Self-mixing interferometry measuring method and measurement system
The use of a classifier to categorize SMI signals and eliminate ambiguous solution sets in Self-Mixing Interferometry measurements addresses the ambiguity in determining distance and radial velocity, achieving high accuracy and precision in distance and velocity measurements.
Patent Information
- Application Number
- PCT/EP2024/062233
- Authority / Receiving Office
- WO · WO
- Patent Type
- Applications
- Current Assignee / Owner
- Priority Date
- 2023-05-08
- Filing Date
- 2024-05-03
- Publication Date
- 2024-11-14
AI Technical Summary
Existing methods for determining distance and radial velocity using Self-Mixing Interferometry (SMI) signals face ambiguity due to unknown contributions of distance and speed to the beat frequency, leading to multiple solution sets without clear identification of the actual values.
A classifier, potentially a machine learning algorithm like Random Forest, is employed to categorize SMI signals based on fringe orientation and eliminate solution sets, ensuring only physically valid solutions are considered, thereby selecting the true distance and radial velocity from multiple calculated sets.
This approach achieves high accuracy and precision in determining distance and radial velocity by reducing ambiguity and eliminating physically impossible solutions, allowing for accurate measurements across various conditions and noise levels.
Smart Images

Figure EP2024062233_14112024_PF_FP_ABST
Abstract
Description
[0001] Description
[0002] SELF-MIXING INTERFEROMETRY MEASURING METHOD AND MEASUREMENT SYSTEM
[0003] A measuring method and a measurement system are speci fied .
[0004] One task to be solved is , inter alia, to speci fy a measuring method by which a distance to a target and / or a radial velocity of the target is unambiguously determined . Another task to be solved is , inter alia, to speci fy a measurement system that allows such a measurement .
[0005] These tasks are solved, inter alia, by a measuring method comprising the features of independent claim 1 and by a measurement system comprising the features of claim 14 . Advantageous embodiments and further developments are the subj ect-matter of the respective dependent patent claims .
[0006] According to at least one embodiment the measuring method allows for determining a distance to a target and / or a radial velocity of the target using a SMI signal . The SMI signal comprises a modulation, at least one upwards ramp and at least one downwards ramp . The radial velocity is in particular a component of the velocity in the direction of the SMI sensor .
[0007] SMI is short for the Sel f-Mixing Interferometry . For example , Sel f-Mixing interferometry can be carried out using a measurement setup, also referred to as SMI sensor or as measurement system, comprising a laser, an optical element such as a focusing lens , a target and electronics . With such a measurement setup the SMI signal may be in particular recorded i f a laser beam, i . e . laser radiation, emitted by the laser strikes the target , which scatters the laser radiation . The laser radiation comprises preferably a wavelength of the infrared wavelength range of the electromagnetic spectrum . For example , the wavelength of the laser radiation is between 900 nm and 1000 nm . A component of the scattered radiation may inj ect back into the laser and superimposes on the existing field, i . e . , the radiation of the laser beam, modulating an amplitude and a frequency of the laser . The SMI signal is obtained in this example by observing the voltage of the laser or an output power of a photodiode .
[0008] The modulation in particular determines the upwards ramp and / or the downwards ramp . That is , the modulation is , for example , a triangular modulation achieved by a triangular modulation signal , such that the upwards ramp emerges from a rising flank of the modulation signal and the downwards ramp emerges from a falling flank of the modulation signal .
[0009] According to at least one embodiment of the measuring method an up frequency and down frequency of the SMI signal are determined . In particular, the up frequency is defined by a beat frequency of the upwards ramp and the down frequency is defined by a beat frequency of the downwards ramp . For example , each of the upwards ramp and the downwards ramp of the SMI signal comprise fringes . In particular, the number fringes per unit time gives the respective beat frequency .
[0010] According to at least one embodiment a plurality of solution sets each comprising a calculated distance and a calculated radial velocity obtained by means of the up frequency of the down frequency . For example , the up frequency and to the down frequency are inputs to a set of equations , by means of which the distance and the radial velocity may be calculated . Such a set of equations may give a plurality of solutions for the calculated distance and the calculated radial velocity . That is , from solving the set of equations an ambiguity with respect to the calculated distances and calculated radial velocities emerges , since only one calculated distance and only one calculated radial velocity may physically correctly give the actual distance , respectively the actual radial velocity of the target .
[0011] According to at least one embodiment , one true solution set is selected out of the plurality of solution sets according to a classi fier . In particular, the one true solution sets comprises the actual distance to the target and the actual radial velocity of the target . Preferably, exactly one solution set is the true solution set . That is , exactly one solution set of the plurality of solution sets comprises the actual distance and the actual radial velocity .
[0012] The classi fier may be configured to categori ze the SMI signal and / or the solution sets . That is , the classi fier may be configured to eliminate a solution set from consideration as the one true solution set , for example , i f the solution set comprises a physically impossible solution such as a negative calculated distance . It is also possible that the classi fier may categori ze the SMI signal on the basis of the shape or orientation of the fringes . For example , the classi fier may classi fy the SMI signal in a certain class . This sets the requirement for the solution sets that only these solution sets may be considered as the true solution sets that belong to the same class as the SMI signal . That is, the classifier is in particular configured to reduce the plurality of solution sets to the preferably exactly one true solution set.
[0013] According to at least one embodiment of the measuring method, the classifier comprises a machine learning algorithm. That is, the classifier may perform at least a part of the classification by carrying out a machine learning algorithm. For example, the classification of the SMI signal is carried out using a machine learning algorithm.
[0014] The machine learning algorithm may be Random Forest classifier that is trained with a dataset comprising a plurality of SMI signals with known classification. The dataset for training the machine learning algorithm may comprise simulated SMI signals. For example, the dataset may comprise at least 5,000 or at least 7,000 or at least 10,000 SMI signals to achieve a good true positive value, a good true negative value, a good precision and / or a good accuracy. For example, the accuracy may be at least 0.9 or at least 0.92 or at least 0.95.
[0015] In at least one embodiment the measuring method for determining a distance to a target and / or a radial velocity of the target using a SMI signal comprising a modulation, at least one upwards ramp and at least one downwards ramp, comprises the steps of determining an up frequency defined by a beat frequency of an the upwards ramp of the SMI signal, determining a down frequency defined by a beat frequency of the downwards ramp of the SMI signal, and obtaining a plurality of solution sets, each comprising a calculated distance and calculated radial velocity by means of the up frequency and the down frequency. One true solution set is selected out of the plurality of solution sets according to a classi fier . The classi fier comprises a machine learning algorithm .
[0016] The measuring method described here is based on the following technical considerations . SMI can be applied to distance and velocity measurements . I f a laser of a corresponding SMI sensor is driven with a constant current , a displacement of the target can be measured, as well as its direction . This measurement can be carried out by analyzing a temporal signal , i . e . peak-counting . For an absolute measurement of a distance of the target , a modulation of laser radiation emitted by the laser can be applied using a periodic signal , commonly a triangular waveform . Using this triangular modulation, a velocity of the target at the same time as the absolute distance may be computed . In particular, the velocity calculated will be the radial component , or in other terms a component of the velocity in the direction of the SMI sensor .
[0017] Due to unknown contributions of a distance contribution and a speed contribution to the beat frequency of the SMI signal , a plurality of independent equations , for example three equations , have to be solved in order to obtain the distance and the radial velocity . From each of these equations a solution set for the distance and the radial velocity follows . Since the contributions of the distance contribution and the speed contribution are unknown, it cannot be unambiguously deduced which solution set includes the actual distance and the actual radial velocity to be measured . It is therefore needed to establish a method to discriminate the solution sets in order to find the one true solution set including the actual distance and the actual radial velocity . Known approaches to this problem include separate methods for either displacement or velocity measurement , which are computed using classical methods as fringe counting, Fast Fourier Trans form ( FFT ) based methods or Extended Kalman Filters (EKF) . Furthermore , there are approaches that include the computation of a phase of a 2nd harmonic of a FFT of the SMI signal to discern the direction of a movement of the target . Other approaches focus on, for example , fringe counting for vibration measurements . Another possible approach would be to follow a physical predictive model to estimate the most reasonable speed and distance compared to previous samples ( i . e . Kalman filter ) . This approach is reasonable , but the SMI signal can also detect rotational speed, which is comparably hard to model . Moreover, these models need to be tuned to the system which is in general time-consuming .
[0018] The method described herein is based on the idea of using a classi fier which is configured to eliminate solution sets . In particular the classi fier is configured to classi fy an SMI signal , for example according to its fringes , and compare the class of the SMI signal with a class of the solution sets . Thereby, at least one solution set may be eliminated from consideration, since the class of the SMI signal and the class of the respective solution set do not comply . To perform the classi fication, the classi fier comprises a machine learning algorithm that is , for example , trained with a plurality of SMI signals with known class .
[0019] Advantageously, the classi fication can be carried out in a simple and fast manner once the classi fier is trained .
[0020] Furthermore , a high accuracy and precision may be achieved by providing suf ficient training data for the machine learning algorithm .
[0021] It has also been shown that the classi fier can be trained with a particular set of operation parameters for the SMI measurement setup and represents a simple procedure to retain the model on new parameters . Hence , the classi fier can be applied to a variety of measurement conditions and applications . Operation parameters for the SMI measurement setup may be , for example , a wavelength of the laser radiation emitted by the laser, a coupling factor, also referred to as C- factor or feedback factor, and a modulation frequency of the modulation and / or a frequency excursion . In particular, the coupling factor measures the quantity of radiation reinj ected into the laser and determines the feedback regime in which the laser is working .
[0022] Furthermore , it is possible to adapt the classi fier to di f ferent levels of noise . For example , the classi fier still gives accurate results , even i f the machine learning algorithm is trained with simulated data or data comprising a low noise level .
[0023] According to at least one embodiment of the measuring method, the calculated distance of the true solution set is greater than zero . That is , solution sets that comprise a calculated distance with a negative value may be eliminated from consideration as the true solution set . The determination whether a calculated distance of a certain solution set is greater or smaller than zero may be carried out by the classi fier . According to at least one embodiment of the measuring method the SMI signal comprises a plurality of fringes. The SMI signal is categorized by the classifier based on an orientation of the fringes. The fringes of the SMI signal may be asymmetric. This means in particular that the fringes do not show a sinusoidal shape but, for example, the fringes show a sawtooth shape. In particular the sawtooth-shaped fringes show an orientation. The asymmetry of the fringes of the SMI signal increases with increasing coupling factor. That is, the higher the coupling factor, the more asymmetrical and sawtooth-shaped are the fringes.
[0024] For example, a sawtooth-shaped fringe may have a rising flank, where the fringe, i.e. the SMI-signal, increases essentially linear and a falling flank, where the fringe, i.e. the SMI-signal, decreases abruptly. Such an orientation is also referred to as a right orientation.
[0025] Alternatively, a sawtooth-shaped fringe may have a rising flank, where the fringe, i.e. the SMI-signal, increases abruptly and a falling flank, where the fringe, i.e. the SMI- signal, decreases approximately linear. Such an orientation is also referred to as a left orientation.
[0026] Furthermore, a fringe may show essentially no rising and / or falling flank. In this case the orientation is also referred to as no orientation.
[0027] According to at least one embodiment of the measuring method, the fringes comprise fringes of a first kind at the upwards ramp and fringes of a second kind at the downwards ramp. The classifier classifies the SMI signal in a class I if the fringes of the first kind have a left orientation and the fringes of the second kind have a left orientation or no orientation, in a class II if the fringes of the first kind have a left orientation and the fringes of the second kind have a right orientation, and in a class III if the fringes of the first kind have a right orientation or no orientation and the fringes of the second kind have a right orientation.
[0028] The fringes of the first kind may correspond to the up frequency and the fringes of the second kind may correspond to the down frequency. In particular, the classification of the SMI signal is carried out by the classifier, preferably by the machine learning algorithm of the classifier. That is the classifier, respectively the machine learning algorithm take SMI signal, in particular the SMI time signal as input and infers the orientation and may in turn categorize the SMI signal in class I, class II or class III.
[0029] According to at least one embodiment of the measuring method, the calculated radial velocity of the true solution set is smaller than a negative of a limit velocity if the SMI signal is categorized in class I, greater than the negative of the limit velocity and smaller than the limit velocity if the SMI signal is categorized in class II, greater than the limit velocity if the SMI signal is categorized in class III. The limit velocity determines in particular at which velocity the up frequency is zero. In particular the down frequency is zero at the negative of the limit velocity. The limit velocity may be calculated by the following equation (1) : wherein Vumdenotes the limit velocity, denotes a wavelength of the laser, and faown denotes the down frequency where the up frequency is zero.
[0030] In particular, the negative of the limit velocity determines a cross over between class I and class II. That is, a solution set with a calculated velocity smaller than the negative of the limit velocity is classified in class I. Furthermore, the limit velocity in particular determines a cross over between class II and class III. That is, a solution set with a calculated velocity greater than the limit velocity is classified in class III and a solution set with a calculated velocity between the negative of the limit velocity and the limit velocity is classified in class II. The classification of the solution sets is carried out, for example, by the classifier.
[0031] It is thus possible to determine by the classifier in which class the SMI signal and the solution sets are classified.
[0032] Additionally or alternatively, at least one solution set is eliminated for physical reasons as the distance is, for example, negative for this at least one solution set.
[0033] According to at least one embodiment of the measuring method, the categorization of SMI signal based on the orientation of the fringes is carried out by the machine learning algorithm of the classifier. Preferably, a classification of the solution sets is carried out by the classifier.
[0034] According to at least one embodiment of the measuring method, the SMI signal is generated by radiating the target with a primary radiation emitted by a laser. A second radiation scattered and / or reflected by the target is superimposed with the primary radiation . That is , electromagnetic fields corresponding to the primary radiation and the secondary radiation are superimposed . Furthermore , an interference pattern of the primary radiation and the secondary radiation is observed .
[0035] The primary radiation and / or the secondary radiation preferably each comprise a wavelength in the infrared wavelength range of the electromagnetic spectrum . For example , the wavelength of the primary and / or secondary radiation is between 900 nm and 1000 nm .
[0036] For example , the superimposition of the primary and secondary radiation takes place in the laser . For example , the secondary radiation is coupled into a resonator of the laser, where an electromagnetic field corresponding to the primary radiation is present .
[0037] It is also possible that a part of the primary radiation is guided to a detector element , for example , by a waveguide . In this case the secondary radiation is coupled towards the detector element , where the superimposition of the primary and secondary radiation takes place .
[0038] According to at least one embodiment of the measuring method, the interference pattern is observed by a variation in a driving voltage of the laser . For example , i f the interference pattern emerges in the laser resonator, the driving voltage of the laser changes . From the change of the driving voltage the interference pattern and hence the SMI signal may be obtained . According to at least one embodiment of the measuring method, the interference pattern is observed by a detector element . For example , an output power of the detector element is measured . The detector element is preferably a photodiode . For example , the primary radiation and the secondary radiation are directed to the detector element and superimposed at the detector element . For example , primary radiation is directed to the detector element by a waveguide . It is possible that a beam splitter is arranged between the laser and the target which directs a part of the primary radiation emitted by the laser to the waveguide , respectively the detector element and another part of the primary radiation to the target .
[0039] According to at least one embodiment of the measuring method, three solution sets are obtained by solving the following equations of motion : where fupdenotes the up frequency, fdown denotes the down frequency, / distancedenotes a distance contribution to the beat frequency, and fspeed denotes a speed contribution to the beat frequency . In each case the calculated distance is calculated by multiplying the distance contribution to the beat frequency by the speed of light and dividing the distance contribution to the beat frequency by four times the frequency of the modulation and the frequency excursion . Further in each case the calculated radial velocity is calculated by multiplying the speed contribution to the beat frequency by a hal f of a wavelength of the primary radiation .
[0040] In particular, equation ( 2 ) gives the actual distance and the actual velocity i f the speed contribution to the beat frequency is greater than or equal to the distance contribution of the beat frequency . Equation ( 3 ) gives in particular the actual distance and the actual velocity i f the speed contribution to the beat frequency is smaller than the distance contribution to the beat frequency and greater than or equal to a negative of the distance contribution to the beat frequency . In particular, equation ( 4 ) gives the actual distance and the actual velocity i f the speed contribution to the beat frequency is smaller than to the negative of the distance contribution of the beat frequency . However, in general the speed contribution and the distance contribution to the beat frequency are not known . Therefore , all equations have to be solved, resulting in a plurality of solution sets . In particular, these solution sets are classi fied by the classi fier in order to find the actual distance and the actual radial velocity .
[0041] According to at least one embodiment of the measuring method, the modulation of the SMI signal is generated by a modulated drive current of the laser . That is , by controlling the drive current of the laser, the modulation of the SMI signal may be chosen . Thus , a precise controlling of the modulation is possible .
[0042] According to at least one embodiment of the measuring method, the modulation of the SMI signal is a triangular modulation . Advantageously, a triangle modulation gives well defined upwards ramps and downwards ramps .
[0043] According to at least one embodiment of the measuring method, the up frequency and the down frequency are obtained by performing a Fourier trans formation on the SMI signal . For example , a Fast Fourier trans formation algorithm is applied to the SMI signal . By performing the Fourier trans formation the frequency components such as the up frequency and the down frequency of the SMI signal may be obtained .
[0044] It is possible that the Fourier trans formation is supported by a bandpass filter around the up frequency and / or the down frequency . By such a bandpass filter, perturbing frequency components of the SMI signal that do not contribute to the up frequency and / or the down frequency may be filtered . Such perturbing frequency components may emerge from the modulation or noise .
[0045] Furthermore , a measurement system is described . The measurement system is in particular configured to carry out a measuring method described herein . That is , all features disclosed for the measurement system are also disclosed for the measuring method and vice versa .
[0046] In at least one embodiment the measurement system comprises at least one laser with a resonator axis configured to emit a primary radiation, at least one optical element and a processing unit . The optical element is configured to direct the primary radiation to a target . The processing unit is configured to carry out the measuring method described herein . The measurement system may also be referred to as measurement setup or SMI sensor . In particular, the primary radiation is a laser radiation generated in a laser resonator of the laser . The optical element is , for example , a lens such as a focusing lens . That is , the optical element is in particular configured to focus the primary radiation to the target .
[0047] The processing unit may be additionally configured to supply the laser with a current or to control the supply of current of the laser . For example , the processing unit may vary or control a variation of the current supply of the laser in order to modulate the primary radiation, which in turn leads to the modulation of the SMI signal .
[0048] According to at least one embodiment of the measuring system, a coupling factor of the measurement system is smaller than 1 and / or greater than 1 . For example , the coupling factor is smaller than 10 or smaller than 50 or smaller than 100 or smaller than 1000 . The coupling factor is also referred to as C- factor or feedback factor . The coupling factor in particular measures a quantity of light reinj ected into the laser and determines the feedback regime in which the laser is working . For example , i f the coupling factor is smaller than 1 , the optical feedback is weak and i f the coupling factor is between 1 and 10 , the optical feedback is moderate .
[0049] According to at least one embodiment , the at least one laser is a semiconductor laser comprising an active zone . The resonator axis is perpendicular to the active zone . This means in particular that the resonator axis is perpendicular to a main extension plane of the active zone . For example , the semiconductor laser is a vertical-cavity surface-emitting laser, VCSEL for short . In particular, the semiconductor laser comprises a semiconductor layer sequence that includes the active zone . During operation, electromagnetic radiation, that is the primary radiation, is produced in the active zone . For example , the semiconductor layer sequence of the laser is based on an I I I-V compound semiconductor material . Preferably a laser resonator of the semiconductor laser is arranged at least partially in the active zone .
[0050] According to at least one embodiment the measurement system comprises a plurality of lasers and resonator axes of the lasers are arranged at an angel from each other . Preferably, the lasers are VCSELs . Such a measurement system may register a plurality of di f ferent SMI signals and each SMI signal may be treated using the method described herein . The information of distance and velocity from each sensor will then be used to compute a distance-velocity pair that can be used for di f ferent applications , for example odometry or simultaneous locali zation and mapping, SLAM for short .
[0051] Further advantages and advantageous embodiments and further developments of the measuring method and the measurement system will become apparent from the following exemplary embodiments shown in connection with schematic drawings . Identical elements , elements of the same kind or elements having the same ef fect , are provided with the same reference signs in the figures . The figures and the proportions of the elements shown in the figures are not to be regarded as true to scale . Rather, individual elements may be shown exxageratedly large for better representability and / or for better comprehensibility .
[0052] In the figures : Figure 1 shows a schematic illustration of a measurement system described here according to a first exemplary embodiment ;
[0053] Figure 2 shows a flow chart of a measuring method described here according to an exemplary embodiment ;
[0054] Figures 3 to 5 show examples of SMI signals used in a measuring methods described here ;
[0055] Figures 6 and 7 illustrate di f ferent classes a classi fier used in a measuring method described here is configured to classi fy in;
[0056] Figure 8 shows a confusion matrix of a machine learning algorithm used in the exemplary embodiment of the measuring method;
[0057] Figure 9 shows a measurement system described here according to a second exemplary embodiment .
[0058] Figure 1 illustrates a measurement system according to a first exemplary embodiment . The measurement system comprises a laser 100 , an optical element 101 and a processing unit 102 . The laser 100 is configured to emit a primary radiation 110 in radiation direction 104 . The optical element 101 is configured to focus the primary radiation 110 . For example , the optical element 101 is a focusing lens .
[0059] The laser 100 is preferably a semiconductor laser 100 . The semiconductor laser 100 comprises a semiconductor layer sequence that includes an active zone . During operation, electromagnetic radiation, that is , the primary radiation
[0060] 110 , is produced in the active zone . For example , the semiconductor layer sequence of the laser 100 is based on an
[0061] I I I-V compound semiconductor material . Preferably a laser resonator of the semiconductor laser is arranged at least partially in the active zone . The primary radiation 110 comprises a wavelength in the infrared wavelength range of the electromagnetic spectrum . For example , the wavelength of the primary radiation 110 is between 900 nm and 1000 nm .
[0062] A resonator axis 103 of the laser 100 is perpendicular a main extension plane of the active zone . In particular, the semiconductor laser 100 is a VCSEL .
[0063] During intended operation the laser 100 emits the primary radiation 110 in radiation direction 104 . The optical element 101 focuses the primary radiation 110 to a target 120 . The target 120 is arranged at a distance 1 from the laser 100 . The target 120 furthermore comprises a velocity 21 . The velocity 21 has a radial component , referred to as radial velocity 2 . The radial velocity 2 is the component of the velocity 21 which is in the direction of the radiation direction 104 .
[0064] The primary radiation 110 hits the target 120 and is scattered and / or reflected as a secondary radiation 111 . The secondary radiation 111 at least partially propagates in the negative radiation direction 104 . This component of the secondary radiation 111 is focused by the optical element 101 in the reinj ected into the laser 100 . Thereby, the secondary radiation 111 interferes with an electrical field of the primary radiation 110 present at a resonator of the laser 100 . Due to this interference an interference pattern may be observed by detecting a variation in a supply voltage of the laser 100 . In particular, the interference pattern, respectively the variation in the supply voltage , gives a SMI signal . In particular, this SMI signal is a raw SMI signal 30 .
[0065] The raw SMI signal 30 is processed by the processing unit 102 in order to determine the distance 1 and the radial velocity 2 of the target 120 . The processing unit 102 is further configured to modulate a supply current for the laser 100 . By modulating the supply current for the laser 100 the primary radiation 110 may comprise a modulation 4 . The supply current for the laser 100 is preferably modulated with a triangle modulation 4 . Consequently, the primary radiation 110 and as a result also this secondary radiation 111 and hence the raw SMI signal 30 each comprise a triangle modulation 4 . Due to the triangular modulation 4 , the raw SMI signal 30 comprises at least one upwards and at least one downwards ramp .
[0066] Figure 2 shows a flow chart illustrating a measuring method described herein according to an exemplary embodiment . The measuring method is in particular carried out by the processing unit 102 ( compare figure 1 ) . In a first step of the measuring method the raw SMI signal 30 is processed to a SMI signal 3 . For example , the raw SMI signal 30 is split by periodical modulation . In particular after the first step, the SMI signal comprises one upwards ramp and one downwards ramp ( compare figures 3 to 5 ) .
[0067] In a subsequent step the SMI signal 3 is analyzed in order to obtain the distance 1 and the radial velocity 2 . Thereby, two branches of steps are carried out essentially simultaneously and / or independent from each other . In the first branch ( left branch in figure 2) three solution sets each including a calculated velocity and a calculated radial velocity. First, a Fourier transform 7 is applied to the SMI signal 3 to identify the frequency contributions of the SMI signal 3. In particular, a Fast Fourier Transform is carried out.
[0068] In a next step, an up frequency 31 and a down frequency 32 is obtained. The up frequency 31 is defined by a beat frequency of the upwards ramp and the down frequency 32 is defined by a beat frequency of the downwards ramp (compare figures 3 to 5) . The up frequency 31 and the down frequency 32 are obtained, for example, by peak counting of the Fourier transform of the SMI signal 3.
[0069] From the up frequency 31 and the down frequency 32 three solution sets are calculated by the equations (2) to (5) discussed above. For example, a first solution set comprises a calculated distance of -0.037 and a calculated radial velocity of 0.0032, a second solution set comprises a calculated distance of 0.301 and a calculated radial velocity of -0.00039, and third solution set comprises a calculated distance of 0.037 and a calculated radial velocity of - 0.0032, wherein each calculated distance and each calculated velocity are given in arbitrary units.
[0070] In the right branch of figure 6 a classifier 5 classifies the SMI signal 3 and the solution sets, in order to obtain one true solution sets, which includes a calculated distance and calculated radial velocity describing the actual distance 1 and actual radial velocity 2. In particular, the classifier 5 classifies the SMI signal 3 in a class I, a class II or a class III depending on an orientation 300, 301, 302 of fringes 33, 34 of the SMI signal 3. In turn the classifier 5 may classify the solution sets depending on whether their calculated velocity is smaller or greater than a limit velocity 20 and smaller or greater than a negative of the limit velocity -20.
[0071] To classify the SMI signal 3, fringes 33, 34 of the SMI signal 3 are analyzed. The SMI signal 3 shows fringes of a first kind 33 at the upwards ramp and fringes of a second kind 34 at the downwards ramp (compare figures 3 to 5) . The fringes 33, 34 show an asymmetry determining an orientation 300, 301, 302 of the fringes 33, 34.
[0072] Figures 3 to 5 show examples of SMI signals 3 for different coupling factors. In Figure 3 the coupling factor is 0.5, in figure 4 the coupling factor is 0.99, and in figure 5 the coupling factor is 1.5. By comparing figures 3 to 5 it can be seen that that the coupling factor influences the shape of the SMI signal 3. In particular, the higher the coupling factor, the more asymmetrical and sawtooth-shaped are the fringes 33 , 34.
[0073] Due to their asymmetry the fringes 33, 34 each comprises an orientation 300, 301, 302. The fringes 33, 34 may show a left orientation 301, where a rising flank of the fringe 33, 34 rises abruptly and a falling flank of the fringe 33, 34 decreases approximately linear. Alternatively, the fringes 33, 34 may show a right orientation 302, where a rising flank of the fringe 33, 34 rises approximately linear and a falling flank of the fringe 33, 34 decreases abruptly. It is also possible, that the fringes 33, 34 show rising or falling flank. This case is referred to as no orientation 300 (compare figure 6) . In figures 3 to 5, for example, fringes of the first kind 33 and fringes of the second kind 34 have a right orientation 302. Fringes of the first kind 33 correspond to the upwards ramp and fringes of the second kind 34 to the downwards ramp of the SMI signal 3.
[0074] By determining the orientation of the fringes of the first kind 33 and the fringes of the second kind 34, the SMI signal 3 may be categorized in class I, class II or class III. As illustrated in figure 6, that is, if the fringes of the first kind 33 have a left orientation 301 and the fringes of the second kind 34 have a left orientation 301 or no orientation 300, the SMI signal 3 is classified in class I. If the fringes of the first kind 33 have a left orientation 301 and the fringes of the second kind 34 have a right orientation 302, the SMI signal 3 is classified in class II. If the fringes of the first kind 33 have a right orientation 302 or no orientation 300 and the fringes of the second kind 34 have a right orientation 302, the SMI signal 3 is classified in class III.
[0075] At the same time, the solution sets may be classified depending on their calculated velocities. As is shown in figure 6, in class I the radial velocity 2 is smaller or equal to the negative of the limit velocity -20. In class II, the radial velocity 2 is between the negative of the limit velocity -20 and the limit velocity 20. In class III, the radial velocity is greater or equal to the limit velocity 20. This means that the limit velocity 20 and the negative value of the limit velocity -20 determine the transitions between the classes I, II, III.
[0076] As shown in figure 7, which shows a graphical representation of the up frequency 31 and the down frequency 32 in dependency of the radial velocity, the negative of the limit velocity -20 and the limit velocity 20 are roots of the up frequency 31 and the down frequency 32, respectively. That is, if the up frequency 31 is zero, the radial velocity 2 is the negative of the limit velocity -20, and if the down frequency 32 is zero, the radial velocity 2 is the limit velocity 20.
[0077] The classifier 5 classifies the SMI signal 3 by using a machine learning algorithm. For example, the machine learning algorithm comprises a Random Forest classifier. The machine learning algorithm is trained, for example, on a dataset composed of simulated SMI signals.
[0078] For example, the data is simulated using a coupling factor 0.99, a wavelength of the laser 100 of 940 nm and a triangular modulation of a frequency of 100 Hz. In particular, such parameters may match parameters of an application of a SMI sensor. In total, the training dataset comprises 5600 datapoints, i.e., SMI signals, each containing one modulation period. For example, each datapoint corresponds to a velocity ranging between 0.01 and -0.01 in arbitrary units, and a distance ranging between 0.2 and 0.8 meters. The datapoints were distributed as follows: 14 different distance values, each having 400 velocity values. In particular, each datapoint is demodulated, then a central window containing part of the upwards ramp and part of the downward ramp is fed to the classifier 5.
[0079] For illustration, the machine learning algorithm is trained on a random 80% of the dataset and tested on the remaining 20%. Fig. 5 shows the results of the testing in a Confusion Matrix, where 51 denotes the true class and 52 denotes the predicted class. It can be seen that most test datapoints are correctly classified. The highest error is shown to be velocities of class II being classified as negative. These results could be further improved by an in-depth analysis of the classifier. From the Confusion Matrix the classifier 5 shows an accuracy of 0.925, a precision score for class I of 0.860, for class II of 0.942, and for class III of 0.970. Further, a recall score for class I is 0.939, for class II is 0.883, for class III is 0.975. A Fl score for class I is 0.898, for class II is 0.911, for class III is 0.973.
[0080] In particular, the numbers shown in the Confusion Matrix represent a percentage, so that the total of the grid amounts to 1. Each line in the Confusion Matrix represents a true values of the samples, i.e. label I is class I, label II is class 2 and label III is class III. Each row represents the label predicted by the classification model, i.e. the classifier. As such, by summing the numbers in the rows it may be computed that the samples were divided into the 3 classes as follows: 27.7% in class I, 43.04% in class II and 29.5% in class III.
[0081] Moreover, 26% of total samples are true class I and are classified as class I, 38% of total samples are true class II and are classified as class II and 28% of total samples are true class III and are classified as class III. This means that the 93.8%, 88.3% and 94.9% of class I, class II and Class III respectively are correctly classified. The remaining percentages give the errors for mis-classif led s amp les.
[0082] The machine learning algorithm, respectively the classifier
[0083] 5, represents a simple procedure to retain the model and may therefore be used to classify a SMI signals over a wide range of operation variables with a high precision.
[0084] By classifying the SMI signal 3 and the solution sets the classifier 5 may be used to determine which of the three solution sets obtained by the up frequency 31 and the down frequency 32 include the actual distance 1 and the actual radial velocity 2. It is possible that at least one solution set is additionally eliminated from consideration as the true solution set, as the calculated distance of this solution set is negative.
[0085] Referring back to figure 6, the distance 1 and the radial velocity 2 can be determined by performing the steps of the left branch and the right branch of the flow chart.
[0086] For the example, where a first solution set comprises a calculated distance of -0.037 and a calculated radial velocity of 0.0032, a second solution set comprises a calculated distance of 0.301 and a calculated radial velocity of -0.00039, and third solution set comprises a calculated distance of 0.037 and a calculated radial velocity of - 0.0032, wherein each calculated distance and each calculated velocity are given in arbitrary units, the first solution set may be eliminated from consideration since the calculated distance is negative. In the specific example, the SMI signal 3 is classified in class II, for example. In the example, the absolute value of the calculated velocity of the second solution set is smaller than the limit velocity and the absolute value of the calculated velocity of the third solution set is greater than the limit velocity. Therefore, the second solution set is the true solution set. Figure 7 shows a measurement system 10 according to a secondary embodiment . In di f ference to the first exemplary embodiment of figure 1 , the measurement system 10 according to figure 7 comprises a plurality of lasers 100 . The lasers 100 are preferably VCSELs and are similar to the laser 100 of figure 1 . Resonator axes 103 of the lasers 100 are arranged at an angle from each other . The measurement system 10 may register a plurality of di f ferent SMI signals 3 and each SMI signal 3 may be treated using the measuring method described herein . The information of distance and velocity from each laser 100 is used to compute a distance-velocity pair that can be used for di f ferent applications .
[0087] This patent application claims the priority of German patent application 102023111943 . 5 , the disclosure content of which is hereby incorporated by reference .
[0088] References
[0089] 1 distance
[0090] 2 radial velocity
[0091] 3 SMI signal
[0092] 4 modulation
[0093] 5 classifier
[0094] 7 Fourier transform
[0095] 10 measurement system
[0096] 20 limit velocity
[0097] 21 velocity of the target
[0098] 30 raw SMI-signal
[0099] 31, 32 up / down frequency
[0100] 33, 34 fringes of first / second kind
[0101] 35 signal voltage
[0102] 36 time
[0103] 37 frequency
[0104] 51 true label
[0105] 52 predicted label
[0106] 100 laser
[0107] 101 optical element
[0108] 102 processing unit
[0109] 103 resonator axes
[0110] 104 radiation direction
[0111] 110 primary radiation
[0112] 111 secondary radiation
[0113] 120 target
[0114] 300, 301, 302 no / left / right orientation
[0115] I, II, III classes
Claims
Claims1. Measuring method for determining a distance (1) to a target (120) and / or a radial velocity (2) of the target (120) using a SMI signal (3) comprising a modulation, at least one upwards ramp and at least one downwards ramp, comprising the steps of- determining an up frequency (31) defined by a beat frequency of the upwards ramp of the SMI signal (3) ,- determining a down frequency (32) defined by a beat frequency of the downwards ramp of the SMI signal (3) ,- obtaining a plurality of solution sets, each comprising a calculated distance and calculated radial velocity by means of the up frequency (31) and the down frequency (32) , wherein- one true solution set is selected out of the plurality of solution sets according to a classifier (5) , and- the classifier (5) comprises a machine learning algorithm.
2. Measuring method according to claim 1, wherein the calculated distance of the true solution set is greater than zero.
3. Method according to one of the preceding claims, wherein- the SMI signal (3) comprises a plurality of fringes (33, 34) ,- the SMI signal (3) is categorized by the classifier(5) based on an orientation (300, 301, 302) of the fringes ( 33 , 34 ) .
4. Method according to claim 3, wherein the fringes (33, 34) comprise fringes of a first kind(33) at the upwards ramp and fringes of a second kind(34) at the downwards ramp,- the classifier (5) classifies the SMI signal (3) in- a class I if the fringes of the first kind (33) have a left orientation (301) and the fringes of the second kind (34) have a left orientation (301) or no orientation (300) ,- a class II if the fringes of the first kind (33) have a left orientation (301) and the fringes of the second kind (34) have a right orientation (302) ,- a class III if the fringes of the first kind (33) have a right orientation (302) or no orientation (300) and the fringes of the second kind (34) have a right orientation (302) .
5. Method according to claim 4, wherein the calculated radial velocity of the true solution set is- smaller than a negative of a limit velocity (20) if the SMI signal (3) is categorized in class I,- greater than the negative of the limit velocity (20) and smaller than the limit velocity (20) if the SMI signal (3) is categorized in class II,- greater than the limit velocity (20) if the SMI signal (3) is categorized in class III.
6. Method according to one of claims 3 or 5, wherein the categorization of SMI signal (3) based on the orientation (300, 301, 302) of the fringes (33, 34) is carried out by the machine learning algorithm of the classifier (5) .
7. Method according to one of the preceding claims, wherein the SMI signal (3) is generated by- radiating the target (120) with a primary radiation (110) emitted by a laser (100) ,- superimposing a second radiation (111) scattered and / or reflected by the target (120) with the laser radiation (110) ,- observing an interference pattern of the primary radiation (110) and the secondary radiation (111) .
8. Method according to claim 7, wherein the interference pattern is observed by a variation in a driving voltage of the laser (100) .
9. Method according to claim 7, wherein the interference pattern is observed by a detector element.
10. Method according to one of the preceding claims, wherein three solution sets are obtained by solving the following equations of motion:where fupdenotes the up frequency (31) , fdown denotes the down frequency (32) , fdistance denotes a distance contribution to the beat frequency, and fspeed denotes a speed contribution to the beat frequency, wherein in each case- the calculated distance is calculated by multiplying the distance contribution to the beat frequency by the speed of light and dividing the distance contribution tothe beat frequency by four times a frequency of the modulation and a frequency excursion, and- the calculated radial velocity is calculated by multiplying the speed contribution to the beat frequency by a half of a wavelength of the primary radiation.
11. Method according to one of claims 7 to 10, wherein the modulation is generated by a modulated drive current (4) of the laser (100) .
12. Method according to one of the preceding claims, wherein the modulation is a triangle modulation.
13. Method according to one of the preceding claims, wherein the up frequency (31) and the down frequency (32) are obtained by performing a Fourier transformation (7) on the SMI signal (3) .
14. Measurement system (10) comprising- at least one laser (100) with a resonator axis (103) configured to emit a primary radiation (110) ,- at least one optical element (101) ,- a processing unit (102) , wherein- the optical element (101) is configured to the primary laser radiation (110) to a target (120) ,- the processing unit (102) is configured to carry out the measuring method according to one of the preceding claims .
15. Measurement system (10) according to claim 14, wherein a coupling factor of the measurement system is smaller than 1 and / or greater than 1 and smaller than 10.
16. Measurement system (10) according to claim 14 or 15, wherein- the at least one laser (100) is a semiconductor laser diode comprising an active zone,- the resonator axis (103) of the at least one laser (100) is arranged perpendicular to the active zone.
17. Measurement system (10) according to one of claims 14 to 16, wherein the measurement system (10) comprises a plurality of lasers (100) and resonator axes (103) of the lasers (100) are arranged at an angel from each other .
Citation Information
Patent Citations
Laser velocimetric image scanning
US20070102523A1
DE102023111943A
