METHOD FOR DETECTING STATIC RADAR TARGETS USING A RADAR SENSOR FOR MOTOR VEHICLES

Description

[0001] The invention relates to a method for detecting and classifying static radar targets by means of a radar sensor of a motor vehicle, comprising: detecting an object as a static radar target based on the received radar signals reflected by the object. State of the art

[0002] In driver assistance systems for motor vehicles, for example in automatic distance control systems or collision warning systems, radar sensors are often used to detect the traffic environment. In addition to dynamic objects, the driver assistance systems must also reliably detect static objects as potential collision hazards. However, in a single radar measurement, actual static obstacles and drivable objects, such as manhole covers, empty chip bags, or cola cans, may not be distinguishable from one another. To date, it has been common practice to observe and classify objects as actual static obstacles or drivable objects by observing and classifying the radar signature of a static object over time. The objects are treated as point-like.

[0003] WO 2010 / 127650 A1 describes a method for evaluating sensor data from an environmental detection system for a motor vehicle. Detection points are entered into a two-dimensional occupancy grid, with the state of a grid cell being occupied and therefore "potentially not traversable" and otherwise unoccupied and thus "traversable." The occupancy grid essentially represents the vehicle's surroundings.

[0004] US 2017 / 0269201 A1 describes estimating an environment from observational data.

[0005] DE 102014221144 A1 describes a target detection device for a vehicle.

[0006] Recently, there has been increasing interest in radar sensors that can detect significantly more radar reflections for an object with greater accuracy.

[0007] The object of the invention is to provide a method for a radar sensor that allows static obstacles to be detected more reliably.

[0008] This object is achieved according to the invention with the features specified in the independent claims.

[0009] The core idea of ​​the solution is that for a static radar target that is tracked over time and classified based on radar signatures of the radar reflections assigned to the radar target, an occupancy pattern is additionally entered into an occupancy grid, with the assignment of the occupancy pattern to the radar target being stored. In this process, an occupancy pattern is assigned to a radar target that corresponds to one or more radar reflections assigned to the radar target. Thus, the occupancy pattern is linked to the classification of the radar target based on radar signatures of the radar reflections.

[0010] The occupancy pattern can indicate an occupancy probability for specific grid positions (grid cells). The occupancy grid is stored and updated over time. Over time, radar reflections can be accumulated in the occupancy pattern. The extent of the occupancy pattern allows an estimation of the extent of a static obstacle. If a radar target is classified as traversable, the occupancy pattern can be removed from the occupancy grid, corresponding to setting the occupancy probability to a value indicating "not occupied." A similar procedure can be applied to vehicles that have started moving. The method can complement a conventional method for classifying point targets and detecting actual obstacles.

[0011] It is particularly advantageous that the assignment of an occupancy pattern to an object makes it possible to subsequently classify the underlying object for occupancy probabilities entered in an occupancy grid and to link the information about the occupancy probabilities in the occupancy grid and about the classification, for example during an approach to the corresponding object.

[0012] During an approach to an object, a large number of radar reflections associated with the object may occur, for example, on the order of 100 or more reflections. Assigning an occupancy pattern to an object can, even with a large number of radar reflections associated with the object, enable easy evaluation of the occupancy pattern to estimate the object's position and / or size.

[0013] Furthermore, distinguishing between different groups of static radar targets can make it possible to distinguish between adjacent radar targets in the occupancy grid. This allows for a better mapping of the vehicle's surroundings. One example is the possible distinction between a guardrail and a neighboring tree. Another example is the distinction between stationary vehicles and other static radar targets.

[0014] A particular advantage of assigning the occupancy pattern to the object is that if the object is later estimated to be traversable, the occupancy pattern can be removed from the occupancy grid because the assignment means that the occupancies in the occupancy grid belonging to the object are known. This enables a reliable removal of occupancy probabilities in the occupancy grid. This means that if an obstacle turns out to be traversable while approaching it, the corresponding occupancy in the occupancy grid can be deleted again. An example is a railway track in a street, such as a tram track or a railway track at a level crossing. The railway track represents a static radar target with a large extent, which can be deleted from the occupancy grid after being classified as traversable.

[0015] Accordingly, if a radar target was initially recorded as a static radar target in the occupancy grid and is detected as a moving vehicle, the occupancy pattern can be deleted from the occupancy grid.

[0016] Another advantage is that, based on the occupancy grid, an additional estimation of the extent of objects is possible, especially for objects that were traditionally treated as point targets. This allows for a reduction in the rate of false-positive obstacle detections.

[0017] Even for larger objects with a high number of radar reflections over time, an additional estimation of the extent based on the occupancy pattern is advantageous, since the occupancy pattern can be evaluated much more easily than the parameters of a large number of radar reflections.

[0018] Another advantage is that the occupancy grid still allows for a model-free representation of the vehicle's static environment. This allows for versatile applications, such as estimating guardrails or finding parking spaces. Furthermore, it allows for modeling of the free space within the occupancy grid. The modeling of the free space can also be updated by removing occupancy patterns from objects detected as being traversable. This improves the reliability of an occupancy grid.

[0019] In particular, an object can be detected as a static radar target based on an estimate of an absolute speed of zero, i.e. a relative speed corresponding to the vehicle's own speed with the opposite sign.

[0020] For example, the occupancy pattern associated with a static radar target can be stored, perhaps in the form of a directory of the entries made in the occupancy grid that belong to the occupancy pattern. For example, for each grid cell of the occupancy grid that belongs to an occupancy pattern of a static radar target, an identification / code (ID) of the static radar target to which the occupancy pattern is assigned can be stored.

[0021] Preferably, an occupancy probability is recorded in a respective grid cell of the occupancy grid.

[0022] The occupancy pattern may include one or more grid cells in the occupancy grid.

[0023] The received radar signals reflected by the object can be mapped in one or more grid cells in the occupancy grid, whereby the respective grid cell corresponds to a respective location position of the received radar signals reflected by the object.

[0024] The occupancy grid preferably extends in at least two dimensions. The occupancy grid is preferably a two-dimensional grid, corresponding to a horizontally extending area in the surroundings of the motor vehicle.

[0025] For example, radar signatures of received radar signals can be compared with characteristic features and a static radar target can be classified based on the result of the comparison.

[0026] Characteristic features for classifying the static radar target can be, for example: azimuth angle, elevation angle, radial range, relative velocity, and / or backscatter properties, such as the radar cross section, as well as combinations of several of these features.

[0027] When tracking the static radar target over time, the classification of the static radar target can be updated, for example.

[0028] Advantageous further developments and embodiments of the invention are specified in the subclaims.

[0029] Preferably, several radar reflections originating from different positioning positions are combined into a static radar target based on the spatial proximity of the positioning positions, wherein the generated movement pattern in the occupancy grid comprises several grid cells in which the several radar reflections are mapped. Thus, several radar reflections assigned to the radar target can be recorded in respective grid cells of the occupancy pattern in the form of occupancy probabilities, wherein the respective grid cells correspond to the respective positioning positions of the radar reflections assigned to the radar target.

[0030] The invention also relates to a radar sensor for motor vehicles in which one of the methods described above is implemented.

[0031] In the following, an embodiment example is explained in more detail using the drawing.

[0032] They show: Fig. 1 is a schematic diagram of a radar sensor to which the invention is applicable; Fig. 2 is a schematic representation of radar reflections, an estimated point target, and an occupancy grid; and Fig. 3 is a flowchart of a method for detecting and classifying static radar targets.

[0033] The Fig. 1 The radar sensor shown comprises a sensor unit 10 with transmitting and receiving antennas, which is connected to a control and evaluation unit 12. This controls the sensor unit 10 in respective measurement cycles for emitting radar signals and for receiving radar signals reflected from objects. The received radar signals are evaluated by the control and evaluation unit 12, and parameters of individual radar reflections, i.e., received radar signals reflected from an object 14, are estimated. The estimated parameters of the radar reflections, as well as other data from the raw measurements, which can also be referred to collectively as radar signatures, are output to an object detection and tracking unit 16.The object detection and tracking unit 16 combines the radar reflections based on the spatial proximity of the detection positions of individual radar reflections at suitably approximately matching relative velocities and recognizes them as a single radar target. This is also referred to as clustering. The detected radar targets are tracked over time (object tracking).

[0034] In the case of a radar target identified as static, i.e., a radar target corresponding to a stationary object, an object classification unit 18 connected to the object detection and tracking unit 16 classifies the detected static radar targets based on characteristic features of their radar signatures in a first processing path. Classification can be carried out, for example, using a deep learning approach, such as an artificial neural network. Parameters of the radar reflections that can be used for classification include, for example, azimuth angle, elevation angle, radial distance, relative speed, and backscatter properties, such as the radar cross section. Furthermore, parameters based on multiple radar reflections can be used for classification, such asNumber of radar reflections, arrangement of reflections, pattern of the arrangement, taking into account the respective backscatter properties of the radar reflections. For example, the following can be distinguished in the classification: non-driveable objects (static obstacles) such as stationary vehicles, curbs, guardrails, signs, traffic light posts, trees, and possibly other obstacles not further classified; as well as driveable objects.

[0035] In a second processing path, the estimated parameters of the radar reflections of a detected static radar target are simultaneously entered into an occupancy grid 20 stored in a memory and accumulated over time.

[0036] This is exemplified in Fig. 2 The motion grid 20 represents the detected static environment of the motor vehicle.

[0037] Fig. 2(a) shows the location positions of individual radar reflections. Fig. 2(b) shows a radar target 21 in the form of an estimated point target, which was estimated based on the individual radar reflections. Fig. 2(c) illustrates how the location positions of the individual radar reflections result in occupancy probabilities of corresponding grid cells 22 of the occupancy grid 20. The radar reflections are assigned corresponding grid cells, which are marked as "probably occupied." Different values ​​of occupancy probabilities can be distinguished. Different occupancy probabilities are represented by different markings of the grid cells 22. Fig. 2(d) shows the obtained occupancy pattern 24 in the occupancy grid 20. The occupancy pattern 24 is thus generated by recording a plurality of radar reflections assigned to the radar target in respective grid cells 22 of the occupancy pattern 24 corresponding to the location positions of the radar reflections in the form of occupancy probabilities.

[0038] Additionally, an association between occupied grid cells 22 of the occupancy pattern 24 and the respective radar target is stored. For example, the occupancy probability and an object identification (ID) that identifies the associated radar target can be stored in each respective grid cell 22. Alternatively or additionally, the object detection and tracking unit 16 can store the occupancy pattern 24 in the form of a directory of the grid cells 22 assigned to the radar target.

[0039] The occupancy grid 20 is designed to store the relevant occupancy patterns 24 for several detected static radar targets, wherein the assignments are also stored in each case.

[0040] The object detection and tracking unit 16 is configured to estimate the extent of a radar target based on the associated occupancy pattern 24. The extent can be determined, for example, according to the occupied grid cells 22 of the occupancy pattern 24. The extent thus estimated can be included as an additional parameter of the radar target in the classification of the radar target by the object classification unit 18.

[0041] If, during the tracking of a static radar target, in particular a point target, it turns out that the radar target is not a relevant obstacle but is (re)classified as traversable, the object detection and tracking unit 16 is configured to delete the assigned occupancy pattern 24 in the occupancy grid 20 based on the assignment of the occupancy pattern 24 to the static radar target. An example of this is an object in the form of a manhole cover. Thus, the occupancy grid 20 always only depicts static radar targets classified as potential obstacles, without leaving artifacts in the occupancy grid 20 due to a declassification of a static radar target as an obstacle.

[0042] If, during tracking of a static radar target, it becomes apparent that the radar target has started moving, the object detection and tracking unit 16 is configured to delete the assigned occupancy pattern 24 in the occupancy grid 20 based on the assignment of the occupancy pattern 24 to the radar target. Thus, the occupancy grid 20 always only displays radar targets identified as static, without leaving artifacts in the occupancy grid 20 due to a declassification of a radar target as a static radar target.

[0043] If, during the tracking of two or more static radar targets, it turns out that these belong to a single object, the object detection and tracking unit 16 is configured to merge the occupancy patterns 24 into a combined occupancy pattern 24 of a radar target based on the assignments of the occupancy patterns 24 to the radar targets.

[0044] The object detection and tracking unit 16 with the occupancy grid 20 and the object classification unit 18 are, for example, part of a driver assistance system 30.

[0045] The Fig. 3 The method shown can be implemented in the described radar sensor and is described below. In a step S10, at least one object 14 is recognized as a static radar target based on the received radar signals reflected by the object 14. In step S12, the respective occupancy pattern 24 is generated in the occupancy grid 20 based on the received radar signals reflected by the respective object 14. In step S14, an assignment is stored that assigns the generated occupancy pattern 24 to the static radar target. In step S16, the above-described classification of the static radar target takes place.

[0046] The classification (step S16) can be carried out in parallel with the generation of the occupancy pattern (step S14), or before or after.

[0047] In step S18, a check is performed to determine whether a radar target has been classified as traversable. If so, the occupancy pattern and the stored assignment are deleted in step S22. The method then continues with step S10, which detects and tracks other radar targets.

[0048] If not, a check is performed in step S20 to determine whether the radar target is still detected as a static radar target. If so, the process is repeated, with the radar target continuing to be tracked in step S10 and the occupancy pattern being updated in step S12 by accumulating the radar reflections assigned to the radar target. Otherwise, if the radar target is now detected as moving, the occupancy pattern and the stored assignment are deleted in step S22. The process then continues with step S10 for detecting and tracking other radar targets.

[0049] The steps described can be carried out accordingly for all detected radar targets.

Claims

1. Method for identifying and classifying static radar targets using a radar sensor of a motor vehicle, the method comprising: identifying (S10) an object (14) as a static radar target based on the received radar signals reflected by the object (14), generating (S12) an occupancy pattern (24) in an occupancy grid (20) based on the received radar signals reflected by the object (14), storing (S14) an assignment that assigns the generated occupancy pattern (24) to the static radar target, characterized in that, subsequent to the generation of the occupancy pattern (24) in the occupancy grid (20) and subsequent to the storage of the assignment, classifying (S16) or reclassifying (S16) the static radar target as associated with one of a plurality of groups of static radar targets based on characteristic features of radar signatures of the received radar signals reflected by the corresponding object (14), wherein the characteristic features include: azimuth angle, elevation angle, radial distance, relative velocity and / or backscatter characteristics, or a combination of several of these features.

2. Method according to Claim 1, wherein the plurality of groups of static radar targets comprise at least one group of traversable radar targets and one group of non-traversable radar targets, and wherein, if a static radar target is classified as traversable, the occupancy pattern assigned to the static radar target in the occupancy grid (20) is deleted (S18, S22) based on the assignment of the occupancy pattern (24) to the static radar target.

3. Method according to Claim 1 or 2, wherein the method comprises: tracking (S10) the static radar target over time based on the received radar signals of multiple measuring cycles of the radar sensor reflected by the corresponding object (14), wherein received radar signals from multiple measuring cycles of the radar sensor reflected by the corresponding object (14) are used in the classification (S16) of the static radar target as associated with one of a plurality of groups of static radar targets.

4. Method according to any one of the preceding claims, wherein the method comprises: tracking (S10) the static radar target over time based on the received radar signals of multiple measuring cycles of the radar sensor reflected by the corresponding object (14), updating (S12) the occupancy pattern (24), which is assigned to the static radar target, in the occupancy grid (20), comprising: accumulating occupancy probabilities in one or more grid cells (22) of the occupancy grid (20) based on the received radar signals of the multiple measuring cycles of the radar sensor reflected by the corresponding object (14).

5. Method according to any one of the preceding claims, wherein an expansion of the static radar target is estimated based on an occupancy pattern (24), which is assigned to a static radar target, and the estimated expansion of the static radar target is used in the classification (S16) of the static radar target as associated with one of a plurality of groups of static radar targets.

6. Method according to any one of the preceding claims, wherein a plurality of radar reflexes originating from different locating positions are combined based on a spatial proximity of the locating positions to a static radar target, wherein the generated occupancy pattern (24) in the occupancy grid (20) comprises a plurality of grid cells (22) in which the plurality of radar reflexes are imaged.

7. Radar sensor for motor vehicles, comprising: a memory for storing an occupancy grid (20) corresponding to an area in the vicinity of the motor vehicle, an object identification and tracking device (16) for detecting and tracking static radar targets sensed by the radar sensor, and an object classification unit (18) for classifying a static radar target as associated with one of a plurality of groups of static radar targets, wherein the radar sensor is designed to carry out the method according to any one of Claims 1 to 6.