Method and device for generating navigation data

Abstract

The invention relates to a method for generating navigation data. The method comprises: generating and / or providing (101) the data BD; segmenting (102) a number M of line elements Lm from the data BD; subdividing (103) each of the line elements L m into a predefined number N of equidistant line sections LA m-n , said number being identical for all line elements L m ; for a line element L m , determining (104) similar line elements L m arranged in the environment (I) thereof; in the environment (I), clustering (105) all line elements L m* similar to L m in non-overlapping fashion to form a local cluster CL p = CL p (L m , L m* , (I); repeatedly carrying out (106) steps (104) and (105) for all line elements L m not already assigned to a local cluster CL p ; for each local cluster CL p , on the basis of the similar line elements L p,k clustered therein, determining (107) a line element LE p representing this cluster by averaging MI over all line elements LE p = MI(LE p,1 , LE p,2 ,..., LE pK(p) ); based on the determined line elements LE p and the data BD, determining (108) boundaries of roadways depicted in the data BD; and generating (109) and outputting navigation data comprising the determined roadway boundaries.

Description

[0001] Method and device for generating navigation data

[0002] The invention relates to a method and a device for generating navigation data, in particular from noisy sensor data. Furthermore, the invention relates to a computer system, a digital storage medium, a computer program product, and a computer program.

[0003] It is known in the art to generate navigation data, particularly for creating highly accurate digital navigation maps, based on aerial and satellite imagery. The respective input sensor data (e.g., from imaging processes and 3D measurement data such as from LiDAR sensors) are advantageously georeferenced. Segmentation methods are used to extract lane alignments and lane boundaries; in particular, line element segmentation, such as...

[0004] Segmentation methods based on Hough transforms are well-known. The identification of line elements through segmentation in image data is impaired by noise in the sensor data. This "noise" refers to random fluctuations, for example, in the brightness or color of camera sensors or in the distance measurements of LiDAR sensors. Noise can be particularly disruptive in image data with low signal strength (e.g., in medical images, satellite images), hindering the identification of line elements.

[0005] The object of the invention is to provide an improved method and a corresponding device for generating navigation data from partially noisy sensor data.

[0006] The application of the invention generally relates to the field of spatiotemporal processing of sensor data and data analysis, in particular areas of building mapping, earth sensing, satellite remote sensing, aircraft remote sensing, especially application in the field of autonomously controlled vehicles, and furthermore to the field of generating navigation data for pedestrians, road, rail and / or water vehicles in open terrain or in the interior of buildings.

[0007] The invention is defined by the features of the independent claims. Advantageous further developments and embodiments are the subject of the dependent claims. Further features, applications, and advantages of the invention will become apparent from the following description and the explanation of exemplary embodiments of the invention illustrated in the figures.

[0008] A first aspect of the invention relates to a method for generating navigation data. The method comprises the following steps.

[0009] In a first step, data BD is generated that depicts a portion of the Earth's surface. This portion of the Earth's surface includes, in particular, at least one roadway in the sense of a navigable route, including, for example, paths in natural terrain such as hiking trails in mountains. The data BD is advantageously referenced to a predefined 2D, 3D, or even higher-dimensional, abstract reference system, such that each data point or data voxel in the data BD is assigned a 2D, 3D, or n-dimensional position. Advantageously, the reference system is a georeference system (e.g., WGS84, GRS80). The data BD is advantageously camera images, especially multi- or hyperspectral image or sensor data from satellite, aircraft, or vehicle-borne sensors.

[0010] In an advantageous application, the BD data is generated by a sensor mounted on a vehicle (e.g., a road vehicle) (e.g., a camera sensor, a LiDAR sensor, a radar sensor, an ultrasonic sensor, or a combination thereof). The BD data can also be referenced to a local reference system, e.g., when generating navigation data for building interiors or, for example, to a coordinate center point of the vehicle carrying the sensor.

[0011] In a second step, a number AT of line elements L are segmented. m from the data BD, for example, but not limited to, information derived from multi- or hyperspectral images or from LiDAR sensor data using an AI-supported segmentation method (see, for example, Convolutional Neural Networks or Vision Transformers), using multi- or hyperspectral images or LiDAR sensor data, where each line element L m in the

[0012] Data BD an initial position and a final position as well as a

[0013] Environment UMG(P(L m ) is assigned, where P^L^ is a position of the line element L m BD specifies in the data, with m = 1, 2,

[0014] The term "data BD" generally encompasses digital data, but especially image data, arranged in a two-dimensional or higher-dimensional array / raster / data space (e.g., in a 2 x 2, 3 x 4, or ax b array / raster / data space), with a, be ∈ N\{0}. The term "data BD" specifically includes 2D or 3D image data.

[0015] The term "line element L" m“ describes in this case an (abstract) connected one-dimensional, i.e. line-like structure without transverse extension depicted in the data BD (e.g. a longer, straight, curved or kinked line / stroke; a straight line, a curve, a series of points, a line segment, an outline, an edge, etc.).

[0016] In the data BD there is a number M of such distinguishable lines, i.e., line elements L. m depicted, e.g. seven (M = 7) individual curved lines L m with m = 1, 2, 3, 4, 5, 6, 7.

[0017] The term "segmentation" is used here as is customary in image data analysis. It refers specifically to the creation of conceptually related regions (here, in particular, lines, edges, etc.) in the data BD by grouping adjacent pixels or voxels according to a predefined homogeneity criterion. Methods for segmenting lines, edges, etc., are well known in the art. Reference is made to the relevant literature. Segmentation thus transforms line elements L depicted in or represented by the data BD into distinct, homogeneous regions. m determined.

[0018] Each of the segmented line elements L m has a vectorial initial position in the data array / raster / space and a vectorial end position Furthermore, each line element is assigned L mA position P^L^ representing this line element in the data BD is assigned. The position P^L^ of the line element L m In the data BD, it is advantageous to have the starting position P^L^, the ending position P^L^, or a position between the starting position P^L^ and the ending position P^L^ on the line element, in particular the midpoint between the starting position P^L^ and the ending position P^L^ on the line element. The position P^L^ could also, for example, be a geometric centroid of the line element L. m be and therefore outside the line element L m The positions P^L^, P^L^, P^L^ may be specified, for example, by position vectors of individual image pixels, image voxels, ... in the data BD.

[0019] The term “environment UMG(P(L m ) “of a line element L m Advantageously defines in the present case an area in the data BD in the immediate vicinity of the line element L. m , which is the line element L mIt contains everything. The environment UMG(P(L m ) can advantageously be defined by a predefined radius around the position P^L^. The neighborhood UMG(P(L) m of the line element L m can alternatively be achieved by a radius R around a midpoint of the line element L m It is defined, this can be on the line element or outside the line element L m lie. Of course, all other known definitions of a neighborhood of the line element L. m Including the idea of ​​the inventor.

[0020] In the third step of the process, each of the line elements L is subdivided. m into a predefined line element L determined for all line elements in the data BD m each identical number N of equidistant line segments LA n ", where each line segment LA nn in the BD data through a starting position and an end position x m n E is defined as:

[0021] LA m n : = ( „ „ , x m “ “ ), with n = 1, 2, N, N >1 and L = V LA m n .

[0022] Example: Let the data BD contain a total of five (M = 5) line elements L m The segments have been formed. It is further specified that N = 10. Each of these five line elements L m LA is now divided into ten equidistant line segments. m "subdivided, with each line segment LA m "In the data BD, a starting position x m ^ s and has a final position. The term "equidistant line segments LA" m , “ is understood to mean that each individual one-dimensional line element L m into ten equal line segments LA m " is divided. The length of the line segments, e.g., LA m=1 , n of the line element L m=1 are identical for n = 1, 2, ..., 10. The length of the line segments, e.g., LA, m=1 , n and LA m= ntwo different line elements L m=1 and L m=2 are only identical if the different line elements L m=1 and L m=2 are of equal length. The sum over n of the individual line segments LA m=1 , n This results in the line element L m=1 In a fourth step of the procedure, for a first of the line elements L,“ , e.g. for the line element L m=l , a determination of in its respective environment UMG P(L m y) arranged similar line elements L m *, where a similarity of line elements Lm and L™* is based on a predefined distance metric DM(Lm, L m *) is determined based on a function F of distances d(LA m , n , LA m * n *) of line segments LA m , n and the line elements L m and L™* based:

[0023] DM( m Lm*) — F(d(LAm,m LAm*,n*)), where d(LAm ,n, LA m * n *) a distance between the line segment LA m , n , LA m *, n * and the line segment specifies, and where: m* m ≤ {1, 2, M} and m* m, n*, n {1, 2, N} such that the distance metric DM(L m , Lm*) between L m and L m * with m* m through a bridge of N line segments LA m , n in N line segments for all n*=n is determined by

[0024] ( \X m «*,n*,E r ~ X m «*,n, *S ) ' ' (X m,n,E ~ X m,n,S ' > 0

[0025] 5 and where the line elements L," and L m * only be considered similar if

[0026] DM(Lm, L m *) satisfies a given condition BED.

[0027] Each of the line segments LA m ," and has a vectorial initial position X in the data array / raster / space of the data BD.m n S or and a vectorial end position X m n E or X mt , ni:)E The distance d between line segments LA m , n and is therefore based on the vector distance d^ between the initial positions X mn S or X m * n * s and the vector distance d ± z between the end positions X mn E or X m * )ni:)E As a condition of the distance metric DM(L m , Lm*) must still be the scalar product of the vectors sejn

[0028] The condition BED is advantageous due to the relation DM(Lm, Lm*) < DM max with a predefined, constant value or a value learned through machine learning DM max >0 implemented.

[0029] In this fourth step, therefore, in an environment UMG PL m one of the line elements L min the data BD for the line element L, "similar line elements L m * determined.

[0030] In a fifth step of the procedure, all in the vicinity of UMG PL are m found to L m similar line elements L,“* to a local cluster CL P = CL p (Lm, L m *, UMG P(L m ))) non-overlapping clusters to initiate noise reduction.

[0031] The term "clustering" in this context refers to the fact that similar line elements L m and Lm* in the same cluster CL P possess “similar” properties, i.e., are similar to each other in the sense of the above definition, and differ from objects that are not in the same cluster CL P are distinguished by these similar properties. The term "non-overlapping clustering" here means that each of the line elements L m only a single cluster CL Pis assigned.

[0032] In a sixth step of the process, steps four and five described above are now applied to all clusters not already part of a local cluster CL. P assigned line elements L m repeated until all line elements L m a local cluster CL P are assigned, so that the result is a number P of local clusters CL P is determined where p = 1, 2P and P > 1, and where each local cluster CL P a number K(p) line elements L p , k includes: with k = 1, 2, ..., K(p) and K(p) > 1. This continues the noise reduction of the sensor data.

[0033] In a seventh step of the procedure, CL is calculated for each of the P local clusters. P based on the similar line elements L clustered therein p ,k determining a line element representing this cluster LE PAs a result, a number P of line elements LE are generated. P determined.

[0034] Advantageously, the line element LE representing this cluster is determined P by averaging or weighted averaging MI over all line elements LE P =MI(LE Ptl , LE P 2, ... , LE pK(p) This continues the automatic noise reduction of the sensor data on the device being described.

[0035] In an eighth step of the process, LE is calculated based on the determined line elements. P Determining the boundaries of roadways depicted in the BD data. These roadway boundaries can be defined by lateral road edges, guardrails, fences, natural structures such as riverbanks, interior walls, etc. Determining the roadway boundaries based on the identified line elements LE. PData processing (BD) is advantageously carried out using a suitably trained machine learning algorithm. This further enhances the automatic noise reduction of the sensor data on the device being described.

[0036] In a ninth step of the process, navigation data comprising the determined lane boundaries is generated and output or provided. The generated navigation data is characterized in particular by a high positional accuracy of the lane boundaries specified therein. This advantageously includes the automatic noise reduction of the sensor data on the device being described. Advantageously, the navigation data includes, in addition to the determined lane boundaries, at least parts of the data BD. Advantageously, the navigation data includes the data BD with the P line elements LE superimposed therein. P .

[0037] Advantageously, the navigation data is displayed on an optical output device. This can be done using an LCD / LED / OLED / QLED display, touchscreen, projector, head-up display, AR / VR glasses, or holographic display. A user of the method can then advantageously interact with the automatically denoised, particularly vector-based, navigation data by editing and navigating. This adds a manual, user-interactive noise reduction feature to the sensor data. Furthermore, the P line elements LE P . advantageously provided or output for further use, advantageously in the form of digital data, e.g. in georeferenced GeoJSON format.

[0038] In another application, the generated navigation data, along with lane markings, is used to control a partially or fully autonomous vehicle (SAE Level 1 to 5). Advantageously, the data is acquired using a sensor mounted on the vehicle. Further processing of the acquired data, up to and including the generation and provision of the navigation data, also takes place within the vehicle. This means the navigation data is generated "onboard" and can be used directly to control the autonomous vehicle without requiring additional data transmission bandwidth or extra computing resources ("edge computing"). The reference system to which the data refers in this case is advantageously a local reference system, for example, with a vehicle reference point as the coordinate origin. Of course, data from the vehicle sensor and, for example, other sources can also be used to generate the navigation data.The provided satellite data will be merged.

[0039] The proposed method enables significantly lower-noise segmentation and clustering of line elements L compared to the state of the art. m The provided data includes BD (database mapping), enabling the generation of navigation data with higher positional accuracy for the specified lane boundaries. This navigation data thus allows for greater precision, for example, in the control of autonomous vehicles. Furthermore, vectorizing the navigation data allows for easy interaction with human users via specialized software or standard geodata processing programs with a graphical user interface.

[0040] An advantageous further development of the procedure is characterized by the fact that the distance metric DM(L) m , L m ») is a modified Hausdorff distance metric, where:

[0041] DM(L m , Lm ») — Advantageous are all line elements L," straight lines, where for these line elements L," the following holds: N=l.

[0042] An advantageous further development of the procedure is characterized by the fact that the determination of the respective cluster CL P representing line element LE P with starting positions P a (LA p li n ) of line segments LA p li n and end positions P e (LA p li n ) of the line segments LA p>k n by one with a respective length the line segment elements LA p , k , n weighted averaging MI according to: and

[0043] P a Starting position of the averaged line segment element LA P “P e LA p n y. End position of the averaged line segment element LA p , nfor cluster p such that the averaged line element for cluster / ? is defined as follows:

[0044] LA p , n := (P a (LA Ptn ),P e (LA Ptn )), with n = 1, 2,..., N, N > 1 and

[0045] An advantageous further development of the process is characterized by the ability to use an artificial intelligence approach for each step. For example, an artificial intelligence algorithm, such as a machine learning model, can be used to predict the corresponding cluster for each line element. Similarly, an artificial intelligence method could be used to determine the representative line element of each cluster.

[0046] Another aspect of the invention relates to a device for generating navigation data and for carrying out one of the methods described above. The device comprises: a first means for generating and / or providing data BD in which a part of the Earth's surface is depicted, a segmentation means for segmenting a number AT of line elements L m from the data BD, where each line element L m in the BD data

[0047] Starting position and a final position as well as an environment

[0048] UMG(P(L m ) is assigned, where P^L^ is a position of the line element L m BD specifies in the data, with m = 1, 2, and where that

[0049] Segmentation means are implemented and set up for this purpose:

[0050] - each of the line elements L m into a predefined and for all line elements L meach identical number N of equidistant line segments LA mi "to subdivide, with each line segment LA mi " is defined in the data BD by a start position T " " and an end position T " ":

[0051] - for a line element L m from within its respective environment

[0052] UMG(P(L m ) similar line elements arranged Z m * to determine, whereby a similarity of line elements L m and L m * based on a predefined distance metric DM(L m , L m ») is determined, which is based on a function F of distances d(LA m .., LA mV ) of line segments LA m "and LA m * n * of the line elements L m and L m * based on: DM(L m , L m *) = F(d(LA m "LA m * n *)), where d(LA m , n , LA m »n ») a distance between the line segment LA m , n LA m * n * and the LA line section m * n * indicates, and where:

[0053] LA m ,„: = ( Xm ' n ' s , Xm ' n ' E ) and m* mc {1, 2, ... , M} and m* m, n* ne {I, 2, ..., N} such that the distance metric DM(L m , L m ») between L m and L m * with m * m through a bridge of N line segments LA m "in N line segments LA" m * n * for all n*=n is determined by and wherein the line elements L," and L m * only be considered similar if DM(Lm, L) m *) satisfies a given condition BED; a clustering tool advantageously designed and configured to operate in the environment UMG(P(L) m)) e\n non-overlapping clustering of all line elements similar to L," L m * to a local cluster CL P = CL lr (L m , L m *, UMG P(L m y) ) to execute; where non-overlapping clustering applies to all clusters not already in a local cluster CL P assigned line elements L," is executed, resulting in a number P of local clusters CL P is determined where p = 1, 2P and P > 1, where each local cluster CL P a number K(p) line elements L p , k includes: = A , with k = 1, 2, K(p) and K(p) > P, and for each local cluster CL P based on the similar line elements L clustered therein p ,k one of these cluster's representing line elements LE P to determine; and a second means designed and set up for this purpose, based on the determined line elements LE Pto determine the boundaries of lanes depicted in the BD data; and a third means designed and configured to generate and output navigation data that includes the determined lane boundaries.

[0054] The third method is advantageous because it allows the navigation data to be edited in interaction with a human user.

[0055] The first device advantageously comprises a sensor for acquiring raw / sensor data. The sensor is advantageously a camera sensor, in particular a multispectral or hyperspectral image sensor. The sensor is also advantageously a remote sensing sensor (radar, LiDAR, optical camera, etc.). The sensor can be located on board a satellite, an aircraft, or a vehicle (road vehicle). The first device advantageously includes preprocessing the acquired raw / sensor data into the data BD. This preprocessing advantageously includes data conversion and formatting: e.g., color space transformation (e.g., RGB to grayscale, HSV, or LAB), normalization of the values ​​(e.g.,...)Adjusting pixel values ​​from 0-255 — > 0-1 or -1-1) or image size ("resizing", "cropping", "padding"), converting LiDAR laser measurements into distance values ​​(sensor to reflecting object) with physical units, data calibration, noise reduction and quality improvement: e.g., by smoothing / filtering (e.g., Gaussian filter, median filter), contrast adjustment (histogram equalization), or by sharpening or blurring, geometric correction: e.g., straightening distortions (e.g., by camera lens), rotation, mirroring, scaling, or perspective correction, feature or data preparation: e.g., by extracting relevant image sections (ROI - Region of Interest), by data augmentation (in AI: artificial generation of variants by rotating, mirroring, etc.).) or by feature extraction (edges, corners, textures, color distributions) a correction of lighting effects, background subtraction, and noise from atmospheric processes.

[0056] The second method advantageously uses a suitably trained machine learning algorithm to determine the lane boundaries.

[0057] The third method is advantageously designed to supplement provided navigation data with the determined lane boundaries or to replace lane boundary data contained in the navigation data with the determined lane boundaries.

[0058] The first means of providing the BD data includes, for example, a satellite-borne, an aircraft-borne, or a vehicle-borne sensor, or a combination thereof. If a sensor combination is used, the sensor data acquired are fused to generate the BD data. The third means is advantageously combined with the control system of a partially or fully autonomously controlled vehicle, in particular a road vehicle with a control system according to SAE Level 1 to 5.

[0059] Another aspect of the invention relates to a partially or fully autonomously controllable vehicle, i.e. with a control system according to SAE levels 1 to 5, with a device as described above.

[0060] Another aspect of the invention relates to a computer system with a data processing device, wherein the data processing device is configured such that a method as described above is carried out on the data processing device.

[0061] Another aspect of the invention relates to a digital storage medium with electronically readable control signals, wherein the control signals can interact with a programmable computer system in such a way that a method as described above is carried out.

[0062] Another aspect of the invention relates to a computer program product with program code stored on a machine-readable medium for carrying out the method as described above when the program code is executed on a data processing device.

[0063] Another aspect of the invention relates to a computer program with program codes for carrying out the method as described above, when the program runs on a data processing device, for which the data processing device can be designed as any computer system known from the prior art.

[0064] Further advantages, features, and details will become apparent from the following description, in which—possibly with reference to the drawing—at least one embodiment is described in detail. Identical, similar, and / or functionally equivalent parts are marked with the same reference numerals. It shows:

[0065] Fig. 1 shows a highly schematic process flow of an embodiment of a method according to the invention, and

[0066] Fig. 2 shows a schematic representation of a proposed device.

[0067] Fig. 3a-g shows a schematic representation of a process flow in an exemplary implementation for Earth observation.

[0068] Fig. 1 shows a highly schematic process flow of a method according to the invention for generating navigation data. The method comprises the following steps.

[0069] In step 101, data BD is generated using an aircraft- or satellite-borne multispectral sensor or a LiDAR remote sensing sensor. This data BD depicts a portion of the Earth's surface with a road network. The road network comprises lanes for road vehicles and generally also includes paths in natural terrain. In this embodiment, the data BD consists of multispectral 2D image data. Raw data generated by the remote sensing sensor is advantageously pre-processed and georeferenced before being provided as data BD. The remote sensing sensor could be, for example, a radar sensor, a camera sensor, or a combination of both. The objective is to determine lane boundaries and road markings depicted in the data BD with high accuracy and, in particular, low noise, and to generate highly accurate navigation data based on this determination.

[0070] For this purpose, in step 102 a segmentation of a number AT of line elements L is carried out. m from the data BD using a line segmentation algorithm known in the prior art, wherein each line element L m in the BD data an initial position and a final position as well as an environment

[0071] UMG(P(L m ) is assigned, where P^L^ is a position of the line element L m BD specifies in the data, with m = 1, 2,

[0072] In step 103, each of the M line elements L is subdivided. m into a predefined and for all line elements L m each identical number N of equidistant line segments LA n ", where each line segment LA mi "in the data BD by a starting position x m n s and an end position x m n E is defined as:

[0073] LA m ,":= ( x m n S , X m n E), with n = 1, 2,..., N, N >1 and

[0074] In step 104, for a line element Lm, a determination is made of the properties in its respective environment UMG PL. m )) arranged similar line elements Lm* , wherein a similarity between line elements Lm and L m * based on a predefined distance metric DM(L m , L m *) is determined based on a function F of distances d(LA m ,n, LA m *,n*) of line segments LA m , n and the line elements Lm and L m * based on: DM(L m , Lm*) = F(d(LA m , n , LA m * n *)), where d(LA m , n , LA m * n *) a distance between the line segment LA m , n , LA m *, n * and the LA line section m *, n * indicates, and where:

[0075] LA m ,n := * m - n - S , m ' n ' E ) and LA m *, n := * m *- n *- s , m *'"*' £ ) m* m ≤ {1, 2, ..., M} and m* L m, n*, n {1, 2, ..., N} such that the distance metric DM(L m , Lm*) between L m and L m * with m*tm through a bridge of N line segments LA m , n in N line sections LA m *, n * for all n*=n is determined by and wherein the line elements Lm and L m * are only considered similar if DM(Lm, Lm*) satisfies a given condition BED.

[0076] In step 105, in the environment UMG(P(L) m )) a non-overlapping clustering of all line elements Lm* similar to Lm into a local cluster CL P = CL p (Lm, Lm*, UMG(P(L m )) ).

[0077] In step 106, steps 104 and 105 are executed again for all clusters not already in a local cluster CL. P assigned line elements Lm such that the result is a number P of local clusters CL P is determined where p = 1, 2P and P > 1, and where each local cluster CL P a number K(p) line elements L p , k includes: with k = 1, 2, . K(p) and K( P ) > 1. In step 107, CL is performed for each local cluster P based on the similar line elements L clustered therein p k Determining a line element LE representing this cluster P by averaging MI over all line elements LE P =MI(LE pl , LEp'2, , LE pK ( P )) .

[0078] In step 108, based on the determined line elements LE, Pand the data BD, a determination of the boundaries of the roadways depicted in the data BD. These boundaries advantageously include the lateral edges and boundaries of the roadways by road markings (solid or dashed lines), guardrails, barriers, boundary objects, etc.

[0079] In step 109, navigation data is generated and output, including the determined lane boundaries.

[0080] Advantageously, the generated navigation data includes the data BD with the line elements LE superimposed within it. P The line elements LE P are advantageously displayed in a color that differs from the image data BD, thus improving the visibility of the line elements LE P improved. The line elements L generated according to the proposed robust and low-noise method mDue to the lower noise, they have significantly higher data quality and data reliability.

[0081] The navigation data includes, in particular, the line elements LE, which were determined in a low-noise and robust manner according to the proposed method. P This includes line elements representing guardrails, lane boundaries, lane markings, or similar features. These line elements, displayed in the navigation data, are LE. P They can be used advantageously, for example, as navigation data for autonomously controlled vehicles or, through interaction with a person, to support their orientation in space.

[0082] Fig. 2 shows a schematic proposed device for generating navigation data.

[0083] The device comprises a first means 201 for generating and / or providing data BD in which a portion of the Earth's surface is imaged. In this embodiment, the first means comprises a satellite-borne multispectral image sensor. The device further comprises a segmentation means 202 for segmenting a number M of line elements L. m from the data BD, where each line element L m in the BD data an initial position and a final position as well as a

[0084] Environment UMG(P(L m ) is assigned, where P^L^ is a position of the line element L m BD specifies in the data, with m = 1, 2, and wherein the segmentation means

[0085] (202) Each of the line elements L is designed and set up for this purpose m into a predefined and for all line elements L m each identical number N of equidistant line segments LA mi"to subdivide, with each line segment LA mi "in the BD data through a starting position and an end position x m n E is defined as: with n = 1, 2, N, N > 1 and L = V LA m n and for a line element L m from in its respective environment UMG(P(L m ) similar line elements arranged L m * to determine, whereby a similarity of line elements L m and L m * based on a predefined distance metric DM(L m , L m *) is determined based on a function F of distances d(LA m .., LA mV ) of line segments LA„,,. and LA m * n * of the line elements L m and L m * based on: DM(L m , L m *) = F(d(LA m "LA m * n *)), where d(LA„ ., LA mV ) a distance between the line segment LA m "LA m * n* and the LA line section m » n * indicates, and where:

[0086] LA m , n : = ( Xm ' n ' s , Xm ' n ' E ) and m* mc {1, 2, ... , M} and m* m, n* ne {I, 2, ..., N} such that the distance metric DM(L m , L m ») between L m and L m * with m * m through a bridge of N line segments LA m "in N line segments LA" m * n * for all n*=n is determined by: and wherein the line elements L m and L m » only be considered similar if DM(L m , L m ») a given condition BED satisfies

[0087] The device further comprises a clustering agent 203, which is designed and configured to operate in the environment UMG(P(L) m ) a non-overlapping clustering of all to L msimilar line elements L",* to a local cluster CL P = CL p (Lm, Lm*, UMG PL m to execute; where non-overlapping clustering applies to all clusters not already in a local cluster CL P The assigned line elements Lm are executed, resulting in a number P of local clusters CL. P is determined where p = 1, 2P and P > 1, where each local cluster CL P a number K(p) line elements L p , k includes: k = 1, 2, K(p) and K(p) > r, and for each local cluster CL P based on the similar line elements L clustered therein p ,k one of the line elements representing this cluster LE p to determine.

[0088] The device comprises a second means 204, which is designed and configured to operate based on the determined line elements LE. Pto determine the boundaries of roadways depicted in the data BD; and a third means 205 which is designed and equipped to generate navigation data including the determined roadway boundaries, to enable editing by a device user (manual noise reduction) and generally to output digitally for further processing.

[0089] The resources 201 to 205 are advantageously each implemented as independent systems that interact as described.

[0090] Figs. 3a-f show a schematic representation of a process flow in an exemplary implementation for Earth observation.

[0091] In Fig. 3a, a geographical scene on the Earth's surface 300 with road and path networks, such as cycle and hiking trails 302, roads 303 with bridges 304, navigable waters 305 and corridors 307 in buildings 306 is first depicted by radar, laser, camera or GPS sensors, from which vectorized and georeferenced data BD are acquired and made available on carrier platforms such as: mobile phones 308, satellites 309, road vehicles 310, wheels or people 311, as well as ships 312 (cf. 201).

[0092] These data BD are generally noisy and, after segmentation according to the state of the art, would show correspondingly noisy vector data 312, 313, 316 in a georeference system 301, cf. Fig. 3b. Fig. 3c shows vector data of the line elements LE generated by the proposed method (cf. process steps 102-107). P .

[0093] Fig. 3d shows an advantageous further interactive noise reduction of the line elements LE. P from Fig. 3c by means of an interactive computer system 331 which, in addition to classic elements such as touchscreens, keyboards, virtual reality glasses, etc., may include methods and devices of artificial intelligence such as an AI agent with speech interaction.

[0094] Fig. 3e-g show various navigation data for road network maps 341, hiking maps 342, cadastral maps 343, which can be advantageously supplemented and extended with appropriate (text) annotations.

[0095] Although the invention has been further illustrated and explained in detail by means of preferred embodiments, the invention is not limited by the disclosed examples, and other variations can be derived from them by a person skilled in the art without departing from the scope of protection of the invention. It is therefore clear that a multitude of possible variations exist. It is also clear that the embodiments mentioned as examples are truly only examples and are not to be understood in any way as limiting, for example, the scope of protection, the possible applications, or the configuration of the invention.Rather, the preceding description and the description of the figures enable the person skilled in the art to implement the exemplary embodiments in concrete terms. With knowledge of the disclosed inventive concept, the person skilled in the art can make numerous modifications, for example, regarding the function or the arrangement of individual elements mentioned in an exemplary embodiment, without leaving the scope of protection defined by the claims and their legal equivalents, such as a further explanation in the description. List of reference numerals.

[0096] 101 to 109 procedural steps

[0097] 201 first means of generating and / or providing data BD in which a part of the Earth's surface is depicted

[0098] 202 Segmentation tools

[0099] 203 cluster funds

[0100] 204 second means

[0101] 205 third means

Claims

Patent claims 1. Method for generating navigation data, comprising the following steps: 1.

1. Generating (101) data BD in which a part of the Earth's surface is depicted; 1.

2. Segmenting (102) a number AT of line elements L m from the data BD, where each line element L m in the BD data an initial position and a final position as well as an environment UMG(P(L m ) is assigned, where P^L^ is a position of the line element L m in the data BD, with m = 1.

3. Subdivide (103) each of the line elements L m into a predefined and for all line elements L m each identical number N of equidistant line segments LA n ", where each line segment LA mE in the BD data through a starting position and an end position x m n E is defined as: LA m n : = (T „ „ , x m “ “ ), with n = 1, 2, N, N >1 and 1.

4. for a line element L m Determine (104) of in its respective environment UMG(P(L m ) arranged similar line elements L m * , where a similarity of line elements L m and L m * based on a predefined distance metric DM(L m , L m ») is determined, which is based on a function F of distances d(LA m .., LA mV ) of line segments LA n "and LA m * n * of the line elements L m and L m * based on: DM(L m , L m ») = F(d(LA mE , LA m » n *)), where d(LA mE , LA m » n ») a distance between the line segment LA m "LA m * n * and the LA line section m * n * indicates, and where: m* mc {1, 2, ... , M} and m* m, n* ne {I, 2, ..., N} such that the distance metric DM(L m , L m ») between L m and L m * with m * m through a bridge of N line segments LA m "in N line segments LA" m * n * for all n*=n is determined by E and and wherein the line elements L," and L m * only be considered similar if DM(Lm, L) m *) satisfies a given condition BED; 1.

5. in the vicinity l G(P(L m )) non-overlapping clustering (105) of all to L m similar line elements L m * to a local cluster CL P = CL^Lm, L m *, UMG(P(L )); 1.

6. Repeated execution (106) of steps 1.

4. (104) and 1.

5. (106) for all not already in a local cluster CL P assigned line elements L m, so that the result is a number P of local clusters CL P is determined where p = 1, 2, ..., P and P > 1, and where each local cluster CL P a number K(p) line elements L p , k includes: and K(p) > P 1.

7. for each local cluster CL P based on the similar line elements L clustered therein p ,k Determine (107) one of the line elements representing this cluster LE P , 1.

8. based on the determined line elements LE P and the data BD Determine (108) limits of carriageways depicted in the data BD; and 1.

9. Generating (109) and outputting navigation data including the determined lane boundaries.

2. Method according to claim 1, wherein the generated data BD are georeferenced in step 1.

1.

3. Method according to claim 1 or 2, wherein the position P(L m)of the line element L,“ in the data BD the starting position P fl (L m ) or the final position P e (L m ) or a position between the starting position P fl (L m ) and the final position P e (L m ) on the line element, in particular the midpoint between the starting position P fl (L m ) and the final position P e (L m ) on the line element.

4. Method according to any one of claims 1 to 3, wherein the environment UMG P(L m ) by a predefined radius around position PL m ) is defined.

5. Method according to any one of claims 1 to 4, wherein the condition BED is expressed by DM(Lm, L m *) < DM max with a predetermined, constant value DM max >0 is implemented.

6. Method according to any one of claims 1 to 5, wherein the distance metric DM(L) m, Lm*) is a modified Hausdorff distance metric, where: DM( L m , L m *) — 7. Method according to any one of claims 1 to 6, wherein the determination of the cluster CL P representing line element LE P with starting positions P a (LA p k )n ) of line segments LA p k n and end positions P e (LA p k n ) of the line segments LA p k n by a with a respective length LA p k n the line segment elements LA p , k , n weighted averaging MI according to: and P a (LA p n ) Initial position of the averaged line segment element LA p , n P e (LA p n ) End position of the averaged line segment element LA p , nfor cluster p such that the averaged line element for cluster / ? is defined as follows:

8. Device for generating navigation data, comprising: 8.

3. a first means (201) for generating and / or providing data BD in which a part of the Earth's surface is depicted; 8.

4. a segmentation tool (202) for segmenting a number M of Line elements L m from the data BD, where each line element L m in the BD data an initial position and a final position as well as a Environment UMG(P(L m ) is assigned, where P^L^ is a position of the line element L m BD specifies in the data, with m = 1, 2, and where that Segmentation means (202) is designed and set up for this purpose - each of the line elements L m into a predefined and for all line elements Lm each identical number N of equidistant line segments LA mi "to subdivide, with each line segment LA mi "in the BD data through a starting position and an end position x m “ “ is defined as: - for a line element L m from within its respective environment UMG(P(L m ) similar line elements arranged Z m * to determine, whereby a similarity of line elements L m and L m * based on a predefined distance metric DM(L m , L m ») is determined, which is based on a function F of distances d(LA m .., LA mV ) of line segments LA m "and LA m * n * of the line elements L m and L m * based on: DM(L m , L m *) = F(d(LA m "LA m * n *)), where d(LA m , n , LAm » n ») a distance between the line segment LA m , n LA m * n * and the LA line section m * n * indicates, and where: m* mc {1, 2, ... , M} and m* m, n* ne {I, 2, ..., N} such that the distance metric DM(L m , L m ») between L m and L m * with m * m through a bridge of N line segments LA m "in N line segments LA" m * n * for all n*=n is determined by and wherein the line elements L m and L m » only be considered similar if DM(L m , L m ») a given condition BED satisfies; a cluster resource (203) that is executed and configured for this purpose, in the environment l G(P(L m )) a non-overlapping clustering of all to L m similar line elements L m * to a local cluster CLP = CL,,(L m , L m *, UMG PL m to execute; where non-overlapping clustering applies to all clusters not already in a local cluster CL P assigned line elements L m is executed so that the result is a number P of local clusters CL P is determined where p = 1, 2P and P > 1, where each local cluster CL P a number K(p) line elements L p , k includes: Ep=i ^(p) = M , with k = 1, 2, K(p) and K(p) > 7; and for each local cluster CL P based on the similar line elements L clustered therein p , k one of the line elements LE representing this cluster P to determine; 8.

6. a second means (204) designed and set up for this purpose, based on the identified line elements LE P To determine the boundaries of the lanes depicted in the BD data; and 8.

7. a third means (205) designed and equipped to generate and output navigation data including the determined lane boundaries.

9. Vehicle with a device according to claim 8, wherein the vehicle is semi-automatically or automatically controllable.

10. Computer system comprising a data processing device, wherein the data processing device is configured such that a method according to one of the preceding claims is executed on the data processing device. 1 1 . Digital storage medium with electronically readable control signals, wherein the control signals can interact with a programmable computer system in such a way that a method according to one of the preceding claims is carried out.

12. Computer program product comprising program code stored on a machine-readable medium for carrying out the method according to one of the preceding claims, when the program code is executed on a data processing device.

13. Computer program with program code for carrying out the method according to one of the preceding claims, if the program is based on a data processing device is running, this can be used for the The data processing device may be designed as any computer system known from the prior art.

Images