Method and radar device for radar-based sizing of objects, and corresponding motor

Abstract

The invention relates to a method and a radar arrangement (12, 18) for radar-based dimensional classification of an object (14). The invention also relates to a correspondingly designed motor vehicle (10). In the method, radar signals (24) reflected by an object (14) are detected in a plurality of receiving channels (20). The detected reflected radar signal (24) is then evaluated with respect to its phase profile over the plurality of receiving channels (20). In the event of a phase jump in the phase curve, the respective object (14) is classified as an extended object (14).

Description

Technical Field

[0001] The invention relates to a method and a radar device for radar-based size classification of detected objects and also to a correspondingly designed motor vehicle. Background Art

[0002] Radar systems in motor vehicles can be used to detect objects in the corresponding environment. In this regard, radar systems can have some advantages over other types of sensors, such as a higher effective range and automatic and precise determination of speed by utilizing the Doppler effect. However, conventional radar systems usually have a limited angular resolution, so that even at greater distances, extended objects, such as the rear side of a truck, etc., are only detected as point objects. For example, a greater angular resolution, i.e. a more detailed separation of radar echoes from different parts or regions of an extended object, can be achieved by a larger aperture of the radar antenna used. However, this is not always feasible due to structural space limitations and cost limitations.

[0003] EP 2 215 497 B1 describes an angle-resolving radar sensor as a solution. The radar sensor has an optical lens and an antenna element arranged at a distance from the optical lens, which can move relative to the lens in a direction transverse to the optical axis of the lens. The antenna element can move together with an associated high-frequency module for generating a radar signal to be transmitted in a direction transverse to the optical axis of the lens. Therefore, a radar sensor should be provided that can achieve simple control of directional characteristics and high angular resolution in a simple structure.

[0004] EP 3 161 514 B1 describes a method for locating radar targets using an angle-resolving MIMO-FMCW radar sensor. In this method, a received signal is mixed with a transmitted signal to form a baseband signal, and the angle of the located radar target is determined using the amplitude and phase relationship between the baseband signals, which are obtained for different selections of antenna elements of the radar sensor for transmission and reception. Therefore, a time multiplexing method for a MIMO radar is to be described, which allows a more accurate angle estimation.

[0005] EP 2 270 541 B1 describes a method for determining the angle of incidence and / or the distance of a sensor from an object in space using a synthetic aperture, wherein an echo profile is received at a plurality of aperture points, respectively. In particular, the method there should be able to determine the angle of incidence independently of the distance to the object or transponder. Summary of the invention

[0006] The object of the present invention is to enable improved radar-based environment detection in a particularly efficient manner.

[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 subject matters of the independent claims, if applicable in conjunction with one or more dependent claims, should be regarded at least similarly as the features, advantages and possible embodiments of the corresponding subject matter of the other independent claims and any possible combination of the subject matters of the independent claims.

[0008] The method according to the present invention can be applied to radar-based size classification or size classification or size identification or size estimation of objects detected by radar. In particular, the method according to the present invention can be applied to motor vehicles, but is not necessarily limited to this application. Radar signals or radar pulses can be transmitted respectively by means of a radar device or at least one radar transmitting antenna. This can be part of the method or be performed before the actual method according to the present invention. In the method steps of the method according to the present invention, the radar signals generated or correspondingly reflected by the object are detected in multiple receiving channels or in multiple receiving channels. Such receiving channels can be realized, for example, by real and / or virtual receivers or receiving antennas, virtual receiver arrays or antenna arrays, when necessary, using multiple transmitting antennas, etc. In the latter case, the number of receiving channels can correspond to the product of the number of transmitting antennas and the number of receiving antennas.

[0009] In the sense of the present invention, detecting a reflected radar signal may, for example, mean or include: receiving or measuring the reflected radar signal by at least one receiving antenna or radar device and / or intercepting it via a corresponding interface, or, for example, reading corresponding raw data from a data memory or the like.

[0010] In a further method step of the method according to the invention, the detected reflected radar signals, i.e., the corresponding raw data detected or present in the radar device used, are analyzed or evaluated with respect to their phase curves on a plurality of receiving channels. The phase curves can be provided by phase sequences or phase positions of corresponding single signals. These single signals can correspond to single signals or signal parts of or in or from the individual receiving channels. Different phases or phase positions in different receiving channels can correspond, for example, to reflections of radar signals at points or regions of corresponding objects at different depths (i.e., at different distances from the radar device) and / or reflections of radar signals by corresponding objects from different azimuths and / or elevations.

[0011] In a further method step of the method according to the invention, when there is at least one phase jump in the phase curve on the receiving channel or the single signal, the corresponding object is classified or categorized as an extended object, and therefore in particular not classified or categorized as a point object. Such a classification as an extended object may, for example, mean that a preset minimum size of the object is assumed or output. Similarly, for example, the size of the corresponding object can be estimated with the help of the detected radar signal or with the help of the phase curve or at least one phase jump, so that the corresponding object can be classified more accurately with respect to the size of the corresponding object. For this purpose, for example, a preset assignment table, a preset model or a preset algorithm and / or a correspondingly trained machine learning device, such as a correspondingly trained artificial neural network, etc. can be used.

[0012] The invention is therefore based on the knowledge that the size of the remaining object can be inferred at least to a certain extent by means of the phase curve, i.e. ultimately by means of the phase information originally contained in the radar signal detected in a plurality of receiving channels or encoded. In this context, phase jumps, i.e. discontinuities in the phase curve in a plurality of receiving channels, can in particular characterize extended objects. Thus, in particular extended objects can be identified which, although reflecting radar signals from a plurality of locations or over a certain area, cannot be resolved or identified as extended objects (e.g. in the form of a plurality of radar detections belonging to the same object) due to the limited angular resolution of the radar device used.

[0013] The corresponding single signal of each receiving channel, i.e., physically real and / or virtual receiving antenna, contains information about the corresponding receiving direction of the radar signal in its corresponding phase or phase position. Phase change or phase rotation, i.e., the phase curve on the lateral and / or vertical extension range of the remaining object can produce a Fourier spectrum corresponding to or characterizing the corresponding extended object as the result of the Fourier transform of the detected radar signal. The generation and analysis or evaluation of this Fourier spectrum can, for example, be a part of the corresponding radar signal evaluated about the phase curve. If the complex pointer of each phase or phase position, i.e., the detected radar signal is not constant, or there is no vertical and / or lateral continuous advancement on the extension range of the real and / or virtual aperture of the radar device used, i.e., further rotation, then this can indicate a hidden target or partial target or sub-target or sub-object in the extended target or object according to the knowledge on which the present invention is based, because the corresponding different reflection points or regions distributedly arranged on the corresponding object can contribute to the antenna signal detected at the end. A corresponding evaluation of the deviation of the phase position of the respective complex valuable received signal or individual signals from the receive channel, for example with respect to a constant, identical phase or a continuous or linear phase curve, can thus indicate an extended target, ie an object.

[0014] The corresponding classification or categorization as an extended object (or as a point object if no corresponding phase jump or correspondingly characterized Fourier spectrum, etc., exists) can be used or taken into account in further signal or data processing. For example, even if the corresponding object cannot be identified as an extended object alone by means of Doppler data or distance data determined based on radar or by means of the angular resolution or angular separation of the radar device used, a further classification of the corresponding object can be performed based on this, for example, or a plausibility check can be performed on the further classification. Thus, in the case of a motor vehicle application, when a truck traveling ahead is traveling at a constant distance and at the same speed, the truck can be identified or categorized as a further vehicle based on the classification or categorization as an extended object obtained by means of the method, for example. This enables a correspondingly improved, more suitable, safer reaction of, for example, one or more further assistance systems of the motor vehicle, for example, for at least semi-automatic vehicle guidance, etc.

[0015] The present invention thus enables more detailed, more accurate or more reliable environmental recognition solely through improved processing or use of the actually detected radar signals, in particular without requiring larger antennas or a greater number of antennas than conventional radar systems currently already available for motor vehicles.

[0016] In a possible design of the present invention, multiple physical antennas or virtual antenna arrays are used or used as multiple receiving channels. Multiple physical antennas can be or include multiple receiving antennas and / or multiple transmitting antennas. With the help of multiple receiving antennas, multiple single signals can be directly received, which can correspond to or correspond to each receiving channel. With the help of multiple transmitting antennas, correspondingly adapted or changing radar signals can be sent, and the radar signals can then be received using one or more physical receiving antennas. Therefore, multiple receiving channels can be generated, and the number thereof can correspond to the product of the number of transmitting antennas and the number of receiving antennas. Here, the phase curve can be detected or evaluated on multiple real receiving antennas and / or multiple virtual antennas of the virtual antenna array. Through the design proposed here of the present invention, the antenna used (i.e., the design of the radar device used) can be adapted according to the needs and flexibly adapted, and the method according to the present invention can still be applied to correspondingly different variants or designs. Therefore, the method according to the present invention can be applied to different conditions or applications, such as different available structural spaces and / or different cost budgets, etc.

[0017] In a possible embodiment of the invention, the phase position of the individual receiving channels is determined using the I&Q method (in-phase and quadrature method) and / or the Hilbert transformation, and thus the phase curves on the individual receiving channels are effectively determined. Thus, the different phase positions, i.e. the phase curves, can be determined directly from the detected radar signals or the corresponding raw data using known methods. Thus, the method according to the invention can be implemented particularly simply and efficiently.

[0018] In another possible embodiment of the present invention, in order to detect phase jumps, it is checked whether the phase position of each radar detection and / or each Doppler class (i.e., Doppler unit) grows or changes uniformly or constantly, i.e., in accordance with the individual detection of a point object or a separate object. In particular, a comparison with a preset threshold value can be performed here. When the phase change, i.e., the phase jump between two different receiving channels is greater than the preset threshold, or the phase curve deviates from a constant or uniformly increasing phase curve by at least the preset threshold, the object can be classified as an extended object. In order to determine the phase position, available data can be extracted or derived from a corresponding radar hexahedron or radar cube, i.e., a common three-dimensional data structure containing received radar signals or radar data. This may be particularly relevant to the data before the data is transformed by means of a particularly fast Fourier transform, i.e., further processed to determine or distinguish the angle at which the radar detection or the object belonging to the radar detection occurs from the radar device used. However, with the help of the angle obtained by this Fourier transform, the corresponding object or detection assignment of the detected radar data can be performed so that the phase curve of each object can be reliably obtained. The embodiment of the invention proposed here is based on the basic knowledge that extended objects can be detected in this way based on radar even if they are not detected as extended objects in a conventional manner, for example due to their distance from the radar device and / or the limited angular resolution or angular separation of the radar device.

[0019] In another possible embodiment of the present invention, at least one Fourier transformation, in particular at least one fast Fourier transformation (FFT), is applied to the detected radar signal in order to determine the corresponding detection angle of the object or each object or each radar detection, in which the corresponding object, i.e. the object responsible for the corresponding radar detection or the detected radar signal, is located from the radar device used. In addition, the width of the central peak or main peak of the spectrum generated, i.e. the spectrum obtained as a result of the Fourier transformation or the generated spectrum is determined. Then, the size of the corresponding object is derived or estimated according to the width or based on the width. For example, this can be performed according to the angle difference corresponding to the width, taking into account the distance of the corresponding object also determined by means of the radar signal, with the help of a preset assignment table between the spectral width and the size or extension range of the object, with the help of a corresponding preset model or preset algorithm, with the help of a corresponding trained machine learning device, etc. The embodiment of the present invention proposed here enables a particularly low-cost and effective implementation and application of the method according to the present invention. For example, the Fourier transformation can originally be performed within the scope of conventional data or signal processing of the radar device in order to determine the detection angle, i.e. to locate the detected object in space. The width of the resulting spectrum can then be determined with particularly little additional computing effort.

[0020] In another possible embodiment of the present invention, a statistical evaluation is performed when evaluating the detected reflected radar signal. Here, at least one preset statistical measure is obtained and used, i.e., taken into account for detecting phase jumps. The standard deviation of the phase or phase position on the receiving channel can be obtained as such a statistical measure in particular. Then, the standard deviation or multiple standard deviations can be compared with a preset threshold value. Here, when the preset threshold value is reached or exceeded, the phase jump can be detected and the corresponding object can be classified as an extended object accordingly. The statistical analysis proposed here can be particularly simple to implement and execute, and can flexibly realize the rapid, robust and reliable recognition of extended objects, i.e., the corresponding size classification.

[0021] In a possible extension of the present invention, the value curve of the statistical measure determined for the corresponding object is determined on a plurality of detected radar signals from a plurality of radar measurement cycles and is taken into account for the classification of the corresponding object. This can usually be realized simply because the detected objects are usually tracked, i.e., tracked, independently of the size classification, for example, based on radar or sensors. Only when the value curve is consistent, i.e., for example, the statistical measure indicates an extended object in a plurality of radar measurement cycles, i.e., for example, exceeds the preset threshold value mentioned in other cases multiple times or continuously, can the object be classified as an extended object, for example. By the design scheme of the present invention proposed here, a particularly robust and reliable size classification can be achieved, because single or short-lived outliers compared to the total duration of a plurality of, in particular, consecutive radar measurement cycles, erroneous measurements, interference effects, etc., for example, cannot determine or change the size classification.

[0022] In another possible design of the present invention, the detected reflected radar signal and / or the data derived from the radar signal are provided or transmitted as input, i.e., as input data, to a machine learning device that is trained for the size classification of objects based thereon. Such a machine learning device may in particular be or include an artificial neural network or the like. The data derived from the radar signal may for example be or include a phase curve and / or the statistical measures mentioned in other cases or the like. Based on the provided or transmitted input data, the corresponding object is classified by the trained machine learning device in particular as an extended object or a non-extended object or a point object with respect to the size of the object. The machine learning device may more accurately estimate the size of the extended object, for example by classifying it into one of a plurality of preset different size categories, at least for the extended object.

[0023] The design of the invention proposed here is based on the knowledge that information or patterns that may be related to the size or extension of the corresponding object can be contained in or encoded in the reflected radar signal detected on multiple receiving channels. Therefore, the radar signal reflected by the extended object can have characteristics that characterize this extended object, which can be different from the characteristics that characterize the point object. These information or patterns may be difficult to define or recognize accurately, but can be automatically and particularly robustly and completely learned with the help of machine learning. Therefore, a correspondingly robust and reliable size classification can be performed by a correspondingly trained machine learning device.

[0024] The size classification of objects proposed here with the aid of a trained machine learning device can be particularly accurate, in particular in marginal cases, for example, compared to other methods. In addition, however, such other methods can be performed, for example, evaluation by Fourier transform or the like. For example, such other methods can achieve particularly reliable size classification in standard or general situations, and / or serve as a guarantee or plausibility check for the size classification output by the machine learning device. If multiple size classification methods, i.e., for example, machine learning-based methods and Fourier transform-based methods, are applied for size classification, their results can be combined with one another. For this purpose, for example, maximum likelihood classification or weighting or the like can be applied. By combining multiple size classification methods, size classification can ultimately be performed particularly robustly and reliably.

[0025] The present invention also relates to a radar device, in particular a radar device for a motor vehicle. The radar device according to the present invention has a signal or data processing device and is designed to automatically execute the method according to the present invention in particular. For this purpose, the radar device or the signal or data processing device may include, for example, corresponding circuits and / or processing devices, such as microchips, microprocessors, microcontrollers, etc. and computer-readable data storage devices coupled thereto. In the data storage device, for example, corresponding operation or computer programs and / or artificial neural networks mentioned in other cases as necessary may be stored. The operation or computer program may encode or implement the method steps, measures or processes mentioned in conjunction with the method according to the present invention or corresponding control instructions, and may be executed by means of a processing device in order to execute the method according to the present invention. The radar device according to the present invention may also have at least one radar antenna and a possible signal generating device. Therefore, the radar device according to the present invention may be or correspond to the radar device mentioned in conjunction with the method according to the present invention or the radar system mentioned in conjunction with the method according to the present invention.

[0026] The invention also relates to a motor vehicle having a radar device according to the invention. The motor vehicle according to the invention can also have at least one antenna or at least one radar transmitter and receiver, provided that they are not part of the radar device. Thus, the motor vehicle according to the invention can be designed to carry out the method according to the invention. In particular, the motor vehicle according to the invention can be or corresponds to the motor vehicle mentioned in connection with the method according to the invention and / or in connection with the radar device according to the invention.

[0027] Further features of the invention can be derived from the claims, the drawings and the description of the drawings. The features and feature combinations mentioned previously in the description and the features and feature combinations shown separately 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 without departing from the scope of the invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] The drawing shows in a single figure a schematic diagram of a motor vehicle with a system designed for radar-based recognition of objects in the environment. DETAILED DESCRIPTION

[0029] Conventional vehicle radars detect point targets, such as spheres, at their geometric center in the vertical and lateral extension range. Real extended targets, such as the rear side of the vehicle in front, are routinely resolved as multiple targets by the radar of the following vehicle only when separation can be made vertically and laterally by Doppler velocity, distance or angle separation. When approaching or following the rear wall of a vehicle, the first two criteria are usually excluded, for example, despite the presence of a crossbeam or similar structure. Depending on the longitudinal distance along the direction of travel, the lateral and vertical angle separation capabilities of conventional vehicle radars are also insufficient to show multiple points or multiple point targets. In practice, however, in principle, information about the width or extension range of the corresponding target can be present in a vehicle radar with multiple receivers.

[0030] To explain the use of this information, Figure 1 An exemplary schematic overview of a motor vehicle 10 is shown which is equipped with a radar system 12. Thus, objects present in the detection range of radar system 12, which are referred to here as objects 14, can be detected.

[0031] Radar system 12 comprises an antenna device 16 and a data or signal processing device 18, which is only schematically shown here. Antenna device 16 can detect signals in a plurality of reception channels 20, which are schematically shown here. Thus, a radar pulse transmitted by antenna device 16 can be reflected on extended object 14 at different locations 22. The corresponding reflected radar signal 24 is also schematically shown here. The reflected radar signal 24 can be detected in a plurality of reception channels 20, which each provide a single signal.

[0032] These individual signals from the individual receiving channels 20 contain information about the respective receiving direction from or along which the radar signal 24 arrives at the respective receiving channel 20 in the phase or phase position of the respective individual signal. Accordingly, corresponding phase curves are generated on the plurality of receiving channels 20 .

[0033] Such a phase curve or a corresponding phase rotation over the lateral and / or vertical extension of the object 14 can generate a corresponding or characterizing spectrum of the detected radar signal 24 or individual signals after a Fourier transformation, which can be performed, for example, by the signal processing device 18. A corresponding analysis of the phase curve can be performed by the signal processing device 18. For this purpose, the signal processing device 18 can detect individual signals from the receiving channels 20, for example, via the interface 26 and process them by means of the processor 28 and the data memory 30. Thus, phase jumps that may be present, i.e. deviations of the phases or phase positions of different individual signals from different receiving channels 20, can be detected. Such phase jumps or deviations can indicate a certain extension of the object 14.

[0034] Thus, if a corresponding phase jump or a corresponding deviation of the phases or phase positions of the individual signals from one another is detected, the object 14 can be classified or assigned as an extended object.

[0035] The phase or phase position of each complex-valued received signal or single signal from different receiving channels 20 can be directly obtained or determined, for example, by means of I&Q methods or after applying a Hilbert transform or the like. Before the Fourier transform is applied to determine the detection angle, the corresponding data can also be extracted or derived from the radar cube, for example. Based on this, in each detection in the corresponding detection area and / or, for example, in each Doppler unit (Doppler-Bin), it can be checked whether the phase or phase position of the single signal is the same, whether it has a constant growth, that is, a constant increase on the receiving channel 20, or whether it shows a deviation therefrom. For this purpose, known mathematical methods can be applied, such as determining the width of the spectrum generated by the Fourier transform, statistical methods, such as determining the standard deviation of the phase position on the receiving channel 20, etc. Based on the automatic tracking of the corresponding object 14, the corresponding statistical measure can also be determined over multiple measurements or radar cycles.

[0036] The problem that can be solved by the scheme described here is that conventionally, in order to separate or identify separately, the targets (for example, different parts 22 of the object 14) must be spaced apart from each other in space, so that the corresponding spectra can be clearly distinguished. However, because the fast Fourier transform and other time domain transformation methods are subject to certain window restrictions, for example, due to limited measurement time and the resulting frequency uncertainty or spectral uncertainty, even for the conversion of a single frequency, a spectrum with a certain width or peak width will always be produced. Therefore, the other peaks present next to the main peak can actually be covered by the width of the main peak, and therefore are no longer resolved or detected. In these cases, a method for grading or classifying targets that are horizontally and / or vertically extended by radar can also be successfully applied by the use described here of phase information, so as to identify extended targets or distinguish them from actual point targets.

[0037] Reference numerals list

[0038] 10 Motor Vehicles

[0039] 12 Radar System

[0040] 14 Objects

[0041] 16 Antenna Device

[0042] 18 Signal processing device

[0043] 20 receiving channels

[0044] 22 Part (of subject 14)

[0045] 24 Radar Signal

[0046] 26 Interface

[0047] 28 Processors

[0048] 30 Data storage

Claims

1. A method for radar-based size classification of objects (14), in particular in a motor vehicle (10), wherein - detecting radar signals (24) reflected by the object in a plurality of receiving channels (20), - the detected reflected radar signal (24) is evaluated with respect to its phase curve on a plurality of the receiving channels (20), and If a phase jump is present in the phase curve, the corresponding object (14) is classified as an extended object (14).

2. The method according to claim 1, characterized in that A plurality of physical antennas or a virtual antenna array is used for a plurality of the receiving channels (20).

3. The method according to claim 2, characterized in that The phase position of each of the receiving channels (20) is determined using an I&Q method and / or a Hilbert transform.

4. The method according to any one of the preceding claims, characterized in that In order to detect phase jumps, it is checked, in particular by comparison with a predefined threshold value, whether the phase position increases uniformly for each radar detection and / or each Doppler class.

5. The method according to any one of the preceding claims, characterized in that A Fourier transformation is applied to the detected radar signal (24) in order to determine the detection angle, the width of the resulting spectrum is determined and the size of the corresponding object (14) is derived therefrom.

6. The method according to any one of the preceding claims, characterized in that When evaluating the detected reflected radar signal (24), a statistical evaluation is performed, wherein a predetermined statistical measure, in particular a standard deviation of the phase position on the receiving channel (20), is determined and used to detect phase jumps.

7. The method according to claim 6, characterized in that A value curve of a statistical measure for the respective object (14) is determined over a plurality of detected reflected radar signals (24) from a plurality of radar measurement cycles and is used for classification of the respective object (14).

8. The method according to any one of the preceding claims, characterized in that The detected reflected radar signal (24) and / or data derived from the radar signal are provided as input data to a machine learning device (18) trained for size classification of objects (14) based thereon, and the corresponding objects (14) are classified based on this by the machine learning device (18) with respect to their sizes.

9. A radar device (12, 18), in particular a radar device for a motor vehicle (10), having a signal processing device (18), wherein the radar device (12, 18) is designed to carry out the method according to one of the preceding claims.

10. A motor vehicle (10) having a radar device (12, 18) according to claim 9.

Images