Sensor-based object tracking method and object tracking device, and correspondingly designed motor vehicle

Abstract

The invention relates to a method (9) and an object tracking device (5) for sensor-based object tracking and a correspondingly designed motor vehicle (1). In the method (9), only an object detected by one of the plurality of sensors (3) is identified as a single-sensor object. The respective object trajectory (10, 14), in which object data describing the respective object is stored, is then updated directly by means of the respective current sensor data (11) without using a predetermined filter. The further object trajectories (10, 15), which are not associated with the individual sensor object, are updated by means of the corresponding current sensor data (11) using a predetermined filter.

Description

Technical Field

[0001] The invention relates to a method and an object tracking device for sensor-based object tracking and also to a motor vehicle designed or equipped accordingly. Background Art

[0002] In various technical applications and fields (e.g. not only in the field of vehicles and traffic), the sensory detection and tracking, i.e. tracking, of objects in the corresponding environment can be useful, for example for collision avoidance, trajectory planning, monitoring, etc. Since different types of corresponding sensors can have different advantages and disadvantages, a plurality of different sensors are often used and their sensor data or outputs are combined with one another.

[0003] For example, a sensor data merging system for environmental perception is described in DE 10 2021 113653 B3. Specifically, a method for establishing an environmental model for a highly automated vehicle having at least two sensors for environmental detection is proposed there. The method comprises: projecting the raw data of a first sensor onto an occupancy grid by generating a grid cell size according to an inverse sensor model, and projecting preprocessed object data of the grid cell size according to an inverse sensor object model. The projected data are then merged into an occupancy grid. An environmental model is then established from the extracted grid data, which is provided for highly automated operation of the vehicle. Therefore, it should be possible to establish a corresponding environmental model for a vehicle with a large number of sensors with a high degree of confidence in a resource-minimizing manner.

[0004] As another example, DE 10 2019 115235A1 describes object tracking in a blind spot. The method includes identifying an object (including points on the object) and performing object tracking based on a motion model that includes the relationship between the points. Then, the position of one of the points in the blind spot is output based on the tracking.

[0005] However, in previous methods, the runtime, delay and bandwidth for signal or data processing can often be problematic. To this end, DE 10 2018 105293 A1 discloses a method for networked scene representation and expansion in a vehicle environment in an automated driving system. The method includes obtaining data related to the position of a first object and obtaining a bandwidth and a delay time of a transmission channel. A portion of the data is then transmitted in response to the bandwidth and delay time of the transmission channel. Summary of the invention

[0006] The object of the present invention is to enable reaction-time-optimized object tracking.

[0007] This object is achieved by the subject matter of the independent claims. Further possible embodiments of the invention are disclosed in the dependent claims, the description and the drawings. The features, advantages and possible embodiments described within the scope of the description of one of the subjects of the independent claims should be regarded at least similarly as the features, advantages and possible embodiments of the corresponding subject matter of the other independent claims, as well as any possible combination of the subject matter of the independent claim in combination with one or more of the dependent claims, if necessary.

[0008] The method according to the invention can be applied within the scope of sensor-based object tracking, in particular in correspondingly designed motor vehicles. However, in principle, the method according to the invention is not limited to this application case. In the method according to the invention, sensor data of a sensor system are detected in each measuring cycle. Here, the sensor system includes a plurality of individual sensors, which can be of different types in particular. The individual sensors or the sensor system including these sensors can already detect the object itself and track the object within this time or within a plurality of frames or measuring cycles. This can be performed automatically, for example, by the sensor system or the individual sensors in internal data processing or preprocessing. Similarly, the sensor data from the sensor system or the individual sensors can be detected or processed accordingly, for example, by an object tracking or tracking device designed to execute or apply the method according to the invention, such as a correspondingly designed control unit, etc., and the object tracking or tracking device or the control unit is coupled or can be coupled to the sensor system or the individual sensors.

[0009] Furthermore, in the method according to the invention, the detected objects represented in the sensor data or by these sensor data are respectively assigned to an object trajectory, in which object data describing or characterizing the corresponding object are stored. Such an object trajectory can be a data set that is respectively object-specific or provided for the corresponding detected object. While the corresponding object is being tracked, i.e. tracked, the detected object or its data or characteristics can be maintained or managed in the object trajectory. The object trajectory can also be referred to as an internal trajectory, because it can be managed, for example, inside a sensor system or sensor or in an object tracking device near the sensor along a corresponding signal or data processing pipeline.

[0010] In an existing object trajectory, for example, object properties or object states from the respective last or past measuring cycle can be stored. A new object trajectory can be provided for each newly detected object. By means of the corresponding object recognition by means of the detected sensor data and the assignment to the object trajectory, it is known by means of which of the individual sensors a specific object has been detected. This can also be stored in the object trajectory.

[0011] Furthermore, the object trajectory or the data stored in the object trajectory or information about the object is updated with the sensor data corresponding to the respectively assigned object from the respectively current measurement cycle, i.e. with the respectively latest available or detected sensor data. It can thus be ensured that the respectively current data or information about the respectively assigned object is stored in the object trajectory.

[0012] According to the invention, in the method, objects which have been detected for the first time before the respective current measurement cycle and which have been detected in the respective current measurement cycle only with the corresponding current sensor data of just one of the multiple individual sensors (which have been detected only with or with the sensor data of just this sensor since at least one preset time period containing multiple measurement cycles) are identified or represented or classified as single-sensor objects. Such single-sensor objects are objects which have always been detected only by one and the same single sensor at least in the preset time period. Therefore, only these objects are identified as single-sensor objects. It can be provided that, for example, objects which have not been detected in the preset time period and / or which have only been detected for the first time or newly in the current measurement cycle are also not identified as single-sensor objects, even when the objects have been or are currently detected only by a single sensor. In contrast, other objects which have been detected in the current measurement cycle and / or in the preset time period with the help of multiple and / or alternating sensors of the individual sensors or which have been identified with or with the help of sensor data of multiple different sensors of the individual sensors can be identified, for example, as multi-sensor object tracks. If an object which has previously been identified as a single-sensor object, i.e. until the respective current measuring cycle or point in time, is detected in the current measuring cycle by at least one sensor different from the previous one, the object can therefore immediately, i.e. also in the same current measuring cycle, lose its identification or classification as a single-sensor object.

[0013] The preset time period can respectively last from the current time point to the past, i.e. be understood as a sliding time window. The preset time period can be specified or defined, for example, in milliseconds or as a number of measuring cycles or the like. In applications in the automotive sector, the preset time period can, for example, include approximately 500 ms or 10 measuring cycles or the like. However, other values ​​or lengths of the preset time period can also be used, for example depending on requirements, the application, the typical measuring frequency of the sensor, the speed or clock frequency of the signal or data processing device or the like.

[0014] In addition, according to the present invention, it is provided that, when using a preset filter or merging mechanism, the object trajectory that is not assigned to a single sensor object or the data stored in the object trajectory or the description of the object is automatically updated with the help of the sensor data corresponding to the assigned object from the corresponding current measurement cycle, that is, with the help of the corresponding latest available or detected sensor data. This can, for example, involve all remaining object trajectories in addition to the single sensor object or the corresponding object trajectory that has been assigned to the single sensor object, or at least the corresponding remaining object trajectory that already existed before the current measurement cycle. The preset filter or merging mechanism can be, for example, a Kalman filter, or include or apply such a Kalman filter. However, in principle, other filters or merging mechanisms or data filters can also be applied, especially if they effectively play a low-pass role. Through the preset filter or merging mechanism, the sensor data of multiple different sensors and / or the data already stored in the corresponding object trajectory and the sensor data from the corresponding current measurement cycle can be merged or combined with each other. This filtering or merging may mean or include weighting of old and current data. This can contribute to the advantageously smooth and consistent tracking of objects detected in uniform motion.

[0015] Furthermore, according to the invention, object tracks already assigned to single sensor objects are updated directly, separately, i.e. separately from remaining object tracks not associated with single sensor objects or independently of filtering or merging methods for said object tracks, with the aid of sensor data corresponding to the assigned single sensor object, i.e. with the aid of the respective latest available or detected sensor data from the respective current measuring cycle, without using a predefined filter, in particular without using any filter at all. In other words, depending on whether the respective object is detected by multiple sensors or is detected by a sensor different from the last or initial sensor or whether the respective object only appears in the sensor data of a single sensor, i.e. is identified as a single sensor object, the object or the associated data and the object track can be processed or manipulated differently in this case.

[0016] In previous methods or merging systems (in which sensor data from multiple sensors are merged, for example, by means of a Kalman filter), the assignment, i.e., association and updating of existing object trajectories is usually performed independently of which sensor or sensors detect or have detected the corresponding object. Therefore, there, the sensor data may be partially filtered twice unnecessarily, i.e., once inside the sensor and once when updating the corresponding object trajectory. This may mean an additional data processing workload and generate additional filtering delays in the uneven movement of the object, such as in braking or turning processes, because, for example, an average value or combined value of the current and corresponding sensor data or object data of at least one previous measurement is formed. As a result, the reaction to the corresponding critical events that are usually accompanied by uneven movement can be realized correspondingly later, for example. By additional or double filtering, the actual dynamics of the movement of the corresponding object can only be recognized as attenuated or delayed. In addition, the running time or data processing time may increase nonlinearly with the number of objects detected or possibly to be assigned or with the increase in the number of object trajectories that may be assigned. For this purpose, for example, the Hungarian method (also known as the Hungarian algorithm or the Kuhn-Mankels algorithm) can be applied, which has a complexity O(N) depending on the number of objects N. 3 ).

[0017] This problem can be solved at least partially by the present invention. To this end, in the present invention, single sensor objects or corresponding sensor data are not filtered again when updating the corresponding object trajectory. As a result, uneven movements of the corresponding objects can be detected earlier or more reliably. This in turn enables a correspondingly earlier or faster reaction to single sensor objects, i.e. objects detected by only one sensor, for example in the vehicle field to enable faster initiation of braking maneuvers or lane changes, etc. Therefore, the present invention can contribute to improved safety through faster reaction times or lower delays and reduce the data processing workload and the related energy consumption of object tracking.

[0018] In a possible embodiment of the present invention, the sensor assigns to the objects detected by means of the corresponding sensor data respectively a sensor-specific identifier that remains unchanged for the corresponding object in repeated detections of the same object by means of the corresponding sensor. Such an identifier can be, for example, an identification number or the like. Therefore, the corresponding object and the sensor detecting the corresponding object are identified or described by the corresponding identifier. The corresponding identifier is stored in the corresponding object trajectory. The single sensor object is then assigned to the corresponding object trajectory for the corresponding object, in particular only by means of these identifiers. Therefore, the assignment or association of the single sensor object detected by the sensor with the corresponding object trajectory is achieved here by means of the identifier or by means of a corresponding balance or matching between the identifier assigned to the corresponding object and also identifying the corresponding sensor and the identifier stored in the object trajectory, and the corresponding sensor can keep the identifier constant for a specific detected physical object (during its tracking inside the sensor). If, for example, an already known object, i.e., an object that has been detected at least once at an earlier time point (for which the object trajectory is also stored) is re-detected in the corresponding current measurement cycle, it can be checked whether there is already or always an object trajectory with the same identifier, which is also assigned to the object by the corresponding object trajectory. If this is the case, then the corresponding object can be directly assigned to this object track. This assignment or association can thus be carried out purely on the basis of the identifier, so that, for example, no distance criterion is required or evaluated, and no association or assignment of sensor data to a plurality of objects or an association or assignment of a specific object to a plurality of object tracks, etc., has to be calculated. The embodiment of the invention proposed here thus enables a particularly simple, labor-saving, rapid and clear or consistent assignment of a single sensor object to its object track.

[0019] In another possible embodiment of the invention, detected objects that are not identified as single-sensor objects are assigned to the object trajectory by means of a preset distance metric. One or the Mahalanobis distance can be used in particular as such a distance metric. Thus, the corresponding objects can be assigned, for example, to the object trajectory for which the minimum distance is generated according to the preset distance metric. This can provide a reliable assignment even in complex situations, in which, for example, a plurality of different sensors (possibly with different resolutions or different degrees of scattering) provide data about the object.

[0020] In another possible embodiment of the present invention, in order to update the object trajectory for a single sensor object, the dynamic characteristics of the corresponding single sensor object determined with the help of the relevant current sensor data are directly overwritten in the corresponding object trajectory for the single sensor object with the corresponding values ​​of the dynamic characteristics of the corresponding single sensor object determined with the help of the current sensor data. A direct overwriting of the corresponding data or values ​​in the object trajectory is achieved, in order to replace the updating or merging of the corresponding object trajectory with the help of filters. This can be particularly practical when the individual sensors already perform internal object tracking. Since it is not necessary to take into account the sensor data of additional or other sensors for the single sensor object, the passage of additional filtering or merging mechanisms, etc. does not provide any advantages and can therefore be omitted in order to avoid the corresponding data processing workload and the associated delays. Therefore, by the direct overwriting of the dynamic characteristics in the object trajectory proposed here, the update can be performed very quickly, simply and labor-savingly.

[0021] For example, the dynamic properties can describe or include the state of the corresponding single sensor object or the corresponding state data. For example, the position and / or the speed and / or the acceleration and / or the yaw angle and / or the yaw rate etc. of the corresponding single sensor object can be determined as the dynamic properties. They can be determined or described, for example, based on a predefined, in particular world-fixed coordinate system or based on the corresponding sensor or sensor system or in the case of motor vehicle applications, for example relative to the motor vehicle equipped with the corresponding sensor or relative to the vehicle coordinate system of the motor vehicle. The associated covariance matrix can be manipulated in the same way together with the dynamic properties or the state data.

[0022] In another possible embodiment of the present invention, in order to update the object trajectory for at least a single sensor object that has been detected at least once before the current measurement cycle, at least the static characteristics of the corresponding single sensor object are adopted or retained from the old data, or estimated over time until the corresponding current time point is reached. In the latter case, the corresponding estimated value or characteristic can be written or saved in the corresponding object trajectory. Static characteristics in this sense can be or include, for example, classification, age, classification as a stationary object or a moving object in view of the state of motion, etc. The age of the object can here, for example, indicate at which time the corresponding object was first detected. This can be sensor-independent in particular, because, for example, the first detection of the object can be achieved by means of a first sensor, such as a radar, etc., but the same object can only be detected by means of other second sensors, such as cameras, etc., in the current measurement cycle or in multiple measurement cycles. The age of such an object that is detected only by means of a second sensor in the current measurement cycle can already be determined or measured from the first detection by means of a different first sensor.

[0023] For example, the old data may be or include an object trajectory already existing for the respective single-sensor object or data stored in the object trajectory or information for the respective single-sensor object and / or old sensor data. The old sensor data may come from the same sensor with which the single-sensor object was also detected in the respective current measuring cycle or from one or more other sensors. In the latter case, the old sensor data or the static properties adopted or retained in the respective current updating may possibly come from a time when the respective object was not yet identified or classified as a single-sensor object or was identified or classified as a single-sensor object by other sensors.

[0024] In this sense, therefore, a kind of data merging can also be performed for objects currently identified as single-sensor objects, since the data or descriptions stored in the respective object trajectory can ultimately be based on sensor data from different sensors. However, this type of data merging, in which static characteristics determined from old sensor data of other sensors are adopted into the respective object trajectory or retained in it, can be performed significantly more easily and faster than the otherwise described, for example filter-based merging of current sensor data of a plurality of different sensors, which can be performed, for example, for objects not identified as single-sensor objects in the respective current measurement cycle. Thus, even if the respective single-sensor object is not detected by means of one or more other sensors in the current measurement cycle, previous or old information from other sensors can optionally be used for such static characteristics.

[0025] If the camera, for example, as a first sensor, has already classified the object in advance, then even if the object is currently identified as a single-sensor object that is only detected with the help of another second sensor, such as a radar, this classification, i.e. the corresponding static property, is not covered if necessary. This can avoid corresponding additional or double or multiple workloads, such as for re-determining the corresponding static property. If the previous sensor (which has detected the corresponding object at an earlier time, i.e. before the current measurement cycle) is better suited for determining the corresponding static property than the sensor that has detected the corresponding single-sensor object in the current measurement cycle, then at least this process can be applied. For this purpose, for example, the corresponding quality levels of different individual sensors can be preset for different static properties. If the sensor that detected the corresponding single-sensor object in the current measurement cycle is better suited for determining a specific static property, then this static property can be covered in the corresponding object trajectory if necessary, especially in the event of deviations. As a result, it is possible to achieve a management or maintenance of the static property of the single-sensor object that is always optimal or as good as possible, but on average particularly labor-saving.

[0026] In another possible embodiment of the invention, when the object trajectory is updated for the first time after the identification of the relevant object as a single-sensor object, the correlations with other sensors and / or other objects or object trajectories are respectively eliminated. Thus, one or more characteristics or data fields of the object or the corresponding object trajectory and / or corresponding sensor-internal data, descriptions or associations, such as correlations with other sensors or other sensor data or one or more other objects or object trajectories, etc., can be deleted or reset to a preset standard value, for example. As a result, the corresponding single-sensor object or the corresponding object trajectory of the single-sensor object can be manipulated, i.e., managed or updated, for example, independently and therefore particularly simply, labor-savingly and quickly.

[0027] In another possible embodiment of the invention, the object trajectory for an existing single-sensor object, i.e., one that was already detected at an earlier point in time or in an earlier measuring cycle, is automatically deleted directly if the corresponding single-sensor object is not output, i.e., detected or predicted, by the corresponding sensor within a predefined time. For example, if necessary, further sensor data or their evaluations from other sensors are not waited for. This prevents objects from being further considered or predicted without the corresponding sensor data as a basis. This reduces the data processing or management workload for object tracking and reduces the probability of erroneous reactions to objects that are no longer actually detected or reliably predicted.

[0028] In a possible improvement of the invention, the preset time includes at least one measuring cycle of the corresponding sensor. Thus, existing objects, in particular single-sensor objects or related object trajectories can be retained or maintained or activated by the corresponding sensor or sensor system as a whole, for example in at least one or exactly one measuring cycle in which the corresponding object is not detected. For this purpose, the logic or data management of the corresponding individual sensors, respectively within the sensor, can be used. As soon as the corresponding sensor no longer outputs an object, the corresponding object trajectory can be deleted immediately or directly outside the sensor if necessary. As a result, the described advantages can be achieved particularly safely and reliably, because the individual sensors can usually already have a function for tracking or predicting objects that is optimized for their corresponding operating mode.

[0029] The present invention also relates to an object tracking device, in particular an object tracking device for a motor vehicle. The object tracking device according to the present invention has a data processing device for processing sensor data of a plurality of sensors. Here, the object tracking device according to the present invention is designed to execute, in particular automatically execute, the method according to the present invention. For this purpose, the object tracking device or its data processing device can have a processing device, such as a microchip, a microprocessor or a microcontroller, etc., and a computer-readable data memory coupled thereto. In the data memory, a corresponding operation or computer program can be stored, which implements or encodes the method steps, measures or processes described in conjunction with the method according to the present invention. Then, the operation or computer program can be executed by means of the processing device in order to implement the execution of the corresponding method. The object tracking device according to the present invention can in particular be or correspond to the object tracking device mentioned in conjunction with the method according to the present invention. The object tracking device according to the present invention can, for example, include the sensor system mentioned in conjunction with the method according to the present invention, or have an interface for direct or indirect coupling with such a sensor system or a plurality of individual sensors, for example via the vehicle network of a motor vehicle equipped with the object tracking device.

[0030] The invention further relates to a motor vehicle having a sensor system having a plurality of different sensors for detecting the environment and an object tracking device according to the invention. The motor vehicle according to the invention can in particular be or correspond to the motor vehicle mentioned in connection with the method according to the invention and / or in connection with the object tracking device according to the invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] Further features of the invention can be derived from the claims, the drawings and the description of the drawings. Without departing from the scope of the invention, the features and feature combinations mentioned previously in the description and the features and feature combinations only shown below in the description of the drawings and / or in the drawings can be used not only in the respectively specified combination but also in other combinations or alone. Among them:

[0032] Figure 1 shows a schematic diagram of a motor vehicle equipped with a multi-sensor-sensor system and designed for optimized object tracking; and

[0033] Figure 2 An exemplary schematic diagram is shown for illustrating a method for sensor-based object tracking. DETAILED DESCRIPTION

[0034] Figure 1An exemplary schematic diagram of a motor vehicle 1 equipped with a sensor system 2 for environmental detection is shown. The sensor system 2 includes a plurality of individual sensors 3 and, by way of example, a preprocessing device 4. Such a preprocessing device 4 can also be integrated into each sensor 3. The individual sensors 3 can also be spatially spaced apart from each other, i.e., arranged at different locations of the motor vehicle 1. In addition, the motor vehicle 1 is equipped with a schematically shown object tracking device 5 for object tracking, i.e., for tracking objects detected by means of one or more of the sensors 3 in the corresponding environment of the motor vehicle 1. The object tracking device 5 can be coupled to the sensor system 2 or the individual sensors 3, for example, directly or via the onboard network of the motor vehicle 1 via an interface 6. Similarly, the object tracking device 5 can be integrated into the sensor system 2 or combined therewith. By way of example and schematically, the object tracking device 5 includes a processor 7 and a computer-readable data memory 8 for processing the preprocessed sensor data of the individual sensors 3 provided by the sensor system 2.

[0035] To further illustrate object tracking, Figure 2 An exemplary scheme 9 is shown. Figure 2 , an object trajectory 10 is shown, which can contain data, information, properties, identifiers (which specify the respective object and the sensor 3 detecting the object), etc., for each object detected by means of at least one of the sensors 3. The object trajectory 10 thus represents the detected physical object and its properties, such as its position, speed, orientation, classification, etc. In the same way, the object trajectory 10 can each contain a data field in which it is specified or stored which sensor 3 or sensors 3 detected the respective object, for example, in the last 500 milliseconds or 10 measurement cycles, etc., or the respective object trajectory 10 has been updated within the respective time based on the sensor data of that or those sensors 3. In each frame or measurement cycle, one or more of the sensors 3 can each provide object measurement values ​​that have been tracked over time within the sensor. Thus, the sensor 3 can not only track or estimate directly measured object properties, such as the position or orientation of the respective object, but also dynamic properties or state data, such as the speed or yaw rate of the respective object, etc., over time. These data or properties can also be stored in the object trajectory 10.

[0036] In the current measurement cycle, one or more of the sensors 3 can provide or detect corresponding current sensor data 11. Then, some or all of the data or descriptions stored in the object trajectory 11 can first be predicted, i.e., for example, extrapolated or further developed or continued to be written to the current time point or measurement or recording time point of the current sensor data 11.

[0037] Based on this and the corresponding current sensor data 11, a single-sensor object identification 13 can be performed. In this case, for example, based on the above-mentioned data area, object tracks 10 can be identified, which object tracking should be updated only based on the current sensor data 11 of a single sensor of the sensors 3 and, for example, have been updated only based on the sensor data 11 of just this sensor 3 within at least one predetermined time. In other words, an object detected here that is currently and has been seen or detected only by a single sensor of the sensors 3 within at least one predetermined time period can be identified as a single-sensor object, and the associated object track 10 is identified as a single-sensor object track 14. If, in a measurement cycle, camera data is detected as current sensor data 11, for example, all object tracks 10 can be identified as single-sensor object tracks 14, which have been updated at least once within a predetermined time period from the current time point to the past, for example within the last 10 measurement cycles, etc., but only based on the sensor data 11 of the camera and not based on the sensor data 11 of other sensors 3, or based on other objects or sensor data 11 belonging to other objects, etc.

[0038] In contrast, the remaining object trajectories 10 and, if applicable, the object trajectories 10 newly provided in the current measuring cycle or based on current sensor data 11 may be identified as multi-sensor object trajectories 15 , for example.

[0039] Based on current sensor data 11 and multi-sensor object track 15, the remaining objects that are not identified as single-sensor objects can be respectively assigned to one of multi-sensor object tracks 15 or (when an object is detected for the first time in the current measuring cycle) to a corresponding new object track 10 in a first assignment method 16. In this first assignment method 16, for the respective assignment or association between objects that are not identified as single-sensor objects and multi-sensor object track 15, for example, a predefined distance metric can be evaluated.

[0040] Multi-sensor object track 15 can be updated in a first updating method 17 using current sensor data 11, wherein objects detected in current sensor data 11 are assigned to the multi-sensor object track. In this case, filtering or merging means, such as a predefined Kalman filter or the like, are used in particular.

[0041] Separately, the second assignment method 18 can be performed for the single sensor objects and the corresponding single sensor object trajectories 14 in order to assign a corresponding one of the single sensor objects to exactly one of the single sensor object trajectories 14. In the second assignment method 18, no distance measure is calculated and evaluated. Instead, the second assignment method 18 is based on a direct assignment by means of an identifier, which is stored in the aforementioned data area and is also assigned to each detected object by the sensor 3 or the sensor system 2, for example, the preprocessing device 4. Therefore, the assignment between the single sensor objects and the single sensor object trajectories 14 in the second assignment method 18 can be realized by or based on the corresponding identifier matching. Each object trajectory 10 can, for example, store all identifiers in the aforementioned data area, which are assigned to the object by the sensor 3, which detected the corresponding managed object within the or a predetermined time period. The identifier identifies the corresponding object and the corresponding detecting sensor 3. Thus, if the identifier of a single sensor object corresponds to an identifier that is already stored in one of object trajectories 10 , then in second assignment method 18 the corresponding single sensor object can simply be assigned directly to this single sensor object trajectory 14 .

[0042] If a single sensor object track 14 which is not assigned to a single sensor object remains, this single sensor object track 14 can be deleted automatically directly within the scope of deletion 19. Since there is no corresponding assignment or association of a single sensor object to such a single sensor object track 14, this can mean that the corresponding sensor 3 no longer outputs, i.e., has neither detected nor predicted the corresponding single sensor object, and furthermore, within a predefined time period, e.g., within the last 500 ms or 10 measurement cycles, etc., the sensor data 11 of other sensors of the sensor 3 have not contributed to the corresponding object track 10.

[0043] In contrast, the second updating method 20 can be applied to the single sensor object trajectories 14 assigned to the single sensor objects in the current measuring cycle. Unlike the first updating method 17, the second updating method 20 does not use filters, in particular low-pass filters or damping filters. Instead, the dynamic properties or states or state data of the respective single sensor objects, which have been determined directly from the respective sensor 3, for example, within the scope of preprocessing the respective sensor raw data, including the respective covariance matrix if necessary, can be copied or written directly from the current sensor data 11 into the respective single sensor object trajectories 14.

[0044] Such dynamic properties may be or include various individual properties or values ​​that can be changed dynamically, such as position, speed, acceleration, yaw rate, the on or off state of the flashers, warning lights, tail lights, brake lights, etc. For example, further data or properties may be determined and updated accordingly with the aid of the current sensor data 11 and the data or properties already stored in the corresponding single sensor object trajectory 14. This may relate, for example, to the state of motion, i.e. the classification of the corresponding object as a moving object or a stationary object, the direction of motion, the age, etc. Static properties, such as the classification of the corresponding object and possible semi-static properties that usually change at most slowly compared to the dynamic properties, may also be manipulated or updated (e.g. according to corresponding specifications) or remain unchanged in the corresponding single sensor object trajectory 14.

[0045] When object track 10 is first classified or identified as single-sensor object track 14, existing cross-correlations with other objects, sensors 3 or object track 10 may be deleted or reset, if necessary. For example, correlations for determining measurement uncertainty or correlations with other coordinate systems of other sensors in sensor 3 etc. may be deleted or reset.

[0046] As a result or output of the first updating method 17 and the second updating method 20 , a corresponding updated object trajectory 10 is generated as the starting point for the next measuring cycle.

[0047] In general, the described examples show how efficient single-sensor tracking of objects can be achieved in a multi-sensor architecture or system.

[0048] Reference numerals list

[0049] 1Motor Vehicle

[0050] 2 sensor system

[0051] 3 Sensors

[0052] 4 Pretreatment device

[0053] 5 Object Tracking Device

[0054] 6 interfaces

[0055] 7 Processor

[0056] 8 Data memory

[0057] 9. Solution

[0058] 10 Object Trajectory

[0059] 11 Sensor Data

[0060] 12 Predictions

[0061] 13 Single sensor object identification

[0062] 14 Single Sensor Object Trajectory

[0063] 15 Multi-sensor object trajectories

[0064] 16 First assignment method

[0065] 17First Update Method

[0066] 18 Second assignment method

[0067] 19 Delete

[0068] 20 Second update method

Claims

1. A method (9) for sensor-based object tracking, in which method, - in each measuring cycle, detecting sensor data (11) of a sensor system (2), said sensor system comprising a plurality of individual sensors (3), - the objects represented in each case in the sensor data (11) are respectively assigned to an object trajectory (10), in which object data describing the respective object are stored, and the object trajectory (10) is updated with the aid of the sensor data (11) from the respective current measurement cycle, which corresponds to the associated object, in, - an object which has been detected for the first time before the respective current measurement cycle and which has been detected in the respective current measurement cycle only by means of respective current sensor data (11) of exactly one of the plurality of individual sensors (3) is identified as a single-sensor object, which object has been detected only by means of sensor data (11) of the sensor (3) since at least one predefined time period comprising a plurality of measurement cycles, - updating object trajectories (10, 15) which are not assigned to single-sensor objects by means of the sensor data (11) from the respective current measurement cycle, using a predefined filter, and - directly updating the object trajectories (10, 14) already assigned to the single sensor objects individually with the aid of the sensor data (11) from the respective current measurement cycle without using the predefined filters.

2. The method (9) according to claim 1, characterized in that: The sensor (3) assigns to each object detected by means of corresponding sensor data (11) a sensor-specific identifier for the object which remains unchanged in repeated detections of the same object by means of the corresponding sensor (3), the identifier being stored in the corresponding object trajectory (10, 14, 15), and the single-sensor object is assigned to the corresponding object trajectory (10, 14) by means of the identifier, in particular only by means of the identifier.

3. The method (9) according to any one of the preceding claims, characterized in that Objects that are not identified as single-sensor objects are assigned to the object trajectory (10, 15) using a predefined distance metric, in particular the Mahalanobis distance.

4. The method (9) according to any one of the preceding claims, characterized in that In order to update an object trajectory (10, 14) for a single sensor object, the dynamic properties of the respective single sensor object determined with the aid of the relevant current sensor data (11), in particular its position and / or velocity and / or acceleration and / or yaw angle and / or yaw rate, are directly overwritten in the object trajectory with the values ​​of the respective dynamic properties determined with the aid of the current sensor data (11).

5. The method (9) according to any one of the preceding claims, characterized in that In order to update the object trajectories (10, 14) for the single-sensor objects, the static properties of the respective single-sensor objects are taken from old data or estimated over time until the respective current point in time is reached.

6. The method (9) according to any one of the preceding claims, characterized in that When the object trajectory (10, 14) is updated for the first time after the associated object has been identified as a single-sensor object, correlations with other sensors (3) and / or objects are eliminated.

7. The method (9) according to any one of the preceding claims, characterized in that If the corresponding single sensor object is not output by the corresponding sensor (3) within a predefined time, the object trajectory (10, 14) with respect to the existing single sensor object is automatically and directly deleted.

8. The method (9) according to claim 7, characterized in that The preset time includes at least one measurement cycle of the corresponding sensor (3).

9. An object tracking device (5), in particular for a motor vehicle (1), having a data processing device (5, 6, 7, 8) for processing sensor data (11) of a plurality of sensors (3), wherein: The object tracking device (5) is designed to carry out a method (9) according to any of the preceding claims.

10. A motor vehicle (1) having a sensor system (2) with a plurality of different sensors (3) for detecting the surroundings and having an object tracking device (5) according to claim 9.

Images