• 专利附录
  • 专利说明
  • 权利要求
  • 类似技术
  • 同族专利
  • 引用文献
  • 法律状态
  • 优先权
    • 专利摘要:
    System and procedure to determine the wear on cutting inserts used in milling operations, from a digital image (9) of a cutting head (1) of a milling machine. The method comprises locating a cutting insert (10) in the digital image (9); detecting the main cutting edge (18a); cutting a wear region (19) containing the main cutting edge (18a); dividing the wear region (19) into a plurality of subregions (22); describe each subregion (22) by means of texture descriptors, generating a vector of characteristics for each subregion (22); classify the subregions (22) into categories using the characteristics vector of the subregions (22) and a supervised learning algorithm; determine the state of the cutting insert (10) according to the classification of the subregions (22). (Machine-translation by Google Translate, not legally binding)
    公开号:ES2608469A1
    申请号:ES201631539
    申请日:2016-11-30
    公开日:2017-04-11
    发明作者:Enrique ALEGRE GUTIERREZ；María Teresa GARCÍA ORDÁS；Víctor González Castro；Rocío Aláiz Rodríguez
    申请人:Universidad de Leon；
    IPC主号:
    专利说明:

    Field of the invention The present invention encompasses in the field of wear monitoring systems of the cutting tools of a milling machine, and more specifically the monitoring of wear of the cutting plates mounted on the milling cutters.
    BACKGROUND OF THE INVENTION The wear of the cutting tools used in a milling machine is caused by a wide variety of factors that act on the cutting edge, such as corrosion, abrasion or fatigue. The condition of the cutting plates used in the mills directly influences the quality of the machined part. Therefore, the monitoring of this wear is crucial to be able to replace the cutting tool at the optimum time, both for the cost of the milling head that would have to be replaced in case of breakage, as well as for the costs derived from the unscheduled shutdown of the milling machine or for the cost due to the non-conformity of the manufactured parts. The optimization of these substitution operations entails a significant improvement in terms of efficiency and competitiveness of production systems.
    The most studied techniques for the evaluation of wear are based on the monitoring of signals that have some correlation with the level of wear of the cutting tool. Thus, there are works that propose using force measures to estimate wear in real time (Franci Cus, U. Z., "Real-Time Cutting Tool Condition Monitoring in Milling", Joumal of Mechanical Engineering, 2011). Other studies are based on the measurement of vibrations (Rmili, W., Ouahabi, A., Serra, R & Leroy, R, "An automatic system based on vibratory analysis for cutting tool wear monitoring", Measurement, 2016, 77, 117 -123) or using acoustic signals (T. Suryasekhara Reddy, Dr. C. Eswara Reddy, S. Prabhavathi, "Belief Network Based Acoustic Emission Analysis For Real Time Monitoring In CIM Environment", International Journal of Technology And Engineering System (IJTES ), 2010). However, all these signals are affected by the noise of industrial environments, which reduces the reliability of these wear evaluation systems.
    Other monitoring methods have also been proposed, such as those based on artificial vision, which directly measure the wear of the tool, thus achieving greater precision and reliability. Recent advances in the field of digital image processing have allowed these techniques to be applied by analyzing, for example, the contour of the shape of the wear region (García-Ordás, MT., Alegre, E., González-Castro, V. and GarcíaOrdás, D., ~ aZIBO: A New Descriptor Based in Shape Moments and Rotational Invariant Features ", 22nd International Conference on Pattern Recognition (ICPR), 2014). Other studies analyze the texture of worn areas and compare it with that of intact areas (Samik Dutta, Surjya K. Pal and Ranjan Sen, "Progressive tool flank wear monitoring by applying discrete wavelet transform on turned surface images», Measurement, 2016). Unlike the method proposed in the present invention, this technique is applied to turning processes instead of milling, and the life of the tool is determined completely differently, measuring the average flank wear.
    Description of the invention The present invention relates to an automated system and method for describing and estimating the wear of cutting plates in milling operations using artificial vision techniques. The procedure evaluates the state of the tool from grayscale digital images by estimating the number of subregions that present wear characterizing the surface by means of texture description techniques and classifying it as worn or not, based on a model trained through a supervised learning method.
    The process of the present invention allows to determine the wear on cutting plates used in milling operations from a digital image of a cutting head of a milling machine with at least one cutting plate. The procedure comprises a first stage of locating at least one cutting insert in the digital image and, for at least one cutting insert located:
    - Detect, in the digital image, the main cutting edge of the cutting insert.
    - Trim a wear region that contains the main cutting edge detected.
    - Divide the wear region into a plurality of subregions.
    - Describe each sub-regions using at least one texture descriptor, generating a vector of characteristics for each sub-region. -Classify sub-regions into two or more categories using the sub-regions feature vector and a supervised learning algorithm.
    - Determine the status of the cutting insert based on the classification of the subregions.
    The location of at least one cutting insert in the digital image may comprise the
    5 screw detection of each cutting insert. Screw detection can be performed by an object-based method of shape detection, preferably by the histogram of oriented gradients.
    In a preferred embodiment of the invention, the detection of the main cutting edge of the
    10 cutting insert is done by an edge detector. Both the clipped wear region and the subregions preferably completely cover the main cutting edge and at least partially the adjacent cutting edges. The step of determining the state of the cutting insert may comprise determining the level of wear of the main cutting edge and of the adjacent cutting edges.
    In a preferred embodiment of the invention, the cutting of the wear region comprises the following steps: -Binarization by thresholding of a region of interest where the cutting insert is located.
    20 -Detection and elimination of the central circle of the cutting insert. -Detection of the cutting edges of the cutting insert. -Cutting of a wear region that comprises the entire main cutting edge and at least partially the adjacent cutting edges.
    25 The description of the subregions may include the extraction of texture-based features using the technique of local binary patterns. In a preferred embodiment of the invention, the supervised learning algorithm used in the sub-region classification is the support vector machine algorithm.
    30 Sub-regions can be classified at least as worn or intact. The classification of sub-regions may include training a machine learning system with a labeled training set; and classifying, using the trained system, the subregions represented by their texture-based feature vector, as intact or worn out using a supervised classification method.
    The status of the cutting insert can be determined based on the number or proportion of sub-regions classified as worn, as well as the location of the sub-regions classified as worn. In the stage of determining the status of the cutting insert, it can be considered worn if the number or proportion of subregions classified as worn exceeds a certain threshold. The step of determining the state of the cutting insert may comprise determining the level of wear of the main cutting edge.
    The procedure may comprise a stage of acquiring the digital image. The method may comprise a step of converting the digital image to a grayscale image.
    A second aspect of the present invention relates to a system for monitoring the wear of cutting plates used in milling operations. The system comprises an image capture system configured to acquire a digital image of a cutting head of a milling machine with at least one cutting insert, and a processing system comprising data processing means configured to locate at least a cutting insert in the digital image and, for at least one localized cutting insert:
    - Detect, in the digital image, the main cutting edge of the cutting insert.
    - Trim a wear region that contains the main cutting edge
    detected;
    - Divide the wear region into a plurality of subregions.
    - Describe each sub-regions using at least one descriptor of
    texture, generating a vector of characteristics for each subregion.
    - Classify subregions into two or more categories using the vector of
    characteristics of the subregions and a supervised learning algorithm.
    - Determine the status of the cutting insert based on the classification of
    The subregions
    The image capture system comprises a digital camera. The image capture system may comprise a lighting system. In a preferred embodiment of the invention the lighting system comprises at least one red LED light bar. The processing system may be configured to locate at least one cutting insert in the digital image by detecting the screw of each cutting insert. System
    processing may comprise an edge detector configured to perform the
    detection of the main cutting edge of the cutting insert.
    A further aspect of the present invention relates to a program product comprising means of program instructions for carrying out the previously defined procedure when the program is executed in a processor. The program product may be stored in a program support medium.
    The process of the invention allows the level of wear of the cutting plates to be automatically estimated and, optionally, to classify the insert as "worn ~ or" intact "from digital images taken from the cutting head in milling machines using any system based in a computer that allows to acquire, process and classify the descriptors extracted from these images.The proposed procedure is of special interest when the system is portable, of reduced dimensions and cost, although with the benefits of a mid-range computer, which is connect a digital camera to acquire the images Some examples of portable computers are the Raspberry Pi, the HummingBoard, the Banana Pi, the BeagleBone Black or the Odroid U3. The digital images used are taken directly in grayscale or converted to This format for further processing.
    The process of the present invention determines the level of wear of the cutting plates taking into account the number of subregions classified as worn as well as the location of said wear, after describing by using texture-based techniques. This procedure applies to milling processes in the manufacturing industry, although it could be adapted to other types of manufacturing processes involving this type of inserts.
    In the description of the present invention, the term "image" is generally used to refer to both still images or photographs, as well as to each of the frames present in a sequence of images or video.
    The procedure proposed in the present invention comprises the following steps:
    one. Acquisition of images.
    2. Location of the cutting and cutting inserts of the cutting edge region.
    3. Image division of cutting edge into sub-regions.
    Four. Description of the sub-regions of the cutting edge.
    5. Classification of sub-regions of the cutting edge.
    In the stage of acquiring a set of images, images of the cutting head are captured with the inserts that are to be analyzed, either as still images or from a video sequence, using a digital camera and a lighting system. The camera is connected to a computer, in which the acquired digital image is analyzed. Preferably, the images are captured in grayscale, although the procedure would work analogously by acquiring color images and converting them to grayscale. Each image contains one or several cutting inserts.
    The method may comprise a preprocessing step of the captured image of the cutting head in order to improve the contrast or convert the image to grayscale, if necessary. The image resulting from this first stage will be referred to as "image" or "captured image", regardless of whether the preprocessing has been performed.
    or not.
    Subsequently, the location of the cutting and cutting inserts of the cutting edge region is performed. Obtaining the regions of interest where each cutting insert is located is done automatically. The first step involves detecting in the captured image the central circular region of the cutting insert, in which the screw that fixes it to the cutting head is located, and removing it from the image. Next, the left edge of the insert is detected by edge detectors and morphological operations. Once detected, the rectangular region that goes from the cutting edge to the circular region of the fixing screw is cut, without containing the latter.
    The procedure can incorporate a preprocessing stage of the image with the region of interest where the insert is located or of the rectangular regions where the cutting edges are located, to improve the contrast thereof and facilitate the operation.
    The images resulting in this stage of location of the cutting and cutting inserts of the region of the cutting edge will hereinafter be referred to as "images of the cutting edge" and include the entire surface of the insert present in the previously cut region.
    Next, the division of the images of the cutting edge into sub-regions is done by means of a mesh of the image in which a certain number of sub-regions is obtained, according to multiple possible configurations of the division. Each one of the configurations of the division of the images of the cutting edge has the cutting edge on one side of the image, for example on the left side, and they represent by means of a circular sector, to the right of the cut rectangle, the area of the fixing thyme The regions resulting from the stage of division into sub-regions will be referred to as "sub-regions of the cutting edge".
    Finally, the description and classification of the sub-regions is carried out. The description of the sub-regions of the cutting edge is carried out using texture descriptors applied to gray-scale digital images, thus creating a vector of characteristics for each sub-region. In accordance with a preferred embodiment of the invention, the optimal methods of description are those that use the Local Binary Patterns technique, called Local Binary Patterns (LBP), and methods derived therefrom.
    Each of the subregions into which the clipped region is divided around the cutting edge is classified as worn or intact by supervised classification techniques. In accordance with a preferred embodiment of the invention, the optimal supervised classification methods employed are the support vector machines, called English Support Vector Machines (SVM). Subsequently, the percentage of worn subregions for each cutting insert is determined, as well as the location of said regions, which implies the location of wear.
    The classification procedure includes the determination of the percentage of sub-regions worn for each cutting insert, as well as the location of wear. Likewise, the insert is classified as intact, meaning that it is still in a position to continue being used, if the number of subregions does not exceed a certain threshold and as worn, otherwise.
    Finally, the present invention also extends to computer programs, particularly those stored in a physical medium, adapted to carry out the described procedure. The program may have the form of source code, object code, an intermediate source between code and object code - for example, in a partially compiled form - or in any other form suitable for use in the implementation of the processes according to the present invention. .
    The physical support can be any entity or device capable of supporting the program. For example, the physical media may include a storage medium, such as a ROM, an optical memory (such as a CO ROM, OVO ROM, Blu-Ray), a semiconductor ROM, a flash memory, a memory solid state (SSO)
    or a magnetic recording medium, such as a hard disk. In addition, the physical support can be a transmissible carrier, such as an electrical or optical signal that could be transported through electrical or optical cable, or by any other means. When the program is incorporated into a signal that can be directly transported by a cable or other device or medium, the carrier may be constituted by said cable or other device or means. As a variant, the carrier can be an integrated circuit in which the program is included, said integrated circuit being adapted to execute or to be used in the execution of the corresponding processes.
    BRIEF DESCRIPTION OF THE FIGURES Next, a series of figures that help to better understand the invention and that expressly relate to an embodiment of said invention that is presented as a non-limiting example thereof is described very briefly.
    Figure 1 shows a simplified scheme of a cutting plate wear determination system according to the present invention.
    Figures 2A and 28 show, respectively, a schematic elevational and plan view of the image capture system used in a preferred embodiment of the invention.
    Figure 3 shows a simple representation of the captured image where the head can be seen with the cutting plates mounted.
    Figure 4 represents, for the example of Figure 3, the image where only the individual cutting plates appear.
    Figure 5A shows a cut-out of the image of Figure 4 with an individual cutting insert. Figure 58 in turn illustrates a cutout of Figure SA in a wear region around the cutting edge of the individual cutting insert. Figure SC shows a representation of the sub-regions of the cutting edge.
    Figures 6A-6D show four examples of the division of the wear region into
    subregions
    Figure 7 shows, in a simplified way, an example of subdivision of the region of wear in subregions that works best for machining under severe conditions.
    DETAILED DESCRIPTION OF THE INVENTION Figure 1 shows a simplified scheme of a monitoring system for cutting plate wear according to a possible embodiment of the present invention. The cutting plates 10 whose wear is to be monitored are screwed to the cutting head 1 of a milling machine. An image capture system 2 is responsible for acquiring digital images 9 of the cutting head 1 with the cutting plates 10.
    The image capture system 2 comprises, in a preferred embodiment shown in Figure 2A (elevation view) and Figure 2B (plan view), a digital camera 7 and a lighting system 8 located in the central part, between the digital camera 7 and the cutting head 1. The acquisition of the digital images and the complete processing of the images of the cutting plates is carried out during the moments in which the milling machine is in a state of rest due to the change of the metal parts used in milling processes.
    The digital images 9 captured by the digital camera 7 are sent to a processing system 3, either by means of a USB connection, by means of a firewire connection, or in the way that the digital camera 7 requires. The processing system 3 comprises data processing means, preferably based on microprocessor or microcontroller. The processing system 3 can be implemented in a portable system
    (e.g. Raspberry Pi, laptop, tablet, smartphone), on a desktop computer or on any other electronic device with sufficient data processing capacity.
    The processing system 3 locates the cutting plates 10 in the captured digital images 9 and divides the images into subregions. The processing system 3 is responsible for making a description of the different sub-regions of the cutting edge of the cutting plates 10, obtaining texture descriptors of the sub-regions.
    Said descriptors are stored either in a data file of the description 4, or in an internal RAM of the processing system 3.
    Likewise, the processing system 3 is responsible for the classification of each of the sub-regions of the cutting edge. The results obtained in the classification, together with the trained model, are stored either in a data file of classification 5, or alternatively in an internal RAM of the processing system 3. These results can be shown to the user through a display screen 6. Next, each step of the wear monitoring procedure of the cutting plates of the present invention is described.
    In a first stage of image acquisition, the cutting head 1, which has the cutting plates 10 mounted, is illuminated and focused with the optics of the digital camera 7 so that the image appears sharp. In a preferred embodiment of the invention, shown in Figures 2A and 28, the lighting system 8 comprises two LEO red light bars, since this type of lighting intensifies the contrast between the light and dark areas of the image, highlighting the clear areas of the edge, and produces less glare on the metal parts. Also, in a preferred embodiment of the invention, the image capture system 2 is isolated from the exterior light by means of a screen or is located inside a cabin, so that uniform illumination can be obtained.
    Next, a preprocessing step of the obtained images may be necessary, which is preferably performed by the image capture system 2 itself or, alternatively, by the processing system 3. If the acquired images are color images, then The preferred embodiment of the invention is converted to grayscale images. To improve the contrast, you can use any method of equalizing or improving the contrast of the image.
    The result of this first stage of image acquisition is the image represented in the example of Figure 3, which shows the digital image 9 captured (with or without preprocessing) by the image capture system 2, where the head can be seen of cut 1 with the plates of cut 10 mounted. The cutting plates for milling are usually rhombohedral or triangular, although other geometries may exist for special cases. The cutting plates normally have three or four flat cutting edges to attack the material.
    Next, the processing system 3 performs the location of the cutting plates 10 and the cutting of the region of the cutting edge. First of all, the thymes 11 are located in the center of the cutting insert 10. In a preferred embodiment of the invention, an object-based method of detection is used and requires training, such as use of Histograms of Oriented Gradients (HOG). The detection of the screws 11 can also be performed with any other object detection method specifically trained with the images used. Next, an edge detector is applied that allows to locate the edges of the cutting insert 10 to obtain an image 14 where each of the cutting inserts 10 are located, as seen in Figure 4. This figure is a representation (and therefore the background is not shown) where the plates present in the captured image are located. Subsequently, the main cutting edge is detected and the region of interest where the wear will be studied is cut for each insert. In a preferred embodiment of the invention the Canny edge filter is the method used, although other methods can be used by modifying post-processing.
    From the image 14, where all the individual plaquitas located appear, each cutting insert 10 is separated individually, obtaining an image of individual insert 16 for each cutting insert 10, as shown in the example of Figure SA. Subsequently, the straight lines present in each individual insert 16 image are detected to find the main cutting edge 18a of the different cutting inserts 10. Each cutting insert 10 normally has three or four cutting edges. All cutting edges can be used for machining, depending on the assembly of the cutting insert one or the other cutting edge will be used. A cutting blade 10 with four cutting edges (18a, 18b, 18c, 18d) is shown in Figure 5A. The cutting edge 18a is indicated as the cutting edge that is currently being used for machining the part, depending on the position of the cutting insert 10 and the direction of rotation of the cutting head 1. In the Figure 5A the main cutting edge 18a is the left edge. The rest of the cutting edges (upper cutting edge 18b, lower cutting edge 18c and opposite cutting edge 18d), are those edges that are not currently being used for machining. The main cutting edge 18a is the one that should preferably be inspected because it is currently being used and may collapse. The main cutting edge 18a is therefore the edge that must be determined to be able to continue cutting. If it is determined that it is not in good condition to continue cutting, the cutting insert 10 can be rotated to select another cutting edge (18b, 18c, 18d) in good condition, which in that case would become the main cutting edge .
    Alternatively, the cutting insert 10 can be replaced if no cutting edge remains intact on the insert.
    For this a line detection method is used, such as the standard Hough transform (SHT). Finally, each image of individual insert edges 16 is trimmed in a wear region 19 around the main cutting edge 18a, which is obtained by reference to the line with greater angle to the horizontal axis located more to the left of each cutting plate 10, which corresponds to the cutting edge due to the direction of rotation of the milling machines, generating the image of the main cutting edge 20 for each cutting plate 10. The result of this cutting is shown in a simplified manner in Figure 5S.
    Once the image of the main cutting edge 20 is obtained, the next step is to divide this image into several subregions 22 to later describe each of them independently, as shown in Figure 5e (in this case, it has been divided in four subregions 22). It is intended to characterize in this way the wear present in different subregions 22 of the cutting insert 10 by describing its texture. Sub-regions 22 may have different sizes and orientations. The choice of the specific configuration of number, sizes and orientations of the subregions is made according to the machining problem to which it is necessary to adapt.
    Figures 6A-6D show, by way of example, four different configurations of the division of the wear region into subregions 22. Each of the configurations shown in said figure have the main cutting edge 18a on the left side. On the right side of the cut rectangle a semicircular region is represented where the screw 11 that fixes the cutting insert 10 to the head 1 is located. Tangentially to said semicircular region, the boundary of the surface to be represented is represented by dashed line 23 analyze for each cutting insert 10. The different rectangles are preferred configurations of the arrangement and size of the subregions 22 of the wear region 19 around the main cutting edge 18a.
    Figure 7 illustrates the scheme that, based on the experiments performed, works best for machining under severe conditions. In this example, in the image of the main cutting edge 20, a vertical strip corresponding to the area of the main cutting edge 18a has been selected, which has been divided into nine sub-regions 22. The areas corresponding to the cutting edge have also been selected upper and lower to evaluate the wear information found in these upper and lower zones, resulting in the final division into eleven sub regions 22 whose simplified scheme appears in said figure. It can be seen that the left strip is subdivided first into nine
    5 regions of equal width and area and subsequently two new sub regions are created, at the top and bottom, these being wider and narrower than the previous ones.
    Subsequently, each of the subregions 22 is described by a vector of characteristics that represent them. The descriptor used in the preferred embodiment of this
    The invention is that of local binary patterns (LBP). This descriptor is a texture operator that labels each pixel of the image by analyzing its neighborhood, studying whether the gray level of each pixel exceeds a certain threshold and encoding said comparison by a binary value. To calculate LBP in a grayscale image, the equation is used:
    Isx ;;, O
    LBPp, R = Is: g, -g,) 2 ', S: x) = {
    "" 0 Osx <O
    Where P is the number of neighbors to be considered, R is the size of the neighborhood and,
    Be: and Bp are the gray level values of the central pixel and each of the P pixels of the
    20 neighborhood respectively.
    At the end of this stage, a vector of characteristics is available for each of the subregions 22 of the cutting edge.
    25 Next, a classification of the subregions 22 of the cutting edge is made. Each of the calculated feature vectors is classified using some previously trained supervised learning method. This classification returns, for each characteristic vector, a binary value, which determines whether the sub-region of the cutting edge has wear or not. After classification, the number of regions is determined
    30 worn, which will be the value that defines the state of the cutting insert 10. Additionally, their location is specified, either in the main cutting edge 18a, or in the corresponding wear region areas 19 to the adjacent cutting edges (upper cutting edge 18b and lower cutting edge 18c), which may also show wear due to the cutting edge used in a previous machining.
    Finally, the procedure may include the identification of the overall status of the insert
    5 cut 10. To this end, the proportion of wear of the cutting insert 10 is determined from the number of wear subregions 22 found in the image of the main cutting edge 20, so that if this proportion of wear exceeds a certain threshold the cutting insert 10 is classified as worn and, otherwise, it is considered to belong to the category of intact cutting insert. Alternatively, for the
    The identification of the overall state of the cutting insert instead of a wear ratio can be considered the number of subregions classified as worn. In a preferred embodiment of this invention, a value of between three and five subregions is used as a threshold.
    权利要求:
    Claims (29)
    [1]
    1. Procedure for determining the wear on cutting plates used in milling operations, from a digital image (9) of a cutting head (1) of a milling machine with at least one cutting plate (10), characterized because it comprises locating at least one cutting insert (10) in the digital image (9) and, for at least one cutting insert (10) located:
    - detect, in the digital image (9), the main cutting edge (18a) of the cutting insert (10);
    - trim a wear region (19) that contains the main cutting edge (18a) detected;
    - divide the wear region (19) into a plurality of subregions (22);
    - describe each sub-regions (22) by at least one texture descriptor,
    generating a vector of characteristics for each subregion (22); -classify sub-regions (22) into two or more categories using the sub-regions feature vector (22) and a supervised learning algorithm; -determine the status of the cutting insert (10) based on the classification of the subregions (22).
    [2]
    2. Method according to claim 1, characterized in that the location of at least one cutting insert (10) in the digital image (9) comprises the detection of the screw (11) of each cutting insert (10).
    [3]
    3. Method according to claim 2, characterized in that the detection of the screw (11) is carried out by a method of object-based object detection.
    [4]
    Four. Method according to claim 3, characterized in that the object detection method used is the histogram of oriented gradients.
    [5]
    5. Method according to any of the preceding claims, characterized in that the detection of the main cutting edge (18a) of the cutting insert (10) is carried out by means of an edge detector.
    [6]
    6. Method according to any of the preceding claims, characterized in that
    both the cut-out wear region (19) and the subregions (22) completely cover the main cutting edge (18a) and at least partially the adjacent cutting edges (18b, 18c).
    [7]
    7. Method according to claim 6, characterized in that the step of determining the state of the cutting insert (10) comprises determining the level of wear of the main cutting edge (18a) and of the adjacent cutting edges (18b, 18c).
    [8]
    8. Method according to any of the preceding claims, characterized in that the cutting of the wear region (19) comprises: -binarization by thresholding of a region of interest where the insert is located
    cutting (10); -detection and elimination of the central circle of the cutting insert (10); -detection of the cutting edges (18a, 18b, 18c, 18d) of the cutting insert (10); -cutting a wear region (19) that completely comprises the cutting edge
    main (18a) and at least partially adjacent cutting edges (18b, 18c).
    [9]
    9. Method according to any of the preceding claims, characterized in that the description of the subregions (22) comprises the extraction of texture-based characteristics using the technique of local binary patterns.
    [10]
    10. Method according to any of the preceding claims, characterized in that the supervised learning algorithm used in the classification of the subregions (22) is that of support vector machines.
    [11]
    eleven. Method according to any of the preceding claims, characterized in that the subregions (22) are classified at least as worn or intact.
    [12]
    12. Method according to any of the preceding claims, characterized in that the classification of the subregions (22) comprises: -training an automatic learning system with a labeled training set;
    - classify, using the trained system, the subregions (22) represented by their vector of texture-based characteristics, as intact or worn out using a supervised classification method.
    [13]
    13. Method according to any of the preceding claims, characterized in that the state of the cutting insert (10) is determined according to the number or proportion of sub-regions (22) classified as worn, as well as the location of the sub-regions (22) classified as worn out
    [14]
    14. Method according to any of the preceding claims, characterized in that in the stage of determining the state of the cutting insert (10), it is considered worn if the number or proportion of subregions (22) classified as worn exceeds a certain threshold.
    [15]
    fifteen. Method according to any of the preceding claims, characterized in that the step of determining the state of the cutting insert (10) comprises determining the level of wear of the main cutting edge (18a).
    [16]
    16. Method according to any of the preceding claims, characterized in that it comprises a step of acquiring the digital image (9).
    [17]
    17. Method according to any of the preceding claims, characterized in that it comprises a step of converting the digital image (9) to a grayscale image.
    [18]
    18. Wear monitoring system for cutting inserts used in operations
    of milling, characterized in that it comprises: an image capture system (2) configured to acquire a digital image
    (9) of a cutting head (1) of a milling machine with at least one cutting insert (10);
    a processing system (3) comprising data processing means configured to locate at least one cutting insert (10) in the digital image (9) and, for at least one cutting insert (10) located:
    - detect, in the digital image (9), the main cutting edge (18a) of the cutting insert (10); -cut a wear region (19) that contains the main cutting edge
    (188) detected; - dividing the wear region (19) into a plurality of subregions (22); -describe each sub-regions (22) by at least one descriptor of
    texture, generating a vector of characteristics for each subregion (22); -classify the sub regions (22) into two or more categories using the vector
    of characteristics of the subregions (22) and a supervised learning algorithm;
    - determine the status of the cutting insert (10) based on the
    classification of sub-regions (22).
    [19]
    19. System according to claim 18, characterized in that the image capture system (2) comprises a digital camera (7).
    [20]
    twenty. System according to any of claims 18 to 19, characterized in that the image capture system (2) comprises a lighting system (8).
    [21]
    twenty-one. System according to claim 20, characterized in that the lighting system (8) comprises at least one LEO red light bar.
    [22]
    22 System according to any of claims 18 to 21, characterized in that the processing system (3) is configured to locate the at least one cutting insert (10) in the digital image (9) by detecting the screw (11) of each cutting insert (10)
    [23]
    2. 3. System according to any of claims 18 to 22, characterized in that the processing system (3) comprises an edge detector configured to detect the main cutting edge (18a) of the cutting plate (10).
    [24]
    24. System according to any of claims 18 to 23, characterized in that for processing the wear region (19) the processing system (3) is configured to:
    - perform a binarization by thresholding of a region of interest where the cutting insert is located (10);
    - detect and eliminate the central circle of the cutting insert (10);
    - detect the cutting edges (18a, 18b, 18c, 18d) of the cutting insert (10);
    - trim a wear region (19) that completely comprises the cutting edge
    main (18a) and at least partially adjacent cutting edges (18b, 18c).
    [25]
    25. System according to any of the preceding claims, characterized in that for processing the sub-regions (22) the processing system (3) is
    configured to extract texture-based features using the technique of local binary patterns.
    [26]
    26. System according to any of claims 18 to 25, characterized in that for processing the sub-regions (22) the processing system (3) is configured to:
    - train a machine learning system with a tagged training set;
    - classify, using the trained system, the subregions (22) represented by their feature vector, as intact or worn out using a supervised classification method.
    [27]
    27. System according to any of claims 18 to 26, characterized in that the processing system (3) is configured to determine the state of the cutting insert (10) based on the number or proportion of sub-regions (22) classified as worn, as well as of the location of the subregions (22) classified as worn.
    [28]
    28. A program product comprising means of program instructions for carrying out the procedure defined in any of claims 1-17 when the program is run on a processor.
    [29]
    29. A program product according to claim 28, stored in a program support medium.
    类似技术:
    公开号 | 公开日 | 专利标题
    García-Ordás et al.2017|A computer vision approach to analyze and classify tool wear level in milling processes using shape descriptors and machine learning techniques
    JP2009529196A|2009-08-13|Method and apparatus for detecting eyes in digital images
    JP5376906B2|2013-12-25|Feature amount extraction device, object identification device, and feature amount extraction method
    US9443129B2|2016-09-13|Methods and apparatus for image analysis using threshold compactness features
    CN107705301B|2021-04-13|Highway marking damage detection method based on unmanned aerial vehicle aerial highway image
    CN108364311B|2020-08-25|Automatic positioning method for metal part and terminal equipment
    Khan et al.2013|Counting clustered cells using distance mapping
    Tien et al.2004|Automated visual inspection for microdrills in printed circuit board production
    CN107093175B|2019-07-19|A kind of power battery welding quality detection method
    ES2608469B1|2018-01-30|SYSTEM AND PROCEDURE FOR DETERMINING WEAR IN CUTTING PLATES USED IN MILLING OPERATIONS
    Powbunthorn et al.2012|Assessment of the severity of brown leaf spot disease in cassava using image analysis
    ES2637034B1|2018-07-30|SYSTEM AND PROCEDURE FOR DETERMINING WEAR IN CUTTING PLATES USED IN MILLING OPERATIONS BY CLASSIFICATION OF FORMAT FORMATORS IN DIGITAL IMAGES
    CN108154132A|2018-06-12|A kind of identity card text extraction method, system and equipment and storage medium
    US9600737B2|2017-03-21|Image feature extraction method, image feature extraction apparatus, and recording medium storing program therefor
    Tian et al.2013|Scene character reconstruction through medial axis
    Deb et al.2011|Automatic vehicle identification by plate recognition for intelligent transportation system applications
    CN105405137A|2016-03-16|Shaft quality detection method based on machine vision
    Hmood et al.2017|Scale and rotation invariant character segmentation from coins
    EP2776974B1|2021-01-20|Methods and apparatus for image analysis using threshold compactness features
    Ranguelova2016|A Salient Region Detector for structured images
    Pant et al.2015|Automatic nepali number plate recognition with support vector machines
    Ieamsaard et al.2015|Vertical edge detection-based automatic optical inspection of HGA solder jet ball joint defects
    Galloway et al.2020|Image processing-based identification of dicentric chromosomes in slide images
    Ieamsaard et al.2018|Automatic Optical Inspection of Solder Ball Burn Defects on Head Gimbal Assembly
    KR101864533B1|2018-06-04|Method and Apparatus for Segmentation of Touching Objects using Concave Sections and Concave Points
    同族专利:
    公开号 | 公开日
    ES2608469B1|2018-01-30|
    引用文献:
    公开号 | 申请日 | 公开日 | 申请人 | 专利标题
    US4700224A|1985-06-20|1987-10-13|Agency Of Industrial Science & Technology|Method of image processing for the measurement of tool wear|
    US4845763A|1987-11-06|1989-07-04|General Motors Corporation|Tool wear measurement by machine vision|
    US6249599B1|1996-04-30|2001-06-19|Komatsu Ltd.|Method and apparatus for detecting tool trouble in machine tool|
    法律状态:
    2018-01-30| FG2A| Definitive protection|Ref document number: 2608469 Country of ref document: ES Kind code of ref document: B1 Effective date: 20180130 |
    优先权:
    申请号 | 申请日 | 专利标题
    ES201631539A|ES2608469B1|2016-11-30|2016-11-30|SYSTEM AND PROCEDURE FOR DETERMINING WEAR IN CUTTING PLATES USED IN MILLING OPERATIONS|ES201631539A| ES2608469B1|2016-11-30|2016-11-30|SYSTEM AND PROCEDURE FOR DETERMINING WEAR IN CUTTING PLATES USED IN MILLING OPERATIONS|
    [返回顶部]