Method for testing device, testing device
By evaluating and clustering the detection signals through machine learning, the problems of complex operation and unreliable results of existing detection equipment are solved, and detailed and reliable detection results and simplified equipment operation are achieved.
Patent Information
- Application Number
- CN202380033372.1
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Priority Date
- 2022-04-11
- Filing Date
- 2023-04-04
- Publication Date
- 2024-11-19
AI Technical Summary
Existing technologies struggle to effectively evaluate detection signals in non-destructive testing to provide detailed and reliable results, and the operation of detection equipment is complex and dependent on the detection context and the movement path of the object.
The detection signal is evaluated using a machine learning process. The sensor unit detects the signal and the computing unit evaluates and outputs the results. The machine learning model is combined to cluster the signal and automatically allocate evaluation parameters, simplifying the operation of the detection equipment.
It enables detailed evaluation of the detection signal, improves the reliability and repeatability of the detection results, simplifies equipment operation, reduces the risk of user error, and provides a range of information.
Smart Images

Figure CN118984953A_ABST
Abstract
Description
Background Art
[0001] A method for a detection device has been proposed, wherein the detection device is arranged to non-destructively detect a detection signal of an object arranged within an inspection context. Summary of the invention
[0002] The invention is based on a method for a test device, wherein the test device is configured to non-destructively detect a test signal of an object arranged within an examination context.
[0003] It is proposed that in at least one method step of the method, based on a machine learning process, the detection signal is assigned to a detection result so as to output the detection result and / or use the detection result to set the detection device. The method preferably includes an application phase. In the application phase, in at least one detection step of the method, the detection signal is detected by a sensor unit of the detection device. In the application phase, in at least one evaluation step of the method, the detection signal is evaluated by an evaluation unit. The evaluation unit is preferably a computing unit of the detection device or alternatively an external computing system having a data connection with the detection device, especially a radio wave-bound data connection, especially an Internet server. In the application phase, the detection result is preferably output to a user of the detection device by an output unit of the detection device and / or by an external output device (such as a smart phone, a tablet computer, etc.). Additionally or alternatively, in the application phase, in the setting step of the method, the detection result is used by a computing unit of the detection device in order to change the setting of the detection device, especially in order to repeat the detection step with a setting of the detection device, especially a sensor unit, that matches the object and / or background.
[0004] The method preferably includes a learning phase. In the learning phase, in the data acquisition step of the method, sample values (so-called training data) of the detection signal are collected. These sample values can be detected by the sensor unit of the detection device, detected by the sensor unit of another detection device, generated by simulation, read out from the recorded data set, etc. In the learning step of the learning phase, these sample values are processed into a model by a machine learning process, and the model maps the detection signal to the detection result. The machine learning process is preferably performed by a learning unit. The learning unit is preferably an external computing system, especially the external computing system already mentioned, alternatively a computing unit of the detection device or another external computing system, such as a dedicated server. Optionally, the method includes a providing step, in which, after the learning phase, the model generated by the machine learning process is transmitted to the detection device and is stored in the memory of the evaluation unit (especially the computing unit of the detection device). Preferably, in the application phase, the evaluation unit uses the model generated by the machine learning process to perform the evaluation step. The learning phase can be completed before the application phase or can overlap in time with the application phase, wherein the example values on which the model is based are supplemented in particular with measured values of the detection signal detected in the application phase.
[0005] The machine learning process is an algorithm from the field of machine learning, which is used to create the model preferably based on these example values. The machine learning process preferably includes a neural network and / or a classification method. The machine learning process may include supervised learning and / or unsupervised learning. The method preferably includes an assignment step, in which at least one evaluation parameter is assigned to the example value of the detection signal. The detection result may be the same as an evaluation parameter, including multiple evaluation parameters or determined according to these evaluation parameters. In the case of supervised learning, the evaluation parameters for the example values, especially the overall evaluation parameters, indicate during the learning process: which detection result the model should map the example value to. In the case of unsupervised learning, the evaluation parameters are then assigned to the structure in these example values determined by the learning process. The assignment step may be performed manually or automatically during the data acquisition step. The evaluation parameters may, for example, describe the nature of the inspection background, the type of object, the size of the object, the depth of the object in the inspection background, etc. For example, when the example value of the detection signal is detected on a known inspection background with a known object, the evaluation parameters may be known. The inspection context and / or the object can be specified, for example, by reconstructing the inspection situation in a laboratory with the inspection device or by a computer-based simulation of the inspection situation and / or can be obtained from the construction plan. The evaluation parameters can be detected, for example, by an external sensor device, which includes other sensors that are in particular invasive and / or cannot be integrated into the inspection device due to the limited installation space of the inspection device and / or are more accurate than the sensor unit of the inspection device.
[0006] The detection device is particularly configured to be placed on and / or moved on the inspection background during the application phase in order to detect the object. "Non-destructive" should be understood in particular as no permanent changes to the inspection background and the object, in particular no damage. The detection device detects the object in particular non-invasively. Preferably, the sensor unit generates a signal in the detection step and sends the signal to the inspection background, in particular as an electromagnetic wave or as an electromagnetic field, wherein the sensor unit detects changes in the signal caused by the inspection background and / or the object, in particular backscattering, as a detection signal. Alternatively or additionally, the detection device is configured to detect a signal emitted by the object as a detection signal. The detection device is, for example, configured to detect metal objects, non-metal objects, wires, in particular low-voltage wires, single-phase AC wires or three-phase power lines, wooden beams, metal supports or plastic pipes, in particular water-filled plastic pipes or non-water-filled plastic pipes, etc., as objects. The inspection background is, for example, a wall, ceiling, floor or furniture of a building.
[0007] By following the design of the present invention, the detection signal can be advantageously evaluated in detail. In particular, additional information contained in the detection signal can be advantageously and reliably extracted from the detection signal, which additional information exceeds the mere presence or absence of an object. By using the additional information, the operation of the detection device can be advantageously kept simple, and / or an advantageously large amount of information can be supplied to the user of the detection device, in particular by only performing a single measurement with the detection device. In addition, an advantageously reliable repeatability of the detection result can be achieved, and in particular the dependence of the detection result on the moving path of the detection device on the detection background can be advantageously kept low.
[0008] It is further proposed that: in the learning phase of the machine learning process, in particular the learning phase already mentioned, the sample values of the detection signal are divided into different groups according to the clustering formation of these sample values. Preferably, the learning phase includes a preprocessing step, in which the learning unit processes the sample data before the learning step. In the preprocessing step, the learning unit preferably performs unsupervised learning in order to divide the sample values into these groups. Preferably, the learning unit divides the sample values into at least two different groups in the preprocessing step. The learning unit preferably determines these groups and whether the sample values belong to these groups by means of a clustering method. Preferably, the learning unit extracts at least one feature, preferably multiple features, from each sample value to be processed in the preprocessing step. The feature can be, for example, a physical quantity, a statistic or an abstract indicator. If the sample values have similar values for the feature / all features, the learning unit preferably assigns these sample values to the same group. If the sample values have different values for the feature / one of these features, the learning unit preferably assigns these sample values to different groups. Here, whether the values of these features are similar or different is defined by the clustering method used. By the design according to the invention, example values that are similar to each other in terms of the extracted features can be aggregated and processed together. In particular, in the case of an advantageously low risk of error, the same evaluation parameter can be assigned to all example values in the same group. In particular, the duration of the execution of the assignment step can be advantageously kept short. In particular, example values detected in an unknown detection situation can be assigned the evaluation parameters of the same group of example values detected in a known detection situation.
[0009] It is also proposed that the machine learning process, especially in the preprocessing step, uses the nearest neighbor classification to perform the division into different groups. Particularly preferably, the learning unit uses the K nearest neighbor algorithm (KNN) to perform the division into different groups. In particular, the learning unit assigns the example value to a group, more precisely according to which group the majority of adjacent example values are assigned to. Adjacent example values are a specified number of example values that have the minimum distance from the example value to be divided in the parameter space expanded by these features. The distance between two example values in the parameter space expanded by these features can be determined based on the Euclidean metric, the Manhattan metric or other metrics. The number of adjacent example values used can be specified by the operator of the learning unit and / or determined by means of an optimization algorithm. Optionally, in the case of a majority decision of the adjacent example values, the voting share of the example value is weighted, especially using the corresponding distance of the adjacent example value to the example value to be divided. Alternatively, the example values are divided into different groups by means of a cuboid classifier, a distance classifier, a polynomial classifier or other classifiers. By the design scheme according to the present invention, the example values can be advantageously and reliably divided into these groups.
[0010] It is also proposed that: in the learning phase of the machine learning process, evaluation parameters, especially one / more evaluation parameters mentioned, are assigned in groups to the sample values of the detection signal, especially the sample values already mentioned. Preferably, at least one sample value is selected from each group as a representative of the group. The representative is an example value whose evaluation parameters are known by reconstruction, simulation or from other sources. Preferably, the learning unit assigns the evaluation parameter or these evaluation parameters to all sample values within the representative group, and the evaluation parameters are assigned to the representative. In the data acquisition step, the representative can be added to the remaining sample values, or determined after the remaining sample values are divided into these groups. Optionally, the learning unit uses the representative and / or domain knowledge as a sampling point for the nearest neighbor classification. Alternatively, the number of groups is automatically determined, for example, by changing the initial cluster center and the minimum sum distance. By following the design scheme of the present invention, the duration of the allocation step can be advantageously kept short. In particular, a large number of sample values can be advantageously used for the machine learning process, and the evaluation parameters of these sample values have not been detected in detail.
[0011] It is also proposed that in the learning phase of the machine learning process, in particular the learning phase already mentioned, the data set consisting of the example values of the detection signal is at least substantially composed of user data. "Substantially" should be understood in particular as more than 30%, preferably more than 60%, particularly preferably more than 90%, and very preferably more than 99%. Preferably, in addition to the user data, the data set includes representatives of these groups. "User data" should be understood in particular as detection signals that are detected by the user using the detection device or another detection device during the application phase, especially in real, i.e. non-reconstructed or simulated detection situations. Preferably, user data is collected from a plurality of detection devices in order to form the data set. In particular, the user data is automatically transmitted from the data interface of the detection device to the learning unit. By the design according to the present invention, the time and cost expenditure for reconstructing and / or simulating detection situations can be advantageously kept low. In addition, these example values advantageously include a plurality of real and not only ideal detection situations, which cannot be reconstructed in the laboratory, especially in terms of their complexity and / or diversity. In particular, a machine learning process can be used to create an advantageously robust model.
[0012] It is also proposed that in at least one method step, an operating mode of the detection device is automatically selected based on a machine learning process. The detection device, in particular a sensor unit, comprises: at least one operating mode, which is set for a specific detection situation; and at least one further operating mode, which is set for another detection situation. The detection situation differs, for example, due to different materials or structural types of the inspection background and / or due to different types of detected objects. The operating modes differ, for example, due to the sensor elements used by the sensor unit and / or due to signal parameters of the emitted signal (for example, intensity, frequency, duration, etc.). Preferably, the sensor unit detects an initial detection signal in the detection situation, which is evaluated by the evaluation unit to generate an initial result. Preferably, the evaluation unit, in particular a computing unit of the detection device, selects one of the operating modes based on the initial result, so that all subsequent detection steps are performed in the same detection situation in the selected operating mode. Preferably, in a memory of the evaluation unit, in particular a computing unit of the detection device, a table is stored, which assigns the operating modes to the initial result and is read out by the evaluation unit to select the operating mode. Alternatively, the detection result includes the operating mode provided for the detection situation as an evaluation parameter and is learned in the process of the machine learning process. By following the design scheme of the present invention, it is possible to achieve advantageously simple and intuitive operation of the detection device. In particular, the user can save the setting of the detection device. In particular, it is advantageously possible to keep the number of menu items of the required operating elements of the detection device and / or the graphical interface of the detection device small. In particular, it is advantageously possible to keep the risk of the user's erroneous operation of the detection device low.
[0013] It is also proposed that in at least one method step, a zero reference of the detection signal is automatically selected according to a machine learning process. The zero reference is a measured value of the detection signal when the detection background has no detectable object. Preferably, the evaluation unit evaluates the difference between the detection signal and the zero reference in order to determine whether an object is present in the inspection background. Under normal circumstances, the zero reference depends on the material and structure type of the inspection background. In particular, different operating modes of the detection device have different zero references. The zero reference is preferably part of a model created by a machine learning process. In particular, in the assignment step, a zero reference is assigned to each example value, in particular each group of example values. Alternatively, the zero reference is stored as a table in a memory of a computing unit and is read out by the computing unit according to the selected operating mode. By the design scheme according to the present invention, the user of the detection device can be saved from the individual determination of the zero reference. In particular, the detection device can be advantageously used to detect an object quickly after being switched on. In particular, the risk of incorrect operation of the detection device can be advantageously kept low.
[0014] It is further proposed that in at least one method step, the detection device outputs a group of example values of the detection signal, in particular the evaluation parameters of the group mentioned above, in particular the evaluation parameters mentioned above, when assigning the measured value of the detection signal to the group. The evaluation unit preferably assigns the evaluation parameters to the measured value with the help of the model. Preferably, the output unit outputs at least: whether the object is detected or whether the object is not detected. Preferably, the output unit outputs the evaluation parameters in addition to the information whether the object is detected. For example, the output unit outputs the material or structure type of the inspection background as the evaluation parameter. For example, the output unit outputs the type of the detected object, the size of the detected object, in particular the maximum extension of the detected object parallel to the surface of the inspection background, the distance of the detected object from the surface of the inspection background, the distance of the edge of the detected object from the current measuring point of the detection device, in particular the distance parallel to the surface of the inspection background, etc., as the evaluation parameters. By the design according to the present invention, a plurality of information can be advantageously provided to the user using, in particular, a single measurement.
[0015] A detection device is also proposed, which has at least one sensor unit, in particular the sensor unit mentioned above, for detecting a detection signal, in particular the detection signal mentioned above; and the detection device has at least one calculation unit, in particular the calculation unit mentioned above, for executing the method according to the present invention. The sensor unit includes at least one sensor element. Optionally, the sensor unit includes a plurality of sensor elements, in particular differently designed. Examples of sensor elements of the sensor unit include: radar, in particular narrowband radar, in particular in the frequency range of 2.4 GHz to 2.4835 GHz, and / or ultra-wideband radar; inductive sensor; capacitive sensor; AC sensor, in particular 50 Hz AC sensor and / or 60 Hz AC sensor, etc. "Calculation unit" should be understood in particular as a unit with information input, information processing and information output. Advantageously, the calculation unit has at least a processor, a memory, input and output devices, other electrical components, an operating program, a regulation routine, a control routine and / or a calculation routine. Preferably, the components of the calculation unit are arranged on a common circuit board and / or advantageously arranged in a common housing. Preferably, the detection device includes an output unit for outputting the detection result. The output unit for example comprises a display, a microphone, a vibration alarm or the like, so as to output the detection result. Preferably, the detection device comprises a data interface. The data interface can have an interface element of a wired interface element and / or radio wave binding, in particular a Bluetooth interface, a WLan interface, an Ethernet interface or the like. The data interface is arranged to output the detection result on an external output device, to forward the detection signal to an external computing system and / or to receive the model from an external computing system. By according to a design of the present invention, a detection device can be provided, which is advantageously easy to operate and / or advantageously provides a plurality of information.
[0016] In this case, the method according to the invention and / or the detection device according to the invention should not be limited to the above-mentioned applications and embodiments. In particular, the detection device according to the invention and / or the method according to the invention can have a number of individual elements, components and units and method steps that differs from the number mentioned in the text in order to meet the working principle described in the text. In addition, in the case of value ranges specified in the present disclosure, the values within the mentioned limits should also be regarded as public and can be used arbitrarily. BRIEF DESCRIPTION OF THE DRAWINGS
[0017] Further advantages are obtained from the following description of the drawings. Embodiments of the invention are shown in the drawings. The drawings, the description and the claims contain a large number of combined features. It is appropriate that a person skilled in the art will also consider these features individually and combine them into meaningful other combinations.
[0018] in:
[0019] Figure 1 shows a schematic diagram of a detection device according to the present invention;
[0020] Figure 2 A schematic flow chart showing a method according to the present invention; and
[0021] Figure 3 A schematic division of a data set of example data during a machine learning process according to the method of the invention is shown. DETAILED DESCRIPTION
[0022] Figure 1 A detection device 12 is shown. The detection device 12 is shown, which is arranged on the surface of a detection background 14 (for example a wall). The detection device 12 is configured to non-destructively detect detection signals of an object 16 arranged within the detection background 14. The detection device 12 is preferably designed to be hand-held, preferably can be held in one hand and in particular can be operated with the same hand. Particularly preferably, the detection device has a diameter of less than 7000 cm 3 , preferably less than 5000cm 3 , particularly preferably less than 3000cm 3 The detection device 12 comprises a sensor unit 24 for detecting the detection signal. The sensor unit 24 preferably comprises at least one sensor element, which is configured to: emit an electromagnetic signal into the inspection background 14; and detect a portion of the emitted signal backscattered by the object 16 as a detection signal. Alternatively or additionally, the sensor unit 24 comprises a capacitive sensor and / or an inductive sensor for detecting the detection signal. The detection device 12 comprises a calculation unit 26 for evaluating the detection signal to generate a detection result.
[0023] The detection device 12 preferably comprises an output unit 28 for outputting the detection result. The output unit 28 preferably comprises a display for displaying the detection result. Preferably, the detection device 12 comprises at least one data interface 32 for exchanging data with an external computing system and / or an external output device. Preferably, the detection device 12 comprises an energy supply unit 34 for supplying electrical energy to the sensor unit 24, the computing unit 26, the output unit 28 and / or the data interface 32. The energy supply unit 34 is particularly configured to accommodate at least one battery and / or storage battery. Preferably, the detection device 12 comprises at least one operating element 30, such as a switch, a button, a slider, etc., for operating the detection device 12, in particular for triggering detection using the detection device 12. The detection device 12 comprises a housing 36, within which at least the sensor unit 24 is arranged. Preferably, the computing unit 26, the data interface 32 and / or the energy supply unit 34 are arranged in the housing 36. Preferably, the output unit 28 and / or the operating element 30 are arranged on the housing 36 or embedded in the housing 36. The external computing system and / or the external output device are preferably designed independently of the detection device 12 (not shown here).
[0024] Figure 2 A flow chart of a method 10 for detecting a device 12 is shown. The method 10 preferably comprises a detection step 40, in which the sensor unit 24 detects the detection signal. The method 10 comprises an evaluation step 42, in which the detection signal is evaluated according to a machine learning process to generate a detection result. The method 10 particularly comprises an output step 44, in which the detection result is output. The method 10 preferably comprises a setting step 46, in which the computing unit 26 changes the settings of the detection device 12, in particular the sensor unit 24 and / or the computing unit 26, according to the detection result.
[0025] The method 10 preferably comprises an application phase 38. Preferably, the application phase 38 is fully executed by means of the detection device 12, alternatively partially executed. The application phase 38 preferably comprises a detection step 40, an evaluation step 42, an output step 44 and / or a setting step 46. In particular, the evaluation step 42 is performed by the computing unit 26. Alternatively, the evaluation step 42 is performed by an external computing system, in particular the external computing system already mentioned or another external computing system. The output step 44 is preferably performed by the output unit 28. Alternatively, the output step 44 is performed by an external output device. In particular, in an alternative design, the data interface 32 transmits the detection signal to the other external computing system, receives the detection result from the external computing system and / or sends the detection result to the external output device.
[0026] The method 10 preferably comprises a learning phase 18. Preferably, the learning phase 18 is performed by the external computing system, alternatively by the computing unit 26. In the learning phase 18, a model is created by means of a machine learning process, which maps the detection signal to the detection result. In a providing step 58, the model is stored in a memory of the computing unit 26 of the detection device 12. In particular, in an evaluation step 42, the computing unit 26 applies the model in order to transform the measured values of the detection signal detected by the sensor unit 24 into a detection result.
[0027] The learning phase 18 preferably includes a data collection step 48. In the data collection step 48, the example values 22 are detected (see Figure 3 ), these example values are processed to create the model. Preferably, the method 10 includes a collection step 56, in which the measured values of the detection signal detected by the sensor unit 24 are collected as example values 22. Preferably, in the collection step 56, the data interface 32 transmits the measured values (especially collected or individual measured values) detected by the sensor unit 24 to the external computing system. Particularly preferably, the external computing system collects measured values from a plurality of detection devices, especially within the framework of the Internet-of-Things (IoT) concept. The data set consisting of the example values 22 of the detection signal consists at least essentially of user data. User data are measured values of the detection signal, which are detected by the user in the application phase 38 with the aid of the detection device 12. Preferably, in the data acquisition step 48, the example values 22 are additionally determined by the manufacturer of the detection device 12 as a representative. These representatives are determined, for example, by carrying out detections with the detection device 12 on replicas of different examination backgrounds 14 including different objects 16 and / or are created by simulating the detection step 40 on different examination backgrounds 14 including different objects 16. Preferably, the data set comprises at least 10 times, preferably at least 100 times, particularly preferably at least 1000 times more user data than the representatives.
[0028] The learning phase 18 preferably comprises a preprocessing step 50. In the preprocessing step 50, the example value 22 is evaluated with respect to at least one feature 60, 62, preferably a plurality of features 60, 62 (see Figure 3). The at least one feature 60, 62 is preferably provided as a distinguishing criterion, according to which the external computing system evaluates whether two example values 22 are the same or different. The features 60, 62 of the example values can be, for example, the intensity of the detection signal, the ratio of the backscattered part to the emitted part of the detection signal, the average value of the detection signal, the extreme value of the detection signal, the duration of the amplitude modulation of the detection signal caused by the object, the fluctuation amplitude of the detection signal, and the like. The example values 22 of the detection signal are divided into different groups 20 by the external computing system according to the cluster formation of the example values 22. The external computing system preferably evaluates the cluster formation of the example values 22 based on the determined features 60, 62. In particular, the external computing system determines whether the distribution of the example values 22 within the parameter space expanded by the at least one feature 60, 62 has clustering. The dimension of the parameter space is equal to the number of different features 60, 62 for each example value 22. The distribution of the example values 22 in the two-dimensional parameter space expanded by the feature 60 and the further feature 62 is exemplarily in Figure 3 Preferably, the external computing system assigns the example values 22 forming the clusters to the same group 20. Preferably, the external computing system assigns the example values 22 forming different clusters to different groups 20. Figure 3 In FIG. 2 , the example values 22 are exemplarily divided into ten groups 20 . The external computing system uses nearest neighbor classification in order to perform the division of the example values 22 into the different groups 20 .
[0029] The method 10 preferably comprises an assignment step 52. In the assignment step 52, at least one evaluation parameter is assigned to the example values 22 of the detection signal in groups. Preferably, at least one evaluation parameter is assigned to each group 20. Optionally, a plurality of evaluation parameters are assigned to at least one group 20. Two different groups 20 differ from each other in at least one evaluation parameter assigned to them. The at least one evaluation parameter is processed by a machine learning process, in particular as an intermediate result to be learned or as an end result to be learned. Preferably, background parameters are assigned to these groups 20 as evaluation parameters, which background parameters describe or characterize the inspection background 14. The background parameters include, for example, the following values: concrete, lightweight structure, drywall, in particular the thickness and / or number of plasterboards and / or wooden boards of lightweight structure or drywall, brick wall, in particular the type of stone of the brick wall, the presence of floor heating or wall heating, etc. Preferably, object type parameters are assigned to these groups 20 as evaluation parameters, which object type parameters describe or characterize the type of object 16. The object type parameter includes, for example, the following values: metal, non-metal, current-carrying, non-current-carrying, magnetic, non-magnetic, low-voltage cable, single-phase AC cable, especially between 110V and 230V, multi-phase AC cable, three-phase cable, wooden beam, metal support, plastic pipe, water-filled pipe, especially fresh water pipe, non-water-filled pipe, especially sewage pipe. Preferably, a depth parameter is assigned to these groups 20 as an evaluation parameter, which describes or characterizes the distance of the object 16 from the surface of the inspection background 14. The depth parameter can indicate the distance continuously or locally. Preferably, a shape parameter is assigned to these groups 20 as an evaluation parameter, which describes or characterizes the shape or extension of the object 16. The shape parameter indicates, for example, the diameter of the object 16. The assignment step 52 is performed by the external computing system with the aid of a representation. The evaluation parameters of these representatives can be detected on replicas by the manufacturer of the detection device 12, for example, using a sensor device independent of the detection device 12, read out from a data table about these replicas or read out from a simulation program, and stored in the memory of the external computing system. Preferably, the external computing system determines at least one representative for each group 20, reads out the evaluation parameters of the representative from a memory of the external computing system, and assigns these evaluation parameters to all example values 22 in the same group 20 as the representative. If there are multiple representatives with different evaluation parameters in the same group 20, the external computing system optionally performs a further division of the groups 20 into subgroups.
[0030] The learning phase 18 comprises in particular a learning step 54. In the learning step 54, the external computing system creates the model by means of the machine learning process. In particular, the external computing system uses the example values 22 as input values of the model and uses the evaluation parameters assigned to the example values 22 as output values of the model. The detection result in particular consists of the totality of the evaluation parameters or is determined as a function of these evaluation parameters. In a provision step 58, the external computing system transmits the model to the computing unit 26 of the detection device 12 via the data interface 32.
[0031] In an application phase 38, the computing unit 26 of the detection device 12 uses the model in order to automatically select an operating mode of the detection device 12. Preferably, in a detection step 40, the sensor unit 24 detects an initial detection signal, which is evaluated in an evaluation step 42 to generate an initial result. Based on the initial detection signal, the computing unit 26 determines in particular the background parameter with the aid of the model in order to automatically select a zero reference for the detection signal according to a machine learning process in a setting step 46. The computing unit 26 determines in particular the object type parameter, the depth parameter and / or the shape parameter based on the initial detection signal with the aid of the model in order to increase the sensitivity of the sensor unit 24 to the object 16 and / or to avoid oversaturation of the sensor unit 24 in the setting step 46.
[0032] In the application phase 38, the detection device 12 outputs in an output step 44 at least one evaluation parameter of the group 20 of the example value 22 of the detection signal when the model assigns the measured value of the detection signal to the group 20. The output unit 28 preferably outputs: whether the object 16 is detected. The output unit 28 preferably outputs the object type parameter and the background parameter. Optionally, the output unit 28 outputs the depth parameter and / or the shape parameter.
Claims
1. A method for detecting a device, in, The detection device is configured to non-destructively detect a detection signal of an object (16) arranged within an inspection background (14), characterized in that in at least one method step, based on a machine learning process, the detection signal is assigned to a detection result in order to output the detection result and / or use the detection result to set the detection device.
2. The method according to claim 1, It is characterized in that In a learning phase (18) of the machine learning process, the example values (22) of the detection signal are divided into different groups (20) according to cluster formation of the example values (22).
3. The method according to claim 2, It is characterized in that The machine learning process uses nearest neighbor classification in order to perform the division into different groups (20).
4. The method according to any one of the preceding claims, It is characterized in that In a learning phase (18) of the machine learning process, at least one evaluation parameter is assigned in groups to the example values (22) of the detection signal.
5. The method according to any one of the preceding claims, It is characterized in that In a learning phase (18) of the machine learning process, the data set consisting of example values (22) of the detection signal consists at least substantially of user data.
6. The method according to any one of the preceding claims, It is characterized in that In at least one method step, an operating mode of the detection device is automatically selected based on the machine learning process.
7. The method according to any one of the preceding claims, It is characterized in that In at least one method step, a zero reference for the detection signal is automatically selected based on the machine learning process.
8. The method according to any one of the preceding claims, It is characterized in that In at least one method step, the detection device outputs an evaluation parameter of a group of example values (22) of the detection signal when a measured value of the detection signal is assigned to the group (20).
9. The method according to any one of the preceding claims, It is characterized in that At least one background parameter, at least one object parameter, at least one depth parameter and / or at least one shape parameter are assigned to the evaluation parameter.
10. A detection device, comprising at least one sensor unit (24) for detecting a detection signal, and comprising at least one calculation unit (26) for executing the method according to any one of the preceding claims.
