisite for the following intended optimization of the vehicle's positioning map.

[0039] Figure 2According to one embodiment, a flow chart of a method for deleting at least one landmark position of a landmark in a radar map is shown. Figure 1 A method 200 is described that may be performed or controlled by a device.

[0040] Method 200 includes at least one reading step 205 and a deleting step 210. In reading step 205, at least one road position of at least one road segment, imaged in a radar map, is read. In deleting step 210, a landmark position is deleted from the radar map when the landmark position is at least a predetermined distance from the road position.

[0041] The method 200 optionally further comprises an identification step 215 , a classification step 220 , a provision step 225 and a elimination step 230 .

[0042] Identification step 215 is performed before reading step 205. In identification step 215, a road position is identified by a matching method between the radar map and a topological road map, which contains or images at least one further road position of a road segment. According to this embodiment, if a further road position imaged on the topological road map in the matching method corresponds to a road position on the radar map within a certain tolerance range, then this road position is identified in identification step 215, in particular, if this road position is represented in a darker color.

[0043] In classification step 220, at least the landmark imaged at the landmark location or another landmark imaged at another landmark location on the radar map is classified using a classifier. According to this embodiment, in classification step 220, the landmarks are classified using a classifier that was trained by creating a laser point cloud map including at least one laser point cloud. In this case, in classification step 220, if the landmark location or the other landmark location is located within a certain tolerance range in the laser point cloud region representing a pillar or in the region of another laser point cloud location in another laser point cloud representing a pillar, the landmark located at the landmark location or the other landmark location is classified using another matching method between the radar map and the laser point cloud map, in particular, the landmark or the other landmark is classified as a pillar.

[0044] If the landmark is classified as unclassifiable in the classification step 220 , then at least the landmark position or the further landmark position is provided for the vehicle positioning map in a providing step 225 .

[0045] If the landmark or the further landmark located at the landmark position is classified as unclassifiable in the classification step, the landmark position or the further landmark position is deleted from the radar map in an elimination step 230 .

[0046] The method 200 described is described in more detail below:

[0047] The method 200 presented here can be represented as a method for extracting pillar base points from high-precision radar satellite measurements in order to create a high-precision positioning map at low cost. This extraction can be understood in the sense of providing or highlighting.

[0048] The method steps or processing steps described herein are used to extract, i.e., provide a highly accurate global position of pillar objects from a radar map. Furthermore, according to this embodiment, in an identification step 215, dark areas (primarily asphalt roads) in the radar map are matched to lane alignments in a topological road map. This allows the approximate road alignment in the radar map to be known. In a subsequent deletion step 210, landmarks that are too far from the road are removed from the radar map. This leaves significantly fewer landmarks near the lanes, which are more likely to be classified by a vehicle than the visible landmarks. The radar map is then segmented, and in a classification step 220, a classifier is applied to the resulting areas. According to this embodiment, a training set for classifier training is created by creating a map composed of a Velodyne laser point cloud. The resulting laser point cloud map provides the positions of the pillars. Map segments of the laser point cloud map are then matched to a satellite radar map. After registration, points in the radar map that are compactly located around the pillar locations are labeled as pillars. The result is a radar map that contains the labeled pillar locations for smaller areas. The classifier can then be trained using known machine learning methods (e.g., deep learning) from this labeled data. As a result, the method 200 produces a radar map with high-precision pillar positions, which can then be used to optimize the global positioning map.

[0049] According to this embodiment, in providing step 225, unclassifiable landmarks or objects are additionally provided as null-type objects for the global positioning map to be optimized, or referenced into the positioning map. If these unclassified landmarks are detected or classified by different vehicles, this information can be aggregated in a backend server, and the relevant landmarks can be assigned a type later. These now-classified landmarks can then be used to further optimize the global positioning map for accuracy. Furthermore, it is possible to remove landmarks that are not visible to the vehicle's sensor system from the map and adjust their type.

[0050] In short, the method 200 presented here enables the automatic extraction of highly accurate pillar base points from radar satellite measurements for cost-optimized creation of positioning maps and a simple method for training the required classifiers.

[0051] The method steps described here can be performed repeatedly and in an order different from that described.

[0052] Figure 3According to an embodiment, a schematic diagram of a matching method of the radar map 110 and the topological road map 300 is shown. Figure 2 Schematic diagram of the deletion steps.

[0053] The topological map 300 is matched with the radar map 110 (radar raw map) to identify roads 130 and discard the associated landmark positions 102 of landmarks that are beyond a maximum distance 135 from the roads 130. The associated landmark positions 102 of landmarks located beyond the distance 135 may also be represented as blobs or radar blobs.

[0054] Figure 4 According to one embodiment, a schematic diagram of the application of a classifier on a radar map 110 is shown. Figure 2 Schematic diagram of the classification step, in which the radar map 110 is matched to the laser point cloud map 400 by means of another matching method.

[0055] The previously trained classifier is applied to the remaining number of landmark positions, i.e., the other landmark positions 142 of the other landmark that are not located outside the distance. In another matching method, the other landmark positions located in the laser point cloud area representing the pillar are classified as pillars 405, which can also be represented as classified radar spots.

[0056] Figure 5 1 shows a schematic diagram of a radar map 110. In the unprocessed radar map 110, which contains many artifacts in addition to high-precision grid points (such as pillars), a road 130 is imaged as a dark color.

[0057] Figure 6 A schematic diagram of a topological road map 300 is shown. Figure 3 The topological road map 300 is shown. Figure 5 In comparison to the radar map shown, an imprecise topological road map 300 is used to overlay the radar map to identify the road 130 in the radar map.

[0058] If an exemplary embodiment includes the conjunction “and / or” between a first feature and a second feature, this can be interpreted as meaning that the exemplary embodiment has both the first and the second feature according to one exemplary embodiment, and has only the first feature or only the second feature according to another exemplary embodiment.

Claims

1. A method (200) for deleting at least one landmark position (102) of a landmark (105) in a radar map (110), wherein the method (200) comprises at least the following steps: - reading (205) at least one road position (125) of at least one road section of a road (130) imaged in the radar map (110); and - when the landmark position (102) is at least a predetermined distance (135) from the road position (125), deleting (210) the landmark position (102) from the radar map (110), The method further comprises an identification step (215) performed before the reading step (205), in which the road position (125) is identified by a matching method of the radar map (110) with a topological road map (300), the topological road map having or imaging at least one further road position of the section of the road (130).

2. The method (200) according to claim 1, wherein the road position (125) is identified in the identification step (215) if a further road position imaged on the topological road map (300) in the matching method corresponds within a tolerance range to the road position (125) on the radar map (110).

3. The method (200) according to claim 1 or 2, comprising a classification step (220), wherein the landmarks (105; 145) imaged on the radar map (110) at least at the landmark position (102) or at another landmark position (142) are classified by a classifier.

4. The method (200) according to claim 3, wherein in the classification step (220) the landmarks (105; 145) are classified by using a classifier, which has been trained by creating a laser point cloud map (400) comprising at least one laser point cloud.

5. The method (200) according to claim 4, wherein if the landmark position (102) or the other landmark position (142) is located in the laser point cloud area representing a pillar or in another laser point cloud position area of another laser point cloud representing a pillar in the laser point cloud map within a tolerance range, then in the classification step (220), the landmark (105; 145) located at the landmark position (102) or at the other landmark position (142) is classified by another matching method of the radar map (110) and the laser point cloud map (400).

6. The method (200) according to claim 3, comprising the step of providing (225) at least the landmark position (102) or the further landmark position (142) for a positioning map of the vehicle (150) if the landmark (105; 145) is classified as unclassifiable in the classification step (220).

7. The method (200) according to claim 3, comprising the step of eliminating (230) the landmark position (102) or the further landmark position (142) from the radar map (110) if the landmark or the further landmark (105; 145) located at the landmark position (102; 142) is classified as unclassifiable in the classification step (220).

8. The method (200) of claim 2, wherein the road location (125) is represented in a dark color in the radar map (110).

9. The method (200) of claim 5, wherein the landmark (105) or the further landmark (145) is classified as a pillar (405).

10. An electrical device (100) configured to execute and / or control the steps of the method (200) according to any one of claims 1 to 9 in respective units (115; 120).

11. A machine-readable storage medium having a computer program stored thereon, the computer program being configured to execute the method (200) according to any one of claims 1 to 9.

Images