Device for estimating the pose of a magnetic-field-measuring object

Abstract

The invention relates to a device for magnetic pose estimation of an object in a measurement volume to which a reference magnetic field is applied, comprising: at least four electrically activatable magnetic-field generators having various predetermined poses in the edge region of the measurement volume; at least one magnetic-field sensor rigidly connected to the object; a control unit for predeterminable electrical activation of the magnetic-field generators; and an evaluation unit for receiving and evaluating the electrical output signal from the at least one magnetic-field sensor; a. the at least one magnetic-field sensor being in the form of a magnetoelectric bending-beam sensor having a known mechanical resonant frequency for simultaneously measuring three linearly independent components of the reference magnetic field; and b. the control unit being designed to perform the electrical activation of the magnetic-field generators in a band-limited manner close to the mechanical resonant frequency of the magnetic field sensor; and c. the evaluation unit being designed to estimate from the output signal from the at least one magnetic-field sensor, using predetermined and tabulated matched filters, the pose of the object in the measurement volume; d. at least four of the magnetic-field generators being in the form of magnetoelectric transmitters.

Description

[0001] DEVICE FOR POSE ESTIMATION OF A MAGNETIC FIELD MEASURING OBJECT

[0002] The invention relates to a device for magnetically locating an object in a measurement volume subjected to a reference magnetic field. Both the position of the object—using three spatial coordinates—and the orientation of the object—using three angles, e.g., Euler angles—are estimated from magnetic field measurements on the object. The aforementioned six degrees of freedom of the object are collectively referred to below as the pose of the object. The invention thus relates to a device for estimating the pose of an object measuring a magnetic field.

[0003] The object can be any tool or aid used by a surgeon to perform an activity. The tool or aid is generally moved freely within a workspace during this activity. The workspace, in turn, lies within the measurement volume in which magnetic pose estimation can be performed. In addition to its primary function—e.g., as a measuring device for any quantity—the object should also be capable of magnetic field measurement by rigidly attaching at least one magnetic field sensor to the object. The magnetic field sensor should be designed to simultaneously measure three linearly independent magnetic field components.

[0004] Through repetitive magnetic field measurements and the resulting repetitive pose determination of the object during the operator's activity, the six-dimensional trajectory of the object can be fully automatically recorded as a function of time and electronically recorded.

[0005] Magnetic pose estimation is particularly suitable for use in medical diagnosis and therapy. The living patient body is virtually transparent to magnetic fields, so the object can be tracked just as easily when inserted inside the patient's body as it can outside. Known alternative tracking methods based on camera image analysis cannot achieve this and are also susceptible to other image occlusions, such as those caused by the surgeon. Furthermore, many patients are generally uncomfortable with cameras during a medical examination.

[0006] Magnetic positioning for medical workspaces is the subject of, for example, the publications US 2006 / 025668 A1 and US 2011 / 224537 A1.

[0007] In US 2011 / 224537 A1, precise knowledge and control of the reference magnetic field plays an important role in pose estimation. There, the reference magnetic field is generated via a large number of generator coils and is thus always predetermined according to the Biot-Savart law. The sensors are either coils or other known field sensors such as flux gates or SQUIDs. Both the shape of the field, or rather, the geometric field profile, as well as the temporal behavior of the field are determined by the current supply to the generator coils. This allows the magnetic field measured on the object to be assigned to the generator coils with pre-known poses, thereby inferring the object's pose.

[0008] The key concept of magnetic pose estimation is the mapping of the components of the magnetic field sensor's electrical output signal to the predetermined electrical input signals of the predeterminably positioned reference magnetic field generators. Determining signal components from a detected sum signal is a classic inverse problem that is generally well solvable for a completely linear transmission system.

[0009] However, in practical implementation, some technical difficulties must be taken into account:

[0010] - If the reference magnetic field is to extend over a working space dimensioned to accommodate a human patient, then a volume of approximately 2 m x 1 m x 0.5 m = 1 m 3 as a possible measurement volume.

[0011] - If one does not want to supply the generator coils for a measuring volume of the order of cubic meters at prohibitively high prices, then one must be able to manage with a reference magnetic flux density of the order of nanotesla or less in the working space.

[0012] Although correspondingly sensitive magnetic field sensors are known, they must also be miniaturizable - preferably integrated into the object - and low-maintenance, cost-effective, and shielded against the much stronger Earth's magnetic field.

[0013] The latter requirements are met by magnetoelectric bending beam sensors, or ME sensors for short. Such sensors generally comprise at least two mechanically rigidly coupled, usually directly connected material phases, namely one made of a magnetostrictive (MS) material and one made of a piezoelectric (PE) material. In the simplest case, the material phases are arranged as layers on a common substrate (e.g. wafer), for example directly contacted with one another or one on the front and back of the substrate. For ME sensors, the substrates are usually strip- or beam-shaped, or they are singulated into these shapes. The PE material layer is electrically contacted and usually poled by means of electrodes along the longitudinal axis of the beam. The MS material layer is pre-magnetized along the same longitudinal axis in such a way that the largest possible piezomagnetic coefficient is created, i.e.This means that when a magnetic field component acts on the MS material layer along this longitudinal axis, it results in the largest possible change in length of the MS layer. This change in length is transferred to the polarized PE material layer through the mechanical coupling and, through the PE effect, generates an electrical voltage that can be tapped via the electrodes.

[0014] A temporally constant magnetic field can thus cause the ME sensor beam to bend, but not a persistent electrical output signal. ME sensors only respond to time-varying magnetic fields with an electrical voltage signal during a repetitive movement, thus acting as oscillators. They are intrinsically insensitive to, among other things, the Earth's magnetic field. However, they are particularly sensitive to magnetic fields with a frequency close to the mechanical resonance frequency of the oscillating bending beam. When excited at resonance, the oscillation amplitude and the electrical output signal of the ME sensor are increased by several orders of magnitude, so that even very small magnetic field amplitudes on the order of nanotesla generate clearly measurable voltage signals. The mechanical resonance frequency of ME bending beams depends primarily on their size and is largely freely adjustable, e.g.between 10 Hz and 1 MHz, although miniaturized ME sensors with beam lengths in the millimeter range often exhibit resonances in the low to mid-kilohertz range. Such mini ME sensors are easily integrated into any object as passive components, and their production—for example, using established thin-film technology—is generally simple and cost-effective. It is also known to arrange at least three bending beams along at least three linearly independent (e.g., orthogonal) spatial axes and to simultaneously record three output signals to determine the complete magnetic field.

[0015] Due to their operating principle, ME sensors do not measure linearly. Instead, they generate voltage signals when simultaneously excited by magnetic fields with slightly different frequencies. These signals contain mixed frequency contributions in addition to the direct responses. This phenomenon can even be advantageously exploited for frequency conversion, for example, to determine magnetic field components with frequencies around a few Hz using an ME cantilever with a resonant frequency of a few hundred Hz, as described in WO 2012 / 097796 A2. In magnetic detection, the nonlinear response behavior is less desirable because it certainly does not simplify the assignment of the measurement signals to different magnetic field generators.

[0016] The publication Bald, C.; Schmidt, G., "Processing Chain for Localization of Magnetoelectric Sensors in Real Time", Sensors 2021, 21, 5675. https: / / doi.org / 10.3390 / s21165675, uses ME sensors and the concept of their resonant excitation to demonstrate the feasibility of real-time localization. In this paper, a flat workspace is surrounded by six coils as reference magnetic field generators, with the coils being energized with band-limited input signals close to the ME resonance frequency at approximately 7.7 kHz. However, the specific example serves to illustrate a universal concept: To compensate for different conversion factors (ratio of magnetic flux density to current strength) of the individual generator coils, a one-time adjustment of the current levels can be performed before the localization measurement ("equalizing").The input signals are arranged orthogonally, so that the magnetic field detected by the sensor can be decomposed into orthogonally generated field components and traced back to the individual generators. The amplitudes of the field components are determined and calculated using the known 1 / r. 3 The decay of the magnetic dipole field is converted into distance estimates to the generator coils; from the ratios of the magnetic field components to each other, the orientation of the magnetic field sensor to each of the generator coils can then, in principle, be deduced.

[0017] Pose estimation can now be made real-time by iterating for the best hypothesis to explain the magnetic field measurements at a given time. To efficiently estimate a position-orientation pair (here: a pose), the workspace is discretized into cells. The cell that represents the most probable estimate for the current sensor location can be further discretized into subcells in subsequent iteration steps and re-evaluated until a hypothesis is obtained that meets a predetermined accuracy criterion. This final hypothesis is then identified with the estimated pose of the magnetic field sensor at the measurement time.

[0018] The hypotheses to be tested are linked to the cells and subcells – in this context, also called indexed – and stored electronically in an evaluation unit in the form of signal-matched filters. The filters are time-inverted measurement signals from a sensor, averaged over several measurement cycles, which arise from the specification of predetermined, time-limited electrical input signals to the individual magnetic field generators. They can be designed as matched filters, which maximize the signal-to-noise ratio for the recognition of noisy input signals. Such matched filters are referred to below as optimized "matched filters" or, for short, as "matched filters" or simply as MF.Indexing the matched filters with cells and subcells means that a sensor detects the field of a single magnetic field generator with knowledge of a previously known sensor-magnetic field generator pose pair, and then a correspondingly indexed matched filter is formed from the measurement signal. The MFs can be viewed as components of hypothesis vectors, where the vector length is the number of sensors multiplied by the number of field generators. A large number of MF hypothesis vectors can be provided during a one-time calibration procedure prior to pose estimation.

[0019] During pose estimation, a numerical comparison of the currently measured output signal vector with a selection of MF hypothesis vectors is then performed by calculating a scalar product. The scalar product can also be referred to as the overlap of the output signal vector with the predetermined MF hypothesis vectors. Orthogonal input signals at the magnetic field generators cause corresponding orthogonal components of the reference magnetic field and lead to predominantly vanishing scalar products, except for those indicated MF hypothesis vectors that overlap with the measured output signal vector. A large overlap with an MF hypothesis vector signals a hit or agreement, i.e., the corresponding hypothesis for the pose estimation is confirmed. For further precision, the above-mentioned subcell iteration is then used, i.e.,the pre-stored MF hypothesis vectors are hierarchically organized, and not all MF hypothesis vectors are used in each measurement cycle.

[0020] It should be noted that Bald and Schmidt universally propose the use of orthogonal input signals at the magnetic field generators and matched filters for analyzing the magnetic field sensor output signal. The matched filters serve as a pulse compression method to improve the signal-to-noise ratio. Furthermore, the MFs are linked to cells and subcells, allowing an iterative search of the magnetic field sensor in the workspace in real time, i.e., within a time period limited by the length of the measurement cycle and potentially several times per second.

[0021] In the specific example of the cited work, the magnetic field is measured by an ME sensor with very limited band sensitivity. The authors therefore opt for time-division multiplexing to orthogonalize the input signals, meaning that only one of the magnetic field generators is active at any given time. This also prevents the non-linear ME sensor from mixing magnetic fields from different generator coils.

[0022] In the publication by C. Bald et al., "Automatic Localization of an Ultrasound Probe with the Help of Magnetic Sensors", Current Directions in Biomedical Engineering 2022;8(2): 317-320, https: / / doi.org / 10.1515 / cdbme-2022-1081, the application of magnetic pose estimation to improve ultrasound imaging with a movable probe is presented. The authors of the paper use a flux gate sensor to measure the reference magnetic field and frequency division multiplexing to orthogonalize the input signals of the generator coils. Otherwise, the real-time process of the aforementioned paper by Bald and Schmidt is also implemented there.

[0023] In the search for an energy-efficient generation of time-varying magnetic fields for the purpose of signal transmission, one can turn, for example, to the work of Junran Xu et al., "A Low Frequency Mechanical Transmitter Based on Magnetoelectric Heterostructures Operated at Their Resonance Frequency," Sensors 2019, 19, 853; doi:10.3390 / s19040853. This work presents magnetoelectric transmitters, or ME transmitters, that operate in a sense inversely to an ME sensor. ME transmitters also have an MS and a PE material phase in a mechanically rigid coupling, whereby the PE material must be electrically contacted and poled. The material phases can also be multilayered; for example, two MS layers can be arranged on top of one PE layer ("trilayer sandwich"), and these can also be laminated to form stacks.By applying a predeterminable input voltage signal, the length of the PE material is varied in a targeted manner, with the length change being transferred to the MS material. The MS material can exhibit a random magnetic order in a ground state—e.g., without an input signal—and be converted into a state with a high degree of magnetization by the enforced length change. Accordingly, the externally measurable magnetic field, which is generated here by the material-inherent order of magnetic domains and not by the Biot-Savart law, varies between approximately zero and a structure-dependent maximum value in time with the input signal. Junran Xu et al.In their work, they found that the measured magnetic flux density of a resonantly excited M E transmitter behaves qualitatively roughly like the dipole field of a powered conductor loop with increasing distance from the transmitter, but is up to two orders of magnitude larger than a purely electrically generated magnetic field of a conductor loop at the same signal input power. Specifically, the flux density of their experimental transmitter drops to approximately 1 nT at a distance of 1 meter, 1 pT at a distance of 10 meters, and 1 fl at (extrapolated) a distance of 200 m (see Fig. 7 of the source). This makes M E transmitters candidates for transmitting antennas in very low frequency (VLF) communication, where limited ranges and the ability to transmit through electromagnetically absorbing media such as seawater are important characteristics.

[0024] Despite their cost-effective design and energy efficiency, M E transmitters are not yet used for location purposes because

[0025] • M E-transmitters operate non-linearly and generate magnetic fields whose timing does not exactly match that of the input signals;

[0026] • the magnetic field amplitudes are built in during production and cannot be adjusted at will afterwards;

[0027] • Even nominally identical sensors show small differences in resonance, amplitudes and non-linear signal distortion.

[0028] This eliminates the need for precise knowledge of the reference magnetic field, which is usually required for magnetic positioning or pose estimation, and its easy adaptability, e.g. by equalizing.

[0029] Against this background, the invention has for its object to propose a device for pose estimation of an object measuring a magnetic field, which device has advantages over the prior art. The object is achieved by a device for magnetic pose estimation of an object in a measuring volume subjected to a reference magnetic field, comprising at least four electrically controllable magnetic field generators with different predetermined poses in the edge region of the measuring volume, at least one magnetic field sensor in rigid connection with the object, a control unit for the predeterminable electrical control of the magnetic field generators and an evaluation unit designed to accept and evaluate the electrical output signal of the at least one magnetic field sensor, characterized in that a.the at least one magnetic field sensor is designed as a magnetoelectric bending beam sensor with a previously known mechanical resonance frequency for the simultaneous measurement of three linearly independent components of the reference magnetic field and b. the control unit is designed to carry out the electrical control of the magnetic field generators in a band-limited manner close to the mechanical resonance frequency of the magnetic field sensor and c. the evaluation unit is designed to estimate the pose of the object in the measurement volume from the output signal of the at least one magnetic field sensor using predetermined and tabulated matched filters, wherein d. at least four of the magnetic field generators are designed as magnetoelectric transmitters.

[0030] The subclaims specify advantageous embodiments of the device.

[0031] The main objective of the invention is the energetically more efficient generation of the reference magnetic field, which also enables the application of magnetic pose estimation in larger measurement volumes and workspaces than previously technically feasible. A key component is the tuning of the at least four M E transmitters and the at least one M E sensor to the same resonant frequency, which must be taken into account during the manufacturing of these components.

[0032] The essential idea of ​​the invention is the recognition that the above-mentioned disadvantages of M E transmitters for magnetic location are based on the prejudice of the expert that he must have precise knowledge of the reference magnetic field in order to

[0033] It is believed that magnetic field measurements can be used to determine sensor poses. However, this is not the case at all.

[0034] The aforementioned work by Bald and Schmidt also assumes the necessity of providing a "well-defined" reference magnetic field, as can be seen, for example, in the equalizing process steps. It fails to recognize that the use of matched filters indexed by cells and subcells already offers the opportunity to address systemically present, not precisely quantifiable "quality deficiencies" in the reference magnetic field.

[0035] In fact, the method proposed by Bald and Schmidt for real-time pose estimation by dividing the workspace into cells and subcells requires a considerable calibration effort to predetermine the indexed MFs, which the authors do not yet fully exploit in their work. During the generation of the MFs, any "deficiencies" in the reference magnetic field can be captured without requiring explicit identification of such deficiencies. The MFs serve solely to compare previously measured (averaged) signals with current measurement signals from the ME sensor; however, how these are generated is no longer important if the pose estimation is performed via the predetermined indexing of the MFs anyway.

[0036] A particularly advantageous embodiment of the device according to the invention is considered to be that at least two magnetic field sensors are rigidly connected at different locations on the object, with the evaluation unit being designed to simultaneously receive and process the output signals of the magnetic field sensors. This is technically easy to accomplish if the object is typically a handheld tool several centimeters in size and the mini ME sensors are only a few millimeters in size, preferably by integration into the housing of the tool. Many tools already have supply lines for power supply and / or data transfer, so that the signal outputs of the ME sensors can also be arranged therein.

[0037] With two ME sensors and at least four ME transmitters, the pose of the object can be estimated with quite high accuracy. In such an embodiment, when applying the method of Bald and Schmidt, the two sensors are preferably already installed in the object during the calibration procedure, i.e., the MF hypothesis vectors and the current output signal vector each comprise at least eight components; a large overlap means a high degree of agreement in all components and thus also a strong confirmation of a hypothesis for pose estimation. The fixed relative pose of the two sensors to each other is thus preferably used implicitly. Its explicit use in the numerical evaluation of the output signal vectors can also be considered.In particular, the evaluation unit can be designed to use the previously known relative pose of the at least two magnetic field sensors to each other to refine the pose estimation of the object.

[0038] As already proposed by Bald and Schmidt, it is considered particularly advantageous for the control unit to be designed to control the at least four magnetoelectric transmitters in a time-staggered manner so that they only vary the reference magnetic field individually. This corresponds to the orthogonalization of the input signals to the magnetic field generators by means of time-division multiplexing and still results in orthogonal components of the reference magnetic field relative to the individual M E transmitters. This simplifies the assignment of the output signal to the sources.

[0039] The device according to the invention preferably comprises a pre-calibration such that the evaluation unit subsequently comprises non-volatilely stored, tabulated matched filters as components of hypothesis vectors, wherein the individual matched filter hypothesis vectors are indexed to different predetermined poses of a magnetic field sensor. A conventional programmable microprocessor can be used as the evaluation unit.

[0040] This calibration proceeds in the same way as explained by Bald and Schmidt, although this time a well-defined reference magnetic field cannot be assumed. Rather, situations may increasingly arise in which, when evaluating the sensor signal during pose estimation, several indexed MF hypothesis vectors show a non-vanishingly small overlap with the sensor signal vector. If this occurs, then a preferred embodiment is for the evaluation unit to be designed to select the overlaps according to their size and to estimate a pose of the magnetic field sensor from a measured output signal by interpolating the indexes of the matched filter hypothesis vectors with the greatest overlap with the output signal vector. In these cases in particular, the explicit use of the relative pose of an ME sensor pair can be very helpful.

[0041] To ensure the real-time capability of pose estimation, a hierarchical discretization of the measurement volume into cells and subcells should also be carried out, analogous to Bald and Schmidt, which enables an iterative search. For this purpose, it is preferred that the evaluation unit is designed to select the matched filters to be applied for the subsequent pose estimation based on the previous pose estimation. The selection should, for example, take into account the hierarchical order of the cells and the subcells defined therein in order to refine the analysis of the magnetic measurement signal. Another selection can be aimed at analyzing movement patterns in the guidance of the object by the user and using this to deduce the probable current location of the object, i.e., the cell. This can be useful for accelerating pose estimation in the case of uniform or slow movements of the object.

[0042] It is possible and advantageous if the evaluation unit is designed to perform the pose estimation of the object at least at 5 Hz, preferably at least at 10 Hz, particularly preferably at least at 20 Hz. This means that a measurement cycle ending with the output of a pose estimation requires no more than 200 ms, preferably less than 100 ms, particularly preferably less than 50 ms. The pose estimation can then be repeated as frequently as required in order to record and log 6D trajectories of the object moving in the workspace over time intervals selectable by the user. A time index of an internal clock of the evaluation unit can preferably be added to the log, so that an assignment to other measurement data with a time index - in particular recorded with the object - is possible in post-processing.

[0043] A particularly advantageous embodiment of the invention is considered to be that the device comprises a patient bed with a patient support surface, wherein the measurement volume extends along the patient bed by several tens of centimeters beyond the height of the patient support surface, wherein the at least four magnetoelectric transmitters are arranged in each leg of the patient bed. A device designed in this way is suitable for implementing the concept for automated ultrasound measurement on patients described in C. Bald et al., "Automatic Localization of an Ultrasound Probe with the Help of Magnetic Sensors", Current Directions in Biomedical Engineering 2022;8(2): 317-320, https: / / doi.org / 10.1515 / cdbme-2022-1081.As is known, an ultrasound imaging system with a movable measuring head can be used to non-invasively examine a living patient, wherein the pose of the measuring head is generally selected and varied by the physician based on the ultrasound image acquired in real time in order to view the medically relevant structures. By means of the invention, it is now advantageously possible for the at least one magnetic field sensor to be integrated into a hand-held measuring head of an ultrasound imaging measuring device and for the evaluation unit to be designed for data-exchanging communication with the ultrasound measuring device and for the pose estimate of the measuring head to be assigned to the acquired ultrasound images in a time-synchronized manner. The overall measuring system can thus provide measurement data that can, for example, also be used to automatically generate a three-dimensional view of the examined area of ​​the patient.

[0044] In principle, a pose estimation system based on ME sensors and ME transmitters with a common resonance frequency (= reference magnetic field frequency) in the low to mid-kilohertz range, i.e., a few kHz to a few hundred kHz, is quite suitable for a medical-technical environment such as a medical examination room. This is because numerous power supply lines to the equipment used there produce magnetic interference fields at 50 Hz and higher harmonics, which can be neglected above about 1 kHz. In contrast, digital circuits and processors today typically operate at high MHz to GHz frequencies, so interference fields from these frequencies are also not detected by mini ME sensors. The ME transmitters, in turn, generate only very low field magnitudes themselves—when activated—which cannot interfere with the surrounding electronics.

[0045] Finally, it should be noted that while time-multiplexing the input signals for magnetic field generators is particularly simple and therefore preferred for measuring magnetic fields with ME sensors and for distinguishing between generators, it is not the only option. Magnetoelectric bending beams generally exhibit a plurality of natural vibration modes, for example, bending modes with at least one nodal line perpendicular to the longitudinal axis of the beam, or torsional modes. These modes generally differ in their natural frequencies and can therefore be specifically resonantly excited with different magnetic field frequencies. This represents an additional option for improving pose estimation, either by using a larger number of reference magnetic field generators or, alternatively, by shortening the measurement cycles and increasing the temporal density of the 6D trajectories.

Claims

A N S P R Ü C H E 1. A device for magnetic pose estimation of an object in a measurement volume subjected to a reference magnetic field, comprising at least four electrically controllable magnetic field generators with different predetermined poses in the edge region of the measurement volume, at least one magnetic field sensor in rigid connection with the object, a control unit for the predeterminable electrical control of the magnetic field generators and an evaluation unit designed to accept and evaluate the electrical output signal of the at least one magnetic field sensor, characterized in that a. the at least one magnetic field sensor is designed as a magnetoelectric bending beam sensor with a previously known mechanical resonance frequency for the simultaneous measurement of three linearly independent components of the reference magnetic field and b.the control unit is designed to carry out the electrical control of the magnetic field generators in a band-limited manner close to the mechanical resonance frequency of the magnetic field sensor, and c. the evaluation unit is designed to estimate the pose of the object in the measurement volume from the output signal of the at least one magnetic field sensor using predetermined and tabulated matched filters, wherein d. at least four of the magnetic field generators are designed as magnetoelectric transmitters.

2. Device according to claim 1, characterized in that at least two magnetic field sensors are arranged rigidly connected at different locations of the object, wherein the evaluation unit is designed to simultaneously accept and process the output signals of the magnetic field sensors.

3. Device according to claim 2, characterized in that the evaluation unit is designed to use the previously known relative pose of the at least two magnetic field sensors to each other to refine the pose estimation of the object.

4. Device according to one of the preceding claims, characterized in that the control unit is designed to control the at least four magnetoelectric transmitters with a time offset such that they only vary the reference magnetic field individually.

5. Device according to one of the preceding claims, characterized in that the evaluation unit comprises non-volatilely stored, tabulated matched filters as components of hypothesis vectors from a calibration procedure, wherein the individual matched filter hypothesis vectors are indexed with different predetermined poses of a magnetic field sensor.

6. Device according to claim 5, characterized in that the evaluation unit is designed to estimate a pose of the magnetic field sensor from a measured output signal vector by interpolating the indices of the matched filter hypothesis vectors with the greatest overlap to the output signal vector.

7. Device according to one of the preceding claims, characterized in that the evaluation unit is designed to select the matched filters to be used for the subsequent pose estimation on the basis of the previous pose estimation.

8. Device according to one of the preceding claims, characterized in that the evaluation unit is designed to carry out the pose estimation of the object at least at 5 Hz, preferably at least at 10 Hz, particularly preferably at least at 20 Hz.

9. Device according to one of the preceding claims, characterized in that the device comprises a patient bed with a patient receiving surface, wherein the measuring volume extends along the patient bed beyond the height of the patient receiving surface by several 10 centimeters, wherein the at least four magnetoelectric transmitters are arranged in each of the supporting legs of the patient bed.

10. Device according to claim 9, characterized in that the at least one magnetic field sensor is integrated in a hand-held measuring head of an imaging ultrasound measuring device and that the evaluation unit is designed for data-exchanging communication with the ultrasound measuring device and a temporally step-by-step assignment of the pose estimate of the measuring head to the acquired ultrasound images takes place.