EP4327302A1 - Procédé d'identification de panneaux à base de bois - Google Patents

Procédé d'identification de panneaux à base de bois

Info

Publication number
EP4327302A1
EP4327302A1 EP22723611.4A EP22723611A EP4327302A1 EP 4327302 A1 EP4327302 A1 EP 4327302A1 EP 22723611 A EP22723611 A EP 22723611A EP 4327302 A1 EP4327302 A1 EP 4327302A1
Authority
EP
European Patent Office
Prior art keywords
wood
comparison
data set
test
image
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Pending
Application number
EP22723611.4A
Other languages
German (de)
English (en)
Inventor
Ralf Niederer
Christian Maurer
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Fritz Egger GmbH and Co OG
Original Assignee
Fritz Egger GmbH and Co OG
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Fritz Egger GmbH and Co OG filed Critical Fritz Egger GmbH and Co OG
Publication of EP4327302A1 publication Critical patent/EP4327302A1/fr
Pending legal-status Critical Current

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/0002—Inspection of images, e.g. flaw detection
    • G06T7/0004—Industrial image inspection
    • G06T7/001—Industrial image inspection using an image reference approach
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V20/00—Scenes; Scene-specific elements
    • G06V20/80—Recognising image objects characterised by unique random patterns
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
    • G06V10/82—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using neural networks
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/20—Special algorithmic details
    • G06T2207/20084—Artificial neural networks [ANN]
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/30—Subject of image; Context of image processing
    • G06T2207/30108—Industrial image inspection
    • G06T2207/30161—Wood; Lumber
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V2201/00—Indexing scheme relating to image or video recognition or understanding
    • G06V2201/06—Recognition of objects for industrial automation

Definitions

  • the invention relates to a method for identifying wooden panels.
  • Wood-based panels in particular chipboard, OS B panels (oriented strand board
  • HD F boards High Density Fiber - HDF
  • MDF boards Medium Density Fibre
  • the manufacturing parameters present during production such as the types of wood-based material elements such as chips or fibers, the size distribution of the wood-based material elements, the type of resin used and any additives, the process temperatures and process pressures are critical parameters for the quality of the wood-based material boards.
  • logistical information such as the date of manufacture, place of manufacture, transport routes and storage locations.
  • the wood-based panels are printed with an optical code in the form of a barcode or a QR code, with information content of the code being linked to manufacturing parameters or other parameters in a database. If a wood-based panel is to be tested at a later point in time, the optical code is scanned and the manufacturing parameters or other parameters of the tested wood-based panel are assigned via the information content of the code and with the addition of the database. The tested wood-based panel can therefore be characterized.
  • the problem with the method described above is that the code is applied to the side edge of the wood-based panel and either a carrier material such as a sticker has to be applied or the code is printed directly onto the side edge. In either case, the code can become corrupted and thus unusable, nor can it be recovered. A code printed directly on the edge of the page may also be difficult to read.
  • the present invention is therefore based on the technical problem of improving and simplifying the identification of wood-based panels.
  • the technical problem mentioned above is solved according to the invention by a method for identifying wood-based material boards, in which at least one comparison image is recorded for a predetermined section of a side edge for a plurality of wood-based material boards, in which a comparison data record from the comparison image is based on the structure, in particular on the basis the geometry, the topography and/or the distribution of the wood-based material elements, is determined and stored, in which a test image for the specified section of a side edge of a wood-based material panel is recorded at least at a later point in time, in which a test data record from the test image is based on the structure, in particular on the basis of the geometry, topography and/or the distribution of the wood-based material elements, is determined in the same way as with the comparative data sets, in which the test data set of the tested wood-based material board is compared with the comparative data sets of several wood material boards is compared and in which the tested wood-based material board is identified as the wood-based material board for which the comparison data set was generated if
  • Wood material elements and optionally also from the resins connecting them and possibly additives, both through different starting materials and through random variations in the nature of the wood material that are not controlled during the manufacturing process, is structured so differently that this structure can be used as a feature for identification. Individual sections of the side edges can thus be used as an identification feature without any application of labels or imprinting of codes or other markings.
  • the structure is determined here, for example, by the shape properties of the side edge such as the properties of the wood-based material elements, i.e. their size, their individual and/or common alignment, the variability of size, the existence and/or number of different wood-based material element types, the dimensions of the distances between wood-based material elements or the presence and distribution of unfilled voids between the wood-based material elements. Not only individual wood-based material elements can be distinguished, but the overall picture of the side edge can be viewed as a structure.
  • the structure can additionally or alternatively be determined by the depth profile of the side edge.
  • the structure can also be determined by the color and/or the brightness of the side edge. This can be the basic color and basic brightness of the side edge, but also the variability of the color and brightness, the color of the wood-based material elements or the additives.
  • comparison data sets and the test data set are calculated in the same way, for example by using the same algorithm or generally the same calculation rule. This ensures that comparison images and test images recorded at different points in time of the same predetermined section result in comparison data records and test data records that are as similar as possible or identical.
  • the data sets can also be calculated in the same way, for example by using slightly different ones Calculation rules are applied if it can be ensured that the
  • the identification can then start from the point in time after the actual production of the material panel and then during the possible
  • the step of determining the comparison data set is carried out, preferably as early as possible, after the production of the panels after they have left the pressing plant.
  • the relevant section must be found when the wood-based panel is checked later.
  • the section can be found by the position of a corner of the wood-based panel as a delimitation of the side edge.
  • the section can also be found by means of an optical marking.
  • the marking can be printed as a rectangle that borders the section.
  • the marking can also be a logo that delimits the section laterally, preferably by a line. In any case, such a marking is only intended to determine the position of the relevant portion and does not serve to enable identification of the plate itself provided with it.
  • comparison images can be recorded for a number of sections on at least one of the side edges, in particular for at least one section on each side edge, of the wood-based material board and a comparison data set can be determined for each comparison image.
  • This increases the number of sections that can be checked later and makes it easier to carry out the check on site.
  • all sections located adjacent to each corner in either direction can be used to capture comparison images. Then no markings are necessary, since the Positioning of the sections is dictated by the physical limitations of the engineered wood panel.
  • the length of the comparison images along the side edge can be specified here, so that the end of the section facing away from the corner is also known without the need for a marking. This length can be specified in absolute terms or also relative to the plate thickness.
  • a test image is recorded and a test data set is determined for a part of the multiple sections with comparison data sets, in particular on a section of a plurality of sections with comparison data sets. It is therefore sufficient, in the case of a wood-based panel for which several comparison data sets have been created and stored, to create a test data set for the comparison only for some of the sections or even for just one section.
  • a further refinement of the method consists in that comparison images are recorded for a plurality of sections at predetermined intervals along at least part of a side edge and comparison data records are generated for the comparison images and that at the at least one later point in time a test image is produced in the area of the plurality of sections of the Side edge is recorded and a test data set for comparison with the comparison data sets is determined.
  • the sections may also overlap if part or all of the side edge is divided into sections at the predetermined intervals. The distances can even be selected to be so small that two sections are spaced apart from one another by one raster unit of the resolution of the subsequent image evaluation.
  • An entire side edge on one or more edges, in particular the entire circumferential side edge including all side edges of the plate, can also be recorded as a coherent comparison image.
  • Individual comparison images can be generated from this coherent comparison image.
  • a comparison between a test image and the coherent comparison image can also take place directly, without prior division of the coherent comparison image. This is possible, for example, through template matching, in which the test image is searched for in the coherent comparison image. Conversely, a coherent test image of all side edges can be recorded and comparison images can be searched for therein.
  • processing means in particular the removal of material by surface treatment, which changes the structure of the wood-based material that is visible on the surface.
  • the resulting at least two partial formats are then independently linked to an original wood-based panel, so that each partial format can be assigned to the original wood-based panel.
  • the previously explained recording of the comparison images preferably takes place after the production of the wood-based material board, after it has been cut to the formats to be delivered and more preferably before further processing such as coating with paint applications or with layers of paper and/or impregnated materials.
  • it can also be advantageous to record the comparison images after the coating that is to say in the state in which the wood-based material board is taken to a warehouse for delivery or is already ready for transport.
  • the wood-based panel has its final structure of the side edges, independent of further coatings.
  • the wood-based material board is preferably coated with one layer on one upper side or with different layers on both upper sides, which can be distinguished in a side view. It can then be advantageous to generate the at least one comparison image only after the coating, so that the orientation of the wood-based panel can be determined and taken into account on the basis of the laterally visible coating or coatings.
  • comparison images can be recorded at a number of stations in the manufacturing process, preferably always shortly before and/or shortly after the appearance of the side edge could have changed as a result of a manufacturing step. In this way, traceability can be implemented along the entire production chain.
  • the aforementioned comparison images and test images can be recorded with conventional camera systems. All 2D image cameras such as CCD cameras, single-lens reflex cameras (DSLR), mirrorless system cameras (DSLM) or smartphone cameras can be used as cameras, the latter in particular when creating the test images.
  • spectral cameras are available that break down and store each pixel into a multi-channel spectrum using a spectral measurement method. where a brightness and color value in the RGB color space with the values for red (R), green (G), blue (B) or in the L*a*b*-
  • L* blueness
  • a* red-green value
  • b* blue-yellow value
  • the aforementioned comparative images and test images can be taken under artificial lighting.
  • the lighting can preferably be substantially identical for each individual shot and can be located directly in the vicinity of the camera.
  • Other light sources such as solar radiation through a window, the lighting in the production hall, etc. should be shielded so that only the lighting illuminates the edge area to be recorded.
  • the edge can also be recorded in a closed chamber or the camera, lighting and side edge are at the moment of the Recording together enclosed by an opaque housing. In this way, it is ensured that only differences in the side edge lead to differences in the comparison and test images and that any interference from surrounding light sources does not also falsify the recording. Thus, even small differences in color, brightness, structure, etc. of the side edges can be reliably distinguished.
  • the light source can be adapted to the nature of the side edge and/or the camera. Likewise, multiple images can be recorded and combined under different illumination, for example at different wavelengths.
  • a comparison data record can be determined from the comparison images in different ways. In principle, any determination method that leads to a value or values that can be used in a subsequent comparison algorithm is suitable for use in the method described.
  • a first exemplary embodiment of a determination method consists in analyzing the recorded structure of the wood-based material in each image, preferably using one or more image analysis programs known per se. For example, OpenCV, scikit-image, Google Cloud Vision API, Microsoft Computer Vision API, vlfeat, dlib, Matlab Image Processing Toolbox can be mentioned here.
  • structural values are then determined by the shape, the circumference, the area, the length, to be determined as the longest extension of the form, and/or the orientation, to be determined as the direction of the longest extension of the form, of the wood-based material elements determined in relation to the dimensions of the captured portion of the side edge.
  • the comparison data set and the test data set are then determined as a sequence of structural values for a specified proportion of the wood-based material elements.
  • the necessary number of structure values for each image depends on the individuality of the wood-based material elements and is determined in relation to the application. In any case, the structural values are decisive for the identifiability of the section and thus of the wood-based panel. In this case, their order in the sequence of structure values can be decisive or irrelevant.
  • the determination method consists in that each image corresponds to a predetermined grid in relation to the orientation of the section of the recorded side edge is divided into pixel areas, in which each pixel area is calculated from the individual pixels, preferably averaged
  • pixel value i.e. a brightness value and/or at least one color value
  • the comparison data set or the test data set is determined as a sequence of pixel values in a predetermined order from the pixel areas.
  • a grid of, for example, N ⁇ M pixels with a gradation of 256 gray scale values is thus determined, the pixels of which each represent the named pixel areas.
  • the respective gray level value of a pixel area can be determined by averaging or another mathematical method and represents an average gray level of the brightness information or one of the color information of the pixels. which can be used as a basis for the subsequent comparison step.
  • N x M vector components which can then also be understood as a matrix.
  • a vector can then be generated from the matrix by averaging each row.
  • the images can be prepared or pre-processed beforehand using an optical filter or a software filter.
  • an edge filter can be used to emphasize the alignment of wood fibers.
  • a further refinement of the method described is that the images, ie the comparison images and the test images, are each recorded with a 3D camera.
  • An example of a 3D camera is a LIDAR camera as a standalone camera or as part of a smartphone.
  • LIDAR is the abbreviation of Light Detection And Ranging.
  • Another example of a 3D camera has two laterally offset cameras.
  • other contactless technologies can also be used to record the 3D structure of the surface or its roughness. These are Radio Detection And Ranging (RADAR),
  • RADAR Radio Detection And Ranging
  • SODAR Sound DetectingAnd Ranging
  • a digital representation of the material surface of one side of the board is generated and stored in one-dimensional or multi-dimensional resolution. Regardless of the measurement method used, the digital surface information obtained in this way can be converted as a comparison data set and test data set.
  • the method step for determining the image data record is initially carried out in that the topography, ie a height profile perpendicular to the surface orientation of the side edge, is determined for each image.
  • a contour line profile, sections of the height profile, a triangular mesh or sections of the triangular mesh are then derived from the topography as topography values and the comparison data set or the test data set is determined from the topography as a sequence of topography values in a predetermined order.
  • the comparison data records and the test data records consist of a sequence of characteristic parameters of the topography, which enable the wood-based panel to be identified.
  • the sequence of parameters can be processed in the comparison step using a suitable algorithm.
  • the topography can also be determined by mechanically scanning the surface, for which various methods are known. For example, the surface can be scanned along a given track with a button, e.g.
  • Shape of a wand, to be traversed As the probe moves along the track, the different heights depending on the distance covered are recorded as digital values and stored mathematically as a vector. Additional vectors can be determined by arranging tracks next to each other, which can then also be saved as a matrix. Such a measurement method has been developed by Keyence Corporation to measure surface roughness.
  • An atomic force microscope can also be used instead of a probe in the form of a needle.
  • the position of the measurement must be precisely specified, especially when multiple measurements are made, even on different sides of the plate.
  • the position of the measurement for the comparison data set on the surface is defined and saved in such a way that it can be repeated for recording the test data set.
  • the starting point of the measurement can be specified by specifying the distance from a corner of a disk and the direction of the track to be scanned.
  • a comparison is made with the comparison data records stored in a database in order to identify the wood-based material panel if there is sufficient agreement, possibly within predetermined parameter limits.
  • the comparison between the test data set and the comparison data sets is carried out using mathematical methods.
  • a mathematically defined distance between the comparison data sets and the test data sets is calculated using an algorithm, and the distance is then compared with a threshold value. If the calculated distance is less than the threshold, a match can be determined.
  • a deterministic method is thus carried out, which is specified by a direct specification of the selected algorithm. Without a threshold value, the comparison data record that has the smallest distance from the test data record of all comparison data records can also be selected as the best match.
  • Algorithms consist of a finite number of defined individual steps. In particular, they can also be implemented in a computer program for execution.
  • a specific input is converted into a specific output.
  • a trained neural network is then used instead of an algorithm mentioned above.
  • a fully trained network is then also a defined system, but a description as an algorithm is not given here.
  • At least one neural network of a system for machine learning is used.
  • a system for machine learning allows the use of pattern analysis and pattern recognition.
  • the visual intelligence used makes it possible to recognize and analyze images or shapes. Handwriting recognition or the identification of people through face recognition are generally mentioned here as application examples.
  • Neural networks are networks of artificial neurons and represent a branch of artificial intelligence.
  • the topology of a neural network consists of the Assignment of connections to nodes and is trained for the specific task, whereby the network can be trained by the following methods.
  • a neural network will not independently make changes to its own topology during training. Instead, you try out different topologies, some of them automatically, and let the neural network learn by adjusting certain parameters within the network, which ultimately corresponds to a learning step in each case.
  • the learning behavior changes when the activation function of the neurons or the learning rate of the neural network changes.
  • a network learns mainly by modifying the weights of the neurons.
  • the threshold value can be adjusted in parallel in this case.
  • the topology describes the structure of the network. This generally means how many artificial neurons are on how many layers, and how they are connected to each other. Artificial neurons can be connected in a variety of ways to form an artificial neural network. In many models, neurons are arranged in consecutive layers.
  • a neural network is trained with as large a number of comparison images or comparison data as possible, for example 1,000 or 10,000 pictures.
  • the entire image data or a mathematical representation of the image data can be used as input data in the neural network for the training.
  • the gray scale data in the grid of the comparison images are preferably fed into the neural network as comparison data sets.
  • the gray levels are assigned in a value range of, for example, 256 gray levels
  • a neural network preferably a convolutional neural network (CNN)
  • CNN convolutional neural network
  • the images of the ⁇ side edges are converted into a feature representation in an encoder, the features of which are then learned so that, starting from them, the original image can be recovered again in a decoder. In this way, an error-tolerant feature representation of the wood structure is generated.
  • CNN convolutional neural network
  • Comparative data set and each test data set is then converted into the characteristic representation by the encoder.
  • the test sets can thus be compared with the comparative data sets in a characteristic set adapted to the wood structure, without characteristics having to be defined manually.
  • autoencoders and decoders can also be used in the method described, which compensate for any errors in the original image and thus pass on an error-reduced image variant to the actual neural network for training.
  • a neural network As an alternative to a neural network, other methods of machine learning (e.g. support vector machines) or mathematical methods can also be used. Furthermore, for example, a cross-correlation between two images, that of the test data and that of the comparison data, can be calculated. Alternatively, other metrics such as root mean square error or cosine similarity can be used. If the similarity is sufficiently high, a match can be recognized. For the use of some metrics it can make sense to convert the pixels from a two-dimensional matrix arrangement into a one-dimensional vector arrangement using a specified operation.
  • machine learning e.g. support vector machines
  • a cross-correlation between two images, that of the test data and that of the comparison data can be calculated.
  • other metrics such as root mean square error or cosine similarity can be used. If the similarity is sufficiently high, a match can be recognized. For the use of some metrics it can make sense to convert the pixels from a two-dimensional matrix arrangement into a one-dimensional vector arrangement using a specified operation.
  • the test and comparison data can also be characterized using predefined image features such as entropy, a distribution of contiguous areas or a dominant fiber orientation. Using these features, a comparison can then be made between the feature values of an existing test image and different comparison images.
  • the technical problem outlined above is now solved according to the invention by a method for identifying wood-based panels, in which at least one classification is determined and stored for a plurality of wood-based panels for each wood-based panel, in which at least one comparison image is recorded for a predetermined section of a side edge of each wood-based panel , in which a comparison data set is determined and stored from the comparison image based on the structure, in particular on the basis of the geometry, the topography and/or the distribution of the wood-based material elements, in which a neural network of a system for machine learning trains with the comparison data sets with the specification that each comparative data set fed in is assigned to a classification, in which a test image for the specified section of a side edge of a wood-based panel is recorded at least
  • a classification for example a place of manufacture or a period of manufacture, can be assigned to a plurality of comparison images during the training of the neural network.
  • a test image is then assigned to a classification using the trained neural network.
  • the identification consists only in the classification, without a clear assignment with a specific wood-based panel.
  • the technical problem outlined above is also solved according to the invention by a method for identifying wood-based material boards, in which at least one comparison image is recorded for a predetermined section of a side edge of each wood-based material board/in which a comparison data set from the comparison image is based on the structure, in particular on the basis of the geometry ,
  • the topography and / or the distribution of the wood-based material elements is determined and stored, in which a neural network of a system for machine learning is trained with the comparison data sets with the specification that each fed-in comparison data set is different from all other comparison data sets, in which after At the end of the training, a mathematical comparison value, for example a number or a vector, is generated and stored for all comparison data sets, at which point a test image for the specified section of a side channel is produced at least at a later point in time te of a wood-based panel is recorded, in which a test data set is determined from the test image using the structure, in particular on the basis of the geometry, topography and/or the distribution
  • a Siamese neural network consists of two identical sub-networks, also known as twin networks, which are connected at their outputs.
  • the twin networks not only have an identical architecture, but also share the same thresholds and weights. They work in parallel and are responsible for creating vector representations for the inputs.
  • the numerical values or vectors calculated by the neural network are compared with one another, which are calculated by the neural networks - similar to an algorithm.
  • the two neural networks cannot be used at the same time; instead, a neural network is first trained with a large number of comparison images Connections between the neurons, is stored and can be transferred to another neural network or the other of the Siamese neural networks.
  • the comparison values in the form of numbers or vectors determined at intervals after the end of the training are stored in a database for later comparison with test values.
  • each new wood-based panel is recorded with comparison images during ongoing production, the calculation of the numerical values or vectors with the trained one Network carried out and saved. The same procedure is used for the comparison images that were used when training the neural network.
  • the numerical values or vectors for these wood-based material panels are calculated and stored in the same way When checking whether a test image corresponds to a comparison image, such a neural network constructed in the same way is then used to calculate a test value in the form of a number or a vector from the test image or the test data set. The test value is then compared with the reference values in the database and if there is sufficient agreement, the Siamese neural network identifies a wood-based panel.
  • a combination of a classification by the first-mentioned method and a subsequent identification by the second-mentioned method is preferably used.
  • a wood-based panel to be tested is initially assigned a classification, for example a place of manufacture, by means of the test image. Since the number of possible comparison images is thus limited, a smaller number of
  • Comparison images can be used to train the neural network, which can improve accuracy and the hit rate.
  • a hit rate ie the number of correct assignments as a proportion of the total assignments, in the range of greater than 80%, preferably greater than 90%, is sufficient for good functionality, although the aim is for the greatest possible hit rate.
  • the method is supplemented by the following steps. For the majority of wood-based material plates, preferably at the beginning of
  • a characterization data set is determined and stored, with each comparison data set being linked to the characterization data set for the respective wood-based panel and in which the tested wood-based panel is assigned the characterization data set linked to the comparison data set if the test data set matches a comparison data set.
  • Both data sets can be stored separately and linked by means of pointers, but the data can also be stored as one data set and linked to that extent.
  • the tested wood-based panel is thus characterized by additional data and information, and the information stored in the characterization data record can be used for further logistics or further processing.
  • the identification described above can thus lead to an additional assignment of at least one item of characterization data to a tested wood-based panel.
  • the assignment can therefore be a classification in to which a production location, production date, production time and production facility and/or a wood species or another characteristic property of the tested wood-based panel is assigned. This property can be assigned to a large number of wood-based panels, which are thus classified.
  • the characterization data record is supplemented by at least one piece of data that characterizes the processing step of separating the wood-based material board.
  • the at least one additional date can contain the time of processing and/or the new dimensions of the partial formats.
  • the other unchanged side edges remain linked to the previous comparison images and comparison data sets, which, however, must also be linked to the changed characterization data set.
  • the characterization file assigned to a tested wood-based panel with the classifying and/or further identifying information can be linked and presented in various ways with further data in the utilization chain from the manufacturer to the end customer.
  • the recording of the test image is recorded on site, in particular at an end customer's premises, preferably with a mobile device such as a smartphone.
  • the application (app) used to record the test image is provided by the manufacturer of the wood-based panel so that a direct data connection can be established with one or more of the manufacturer's servers.
  • the wood-based panel is then characterized or identified, so that the manufacturer can record data on the current location, the registered user and/or the type of use.
  • the manufacturer can determine further data on this wood-based panel and the user via the application (app). present. In both cases, there is an added value that is made possible by using one of the methods described.
  • Further product suggestions suitable for the wood-based panel can preferably be submitted to the user.
  • suggestions for other suitable products can be displayed that can be used with the present wood-based panel, in particular taking into account the decor associated with the wood-based panel.
  • the proposals can have further wooden boards, laminates or floor panels.
  • inspiration for the user can be displayed, for example photos of room situations and/or furniture in which the wood-based panel has been or is to be installed.
  • business information can also be processed with the
  • Characterization or, where appropriate, identification of the wood-based panel are connected.
  • the transmission of an invoice from the dealer or the manufacturer can be linked to the recording and analysis of the test image or the processing of a complaint.
  • proof of originality can be sent in the form of a certificate of authenticity.
  • the purchasing behavior of the end customer can be recorded when the wood-based panel has been scanned or recorded. This can be used to send individualized offers, new products or the like. As a result, the manufacturer or a retailer can determine data about the use, the time of use and the type of use of the wood-based panels. Thus, among other things, information about the ordering behavior and storage times at dealers and end users can be recorded until processing.
  • the invention is explained below using exemplary embodiments with reference to the drawing. Show in the drawing
  • FIG. 1 to 4 a schematic representation and tables for an embodiment of a method according to the invention applied to a section of a side edge of a wood-based panel and evaluation of geometric variables of the chip distribution,
  • FIG. 5 shows a schematic representation of a further exemplary embodiment of a method according to the invention using four sections of a side edge of a wood-based material board
  • FIG. 6 shows a schematic representation of a further exemplary embodiment of a method according to the invention using two sections at each corner of a wood-based material panel
  • FIG. 7 shows a schematic representation of a further exemplary embodiment of a method according to the invention using a plurality of sections on a side edge of a wood-based material board
  • FIG. 8 shows a schematic representation of a further exemplary embodiment of a method according to the invention, applied to a section on a side edge of a wood-based material board and evaluating a gray scale grid of the chip distribution
  • 9 shows a schematic representation of a further exemplary embodiment of a method according to the invention applied to a section on a side edge of a wood-based material board and evaluating height profile data of the chip distribution
  • FIG. 10 shows a schematic representation of a further exemplary embodiment of a method according to the invention with an evaluation of data sets of the chip distribution using training of a neural network with a plurality of comparison data sets,
  • FIG. 11 shows the neural network shown in FIG. 10 during a check of a
  • FIG. 12 shows a schematic representation of a further exemplary embodiment of a method according to the invention with an evaluation of data sets of the chip distribution using training of a neural network with a plurality of comparison data sets,
  • FIG. 13 shows the neural network shown in FIG. 12 during a check of a test data set
  • FIG. 14 shows an example of a detection of the surface structure by means of a mechanical probe
  • FIG. 15 shows a plate to be measured with coordinates drawn in.
  • 1 to 4 show a first exemplary embodiment of a method for identifying wood-based panels.
  • a photograph of a piece of a side edge 2 of a wood-based panel is shown in a side view.
  • a right-angled corner 4 is arranged, which can be used as a reference for a dimension "da" to determine the position of a section 6.
  • the section 6 is shown with a dashed auxiliary line, but this is not placed on the side edge itself.
  • the side edge 2 is not further printed or otherwise marked Fig. 2 shows section 6 in a separate view
  • a comparison image for section 6 of side edge 2 is now recorded with a camera (not shown) and evaluated.
  • Conventional image analysis software such as OpenCV, scikit-image, Google Cloud Vision API, Microsoft Computer Vision API is used for this purpose , vlfeat, dlib, matlab image processing toolbox, etc. and a number of individual wood chips are detected.
  • the outlines of the detected chips are partially determined by the software using specified criteria such as size or area. This is shown as an example for a long chip in FIGS. 1 and 2, the surface of which is marked by a constant gray coloring.
  • Fig. 4 shows the graphical representation of the data according to the table in Fig. 3, where the length dl in the direction of the x-axis and the angle a relative to the x-axis are plotted as vectors.
  • the corresponding vectors vi, V2, V3 and V20 are shown in Fig. 4.
  • One advantage of forming the absolute value for the values dx(N) is that the value is the same, regardless of the orientation of the wood-based panel
  • test image is taken for the specified section of the side edge of a wood-based panel recorded.
  • test image is subjected to the same analysis as explained above with reference to FIGS. 1 to 4, and a test data set is determined from the test image using the geometry of the wood-based material elements (chips, fibers) in the same way as with the comparison data sets.
  • the test data set is then compared with the comparison data sets stored in the database for the large number of wood-based panels. If the test data set matches a comparison data set, the tested wood-based material panel is then identified as the wood-based material panel for which the comparison data set was generated.
  • the test image is preferably recorded with the same camera and in the same exposure situations.
  • the process can also be so robust to fluctuations that, for example, a permanently installed camera records the comparison image after the wood-based panel has been manufactured, while the Test image can be recorded at a later time with a mobile camera, in particular using a camera of a smartphone.
  • the database is generally stored on a server which is connected to the processing units for creating comparison data sets and the processing units for creating test data sets via a local network of the manufacturing facility and/or via the Internet and enables data exchange.
  • Fig. 5 shows a further exemplary embodiment of a method for identifying wood-based panels, in which, in comparison to the example according to Figs and da4 are spaced from the corner 4.
  • comparison images are recorded for the four sections 6a, 6b, 6c and 6d on the side edge 2 and for each section 6a, 6b, 6c and 6d, a sum and is calculated in the manner explained above educated.
  • the numerical values of four totals are then entered in the order of the sections a 4-dimensional vector V is written, which forms the comparison data set.
  • the four sections positioned in the predetermined sections dai, da2, da3 and da4 are then photographed in the same way on a predetermined side edge and, as described above, evaluated and a vector V is formed as a test data set.
  • This test data record is then compared with the comparison data records stored in the database. If the test data set and one of the comparison data sets match, the tested wood-based material board is then identified as the wood-based material board for which the comparison data set was generated.
  • the comparison data record and the test data record are thus a sequence of structural values been determined for a given proportion of the wood-based material elements.
  • FIG. 6 shows another embodiment of the method for identifying wood-based material panels, in which at each of the four corners of a rectangular wood-based material panel at each of the two adjacent side edges, in this case adjacent to the corner with a value for the distance da equal to zero section defined.
  • Eight comparison images are thus generated for eight sections 6a to 6h and processed further.
  • each wood-based panel is so often linked to a comparison image that each wood-based panel can be identified in a simple manner, regardless of its position. If a test image is recorded and a test data set is determined for a test process for only one of the eight sections with comparative data sets, then the wood-based material board can still be identified.
  • comparison images are recorded for a plurality of sections 6a to 6e at predetermined distances D along at least part of a side edge 2 and comparison data records are generated for the comparison images.
  • the predetermined distance can be adapted to the resolution of the comparison images or to the spatial resolution of a grayscale distribution determined from the comparison images, see the discussion of FIG. 8 below.
  • a test image can then be recorded in the area of the plurality of sections of the side edge in accordance with the grid, ie preferably randomly, without the need for exact positioning of the recorded test image.
  • a test data set for the comparison with the comparison data sets is then determined again. Further processing then corresponds to the previous examples. Since the structures of the side edges themselves are used for identification, it is necessary for identification that the structures are preserved or only changed to a small extent. However, if the structures are changed too much, comparison images and comparison data sets must be created again.
  • machining means in particular removing material, for example milling or grinding.
  • this also applies when the wood-based panel is separated and two or more partial formats are cut. Then, in a further embodiment of the method, it is provided that after one of the wood-based panels has been separated into at least two partial formats, at least one comparison image is recorded for each partial format created on at least one newly created side edge of the partial formats for at least one section and a comparison data record is determined.
  • the configurations of the method have been described, in which 2D cameras have been used and in which the recorded structure of the wood-based material is analyzed in each image, in which the structure of the wood-based material elements are determined by the length as the longest extent in the form and structural values the comparison data sets and test data sets are determined by the alignment, ie the direction of the longest extension in the shape of the wood-based material elements in relation to the dimensions of the recorded section of the side edge.
  • information for the comparison data sets and test data sets can also be obtained through the shape, through the circumference and/or through the area of the wood-based material elements.
  • the 2D images i.e. the comparison images and the test images, can be recorded with conventional cameras, for example generally CCD
  • DSLR single lens reflex cameras
  • DSLM mirrorless system cameras
  • a 2D image can also be determined with a spectral data for each pixel
  • the image can be in the RGB color space with the values for red (R), green (G) and blue (B) or in the L*a*b* color space with pixel values L*, a*, or b* for brightness ( L*), red-green color intensity (a*) and blue-yellow color intensity (b*) or in other color spaces.
  • Another embodiment has a different image analysis, in which not the existing structures are determined, but in which - as shown in Fig. 8 - each image of the section 6 adjacent to the corner 4 of the side edge 2 according to a predetermined grid with respect to the in Fig. 8 horizontal alignment of section 6 is divided into pixel areas.
  • Each square of the grid comprises a plurality of pixels of the recorded image and represents a pixel area.
  • a pixel value calculated from the individual pixels and averaged from the brightness values is then assigned to each pixel area and the comparison data set or the test data set is then processed as a sequence of averaged pixel values 10, 75, 190, 205, . . . in a value range of (0 a series of values is shown in sections.
  • At least one color value of a color channel can also be calculated as a pixel value.
  • the comparison data set determined in this way and the test data set are then processed in the manner described above in order to carry out the method.
  • Wood-based panels is shown in FIG. Shown is again the side edge 2 of Wood-based panel as in Figure 1, the selected portion 6 is located adjacent the corner 4 of the wood-based panel.
  • the images i.e. the comparison images and the test images, are each recorded with a 3D camera, which is designed as a LIDAR camera as an independent camera or as part of a smartphone or as a stereoscopic camera having two cameras.
  • the topography is then determined for each image and a profile that reproduces this topography is generated from the topography.
  • Different parameters such as the contour line profile as a whole or in sections, a triangular mesh or sections of the triangular mesh can be derived as topography values from the profile and the comparison data set or the test data set are then created as a sequence of topography values in a specified order from the topography determined.
  • the shape of the contour lines can be analyzed directly.
  • a number of black and white images can be generated, with a height threshold being specified for each image, and areas whose height is below the threshold are colored black and those whose height are greater than or equal to the threshold are colored white. From this, properties of the binary image elements can be evaluated, such as the largest connected area, the distribution of the sizes of individual areas and/or the ratio of the main to the secondary axes in individual areas.
  • a depth image can be converted into a grayscale image composed of pixels that each describe a depth of the surface profile.
  • a low value close to 0 can indicate areas that are close to the surveying camera, i.e. to an imaginary perfectly smooth surface.
  • Higher values >0 can indicate a depression such as a space between two fibers.
  • the scale can also be reversed.
  • the three-dimensional structure can also be analyzed directly, without being converted into a two-dimensional equivalent.
  • Various 3D parameters can be used for this, in particular according to the standard EN ISO 25178. To evaluate a three-dimensional surface, parameters and measurement methods are specified, by means of which the surfaces can be measured and thus used for identification.
  • a digital model can also be created from a large number of individual points that represent the surface of the structure of the side edge. This can be done either by meshing a triangle (Triangulated Irregular Network) or by approximating a rectangular grid network. This elevation model is intersected with a horizontal plane at the desired elevation, the result is the contour line.
  • a triangle Triangulated Irregular Network
  • Amplitude characteristics can be used as further parameters, which describe, for example, the height structure of the surface with individual values and are based on a 2D profile section evaluation. They are transferred from the linear 2D to the areal 3D evaluation.
  • Other parameters are an area-related arithmetic average roughness (Sa), an area-related core roughness depth (Sk), an area-related mean roughness depth (Sz) or an area-related square mean roughness value (Sq). Since the surface of the side edge of a wood-based panel can assume different values in these parameters, one or more of these parameters can be determined in a grid corresponding to the grid according to FIG. 8 for each field of the grid, the numerical values of which are written together in a specific vector.
  • the autocorrelation function ACF is used for this and the parameters can be again preferably in a raster, the autocorrelation length (Sal) and/or the texture aspect ratio (Str) are used.
  • test data records with the mean square surface gradient (Sdq) and/or the ratio of the real surface to the measurement area (Sdr), again preferably arranged in a grid, can be used as parameters.
  • Sdq mean square surface gradient
  • Sdr ratio of the real surface to the measurement area
  • Comparison data records are compared, with a mathematically defined distance between the comparison data records and the test data records preferably being calculated using an algorithm, at which the distance is compared to a threshold value and at which a match is established if the distance is less than the threshold value.
  • a first common distance metric is a root mean square error (RMSE), in which a difference in the feature values between the test and comparison data is calculated for each element, this difference is squared, the squared differences for all features are added up and then the root is calculated the sum is drawn
  • RMSE root mean square error
  • a cosine similarity C can be used, specifically as follows. Values of -1 indicate a large difference, values close to 1 indicate a high level of agreement.
  • the algorithms described above implement a direct mathematical
  • a machine-learning neural network can also be used.
  • Various types of neural networks can be used in the method described, for example also convolutional neural networks (CNN).
  • CNN convolutional neural networks
  • artificial neural networks can be based on the networking of many McCulloch-Pitts neurons or slight modifications thereof.
  • other artificial neurons can also be used in artificial intelligence, for example the high-order neuron.
  • the input layer is also represented as a layer of nodes.
  • Machine learning methods are then used to modify a neural network in such a way that it generates associated output patterns (e.g. agreement yes/no) for specific input patterns (comparison data sets).
  • output patterns e.g. agreement yes/no
  • the artificial neural network is given an input pattern and the output that the neural network produces in its current state is compared with the value that it is supposed to output.
  • conclusions can be drawn about the changes to be made to the network configuration and the neural network can be run through again in the opposite direction. Training a new neural network with a large amount of data is computationally intensive and time-consuming.
  • Enormous computing power is required, particularly in the area of detecting objects in images with deep leaming, in order to train the network, for example with millions of images, for many training cycles (epochs).
  • a convolutional neural network (CNN) first learns to distinguish between the relevant structures and shapes and then derives and recognizes less abstract objects from them.
  • the fully trained layers of a neural network are taken over and only the output layer is adapted to the number of object classes to be recognized in the new network and retrained.
  • the advantage is that the neural network can already recognize and distinguish shapes and structures, and it only has to learn a new assignment of the object classes.
  • the fully trained layers are either kept constant and retraining is only carried out on the output layer, or some or all layers are further trained on the basis of the current training status.
  • the method described above can now be implemented with the help of a neural network in such a way that a neural network of a system for machine learning is trained with the comparison data sets with the specification that each supplied comparison data set is assigned to a classification and that each test data set is assigned to the trained neural network assigned to one of the classifications.
  • the same classification is thus assigned to a plurality of comparison images, ie for example a place of manufacture.
  • This procedure corresponds to that of character recognition, in which differently written individual letters can each be assigned to a classification of a specific letter by a trained neural network.
  • the method can also be carried out in such a way that a neural network of a system for machine learning is trained with the comparison data sets with the specification that each fed-in comparison data set is different from all other comparison data sets, in which, after completion of the training for all comparison data sets, a mathematical Comparative value, preferably in the form of a number or a vector, is generated and stored, that for each test data record with the trained neural network, a mathematical test value, also preferably in the form of a number or a vector, is generated and stored and that the tested wood-based panel is Agreement of the mathematical test value with a comparison value is identified as the wood-based panel for which the comparison data set has been generated.
  • a mathematical Comparative value preferably in the form of a number or a vector
  • a so-called Siamese neural network is used here, in which two identically working neural networks generate comparison and test values that can be compared with one another.
  • a test image can be clearly assigned to a wood-based panel using the comparison images. This procedure corresponds to that of face recognition.
  • the two described alternatives of the method are applied one after the other by pre-classifying the
  • test image is performed and by performing the procedure for comparison images of each of the available classifications separately.
  • the wood-based panel can initially be assigned to a place of manufacture or a manufacturing facility by means of a classification. Thereafter, a Siamese neural network can be accessed that on Wood-based panels has been trained, which have been manufactured exclusively on this manufacturing plant. In this way, a gradual and modular identification can be achieved.
  • FIG. 10 now shows a schematic representation of a training of a neural network 10 which consists of an input layer 12 , hidden layers 14 and an output layer 16 .
  • a plurality of comparison datasets V are entered into the input layer 12 via an interface 20, with the training taking place with the proviso that each fed-in comparison dataset V is assigned to a classification A or B.
  • Such training takes place with a large number of comparison datasets, for example 10,000 comparison datasets.
  • the training can be carried out by changing or adapting the parameters of the neural network such as the threshold value of each node or neuron (circles in the hidden layers 14) and the weighting factors wjj from neuron j to neuron i, until a reliable classification of the comparative data sets can be ensured by the neural network.
  • the neural network is thus set up in such a way that, according to FIG. 11, one of the classifications A or B is assigned to a test data set P with the trained neural network.
  • the place of manufacture, the date of manufacture, the panel thickness of the wood-based panel or other general data can be used as classifications A and B.
  • a classification therefore includes a plurality of associated wood-based panels, with the classification being viewed as a possible form of identification.
  • the neural network is trained by training with a large number of comparative data sets in such a way that the wood-based material panels to which the comparative data sets used for training belong can be classified by the trained network.
  • wood-based panels can also be classified are used whose comparison datasets have not been used for training.
  • this property of the trained network applies to wood-based panels that were only produced after the training and can be assigned to one of the classifications.
  • Figures 12 and 13 show another application of neural networks.
  • Fig. 12 shows schematically a similar neural network as Figs. 10 and 11.
  • the neural network is trained with the comparison data sets V with the stipulation that each supplied comparison data set V is different from all other comparison data sets.
  • two identical images can also be fed into the neural network with the specification that these are two identical images.
  • the number of comparison data sets is large and is in the range of more than 10,000, for example.
  • the neural network is changed in such a way that different comparison data sets can be distinguished and that after the training is complete for all comparison data sets V a mathematical comparison value Z different from all other comparison values is generated and stored.
  • the comparison values Z of the corresponding wood-based panels are linked and saved.
  • the neural network was trained to compare a capture of a proof image with the existing captures and determine whether the captured proof image shows a pre-existing panel. According to FIG. 13, this is done in that a mathematical test value Y is generated with the trained neural network for the test data set generated from the test image of a wood-based panel to be tested and stored at least temporarily. The tested wood-based panel is then identified as the original wood-based panel linked to the comparison value Z if the mathematical test range Y matches a comparison value Z.
  • the greyscale data sets of the comparison images and test images are preferably fed to the neural network as comparison data sets and test data sets, as they have been explained in connection with FIG. 8, for example.
  • FIG. 14 and 15 show a further exemplary embodiment of the method described above, in which the surface structure of a wooden panel is mechanically scanned and the topography is thus determined as a contour line profile. Devices that are also used for measuring surface roughness can be used for this purpose.
  • the deflection of the probe tip 32 is measured at predetermined time intervals and quantified—here in units of 0.5. This results in a depth profile that is characteristic of the surface structure, which is recorded and stored graphically as an image and mathematically as a series of numbers (shown below in FIG. 14).
  • FIG. 15 shows the surface of a side edge 2 with drawn-in lines for determining the position and the measurement section.
  • a matrix can be mathematically formed from the n vectors from the n different height profiles.
  • Comparison images and test images or the associated comparison data records and test data records can then be generated from individual vectors or from a plurality of vectors.
  • a characterization data record is preferably determined and stored in one method step for the plurality of wood-based panels.
  • Each comparison data set is then linked to the characterization data set for the respective wood-based panel and the tested wood-based panel is assigned the characterization data set linked to the comparison data set if the test data set matches a comparison data set.
  • Characteristic data can thus be assigned to the tested wood-based panel by identifying the wood-based panel.
  • Characterization dataset stored, either as separate and pointer-linked datasets or as one combined dataset.
  • characterization data set At least one datum selected from the non-exhaustive list is stored in the characterization data set:
  • the associated wood-based material board can be characterized with this data of the characterization data set, with further data being able to be added in the course of the further treatment of the wood-based material board.
  • the characterization data can thus be determined after the wood-based panel has been manufactured and can be taken into account during further processing and treatment.
  • the characterization data set of the separated wood-based material board is linked to the respective comparison data set for each partial format.
  • the wood-based panel is linked to the generated partial formats and, if necessary, data for separating the wood-based panel is added to the characterization data. For example, the processing time and the new dimensions of the part format are added to the characterization data. The further comparison data of the unchanged side edges of the partial format are then likewise linked to the changed characterization data record.

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Physics & Mathematics (AREA)
  • Evolutionary Computation (AREA)
  • Multimedia (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • General Health & Medical Sciences (AREA)
  • Medical Informatics (AREA)
  • Software Systems (AREA)
  • Databases & Information Systems (AREA)
  • Computing Systems (AREA)
  • Artificial Intelligence (AREA)
  • Health & Medical Sciences (AREA)
  • Quality & Reliability (AREA)
  • Image Analysis (AREA)
  • Investigating Strength Of Materials By Application Of Mechanical Stress (AREA)

Abstract

L'invention concerne un procédé d'identification de panneaux à base de bois. Au moins une image de comparaison pour une section prédéfinie d'un bord latéral est enregistrée pour une multiplicité de panneaux à base de bois. Un ensemble de données de comparaison est déterminé à partir de l'image de comparaison sur la base de la structure, en particulier sur la base de la géométrie, la topographie et/ou la répartition des éléments à base de bois et stocké. Une image de test pour la section prédéfinie d'un bord latéral d'un panneau à base de bois est enregistrée à au moins un instant ultérieur. Un ensemble de données de test est déterminé à partir de l'image de test sur la base de la structure, en particulier sur la base de la géométrie, de la topographie et/ou de la répartition des éléments à base de bois de la même manière que pour les ensembles de données de comparaison. L'ensemble de données de test du panneau à base de bois testé est comparé aux ensembles de données de comparaison de multiples panneaux à base de bois et le panneau à base de bois testé est identifié en tant que panneau à base de bois pour lequel un ensemble de données de comparaison a été généré dans le cas où l'ensemble de données de test correspond à l'ensemble de données de comparaison. Le procédé résout le problème technique d'amélioration et de simplification de l'identification de panneaux à base de bois.
EP22723611.4A 2021-04-22 2022-04-20 Procédé d'identification de panneaux à base de bois Pending EP4327302A1 (fr)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
EP21169861.8A EP4080405A1 (fr) 2021-04-22 2021-04-22 Procédé d'identification de plaques en matériau dérivé du bois
PCT/EP2022/060318 WO2022223563A1 (fr) 2021-04-22 2022-04-20 Procédé d'identification de panneaux à base de bois

Publications (1)

Publication Number Publication Date
EP4327302A1 true EP4327302A1 (fr) 2024-02-28

Family

ID=75639786

Family Applications (2)

Application Number Title Priority Date Filing Date
EP21169861.8A Pending EP4080405A1 (fr) 2021-04-22 2021-04-22 Procédé d'identification de plaques en matériau dérivé du bois
EP22723611.4A Pending EP4327302A1 (fr) 2021-04-22 2022-04-20 Procédé d'identification de panneaux à base de bois

Family Applications Before (1)

Application Number Title Priority Date Filing Date
EP21169861.8A Pending EP4080405A1 (fr) 2021-04-22 2021-04-22 Procédé d'identification de plaques en matériau dérivé du bois

Country Status (5)

Country Link
US (1) US20240185411A1 (fr)
EP (2) EP4080405A1 (fr)
CA (1) CA3215993A1 (fr)
DE (1) DE112022002283A5 (fr)
WO (1) WO2022223563A1 (fr)

Family Cites Families (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2012083769A1 (fr) * 2010-12-24 2012-06-28 深圳兆日科技股份有限公司 Procédé et système anti-contrefaçon utilisant des caractéristiques physiques d'une substance
DE102013004332B4 (de) 2013-03-14 2016-05-12 Fritz Egger Gmbh & Co. Og Verfahren und Vorrichtung zur Herstellung eines Lignocellulose enthaltenden, plattenförmigen Werkstoffs

Also Published As

Publication number Publication date
CA3215993A1 (fr) 2022-10-27
WO2022223563A1 (fr) 2022-10-27
US20240185411A1 (en) 2024-06-06
EP4080405A1 (fr) 2022-10-26
DE112022002283A5 (de) 2024-03-14

Similar Documents

Publication Publication Date Title
DE102018129425B4 (de) System zur Erkennung eines Bearbeitungsfehlers für ein Laserbearbeitungssystem zur Bearbeitung eines Werkstücks, Laserbearbeitungssystem zur Bearbeitung eines Werkstücks mittels eines Laserstrahls umfassend dasselbe und Verfahren zur Erkennung eines Bearbeitungsfehlers eines Laserbearbeitungssystems zur Bearbeitung eines Werkstücks
EP1290408B1 (fr) Procede et dispositif permettant la detection photogrammetrique de la forme d'un objet dans l'espace
DE102013205762B4 (de) Informationsverarbeitungsvorrichtung, die eine dreidimensionale Formvermessung durchführt, Informationsverarbeitungsverfahren und Speichermedium
EP2144036A2 (fr) Procédé et dispositif destinés à la numérisation en 3D d'un objet
DE102009023756B4 (de) Verfahren zum Herkunftsnachweis und zur Urheberschaft von Bildern
DE102017215334A1 (de) Verfahren, Computerprogrammprodukt und Messsystem zum Betrieb mindestens eines Triangulations-Laserscanners zur Identifizierung von Oberflächeneigenschaften eines zu vermessenden Werkstücks
DE112014004190T5 (de) Positurschätzverfahren und Roboter
DE4309802A1 (de) Produktionsnahe Farbkontrolle mit bildgebenden Sensoren
EP4072808B1 (fr) Procédé assisté par ordinateur et dispositif pour contrôler la qualité d'un béton apparent
DE102014114440A1 (de) Verfahren zum Klassifizieren eines bekannten Objekts in einem Sichtfeld einer Kamera
EP3671645A1 (fr) Procédé et dispositif de génération d'une reconstruction tridimensionnelle d'un objet
DE112010002677T5 (de) Verfahren und vorrichtung zum bestimmen einer formübereinstimmung in drei dimensionen
EP2641539A1 (fr) Procédé destiné à la détermination des dimensions d'une partie du corps; procédé et dispositif destiné à la détermination des dimensions d'un vêtement
DE102016100134B4 (de) Verfahren und Vorrichtung zum Untersuchen eines Objekts unter Verwendung von maschinellem Sehen
DE102016100132B4 (de) Verfahren und Vorrichtung zum Untersuchen eines Objekts unter Verwendung von maschinellem Sehen
DE102017219244B4 (de) Inspektionsbedingungsbestimmungseinrichtung, inspektionsbedingungsbestimmungsverfahren und inspektionsbedingungsbestimmungsprogramm
DE112019006855T5 (de) Simulationsvorrichtung und simulationsverfahren
DE102005043753A1 (de) Verfahren zur Bestimmung der Raumkoordinaten von Oberflächenpunkten eines Prüflings und Vorrichtung zur Durchführung des Verfahrens
WO2022223563A1 (fr) Procédé d'identification de panneaux à base de bois
DE102018218611B4 (de) Verfahren und Recheneinrichtung zum Erzeugen eines dreidimensionalen CAD-Modells
DE102019100011B4 (de) Verfahren zur 3D-Rekonstruktion eines Objekts
DE102014108924B4 (de) Ein halbüberwachtes Verfahren zum Trainiern eines Hilfsmodells zur Erkennung und Erfassung von Mehrfachmustern
EP1435065A1 (fr) Determination automatique de modeles geometriques pour des reconnaissances optiques partielles
WO2020126926A1 (fr) Procédé d'étalonnage pour améliorer la vérification de motifs d'authentification par l'intermédiaire d'appareils d'acquisition d'images numériques
AT514749A1 (de) Verfahren zum Prüfen der Oberfläche von Gegenständen und Münze mit Echtheitsmerkmal

Legal Events

Date Code Title Description
STAA Information on the status of an ep patent application or granted ep patent

Free format text: STATUS: UNKNOWN

STAA Information on the status of an ep patent application or granted ep patent

Free format text: STATUS: THE INTERNATIONAL PUBLICATION HAS BEEN MADE

PUAI Public reference made under article 153(3) epc to a published international application that has entered the european phase

Free format text: ORIGINAL CODE: 0009012

STAA Information on the status of an ep patent application or granted ep patent

Free format text: STATUS: REQUEST FOR EXAMINATION WAS MADE

17P Request for examination filed

Effective date: 20231005

AK Designated contracting states

Kind code of ref document: A1

Designated state(s): AL AT BE BG CH CY CZ DE DK EE ES FI FR GB GR HR HU IE IS IT LI LT LU LV MC MK MT NL NO PL PT RO RS SE SI SK SM TR