Method for analyzing image information and method for analyzing acoustic information using scalar values

Abstract

The invention relates to a method for analyzing image information (134, 234) using specified scalar values ​​(119, 129, 219, 229), comprising the method steps of: a.) detecting a first image (134) of an object (132, 232) or a scene (130) and detecting at least one first scalar sensor value (129) associated with the object (132) or the scene (130), b.) detecting a second image (234) of the object (132, 232) or the scene (130) and detecting at least one first scalar sensor value (129) associated with the object (132, 232) or the scene (130), c.) introducing the first image (134) and the at least one first scalar sensor value (129) as a consistent representation into a first data structure (159), and introducing the second image (234) and the at least one second scalar sensor value (229) as a consistent representation into a second data structure (259), d.) comparing the first data structure (159) and the second data structure (259), and e.) outputting information if the comparison results in a difference that corresponds to a predefined or predefined standard.

Description

Technical Field

[0001] The present invention relates to a method for analyzing image information or acoustic information using a specified scalar value, the method comprising the following method steps:

[0002] a.) detecting a first image or first acoustic information associated with an object or a scene, and detecting at least one first scalar sensor value associated with the object or the scene,

[0003] b.) Detecting a second image or second acoustic information related to the object or the scene, and detecting at least one second scalar sensor value related to the object or the scene. Background Art

[0004] Such methods are known from the prior art. For example, US Patent Application Publication No. US 2017 / 0032281 A1 discloses a method for monitoring a welding system, in which various sensor or image data are detected to train a neural network. Furthermore, this publication discloses a method in which the properties of the resulting weld seam or welding process are predicted using the trained neural network and image or sensor data recorded within the context of the welding process.

[0005] A disadvantage of the aforementioned prior art is that creating and using neural networks is very complex. Creating and setting up neural networks, but especially training them, is very complex and requires a large amount of training data for reliable operation. This training data is often not available, or at least not in sufficient quantities. Summary of the Invention

[0006] It is therefore an object of the present invention to provide a method for evaluating sensor data, with which method even relatively complex sensor data can be evaluated with reduced effort.

[0007] This object is achieved by a method having the features of the invention.

[0008] This method is designed and configured for analyzing image information using a specified scalar value and comprises the following method steps:

[0009] a.) detecting a first image of an object or scene and detecting at least one first scalar sensor value associated with the object or scene,

[0010] b.) detecting a second image of the object or scene and detecting at least one second scalar sensor value associated with the object or scene,

[0011] c.) introducing the first image and the at least one first scalar sensor value as a consistent representation into a first data structure, and introducing the second image and the at least one second scalar sensor value as a consistent representation into a second data structure,

[0012] d.) comparing the first data structure to the second data structure, and

[0013] e.) If the comparison results in a difference corresponding to a predefined or predefinable standard, a message is output.

[0014] By creating a consistent representation of the corresponding data structures, it is possible to combine image information and associated scalar information within a consistent representation or data structure, thereby implementing entirely new mechanisms by comparing two or more such data structures. This makes it possible to analyze even more complex sensor data without necessarily having to use neural network evaluation.

[0015] In one possible embodiment, the comparison of the first and second data structures or the check whether a predetermined or predefinable criterion is met can be performed not only manually or using automated analysis methods, but also using a neural network. However, such a neural network may be easier to design and train than a neural network that is directly required for analyzing image and sensor data.

[0016] Here, a scalar value can be, for example, an alphanumeric value, an alphanumeric string, an integer value, a floating point value, a Boolean value, a string value, or the like. A scalar value represents a 0-dimensional arrangement of values, in contrast to a vector, which represents a 1-dimensional arrangement of values, and a matrix, which represents a 2-dimensional or even higher-dimensional arrangement of values.

[0017] For the purposes of this specification, image information or images are considered to be matrices, i.e., images are values ​​constructed in a matrix, wherein the individual elements of the image are, for example, its pixels, which can be characterized by lengths in two or more dimensions and one or more brightness and / or color values.

[0018] In this case, the detected images can each be designed and configured, for example, as two-dimensional or higher-dimensional pixels or vector graphics.

[0019] The image information or image can also have more than two dimensions, for example by superimposing further two-dimensional information layers onto the two-dimensional image layer. Thus, for example, a further layer can be superimposed onto the two-dimensional grayscale, brightness value, and / or color value pixel image, in which further layer, for example, material density values, X-ray information values, or similar other characteristic values ​​are assigned to the individual pixels.

[0020] For example, the first image and / or the second image can be recorded using any type of suitable camera. For example, the recording can be performed in different optical frequency ranges, i.e., in the visible range, the infrared range, the ultraviolet range, and / or the X-ray range, or in other suitable spectral ranges. The detection can be performed using suitable image detection methods and devices or cameras, respectively. Furthermore, the image can be detected using other suitable image detection methods, such as thermography methods (i.e., methods for two-dimensional temperature detection), MRI methods (magnetic resonance imaging), X-ray structure analysis methods, or other methods suitable for generating two-dimensional pixel or vector images.

[0021] Furthermore, the image can also be detected, for example, by means of a scanner, a screenshot (=a digital image of a two-dimensional representation displayed on a screen, a display or a monitor) or in a similar manner.

[0022] The first image and the second image may be, for example, temporally consecutive images of a scene or an object.

[0023] In addition, the first image and the second image can also be images of an object or scene recorded at a specific time point, and a corresponding reference image of the object or scene. Such a reference image can, for example, represent an original, target or desired design of the object or scene.

[0024] The image of a scenario can be, for example, an image of a specific spatial area or a specific logical scenario. A logical scenario can be characterized, for example, by specific criteria or triggering features or events. Thus, for example, within the context of a process flow or the production of a specific product, a specific process step or production step or a result of a specific process step or production step (e.g., an intermediate or final product or result) can be such a logical scenario.

[0025] For example, the context can also be defined by a geographical location and / or other characteristic properties, such as clock time, time range, specific brightness values, specific detected sensor data such as temperature, person identification or similar properties.

[0026] Thus, a scenario can be, for example, a specific situation in road traffic, such as a specific situation given by a specific location and spatial perspective, or a production scenario given by a specific production machine or a specific production step. Furthermore, a scenario can also be given by corresponding process flows, such as a combustion process, a transport process, interactions between different people, or similar processes.

[0027] An object may be any type of object, device, facility, machine, living being or other material object detectable with the aid of optical or other sensors.

[0028] The sensor value may be a value from a sensor associated with the production process, used to detect physical variables, material properties, chemical properties, identification information or other information about a product or a production facility or a portion thereof in the production process.

[0029] Here, a scalar sensor value may be any type of scalar value, such as a numeric value, an alphanumeric value, an alphanumeric string, an integer value, a floating point value, a Boolean value, a string value, or the like.

[0030] The scalar sensor value can be, for example, a number, one or more letters or words or other alphanumeric data. In this case, the sensor value can be output by the corresponding sensor.

[0031] In this case, the detection of the at least one scalar sensor value can be designed and configured such that only one scalar sensor value is detected.

[0032] Furthermore, a plurality of values ​​provided by a sensor at different points in time can also be detected as at least one scalar sensor value (e.g., a so-called "time series"). Within the context of creating a corresponding data structure, for example, one, individual, or all of these values ​​can then be used, or even, for example, an average value of all or individual values ​​of these values ​​can be used.

[0033] The at least one first and second scalar sensor values ​​can also, for example, each relate to different aspects or parts of an object or a scene (e.g., the temperature at different locations on an object). If, for example, multiple sensor values ​​are detected as the at least one first or second sensor value, then one, some, or all of these values ​​can be used within the scope of creating the corresponding data structure, or an average of multiple or all values ​​can also be used.

[0034] The fact that both the image and the at least one scalar sensor value can relate to the same object can mean, for example, that both the image and the at least one scalar sensor value relate to at least one subregion of the object, wherein the subregion with respect to the image can differ from the subregion with respect to the sensor value. Thus, for example, an image can be recorded from a first part of the object, and a corresponding sensor value, such as a temperature value, can be recorded at another part of the same object.

[0035] The image and at least one scalar sensor value both relate to the same scene, which can mean, for example, that the corresponding sensor value is associated with an object and / or spatial region that is included or at least partially included in the recorded image. Thus, for example, if the scene corresponds to a specific road traffic scene, the sensor value can be, for example, the air temperature of the air present in the recorded area or a brightness value detected within the recorded area. Furthermore, sensor values ​​can also be logically assigned to the scene by, for example, assigning the clock time prevailing at the time of the recording as the sensor value to the corresponding traffic scene.

[0036] Taking production monitoring as an example, the image information may relate to an intermediate product, for example, and the associated sensor values ​​may relate to the production steps that lead to the intermediate product (e.g., the temperature or temperature profile during a production step, or the winding speed when producing a wound multi-layered battery). Such "intermediate products" are also examples of scenarios within the meaning of this specification. Thus, for example, a first image of the first intermediate product may be acquired using the sensor values ​​relating to this first intermediate product, and a second image of the second intermediate product may be acquired together with the sensor values ​​relating to the second intermediate product.

[0037] Furthermore, within the scope of production monitoring, for example, an image of a specific plant part or a specific production machine can be recorded, and corresponding sensor values ​​of sensors of this machine or this plant part can be recorded.

[0038] The corresponding scenario may also involve, for example, monitoring a vehicle, in particular an autonomously driven vehicle. Thus, for example, a camera located at or in the vehicle may record a specific image of the surroundings, and the corresponding sensor values ​​may be environmental parameters of the recorded environment, such as a humidity value for the roadway, weather or climate information, measured parameters regarding objects or persons in the environment, temperature, time of day, etc. Furthermore, in this context, the sensor parameters may also be information obtained by analyzing text or words in the recorded images, such as street and / or road sign markings.

[0039] The corresponding objects and / or scenes can also come from medical diagnosis, material testing, or other applications of the image evaluation method. Thus, for example, within the scope of medical diagnosis, sensor values ​​for a patient's blood values, their body temperature, acoustic parameters (e.g., for breathing noise, coughing noise, or the like) can be assigned to the recorded image.

[0040] In the context of this specification, a data structure is generally understood to be a representation of information in a computer or a corresponding electronic storage device ("in units of bits or bytes"). This also corresponds to the definition of the term "data structure" in computer science.

[0041] In this case, introducing the respective image and the corresponding at least one sensor value as a consistent representation into the first data structure means, for example, that the respective data sets are designed and configured such that all values ​​included in the respective data structure are characterized by one or more variables of a respectively uniform scale.

[0042] For example, the corresponding data structure can be designed and configured so that all included values ​​are represented by uniformly scaled variables. For example, a graphical representation as a number axis, a linear axis, or a comparable one-dimensional representation with the corresponding values ​​plotted thereon can be assigned to or associated with the data set designed in this way.

[0043] Furthermore, the data structures can also be designed and configured in such a way that all included values ​​are represented by two variables, each with a uniform scale. For example, a graphical representation as a two-dimensional diagram with two linear axes or two axes corresponding to the displayed values, or a comparable two-dimensional representation, such as a graph with the values ​​represented therein, can be assigned to such a data structure.

[0044] Furthermore, the data structures can also be designed and configured so that all included values ​​are represented by three or more variables that are scaled uniformly. A corresponding three-dimensional or higher-dimensional representation can then be assigned to or associated with such a data set.

[0045] The corresponding image and one or more sensor values ​​can be incorporated into a consistent representation of the data structure, for example, by means of at least the method flow described below. Thus, for example, in a first step, the recorded image can be subjected to corresponding image processing. In addition to adapting the color, brightness, contrast, and similar image parameters, this image processing can also include corresponding transformations and / or distortions or corrections of the image. Such transformations can include, for example, spatial transformations or also transformations into the frequency domain. Subsequently, for example, specific image points and / or different frequency components can be selected for incorporation into the data structure, or the entire image can be incorporated into the data structure.

[0046] Within the scope of introducing at least one sensor value into a consistent representation of the data structure, a first step can also include, for example, specific processing of the detected value, such as normalization or adaptation to a predetermined scale on which the representation is based. The corresponding value can then be introduced into the data structure accordingly, if necessary, after further adaptation of the scale or adaptation of the representation of the data structure.

[0047] In this way, a data structure is formed which combines the data of the captured image and the data of at least one sensor value captured in this respect within a unified representation.

[0048] For example, the first and second data structures can be compared by comparing individual or even all data points of the first data structure with the corresponding associated data of the second data structure. For example, the determined deviations of the points can then be added together, or an average value of the deviations can also be determined or calculated.

[0049] Furthermore, specific segments of the corresponding images can also be part of the comparison of the data structures. In this way, for example, specific parameters of individual segments of the first image can be compared with corresponding parameters of the corresponding segment of the second image, and associated deviations can be determined if necessary.

[0050] Part of the comparison of the first and second data structures may also include clustering of the corresponding data structures, wherein after the clustering, the determined cluster structures of the first and second data structures are then compared. This will be discussed in more detail below.

[0051] In this case, method step e.) can be designed and configured, for example, such that the output information is designed and configured as a warning message and the predefined or predefinable criterion corresponds to an error criterion.

[0052] A predefined or predefinable criterion for outputting information can be associated with each comparison value determined within the scope of the comparison of the first and second data structures.

[0053] Therefore, as already mentioned above, the cumulative or average deviation of the values ​​of the first data structure from the corresponding values ​​of the second data structure can serve as a predetermined or predefinable criterion. Furthermore, within the scope of the differences determined in the segments of the first and second data structures, or also within the differences in the determined cluster structures of the first and second data structures, specific limit values ​​can also correspond to predetermined or predefinable criteria for outputting information. Thus, for example, within the scope of clustered first and second data structures, a predetermined or predefinable criterion can be that the number of clusters determined in the first and second data structures differs, or that the clustering weights of the first and second data structures also differ by at least a specific predetermined value.

[0054] The output of information can, for example, be an indication that a corresponding significant change has been detected between the scene or object detected in the first image and the second image, respectively. This information can then, for example, be an instruction to the user to check the corresponding scene, production facility, or corresponding object, or to trigger similar inspection or intervention measures, or to forward corresponding information.

[0055] Furthermore, the output of information can also be the output of control commands or also the output of alarms or control messages. Here, the control commands or corresponding alarms or control messages can be forwarded to corresponding controllers, control devices, hosts, computers or similar devices, for example, which can automatically trigger corresponding actions.

[0056] In addition, the method according to the present specification can be designed and configured so that in order to create the first data structure and the second data structure, the detected images are respectively converted into the frequency domain. In particular, in order to create the first data structure and the second data structure, the detected images are respectively transformed into the frequency domain using Fourier analysis.

[0057] In this case, the Fourier analysis method can be designed and configured, for example, as a so-called "discrete Fourier transform," a so-called "fast Fourier transform," or also as a so-called "discrete cosine transform" (DCT). The "discrete cosine transform" (DCT) is also used for images within the scope of so-called JPEG compression (JPEG: Joint Photographic Experts Group) and is therefore an established method for converting images into the frequency domain.

[0058] The advantage of frequency transforming an image is that it makes it easier to identify certain structural features, such as clearly defined objects, line structures, and edges. Furthermore, frequency transforming allows for the representation of image information based on amplitude information associated with a specific frequency comb. By representing the individual waves of the frequency comb along spatial axes, image information can be represented along one or more spatial axes, allowing sensor values ​​to be plotted in the frequency domain relative to individual spatial sampling points. This allows for a unified representation of data structures in a simplified manner.

[0059] The method according to this specification can also be designed and configured so that

[0060] In method step a.), at least one first scalar parameter value associated with the object or the situation is also detected, and

[0061] In method step b.), at least one second scalar parameter value is also detected, which is associated with the object or the situation.

[0062] Therein, the first data structure and the second data structure are created by using at least one first scalar parameter value and / or at least one second parameter value.

[0063] Within the scope of method step c.), at least one first parameter value is also introduced into a consistent representation of the first data structure and, if necessary, is also converted into a corresponding representation or adapted thereto. Similarly, at least one second parameter value is also introduced into a consistent representation of the second data structure and, if necessary, is also converted into a corresponding representation or adapted thereto.

[0064] In this way, various other sensor values ​​or other values ​​related to the object or the situation can be introduced into the corresponding analysis, thereby improving the analysis and / or making it more robust or more sensitive—depending on the choice of parameters or comparison or evaluation methods.

[0065] A parameter value can be, for example, a sensor value, such as a value output by a sensor assigned to an object or context. Furthermore, a parameter value can also be any other value assigned to an object or context. Such a value assigned to an object or context can, for example, describe or relate to a characteristic, state, behavior, feature, or similar information associated with the object or context. A scalar parameter value can, for example, be a numeric value, an alphanumeric value, a Boolean value, and / or a string value.

[0066] Furthermore, the method according to the present description may be designed and configured such that the comparison between the first data structure and the second data structure is achieved using a neural network.

[0067] Here, the neural network can be designed and configured as a trained neural network.

[0068] The training of such a neural network can be performed so that a corresponding data structure with a uniform representation is created for each of the multiple combinations of image information with associated sensor or parameter information. This data structure can then be manually assigned to the corresponding evaluation results, for example. According to the present description, such evaluation results can, for example, correspond to comparative values ​​for comparing the corresponding data structure with corresponding other data structures. Furthermore, such evaluation results can correspond to good / bad or improvement / deterioration analyses, for example, for monitoring production or process flows within the context of using the method according to the present description.

[0069] Such a neural network can then be used, for example, to compare the first and second data structures, so that the first and second data structures are transmitted to the neural network in a correspondingly suitable manner according to methods known from the prior art, and the neural network outputs corresponding comparison values ​​or also good / bad values ​​or also improved / deteriorated values.

[0070] At least in the context of this description, a neural network is understood to be an electronic device comprising a network of so-called nodes, wherein each node is generally connected to a plurality of other nodes. Nodes are also referred to as neurons, units or cells, for example. In this case, each node has at least one input connection and one output connection. Such a node is understood to be an input node for a neural network, which can receive signals (data, stimuli, patterns or the like) from the outside world. An output node of a neural network is understood to be a node that can forward signals, data or the like to the outside world. So-called "hidden nodes" ("hidden nodes") are understood to be nodes of a neural network that are neither constituted as input nodes nor as output nodes.

[0071] Neural networks can generally be trained by determining parameter values ​​for individual nodes or their connections using various known learning methods by feeding input data into the neural network and analyzing the corresponding output data in the neural network. In this way, neural networks can be trained in a manner known per se using known data, patterns, stimuli, or signals, so that the network trained in this way can then be used to analyze further data.

[0072] In this case, for example, the neural network can be designed as a so-called deep neural network ("deep neural network" (DNN)). Such a "deep neural network" is a neural network in which the network nodes are arranged in layers (wherein the layers themselves can be one-dimensional, two-dimensional, or even higher-dimensional). In this case, the deep neural network includes at least one or two so-called hidden layers, which only include nodes that are not input nodes or output nodes. This means that the hidden layers have no connections to the input or output signals.

[0073] In this context, so-called “deep learning” is understood to mean, for example, a class of machine learning techniques that uses multiple layers of nonlinear information processing for supervised or unsupervised feature extraction and transformation as well as for pattern analysis and classification.

[0074] A deep neural network may also have a so-called autoencoder structure, which is explained in more detail in the course of this specification. Such an autoencoder structure may be suitable, for example, for reducing the dimensionality of data and thus for example for identifying similarities and commonalities.

[0075] Deep neural networks can also be designed, for example, as so-called classification networks, which are particularly suitable for classifying data into categories. Such classification networks are used, for example, in conjunction with handwriting recognition.

[0076] Another possible structure of a neural network with a deep learning architecture may be, for example, a design as a so-called “deep belief network”.

[0077] For example, a neural network with a deep learning architecture may also have a combination of multiple structures among the above structures. Thus, for example, a deep learning architecture may include an autoencoder structure to reduce the dimensionality of the input data, and the deep learning architecture may then be combined with other network structures to, for example, identify peculiarities and / or anomalies within the reduced dimensionality of the data, or to classify the reduced dimensionality of the data.

[0078] For example, a neural network with a deep learning architecture can be trained using one of the so-called "supervised learning" methods. Here, the network is trained with the results or capabilities assigned to these data by training with the corresponding training data. In addition, a neural network can also be trained using the so-called unsupervised training method. This algorithm, for example, generates a model for a given amount of input, which describes the input and makes predictions from it. Here, for example, there is a clustering method that divides the data into different categories if the data differ from each other, for example, due to a representation pattern.

[0079] When training neural networks, it is also possible to combine supervised and unsupervised learning methods, for example if trainable properties or abilities are assigned to one part of the data, but not to another part of the data.

[0080] Furthermore, at least among other things, so-called reinforcement learning methods can also be used for training neural networks.

[0081] Generally, the training of a neural network is understood to mean that the data used to train the neural network are processed in the neural network with the aid of one or more training algorithms in order to calculate or change so-called bias values ​​("Bias"), weight values ​​("weights") and / or transfer functions ("Transfer Functions") of the neural network or of individual nodes of a connection between two nodes within the neural network.

[0082] The values ​​describing the individual nodes and their connections, including other values ​​describing the neural network, can be stored, for example, in a value set describing the neural network. This set then represents, for example, a design of the neural network. If this value set is stored after training the neural network, it then stores, for example, the design of the trained neural network. Thus, for example, it is possible to train the neural network using appropriate training data in a first computer system, then store the corresponding value set assigned to the neural network, and transfer it to a second system as the design of the trained neural network.

[0083] For example, training requiring a relatively high computing power of a corresponding computer can be performed on a high-performance system, while further work or data analysis can then be performed using the trained neural network on a low-performance system. Such further work and / or data analysis using the trained neural network can be performed, for example, on an auxiliary system according to the present description and / or a control device, a programmable logic controller, or a modular programmable logic controller.

[0084] In addition, the method according to the present specification can also be designed and configured so that the first data structure and the second data structure respectively have a two-dimensional or higher-dimensional chart structure or a two-dimensional or higher-dimensional graph structure, or the first data structure and the second data structure are respectively represented or can be represented as a two-dimensional or higher-dimensional chart or a two-dimensional or higher-dimensional graph.

[0085] In this context, within the scope of creating the first and second data structures, for example, a corresponding chart, a corresponding graph, or a corresponding chart or graph structure can be created first, and then a clustering method can be applied thereto. Furthermore, clustering can also be performed in parallel with the creation of the chart, graph, chart structure, or graph structure. As a result, the first and second data structures then already include the corresponding chart, graph, chart structure, or graph structure, as well as the corresponding clustering structure.

[0086] Alternatively or additionally, within the scope of the comparison of the first and second structures, a clustering method can be applied to the corresponding diagram, graph, diagram structure, or graphical structure. This can be done both when clustering has not yet been applied to the corresponding diagram, graph, diagram structure, or graphical structure, and when a clustering method as described above (so-called "hierarchical clustering") has already been applied to it within the scope of the data structure creation.

[0087] A chart structure or chart is understood to mean any structure that can be represented in a corresponding coordinate system, in particular any structure that can be represented as individual data points in a corresponding coordinate system. In this context, an N-dimensional chart structure or N-dimensional chart corresponds to a structure that can be represented as individual data points in an N-dimensional chart, for example.

[0088] In this context, a graph structure or graph is understood to mean any structure that can be represented as a corresponding graph. Such a graph or such a graph structure can be designed and configured such that, for example, the nodes of the graph correspond to individual values ​​of a data structure and are connected by corresponding connections, so-called "edges" of the graph. Such connections or edges can be designed and configured such that, for example, all points are connected to all other points, a point is only connected to a maximum number of arbitrary points, and / or a (sampling) point from the frequency domain is necessarily connected to all samples of other sensor values, in particular to a maximum predetermined or predeterminable number of other samples of the frequency domain.

[0089] Furthermore, in a corresponding graph or a corresponding graph structure, a part of the value of a specific node can also correspond to a part of the data structure, and another part of the data structure to the associated edge.

[0090] This embodiment of the method has the advantage that there is a very efficient and established method for evaluating such diagrams and / or graphs and comparing them with one another. In this way, the comparison of data structures and the corresponding matching of comparison results with predeterminable or predefined criteria can be further simplified.

[0091] Cause analysis ("root cause analysis" in English) is very important, for example, for industrial production processes and can be much simpler within the scope of this method, precisely when using the method according to the present description, in particular when using the clustering method according to the present description. Within the scope of such an application, for example, changes in the clusters can then be understood, for example by tracking node movements within the corresponding graph. Within the scope of the method according to the present description, nodes can indeed be derived directly from sensor values ​​or harmonics in the frequency domain. And then, for example, such nodes can be assigned to corresponding sensor values ​​or harmonics in the frequency domain from which the nodes are derived. If, for example, a specific node causes a change in the cluster structure, then, for example, the time series / image portion behind it can be directly identified as the cause.

[0092] In general, a graph is a mathematical structure consisting of so-called "nodes" and so-called "edges" connecting two nodes. A graphical representation of such a graph can be, for example, a representation in which nodes are represented as points or circles and edges are represented as lines connecting the circles.

[0093] Here, an edge can be, for example, a so-called "undirected edge," in which no logical direction is assigned to the connection of the corresponding node. Furthermore, an edge can also be designed as a so-called "directed edge," in which a logical direction or meaning is assigned to the connection of the corresponding node.

[0094] The method according to the present description can also be designed and configured so that, within the scope of creating the first data structure and the second data structure according to method step c.), the clustering method is applied or has been applied to the corresponding charts, graphs, chart structures or graphic structures of the first data structure and the second data structure, respectively, or, within the scope of comparing the first data structure and the second data structure according to method step d.), the clustering method is applied to the corresponding charts, graphs, chart structures or graphic structures of the first data structure and the second data structure, respectively.

[0095] The comparison of the first and second data structures according to method step d.) or the creation of the first and second data structures according to method step c.) can be performed, for example, by applying one or more clustering methods to corresponding charts, graphs, chart structures or graph structures of the first and second data structures.

[0096] After such a clustering, the clusters, cluster structures, cluster properties or the like identified on the basis of the above clustering can then also be compared, for example within the scope of method step d.).

[0097] Furthermore, in this context, the preset or pre-settable criteria for triggering the output of information according to method step e.) may include either a criterion regarding the difference between a plurality of identified clusters, a criterion regarding one or more positional differences between the identified clusters, or also a criterion regarding other differences in the characteristics, number and orientation of the respectively identified clusters.

[0098] The application of clustering methods can be, for example, automatic clustering methods. In this case, data structures can be clustered, for example, by means of corresponding software that automatically implements the process of clustering methods. In this case, one or more clustering algorithms can be implemented, for example, within the scope of software.

[0099] Furthermore, the clustering method can also be applied as a semi-automatic clustering method, for example. This can be achieved, for example, by means of corresponding software that semi-automatically implements the clustering method process. This can be achieved, for example, by requiring the software to expect corresponding user input at certain points in the clustering method process.

[0100] The application of clustering methods may include, for example, the application of one clustering algorithm or the application of multiple clustering algorithms, for example, used in sequence. Such clustering algorithms may be, for example, so-called "K-Means clustering", so-called "Mean-Shift clustering", so-called "Expectation Maximization (EM) clustering using a Gaussian Mixture Model (GMM)", so-called "Agglomerative Hierarchical Clustering" and / or so-called "Density-Based Spatial Clustering", such as Density-Based Spatial Clustering of Applications with Noise (DBSCAN). Other examples of clustering algorithms may include, for example, the following algorithms: "Mini Batch K-Means", "Affinity Propagation Clustering", "Mean Shift", "Spectral Clustering", "Ward Sum of Squares Method", "Agglomeration Clustering", "Hierarchically Balanced Iterative Clustering (Birch)", and "Gaussian Mixture".

[0101] Thus, a cluster is considered to be a group of similar data points or data groups formed by a corresponding cluster analysis or a corresponding cluster.

[0102] Clustering is generally understood to be a so-called "machine learning" technique in which data or data points are grouped into so-called "clusters." Within the context of a set of data or data points, cluster analysis, clustering methods, or clustering algorithms can be used, for example, to classify each data point or each data point, or even individual data points, into specific groups. These groups are then referred to as "clusters." Data or data points in the same group (i.e., the same cluster) have similar properties and / or characteristics, while data points in different groups have very different properties and / or characteristics.

[0103] Mathematically, a cluster consists of objects that are closer to each other (or, conversely, more similar) than objects in other clusters. Clustering methods can be distinguished based on the distance or proximity metric used between clustered objects, but also between entire clusters. Additionally or alternatively, clustering methods can also be distinguished based on the corresponding calculation rules for such distance metrics.

[0104] Cluster analysis or clustering methods are understood to be methods for discovering similar structures in large databases. This includes, for example, supervised or unsupervised machine learning (e.g., k-means or DBSCAN). The result of cluster analysis is clusters. The advantage here is that the data analysis can be performed fully automatically. Supervised learning is useful when the data is already contextualized. Unsupervised learning algorithms can also find similar structures in data that has not yet been contextualized. The found clusters can then be analyzed by domain experts.

[0105] Here, when performing a clustering method or clustering algorithm according to the present specification, various common distance metrics or similarity metrics can be used on numerical data, binary data, string data, categorical data, text data, and / or time series data, depending on the type of data category used.

[0106] Examples of such clustering methods or algorithms are:

[0107] - so-called "unsupervised clustering",

[0108] - the so-called K-means clustering method,

[0109] - a method from image processing for identifying commonly shared structures within a current image or image information,

[0110] - A combination of the above methods.

[0111] The clustering method used can be selected in this case in a manner adapted to the type of data within the current data.

[0112] Furthermore, the method according to the present description can be designed and configured such that the method for monitoring a method process or a production process is designed and configured such that

[0113] - the detection of the first image and / or the second image is designed and configured to detect the first image and / or the second image of an object related to the production process or a scene related to the production process, and

[0114] - the detection of at least one first scalar sensor value and / or at least one second scalar sensor value is designed and configured as the detection of at least one first scalar sensor value and / or at least one second scalar sensor value relating to an object related to the production process or a situation related to the production process.

[0115] With the aid of such a method, for example, the course of a method or production or specific production steps can be efficiently analyzed and / or monitored. Such a method allows image information and also sensor information relevant to the method or production to be included in the analysis, thereby enabling a very good and / or comprehensive detection and characterization of the method or selected method steps or production steps.

[0116] Furthermore, for example, within the scope of the comparison of the data structures determined here, the data can be analyzed completely using a neural network, but this neural network is not absolutely necessary. This method thus further simplifies the analysis of the production method compared to methods known from the prior art.

[0117] In this context, the predefined or predefined criteria for outputting the corresponding information can be selected, for example, so that information is only output when a faulty state or faulty product, or even a dangerous state, is likely to occur within the process or production process. The information can then be, for example, a corresponding warning message or a corresponding control command, such as shutting down a specific process part or area or production section, or switching to a safe state. Furthermore, the corresponding information can also be an alarm message, which can then be output, processed, and / or disseminated, for example, via a corresponding alarm system.

[0118] In this context, the detected image can be designed as an image of the final product or intermediate product of the production method. In addition, the image can be designed as an image of a facility or a facility or a facility part or a facility or a facility part of the equipment used within the scope of the production method.

[0119] The corresponding sensor values ​​can be, for example, sensor values ​​that characterize the production method, such as the temperature or temperature trend of a furnace, the transport speed of a track processed within the scope of the production method, power values ​​or consumption values ​​of a method step running within the scope of the production method, or similar sensor values ​​related to the production method or production steps that, for example, result in an intermediate product or final product.

[0120] Furthermore, the image can be assigned to a method sequence, for example, an image already recorded or recorded in the combustion chamber of a gas turbine can be assigned to a method sequence. With the aid of such an image, anomalies in such a combustion process can be detected using the method according to the present description, for example.

[0121] In this context, for example, a time series of the gas turbine power can be used as the at least one assigned sensor value. In general, any sensor value that characterizes or is derived from the method sequence can also be used as the at least one assigned sensor value. This can be, for example, the time series already mentioned, or individual values ​​of power, speed, temperature, or other measured variables that characterize a method step.

[0122] Furthermore, the method according to the present description may be designed and configured such that a production process is designed and configured for manufacturing a product and comprises a sequence of production steps, wherein after completion of a selected production step from the sequence of production steps, an intermediate product exists,

[0123] - and also designing and configuring the first image as a first intermediate product image,

[0124] - at least one first scalar sensor value relates to a selected production step,

[0125] - designing and configuring the second image as a second intermediate product image, and

[0126] - At least one second scalar sensor value relates to a selected production step.

[0127] Furthermore, the above object is achieved by a method for monitoring a production process for producing a product, wherein the production process comprises a sequence of production steps, wherein an intermediate product is present after expiration of a production step selected from the sequence of production steps, the method comprising the following steps:

[0128] a.) detecting an image of a first intermediate product and detecting at least one first scalar sensor value associated with a selected production step;

[0129] b.) detecting an image of a second intermediate product and detecting at least one second scalar sensor value associated with a selected production step;

[0130] c.) importing the first image and the at least one first scalar sensor value as a consistent representation into a first data structure, and importing the second image and the at least one second scalar sensor value as a consistent representation into a second data structure;

[0131] d.) comparing the first and second data structures, and

[0132] e.) If the comparison results in a difference corresponding to a predefined or predefinable standard, a message is output.

[0133] In this case, method step e.) can be designed and configured, for example, such that the outputted information is designed and configured as a warning message and / or a predefined or predefinable criterion corresponds to an error criterion.

[0134] Furthermore, the first image may be designed and configured as a first intermediate product image of a first intermediate product existing at a first time point, and the second image may be designed and configured as a second intermediate product image of a second intermediate product existing at a second time point.

[0135] In the context of discrete production of a continuous series of individual products, for example, the first and second points in time can be selected such that a first intermediate product is present at the first point in time and an intermediate product immediately following the first intermediate product is present at the second point in time, so that each intermediate product is compared with the preceding intermediate product within the production sequence. The first and second points in time can also be selected such that not every intermediate product is compared with the preceding intermediate product, but rather only every second, fifth, tenth, or other intermediate product is considered.

[0136] In the context of continuous production, the first and second points in time can be spaced apart, for example, so that they correspond to typical times for corresponding process sequence changes. This can be, for example, a control time constant for a process involved in the method or also for an apparatus involved in the method, such as a furnace, a heating device, a cooling device, a combustion device, a conveyor belt, a machine tool, a processing machine, or the like.

[0137] Based on the image detection of the first image and the second image at the first and second time points, at least one first scalar sensor value can relate to a selected production step, and at least one further first time point can be detected based on the first time point. Furthermore, at least one second scalar sensor value can relate to a selected production step, and at least one further second time point can be detected based on the second time point.

[0138] For example, at least one further first point in time can be selected based on the first point in time so that the corresponding recorded sensor value is detected, for example, within the context of the production of the intermediate product detected in each case via the image. Furthermore, the at least one further first point in time can be selected so that, for example, during the course of the analysis method, sensor values ​​are recorded at multiple points in time that are causally related to the state detected in the image. The same applies to at least one further second point in time based on the second point in time with respect to the relationship between the second sensor value and the detection of the second image.

[0139] More generally, at least one further first point in time or second point in time based on the first point in time or the second point in time can be selected such that at least one sensor value detected respectively is causally related to the scene or the object detected using the first image and / or the second image.

[0140] A sequence of production steps may, for example, consist of one or more production steps.

[0141] In this case, the production steps of a production process can be designed and configured, for example, so that each production step is performed by a respective production machine and / or production device. Furthermore or additionally, a production step can be characterized, for example, by a specific setting of parameter values ​​or a specific sequence of parameter values ​​(e.g., a heating process or a cooling process).

[0142] The sequence of production steps allows, for example, a so-called discrete process, in which a series of individual products (e.g., cars, mobile phones, etc.) are manufactured. In this context, an intermediate product can be, for example, an intermediate product that is present after a specific production step and has completed the corresponding production step at a specified point in time.

[0143] Furthermore, a so-called continuous process can be achieved by a sequence of production steps, in which a specific material or substance is continuously produced, for example. In such a designed method, an intermediate product can then be an intermediate product that is present at a specified time point after or during a method step.

[0144] In addition, a so-called batch process can also be realized by the sequence of production steps, which represents a mixture of a discrete process and a continuous process to a certain extent. Here, intermediate products can be used according to the above explanation about the discrete process and the continuous process.

[0145] Furthermore, the method according to the present description can be designed and configured such that in method step c.) the first data set is created using at least one further first production parameter value and the second data set is created using at least one further second production parameter value.

[0146] Within the scope of method step c.), at least one first production parameter is also introduced into the consistent representation of the first data structure and, if necessary, converted into a corresponding representation or adapted thereto. Similarly, at least one second production parameter is also introduced into the consistent representation of the second data structure and, if necessary, converted into a corresponding representation or adapted thereto.

[0147] In this way, various other sensor values ​​or other values ​​relating to the production sequence, the process sequence, the corresponding production steps or equipment, starting materials or conditions or other parameters characterizing the production or the process can be incorporated into the corresponding analysis.

[0148] Thus, for example, an analysis according to the present description can be further improved and / or made more robust or more sensitive—for example by means of the selection or comparison or evaluation method of corresponding parameters.

[0149] A production parameter may be any characteristic of a product of a treatment or processing or also of a production facility that is present, for example, at a first or second point in time during a selected production step.

[0150] For example, at least one further first or second production parameter can also relate to a selected production step and / or be recorded or present at at least one further first or second point in time.

[0151] Furthermore, at least one further first or second production parameter can relate, for example, to a selected production step and / or be detected or present at at least one further first or second point in time.

[0152] In this case, the production parameters can be detected, for example, as measured values ​​or time series during the execution of a selected production step. Furthermore, the production parameters can also be other values ​​that are related to the selected production step, the product processed or handled during the selected production step, or the production facility, each during the execution of the selected production step within a time period.

[0153] Furthermore, the method according to the present specification can be designed and configured to monitor the movement of a vehicle so that

[0154] - the detection of the first image and / or the second image is designed and configured as the detection of the first image and / or the second image of a portion of the surroundings of the vehicle, and

[0155] The detection of the at least one first sensor value and / or the at least one second sensor value is designed and configured as the detection of the at least one first sensor value and / or the at least one second sensor value related to the vehicle or the environment of the vehicle.

[0156] In this case, the vehicle can be designed and configured as a so-called autonomous vehicle, for example. Such an autonomous vehicle can be designed and configured as a so-called "AGV" ("Automated Guided Vehicle"). Generally, an autonomous vehicle is considered to be a vehicle that moves without or substantially without the continuous action of a human driver, or even without a human driver at all.

[0157] Within the context of monitoring the movement of a vehicle, the method according to the present description can be designed and configured, for example, so that, for example, an autonomous vehicle records images of its environment or a portion of its environment (e.g., in the forward direction) at regular time intervals, and simultaneously uses, for example, measured values ​​from corresponding proximity sensors of the autonomous vehicle as sensor values. In this way, it is possible, for example, to detect and analyze the presence or appearance of objects, obstacles, lane markings, or similar objects or markings in the vehicle's environment through continuous such image / sensor recordings, and to derive corresponding results, particularly also with respect to the vehicle's direction of travel.

[0158] Thus, for example, within the scope of comparing the first and second data structures created accordingly and subsequently outputting corresponding information according to corresponding predefined criteria, the method can be designed and configured so that, for example, the vehicle can avoid corresponding obstacles or, for example, can follow corresponding signs.

[0159] In this way, for example, the safe forward movement of an autonomous vehicle can be achieved or supported within the scope of industrial facilities or applications or also in public road traffic. In addition, the vehicle driver can also be supported in a similar manner, for example in the event of an unexpected obstacle or to maintain the corresponding lane.

[0160] In the context of monitoring vehicle movement, for example, predefined or predefined criteria for outputting information can be selected so that corresponding information is output when a dangerous situation occurs, such as when an object or person appears in the planned route. This information can be, for example, a warning message to the driver or a control message to the vehicle or vehicle control unit, which can be designed to, for example, stop the vehicle, reduce its speed, and / or change the route.

[0161] Furthermore, the method according to the present specification may be designed and configured such that, in order to analyze an image of an object or a living being, the method is designed and configured such that

[0162] - the detection of the first image and / or the second image is designed and configured as the detection of the first image and / or the second image of an object or a living being, or as the detection of a region of an object or a living being, respectively, and

[0163] - The detection of at least one first sensor value and / or at least one second sensor value is designed and configured as the detection of at least one first sensor value and / or at least one second sensor value related to an object or a living being, or is designed and configured as the detection of an area of ​​an object or a living being, respectively.

[0164] For example, an image of an object can be used to check its quality, for example, by checking for adherence to a correct shape or for variations in the object's shape. Furthermore, an object can be checked for counterfeit status, for example, by checking for corresponding differences from a similarly existing authentic object or an image of an authentic object using the method according to the present description. In this context, the corresponding sensor value can be, for example, the object's temperature, color value, surface properties, or other properties. Furthermore, it can be, for example, a corresponding optical image of the object or an X-ray image or other image.

[0165] The analysis of biological images can be used, for example, within the context of medical diagnosis of humans or animals. For example, optical images, X-ray images, multispectral refractive topography (MRT) images, or similar images can be used. Corresponding sensor values ​​can be, for example, body temperature, pulse rate, pulse, acoustic values ​​(e.g., respiratory or lung noise values, acoustic values ​​recorded during coughing), electrocardiogram (EKG) values, skin color, blood supply values, or similar sensor values.

[0166] Within the scope of methods for analyzing images of objects or living beings, predefined or predefined criteria for outputting information can be designed and configured, for example, so that corresponding warning messages are output if, for example, an object is defective or in an unnatural state, or if a living being is ill. This information can, for example, be or include corresponding warning messages for a user or corresponding automatic warning messages for an alarm system. Furthermore, the information can also be corresponding control commands or control messages that, for example, automatically generate a specific state or shut down a critical device.

[0167] The above object is also achieved by a method for analyzing acoustic information using a specified scalar value, the method comprising the following method steps:

[0168] a.) detecting first acoustic information related to a sound source and detecting at least one first scalar sensor value related to the sound source,

[0169] b.) detecting second acoustic information associated with the sound source or the second sound source and detecting at least one second scalar sensor value associated with the sound source or the second sound source,

[0170] c.) transforming the first acoustic dataset into the frequency domain and introducing the frequency transformed first acoustic dataset and the at least one first scalar sensor value as a consistent representation into a first data structure, and transforming the second acoustic dataset into the frequency domain and introducing the frequency transformed second acoustic dataset and the at least one second scalar sensor value as a consistent representation into a second data structure,

[0171] d.) comparing the first data structure to the second data structure, and

[0172] e.) If the comparison results in a difference corresponding to a predefined or predefinable standard, a message is output.

[0173] Acoustic information can be, for example, any acoustic recording and / or all other acoustic data that is stored or can be stored in an electronic storage device or can also be converted into an electronically storable format. Such acoustic information can be, for example, a sound recording, acoustic data detected by a microphone, or acoustic data detected in another way (e.g., optically or by pressure measurement).

[0174] The storage of such acoustic information can be carried out in any suitable format, for example in the so-called “WAV” format (or as a “.wav” file), the so-called “MP3” format (MPEG-1 Audio Layer 3; or as a “.mp3” file), the so-called “WMA” format (Windows Media Audio; or as a “.wma” file), the so-called “AAC” format (Advanced Audio Coding; or as a “.aac” file)), the so-called “OGG” format (Ogg Vobis; or as a “.ogg” file), the “FLAC” format (Free Lossles Audio Codec; or as a “.flac” file), the “RM” format (Real Media; or as a “.rm” file) or any suitable, similar format.

[0175] The sound source or second sound source may be, for example, a machine, a motor, a specific situation, a geographical location, one or more living beings or any other source of acoustic information.

[0176] The corresponding acoustic information can be, for example, a sound recording associated with one of the sound sources or a specific portion thereof. This sound recording can be detected, for example, using a corresponding microphone. Furthermore, the acoustic information can also be detected using other suitable means, such as optical methods for vibration detection or methods for detecting pressure fluctuations.

[0177] For example, the corresponding sound recording can also be recorded only within a specific acoustic frequency range or using a specific acoustic filter. In this case, the acoustic data can originate from or be recorded by an acoustic sensor, or can consist of data from multiple sensors related to the same source.

[0178] The correspondingly assigned sensor data may be, for example, corresponding temperature data or power data (e.g., current or power consumption) of the corresponding machine or corresponding motor. Furthermore, the corresponding sensor data may be, for example, other sensor data related to the corresponding context (e.g., clock time, brightness value, temperature, humidity value, etc.), other sensor data related to geographic location (e.g., GPS value, location information, address, clock time, brightness value, temperature, humidity value, etc.), or other sensor data related to one or more living beings (e.g., quantity, temperature, clock time, etc.).

[0179] Here, at least one first and second scalar sensor value, a consistent representation of the first and second data structures, a comparison of the first and second data structures and an output of information (if the comparison results in a difference corresponding to a preset or pre-settable standard) can be configured and designed according to this description.

[0180] Furthermore, the above method may be designed and configured such that the method is also designed and configured according to one or more features of the present invention.

[0181] Furthermore, the above method can be designed and configured to monitor a method process or a production process so that

[0182] - the detection of the first acoustic information and / or the second acoustic information is designed and configured as an acoustic recording of a situation involving an object participating in the production process or related to the production process or the process flow, and

[0183] The detection of the at least one first scalar sensor value and / or the second scalar sensor value is designed and configured as detection of at least one value by means of a sensor, which at least one value relates to an object involved in the production process or a context related to the production process or the method flow.

[0184] In this case, sensors, the detection of corresponding sensor values, at least one first or second scalar sensor value, objects involved in the production process, and contexts related to the production or method process can be designed and configured according to this description.

[0185] Furthermore, objects involved in the production process may be, for example, devices or machines involved in production. Within the scope of the present method, corresponding acoustic information is recorded from these devices or machines, for example, via corresponding microphones or similar devices. Parallel to this acoustic recording, other parameters of the machine or equipment, such as power data, temperature, speed, control parameters, etc., can then be detected using corresponding sensors.

[0186] Furthermore, starting products, intermediate products, and / or end products present in the production process may be, for example, objects involved in the production process. For example, corresponding acoustic information (e.g., vibration data or also boiling or flow noise) of such intermediate products may be detected, and then, within the scope of the method according to the present description, corresponding sensor data (e.g., regarding the temperature, flow rate, chemical composition of the corresponding intermediate product) may be assigned to this acoustic information.

[0187] An object involved in a production process can be designed, for example, as a final product or intermediate product of a production method. In addition, an object involved in a production process can be designed and configured, for example, as a facility or equipment component, or as a facility or equipment part of a facility or equipment used within the scope of a production method.

[0188] Furthermore, scenarios related to the production or process flow can be assigned to the process flow, for example, scenarios within the combustion chamber of a gas turbine. Using corresponding acoustic information (e.g., combustion noise), anomalies within such a combustion process can be detected using the method according to the present description. In this scenario, for example, a time series of the gas turbine power output can be used as at least one assigned sensor value.

[0189] In addition, the method for analyzing acoustic information using specified scalar values ​​according to the present specification can also be designed and configured so that the method is designed and configured for analyzing medical acoustic data, such as biological acoustic data (such as lung noise or cough noise). In this scenario, the first and second acoustic information can be, for example, acoustic information of a human or animal (such as lung noise or cough noise) and the corresponding other scalar sensor values ​​are other, for example, medical sensor values ​​related to the human or animal, such as body temperature, pulse rate, or the like.

[0190] Furthermore, the method for analyzing acoustic information using a specified scalar value according to the present specification can also be designed and configured for use in person recognition or authentication. In this scenario, the first and second acoustic information can be, for example, a sound recording associated with a person (e.g., a voice recording), and the first and second sensor values ​​can also be values ​​characteristic of the person, such as a location value (e.g., GPS coordinates), an input code, eye color, or the like.

[0191] Further advantageous embodiments result from the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0192] The present invention is explained in more detail below by way of example with reference to the accompanying drawings.

[0193] The accompanying drawings show:

[0194] Figure 1A diagram showing some stages of an exemplary process sequence for winding battery cells and the determination of associated process parameters for subsequent analysis;

[0195] Figure 2 An exemplary diagram shows a process for creating an evaluation graph for a cluster of battery cells of a first winding;

[0196] Figure 3 An exemplary diagram shows a process for creating an evaluation graph for a cluster of second battery cells;

[0197] Figure 4 An exemplary diagram shows the sequence of a comparison of a first and a second evaluation graph. DETAILED DESCRIPTION

[0198] exist Figures 1 to 4 An embodiment of a method according to the present description is shown in FIG, wherein the method is used for quality control within the context of a battery manufacturing process. During the production process, indicators of variations in production quality can be predicted based on the analysis of images and sensor data from various manufactured battery cells, which represent intermediate products in the battery manufacturing process. This example thus illustrates a method according to the present description that utilizes assigned, context-dependent scalar values ​​for analyzing image information.

[0199] Figure 1 The left-hand part of the figure shows three process stages of a process for winding film layers 112 , 114 , 116 to produce a wound battery cell 132 , which represents an intermediate product within the scope of corresponding battery production.

[0200] exist Figures 1 to 4 Within the scope of , it will now be explained based on this example how, with the aid of an exemplary embodiment of the method according to the present description, a quality check of the winding of film layers 112 , 114 , 116 for producing a wound battery cell 132 can be achieved.

[0201] Figure 1 The three process stages shown in the left part of the diagram consist of a first stage 110, which includes providing the input materials or raw materials for producing battery cells within the scope of the process flow shown. In this case, an anode film 112, a separator film 114, and a cathode film 116 are provided as input materials, which are provided as a layer structure to a subsequent winding process 120, which is Figure 1 is shown in the middle of the left part of the figure.

[0202] Within the scope of this subsequent winding process 120 , the provided membranes, namely the separator membrane 114 , the anode membrane 112 , the further separator membrane 114 and the cathode membrane 116 , are shown, as well as the battery cell 132 produced by winding these membranes 114 , 112 , 116 .

[0203] exist Figure 1 The third process stage 130 shown in the left part is the finished intermediate product, which represents a wound battery cell 132. This battery cell is now further processed to complete a battery that can be put into use. Figure 1 It is not shown and is not part of the presented example.

[0204] exist Figure 1 The middle column schematically illustrates how the corresponding process and sensor parameters are determined within the scope of the described process stages 110, 120, and 130. In a first determination step 118, a parameter is determined, which includes whether an additional separating membrane 114 is used when winding the membranes 112, 114, and 116 in method step 120. If such an additional separating membrane 114 is used, the corresponding process parameter is set to 1, otherwise to 0.

[0205] exist Figure 1 The middle image 128 in the middle column shows the determination of at least one winding speed 128 for the winding process 120, which is an example of a sensor value according to the present description. To determine the winding speed, a time series is detected, in which the winding speed of the battery cells 132 is detected at different points in time during the winding process 120 while the battery cells 132 are being wound within the scope of the winding process 120.

[0206] Figure 1 The lower illustration 138 also shows a detection 138 of an image 134 of the front side of a wound battery cell 132. This image 134 is recorded in such a way that the layer structure of the wound can be recognized.

[0207] Then, in Figure 1 The right part of the figure shows diagrams 119 , 129 , 139 , which respectively show a diagram of a detected process parameter 119 , a sensor parameter 129 and an image information 139 .

[0208] The process parameters determined in method step 118 are shown in a diagram 119 in the vertical direction over the time plotted in the horizontal direction.

[0209] Furthermore, the winding speed detected in method step 128 is plotted on the vertical axis in a corresponding diagram 129 over the respectively determined time plotted on the horizontal axis.

[0210] exist Figure 1 The diagram 139 at the bottom right of FIG. 1 shows a schematic diagram 139 of the result of the frequency transformation of image 134. In this exemplary embodiment, a so-called discrete cosine transform (DCT) is used for the frequency transformation of image 134, as is also standardly used for frequency transformation of compressed images, for example, within the context of so-called JPEG compression of images. In the diagram 139 of the frequency-transformed image, only one of the harmonics in one direction is shown symbolically.

[0211] Figures 2 to 4 The evaluation step which enables the above-described quality check of the battery cells is now presented.

[0212] Here, according to Figure 2 The present invention describes how a first evaluation diagram 159 is created for first wound battery cells 132 based on images 134 detected for first wound battery cells 132 and process parameters 119 and winding speed 129 determined within the context of the respective winding. First wound battery cells 132 are an example of a first intermediate product according to the present description, and first evaluation diagram 159 is an example of a consistent representation of a first data structure according to the present description.

[0213] Then, according to Figure 3 It is explained that a second evaluation diagram 259 is created for the second wound battery cell 232 based on the corresponding recorded images 238 and using the process parameters 219 and the winding speed 229 determined during the winding process. The second wound battery cell 232 is an example of a second intermediate product according to the present description, and the second evaluation diagram 259 is an example of a consistent representation of the second data structure according to the present description.

[0214] Therefore, according to the following Figure 4 The comparison of the evaluation diagrams 159 , 259 created for checking the quality of the wound battery cells 132 , 232 is explained in more detail.

[0215] Now right Figure 2 The method flow shown in FIG is described in detail. As already mentioned, Figure 2 Show about Figure 1 . This is a diagram illustrating the creation of a first evaluation diagram 159 for process values, sensor values, and image values ​​determined within the scope of the process shown in FIG. Taking the process values, sensor values, and image values ​​determined within the scope of the process sequence presented in the figure or determined therein as an example, this evaluation diagram 159 is an example for a consistent representation of the first data structure according to the present description.

[0216] In a first creation step 140 for creating a first evaluation diagram 159, the parameter values ​​119 and the winding speed values ​​129 determined for the winding of the first battery cell 132 are initially placed in a unified parameter data diagram 149. For this purpose, the parameter values ​​119 and the winding speed 129 are correspondingly normalized. In the present example, this can be provided, for example, such that the value "1" for the parameter value "additional membrane present" is equivalent to the average value of the angular velocities determined in 129.

[0217] In the unified parameter data diagram 149 , the normalized winding speed 340 is now plotted on the time axis, and the parameter value 330 present during the winding of the first battery cell 132 is also plotted on the time axis in a correspondingly normalized manner.

[0218] Now, in a subsequent second creation step 150 , the just created unified parameter data diagram 149 and the data from the frequency-transformed image 139 are placed in a unified format in order to subsequently create a first evaluation diagram 159 .

[0219] Here, a representation is selected for the frequency-transformed image 139 of the first battery cell 132, wherein the individual harmonics 320, 310 of the frequency transformation are shown on a spatial axis 305. The spatial axis 305 can generally correspond to a section through the detected image 134 along a predetermined direction or also to a section thereof. Figure 2 A section through a horizontal section of the recorded image 134 is shown in .

[0220] Figure 2 The illustration of the frequency-converted image 139 on the left side of FIG. 1 is a simplified illustration of a first frequency-converted image 139 of a first battery cell 132 which has been already in FIG. 1 . Figure 1 The present invention is made within the scope of the process flow described. This simplification is made for the sake of clarity. The simplification is that, with respect to the DCT transformation used for the image, which is usually performed in two directions, only one component is shown in the diagram 139. A further simplification is that only one first harmonic frequency 310 and one second harmonic frequency 320 of the DCT transformation performed are shown in the diagram 139. The two harmonics 310, 320 shown by way of example are shown in FIG. Figure 2 In the diagram 139 in FIG. 1 , the phase shifts and the period lengths of the harmonics 310 , 320 are plotted on the spatial axis 305 , wherein the phase shifts and the period lengths of the harmonics 310 , 320 are respectively the results of the DCT transformation.

[0221] To create the first evaluation diagram, the data of frequency-converted image 139 are scanned at a specific spatial sampling frequency or spatial scanning frequency, and these scanned values ​​are then used to create first evaluation diagram 159. To this end, a second spatial axis 300 is also plotted in frequency-converted image 139, on which a selection of scanning points 301 is symbolically shown, at which the individual harmonics 310, 320 of the image converted into the frequency domain are scanned. For example, twice the spatial frequency of the highest harmonic vibration shown in the frequency-converted image can be used as the spatial scanning frequency.

[0222] The sampled values ​​for first harmonic 310 are marked in first evaluation diagram 159 by data points 352, which are represented by crosses. Similarly, the sampled values ​​for second harmonic 320 are marked in first evaluation diagram 159 by data points 350, which are represented by crosses. Furthermore, within the scope of second creation step 150, parameter values ​​360 and winding speed 362 are each extracted from unified parameter data diagram 149 at a predetermined or predeterminable time T0. Time T0 can, for example, correspond to the time at which image 134 of wound battery cell 132 was recorded, or even before that. This prior time can, for example, be selected so that the corresponding parameter or winding speed values ​​were already present during the winding of battery cell 132.

[0223] In order to integrate this parameter value and the winding speed value into the first evaluation diagram 159, an upper limit line and a lower limit line 352 are now determined. This is done by determining the maximum and minimum sum of the amplitudes of all harmonics of the frequency-transformed image and using this as the upper limit and lower limit 352. Figure 1 In the first evaluation diagram 159, a dashed line 352 is marked. The boundary line 352 prescribes a measure within which the parameter value 360 ​​present at the time T0 and the winding speed value 362 present at the time T0 are plotted. In this case, each value 360, 362 is repeatedly plotted for each scanning point 301 on the horizontal spatial axis of the first evaluation diagram 159, as shown in FIG. Figure 2 Some examples are shown in the diagram of the first evaluation diagram 159 .

[0224] In this way, the first evaluation diagram 159 now includes both the data of the image 134 recorded from the wound battery cell 132 and also data about the presence of the additional intermediate film 114 and the winding speed when winding the battery cell 132 .

[0225] To prepare a comparison of first evaluation diagram 159 with second evaluation diagram 259, a clustering method according to the prior art is now also applied to the entire set of data points of first evaluation diagram 159. In this example, only one cluster 370 is determined, which is marked as a dashed line in first evaluation diagram 159.

[0226] Figure 3 The creation of a second evaluation diagram 259 is now shown, which is created based on the data of the recorded second image 234 of the second wound battery cell 232, as well as parameter data 219 regarding the presence of an additional intermediate film 114 during the production of the second wound battery cell 232, and a winding speed value 229, wherein the winding speed value was already recorded during the winding of the battery cell 232. The creation of the second evaluation diagram 259 is carried out in this case in a manner corresponding to the creation of the first evaluation diagram 159. The second battery cell 232 represents a second intermediate product within the scope of the present exemplary embodiment.

[0227] Here, again to create second evaluation diagram 259, parameter values ​​219 and winding speed values ​​229 are standardized in first creation step 140 and converted into a corresponding, unified data diagram 249. In this data diagram, standardized parameter values ​​430 and standardized winding speed values ​​440 are plotted over time.

[0228] The recorded image 234 of the battery cell 232 is then frequency transformed again using the DCT method, which Figure 3 239 in a frequency-transformed image. Here, the first harmonic 410 and the second harmonic 420 are again shown by way of example on the spatial axis 405, wherein the harmonics 410, 420 represent the selection of waves or frequencies considered within the scope of the DCT method. Here, the spatial axis 405 can also generally correspond to a section through the detected image 234 in a predetermined direction or also to a portion thereof. Figure 3 2 shows again a part of a horizontal section through the recorded image 234. The two exemplary harmonics 410, 420 are shown in FIG. Figure 3 In the diagram 239 in FIG. 4 , the phase shift and the period length of the harmonics 410 , 420 are plotted on the spatial axis 405 , wherein the phase shift and the period length of the harmonics 410 , 420 are respectively the result of the DCT transformation.

[0229] As already mentioned above, here, for the sake of clarity, only one dimension of the frequency transformation is shown, as already in the illustration of the frequency-transformed image 139 for the first intermediate product 132 .

[0230] Now, in order to create the second evaluation diagram 259, the two harmonics 410, 420 shown are again scanned at the corresponding scanning points 401, wherein the scanning points 401 are selected in Figure 3 This is shown in the diagram of a frequency-transformed image 239 along a further spatial axis 400 .

[0231] The data points determined within the scope of this scan for first harmonic 410 are plotted in second evaluation diagram 259 as data points marked with crosses 452. The data points determined using this scan for second harmonic 420 are shown in second evaluation diagram 259 as data points marked with crosses 450.

[0232] Furthermore, to create second evaluation diagram 259, upper and lower limit lines 452 are now again found for integrating parameter values ​​and winding speed values ​​by determining the maximum and minimum amplitude sums of the harmonics determined by the DCT transformation. Within this range, parameter values ​​460 obtained from unified data diagram 249 at time T1 and winding speed values ​​262 obtained from unified data diagram 249 at time T1 are plotted. In this case, parameter values ​​460 and winding speed values ​​262 are plotted multiple times in second evaluation diagram 259 at corresponding scanning points 401 as in first evaluation diagram 159.

[0233] In order to prepare a comparison of the first and second evaluation diagrams, the clustering method is again applied to the entirety of the data points 450, 452, 460, 462 of the second evaluation diagram 259. In this case, two clusters are obtained, which are Figure 3 Dashed lines 470 , 472 are marked in the second evaluation diagram 259 .

[0234] Figure 4 Now, a first evaluation diagram 159 and a second evaluation diagram 259 are shown, wherein the reference numerals in the two evaluation diagrams 159, 259 correspond to the reference numerals in the two evaluation diagrams 159, 259. Figure 2 and Figure 3 The reference numerals correspond to those in FIG.

[0235] Figure 4 Now, a cluster analysis 510 derived from the clustering with respect to the first evaluation diagram 159 is shown, from which the number of clusters, their average area and the coordinates of the respective center or centroid of each determined cluster are determined. Figure 4 A cluster analysis 520 derived from the clustering with respect to the second evaluation diagram 259 is shown, from which the number of clusters, their average area and the coordinates of the respective center or centroid of each determined cluster are also derived.

[0236] In this example, the production of first battery cell 132 corresponds to the prescribed production, while the winding speed was too low during the production of second battery cell 232. Furthermore, in this example, a predetermined criterion for determining a possible error is that the number of identified clusters differs between first and second evaluation diagrams 159 , 259 . Other predetermined or predefinable criteria for outputting information can be, for example, that the position of at least one cluster changes by at least one predetermined or predefinable value, or that the area covered by the clusters, on average or as a whole, changes by a predetermined or predefinable value.

[0237] By comparing cluster analysis 510 of first evaluation diagram 159 with cluster analysis 520 of second evaluation diagram 259, it can now be determined, for example by a computer or a user, that the number of clusters in the second evaluation diagram has increased to two. Based on predefined criteria, appropriate information is then output, such as a warning message to the user or a corresponding alarm or control message or control command to the production facility or its host computer. Such control commands or alarms or control messages can, for example, trigger a check of various or all machine parameters or, if necessary, an emergency stop of the facility or a part of it.

[0238] In an alternative embodiment, the cluster structure 510 of the first evaluation diagram 159 and the cluster structure 259 of the second evaluation diagram 259 can also be evaluated via the neural network 600. In this case, the corresponding cluster analysis 510, 520 is input into the neural network, and the neural network outputs a result that only triggers a corresponding information message if the result corresponds to the corresponding error criterion. The neural network can also directly output whether a corresponding information message should be triggered.

Claims

1. A method for analyzing image information using a specified scalar value, the method comprising the following method steps: a.) detecting a first image of an object or scene, and detecting at least one first scalar sensor value associated with the object or scene; b.) detecting a second image of the object or scene, and detecting at least one second scalar sensor value associated with the object or scene; c.) importing the first image and the at least one first scalar sensor value as a consistent representation into a first data structure, and importing the second image and the at least one second scalar sensor value as a consistent representation into a second data structure; d.) comparing the first data structure and the second data structure; and e.) If the comparison results in a difference corresponding to a predefined or predefinable standard, a message is output.

2. The method according to claim 1, It is characterized in that In order to create the first data structure and the second data structure, the respectively detected images are respectively transformed into the frequency domain.

3. The method according to claim 2, It is characterized in that In order to create the first data structure and the second data structure, the respectively detected images are transformed into the frequency domain using Fourier analysis.

4. The method according to any one of claims 1 to 3, It is characterized in that In method step a.), at least one first scalar parameter value associated with the object or the situation is also detected, and In method step b.), at least one second scalar parameter value associated with the object or the situation is also detected. The first data structure and the second data structure are created by further utilizing the at least one first scalar parameter value and / or the at least one second parameter value.

5. The method according to any one of claims 1 to 3, It is characterized in that The comparison between the first data structure and the second data structure is achieved using a neural network (600).

6. The method according to any one of claims 1 to 3, It is characterized in that The first data structure and the second data structure respectively have a two-dimensional or higher-dimensional chart structure or a two-dimensional or higher-dimensional graph structure, or the first data structure and the second data structure respectively represent or can be represented as a two-dimensional or higher-dimensional chart or a two-dimensional or higher-dimensional graph.

7. The method according to claim 6, It is characterized in that Within the scope of the creation of the first data structure and the second data structure according to method step c.), a clustering method is or has been applied to corresponding charts, graphs, chart structures or graph structures of the first data structure and the second data structure, respectively. Alternatively, within the scope of the comparison of the first data structure and the second data structure according to method step d.), a clustering method is applied to corresponding charts, graphs, chart structures or graph structures of the first data structure and the second data structure, respectively.

8. The method according to any one of claims 1 to 3, It is characterized in that The method is designed and configured to monitor a method process or a production process so that - designing and configuring the detection of the first image and / or the second image as the detection of a first image and / or a second image of an object related to the production process or a scene related to the production process, and - the detection of the at least one first scalar sensor value and / or the second scalar sensor value is designed and configured as the detection of at least one first scalar sensor value and / or at least one second scalar sensor value relating to an object related to the production process or a situation related to the production process.

9. The method according to claim 8, It is characterized by: The production process is designed and configured for manufacturing a product and comprises a sequence of production steps, wherein after completion of a selected production step from the sequence of production steps, an intermediate product exists, - and also designing and configuring the first image as a first intermediate product image, - said at least one first scalar sensor value relates to a selected said production step, - designing and configuring the second image as a second intermediate product image, and - said at least one second scalar sensor value relates to a selected said production step.

10. The method according to claim 8, It is characterized by: In method step c.), a first data set is created using at least one further first production parameter value, And a second data set is created using at least one additional second production parameter value.

11. The method according to any one of claims 1 to 3, It is characterized by: The method is designed and configured to monitor the movement of a vehicle so that - the detection of the first image and / or the second image is designed and configured as the detection of a first image and / or a second image of a portion of the vehicle's environment, and - the detection of the at least one first scalar sensor value and / or the at least one second scalar sensor value is designed and configured as the detection of at least one first scalar sensor value and / or at least one second scalar sensor value related to the vehicle or the environment of the vehicle.

12. The method according to any one of claims 1 to 3, It is characterized by: The method is designed and configured for analyzing images of an object or organism such that - the detection of the first image and / or the second image is designed and configured as the detection of the first image and / or the second image of the object or living being, or as the detection of a region of the object or living being, respectively, and - the detection of the at least one first scalar sensor value and / or the at least one second scalar sensor value is designed and configured as the detection of at least one first scalar sensor value and / or at least one second scalar sensor value related to the object or the living being or as the detection of an area of ​​the object or the living being, respectively.

13. A method for analyzing acoustic information using a specified scalar value, the method comprising the following method steps: a.) detecting first acoustic information related to a sound source, and detecting at least one first scalar sensor value related to the sound source; b.) detecting second acoustic information related to the sound source or a second sound source, and detecting at least one second scalar sensor value related to the sound source or the second sound source; c.) transforming a first acoustic data set into the frequency domain, and introducing the frequency transformed first acoustic data set and the at least one first scalar sensor value into a first data structure as a consistent representation; and transforming a second acoustic data set into the frequency domain and introducing the frequency transformed second acoustic data set and the at least one second scalar sensor value into a second data structure as a consistent representation; d.) comparing the first data structure and the second data structure; and e.) If the comparison results in a difference corresponding to a predefined or predefinable standard, a message is output.

14. The method according to claim 13, It is characterized in that The method is also designed and configured according to the method for analyzing image information using a specified scalar value according to any one of claims 4 to 7.

15. The method according to claim 13, It is characterized in that The method is designed and configured to monitor a method process or a production process so that - designing and configuring the detection of the first acoustic information and / or the second acoustic information as an acoustic recording of a situation involving an object participating in a production process or related to a production process or a method flow, and - the detection of the at least one first scalar sensor value and / or the at least one second scalar sensor value is designed and configured as the detection of at least one value by means of a sensor, the at least one value relating to an object involved in the production process or a situation related to the production process or the method flow.

Images