Method and device for determining measurement information and LiDAR device

Abstract

Method (100) for determining measurement information based on a plurality of measured values (221) from a measured value range (223), the method comprising: Obtaining (120) a frequency distribution (222) of a plurality of measured values (221), wherein the measured values (221) of the frequency distribution (222) are each assigned to a class of a plurality (228) of classes of the frequency distribution (222), and wherein a frequency value (224) of a class (226) describes a number of measured values assigned to the class (226), Dividing (140) the frequency distribution (222) into a plurality of ranges (344), wherein each of the ranges (344) represents an interval of the measured value range (223) and contains one or more classes (226) of the frequency distribution (222), Selecting (160) one class of a respective area (344) as a selected class (366) of the respective area (344) based on a selection rule, wherein the areas (344) are each assigned an area feature (364) based on the selection rule, Determining (180) a probability value (384; 387) for one of the selected classes (366) based on the range features (364), wherein the probability value (384; 387) represents an estimate of the probability with which the selected class represents a value of a useful signal, wherein the determination of the probability value is based on machine learning methods.

Description

Technical area

[0001] Embodiments of the invention relate to a method and a device for determining measurement information based on a plurality of measured values from a measured value range. Further embodiments relate to a light detection and ranging (LiDAR) device. Some embodiments relate to a method for improving the reliability of LiDAR measurements using machine learning and block-correlated analysis. Some embodiments relate to processing TCSPC histograms in a digital process stage of LiDAR systems. background

[0002] When measuring a desired signal or a wanted signal, it is often difficult to distinguish whether a measured value is based on the desired signal or whether the measured value is the result of a background signal. For example, a measuring device used to measure the desired signal can be influenced or triggered by an interfering signal of the same type as the desired signal, so that a measured value generated by the measuring device is attributable to the interfering signal. Furthermore, the background signal can be due to noise from the measuring device, for example. Especially with very sensitive measuring devices or in cases where the desired signal is very weak, resulting in a particularly small signal-to-noise ratio, it can be difficult to derive the desired information from the measured values.

[0003] An example of a particularly sensitive detector is a single-photon avalanche diode (SPAD), which is capable of detecting the energy of a single photon. One application of the SPAD is depth imaging in LiDAR systems. The distance is determined by the time interval between the emitted light pulse and the received echo on the SPAD. Due to this characteristic, the SPAD provides accurate imaging with high photon efficiency. While the SPAD benefits from its high sensitivity, it leads to significant false detections due to ambient light, which degrades the reliability of the single measurement and limits the resistance to ambient light. Time-correlated single photon counting (TCSPC) is one way to deal with false detections.For example, a large number of time duration measurements are determined per block or per frame and collected in a histogram as a frequency distribution. However, in advanced driver assistance systems (ADAS), depth information must be available reliably and simultaneously in real time under changing ambient light. Therefore, the random factor caused by the high block rate or frame rate—i.e., a limited number of measurements per block—and the extent of the prevailing lighting situation must also be taken into account.

[0004] Generally, the digital processing used in LiDAR systems runs a digital filter through the TCSPC histogram and then captures the global maximum as an absolute prediction, for example, for the most likely value for the time of detection or the distance traveled by the desired light pulse (in strong ambient light, estimation and noise removal may be involved). To increase the reliability of the prediction, considerable effort has been invested in various processing stages of LiDAR systems: A fundamental approach is the use of optical bandpass filters. This approach removes most of the ambient light. However, choosing a narrow filter bandwidth range is difficult because the influence of manufacturing and temperature-related variations is difficult to estimate.

[0005] A second approach is to use a scanning LiDAR, which illuminates a point or line of a scene at a time. This approach increases the optical power intensity at the expense of the field of view. This approach makes it easier to distinguish the light pulse from the ambient light, but results in asynchronous detection and a reduction in the scene frame rate.

[0006] A common approach to suppressing ambient light is coincidence detection. This approach involves multiple detectors within a pixel. These detectors operate in parallel. When two or more detectors are triggered within a defined time interval, a coincidence event is generated. This approach prevents the SPAD from becoming saturated with strong ambient light. Since the emitted light pulse is limited by eye safety and emitter technology, an appropriate coincidence level should be carefully selected, which strongly depends on the ambient light intensity, the intensity of the received target light pulse, the target reflectance, and the target distance. Unfortunately, these factors always vary. Therefore, this approach usually works reliably in one or more specific scenarios.

[0007] Another approach is time-gating. This method uses fast electrical shutters to control the sensor's sensitivity to photons by limiting the time interval during which the detectors are activated for photon detection. Typically, this approach requires knowing approximately where the object is located.

[0008] For example, DE 102017220774 A1 describes a LiDAR method for distance measurement, according to which a first histogram is first recorded, based on which a first, rough time-of-flight determination is performed. Subsequently, a second time-of-flight determination is performed using a second histogram, wherein the second histogram comprises a temporal section of the first histogram and thus enables a finer time-of-flight determination than the first histogram.

[0009] Another approach introduces spatial analysis using various algorithms, such as deep learning, on the point cloud data. By exploiting the potential relationships between different measurement points, this approach can also improve noise resistance to a certain extent. However, reliable analysis requires a certain quality of the point cloud data, which is not always guaranteed by front-end processing.

[0010] DE 102018203533 A1 shows a receiving arrangement for receiving light signals, in particular for LiDAR applications, which carries out a time-correlated photon count at least twice, i.e. redundantly.

[0011] US 2016 / 0238695 A1 shows an example of a method for correcting environmental influences on spatially resolved signals, such as radar or LiDAR signals.

[0012] For example, embodiments of the present disclosure may be used in the context of TCSPC or LiDAR and may use the approaches described above.

[0013] What would be desirable is a concept for determining measurement information based on a large number of measured values, which offers an improved compromise between high reliability in detecting a useful signal in the large number of measured values, a small number of measured values required for this purpose, high accuracy of the determined information and low implementation effort.

[0014] Such a concept can be realized by means of the method and devices according to the independent claims. Embodiments and further advantageous aspects are mentioned in the respective dependent claims. overview

[0015] Embodiments of the present disclosure relate to a method for determining measurement information based on a plurality of measured values from a measured value range. The method includes obtaining a frequency distribution of a plurality of measured values (of the plurality of measured values), wherein the measured values of the frequency distribution are each assigned to a class of a plurality of classes of the frequency distribution, and wherein a frequency value of a class of the plurality of classes describes a number of measured values assigned to the class (e.g., absolute or relative with respect to a total number of the plurality of measured values of the frequency distribution). For example, one of the classes of the frequency distribution represents an interval from the measured value range around a position of the class in the measured value range, ie, for example, a position of a class represents the interval from the measured value range represented by the class.Thus, a width of the interval can indicate an accuracy of the position of a class. For example, the classes represent adjacent intervals of the measured value range. The method further includes dividing the frequency distribution into a plurality of ranges, for example, adjacent ranges, wherein each of the ranges represents an interval of the measured value range and contains one or more classes of the frequency distribution. The method further includes selecting a class of a respective range(s) as a selected class of the respective range based on a selection rule, wherein each of the ranges is assigned a range feature based on the selection rule.For example, the selection rule states that the class with the largest or smallest frequency value of a range is to be selected as the selected class of the range, and that the frequency value is to be assigned to the selected class as the range feature. In further examples, the selected class and the range feature are determined based on the classes of a respective range using a function. The method includes determining a probability value (or certainty value) for one of the selected classes based on the range features, wherein the probability value represents an estimate of the probability with which the selected class represents a value of a useful signal, wherein the determination of the probability value is based on machine learning (MML) methods, e.g., an artificial neural network (ANN).For example, a probability value can be determined for one, several, or all of the selected classes. For example, the useful signal is based on some of the measured values, so that the value of the useful signal is represented by the measured values underlying the useful signal.

[0016] Examples of the present disclosure are based on the idea that the measurement information can be determined particularly reliably based on the plurality of measured values using the MML. This means, for example, that measurement accuracy can be increased so that the determined measurement information represents the useful signal with a particularly high probability. This is achieved by allowing the MML to rely on empirical values, which have been learned, for example, from training data sets, to determine the probability value. Furthermore, the use of the MML makes it possible to achieve a high degree of robustness in determining the measurement information.For example, by using MML, measurement information can be reliably determined even when the total number of measured values in the frequency distribution is small, or the intensity of the desired signal is low compared to a background signal that may underlie some of the multitude of measured values. By dividing the frequency distribution into ranges, each of which is assigned a range feature, determining the probability values using MML can be particularly computationally inexpensive and requires little storage space. For example, because the number of ranges may be smaller than the number of classes in the frequency distribution, and thus a smaller amount of data needs to be processed using MML.A model underlying the MML also requires little storage space, as the method allows the MML to be based only on the range characteristics of a plurality of frequency distributions, and not, for example, on the majority of measured values of the frequency distributions. Furthermore, by selecting the selected class for a range and assigning the range characteristic to the range based on the selection rule, it is possible to assign a probability value determined based on the range characteristics to a selected class. This makes it possible to base the accuracy or resolution of the measurement information, for example, a value of the useful signal to be determined, on the accuracy or resolution of a selected class, even if a classification of the MML is less accurate than the resolution of the classes of the frequency distribution.The measurement information can therefore be determined more precisely than the resolution of the MML classification allows. Determining the probability value based on the area features of the areas obtained by division, each of which is assigned a selected class, thus allows the measurement information to be determined both computationally and accurately.

[0017] According to one embodiment, the frequency distribution is part of a series of frequency distributions of a respective plurality of measured values (the plurality of measured values), wherein the method further includes comparing the selected classes with the selected classes of a previous frequency distribution of the series of frequency distributions (wherein the selected classes of the previous frequency distribution are selected from corresponding ranges of the previous frequency distribution based on the selection rule) to adapt or maintain one or more of the range features depending on the comparison and providing the range features for determining the probability value.Thus, the range features used to determine the probability value may be identical to the range features assigned to the ranges based on the selection rule, or one or more or all of the range features may be adjusted depending on the comparison. In examples, the selected classes of the frequency distribution may also be compared with the selected classes of several previous frequency distributions in the series of frequency distributions.

[0018] Through the comparison, the previous frequency distribution can be used as prior knowledge to assess the probability with which the selected class represents the desired signal. For example, the comparison can be based on knowledge or an estimate of a change in the desired signal between the generation of the previous frequency distribution and the frequency distribution. By adjusting a range feature based on the comparison, the prior knowledge can be used to influence the probability value for the selected class accordingly. Using prior knowledge from one or more previous frequency distributions to potentially adjust the range features that serve as input data for determining the probability value can thus increase the accuracy or reliability of determining the probability values.Compared to using a single prior frequency distribution for comparison, using multiple prior frequency distributions for comparison can reduce measurement uncertainty, e.g., due to statistical changes in a background signal. Furthermore, additional information can be obtained, e.g., based on a temporal change in the wanted signal. This additional information can be provided as part of the measurement information or can also be used to further increase the reliability of determining the probability value using the MML.

[0019] According to one embodiment, comparing the selected classes with the selected classes of the previous frequency distribution of the series of frequency distributions includes comparing a position of one of the selected classes of the frequency distribution with the positions of one or more of the selected classes of the previous frequency distribution.

[0020] For example, the comparison can be based on checking where the position of the selected class is in relation to a position expected based on the position of the previously selected class, and adjusting or maintaining the range feature depending on this. For example, the type or strength of a possible adjustment can also be based on the comparison. Thus, based on a comparison, a measure of a correlation or coincidence between the selected classes and the frequency distribution and the previous frequency distribution can be determined, which can be used as a measure of a possible adjustment of the range feature. Thus, the range feature can be adjusted sensitively, for example, to a relative position of the selected class in relation to the selected class of the previous frequency distribution, whereby the adjustment can be carried out in such a way that the probability value can be determined particularly reliably.

[0021] According to one embodiment, comparing the selected classes with the selected classes of a previous frequency distribution of the series of frequency distributions includes selectively adjusting the range characteristic of the range of a selected class (i.e., the range to which the selected class belongs) taking into account a previous selected class if the previous selected class is in a correlation interval, the previous selected class being one of the selected classes of the previous frequency distribution.

[0022] For example, the range feature is adjusted if the previously selected class is within the correlation interval and not adjusted if the previously selected class is not within the correlation interval. Checking whether the previously selected class is within the correlation interval can be performed with minimal computational effort and can simultaneously be a reliable indicator of the probability value of the selected class.

[0023] According to one embodiment, comparing the selected classes with the selected classes of a previous frequency distribution of the series of frequency distributions includes determining a comparison position based on positions of each of one or more of the selected classes of a plurality of previous frequency distributions of the series of frequency distributions, and selectively adjusting the range feature of the range of one of the selected classes of the frequency distribution taking into account the selected classes used to determine the comparison position if the comparison position is located in a correlation interval.

[0024] By using multiple prior frequency distributions, the amount of prior knowledge can be increased, thus increasing the reliability of the correctness of a range coefficient adjustment. For example, based on the positions of one or more of the selected classes of multiple prior frequency distributions, a reliability of the comparison position can be determined, which can be considered for the comparison. This allows statistical uncertainties to be compensated.

[0025] According to one embodiment, adjusting one of the range features is based on an adjustment coefficient, wherein the adjustment coefficient is based on the probability value and / or the frequency value and / or a position of one or more selected classes of the prior frequency distribution.

[0026] For example, the adaptation coefficient represents a measure of the adaptation of the range feature. Thus, the degree of adaptation of the range feature can be adapted according to the probability value and / or the frequency value and / or a position of one or more selected classes of the previous frequency distribution, thereby increasing the reliability of the adaptation. For example, a high probability value or a high frequency value of a previously selected class can indicate that the previously selected class represents the value of the useful signal in the previous frequency distribution with a high probability, so that in combination with an expected change, a reliable prediction can be made for the probability value of a selected class of the frequency distribution, and the range feature can be adapted accordingly.Taking into account the position of a selected class of the previous frequency distribution makes it possible to weight a consideration of the selected class of the previous frequency distribution for the adjustment coefficient according to a relative position of the selected class of the previous frequency distribution with respect to a position of a selected class of the frequency distribution, and thus to increase the reliability of the adjustment of the range feature.

[0027] According to one embodiment, the adaptation of the range feature is based on an adaptation coefficient, wherein the adaptation coefficient and / or the correlation interval is based on the probability value and / or the frequency value and / or a position of the previously selected class being considered. In addition to the advantages described with respect to the adaptation coefficient, taking into account the probability value and / or the frequency value and / or the position of a previously selected class for the correlation interval allows the previously selected classes to be considered based on these parameters, thereby providing a low-computational way to adapt a measure for adapting the range feature based on these parameters.

[0028] According to one embodiment, comparing the selected classes with the selected classes of a previous frequency distribution of the series of frequency distributions includes determining a comparison position based on positions of one or more of the selected classes of a plurality of previous frequency distributions of the series of frequency distributions, and determining the adjustment coefficient based on the probability values and / or the frequency values and / or the positions of the selected classes used to determine the comparison position.

[0029] According to one embodiment, the adaptation of the range feature is based on an adaptation coefficient, wherein the adaptation coefficient and / or the correlation interval is based on the probability value and / or the frequency value and / or a position of the selected classes used to determine the comparison position.

[0030] According to one embodiment, the method further includes providing the adaptation coefficient as part of the measurement information. For example, a value of the adaptation coefficient may be an indicator of a change or a rate of change in the value of the useful signal, thereby expanding the information content of the measurement information.

[0031] According to one embodiment, a position of the correlation interval in the measured value range is based on a position of the selected class in the measured value range, and a width of the correlation interval is based on an expected change in the value of the useful signal. The expected change represents additional information, taking into account which can improve the effect of comparing the frequency distribution with the previous frequency distribution.

[0032] According to one embodiment, a position of the correlation interval in the measured value range is based on a position of the selected class in the measured value range and on an expected change in the value of the wanted signal, and furthermore, a width of the correlation interval is based on an expected change in the value of the wanted signal. By additionally using information about a change in the expected change, the effect of comparing the frequency distribution with the previous frequency distribution can be further improved.

[0033] According to one embodiment, the method further includes determining the correlation interval and / or the adjustment coefficient using an artificial neural network (wherein the artificial neural network may be part of an artificial neural network for determining the probability value or may be a separate network).

[0034] For example, the correlation interval and / or the adjustment coefficient can be determined based on a correlation interval and / or an adjustment coefficient of a previous frequency distribution or based on the range characteristics of the frequency distribution. This allows learned empirical values to be used to determine the correlation interval and / or the adjustment coefficient, and thus these parameters can be selected in such a way that the reliability of determining the probability value is increased.

[0035] According to one embodiment, the method further includes determining the class with the highest probability value from the selected classes as a useful signal class, and providing a position represented by the useful signal class in the measured value range as part of the measurement information. This selection of the useful signal class increases or maximizes the probability that the useful signal class represents the value of the useful signal. Thus, the method makes it possible to provide a probable value of the useful signal from the plurality of measured values as part of the measurement information.

[0036] According to one embodiment, the method further includes providing the probability value of the wanted signal class as part of the measurement information. Based on this information, for example, a reliability assessment can be made as to whether the wanted signal class represents the value of the wanted signal.

[0037] According to one embodiment, selecting the selected classes includes selecting the class of one of the ranges that has the highest frequency value as the selected class of the range, and wherein the range feature associated with the range is based on the frequency value of the selected class of the range. Selecting based on the highest frequency value represents a low-computational implementation of the selection rule and can still provide a high probability that the selected class of the range is the class of the range that has the highest probability of representing the wanted signal. Thus, the probability that one of the selected classes of the frequency distribution represents the wanted signal is increased.

[0038] According to one embodiment, the regions are equidistant and adjacent. By choosing adjacent regions, the frequency distribution can be mapped coherently without redundancies, and equidistant regions can be easily determined, allowing the method to be implemented with minimal computational effort.

[0039] According to one embodiment, a measured value of the plurality of measured values represents a time duration between an emission of a light pulse and a detection of a photon (or the first detection of a photon after the emission of the light pulse), wherein the measured value represents either a wanted signal value if the photon is based on the light pulse (e.g., based on an echo or a reflection of the light pulse), or a background signal value (e.g., if the measured value is based on detector noise or on a detection of ambient light), and wherein the wanted signal is based on one or more wanted signal values. For example, a class of the frequency distribution can include both wanted signal values and background signal values.Due to the way the measurement information is determined, the measurement information can be determined very reliably and accurately even when the number of useful signal values is very small compared to the number of background signal values. This makes it possible to reliably determine the measurement information even in strong ambient light and / or with a low light pulse intensity.

[0040] According to one embodiment, dividing the frequency distribution into the plurality of ranges involves selecting a width of one of the ranges such that the class of the range into which a measured value falling within the range falls with the highest probability represents a class of the wanted signal if the wanted signal falls within the range. The probability distribution for the classes can be based, for example, on average or expected rates of the wanted signal and a background signal, wherein a signal fluctuation range within which the rates of the wanted signal and the background signal can fluctuate can optionally also be taken into account.In combination with the selection of the class of a range with the highest probability value as the selected class of the range, this type of division of the frequency distribution ensures that the selected class of a range is most likely the class of the range with the highest probability of representing the desired signal. Given this criterion, this type of division simultaneously ensures that the frequency distribution is divided into as few ranges as possible, thus saving computing power and storage capacity.

[0041] According to one embodiment, the machine learning methods include an artificial neural network, and the method involves training the artificial neural network based on the area features associated with the areas of the frequency distribution. For example, training data sets for which the value of the useful signal is known can be used for training. Through the training, information about the area features can be used as additional empirical value for determining a probability value of a subsequent frequency distribution, so that the training can improve the reliability for determining the probability value.

[0042] According to one embodiment, the method further includes deconvolving the frequency distribution with one or more convolution kernels before selecting the selected class, for example before dividing the frequency distribution into the regions, wherein the convolution kernel describes, for example, an influence of a measuring device providing the plurality of measured values on the frequency distribution.

[0043] By deconvolution with the convolution kernel, the influence of measuring instruments on the measured values or the frequency distribution can be reduced or compensated, so that systematic errors can be reduced.

[0044] According to one embodiment, the plurality of measured values represent a series of measured values, and the method involves collecting a plurality of consecutive measured values of the series of measured values to obtain a frequency distribution of the series of frequency distributions. For example, the frequency distribution can be divided into ranges such that the ranges for the frequency distributions of the series of frequency distributions are identical, which is particularly advantageous for determining the probability values using the MML, since the input data for the MML are thus equally spaced.

[0045] According to one embodiment, the method includes emitting a light pulse by means of a light source, determining a time period between the emission of a light pulse and the detection of a photon by means of a detector, and providing the time period as one of the plurality of measured values. For example, the time period can be determined by means of a correlator based on a first signal emitted by the light source in connection with the emission of the light pulse and a second signal emitted by the detector as a result of the detection of the photon. Due to the manner in which the measurement information is determined, the light source can have a low intensity so that the light pulse, for example, does not pose a danger to an eye.Since the measurement information can be reliably determined even if the signal-to-noise ratio of the large number of measured values is small, a particularly sensitive detector, for example a SPAD, can be used, whereby a low intensity of the light source is sufficient to produce a sufficiently high rate of the useful signal at the detector.

[0046] According to one embodiment, the method includes obtaining a plurality of measured value series in parallel and determining a contribution to the measurement information based on respective frequency distributions of a respective plurality of measured values of the plurality of measured value series. The method includes determining the selected class with the highest probability value from the selected classes of the respective frequency distributions as a useful signal class and providing a position represented by the useful signal class in the measured value range as part of the respective contribution to the measurement information. For example, one of the measured value series is provided by a respective detector or a pixel of a detector array, so that image information can be obtained based on respective contributions to the measurement information.In examples, the respective contributions to the measurement information further include an adaptation coefficient and / or the probability value of the respective useful signal class. For example, spatial information can be obtained based on the measurement information. By combining the positions with probability values, the accuracy of the spatial information can be increased; for example, a distinction can be made between depth information and information about a surface texture (e.g., reflectivity) of an object. By combining the positions with the adaptation coefficients, motion information can be obtained, for example.

[0047] A further embodiment provides a device for determining measurement information based on a plurality of measured values from a measured value range. The device is designed to obtain a frequency distribution of a plurality of measured values, wherein the measured values of the frequency distribution are each assigned to a class of a plurality of classes of the frequency distribution, and wherein a frequency value of a class describes a number of measured values assigned to the class. Furthermore, the device is designed to divide the frequency distribution into a plurality of ranges, wherein each of the ranges represents an interval of the measured value range and contains one or more classes of the frequency distribution.The device is designed to select a class of a respective area as a selected class of the respective area based on a selection rule, wherein a range feature is each assigned to the areas based on the selection rule. The device is further designed to determine a probability value for one of the selected classes based on the range features, wherein the probability value represents an estimate of the probability with which the selected class represents a value of a useful signal, wherein the determination of the probability value is based on a statistical model.

[0048] A further embodiment provides a LiDAR device comprising the device for determining measurement information. Furthermore, the LiDAR device comprises a light source configured to emit a light pulse and, in conjunction with the emission of the light pulse (e.g., in a temporal context, e.g.,simultaneously or subsequently) to provide a first signal, a detector which is designed to detect a photon and to provide a second signal following detection of a photon, and a correlator which is designed to determine, based on the first signal and the second signal, a time period between the emission of the light pulse and the detection of the photon, and to provide the time period as a measured value of the plurality of measured values, wherein the measured value represents a useful signal value when the photon is based on an echo of the light pulse, wherein the useful signal is based on one or more useful signal values, and wherein the measurement information includes a position of the selected class of the frequency distribution with the highest probability value.The position of the selected class can therefore represent, with the certain probability, a time duration for the echo or a distance of the LiDAR device to an object on which the echo is based.

[0049] According to one embodiment, the range feature of the range of a selected class is selectively adjusted taking into account a previously selected class if the previously selected class is located in a correlation interval, wherein the previously selected class is one of the selected classes of the previous frequency distribution, and wherein a position of the correlation interval in the measured value range and / or a width of the correlation interval is based on an expected change in the value of the wanted signal, and wherein the expected change is based on a speed and / or an acceleration of the LiDAR device. For example, the LiDAR device is arranged on a movable device, e.g., a vehicle, whose speed or acceleration is provided to the LiDAR device as an input signal.Alternatively, the LiDAR device may be able to determine or estimate the velocity and / or acceleration based on the series of frequency distributions. Considering the velocity and / or acceleration, it is possible to accurately assess the probability that a selected class represents the distance of an object, since the previous frequency distribution can serve as an estimate of the object's location at an earlier time. Thus, a range feature can be adjusted to incorporate this prior knowledge into determining the probability value.

[0050] According to one embodiment, the LiDAR device has a plurality of detector units (e.g., a detector array), wherein the LiDAR device is designed to obtain a plurality of measured values using a respective detector unit, wherein the device for determining the measurement information is designed to determine, based on the respective plurality of measured values, a contribution to the measurement information assigned to the respective detector unit.

[0051] The device and the LiDAR device are based on the same considerations as the method explained above. Furthermore, it should be noted that the device can be supplemented with all the features, functionalities, and details described herein with regard to the method according to the invention for determining measurement information. The device can be supplemented with the aforementioned features, functionalities, and details both individually and in combination.

[0052] A further embodiment provides a computer program with a program code for carrying out one of the previously explained methods when the program runs on a computer. Short description of the characters

[0053] Examples of the disclosure are described below with reference to the accompanying figures. They show: Fig. 1 a flowchart of a method for determining measurement information according to an embodiment, Fig. 2 a schematic diagram of an example of a frequency distribution, Fig. 3A a diagram of another example of a frequency distribution, Fig. 3B is a diagram showing an example of a frequency distribution after applying a convolution kernel, Fig. 3C a diagram with an example of selected class and range features of a frequency distribution, Fig. 3D a diagram with an example of probability values of selected classes, Fig. 4 a block diagram of a method for a recording-correlated learning method according to an embodiment, Fig. 5 diagrams illustrating a time-correlated analysis according to an embodiment, Fig. 6 shows a representation of a comparison of a frequency distribution with a previous frequency distribution according to an embodiment, Fig. 7 a representation of an example of determining a correlation interval, Fig. 8 a representation of another example of determining a correlation interval, Fig. 9 an example of a probability density function of the first incoming photon, Fig. 10 an example of spatial depth information and certainty information, Fig. 11 an example of a contribution to the measurement information based on the adjustment coefficient, Fig. 12 a schematic representation of a device for determining measurement information according to an embodiment, Fig. 13 a schematic representation of a LiDAR device according to an embodiment, Fig. 14 several diagrams with examples of previous selected classes. Detailed description

[0054] Examples of the present disclosure are described in detail below using the accompanying descriptions. Many details are described in the following description to provide a more thorough explanation of examples of the disclosure. However, it will be apparent to those skilled in the art that other examples may be implemented without these specific details. Features of the various described examples may be combined with one another unless features of a corresponding combination are mutually exclusive or such a combination is expressly excluded.

[0055] It should be noted that identical or similar elements, or elements having the same functionality, may be provided with identical or similar reference symbols or be designated alike. Repeated descriptions of elements having identical or similar reference symbols or being designated alike are typically omitted. Descriptions of elements having identical or similar reference symbols or being designated alike are interchangeable.

[0056] Fig. 1 shows a method 100 for determining measurement information based on a plurality of measured values from a measured value range according to an exemplary embodiment. The method comprises the following steps: Obtaining 120 a frequency distribution of a plurality of measured values, wherein the measured values of the frequency distribution are each assigned to a class of a plurality of classes of the frequency distribution, and wherein a frequency value of a class describes a number of measured values assigned to the class; Dividing 140 the frequency distribution into a plurality of ranges, wherein each of the ranges represents an interval of the measured value range and contains one or more classes of the frequency distribution; Selecting 160 a respective class of a respective range as a selected class of the respective range based on a selection rule, wherein a range feature is each assigned to the ranges based on the selection rule;and determining 180 a probability value for one of the selected classes based on the range features, wherein the probability value represents an estimate of the probability that the selected class represents a value of a useful signal, wherein the determination of the probability value is based on machine learning methods;

[0057] Fig. 2 shows a schematic histogram of an example of a frequency distribution 222 comprising a plurality of measured values 221. In examples, the frequency distribution 222 is a result of step 120. A value of the respective measured values 221 is indicated by the abscissa. The values of the measured values are located in a measured value range 223 comprising a plurality 228 of classes, for example, class 226, wherein a class represents an interval in the measured value range 223. The class 226 is located at a position 230 representing a value from the measured value range 223. A number of measured values falling into class 226 are represented by a frequency value 224, which is indicated on the ordinate of the frequency distribution 222.

[0058] Fig. Figure 3A shows a histogram of another example of the frequency distribution 222. In this example, the measured values represent distance values, which are plotted along the abscissa, and the ordinate, which indicates the frequency values of the classes, is plotted as the absolute number of measured values. In this example, the measured value range is 0 m to 100 m.

[0059] According to one embodiment, the method 100 includes an optional step 130, which includes deconvolving the frequency distribution 222 with one or more convolution kernels. The step 130 may, for example, be performed before selecting 160 the selected class, or, as with respect to the Fig. 3A to Fig. 3D shown, before dividing 140.

[0060] For example, the convolution kernel can contain information about the measurement system or act similarly to a mean filter.

[0061] In further examples, noise can be removed by applying an additional filter, for example based on an estimated intensity of a background signal. This noise suppression can occur before selection 160. This can prevent incorrect selection of a class as the selected class, for example if the selection rule selects the selected class according to the maximum of the frequency values of the range. Reliable selection of the selected classes of the ranges can increase the probability that a useful signal class determined from the selected classes particularly closely approximates the value of the useful signal. Alternatively, noise suppression can also occur after selection 160.

[0062] Fig. 3B shows a histogram showing an example of a result of optionally applying 130 a convolution kernel to the Fig. 3A. In the example shown, the classes 228 of the frequency distribution 222 remain unchanged by the application 130 of the convolution kernel, whereby the frequency values of one or more of the classes 228 have been changed. Fig. Figure 3B is an example of a representation of the frequency values as a normalized number.

[0063] Fig. 3C shows a diagram 362 illustrating an example of sharing 140 of the Fig. 3A or Fig. 3B into regions 344 and the selection 160 of selected classes 366 for the regions 344. In the example shown, the regions 344 are equidistant and adjacent, but the regions 344 can also be selected differently, regardless of other features shown. For example, a class 326 is the selected class of the region 345 of the plurality of regions 344. Alternatively, the division 140 can also be applied directly to the Fig. 3A can be applied. The ordinate of the diagram 362 indicates a value for range features 364 associated with the selected classes 366.

[0064] In examples such as the one in Fig. 3C, selecting 140 the selected classes 366 includes selecting the class 326 of one of the ranges 345 as the selected class of the range 345 that has the largest frequency value 324, and wherein the range feature 364 associated with the range is based on the frequency value 324 of the selected class 326 of the range 345. For example, the frequency value 324 of the exemplary class 326 (see Fig. 3A, Fig. 3B, Fig. 3C) represents the maximum of the frequency values of all classes of the range 345, and the range feature 367 of the selected class 366 corresponds to the frequency value 324.

[0065] Fig. 3D shows a diagram 382 illustrating an example of determining 180 probability values 384 for the selected classes 366 based on the Fig. 3C. The ordinate of the diagram 382 indicates the value of the respective probability value. For example, the probability value 385 of the exemplary selected class 326 of the area 345 represents the probability or certainty with which the position 231 of the selected class 326 corresponds to a distance indicated by a useful signal.

[0066] In examples, the method 100 further includes determining from the selected classes 366 the one with the highest probability value 384,387 as a payload class 386, and providing a position 388 represented by the payload class 386 in the measurement value range 223 as part of the measurement information.

[0067] In examples, the method 100 further includes providing the probability value 387 of the payload class 386 as part of the measurement information.

[0068] In the Fig. In the example shown in Figure 3D, the probability values 384 were determined using a neural network, but alternatively, other MMLs that can be learned using training data and that are suitable, for example, for classifying the area features 364 can be used.

[0069] In examples, the probability value 384 of the selected class 366 is further determined at the position 230 of the selected class 366. For example, the position 230 may be provided along with the range feature 364 of the selected class 366 to determine the probability value 384.

[0070] In examples, the machine learning methods include an artificial neural network, and the method 100 includes training the artificial neural network based on the region features 364 associated with the regions 344 of the frequency distribution 222.

[0071] In examples, the frequency distribution 222 is part of a series of frequency distributions of a respective plurality of measured values. For example, a piece of measurement information can be determined using the consecutive frequency distribution method 100.

[0072] For example, the plurality of measurements 221 represents a series of measurements 221, and the method includes collecting a plurality of consecutive measurements 221 of the series of measurements 221 to obtain a frequency distribution of the series of frequency distributions.

[0073] Fig. 6 shows a diagram 662 with a further example of areas 344 each with a selected class 366 and associated area features 364, similar to diagram 362 of Fig. 3C, where the underlying frequency distribution is part of a series of frequency distributions. Furthermore, Fig. 6 shows a further diagram 672 with an example of selected classes 676 and probability values 674 assigned to them, which are determined based on a previous or earlier frequency distribution of the series of frequency distributions.

[0074] In examples, the method 100 includes comparing the selected classes 366 with the selected classes 676 of a previous frequency distribution of the series of frequency distributions to adjust or maintain one or more of the range features 364 depending on the comparison, and providing the range features 364 for determining the probability value 384, 385, 387. A selected class 676 of a previous frequency distribution is also referred to below as a previous selected class 676, wherein a selected class 366 of the frequency distribution 222 may also be referred to as a current frequency distribution.

[0075] For example, comparing the selected classes 366 with the selected classes 676 of the previous frequency distribution of the series of frequency distributions involves comparing a position 230 of one of the selected classes 366 of the frequency distribution with the positions 230 of one or more of the selected classes 676 of the previous frequency distribution.

[0076] In Fig. 6 shows an example of a selected class 666, which is selected for illustration from the selected classes 366. The selected class 666 is assigned the area feature 664. Furthermore, Fig. 6 shows an example of a correlation interval 677 that can be used to compare the position 230 of the currently selected class 666 with the positions of the previously selected classes 676. The correlation interval can include the position 230 of the selected class 666, as in Fig. 6 shown, or not included (cf. Fig. 7, Fig. 8).

[0077] In examples, comparing the selected classes 366 with the selected classes 676 of a prior frequency distribution of the series of frequency distributions includes selectively adjusting the range feature 664 of the range of a selected class 666 considering a prior selected class 676 if the prior selected class 676 is in a correlation interval 677, wherein the prior selected class 676 is one of the selected classes of the prior frequency distribution.

[0078] In examples, several previous frequency distributions of the series of frequency distributions are considered for comparison, whereby a selected class of one of the previous frequency distributions is understood as a previously selected class. For example, the influence of a previously selected class can be weighted depending on how far back the corresponding frequency distribution lies in the series of frequency distributions compared to the current frequency distribution 222.

[0079] It should also be noted that the comparison of the positions 230 of the current selected classes 366 with the positions 630 of the previous selected classes 676 can be carried out by considering a relative position, i.e. a distance, which is why it can be equivalent to check whether the position 230 of a current selected class 366 lies in a correlation interval 677 whose position was chosen with reference to the position 630 of a previous selected frequency distribution, or whether the position 630 of a previous selected class 676 lies in a correlation interval 677 whose position was chosen with reference to the position 230 of a previous selected frequency distribution.

[0080] In examples, comparing the selected classes 366 with the selected classes 676 of a previous frequency distribution of the series of frequency distributions includes determining a comparison position based on positions 630 of one or more of the selected classes 676 of a plurality of previous frequency distributions of the series of frequency distributions, and selectively adjusting the range feature 664 of the range of one of the selected classes 666 of the frequency distribution, taking into account the selected classes used to determine the comparison position, if the comparison position is within a correlation interval 677. For example, a comparison position may be determined for each selected class 366.Alternatively, for example, based on the previously selected classes 676, one or more comparison positions may be determined, which are compared, for example, with the position 230 of a selected class 666 to selectively determine an adjustment of the range feature 664 of the selected class 666. In other words, the number of comparison positions may be the same as or different from the number of selected classes 366. Determining a comparison position provides better filtering of statistical uncertainties than comparing the position 230 with the individual positions 630 of the previously selected classes.

[0081] For example, to determine the comparison position, a weighted mean and / or a standard deviation may be determined from the positions 630 of several previously selected classes 676 of different previous frequency distributions, wherein the weighting may be based, for example, on the positions 630, the frequency values 224, and / or the probability values 676 of the previously selected classes 676. In examples, the correlation interval is further determined based on the positions 630 (e.g., a weighted mean and / or a standard deviation), the frequency values 224, and / or the probability values 676 of the previously selected classes 676.

[0082] In examples, adjusting one of the range features 364 is based on an adjustment coefficient, where the adjustment coefficient is based on the probability value 674 and / or the frequency value 224 and / or a position 630 of one or more selected classes 676 of the prior frequency distribution.

[0083] For example, the adjustment coefficient is a value added to or subtracted from the range characteristic, or a factor by which the range characteristic is multiplied. When adjusting a range characteristic, the value of the range characteristic can also be taken into account.

[0084] In examples, adjusting the range feature 664 is based on an adjustment coefficient, where the adjustment coefficient and / or the correlation interval 677 is based on the probability value and / or the frequency value and / or a position 630 of the prior selected class being considered.

[0085] In examples, comparing the selected classes 366 with the selected classes 676 of a previous frequency distribution of the series of frequency distributions includes determining a comparison position based on positions 630 of each of one or more of the selected classes of a plurality of previous frequency distributions of the series of frequency distributions, and determining the adjustment coefficient based on the probability values and / or the frequency values and / or the positions of the selected classes used to determine the comparison position.

[0086] In examples, adjusting the range feature 664 is based on an adjustment coefficient, where the adjustment coefficient and / or the correlation interval 677 is based on the probability value and / or the frequency value and / or a position of the selected classes used to determine the comparison position.

[0087] Fig. Figure 14 shows examples of histograms 672-1, 672-2, 672-3 with previously selected classes 676-1, 676-2, 676-3 of several previous frequency distributions of the series of frequency distributions. The previously selected classes 676-1, 676-2, 676-3 have positions 630-1, 630-2, 630-3 and associated frequency values or probability values 674-1, 674-2, 674-3, and may correspond to the previously selected class 676 with position 630 and probability value 674.

[0088] Although in Fig. 14 the measured value range represents a distance, the measured value range can also represent other quantities, e.g. a time. This applies equivalently to Fig. 7, Fig. 8, Fig. 5 and Fig. 3A-D.

[0089] For example, a mean and / or a standard deviation of the positions 630 of several previously selected classes 676 is taken into account for the adjustment coefficient and / or the correlation interval. For example, the standard deviation of several positions 630 of previously selected classes 676 may be small if the previously selected classes 676 represent wanted signal classes, and large if the previously selected classes 676 represent a background signal, so that the adjustment coefficient can be adjusted accordingly.

[0090] In other words, the comparison position, the correlation interval, and / or the adjustment coefficient may be influenced by information from previous frequency distributions, for example, by positions, probability values, frequency values, and / or adjustment coefficients of previously selected classes. That is, the relationship between the comparison position, adjustment coefficient, and correlation interval may be parallel. The comparison position, adjustment coefficient, and correlation interval may be based on information from the plurality of previous frequency distributions.

[0091] Fig. 7 and Fig. 8 each illustrate an example of determining the correlation interval 677. In the diagrams shown, the measured value range 223 is plotted along the abscissa, the ordinate shows, for example, a value of the frequency value 224, 324 or, as shown, a probability, or another value assigned to a selected class.

[0092] In examples such as Fig. 7, a position 678 of the correlation interval 677 in the measured value range 223 is based on a position 630 of the selected class 676 in the measured value range 223, and a width 679 of the correlation interval 677 is based on an expected change in the value of the useful signal.

[0093] In further examples, as in Fig. 8, wherein a position 678 of the correlation interval 677 in the measured value range 223 is based on a position 630 of the selected class 676 in the measured value range 223 and on an expected change in the value of the useful signal, and wherein a width 679 of the correlation interval 677 is based on the expected change in the value of the useful signal.

[0094] The consideration of an expected change in the useful signal can also be combined with the consideration of several previous frequency distributions and can further be combined with a determination of the correlation interval based on the positions 630, the frequency values 224, the adjustment coefficients and / or the probability values 676 of the previously selected classes 676.

[0095] In examples, the method 100 further includes determining the correlation interval and / or the adjustment coefficient using an artificial neural network.

[0096] According to one embodiment, the method 100 further includes emitting a light pulse by means of a light source, determining a time period between emitting a light pulse and detecting a photon by means of a detector, and providing the time period as a measured value 221 of the plurality of measured values.

[0097] In examples, a measurement value 221 of the plurality of measurement values 221 represents a time period between an emission of a light pulse and a detection of a photon, wherein the measurement value 221 represents either a wanted signal value if the photon is based on the light pulse or a background signal value, and wherein the wanted signal is based on one or more wanted signal values.

[0098] For example, in a TCSPC recording, a plurality of measured values of the time duration can be determined in order to obtain the frequency distribution 222, wherein one or more classes 226 of the frequency distribution 222 can contain useful signal values.

[0099] Fig. Figure 9 shows an example of a probability density function of the first incoming photon, for example, the first photon after the light pulse has been emitted. After the first photon has arrived, a measurement of the time duration can be terminated, and the determination of a further time duration can be started with the emission of another light pulse. The probability with which a photon arrives at a specific time after the emission of a light pulse can be determined from the rates of the background signal and the light pulse, and further from the time of flight (TOF) of the light pulse, which in turn can be determined from the distance of an object at which the light pulse is reflected. Fig. Figure 9 shows an example of the probability density function for the arrival of a photon. The diagram shows a region 929. Classes 928 represent a useful signal, with class 925 having the highest probability 949 of all classes in region 929.

[0100] In examples, dividing 140 the frequency distribution into the plurality of ranges 344 includes selecting a width of one of the ranges 929 such that the class of the range into which a measured value falling within the range 929 falls with the highest probability represents a class of the useful signal if the useful signal falls within the range.

[0101] The probability density function for the first incoming photon for constant ambient light follows the function F1: F1=rBe−rBt, where r B is the background photon rate. Taking into account the emitted laser pulse, the overall function F is: F={rBe−rBtfor 0≤t <TTOFrLBe−rLB(t−TTOF)e−rBTTOF for TTOF≤t<TTOF+TprBe−rBte−rLTPfor TTOF+TP≤t

[0102] Here are r B , r L and r LB for background, emitted laser, or total photon rate on the receiver side. T TOF is the arrival time of the first photon of the emitted light. T p is the width of the emitted light pulse. To fulfill the principle that the selected local maxima (e.g., the selected classes 366 of a range 344) have a high or the highest probability of containing the target information, the probability of the target class group should always be the maximum. Considering the worst case, ie, the laser pulse returns at the very end of the range, and at a high or maximum background photon rate, as in Fig. 9, it follows that: 1−e−rBTpe−rBTTOF(1−e−rLBTp)<1 Assuming that the maximum background photon rate and the received laser photon rate are both 10 MHz, the width of the leftmost region can be derived as: TTOF+Tp<6.69∗10−8s+5∗10−9s=7.19∗10−8s

[0103] That means: Width range <10.78 m

[0104] This is a simplified calculation. In practice, more factors such as distance attenuation (e.g., a decrease in the reflected power of the light pulse with increasing TOF) or the number of measured values of the frequency distribution should be considered.

[0105] Fig. 4 shows a method model of a method 400 for a recording-correlated learning method for TCSPC histograms at the pixel level according to an embodiment. The method 400 may be similar to the method 100 of Fig. 1. The method 400 includes feature extraction 460, which corresponds, for example, to steps 140 and 160 of Fig. 1, training and time-correlated analysis 470. The method 400 includes obtaining time-correlated information, e.g. in the form of a histogram, which, for example, corresponds to the frequency distribution 222 from Fig. 2, Fig. 3. The method includes a prediction 480, which may, for example, correspond to step 180. The output information 490 includes depth information and optionally certainty information, which may, for example, correspond to the position and probability value of the payload. In the following, steps of the method 400 are explained using the detailed example of time-correlated single-photon counting (TCSPC). However, this is not the only way to implement the method 400: - Data sampling: Creating raw TCSPC histograms. Histograms are collected at different intervals and under different ambient light intensities and used as training data, for example, for training the MML or ANN. - Feature extraction 460 according to the process as in the Fig. 3A-D. A filter, e.g., a convolution kernel, is applied to the raw data of the histogram, e.g., the frequency distribution 222, to form a new histogram, e.g., as shown in Fig. 3B, which aims to highlight the groups of classes that are similar to the target shape, e.g., the shape of an output light pulse. In the example shown, a one-dimensional convolution kernel is used as a filter. A template is set according to the ideal pulse shape and pulse width. If the pulse width of the laser is narrow enough, the layer can be discarded. Next, the new histogram is divided into several regions 344. Specific local features (e.g., region features 364), e.g., local maxima, are selected from each of the regions 344. The width of a region is defined, for example, by two principles (a specific example is given in connection with Fig. 9): 1) The selected local maxima have a high or the highest probability of containing the target information (e.g., the useful signal); 2) the total number of local maxima (e.g., the total number of regions 344 into which the frequency distribution 222 is divided) is as small as possible. The selected local maxima and the correlated positions represent the entire histogram. After feature extraction, other histogram information can be released from memory. - Prediction 480: A classifier, e.g., a feedforward neural network, is trained through supervised learning using the local maxima (e.g., the range features 364). The classifier outputs each represent the certainties (which may, e.g., correspond to the probability values 384) of the correlated local maxima. After training, the weights can be frozen or fine-tuned. The final prediction may include depth information and its certainties. - Shot-correlated analysis 470: The previous certainty information (e.g., probability values 384) is used as feedback to support the analysis of the current shot. Since the target cannot suddenly disappear, more attention should be paid to the coincidence positions of potential peaks in different shots. One way to realize the analysis is to calculate the information gain (which can correspond, for example, to the adjustment coefficient). It should be noted that situations that cause a sudden disappearance of the previous object (e.g., when the object moves out of the field of view or the near object blocks the distant object) have little negative impact because the analysis is based on the histogram of the current shot.

[0106] In examples, information gain includes two aspects: gain coefficient and gain width. The gain coefficient is proportional to the certainty of the local feature and, when multiple recordings (e.g., multiple previous frequency distributions) are involved, is inversely proportional to the standard deviation of class positions (e.g., positions of selected classes) from different recordings in a region. The gain width, for example, is proportional to the dynamic degree of the scenario (e.g., the radial velocity of the camera, the variance rate of the acquired distance history). The information gain can be used internally in the method 100, 400.

[0107] Fig. 5 illustrates an example of a time-correlated analysis, which may, for example, correspond to the uptake-correlated analysis 470 or to comparing the selected classes 366 with the selected classes of a previous frequency distribution of the series of frequency distributions, or an example of an implementation of information gain. The middle row of diagrams in Fig. 5 shows a histogram 222 with raw data (e.g. measured values), a diagram 562 with area features and a diagram 582 with probability values, which, for example, correspond to the Fig. 3A, Fig. 3C, and Fig. 3D shown diagrams 222, 362, 382, and which according to the description of the Fig. 3A to 3D. The histogram 222 represents a current recording, e.g., a current plurality of measured values 221, or a current TCSPC histogram, based on which a measurement information, e.g., depth information, is to be determined. The first row of diagrams in Fig. 5 shows a histogram 512 with raw data, similar to the histogram 222. The raw data of the histogram 512 (e.g., the majority of measured values) are part of a previous or last recording, e.g., an earlier plurality of measured data that were analyzed, for example, at an earlier point in time using the method 100, 400. For example, the histogram 512 contains a previous frequency distribution 513. Diagram 516 shows the selected classes 517 with the associated range features of the previous frequency distribution 513. The lower row of diagrams shows a time-correlated analysis of the range features of the diagram 512. In diagram 518 with probability values of the selected classes 517 of the previous frequency distribution 513, a correlation interval 677, 677' is shown for two selected classes 526, 527 as an example.In the diagram 562 with the selected classes 366 of the frequency distribution 222, a selected class 528 lies in the correlation interval 677, while none of the selected classes 366 of the frequency distribution 222 lies in the correlation interval 677'. Therefore, a gain amplitude 581 determined for the selected class 528, which may correspond, for example, to the adjustment coefficient, is greater than zero, so that an amplitude, e.g., the range characteristic, of the selected class 528 is increased.

[0108] As described with respect to the prediction 480, by learning the relationship between multiple potential target peaks, the method 100, 400 has the potential to operate in a wider range of ambient light intensity than digital processing. Furthermore, due to the aforementioned shot-correlated analysis, the measurement reliability is higher than when analyzing a single shot and requires only a small amount of additional storage space to save previous local maxima and indices, which may correspond to the area features 364 and the positions 230 of the selected classes 366. Considering various desired LiDAR applications, this may result in a smaller number of required measurements per shot, a wider measurement range, less sensitivity to ambient light, or fewer requirements for the light emitter. The depth resolution is retained in the output, e.g.Because the positions 230 of the selected classes are stored, the position 230 of a selected class identified as a useful signal can be used as depth information. The final outputs can include multiple types of information, such as the depth map and the certainty map. Spatial analysis can benefit from these maps. Furthermore, since current technologies such as coincidence detection and time gating also generate similar histograms, combining the method and these technologies can further increase the robustness of the system.

[0109] Instead of a feedforward network, another algorithm can be used, which is based on the same working principle of the method.

[0110] Different technologies can be used as the filter to support feature extraction, or without the filter if the raw data quality is sufficient for preprocessing, e.g., for selecting the selected classes.

[0111] Other algorithms for uptake-correlated analysis can be implemented, or another machine learning algorithm can be used that includes both training and uptake-correlated analysis.

[0112] In examples, the method 100, 400 processes the local maxima from a TCSPC histogram, analyzes the coincidence information on different frames of TCSPC histograms, or provides a certainty or probability of the measurement by using algorithms on TCSPC histograms.

[0113] In examples, the method 100 is applied to a LiDAR system or for applications that benefit from prediction, certainty of prediction, and variance of potential information from different blocks.

[0114] In examples, the method 100, 400 includes obtaining a plurality of measured value series in parallel and, based on respective frequency distributions of a respective plurality of measured values of the plurality of measured value series, determining a respective contribution to the measurement information. The method includes determining the selected class with the highest probability value from the selected classes of the respective frequency distributions as a useful signal class and providing a position represented by the useful signal class in the measured value range as part of the respective contribution to the measurement information. In other words, the method 100, 400 can be applied in parallel to one another to a plurality of measured values that are generated, for example, based on signals from a plurality of detectors, e.g., a plurality of pixels of a detector array.In this way, image information can be obtained, for example, in the case where the useful signal represents a travel time or depth information, spatial depth information can be obtained.

[0115] Fig. 10 shows an example of measurement information 1094 that includes spatial depth information 1091 with a plurality of contributions. The depth information is based, for example, on the positions of useful signal classes determined by the method 100, 400 based on a frequency distribution 222. Furthermore, Fig. 10 shows an example of measurement information 1095 containing the spatial depth information and additionally certainty information 1092, for example, the respective probability values of the useful signal classes underlying the depth information 1091. For example, the measurement information 1094 can be interpreted such that an object underlying the distance values has a rough surface. Since the measurement information 1095 additionally contains the certainty information 1092, the measurement information 1095 suggests the interpretation that the object has a flat surface with different surface reflectivities.

[0116] In examples, the method 100 includes providing the adaptation coefficient as part of the measurement information. This allows, for example, a certainty to be assessed as to whether a class selected as a wanted signal class represents a true value of the wanted signal.

[0117] Fig. Figure 11 shows an example of an application of the adaptation coefficient as contribution 1096 to the measurement information. Contribution 1096 contains information based on the adaptation coefficients of the useful signal classes underlying the depth information 1091. For example, contribution 1096 allows conclusions to be drawn about the movement of an object detected based on the depth information 1091.

[0118] Fig. 12 shows a device 1200 for determining measurement information 1290 based on a plurality of measured values 221 from a measured value range 223 according to an exemplary embodiment. The device 1200 is designed to: obtain a frequency distribution 222 of a plurality of measured values 221, wherein the measured values 221 of the frequency distribution 222 are each assigned to a class of a plurality 228 of classes of the frequency distribution 222, and wherein a frequency value 224 of a class 226 describes a number of the measured values assigned to the class 226; to divide the frequency distribution 222 into a plurality of ranges 344, wherein each of the ranges 344 represents an interval of the measured value range 223 and contains one or more classes 226; 326 of the frequency distribution;to select a class of a respective area 344 as a selected class 366 of the respective area 344 based on a selection rule, wherein an area feature 364 is each assigned to the areas 344 based on the selection rule; and to determine a probability value 384; 387 for one of the selected classes 366 based on the area features 364, wherein the probability value 384; 387 represents an estimate of the probability with which the selected class represents a value of a useful signal, wherein the determination of the probability value is based on a statistical model;

[0119] Fig.13 shows a LiDAR device 1300 according to an embodiment. The LiDAR device 1300 has the device 1200, for example as a signal processing unit, and further includes: a light source 1302, which is designed to emit a light pulse 1303 and, in connection with the emission of the light pulse 1303, to provide a first signal 1304, e.g., an electrical signal; a detector 1305, which is designed to detect a photon 1306 and, following detection of a photon, to provide a second signal 1307, e.g., an electrical signal; a correlator 1308, e.g.,TCSPC electronics configured to determine, based on the first signal 1304 and the second signal 1307, a time period 1309 between the emission of the light pulse and the detection of the photon, and to provide the time period 1309 as a measured value of the plurality of measured values, wherein the measured value represents a useful signal value when the photon 1306 is based on an echo of the light pulse 1303, wherein the useful signal is based on one or more useful signal values. Furthermore, the measurement information 1290 may include a position 230 of the selected class 344 of the frequency distribution 222 with the highest probability value.

[0120] In examples, the range feature 364 of the range 344 of a selected class 366 is selectively adjusted taking into account a previous selected class if the previous selected class is in a correlation interval, wherein the previous selected class is one of the selected classes of the previous frequency distribution, and wherein a position of the correlation interval in the measurement range and / or a width of the correlation interval is based on an expected change in the value of the wanted signal, and wherein the expected change is based on a speed and / or an acceleration of the LiDAR device 1300.

[0121] In examples, the LiDAR device 1300 comprises a plurality of detector units, wherein the LiDAR device 1300 is configured to obtain a plurality of measured values using a respective detector unit, wherein the device 1200 for determining the measurement information is configured to determine a contribution to the measurement information 1290 associated with the respective detector unit based on the respective plurality of measured values.

[0122] Although some aspects of the present disclosure have been described as features associated with a device, it is clear that such a description may also be considered a description of corresponding method features. Although some aspects have been described as features associated with a method, it is clear that such a description may also be considered a description of corresponding features of a device or the functionality of a device.

[0123] Some or all of the method steps may be performed by (or using) a hardware device, such as a microprocessor, a programmable computer, or an electronic circuit. In some embodiments, some or more of the essential method steps may be performed by such a device.

[0124] Depending on specific implementation requirements, embodiments of the disclosure may be implemented in hardware or in software, or at least partially in hardware or at least partially in software. The implementation may be performed using a digital storage medium, for example a floppy disk, a DVD, a Blu-ray disc, a CD, a ROM, a PROM, an EPROM, an EEPROM, or a FLASH memory, a hard disk, or other magnetic or optical storage device storing electronically readable control signals that can interact or interact with a programmable computer system to perform the respective method. Therefore, the digital storage medium may be computer-readable.

[0125] Some embodiments according to the disclosure thus comprise a data carrier having electronically readable control signals capable of interacting with a programmable computer system such that one of the methods described herein is carried out.

[0126] In general, embodiments of the present disclosure may be implemented as a computer program product having program code, wherein the program code is operable to perform one of the methods when the computer program product is run on a computer.

[0127] The program code can, for example, also be stored on a machine-readable medium.

[0128] Other embodiments include the computer program for performing one of the methods described herein, wherein the computer program is stored on a machine-readable medium. In other words, one embodiment of the method according to the disclosure is thus a computer program that has program code for performing one of the methods described herein when the computer program is executed on a computer.

[0129] A further embodiment of the methods according to the disclosure is thus a data carrier (or a digital storage medium or a computer-readable medium) on which the computer program for performing one of the methods described herein is recorded. The data carrier or the digital storage medium or the computer-readable medium is typically tangible and / or non-transitory.

[0130] A further embodiment of the method according to the disclosure is thus a data stream or a sequence of signals that represents the computer program for performing one of the methods described herein. The data stream or the sequence of signals can be configured, for example, to be transferred via a data communication connection, for example, via the Internet.

[0131] A further embodiment comprises a processing device, for example a computer or a programmable logic device, which is configured or adapted to perform one of the methods described herein.

[0132] A further embodiment comprises a computer on which the computer program for performing one of the methods described herein is installed.

[0133] A further embodiment according to the disclosure comprises a device or system configured to transmit a computer program for performing at least one of the methods described herein to a recipient. The transmission may, for example, be electronic or optical. The recipient may, for example, be a computer, a mobile device, a storage device, or a similar device. The device or system may, for example, comprise a file server for transmitting the computer program to the recipient.

[0134] In some embodiments, a programmable logic device (e.g., a field-programmable gate array, an FPGA) may be used to perform some or all of the functionality of the methods described herein. In some embodiments, a field-programmable gate array may interact with a microprocessor to perform any of the methods described herein. In general, in some embodiments, the methods are performed by any hardware device. This may be general-purpose hardware, such as a computer processor (CPU), or method-specific hardware, such as an ASIC.

[0135] In the foregoing Detailed Description, various features have been grouped together in examples in order to streamline the disclosure. This manner of disclosure should not be interpreted as intending that the claimed examples include more features than are expressly recited in each claim. Rather, as the following claims reflect, the subject matter may lie in fewer than all of the features of a single disclosed example. Accordingly, the following claims are hereby incorporated into the Detailed Description, with each claim being capable of standing as its own separate example.While each claim may stand as its own separate example, it should be noted that although dependent claims in the claims refer to a specific combination with one or more other claims, other examples also include a combination of dependent claims with the subject matter of any other dependent claim or a combination of any feature with other dependent or independent claims. Such combinations are intended to be encompassed unless it is stated that a specific combination is not intended. Furthermore, it is intended to encompass a combination of features of a claim with any other independent claim, even if that claim is not directly dependent on the independent claim.

[0136] The above-described embodiments are merely illustrative of the principles of the present disclosure. It is understood that modifications and variations of the arrangements and details described herein will be apparent to others skilled in the art. Therefore, it is intended that the disclosure be limited only by the scope of the following claims and not by the specific details presented in the description and explanation of the embodiments herein.

Claims

[1] Method (100) for determining measurement information based on a plurality of measured values (221) from a measured value range (223), the method comprising: Obtaining (120) a frequency distribution (222) of a plurality of measured values (221), wherein the measured values (221) of the frequency distribution (222) are each assigned to a class of a plurality (228) of classes of the frequency distribution (222), and wherein a frequency value (224) of a class (226) describes a number of measured values assigned to the class (226), Dividing (140) the frequency distribution (222) into a plurality of ranges (344), wherein each of the ranges (344) represents an interval of the measured value range (223) and contains one or more classes (226) of the frequency distribution (222), Selecting (160) one class of a respective area (344) as a selected class (366) of the respective area (344) based on a selection rule, wherein the areas (344) are each assigned an area feature (364) based on the selection rule, Determining (180) a probability value (384; 387) for one of the selected classes (366) based on the range features (364), wherein the probability value (384; 387) represents an estimate of the probability with which the selected class represents a value of a useful signal, wherein the determination of the probability value is based on machine learning methods. [2] Method according to claim 1, wherein the frequency distribution is part of a series of frequency distributions of a respective plurality of measured values, wherein the method (100) further includes comparing the selected classes (366) with the selected classes (517; 676) of a previous frequency distribution (513) of the series of frequency distributions to adjust or maintain one or more of the range features (364) depending on the comparison, and providing the range features (364) for determining the probability value (384). [3] The method of claim 2, wherein comparing the selected classes (366) with the selected classes (517; 676) of the previous frequency distribution (513) of the series of frequency distributions includes comparing a position (230) of one of the selected classes (366) of the frequency distribution (222) with the positions (230) of one or more of the selected classes (676; 517) of the previous frequency distribution. [4] The method of claim 2 or 3, wherein comparing the selected classes (366) with the selected classes (676) of a previous frequency distribution of the series of frequency distributions includes selectively adjusting the range feature (664) of the range of a selected class (666) taking into account a previous selected class (676) if the previous selected class (676) is in a correlation interval (677), the previous selected class (676) being one of the selected classes of the previous frequency distribution. [5] The method of claim 3, wherein comparing the selected classes (366) with the selected classes (676) of a previous frequency distribution of the series of frequency distributions includes determining a comparison position based on positions (630) of one or more of the selected classes (676) of a plurality of previous frequency distributions of the series of frequency distributions, and selectively adjusting the range feature (664) of the range of one of the selected classes (666) of the frequency distribution taking into account the selected classes used to determine the comparison position if the comparison position is located in a correlation interval (677). [6] Method according to one of claims 2 to 5, wherein adjusting one of the range features (364; 664) is based on an adjustment coefficient, wherein the adjustment coefficient is based on the probability value (674) and / or the frequency value (224) and / or a position (630) of one or more selected classes (676) of the previous frequency distribution. [7] The method of claim 4, wherein the adaptation of the range feature (664) is based on an adaptation coefficient, wherein the adaptation coefficient and / or the correlation interval (677) is based on the probability value (384) and / or the frequency value (224) and / or a position (230) of the considered previous selected class. [8] The method of claim 6, wherein comparing the selected classes (366) with the selected classes (676) of a previous frequency distribution of the series of frequency distributions includes determining a comparison position based on positions (630) of one or more of the selected classes of a plurality of previous frequency distributions of the series of frequency distributions, and determining the adjustment coefficient based on the probability values (384) and / or the frequency values (224) and / or the positions (230) of the selected classes used to determine the comparison position. [9] The method of claim 5, wherein the adjusting of the range feature (664) is based on an adjustment coefficient, wherein the adjustment coefficient and / or the correlation interval (677) is based on the probability value (384) and / or the frequency value (224) and / or a position (230) of the selected classes used to determine the comparison position. [10] A method according to any one of claims 6 to 9, wherein the method further includes providing the adaptation coefficient as part of the measurement information. [11] The method of any one of claims 4, 5, 7 or 9, wherein a position (678) of the correlation interval (677) in the measurement value range (223) is based on a position (630) of the selected class (676) in the measurement value range (223), and wherein a width (679) of the correlation interval (677) is based on an expected change in the value of the useful signal. [12] The method of any one of claims 4, 5, 7 or 9, wherein a position (678) of the correlation interval (677) in the measurement value range (223) is based on a position (630) of the selected class (676) in the measurement value range (223) and on an expected change in the value of the useful signal, and wherein a width (679) of the correlation interval (677) is based on the expected change in the value of the useful signal. [13] The method of any one of claims 4 to 12, wherein the method further comprises determining the correlation interval and / or the adjustment coefficient using an artificial neural network. [14] Method according to one of the preceding claims, wherein the method further comprises determining from the selected classes the one with the highest probability value as a useful signal class, and providing a position represented by the useful signal class in the measured value range as part of the measurement information. [15] The method of claim 14, wherein the method further includes providing the probability value of the useful signal class as part of the measurement information. [16] A method according to any one of the preceding claims, wherein selecting the selected classes includes selecting the class of one of the regions having the highest frequency value as the selected class of the region, and wherein the region feature associated with the region is based on the frequency value of the selected class of the region. [17] Method according to one of the preceding claims, wherein the regions are equidistant and adjacent. [18] Method according to one of the preceding claims, wherein a measured value (221) of the plurality of measured values (221) represents a time period between an emission of a light pulse and a detection of a photon, wherein the measured value (221) represents either a wanted signal value if the photon is based on the light pulse, or a background signal value, and wherein the wanted signal is based on one or more wanted signal values. [19] The method of claim 18, wherein dividing (140) the frequency distribution into the plurality of ranges includes selecting a width of one of the ranges such that the class of the range into which a measured value falling within the range falls with the highest probability represents a class of the useful signal if the useful signal falls within the range. [20] A method according to any one of the preceding claims, wherein the machine learning methods include an artificial neural network, and wherein the method includes training the artificial neural network based on the area features associated with the areas of the frequency distribution. [21] A method according to any one of the preceding claims, wherein the method further comprises deconvolving the frequency distribution with one or more convolution kernels prior to selecting the selected class. [22] A method according to any one of the preceding claims, wherein the plurality of measured values (221) represents a series of measured values (221), and wherein the method further includes collecting a plurality of consecutive measured values (221) of the series of measured values (221) to obtain a frequency distribution (222) of the series of frequency distributions. [23] Method according to one of the preceding claims, comprising: Emitting a light pulse using a light source, Determining a time period between emitting a light pulse and detecting a photon by a detector, and Providing the duration as one of the plurality of measured values. [24] Method according to one of the preceding claims, wherein the method includes obtaining a plurality of measured value series in parallel, and determining a contribution to the measurement information based on respective frequency distributions of a respective plurality of measured values of the plurality of measured value series, wherein the method includes determining the selected class with the highest probability value as a useful signal class from the selected classes of the respective frequency distributions, and providing a position represented by the useful signal class in the measured value range as part of the respective contribution to the measurement information. [25] Device (1200) for determining measurement information (1290) based on a plurality of measured values (221) from a measured value range (223), wherein the device (1200) is designed to obtain a frequency distribution (222) of a plurality of measured values (221), wherein the measured values (221) of the frequency distribution (222) are each assigned to a class of a plurality (228) of classes of the frequency distribution (222), and wherein a frequency value (224) of a class (226) describes a number of measured values assigned to the class (226), to divide the frequency distribution (222) into several areas (344), wherein each of the areas (344) represents an interval of the measured value range (223) and contains one or more classes (226; 326) of the frequency distribution, to select a class of a respective area (344) as a selected class (366) of the respective area (344) based on a selection rule, wherein the areas (344) are each assigned an area feature (364) based on the selection rule, and to determine a probability value (384; 387) for one of the selected classes (366) based on the range features (364), wherein the probability value (384; 387) represents an estimate of the probability that the selected class represents a value of a useful signal, wherein the determination of the probability value is based on a statistical model. [26] LiDAR device (1300) comprising the apparatus (1200) according to claim 25, and further: a light source (1302) configured to emit a light pulse (1303) and to provide a first signal (1304) in conjunction with the emission of the light pulse (1303), a detector (1305) configured to detect a photon (1306) and to provide a second signal (1307) following detection of a photon, a correlator (1308) configured to determine, based on the first signal (1304) and the second signal (1307), a time period (1309) between the emission of the light pulse and the detection of the photon, and to provide the time period (1309) as a measured value of the plurality of measured values, wherein the measured value represents a useful signal value when the photon (1306) is based on an echo of the light pulse (1303), wherein the useful signal is based on one or more useful signal values, and wherein the measurement information (1290) includes a position (230) of the selected class (366) of the frequency distribution (222) with the highest probability value. [27] The LiDAR device (1300) of claim 26, wherein the range feature (364) of the range (344) of a selected class (366) is selectively adjusted taking into account a previous selected class if the previous selected class is located in a correlation interval, the previous selected class being one of the selected classes of the previous frequency distribution, and wherein a position of the correlation interval in the measurement value range and / or a width of the correlation interval is based on an expected change in the value of the useful signal, and wherein the expected change is based on a speed and / or an acceleration of the LiDAR device. [28] LiDAR device (1300) according to one of claims 26 or 27, wherein the LiDAR device (1300) comprises a plurality of detector units, wherein the LiDAR device is designed to obtain a plurality of measured values using a respective detector unit, wherein the device (1200) for determining the measurement information is designed to determine a contribution to the measurement information assigned to the respective detector unit based on the respective plurality of measured values. [29] Computer program with a program code for carrying out the method according to one of claims 1 to 24 when the program runs on a computer.

Images