SEGMENTING TRAJECTORY DATA

Description

[0001] The invention relates to a method for segmenting trajectory data. Furthermore, the invention relates to a control unit, a computer program, and a machine-readable storage medium. State of the art

[0002] The document "Dynamics Based Trajectory Segmentation for UAV videos" by Banerjee et al. (Advanced Video and Signal-Based Surveillance (AVSS), 2010, 7th International Conference on, IEEE, Piscataway, NJ, USA, August 29, 2010, pages 345-352, XP031772045) discloses a specific representation of vehicle trajectories for applications in trajectory analysis and activity detection.

[0003] The document "Improving persistence based trajectory simplification" by Laass et al. (2021 22nd IEEE International Conference on Mobile Data Management (MOM), IEEE, June 15, 2021, pages 157-162, XP033938584) reveals an online algorithm for simplifying spatial trajectories in linear time.

[0004] The document "Noise filtering, trajectory compression and trajectory segmentation on GPS data" by Lin et al. (2016 11th International Conference on Computer Science & Education (ICCSE), IEEE, August 23, 2016, pages 490-495, XP032975301) reveals an improved method for noise filtering, trajectory compression and trajectory segmentation based on the Kalman filter and the Douglas-Peucker algorithm.

[0005] US patent 2019 / 086925 A1 discloses a computer-implemented method for optimizing a trajectory for an autonomous vehicle.

[0006] A trajectory refers to a sequence of timestamped positions recorded in a specific coordinate system. An example is a vehicle trajectory recorded using sequential geographic coordinates via GNSS. Trajectory segmentation aims to divide given trajectories into parts or blocks that exhibit consistent geometric properties, such as direction and linearity. Segmentation is applied in various fields, such as the creation of planning maps using crowdsourced trajectory data, where trajectory segmentation allows for scaling over any desired area.

[0007] Trajectory segmentation typically focuses on identifying trajectory points or positions whose geometric properties become inconsistent compared to their neighboring points. These points are then registered as segmentation points. The segmentation points form a boundary between two segmentations or blocks. Methods for segmenting trajectories are already known. For example, trajectories can be divided or segmented into segments of equal length. Furthermore, the linearity of the given trajectories can be exploited to determine the minimum number of segmentation points.

[0008] However, a problem with the known methods is that segmentation points are often defined within curves, or trajectory sections are divided into too many segments. Disclosure of the invention

[0009] The object underlying the invention can be seen as proposing a method for dividing trajectory data into segments, which avoids the segmentation of curve sections.

[0010] This problem is solved by means of the respective subject matter of the independent claims. Advantageous embodiments of the invention are the subject matter of dependent claims.

[0011] According to one aspect of the invention, a method for segmenting trajectory data, particularly by a control unit, is provided. In one step, trajectory data with a plurality of points is received. The points can be position points, each with a timestamp.

[0012] To identify essentially straight segments separated by segmentation points, an algorithm for curve smoothing and / or generalization, such as a Douglas-Peucker algorithm, is applied to the trajectory data. This simplifies the shape or geographical course of the trajectory data, allowing segmentation points to be formed to divide the trajectory data into curve segments.

[0013] In a further step, an algorithm for detecting curves, particularly in the form of curve start and end points, is applied to the trajectory data. This algorithm detects curve segments and their lengths.

[0014] At least one segmentation point determined within an interval between a start point and an end point of a curve is replaced by the start point and the end point. This prevents the segmentation of trajectory data within curve segments. Such segmentation points within a curve segment are therefore deleted and replaced by the start point and end point of the curve segment.

[0015] The trajectory data is segmented at the identified segmentation points, the starting points and the endpoints.

[0016] This method combines algorithms for curve smoothing and / or generalization with algorithms for detecting curves or curve segments. The results of the respective algorithms can be fused to prevent the formation of segmentation points within curve segments and within straight sections of the trajectory data. In particular, the curve detection algorithm can prevent a curve from being segmented in its middle by taking into account the consistency of the trajectory data's course changes.

[0017] The method according to the invention is applied separately to each set of trajectory data and can therefore process several sets of trajectory data simultaneously. A set of trajectory data can be a group of measurement data from a vehicle, a group of measurement data from a geographical area, a predefined number of measurement data, and the like.

[0018] According to a further aspect of the invention, a control unit is provided, wherein the control unit is configured to execute the method. The control unit can, for example, be a vehicle-integrated control unit, an external vehicle-integrated control unit, or an external vehicle-integrated server unit, such as a cloud system. The control unit can preferably receive and process the trajectory data. For this purpose, the control unit can have internal or external memory to be able to temporarily or permanently store the trajectory data and the results of the method.

[0019] Furthermore, the control unit can have an integrated communication unit or an external communication unit to receive the trajectory data and to send the segmented trajectory data as a possible result of the procedure.

[0020] Furthermore, according to one aspect of the invention, a computer program is provided which includes instructions that, when executed by a computer or control unit, cause it to execute the method according to the invention. According to another aspect of the invention, a machine-readable storage medium is provided on which the computer program according to the invention is stored.

[0021] In one embodiment, the curve detection algorithm determines a heading for each point and subsequently the difference in heading between any two consecutive points. A point in the trajectory data is assigned to a curve or curve segment if the difference in heading between that point and the next exceeds a predefined threshold. This allows for the technically simple detection of curve segments. Points in the trajectory data on straight sections are filtered out as points with relative headings below the threshold.

[0022] In another embodiment, the heading is determined as a vector between a first point and a subsequent second point, using a yaw angle. For vehicle trajectory data, this corresponds to the vehicle's direction of travel from the first point towards the second point. The yaw angle can be determined relative to a predefined compass direction or a reference direction.

[0023] According to another embodiment, the difference in course direction is standardized by a distance between the first point and the second point. This measure allows for additional curve smoothing and comparability of the determined course directions from the different points.

[0024] In another embodiment, points exceeding the predefined threshold are defined as starting or ending points of a curve, and these points are used as segmentation points to divide the trajectory data into sections. This allows for the determination of a unique starting point and endpoint for a curve or curve segment. As soon as there are no or only minor differences in direction between two adjacent points in the trajectory data, they will not exceed the threshold and will be assigned a straight segment rather than a curve segment by the method.

[0025] According to a further embodiment, the trajectory data, segmented at the determined segmentation points, the start points, and the end points, are received at least segment by segment and used to control a vehicle. The trajectory data can be received segment by segment and used for vehicle control. Multiple segments can also be received simultaneously, for example, to map a planned trajectory or a portion thereof.

[0026] In the following, preferred embodiments of the invention are explained in more detail with reference to highly simplified schematic representations. These show Fig. 1 shows a representation of segmented trajectory data according to the prior art, Figs. 2-4 show schematic top views to illustrate a method according to the invention in one embodiment, and Fig. 5 shows schematic top views to illustrate an algorithm for detecting curves.

[0027] In the Figure 1 A representation of segmented trajectory data according to the state of the art is shown. The straight sections 2 and the curved sections 4 of the trajectory data 1 are divided into different segments 6. The trajectory data 1 can then be further processed in the form of segments 6 or made available to road users not shown.

[0028] To subdivide the trajectory data 1, segmentation points 8 must be detected, at which boundaries between two segments 6 are drawn. As in the Figure 1As can be seen, the trajectory data 1 is segmented into uniform, elongated segments 6, which also have boundaries within curved sections 4. If a road user utilizes such segmented trajectory data 1, processing delays can occur. This can have a detrimental effect on road safety in complex traffic situations, such as on curves. Furthermore, the straight sections 2 are unnecessarily divided into a large number of segments 6.

[0029] In the Figures 2 to 4 The schematic top views illustrate a method according to the invention in one embodiment. The diagrams show... Figure 2A direction of movement is indicated by an arrow P, along which the trajectory data 1 were determined and along which the processing of the trajectory data 1 by the method takes place. The method can be executed by a control unit (not shown), which can be located inside or outside the vehicle.

[0030] In a first step, the trajectory data 1 with a large number of points 3 are received. For this purpose, in the Figures 2 to 4 A detailed view is shown that displays only one trajectory. However, the trajectory data 1 can also be composed of a multitude of trajectories and may contain a multitude of overlapping points 3.

[0031] To identify essentially straight segments separated by segmentation points 8 and 9, a curve smoothing and / or generalization algorithm, such as a Douglas-Peucker algorithm, is applied to the trajectory data 1. The corresponding result is shown in the Figure 2 The illustration shows two segmentation points, 8 and 9, as examples.

[0032] In a second step, which takes place in the Figure 3 As has been clarified, an algorithm for detecting curves or curve segments 4, in particular in the form of starting points 10 and endpoints 12 of curve segments 4, is applied to the trajectory data 1.

[0033] The algorithm for detecting curve segments 4 determines a course direction for each point 3 and subsequently a difference in course direction between any two consecutive points. A point 3 of the trajectory data 1 is assigned to a curve or a curve segment 4 if the difference in course direction between point 31 and the following point 32 exceeds a predefined threshold. This is described in the Figure 5 explained in more detail.

[0034] In a further step of the procedure, at least one segmentation point 8, 9 determined in an interval between a start point 10 and an end point 12 of a curve segment 4 is replaced by the start point 10 and the end point 12. The corresponding segmentation point 9 is then deleted, or at least no more than one segmentation point 9 is used.

[0035] Subsequently, the trajectory data 1 are segmented into segments 6 at the determined segmentation points 8, the starting points 10 and the endpoints 12.

[0036] The Figure 5 Figure 4 shows schematic top views to illustrate an algorithm for detecting curve segments. At a first point 31 with a first timestamp tn, a course direction in the form of a yaw angle is determined. The second point 32 has a second timestamp tn+1.

[0037] The trajectory data 1 are determined using an example by a vehicle 14 traveling a route. The different points 31, 32 are visited at different times t and assigned corresponding timestamps.

[0038] In the illustrated embodiment, the yaw angle yaw is measured relative to a horizontal axis x, which can be, for example, an east-west line. In this embodiment, the yaw angle yaw can be calculated by the arctangent of the quotient of a vertical offset Δy and a horizontal offset Δx to a subsequent second point 32.

[0039] The yaw angles are calculated for each point 3 with a subsequent point. In a further step, relative course changes Δyaw between the respective points 3 are calculated. This is achieved by calculating the difference between the respective yaw angles.

[0040] The relative course changes Δyaw of each trajectory point 3 can be accumulated with the values ​​of those points 3 that lie within a predefined distance parameter. The accumulated angle value of each trajectory point can then be smoothed with a predefined averaging width. These two steps allow points 3 exhibiting continuous course changes to be highlighted, while points 3 in the trajectory data 1 with a relatively constant direction of movement, such as on straight sections 2, are suppressed. By marking the points whose smoothed relative course directions or course changes Δyaw exceed a certain angle threshold, curves in the given trajectory can be identified.

Claims

1. Method for segmenting trajectory data (1), in particular by means of a control unit, wherein - trajectory data (1) comprising a multiplicity of points (3) are received, - a curve smoothing and / or generalization algorithm, for example a Douglas-Peucker algorithm, is applied to the trajectory data (1) to determine substantially straight sections (2) separated by segmentation points (8, 9), - an algorithm for detecting curves (4), in particular in the form of starting points (10) and end points (12) of curves (4), is applied to the trajectory data (1), - at least one segmentation point (9) determined in an interval between a starting point (10) and an end point (12) of a curve (4) is replaced by the starting point (10) and the end point (12), the trajectory data (1) being segmented at the determined segmentation points (8), the starting points (10) and the end points (12), the algorithm for detecting curves (4) being used to determine a path direction for each point (3) and then a difference in the path direction between two particular consecutive points (31, 32), a point (3) in the trajectory data (1) being assigned to a curve (4) if the difference in the path direction between the point (31) and the next point (32) exceeds a predefined threshold value, characterized in that the difference in the path direction is standardized by a distance between the first point (31) and the second point (32).

2. Method according to Claim 1, wherein the path direction is determined in the form of a vector between a first point (31) and a subsequent second point (32) comprising a yaw angle (yaw).

3. Method according to either of Claims 1 and 2, wherein points (3) in the trajectory data (1) that do not exceed the predefined threshold value and are adjacent to the points (3) that do exceed the threshold value are defined as the starting points (10) or end points (12) of a curve (4) and used as segmentation points (8) for splitting the trajectory data (1) into segments (6).

4. Method according to one of Claims 1 to 3, wherein the trajectory data (1) segmented at the determined segmentation points (8), the starting points (10) and the end points (12) are received at least segment by segment and used for controlling a vehicle (14).

5. Control unit, the control unit being designed to carry out the method according to one of Claims 1 to 4.

6. Computer program comprising instructions that, when the computer program is executed by a computer or a control unit, cause same to carry out the method according to one of Claims 1 to 4.

7. Machine-readable storage medium on which the computer program according to Claim 6 is stored.