Methods for detecting the movement of an object

Abstract

Method for detecting the movement of an object, in which the object is detected by means of a radar sensor by reflecting a radar signal emitted by the radar sensor from the object and subsequently receiving it back from the radar sensor, whereby the object is divided into at least one reference area (7) and at least one sub-area (6), a Doppler signature of the radar signal is created for the reference area (7) and the sub-area (6), the frequencies of the Doppler signatures are recorded and The movement of the object is detected based on a frequency deviation between the frequencies of the Doppler signatures, and for each sub-area (6) and reference area (7) at least one trajectory is created, wherein The type of movement of the object is determined based on the trajectories, Movement patterns for different movements of the object are stored in a memory, and by comparing the movement patterns and the determined movement, conclusions are drawn about the type of movement of the object, whereby the trajectory has a definable line thickness and the line thickness is determined based on the energy of the Doppler signature of the radar signal, the line thickness is used to detect the movement or type of movement of the object, and The movement of the object is used to predict the behavior of the object.

Description

[0001] The present invention relates to a method for detecting a movement and / or change in the movement of an object and to a sensor system for detecting a movement and / or change in the movement of an object, in which the object detection is carried out using the method according to the invention. Technological background

[0002] Modern vehicles such as cars and motorcycles are increasingly equipped with driver assistance systems (ADAS). These systems use sensors to perceive the environment, recognize traffic situations, and support the driver, for example, by applying the brakes or steering, or by issuing visual or audible warnings. Radar sensors, lidar sensors, cameras, and similar devices are commonly used for environmental perception. The sensor data collected by these devices allows for inferences to be drawn about the surroundings. Radar sensors are used for environmental perception by emitting focused electromagnetic waves and reflecting them back, for example, from other road users, obstacles on the road, or roadside structures.The detection of animals or pedestrians is often carried out using optical sensor systems such as camera sensors or lidar sensors. However, these sensor systems exhibit insufficient functionality, especially in poor visibility conditions or during weather events like snow, fog, or rain. As a result, the detection of an object's movement may be either impossible or unsatisfactory, meaning that sensor systems based on such optical sensors are only of limited use in poor visibility. Therefore, it is necessary to create redundant systems for animal and pedestrian detection.

[0003] Prior art methods exist for detecting and classifying objects, such as people or animals, using radar sensors. These methods utilize, among other things, micro-Doppler and distance information acquired by the radar sensor. Furthermore, Zhang et al. in "Dynamic Hand Gesture Classification Based on Radar Micro-Doppler Signatures" (2016 CIE International Conference on Radar, RADAR 2016) and Yuliang et al. in "Gesture Classification with Handcrafted Micro-Doppler Features using an FMCW Radar" (2018 IEEE MTT-S International Conference on Microwaves for Intelligent Mobility) propose using a radar sensor to capture hand gestures by evaluating micro-Doppler signatures.

[0004] In "Detection and Analysis of Human Motion by Radar" (in: IEEE Radar Conference, 1-4244-1539-X08, 2008), V.C. Chen discusses radar backscatter from a "walking human" model and derives motion trajectories and velocity patterns of human body parts based on this model. He also describes in detail how radar backscatter is analyzed and how the time-frequency transformation can be used to analyze micro-Doppler signatures of human motion. Furthermore, he uses an example to demonstrate how radar backscatters from complex human arm and leg movements and how the micro-Doppler signature of these complex movements can be characterized.

[0005] Furthermore, D. Tahmoush et al., in "Angle, Elevation, PRF, and Illumination on Radar Micro Doppler for Security Applications" (in: IEEE Antennas and Propagation Society Symposium 2009, Print ISBN: 978-1-4244-3647-7), already describe the approach of using radar to measure the direction, distance, and radial velocity of a walking person as a function of time. Detailed radar processing can then identify further characteristics of the walking person, as parts of the human body do not move at a constant radial velocity. The resulting small micro-Doppler signatures vary over time, allowing analysis techniques to be used to obtain a more detailed characterization. Printed state of the art

[0006] German patent DE 10 2007 054 507 A1 discloses a method for detecting a living being in which an object is detected using radar beams. Radar data on the detected object are collected and evaluated in such a way that at least one property of the object is classified. This property can then be compared with characteristic features of known living beings to determine the class of living being the object. Based on this determination, the living being can then be identified and classified, for example, as a human, animal, or the like. Furthermore, the method includes the evaluation and comparison of movement patterns. The movement of individual limbs can be extracted from the micro-Doppler signature. However, this type of extraction is only possible to a limited extent, since the radar signal or...Since the microdoppler signature of individual limbs can only be extracted to a limited extent using conventional evaluation methods, the detection of the organism's movement can only be inadequate.

[0007] Furthermore, DE 10 2016 215 102 A1 describes a method for pedestrian detection in which information about the size, speed, and distance of a pedestrian's body parts, obtained from a radar wave, is used to identify the pedestrian. For example, due to the different movement speeds of a pedestrian's extremities, radar reflection points from the arms and legs can generate a micro-Doppler effect in the radar signal, enabling the object to be classified as a pedestrian. The different distances between the radar sensor and at least the arms, legs, and torso of the pedestrian can be used to classify the object as a pedestrian. Object of the present invention

[0008] The present invention is therefore based on the objective of providing an improved method for detecting the movement of an object, in which the disadvantages of the prior art are overcome and in which motion detection is improved. Solution to the task

[0009] The aforementioned problem is solved by the entire teaching of claim 1 and the dependent claim. Advantageous embodiments of the invention are claimed in the dependent claims.

[0010] In the inventive method for detecting the movement of an object, the object is first detected using a radar sensor. A radar signal emitted by the radar sensor is transmitted as a primary signal, reflected by the illuminated object, and subsequently received by the radar sensor as a secondary signal. The object, or an object image generated by the radar data, is then divided into at least one reference area and at least one sub-area. A Doppler signature of the radar signal can be created for both the reference area and the sub-area, with the corresponding frequency being recorded for each Doppler signature. This serves to determine a frequency shift or frequency deviation between the frequencies of the Doppler signatures in order to detect the object's movement. By using a radar sensor for object detection, the method...When motion detection is used, reliable object detection can be achieved even under adverse weather conditions, when optical systems such as cameras and lidar sensors have limited functionality. This method thus creates redundancy for optical motion detection systems. Furthermore, by dividing the object or object image into a reference area and a sub-area, the object's movement can be extracted particularly effectively. This is because not only is the movement itself detected, but also the relative movements occurring in the sub-areas (e.g., limbs) in relation to a reference area (e.g., torso), which is typically less prone to movement. This results in the significant improvement of both motion detection and motion classification.Furthermore, the method is particularly easy to implement as an algorithm and can therefore be implemented or retrofitted cost-effectively into existing sensor or assistance systems. Additionally, the behavior of the object is predicted. Such a prediction can be easily made based on the object's movement. For example, a rapid change in gait (from slow steps to running / galloping) can indicate a startled or bolting animal. This can lead to dangerous situations, to which, for example, regulatory and / or warning assistance functions can then be used. The prediction can be based on a comparison of currently generated trajectories with stored reference patterns and the associated statistical models.

[0011] Preferably, the object is a living being with a torso and limbs, in particular an animal such as cattle, horses, wild animals, and the like. The areas are divided by, for example, assigning sub-areas to the limbs and the reference area to the torso. In a further development of the invention, by refining the methods presented here and with the availability of increased computing power, smaller animals such as dogs or wild boars can also be taken into account. Higher frequencies in the range of 100 GHz to 300 GHz can also be used for scanning.

[0012] According to a preferred embodiment of the invention, each of the sub-areas can be individually illuminated by the radar sensor. While this requires a high-performance radar sensor capable of, for example, beamforming, it significantly improves the reliability of motion detection.

[0013] Furthermore, at least one trajectory can be created for each sub-area and each reference area. This allows for the creation of a movement trajectory for each area (sub-area and reference area), for example, a trajectory for the movement of limbs or legs in the respective sub-areas and of the torso in the reference area. These trajectories make it particularly easy to relate the movements in the areas to one another, and especially to compare them graphically with existing reference patterns.

[0014] Advantageously, each sub-area can be divided into at least two, and in particular at least four, sections or zones, with at least one trajectory being created for each section. For example, a trajectory can be created for a sub-area for a left forward value, a left rear value, a right forward value, and a right rear value. The value used for each value could be, for example, the signal strength of the radar signal or the energy value of the Doppler signature. This would create, for example, a front left, a rear left, a front right, and a rear right section of the sub-area. Alternatively or additionally, further sections can be inserted between the left and right sections.

[0015] Furthermore, a radar signature of the torso can be generated, allowing for the detection of frequency deviations or shifts between the radar signature of the limbs and the radar signature of the torso. This enables the extraction of movements in individual areas in a particularly simple manner.

[0016] Preferably, for each sub-area, the minimum and maximum energy of the frequency deviation is determined, with frequency deviations at a higher energy level being assigned to limbs that are closer to the radar sensor, and frequency deviations at a lower energy level being assigned to limbs that are farther away from the radar sensor. For an animal that is measured or illuminated from the front or rear, for example, for a left zone of a sub-area, the anterior left value is the energy value of the micro-Doppler signature from the leftmost zone with maximum energy, and the posterior left value is the energy value of the micro-Doppler signature from the leftmost zone with minimum energy.Similarly, for a right-hand zone of the sub-area, the anterior right value is the energy value of the micro-Doppler signature with maximum energy from the right-hand zone, and the posterior right value is the energy value of the micro-Doppler signature with minimum energy from the right-hand zone. Here, the minimum energy values ​​correspond to the ejected or hind limbs (i.e., the legs furthest from the radar sensor), and the maximum energy values ​​correspond to the retracted or forelimbs (i.e., the legs closer to the radar sensor). This allows for the detection of particularly complex animal movements, such as whether the animal is performing fast or slow movements, or how the legs move relative to each other (e.g., alternately, in pairs front and / or rear, or asynchronously).

[0017] It is useful to determine or classify the object's mode of movement. This can then be easily determined based on the identified trajectories. For example, by characterizing the movement, one can infer a gait, such as the gaits of horses (trot, walk, tölt, canter, pace). Furthermore, more general statements can also be made, such as slow / fast movement, crawling, running, walking, and the like.

[0018] According to the invention, motion patterns for different movements of the object are stored in a memory. This allows a kind of motion database to be created, so that the respective sensor or sensor / assistance system is aware of many motion patterns. Furthermore, motion patterns can also be subsequently stored in the memory through updates or through movements "learned" in specific situations, thus making the motion database expandable.

[0019] By comparing the movement patterns with the detected movement, the type of movement of the object can be determined. In particular, for the purposes of the invention, in addition to running, walking, crawling, and the like, different gaits of animals are considered types of movement, such as trot, gallop, tölt, walk, pace, and the like. This offers the advantage that changes in gait of animals in traffic can be detected. For example, this allows the detection of a horse bolting, such as when the horse changes from a walk to a gallop.

[0020] Furthermore, movement patterns can also be generated or learned and stored in memory using a hidden Markov model or a neural network. This makes it possible to generate training scenarios that have not been experienced in real life. This type of machine learning (deep learning) can further enhance movement recognition, as the stored movement patterns are automatically augmented. This significantly improves movement recognition because, in particular, comparable or matching movement patterns are stored even for very rare scenarios.

[0021] The trajectory has a definable line thickness, which is determined by the energy of the radar signal's Doppler signature. For example, a high energy value corresponds to a strong or large line thickness, and a low energy value to a weak or small line thickness.

[0022] According to the invention, the line thickness is used to detect the movement or type of movement of the object. This allows, for example, a handwriting recognition method to be easily implemented in the process to extract or capture the movements of the object or an animal. If one considers, for example, the shifts or deviations in frequency of the microDoppler signal of leg movements compared to the frequency of the microDoppler signal of the torso, this is similar to the deviation of parts of a letter over a continuous baseline. Compared to cursive or Arabic script, the arcs or deviations of the trajectories created by the microDoppler signatures can also be compared analogously to the ligatures of the scripts. The graphically represented trajectories of the leg movements can, for example, be...The object's movement or type of movement can be determined using a method for character recognition in relation to the trajectory of the torso, so to speak the baseline.

[0023] The radar signal can also be amplified or filtered as needed. Alternatively or additionally, parameters such as amplitude, phase, polarization, travel time, angle of incidence, and the like can be used to determine the Doppler signature and / or to characterize the radar signal or Doppler signatures, in addition to the frequency.

[0024] Furthermore, the present invention comprises a sensor system for detecting the movement of an object, which includes a radar sensor and is characterized in that it performs the detection of the object's movement using the method according to the invention. Practically, the sensor system can be part of a (driver) assistance system that detects the object's movement and, depending on the movement, issues a warning to the driver, performs steering and / or braking interventions, or adjusts the speed and / or acceleration, particularly automatically.

[0025] The invention also expressly includes combinations of features or claims that are not explicitly mentioned, so-called sub-combinations. Description of the invention using exemplary embodiments

[0026] The invention will now be described in more detail using practical embodiments. The figures show: Fig. 1 Four simplified representations (AD) of traffic situations involving a vehicle and an animal; Fig. 2 a simplified schematic representation of a division of the front or back view of an animal into zones using a grid; Fig. 3 a simplified schematic representation of a trajectory for leg movement with support points and a comparison trajectory for the trunk, as well as suggestions for a finer subdivision into several sections to obtain additional support points for additional trajectories; Fig. 4 a simplified schematic representation of a trajectory for leg movement with support points and a comparison trajectory for the torso; Fig. 5 a simplified schematic representation of eight trajectories for two areas, each with the microdoppler signal from the trunk and leg of an animal, as well as Fig. 6 a simplified schematic representation of trajectory segments, the line thickness of which varies depending on the signal strength.

[0027] Reference number 1 in Fig. 1 denotes a vehicle moving along a road in the direction of the arrow (light arrow). Reference numeral 2 denotes an animal also moving in the direction of the arrow (dark arrow). The individual representations AD in Fig. Figure 1 shows different scenarios or traffic situations: In Fig. 1(A) Vehicle 1 and Tier 2 are moving in the same lane in the same direction, in Fig. 1(B) Vehicle 1 and Tier 2 are moving in opposite directions on the same lane, in Fig. 1(C) Vehicle 1 and Tier 2 are moving in opposite directions on different lanes and in Fig. In scenario 1(D), vehicle 1 and Tier 2 are moving in the same direction on different lanes. Therefore, from vehicle 1, either the front or the rear of Tier 2 is visible. Vehicle 1 is equipped with a radar sensor or sensor system for object detection, which uses emitted electromagnetic waves (radar signals) to detect and / or locate objects. This is achieved by the radar sensor emitting a radar signal and receiving and analyzing the signal reflected by the respective object.

[0028] According to the present invention, significant changes in the animal's locomotion are observed, detected by monitoring the movement of the animal's limbs or legs. These leg movements are considered, in the spirit of the invention, to be "gestures" of the respective animal. The evaluation is performed using a Doppler signature or micro-Doppler signature of the radar signal or Doppler signal. The relative movement between the radar sensor and the respective limb can be determined or calculated from the Doppler effect by measuring the shift in the frequency of the reflected radar signal. Since the animal's locomotion typically occurs through leg movement, it is proposed to use scans of individual zones or areas of the animal's view, focusing primarily on the limbs or legs, rather than a complete scan of the animal.

[0029] For example, the front or back of the animal can be divided into zones or areas that are separately recorded when acquiring micro-Doppler signatures, as exemplified in Fig. Figure 2 illustrates this. For example, the front of the animal, comprising the torso 3 and legs or limbs 4, i.e., the side facing the vehicle 1, is covered with a grid 5. The grid creates areas 6, which can be considered input fields (zones as input fields, so to speak). For example, sub-areas 6 can be defined for the legs and a reference area 7 for the torso. Such a sub-area 6 should ideally be shaped like a strip to conform to the (usually elongated) geometry of the legs. To process these individual areas, a powerful radar is required, with which, for example, the beamforming method can be applied to each individual area. This means the radar must be able to illuminate, for example, only one sub-area 6 at a time and then successively apply this to all areas without significant changeover times.By individually illuminating the sub-areas 6, the frontal or rear view of the animal can be divided into strip-shaped sub-areas 6, allowing the Doppler signatures to be evaluated separately. These results are then combined again during later analysis, e.g., using a hidden Markov model or a neural network.

[0030] Furthermore, the radar signature of the animal's torso is measured to compare it, for example, with the radar signature of the limbs. Additional measurements are taken to generate the micro-Doppler signature of the torso, as not all zones or parts of the animal's torso may be included. This allows the frequency deviations of the micro-Doppler signature generated at the moving legs to be determined in relation to the micro-Doppler signature of the torso.

[0031] In the use cases from Fig. From vehicle 1, one sees either the frontal or rear view of the animal. This view is divided into areas or zones according to Fig. The area is divided into two sections, with a multiple radar micro-Doppler signature being created for each section. Furthermore, each section is individually illuminated by beamforming of the radar sensor, specifically to allow the reflected signals to be assigned to a particular area of ​​the illuminated object. This results in a larger area of ​​reflected signals than the original area, as shown in Fig. Figure 3 shows the original area as indicated by the dashed line. To select the measured values, i.e., the frequencies shifted due to the microdoppler effect and their corresponding energies, the values ​​with the minimum and maximum energies (i.e., the greatest frequency deviation and frequency shift) are sought among the shifted frequencies, for each of the left and right halves of the reflections belonging to a sub-area 6, as shown in Figure 3. Fig. 3 shown on the left. In a practical way, the entire front and back view of the animal is scanned or measured by successive scans of the individual sub-areas 6.

[0032] For each section of sub-area 6, the microDoppler frequency shift at high and low energy levels is considered. The high-energy microDoppler frequency shift is attributed to the animal's legs on the side facing the sensor. Conversely, the low-energy microDoppler frequency shift is attributed to the animal's legs on the opposite side. Furthermore, the reflections from the trunk, which typically appear as a high-energy Doppler signal, are also taken into account. This value can be measured separately if no trunk components are present in sub-area 6, as the microDoppler frequency deviation relative to the trunk's Doppler frequency must be determined. These energy differences determine whether the received reflection frequency is attributed to the animal's trunk or its legs / limbs.MicroDoppler frequency shift measurements at a higher energy level (i.e., still below the energy level for the torso) indicate reflections from the legs on the side of the animal facing the vehicle (e.g., the front legs in a frontal view). Conversely, microDoppler frequency shift measurements at a lower energy level indicate reflections from the legs on the side of the animal facing away from the vehicle (e.g., the hind legs in a frontal view).

[0033] Trajectories are expediently generated from the reflected micro-Doppler signals or micro-Doppler signatures. Trajectory generation can be performed in a similar manner to handwriting recognition methods; that is, the handwriting recognition methodology or procedure can be adapted for trajectory planning. For example, so-called contact points or anchor points can be detected, which serve as anchor points for the trajectories yet to be created. Subsequently, the trajectories generated using these anchor points are recorded and output.

[0034] As in Fig. Figure 3 (left) shows that when sub-area 6 is halved into two sections 8a and 8b, the values ​​for the micro-Doppler signals with the minimum and maximum energy are first determined for the left and right halves of each section 8a and 8b, respectively. The front left value in section 8a is the value belonging to the micro-Doppler signals, lies in the left half of section 8a, and has the maximum energy value. Conversely, the front right value in section 8a is the value belonging to the micro-Doppler signals, lies in the right half of section 8a, and has the highest energy value. The front left and front right values ​​can be identical, for example, if both of the animal's front legs are equidistant from the radar sensor.The rear left value in section 8a is the value belonging to the microDoppler signals, located in the left half of section 8a, and exhibiting the lowest energy value. Conversely, the rear right value in section 8a is the value belonging to the microDoppler signals, located in the right half of section 8a, and exhibiting the lowest energy value. The rear left and rear right values ​​can also be identical. This is repeated in the same manner for section 8b. Furthermore, each subsection 6 can also be defined according to... Fig. 3 (middle) and Fig. 3 (right) may be divided into several sections 8a-8h.

[0035] Subsequently, four trajectories can be generated for each section of a sub-area 6. One trajectory is created using the front left values ​​for all measured time points as support points. Another trajectory is created using the front right values ​​for all measured time points as support points. Finally, a trajectory is created using the back left values ​​for all measured time points as support points, and a third trajectory is created using the back right values ​​for all measured time points as support points. For example, in Fig. 4. A trajectory with measured values ​​as support points is shown, where the frequency is plotted against time and the support points were recorded at times t1, t2, t3 ... tn. Furthermore, in Fig. Figure 5 shows the trajectories T1-T8 for two sections 8a and 8b of a sub-area, each with the micro-Doppler signal of the trunk (dashed line) and that of the respective leg (solid line). Here, T1 for section 8a is the anterior left value, T2 for section 8a the anterior right value, T3 for section 8a the posterior left value, T4 for section 8a the posterior right value, T5 for section 8b the anterior left value, T6 for section 8b the anterior right value, T7 for section 8b the posterior left value, and T8 for section 8b the posterior right value.

[0036] The line thickness or line strength of the respective trajectory can correspond to the energy of the received measurements; that is, the line strengths of the trajectories correlate with the energy of the received measurements or the signal strength. In particular, such a line can also disappear if the energy of a reflected microdoppler signal falls below a definable lower threshold. This can occur, for example, with reflections from the animal's far-facing legs, as in Fig. Figure 6 illustrates this. From left to right, it shows by way of example that a lot of energy from the reflected signal results in a high line thickness (left), little energy (middle left and right) causes a normal line thickness, and very little energy (middle right) leads to the disappearance of the line.

[0037] For each segment, trajectories are created for each leg or limb of the animal. These trajectories are derived from the differences between the micro-Doppler frequencies reflected by the legs and the Doppler frequency reflected by the torso, resulting in four trajectories per segment. It is possible for the same value to be assigned to two legs at once, as is the case with a horse galloping, since the front and hind legs move forward and backward almost simultaneously compared to a trot or tölt. The values ​​for the side facing the vehicle can be identified due to the higher energy of the reflected radar signals.

[0038] To recognize specific movement transitions in animals, tools such as Hidden Markov Models or neural networks are used. One important movement transition to be recognized is, for example, the transition from walking to galloping. This transition can be identified based on the provided trajectories. The currently generated trajectories are normalized and then compared with the stored patterns. Stored statistical data can also be used to determine the degree of similarity with the stored patterns. Unlike handwriting recognition of individual letters, handwriting recognition of cursive or Arabic script always involves analyzing a group of letters, i.e., a word. Based on multiple matches, the word is recognized after statistical evaluation. This approach is extended here to include the parallel evaluation of all trajectories.It is particularly important to consider that only the entirety of all sub-areas or sections covers the front or back of an animal, and it is impossible to predict in advance which sub-area will capture which part of the animal or leg. These shifts must be taken into account when comparing the generated trajectories with the stored patterns.

[0039] Hidden Markov models or neural networks can be trained to capture specific movement transitions using simulation. With the aid of established simulation techniques, appropriate patterns for capturing these transitions can be created. The animals are varied in size and speed. Furthermore, statistical models can be developed that must be considered whenever comparing live data with pattern data.

[0040] The trajectories are to be treated as in multiple, parallel handwriting recognition of cursive writing. The trajectories for the number of segments and the number of legs are evaluated in parallel, with each trajectory describing the shift of the micro-Doppler signal of the legs relative to the Doppler signal of the trunk in the direction of time. The Doppler signal of the trunk (the animal is in motion) forms the "zero line." This zero line is regularly crossed by the animals' legs while walking or galloping. Furthermore, the line thickness can be determined according to... Fig. 6, to be taken into account.

[0041] Unlike handwriting recognition, the described method can also be used to monitor an animal's locomotion by detecting periodic leg movements. This allows for the monitoring of changes in the animal's gait. The continuous comparison of the trajectories generated from the measurements with the stored patterns reveals whether an animal, for example, transitions from a walk to a gallop or shies. This makes it possible to predict the animal's behavior.

[0042] According to a preferred embodiment, the radar frequency used should be approximately 80 GHz. The trajectory measurement points should be at a rate of 20 Hz. Furthermore, the view of the animal's front or back can be divided into 50 sections, so that the radar scan of each section should be completed within 1 ms, resulting in a total scan rate of 100 Hz. The distance from the vehicle to the animal would ideally be up to approximately 20 m. The detection accuracy and the vehicle's distance from the animal can be improved, for example, by reducing the size of the zones (i.e., sub-areas and / or sections) and increasing their number. In particular, this can be achieved by using a radar sensor that can perform more precise beamforming on the smaller zones and also achieve a faster sequence of scans of the individual zones.For example, such beamforming can be achieved at least by a radar constructed according to the AESA (Active Electronic Scanned Array) principle. In a simpler way, the radar can comprise an antenna array with suitably aligned antennas, in which the antennas are used alternately. Preferably, however, to improve detection accuracy, the scan rate is not changed, but rather the division of the segments is modified, as shown in [reference]. Fig. 3 described, refined to create additional support points and thus additional trajectories.

[0043] The movement and mode of movement of an animal, such as a horse, can be reliably detected using the method according to the invention. Additionally, outputs can be provided via the car's display devices, such as a screen or head-up display, an audio system, an infotainment head unit, and / or an instrument cluster display (e.g., visual, audible, or haptic), indicating, for example, that the animal or horse currently poses no danger. If a situation occurs that the driver cannot see, such as the horse becoming startled and shying, this change in the horse's condition is reclassified from "safe for traffic" to "problematic," and, for example, the color of an icon symbolizing the horse in the display changes from green / yellow to red. The vehicle is then preferably brought to a stop or swerve, either partially or fully automatically. Furthermore, a warning message regarding the runaway horse can be displayed on the head unit or instrument cluster screen.The invention expressly includes solutions for other means of transport, such as airplanes, helicopters, drones, motorcycles, bicycles, boats, watercraft and the like. REFERENCE MARK LIST 1 vehicle 2 Tier 3 Hull 4-legged 5 grids 6 sub-area 7 Reference range Sections 8a-8h T1 Trajectory Section 8a, front left value T2 Trajectory Section 8a, front right value T3 Trajectory Section 8a, rear left value T4 Trajectory Section 8a, rear right value T5 Trajectory Section 8b, front left value T6 Trajectory Section 8b, front right value T7 Trajectory Section 8b, rear left value T8 Trajectory Section 8b, rear right value

Claims

[1] Method for detecting the movement of an object, in which the object is detected by means of a radar sensor by reflecting a radar signal emitted by the radar sensor from the object and subsequently receiving it back from the radar sensor, whereby the object is divided into at least one reference area (7) and at least one sub-area (6), a Doppler signature of the radar signal is created for the reference area (7) and the sub-area (6), the frequencies of the Doppler signatures are recorded and The movement of the object is detected based on a frequency deviation between the frequencies of the Doppler signatures, and for each sub-area (6) and reference area (7) at least one trajectory is created, wherein The type of movement of the object is determined based on the trajectories, Movement patterns for different movements of the object are stored in a memory, and by comparing the movement patterns and the determined movement, conclusions are drawn about the type of movement of the object, whereby the trajectory has a definable line thickness and the line thickness is determined based on the energy of the Doppler signature of the radar signal, the line thickness is used to detect the movement or type of movement of the object, and The movement of the object is used to predict the behavior of the object. [2] Method according to claim 1, characterized by , that the object is a living being with a torso (3) and limbs, in particular an animal, and the division of the areas is done by assigning sub-areas (6) to the limbs and the reference area (7) to the torso. [3] Method according to claim 1 or 2, characterized by, that each sub-area (6) is illuminated individually by the radar sensor. [4] Method according to any one of the preceding claims, characterized by , that each sub-area (6) is divided into several sections (8a-8h), in particular into two or four sections, with at least one trajectory (T1-T4 or T5-T8) being created for each section (8a-8h). [5] Method according to any one of the preceding claims, characterized by , that a radar signature of the torso (3) is created, so that a frequency deviation of the radar signature of the limbs compared to the radar signature of the torso (3) can be determined. [6] Method according to any one of the preceding claims, characterized by, that for each section (8a-8h) of the sub-area (6) the minimum and maximum energy of the frequency deviation is determined, whereby frequency deviations at a higher energy level are assigned to limbs that are closer to the radar sensor, and frequency deviations at a lower energy level are assigned to limbs that are farther away from the radar sensor. [7] Method according to any one of the preceding claims, characterized by that the movement patterns are evaluated by a hidden Markov model or a neural network. [8] Sensor system for detecting the movement of an object, with a radar sensor (1), characterized by , that the sensor system performs the detection of the object's movement using a method according to one of the preceding claims.

Images