Method for tracking at least one object in a surrounding area using a detection device, computer program product, computer-readable storage medium, and detection device

Abstract

The invention relates to a method for tracking (20) at least one object (22) in a surrounding area (24) using a detection device (12), having the steps of: - emitting and receiving detection radiation (16) into and from the surrounding area (24) using an antenna device (14) of the detection device (12); - analyzing the detected surrounding area (24) on the basis of the emitted and received detection radiation (16) using an electronic computing device (18) of the detection device (12); - determining at least one object (22) in the surrounding area (24) on the basis of the analysis using the electronic computing device (18); - specifying at least one rule system which is specific to the object tracking process (20) using the electronic computing device (18); and - checking the plausibility of the analysis of the object (22) on the basis of the specific rule system and tracking (20) the object on the basis of the plausibility using the electronic computing device (18). The invention additionally relates to a computer program product, to a computer-readable storage medium, and to a detection device (12).

Description

[0001] Description

[0002] Method for performing object tracking of at least one object in an environment by means of a detection device, computer program product, computer-readable storage medium and detection device

[0003] The invention relates to a method for performing object tracking of at least one object in an environment by means of a detection device. Furthermore, the invention relates to a computer program product, a computer-readable storage medium, and a detection device.

[0004] Object tracking is of great importance, for example, in security-relevant, traffic-related, and industrial applications. It is particularly relevant for autonomous driving and transportation, and especially for the associated safety-relevant monitoring tasks. Errors in object tracking can impair the assessment of the actual situation. This can lead to incorrect, poor, or at best unnecessary decisions and have undesirable consequences, such as potential security risks, unavailability of goods and services, or unnecessary and costly efforts in general.

[0005] It is already known from the state of the art, for example, that aircraft tracking has become increasingly important during the development of radar applications. Object tracking based on the Kalman filter, for example, was first successfully used in space applications. Improved methods, such as the extended Kalman filter or the unscented Kalman filter to better account for non-linear state transitions, have been developed. Furthermore, multi-target tracking (MTT) is essential with regard to autonomous driving and requires advanced technologies, such as the multi-hypothesis Kalman filter (MHKF) or the particle filter.In object tracking with multiple hypotheses, which leads to an enormous state space of potential events, which can also be referred to as state explosion due to exponential growth, the computational effort and methods for reducing complexity become increasingly important.

[0006] The object of the present invention is to provide a method, a computer program product, a computer-readable storage medium and a detection device by means of which improved object tracking can be carried out.

[0007] This object is achieved by a method, a computer program product, a computer-readable storage medium, and a detection device according to the independent patent claims. Advantageous embodiments are specified in the subclaims.

[0008] One aspect of the invention relates to a method for carrying out object tracking of at least one object in an environment by means of a detection device. Detection radiation is emitted and received into the environment by means of an antenna device of the detection device. The detected environment is evaluated as a function of the emitted and received detection radiation by means of an electronic computing device of the detection device. At least one object in the environment is determined as a function of the evaluation by means of the electronic computing device. At least one set of rules specific to the object tracking is predetermined by means of the electronic computing device.The plausibility of the evaluation for the object is then checked (plausibility check) based on the specific rules, and object tracking is performed based on the plausibility using the electronic computing device. This allows for improved object tracking.

[0009] In particular, the described method reduces the corresponding negative consequences of object tracking by increasing the accuracy of object tracking. This is achieved by eliminating and correcting errors within the decision-making process during object tracking. Higher object tracking accuracy can also improve the quality of object classification, which is important, for example, for security, traffic, and industrial applications.

[0010] In particular, the proposed method can be used in a variety of areas. For example, the corresponding method can be implemented in industrial plants, in motor vehicles, such as cars or aircraft, or in other specialized areas.

[0011] The environment is considered to be, in particular, the surroundings of the detection device. For example, in the case of a radar device, corresponding electromagnetic radiation can be considered the detection radiation. For example, in the case of a lidar sensor device as the detection device, light can be considered the emitted radiation. Furthermore, in the case of an ultrasonic sensor device, sound signals can be considered the corresponding radiation.

[0012] In particular, it is thus provided that decision errors during object tracking can be corrected. For example, a plausibility check is carried out for each state during a time step within the object tracking. For this purpose, rules are applied to detect implausible or sufficiently improbable states. If this implausible state is detected, for example, at least one decision in the tracking process that led to this implausible state is corrected. Error correction in the decision-making process depends in particular on the type of implausibility identified and on special rules for corrective measures.

[0013] The object tracking is repeated, for example, using this corrected decision and the measurement data available at that time, which requires, in particular, intermediate storage of measurement data. The rules or additional rules are then applied to this alternative object tracking process in order to continue to detect implausible and sufficiently improbable states during object tracking. These steps are repeated until no more implausible states exist, or until an external requirement, such as a timeout, terminates this process step.

[0014] It is self-evident for the expert that if, for example, no object is detected in the environment, then corresponding object tracking will not be carried out.

[0015] In particular, the main difference to the state of the art is that a state explosion, which is also called exponential growth, and thus a high computational effort, as occurs in object tracking with multiple hypotheses, can be avoided, while the performance of conventional algorithms for tracking objects with only one hypothesis, such as Kalman filter-based approaches, is exceeded in suitable applications.

[0016] The computational effort resulting from object tracking with multiple hypotheses may prevent its use in certain real-time applications, or limit it to unfavorable operating conditions or inappropriate levels of accuracy when only a limited number of measurements can be processed. An example is described below for autonomous driving. If the computational capacity is insufficient to evaluate all relevant hypotheses for a certain number of object-related measurements, it may be necessary to generally or temporarily reduce the vehicle speed to ensure timely braking. Alternatively, it may be necessary to discard potentially relevant measurements and thus reduce tracking accuracy, which may also lead to safety risks or the unavailability of services if unnecessary braking is required.

[0017] In contrast to the aforementioned full-fledged multi-hypothesis object tracking methods, the method according to the invention initially evaluates only one object tracking hypothesis. Alternative hypotheses are only considered if an implausible system state occurs. In this case, potential incorrect decisions leading to this implausible hypothesis are re-evaluated and corrected using a database containing corresponding rules and correction procedures.

[0018] Overall, for suitable applications, an increase in object tracking accuracy can be achieved compared to conventional approaches such as Kalman filter-based methods, while avoiding the enormous computational limitations inherent in, for example, so-called brute-force multi-hypothesis tracking methods. This can prevent unnecessary reliability problems, security risks, or adverse operating conditions in the corresponding applications.

[0019] According to an advantageous embodiment, the set of rules is specified depending on the domain of use of the detection device. Domain use can, for example, be considered the location of the detection device. For example, when used in motor vehicles / automobiles or rail transport, rules can be established so that objects that could be classified as aircraft, for example, are not taken into account. Thus, corresponding rules can be established domain-specifically, which increase the reliability of object tracking.

[0020] It is also advantageous if the set of rules is specified depending on the detection type of the detection device. For example, specific properties of radar devices, LIDAR devices, or ultrasonic sensor devices can be specified accordingly. If, for example, corresponding implausible confidence levels of objects occur depending on the detection type, this can be an indication that an error has occurred in the object tracking. This can be prevented from the outset due to the set of rules relating to the detection type. Thus, implausible states during object tracking can be reliably detected.

[0021] It is also advantageous if the set of rules is specified depending on the object type. For example, with the object types it can be that a tracking object is classified as a motor vehicle. A decision can then be made depending on whether, for example, certain speeds or movement patterns of this motor vehicle are not plausible. Furthermore, corresponding object properties, object behavior or even the continuity of the object's existence, in particular with regard to object occlusion, can be taken into account accordingly with the object type. Furthermore, physical properties of the object can also be taken into account accordingly depending on the object type, in particular on the basis of a previous object classification. In this way, improved object tracking can be achieved.

[0022] It has also proven advantageous if, in the event of a negative plausibility check, the reason for the negative plausibility check is taken into account for the re-evaluation of the measurement characteristics of the object. In particular, this provides for the reason to be taken into account in future evaluations in order to carry out a new object evaluation. If, for example, an object is classified as a single object in a first journal and as two objects in a second journal, this reason, namely that the object is two objects, can be taken into account during the new object tracking. In particular, this means that the fact that the object consists of at least two objects can be taken into account during the new object tracking. This way, errors that have already been made during the new object tracking can be prevented.

[0023] It is also advantageous if the new evaluation is carried out with respect to a new object hypothesis depending on the set of rules. In particular, a hypothesis space can be provided for the object. In a first period, the most probable hypothesis can then be used in the evaluation. Should this hypothesis prove to be incorrect, the second most probable object hypothesis can be evaluated accordingly, for example, also taking into account the reason. This allows for reliable object tracking.

[0024] It has also proven advantageous to consider previously evaluated measurement data during re-evaluation. For example, object tracking may be performed and the object tracking may initially be incorrectly assessed. However, further recording and evaluation of the measurement data still takes place. Should the plausibility check reveal that an incorrect hypothesis was assumed, the corresponding measurement data can be considered during re-evaluation, preventing any loss of information regarding the object. This enables improved object tracking.In a further advantageous embodiment, if the plausibility check fails, a time point of a previous incorrect decision is determined during the evaluation, and a new evaluation is performed using a new object hypothesis starting from that specific time point. In particular, this allows a new object hypothesis to be tested starting from the time point at which the error occurred. This prevents a loss of information during object tracking.

[0025] It has also proven advantageous to provide an additional set of rules for correcting the decision in the event of a negative plausibility check, with the evaluation being corrected based on this additional set of rules. In particular, the additional set of rules is required to correct previously detected implausible states of the object during the tracking process. The relevant rules relate to the specific implausible state that was detected. This set of rules contains the necessary corrective measures and can, for example, specify that object tracking is to be repeated from the beginning if corresponding measurement data is available in a database, assuming that, for example, two objects were present the whole time instead of one.The implementation of this assumption can, for example, be enforced, supported by "distorted" pseudo-measurements or, in the case of an implementation based on the Kalman filter, by reducing and / or increasing threshold values ​​for the distance metrics used to distinguish the objects, for example a so-called Mahalanobis distance.

[0026] It is also advantageous if an evaluation is carried out on the basis of a Kalman filter. In particular, the Kalman filter has proven to be extremely robust, which means that advantageous object tracking can be achieved using the Kalman filter. It has also proven advantageous if, in the event of a negative plausibility check, at least one threshold value of the Kalman filter is adjusted. In particular, the advantages of the Kalman filter can thus be used and an adjustment can still be carried out in the event of, for example, a negative plausibility check, so that the Kalman filter shows an improved result when evaluated again.

[0027] It is also advantageous if the process is aborted depending on a predefined event. In many cases, restrictions regarding the availability of measurement data, computing power or reaction time must be taken into account. In particular, appropriate termination conditions are provided for the process. For example, timeouts can be provided as an event. Furthermore, a default hypothesis can be used after the abort. It is also possible to go back even further in time to find the time of the incorrect decision. A completely new measurement can also be carried out for object tracking, in particular independently of the previously carried out object tracking. Furthermore, an object hypothesis can also be continued with the most recent measurement if a rollback with the temporarily stored measurements is not finished in time or if no correction rules are applicable.Furthermore, the object hypothesis can be rejected and a new object hypothesis can be initialized based on the most recent measurement. Another alternative is for the vehicle to switch to a safe state.

[0028] According to a further advantageous embodiment, the evaluation is carried out using artificial intelligence in the electronic computing device. In particular, the rule-based detection of implausible states can thus be supplemented or replaced by data-driven approaches. Artificial intelligence techniques, such as neural networks, can then be used. This allows for more reliable object tracking.

[0029] The method presented is, in particular, a computer-implemented method. Therefore, a further aspect of the invention relates to a computer program product with program code means that cause an electronic computing device, when the program code means are processed by the electronic computing device, to carry out a method according to the preceding aspect.

[0030] Furthermore, the invention also relates to a computer-readable storage medium with the computer program product according to the preceding aspect.

[0031] Yet another aspect of the invention relates to a detection device for performing object tracking of at least one object in an environment, comprising at least one antenna device and an electronic computing device, wherein the detection device is designed to perform a method according to the preceding aspect. In particular, the method is performed by means of the detection device.

[0032] Advantageous embodiments of the method are to be regarded as advantageous embodiments of the computer program product, the computer-readable storage medium, and the detection device. The detection device has material features for this purpose in order to be able to carry out corresponding method steps.

[0033] Here and in the following, an artificial neural network can be understood as software code that is stored on a computer-readable storage medium and represents one or more networked artificial neurons or can simulate their function. The software code can also contain several software code components that can, for example, have different functions. In particular, an artificial neural network can implement a non-linear model or a non-linear algorithm that maps an input to an output, where the input is given by an input feature vector or an input sequence and the output can, for example, contain an output category for a classification task, one or more predicted values ​​or a predicted sequence.

[0034] In the context of the present disclosure, an object recognition algorithm can be understood as a computer algorithm that is able to identify and localize one or more objects within a provided input data set, for example an input image, for example by defining corresponding bounding boxes or regions of interest (ROI) and, in particular, by assigning a corresponding object class to each of the bounding boxes, wherein the object classes can be selected from a predefined set of object classes. The assignment of an object class to a bounding box can be understood in such a way that a corresponding confidence value or a probability that the object identified within the bounding box belongs to the corresponding object class is provided.For example, the algorithm may provide such a confidence value or probability for each of the object classes for a given bounding box. Assigning the object class may, for example, involve selecting or providing the object class with the highest confidence value or probability. Alternatively, the algorithm may simply specify the bounding boxes without assigning a corresponding object class.

[0035] A computing unit / electronic computing device can be understood, in particular, as a data processing device that contains a processing circuit. The computing unit can therefore, in particular, process data to perform computing operations. This may also include operations for performing indexed access to a data structure, for example, a look-up table (LUT).

[0036] The computing unit can in particular contain one or more computers, one or more microcontrollers and / or one or more integrated circuits, for example one or more application-specific integrated circuits (ASICs), one or more field-programmable gate arrays (FPGAs), and / or one or more single-chip systems (SoCs). The computing unit can also contain one or more processors, for example one or more microprocessors, one or more central processing units (CPUs), one or more graphics processing units (GPUs), and / or one or more signal processors, in particular one or more digital signal processors (DSPs). The computing unit can also contain a physical or virtual network of computers or other of the aforementioned units.

[0037] In various embodiments, the computing unit includes one or more hardware and / or software interfaces and / or one or more memory units.

[0038] A memory unit can be a volatile data memory, for example a dynamic random access memory (DRAM) or a static random access memory (SRAM), or a non-volatile data memory, for example a read-only memory (ROM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), an electrically erasable programmable read-only memory (EEPROM), a flash memory or flash EEPROM, a ferroelectric random access memory (FRAM), a magnetoresistive random access memory (MRM), or a memory with a random access frequency of up to 128 MB. Access,MRAM (magnetoresistive random access memory) or as phase-change random access memory, PCRAM (phase-change random access memory).

[0039] An environmental sensor system, in this case in particular the detection device, can be understood, for example, as a sensor system capable of generating sensor data or sensor signals that map, represent, or reproduce an environment of the environmental sensor system. In particular, the ability to detect electromagnetic or other signals from the environment is not sufficient to consider a sensor system an environmental sensor system. For example, cameras, radar systems, lidar systems, or ultrasonic sensor systems can be understood as environmental sensor systems. A combination of sensors is also possible.

[0040] Algorithms for automatic visual perception, which can also be called computer vision algorithms, machine vision algorithms or machine vision algorithms, can be considered as computer algorithms for automatically performing a visual perception task. A visual perception task, also called a computer vision task, can be understood, for example, as a task for extracting visual information from image data. In particular, the visual perception task can, in principle, in some cases be performed by a human who is able to visually perceive an image corresponding to the image data. In the present context, however, visual perception tasks are also performed automatically, without the need for human assistance.

[0041] For example, a computer vision algorithm can

[0042] An image processing algorithm or an image analysis algorithm that is or was trained by machine learning and can be based, for example, on an artificial neural network, in particular a convolutional neural network. The computer vision algorithm can, for example, include an object recognition algorithm, an obstacle detection algorithm, an object tracking algorithm, a classification algorithm, a semantic segmentation algorithm, and / or a depth estimation algorithm.

[0043] Corresponding algorithms can also be carried out analogously based on input data other than images that can be visually perceived by humans. For example, point clouds or images from infrared cameras, lidar systems, etc. can also be evaluated using appropriately adapted computer algorithms. Strictly speaking, the corresponding algorithms are not algorithms for visual perception, since the corresponding sensors can operate in ranges that are visually imperceptible, i.e. not perceptible to the human eye, for example in the infrared range. For this reason, such algorithms are referred to as algorithms for automatic perception in the context of the present invention. Algorithms for automatic perception therefore include algorithms for automatic visual perception, but are not limited to this with regard to human perception.Consequently, an algorithm for automatic perception according to this understanding can include a computer algorithm for automatically performing a perception task, which is or has been trained, for example, by machine learning and can in particular be based on an artificial neural network. Such generalized algorithms for automatic perception can also include object detection algorithms, object tracking algorithms, depth estimation algorithms, classification algorithms and / or segmentation algorithms, for example semantic segmentation algorithms. If an artificial neural network is used to implement an algorithm for automatic visual perception, a frequently used architecture is that of a convolutional neural network, CNN. In particular, a 2D CNN can be applied to corresponding 2D camera images.CNNs can also be used for other automatic perception algorithms. For example, 3D CNNs, 2D CNNs, or ID-CNNs can be applied to point clouds, depending on the spatial dimensions of the point cloud and the details of the processing.

[0044] The result or output of an automatic perception algorithm depends on the specific underlying perception task. For example, the output of an object detection algorithm may contain one or more bounding boxes defining a spatial position and optionally an orientation of one or more corresponding objects in the environment and / or corresponding object classes for the one or more objects. An output of a semantic segmentation algorithm applied to a camera image may contain a pixel-level class for each pixel of the camera image. Analogously, an output of a semantic segmentation algorithm applied to a point cloud may contain a corresponding point-level class for each of the points. The pixel-level or point-level classes, respectively, may define an object type to which the respective pixel or point belongs.

[0045] For use cases or application situations that may arise during the method and are not explicitly described here, it may be provided that, in accordance with the method, an error message and / or a request for user feedback is issued and / or a default setting and / or a predetermined initial state is set. Regardless of the grammatical gender of a particular term, this includes persons with male, female, or other gender identities.

[0046] Further features of the invention emerge from the claims, the figures and the description of the figures. The features and combinations of features mentioned above in the description as well as the features and combinations of features mentioned below in the description of the figures and / or shown in the figures can be encompassed by the invention not only in the respectively specified combination, but also in other combinations. In particular, the invention can also encompass embodiments and combinations of features which do not have all the features of an originally formulated claim. Furthermore, the invention can encompass embodiments and combinations of features which go beyond the combinations of features set out in the references to the claims or deviate from them.

[0047] Showing:

[0048] FIG 1 shows a schematic plan view of an embodiment of a motor vehicle with an embodiment of a detection device according to the invention; and

[0049] FIG 2 shows a schematic flow diagram according to an embodiment of the method.

[0050] The invention is explained in more detail below with reference to specific exemplary embodiments and associated schematic drawings. In the figures, identical or functionally equivalent elements may be provided with the same reference numerals. The description of identical or functionally equivalent elements may not necessarily be repeated for different figures.

[0051] FIG 1 shows a schematic plan view of a motor vehicle 10. In the present example, the motor vehicle 10 is shown as a car. However, the motor vehicle 10 can also be, for example, an aircraft or a ship or a rail vehicle or a driverless transport system (AGV - Automated Guided Vehicle). The motor vehicle 10 has a detection device 12. The detection device 12 has at least one antenna device 14 for emitting and receiving detection radiation 16. Furthermore, the detection device 12 has at least one electronic computing device 18.

[0052] The detection device 12 is provided in particular for carrying out an object tracking 20. In the present exemplary embodiment, three points in time t1, t2 and t3 are shown. FIG. 1 shows in particular that, for example, the object tracking 20 tracked a single object 22 at the first point in time t1 and also a single object 22 at the second point in time t2. At the third point in time t3, it has been determined that the object 22 classified as the single object 22 is, however, two independent objects 22a, 22b. This can have been determined, for example, because the objects 22a and 22b have approached the motor vehicle 10. Thus, for example, a resolution of the detection device 12 can be increased, which is why two objects 22a, 22b are tracked at time t3.

[0053] FIG. 2 shows a schematic flow diagram according to one embodiment of the method. In particular, FIG. 2 shows the situation already described in FIG. 1.

[0054] In a method for object tracking 20 according to the invention, it is provided that the detection radiation 16 is emitted and received, in particular emitted into an environment 24 of the detection device 12. The detected environment 24 is evaluated as a function of the emitted and received detection radiation 16, in particular by means of the electronic computing device 18. The at least one object 22 in the environment 24 is determined as a function of the evaluation by means of the electronic computing device 18. At least one set of rules specific to the object tracking 20 is specified by means of the electronic computing device 18.The plausibility (plausibility check 32) of the evaluation of the object 22 is checked depending on the specific set of rules and the object tracking 20 is carried out depending on the plausibility by means of the electronic computing device 18.

[0055] In this case, it can be provided that the set of rules is specified depending on a domain use of the detection device 12. Furthermore, the set of rules can be specified depending on a detection type of the detection device 12. Furthermore, it can be provided that the set of rules is specified depending on an object type of the object 22.

[0056] FIG 2 shows in particular that at time t1 a first object hypothesis 26 is evaluated. In the first object hypothesis 26 it is assumed in particular that the object 22 is a single object 22. At time t1 and at time t2 this object hypothesis is pursued further. At the third time t3 however it is recognized that the first object hypothesis 26 was incorrect. It can now be provided, for example, that a return is made to the time at which a respective error was recognized in order to carry out a new object hypothesis 28. In the present exemplary embodiment it is thus possible to return from the third time t3 to the first time t1, with the times t2 and t3 then again being evaluated as a function of the further object hypothesis 28.

[0057] In particular, it can be provided that, in the event of a negative plausibility check 30, the reason for the negative plausibility check 30 is taken into account for a renewed evaluation of the measurement characteristics of the object 22. In particular, the new evaluation is then carried out with regard to the new object hypothesis 28 in accordance with the set of rules. In this case, previously evaluated measurement data relating to the object 22 are taken into account in the renewed evaluation of the measurement characteristics.

[0058] Furthermore, it can be provided that in the case of the negative plausibility check 30, a further set of rules is provided for the decision correction, wherein a correction of the evaluation is carried out depending on the further set of rules.

[0059] It can further be provided that the evaluation is carried out based on a Kalman filter. If the plausibility check 30 is negative, at least one threshold value for the Kalman filter can then be adjusted accordingly.

[0060] In particular, it is thus provided that a set of rules for detecting implausible states during object tracking 20 is stored for the electronic computing device 18, which in particular has a database. This set of rules contains relevant criteria and corresponding threshold values ​​and can be area- and application-specific, which can also be referred to as domain-specific. In addition, parameterization can take place in order to comprehensively take into account risks resulting from incorrect decisions. In particular, the costs of an incorrect decision can thus also be taken into account. This set of rules can, for example, take into account object properties, object behavior or the continuity of object existence, for example with regard to object occlusion. This means that both generic criteria, for example physical, and domain-specific restrictions can be used or configured.

[0061] For example, if two objects 22a, 22b approach the detection device 12, the detection device 12 can initially only detect a relatively large object 22. Given sufficient proximity, particularly at time t3, the detection device 12 clearly detects two independently moving objects 22a, 22b. Application-specific rules can indicate that this is implausible and that two objects 22a, 22b must have already been present.

[0062] A second set of rules, in particular a further set of rules, is required to correct previously detected implausible states during object tracking 20. The relevant rules can, for example, also be stored in the electronic computing device 18 and relate to the specific implausible state that was detected. These rules contain the required corrective measures and can, for example, now specify that object tracking 20 is to be repeated from the beginning. This is particularly the case at the first time t1. In particular, this can be provided if, for example, measurement data is available in the corresponding database, it being assumed that two objects 22a, 22b were present the entire time.The implementation of this assumption can be enforced directly, supported by biased pseudo-measurement or, in the case of a Kalman filter-based implementation, by reducing the thresholds for the distance metrics used to distinguish the objects 22a, 22b, for example the so-called Mahalanobis distance.

[0063] As a result, the error correction method according to the invention not only has the potential to improve the object tracking 20 itself, but also the quality of the object classification based on the length of the object observation time and the corrected incorrect decisions. Furthermore, the rule-based detection of implausible states can be supplemented or replaced by data-driven approaches; in particular, artificial intelligence, for example a neural network, can be used for this purpose. In many cases, restrictions with regard to the availability of measurement data, computing power or reaction time must be taken into account. In particular, appropriate termination conditions are provided for the method. For example, timeouts can be provided. Furthermore, after termination, a default hypothesis, for example, can be used.It is also possible to go back even further in time to find the time of the incorrect decision. A completely new measurement can also be carried out for object tracking, particularly independently of the previously carried out object tracking. Furthermore, an object hypothesis can be continued with the most recent measurement if a rollback with the temporarily stored measurements is not completed in a timely manner or if no correction rules are applicable. Furthermore, an object hypothesis can be discarded and a new object hypothesis can be initialized with the most recent measurement. Another alternative is for the motor vehicle to switch to a safe state.

Claims

Patent claims 1. A method for performing object tracking (20) of at least one object (22) in an environment (24) by means of a detection device (12), comprising the steps: - transmitting and receiving detection radiation (16) into the environment (24) by means of an antenna device (14) of the detection device (12); - evaluating the detected environment (24) as a function of the emitted and received detection radiation (16) by means of an electronic computing device (18) of the detection device (12); - Determining at least one object (22) in the environment (24) as a function of the evaluation by means of the electronic computing device (18); - specifying at least one set of rules specific for the object tracking (20) by means of the electronic computing device (18); and - Checking the plausibility of the evaluation of the object (22) depending on the specific set of rules and carrying out the object tracking (20) depending on the plausibility by means of the electronic computing device (18).

2. Method according to claim 1, characterized in that the set of rules is specified as a function of a domain use of the detection device (12).

3. Method according to claim 1 or 2, characterized in that the set of rules is specified as a function of a detection type of the detection device (12).

4. Method according to one of the preceding claims, characterized in that the set of rules is specified as a function of an object type of the object (22).

5. Method according to one of the preceding claims, characterized in that in the case of a negative plausibility check (30), the reason for the negative plausibility check (30) is taken into account for a renewed evaluation of the measurement characteristics of the object (22).

6. Method according to claim 5, characterized in that the new evaluation is carried out with respect to a new object hypothesis (28) in dependence on the set of rules.

7. Method according to one of claims 5 or 6, characterized in that already evaluated measurement data relating to the object (22) are taken into account in the renewed evaluation.

8. Method according to one of the preceding claims, characterized in that in the case of a negative plausibility check (30) a time (tl, t2, t3) of a previous incorrect decision is determined during the evaluation and a new evaluation is carried out under a new object hypothesis (28) from the determined time (tl, t2, t3).

9. Method according to one of the preceding claims, characterized in that in the case of a negative plausibility check (30), a further set of rules is provided for a decision correction, wherein a correction of the evaluation is carried out depending on the further set of rules.

10. Method according to one of the preceding claims, characterized in that an evaluation is carried out on the basis of a Kalman filter and / or in the event of a negative plausibility check (30) at least one threshold value of the Kalman filter is adjusted.

11. Method according to one of the preceding claims, characterized in that Depending on a predefined event, the procedure is aborted.

12. Method according to one of the preceding claims, characterized in that the evaluation is carried out by means of an artificial intelligence of the electronic computing device (18).

13. Computer program product with program code means which cause an electronic computing device (18) to carry out a method according to one of claims 1 to 12 when the program code means are processed by the electronic computing device (18).

14. A computer-readable storage medium comprising a computer program product according to claim 13.

15. Detection device (12) for carrying out object tracking (20) of at least one object (22) in an environment (24), with at least one antenna device (14) and an electronic computing device (18), wherein the detection device (12) is designed to carry out a method according to one of claims 1 to 12.

Images