Method for determining properties of a structure of a sub-region of a track bed

Abstract

The invention relates to a computer-implemented method for identifying different layers and for determining mechanical properties in a measurement image of a track bed, wherein: in a first step, with respect to patterns of a signal property over one first set of points in each case, a pattern recognition with stored reference patterns of signal properties of the entire track bed is carried out in order to identify layers, layer transitions and / or layer boundaries defined in a database, wherein points are assigned to a layer or a layer transition over the entire measurement image; in a second step, for each second set of points which are jointly assigned to a single layer or to a single layer transition, at least one mechanical property is determined by carrying out a pattern recognition with respect to signal properties in the measurement image and the second sets of points are assigned those mechanical properties which are stored for the relevant recognized pattern for said single layer or said single layer transition.

Description

[0001] Method for determining the properties of a structure of a sub-area of ​​a track bed

[0002] The invention disclosed herein relates to a computer-implemented method for determining the mechanical properties of a track bed.

[0003] The invention disclosed herein relates to a method for determining the properties of a track bed structure comprising the artificially produced load-bearing layers and the naturally grown load-bearing layers.

[0004] According to the prior art, it is known to determine the properties of a track bed structure using radar signals. Electromagnetic beams with an output frequency, amplitude, and direction are emitted at a measurement position and at a specific time into a sub-area of ​​the structure by means of one or more transmitters. These beams are emitted as a signal at the surface of the structure. Depending on the local electromagnetic properties of locations within the sub-area, particularly their permittivity, the beams are reflected at these locations or transmitted to deeper locations. The reflected beams are recorded as signals at the surface of the sub-structure at a receiving position at a specific time and contain information about the locations within the sub-area, especially locations along an axis perpendicular to the track bed at the measurement position.By shifting the measuring position in the direction of the track, information about the locations of another sub-area is obtained. A measurement image created from this information by a receiver comprises two axes, representing the horizontal longitudinal direction of the measuring section and the direction perpendicular to the track bed, with each pixel of the measurement image representing a location in the track bed for which at least one signal property of the received radar signal is stored as a measured value.

[0005] According to conventional theory, conclusions about the electromagnetic properties of locations on the track bed, and consequently their mechanical properties, are drawn from the reflected signal.

[0006] The measurement position for outputting the signal and the reception position for receiving the signal can be identical or different. Different measurement positions can be achieved, for example, by using sensors moving along the track. This is a well-established principle.

[0007] The method can be implemented with a transmitter and a receiver as separate units. The transmitter and receiver can also be designed as a single unit. Multiple transmitters and / or receivers can also be used, with the transmitters and / or receivers positioned essentially in the track center and essentially at the track edge, respectively. For example, three transmitters and three receivers, or five transmitters and five receivers, can be arranged above the superstructure, perpendicular to the track direction.

[0008] The transmitter and receiver are positioned at a distance from the surface of the track's substructure.

[0009] The property of locations in the subsurface, as determined according to established doctrine, can be a statement about a discontinuity and thus a mechanical property. Within the scope of the disclosure, the mechanical property of locations can include the statement that the mechanical property of the locations differs from the mechanical property of other locations, from reference locations, or from a target property. The determination of the mechanical property can include the identification of a discontinuity.

[0010] The standard track bed structure typically comprises several layers of ballast or a ballast-like material. An actual track bed structure that deviates from the standard structure may include, for example, contaminated ballast, a section containing mud, water inclusions, or foreign objects such as boulders, concrete fragments, or small iron parts. Discontinuities in the ballast layers can be identified using state-of-the-art methods.

[0011] In the aforementioned prior art method, the layer affiliation of locations is not taken into account when determining the properties. This creates an accuracy problem, as recent research in this field has shown, because the same discontinuity has a different influence on the electromagnetic properties of the layers.

[0012] According to conventional theory, the problem of accuracy is addressed by outputting signals with a high output frequency. However, high-frequency signals experience greater attenuation in the setup than low-frequency signals, which is why only a shallow depth of the setup can be investigated using high-frequency signals.

[0013] The invention disclosed herein aims to achieve the general goal of increasing the accuracy of the investigation of the structure.

[0014] The invention disclosed here, unlike the prior art, is based on the understanding that the affiliation of locations to a layer must be determined and taken into account when ascertaining the mechanical properties of those locations. For example, a determined electromagnetic property of locations in a first layer can indicate a first moisture content, and in a second layer, an indication of a second moisture content.

[0015] According to the invention, the associated technical problem is solved by claim 1.

[0016] In one embodiment, a computer-implemented method is provided for identifying different layers and determining mechanical properties in a measurement image of a track bed. This measurement image is recorded by a radar-emitting and -receiving measuring vehicle traveling along a measuring section. The measurement image comprises two axes, with the horizontal axis representing the longitudinal direction of the measuring section and the vertical axis representing the direction into the track bed. Each pixel of the measurement image represents a location on the track bed for which at least one signal property of the received radar signal is stored as a measured value. In a first step, pattern recognition is performed for each set of locations using stored reference patterns of signal properties of the entire track bed.to identify layers, layer transitions and / or layer boundaries defined in a database, whereby locations are assigned to a layer or layer transition across the entire measurement image, and in a second step, for each second bundle of locations that are jointly assigned to a single layer or layer transition, at least one mechanical property is determined by performing pattern recognition of signal properties in the measurement image and assigning to the second bundles of locations those mechanical properties which are stored for the respective recognized pattern for this single layer or layer transition.

[0017] It is provided that in a first process step, a bundle of locations is assigned to a layer. In a second process step, the mechanical properties of the locations are determined. The method according to the invention is implemented as a computer-based method.

[0018] The measured value can be a reflection signal property of the signal reflected at a location. A pattern of the reflection signal property can be compared with a pattern of a reflection reference property of a reference signal, which is defined as a reference signal for a layer of the track bed structure. The reflection reference property most similar to the pattern of the reflection signal property is determined, which reflection reference property is linked to a reference layer in a database, and which reference layer is assigned to the location.

[0019] In one variant, electromagnetic layer property ranges are defined in a database for individual layers. In the first step of the process, a pattern of the determined electromagnetic properties of a bundle of locations is compared with the electromagnetic layer property ranges, and the locations are assigned to the layer that has a layer property range encompassing the electromagnetic properties of the bundle of locations.

[0020] The electromagnetic properties of locations can be defined by the permeability and / or the reflection and / or by the change in direction of the electromagnetic rays at the locations.

[0021] The electromagnetic property of a location is exemplified here as the property of that location which significantly influences the reflected signal. This electromagnetic property can be determined by other properties such as the density of the material at that location. Determining these other properties is generally considered equivalent to determining the electromagnetic property at that location as described here.

[0022] In one implementation, the signal pattern reflected at a bundle of locations and received by the receiver is compared with reference signals, where the reference signals describe the occurrence of a layer at a bundle of locations. Based on a similarity between the signal property of the signal pattern and the reference signal property of a layer, the occurrence of a layer at the locations is determined, and a layer is assigned to the locations. At least one property of a signal, such as frequency and / or amplitude and / or direction and / or phase, is compared with a suitable reference property.

[0023] The phase property of the reflected signal refers to the phase property of the reflected signal in comparison to the output signal.

[0024] It is planned that in the first process step the assignment of a single location to a layer is carried out by pattern recognition over a bundle of several locations, whereby not individual measured values, but a pattern of measured values ​​or a reference pattern of reference signals is considered.

[0025] The method can be characterized by the fact that a first plurality of reference patterns from reference signals for a ballast bed are defined in a database as a first layer of the structure, a second plurality of reference patterns from reference signals for a sub-ballast bed as a second layer of the structure, and a third plurality of reference patterns from reference signals for a sub-bed as a third layer of the substructure, wherein neighboring locations, at which a pattern of reflected signals is similar to a reference pattern from reference signals of one of the aforementioned layers, are assigned to this layer.

[0026] The pattern mentioned describes a property of the signal, and the reference pattern mentioned describes a reference property of the reference signal, such as frequency and / or amplitude and / or direction and / or phase of the signal or reference signal. A comparison of the pattern and the reference pattern allows for a comparison of at least one property and one reference property.

[0027] The aforementioned first process step can, for example, be implemented according to established principles of artificial intelligence. The invention disclosed here also includes a database of reference patterns.

[0028] Three layers are given as examples. A person skilled in the art can define more or fewer layers. Likewise, the names of the layers are only examples. A person skilled in the art can also identify the layers under consideration with a number.

[0029] Multiple locations can be assigned to each layer. A dataset can be created describing the progression of layers in the substructure. This dataset can include a vector description of the substructure layers.

[0030] The method according to the invention can also include the non-technical step of the user correcting an assignment of a location or group of locations to a layer by means of an input. Applying the established teaching, such an input correcting an assignment of a location or group of locations can trigger self-learning routines. The corrected assignments are stored in the database as reference patterns.

[0031] Based on the assignment of each location or locations to a layer, the mechanical properties of locations are determined in a second process step.

[0032] In one implementation variant, the reference patterns of the second step are available as 2D filters of a filter bank in order to be able to recognize different patterns in the measurement image, whereby the different reference patterns and the different layers of the track bed are defined in a database and a mechanical property is stored for each combination of one of the reference patterns and one of the layers, whereby different mechanical properties are stored for at least one of the reference patterns for at least two different layers.

[0033] The reference patterns of the second step differ from the reference patterns of the first step. The reference patterns of the first step preferably extend over a larger area of ​​pixels or locations. The reference patterns of the first step and the reference patterns of the second step can be represented as rectangular windows of pixels or locations.

[0034] The reference patterns for the first step are preferably taken from actual measurement images where the layer assignment has already been performed. In one execution variant, the first step is carried out by a machine learning model trained specifically for this purpose.

[0035] The reference patterns for the second step are preferably in the form of 2D filters or wavelets. A large number of reference patterns are compared by software or artificial intelligence with a large number of positions in the measurement image to identify matches between the reference patterns and the bundles of locations. If a match is found, mechanical properties stored for the reference pattern are assigned based on the previously established assignment of the bundle of locations to one of the layers.

[0036] The signal properties at locations within a layer identified in the measurement image can be compared with a reference signal property for that layer, and the mechanical properties of the most similar reference signal property can be assigned. This involves comparing the electromagnetic properties of a bundle of locations with electromagnetic reference properties and assigning the mechanical properties of the most similar electromagnetic reference property to the locations.

[0037] A pattern of signal properties at a bundle of locations can be compared with a pattern of reference signal properties for the determined layer of the location, and the mechanical properties of the most similar reference signal property can be assigned to the bundle of locations. This involves comparing the electromagnetic properties of a bundle of locations with electromagnetic reference properties and assigning the mechanical properties of the most similar electromagnetic reference property to the bundle of locations.

[0038] A data set can be determined containing a vector specification of the signal property, electromagnetic property, or mechanical property of the location. Determining the mechanical properties of the location(s) can generally include identifying discontinuities. A discontinuity in the substructure can be a change such as contamination of the layer or the presence of a non-track-related object. In addition to or as an alternative to determining a mechanical property, the location(s) can also be assigned to an object class. The mechanical properties of the objects within each object class are preferably known.

[0039] The method can be used to create a dataset about the mechanical properties or object affiliation of the structure at the locations. The dataset can further include information about the layers assigned to the locations.

[0040] The method according to the invention can be characterized by the fact that a property or a change in a property of the signal reflected at the locations is determined.

[0041] The properties or changes in properties of the signal reflected at the locations are measured values ​​determined by the influence of the electromagnetic properties of the setup, particularly the setup in the region between the transmitter, the location, and the receiver, using established principles. A property of the signal is its frequency and / or amplitude and / or direction and / or phase. A change in property is a change in the aforementioned property of the signal reflected at the locations, which property change occurs at the locations due to a change in the electromagnetic properties between a location and an adjacent location.

[0042] The determined data set can include information about a property of the reflected signal or a change in the property of the reflected signal.

[0043] The method according to the invention can be characterized in that the determined electromagnetic properties are filtered by several property filters and / or by a single property filter. This can be achieved by applying at least one filter to the measurement image before the first step of the method is carried out. Alternatively, at least one filter can be applied to the measurement image after the first step of the method has been carried out and before the second step of the method is carried out.

[0044] The method according to the invention can further be characterized in that a signal property of a said signal is filtered by several signal filters and / or by a signal filter.

[0045] According to the prior art, components of the filter's input signal can be amplified or attenuated by means of one or more filters. In particular, a person skilled in the art can select one filter from among several filters, as is illustrated by example in the embodiment described below.

[0046] The method according to the invention can be characterized in that, for locations that lie at a boundary or in a transition area between the aforementioned layers, this information is stored or drawn at the pixel of the measurement image.

[0047] At a boundary, the change between layers occurs abruptly. In a transition zone, the change between layers occurs gradually, with a beginning and an end to be defined according to established theory.

[0048] The data set generated using the method according to the invention can also include a specification of the boundaries of the layers or the transition regions.

[0049] The method according to the invention can be characterized in that a difference is determined between a determined signal property at the locations and a target signal property of a layer assigned to the locations stored in the database, and / or a difference is determined between a determined mechanical property of the locations assigned to a layer and a target mechanical property of this layer stored in the database, and / or a difference is determined between a determined electromagnetic property or an electromagnetic property change of a signal reflected at the locations assigned to a layer and an electromagnetic reference property or an electromagnetic reference property change of a reference signal of this layer.where a difference exceeding a limit is interpreted as a discontinuity at these locations and a difference falling below the limit is interpreted as a homogeneity at these locations of this layer.

[0050] A property of a signal can be, for example, its amplitude, frequency, direction, or phase. A reference property can accordingly be a reference frequency, amplitude, direction, or phase. A property change or reference property change is a change in the aforementioned properties caused by the reflection of the signal at specific locations.

[0051] The method according to the invention can be characterized in that a difference is determined between a determined mechanical property of locations which are assigned to a layer and a target property of this layer, wherein a difference exceeding a limit value is interpreted as a discontinuity at these locations and a difference falling below the limit value is interpreted as a homogeneity at these locations of this layer.

[0052] The generated data set can include an indication of whether a discontinuity or no discontinuity was detected at the locations.

[0053] The method according to the invention can be characterized in that a difference between a mechanical property at a first bundle of locations of an assigned layer and a mechanical property at a second bundle of locations of this assigned layer is determined, wherein a difference exceeding a limit value is interpreted as a discontinuity in this layer and a difference below the limit value is interpreted as homogeneity in this layer. The method according to the invention can comprise the comparison of a bundle of locations of a layer with a further bundle of locations of this layer.

[0054] In the method according to the invention, a threshold is defined above which a discontinuity in the input values ​​is considered an actual discontinuity in the setup. The person skilled in the art can specify the threshold or train neural networks to specify the threshold depending on the incoming reflection signal properties.

[0055] The method according to the invention can be characterized in that a difference is determined between a reflection signal property such as frequency and / or amplitude and / or direction and / or phase property of the signal reflected at a bundle of locations and a reference signal property such as frequency and / or amplitude and / or direction and / or phase property of a reference signal, wherein, in the case of a difference exceeding a limit value, the reference signal is considered unsuitable and in the case of a difference falling below the limit value, the reference signal is considered suitable.

[0056] This process step allows the suitability of the method according to the invention to be verified by the computing unit. If the limit value is exceeded, the operator can be prompted to confirm the assignment of a location to a shift.

[0057] The method according to the invention can be characterized in that a pattern of difference signal properties is compared with a pattern of reference difference signal properties and / or a pattern of layer properties with a layer property area pattern and / or a pattern of multiple signals with a reference signal pattern.

[0058] The method according to the invention allows the consideration of multiple locations, whereby, when considering multiple locations, the signals at the considered locations or the properties at the locations are described by a pattern. The processing of patterns and reference patterns is fundamentally applicable to all method steps described here. The consideration of patterns generally has the advantage that the result or a partial result of the method according to the invention becomes more stable against local fluctuations of the signal or the properties at a location.

[0059] The method according to the invention comprises a first method step and a second method step, wherein a pattern is considered in the first method step and in the second method step.

[0060] When considering a pattern that describes the difference signal property, this pattern can be compared with the difference signal property range for the respective layer.

[0061] When considering a pattern that describes the electromagnetic properties, this pattern can be compared with the layer property range for the respective layer.

[0062] When considering a pattern that describes the signal property, this pattern can be compared with the layer property range for the respective layer. As explained above, a signal property can be defined by its amplitude and / or frequency and / or direction and / or phase. Similarly, a reference signal property can be defined by its reference signal amplitude and / or frequency and / or direction and / or phase.

[0063] The size of the pattern under consideration can be varied across the locations in order to determine the influence of the pattern size on the assignment of locations to the individual layers.

[0064] A reference pattern can describe a boundary or transition area between two layers, so that by comparing a pattern with a reference pattern a boundary or transition area between two layers can be determined.

[0065] A pattern forms the aforementioned properties at locations in a

[0066] A range with a range size starting from . The range size and, accordingly, a reference range size can be adjusted to detect a boundary or transition between layers.

[0067] As mentioned above, a user can correct the assignment of a location to a shift. The user can also correct the assignment of a subset of locations or a bundle of locations to a shift. The corrected assignment is stored in the database as a reference pattern.

[0068] The second process step can be characterized by comparing a pattern of electromagnetic properties of locations of an assigned layer with a multitude of reference patterns of electromagnetic reference properties of a reference layer corresponding to the assigned layer, which reference patterns of electromagnetic reference properties are stored in the database, and selecting a reference pattern of electromagnetic reference properties from the multitude of reference patterns of electromagnetic reference properties that is most similar to the pattern of electromagnetic properties, and assigning a mechanical reference property to the locations, which mechanical reference property is associated with the selected reference pattern of electromagnetic reference property in the database.

[0069] The second process step can be characterized by the fact that a computing unit compares a reflection signal property pattern of an assigned layer with a multitude of reflection reference signal property patterns of a reference layer corresponding to the assigned layer, which reflection reference signal property patterns are stored in the database, and selects a reflection reference signal property pattern from the multitude of reflection reference signal property patterns which reflection reference signal property pattern is most similar to the reflection signal property pattern, and assigns a mechanical reference property to the location which mechanical reference property is associated with the reflection reference signal property pattern in the database.The method according to the invention can be characterized in that several transmitters and receivers are arranged along an axis transverse to the track direction, wherein a measurement image is created for each receiver, which measurement image comprises two axes, wherein the horizontal axis represents the longitudinal direction of the measuring section and the vertical axis represents the direction into the track bed at the position of the respective receiver, wherein the first and second steps of the method are carried out on each measurement image of the several receivers.

[0070] The invention disclosed herein also relates to a computer program product comprising instructions which, when a computer program is executed by a computer, cause it to carry out the method described in the disclosure.

[0071] The invention disclosed herein also relates to a computer-readable storage medium comprising

[0072] Instructions which, when executed by a computer, cause it to carry out the procedure described in the disclosure, and / or comprising layer property areas and / or reference signals and / or reference properties.

[0073] The invention is further explained with reference to the following embodiments shown in the figures:

[0074] Fig. 1 and Fig. 2 show measurement images and result images of an analysis of the measurement images.

[0075] The embodiments shown in the figures merely illustrate possible embodiments. It should be noted that the invention is not limited to these specifically depicted embodiments, but also includes combinations of the individual embodiments with one another and combinations of an embodiment with the general description given above. These further possible combinations need not be explicitly mentioned, as they are within the knowledge of a person skilled in the art in this technical field, given the teaching of the present invention. The scope of protection is defined by the claims. However, the description and the drawings should be consulted for the interpretation of the claims.Individual features or combinations of features from the different embodiments shown and described can each represent independent inventive solutions. The problem underlying these independent inventive solutions can be found in the description.

[0076] Reference symbol list:

[0077] 100 - 103 layers

[0078] 105 - 108 Filter response

[0079] 110 - 112 borders

[0080] 113 Measurement image

[0081] Figure 1 shows a measurement image 113 of signals. The position in the measurement section is plotted on the abscissa (x-axis). A radar signal (hereinafter referred to as "signal") is emitted into the track bed structure. This signal is reflected within the track bed structure, with the depth of reflection in measurement image 113 being indicated above by the time interval from emission to reflection at a specific location, plotted on the ordinate. The measurement image thus comprises two axes: the horizontal axis represents the longitudinal direction of the measurement section, and the vertical axis represents the direction into the track bed. Each pixel of the measurement image represents a location in the track bed for which at least one signal property of the received radar signal is stored as a measured value.

[0082] In measurement image 113 of Figure 1 above, the amplitude of the signal is indicated by the brightness of the pixel. A white pixel illustrates a large amplitude of the signal reflected at the location symbolized by the pixel. Conversely, a black pixel illustrates a small amplitude.

[0083] The inventive method for determining the properties of a sub-area of ​​a track bed structure based on the amplitude as a property of the signal is thus discussed.

[0084] To obtain such a measurement image, signals comprising electromagnetic beams with output signal properties such as output frequency and / or output amplitude and / or output direction and / or output phase are emitted into the track bed at a specified time by means of a transmitter at the measurement position. A signal is reflected and / or absorbed and / or transmitted at a location in the track bed, depending on the electromagnetic properties of that location. Reflected beams comprising electromagnetic beams reflected at that location with receive signal properties such as receive frequency and / or receive amplitude and / or receive phase are received by a receiver located at a specified time by means of a receiver at a specified time. The measurement position corresponds here, for example, to the receive position, but this is not a limiting factor.A measuring car is moved along the track during the output and reception of the signal.

[0085] Applying standard principles, a processing unit determines the amplitude of the reflected signal at the location and generates the measurement image shown in Figure 1. The measurement image displays the amplitude as a reflection signal property at the locations; it is also conceivable to determine and supplementarily or alternatively display the frequency and / or direction and / or phase properties.

[0086] The measurement image shows the amplitudes of the signals reflected at the locations. According to established theory, patterns can be determined over sub-areas of the measurement image. Such a pattern can, for example, describe the distribution of the brightness of the pixels over a sub-area of ​​the measurement image. Additionally or alternatively, the pattern can describe sub-areas of pixels, in particular contiguous pixels of the measurement image, which sub-areas encompass amplitudes within a specific amplitude range.

[0087] The first step of the method according to the invention relates to the allocation of locations to layers of the structure. According to the prior art, a track bed structure comprises several layers, such as (from top to bottom) a ballast bed, a sub-ballast bed, and a sub-bed.

[0088] The invention provides that a multitude of reference signal properties are stored in a database for each of the aforementioned layers. In particular, a multitude of reference patterns regarding reference properties, especially the distribution of reference signal properties, can be stored in the database. A reference pattern can, for example, and with reference to the embodiment discussed here, comprise a description of the amplitudes over a sub-area of ​​the incoming measurement pattern. The sub-area and the pixels with the amplitudes can be described by relative x-coordinates and y-coordinates of a measurement image.

[0089] Since amplitudes are represented by pixels of varying brightness in the upper measurement image of Figure 1, the embodiment discussed here compares amplitudes as signal properties with reference amplitudes as reference properties. It is also conceivable that frequencies or reference frequencies, or direction or reference directions, or phase properties or reference phase properties are processed.

[0090] In general, processing patterns of signal properties is more useful than processing individual signal properties.

[0091] A pattern of amplitudes to be determined within a range of locations can be compared with a reference pattern of reference amplitudes within a reference range of reference locations. A reference pattern describing a layer can be selected to assign the locations of the range to the layer of the selected reference pattern if the similarity between the pattern to be determined and the reference pattern exceeds a threshold. A reference pattern describing a boundary or transition region between two layers can be selected to assign the locations of the range to such a boundary or transition if the similarity between the pattern to be determined and the reference pattern exceeds a threshold. Overlapping regions can be processed.Basically, the comparison of a pattern with a reference pattern or with several reference patterns is known according to established doctrine; this doctrine is applied.

[0092] The first process step is executed using a computing unit. The computing unit receives the amplitudes at the locations as input values, as shown in the measurement image above in Figure 1 (without layer boundaries).

[0093] The result of the first process step is the determination of the layers of the structure and the assignment of locations to layers 100-102. Alternatively or additionally, the boundaries and / or transition zones between the layers can also be determined. The ballast bed 100, the sub-ballast bed 101, and the sub-bed 102 are shown as examples in the upper measurement image of Figure 1. The designation of layers 100-102 and the definition of layers 101-102 with regard to the ballast present in the respective layer are not relevant for the explanation of the invention.

[0094] Alternatively or in addition to determining the layers 100-102 of the structure, the boundaries 110 to 112 between the layers can also be determined, which is basically equivalent to determining the layers and assigning locations to the layers.

[0095] In the second step of the process, the reflection signal property of the locations of an assigned layer is compared with a multitude of reflection reference signal properties of a reference layer corresponding to the assigned layer. These reflection reference signal properties are stored in the database. A reflection reference signal property is selected from the multitude of reflection reference signal properties of the reference layer corresponding to the assigned layer, specifically the one that most closely resembles the reflection signal property of a bundle of locations. A mechanical reference property is associated with the reflection reference signal property in the database; this mechanical reference property is then assigned to the locations of the bundle as a mechanical property.

[0096] Alternatively, in the second step of the process, the electromagnetic properties of the locations of an assigned layer can be compared with a multitude of electromagnetic reference properties of a reference layer corresponding to the assigned layer, which electromagnetic reference properties are stored in the database. An electromagnetic reference property is selected from the multitude of electromagnetic reference properties that most closely resembles the electromagnetic properties of a bundle of locations, and a mechanical reference property is assigned to the locations of the bundle, which is linked to the electromagnetic reference property in the database.

[0097] Instead of considering individual locations, the second step of the process involves comparing patterns with a large number of reference patterns, as described in the disclosure of the invention.

[0098] The reference patterns for the second step can be a filter bank consisting of several 2D filters, particularly 2D Gabor filters, with different frequencies and orientations to detect different patterns in the measurement image. However, pattern recognition in the measurement image for pattern identification can also be performed using other known methods.

[0099] Each of the different reference patterns in the second step is assigned at least one mechanical property to each layer defined in the database. Thus, a first reference pattern comprises one assigned mechanical property per layer. The mechanical properties assigned to a first reference pattern in a first layer can differ from the mechanical properties assigned to the first reference pattern in a second layer. In general terms, for at least one reference pattern, differing mechanical properties are stored in the database for at least two different layers.

[0100] In the second process step mentioned above, the signal amplitude is processed as a signal property. The electromagnetic property of location is also processed as an example in the second process step mentioned above.

[0101] The second process step is also performed using a computing unit.

[0102] The signal characteristics, such as amplitudes (shown in the measurement image above in Figure 1), can be the input values ​​for the execution of the second process step with the computing unit.

[0103] The input values ​​can be the signal characteristics amplified or attenuated by multiple filters. Figure 1 below shows the result of applying multiple filters to the measurement image in Figure 1 above. A person skilled in the art can apply multiple filters to the measurement image above, which will reveal discontinuities in the layers.

[0104] In the resulting image in Figure 1 below, a filter response 105 can be seen in the uppermost first layer 100 with a particular amplitude value or a particular amplitude pattern. A comparison of these particular measurements or this particular measurement pattern with reference measurements or with a reference measurement pattern from a reference layer corresponding to the first layer 100 yields the result that the filter response 105 is an indication of a structure.

[0105] However, an analysis of the filter response 106 of the second layer 101, which is similar to the filter response 105, after the second process step yields the result that the filter response 106 is a mixture of the material from layer 102 with the material of layer 101.

[0106] Failure to assign the area with filter response 105 to the first layer 100 could lead to a misinterpretation, namely the interpretation of the filter response as a mixture of the material from layer 101 and layer 102. Failure to assign the area with filter response 106 to the second layer 101 could lead to a misinterpretation of filter response 106 as a structure.

[0107] The filter responses 107 of the third layer 102 are a typical mixture of the material from layer 102 and layer 103. A similar filter response 108 of the second layer 101 is a mixture of the material from this layer 101 with a fine material, which significantly reduces the coefficient of friction between the gravel grains of layer 101.

[0108] A selected filter can be applied to the measurement image in Figure 1 to reduce or enhance a selected discontinuity. Figure 2 shows the result of applying a selected filter to the measurement image in Figure 1 above.

[0109] The measurement image in Figure 1 (bottom) shows the filter response of an application of a selected filter to the measurement image in Figure 1 (top). Figure 2 (top) shows the accumulation of water in the layers. This accumulation of water influences the mechanical properties of a layer. Thus, the mechanical properties of a location are determined indirectly.

[0110] The diagram in Figure 2 below shows the sum of the filter responses depicted in the image in Figure 2 above at a vertical range of locations. The y-value of the diagram in Figure 2 below thus indicates the frequency of the discontinuity detected by the filter. This sum of the filter responses can be compared to a threshold value. A filter response exceeding the threshold value is interpreted as insufficient mechanical strength of this layer at that location. A filter response below the threshold value is interpreted as sufficient mechanical strength of this layer at that location.

Claims

Patent claims 1. A computer-implemented method for identifying different layers and determining mechanical properties in a measurement image of a track bed, which measurement image was recorded by a radar-emitting and -receiving measuring vehicle traveling along a measuring section, the measurement image comprising two axes, wherein the horizontal axis represents the longitudinal direction of the measuring section and the vertical axis represents the direction into the track bed, wherein each pixel of the measurement image represents a location in the track bed for which at least one signal property of the received radar signal is stored as a measured value, characterized in that, in a first step, pattern recognition is performed for patterns of the signal property over a first bundle of locations using stored reference patterns of signal properties of the entire track bed in order to identify layers defined in a database.To identify layer transitions and / or layer boundaries, whereby locations are assigned to a layer or layer transition across the entire measurement image, in a second step, for each second bundle of locations that are jointly assigned to a single layer or a single layer transition, at least one mechanical property is determined by performing pattern recognition of signal properties in the measurement image and assigning to the second bundles of locations those mechanical properties which are stored for the respective recognized pattern to this single layer or single layer transition.

2. Computer-implemented method according to claim 1, characterized in that at least one filter is applied to the measurement image before the first step of the method is carried out.

3. Computer-implemented method according to claim 1 or 2, characterized in that at least one filter is applied to the measurement image after the first step of the method. was carried out and before the second step of the procedure is carried out.

4. Computer-implemented method according to one of claims 1 to 10. 3 , characterized in that in the first step of the procedure, for locations that lie on a boundary or in a transition area between the layers, this information is stored or drawn into the image point of the measurement image.

5. Computer-implemented method according to one of claims 1 to 10. 4 , characterized in that in the first step a pattern of amplitudes in an area of ​​one of the first bundles of locations is compared with reference patterns of reference amplitudes of a reference area of ​​reference locations, wherein, if there is a similarity between the pattern and one of the reference patterns exceeding a limit value, the locations of the area are assigned to a layer or a boundary or a transition area of ​​the reference pattern .

6. Computer-implemented method according to one of claims 1 to 5 , characterized in that in the second process step the signal property of the second bundle of locations of an assigned layer is compared with a plurality of patterns of a reference signal property of a reference layer corresponding to the assigned layer, wherein from the plurality of patterns of the reference signal properties of the reference layer corresponding to the assigned layer the one is selected which is most similar to the pattern of the signal property of the bundle of locations .

7. Computer-implemented method according to claim 6, characterized in that a mechanical reference property is associated with the reference signal property in a database, which mechanical reference property is assigned as a mechanical property to the locations of the bundle.

8. Computer-implemented method according to one of claims 1 to 7, characterized in that a difference between a determined mechanical property of a bundle of locations assigned to a layer and a target mechanical property of this layer stored in the database is is determined, whereby a difference exceeding a limit is interpreted as a discontinuity at these locations in this layer and a difference falling below the limit is interpreted as a homogeneity at these locations in this layer.

9. Computer-implemented method according to one of claims 1 to 8 , characterized in that a pattern of difference signal properties is compared with patterns of reference difference signal properties or a pattern of layer properties with layer property range patterns or a pattern of signal properties with reference signal patterns from reference signal properties .

10. Computer-implemented method according to one of claims 1 to 9 , characterized in that several transmitters and receivers are arranged along an axis transverse to the track direction, wherein a measurement image is created for each receiver, which measurement image comprises two axes, wherein the horizontal axis represents the longitudinal direction of the measuring section and the vertical axis represents the direction into the track bed at the position of the respective receiver, wherein the first and second steps of the method are carried out on each measurement image of the several receivers .

11. Computer-implemented method according to one of claims 1 to 10 , characterized in that the reference patterns of the second step are available as 2D filters of a filter bank in order to be able to recognize different patterns in the measurement image, wherein the different reference patterns and the different layers of the track bed are defined in a database and a mechanical property is stored for each combination of one of the reference patterns and one of the layers, wherein at least for one of the reference patterns different mechanical properties are stored for at least two different layers.

12. Computer program product comprising instructions which commands are executed by a computer when a computer executes a computer program. cause the process according to any one of claims 1 to 11 to be carried out.

13. A computer-readable storage medium comprising instructions which, when executed by a computer, cause it to perform the method according to any one of claims 1 to 12. 12, and / or encompassing layer property areas and / or reference signals and / or reference properties.

Images