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
  • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
  • G06T7/00—Image analysis
  • G06T7/30—Determination of transform parameters for the alignment of images, i.e. image registration
  • G06T7/33—Determination of transform parameters for the alignment of images, i.e. image registration using feature-based methods
  • G06T7/337—Determination of transform parameters for the alignment of images, i.e. image registration using feature-based methods involving reference images or patches
• • G—PHYSICS
  • G06—COMPUTING; CALCULATING OR COUNTING
  • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
  • G06T7/00—Image analysis
  • G06T7/90—Determination of colour characteristics
• • 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
  • G06T2207/10—Image acquisition modality
  • G06T2207/10024—Color image
• • 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

Definitions

• the invention relates generally to the field of tire manufacturing, and is more particularly in the context of tire appearance control operations in progress or at the end of the manufacturing process, in order to determine the conformity of said tires with respect to control references.
• the methods used to perform these treatments consist, as a rule, in comparing an image in two or preferably in three dimensions of the surface of the tire to be inspected, with a reference image in two and preferably in three dimensions of the surface of said tire.
• One of the steps of this process is, in a known manner, to acquire the three-dimensional image of the surface of the tire, using, for example, means based on the principle of optical triangulation, for example a 2D sensor coupled to a laser-type lighting source.
• the two-dimensional image is obtained using the same capture means and then consists of a simple photograph of the surface.
• This two-dimensional image can also be obtained by any means of acquisition of the linear camera type or calculated at the same time as the three-dimensional image in the case of optical triangulation from the intensity of the laser line projected on the tire.
• the image of the tire surface is formed by the set of digital data from the capture of the three-dimensional and two-dimensional image.
• This data from the sensor is transmitted to a processor whose memory contains code instructions able, when executed, to process very large data volumes.
• the following steps consist in carrying out a number of pre-treatments making it easier to use the data forming the image of the surface of the tire.
• the process continues to recalibrate the image of the raised surface of the tire to be inspected with the image of the reference relief surface.
• This step proves particularly complex because of the local deformations generated by the release of the stresses at the outlet of the baking mold.
• This reference image can be derived from a CAD model, or from a model that was used to design the mold from which the tire, or even a reference tire considered as free of anomalies.
• the publication WO2009077539 proposes to determine an affine transformation by seeking the coincidence of previously identified characteristic points on the surface of the tire to be inspected and on the reference surface.
• the publication WO2012 / 052300 proposes a method of registration of the patterns of a tread by matching the sequenced basic patterns in a known manner by using the basic patterns comprising a wear indicator forming a characteristic and easily identifiable image.
• the process ends using numerical methods of inspection and troubleshooting, or by simply comparing the image of the tire to be inspected with the reference image.
• the publication WO2013 / 045593 proposes a method for processing two-dimensional images of smooth surfaces by morphological analysis.
• the publication WO2012 / 143197 focuses more particularly on the treatment of non-measurement points generated by the shadow zones formed by the vents.
• the applications FR1462901 or FR1462898 not yet published, propose algorithms for detection and analysis of conformity of the striations present on the surface of the tire sidewall which are less demanding in computation time than traditional methods by Fourier transformation or by analysis. of the three-dimensional image.
• the object of the invention is to make a contribution that makes it possible to reduce the calculation time required for the inspection of tires at the end of the manufacturing process by allowing the identification of distinct areas of interest on which algorithms are used. specific will be used.
• the method of inspecting a surface of a tire to be inspected by comparison with a reference surface of a reference tire, said surfaces comprising markings or elements in relief is characterized in that the stages in which:
• an image of the relief of the surface of a reference tire taken from the manufacture and considered as being in conformity which image is formed by a two-dimensional image in gray level or in color, and a three-dimensional image in gray level in which at each point of the image is assigned a gray level value proportional to the topographic elevation of this point.
• an interaction of an operator with said processor is carried out by parameterizing the main characteristics and mapping the reference surface of the reference tire by cutting the image of the reference surface of the reference tire into a plurality of zones; distinct interests, each of the zones of interest comprising a characteristic shape or relief, the zones of interest being separated by boundaries, and each of the previously identified zones of interest is assigned one or more selected registration and control algorithms. in a collection of algorithms previously constituted and stored in a memory of said processor,
• said processor performs, by executing the coded instructions, the operations in which:
• the method thus provides, in a preparatory manner, a long and sometimes complex parameterization of the reference envelope, so as to define the most optimal parameters forming the main characteristics of the areas of interest of the reference tire, and to choose carefully the algorithm or algorithms best suited for each of these areas of interest.
• the invention therefore makes it possible, for a given computing power of the processor, to reduce the calculation time necessary for estimating the conformity of an envelope during the control phase, and to make it possible to carry out this control at a compatible rate. with the rhythm of industrial manufacturing.
• a surface of the tire is here understood to mean a surface formed by all or part of the surfaces formed by the inner surface, or by the external surface of the tire comprising the sidewall, the shoulders, the bead or the tread. It goes without saying that when the reference surface of the reference tire represents only a part of the tire surface such as the sidewall, the tread or the inner part, the surface to be inspected of the tire to be inspected represents the surface on which the tire is to be inspected. an identical part of the tire is respectively the sidewall, the tread or the inner part.
• the pretreatment of the acquired image comprises one or more treatments chosen from the following algorithms
• the parameterization of the main characteristics of the reference tire comprises data selected from the following characteristics: o a sum of the value of the gray levels of the columns and lines of the raw or filtered three-dimensional image, forming a one-dimensional signature,
• the collection of registration algorithms comprises at least one registration method for registering the relief of an area of interest of the reference surface with the relief of the area of interest of the surface to be inspected.
• the registration method is chosen, in order of increasing complexity, from one or more of the following methods:
• the collection of algorithms includes first level algorithms based on analyzes such as the comparison of the contours of characteristic images, the comparison of geometric measurements between characteristic points, the comparison of geometric moments.
• control algorithms comprises at least one specific method for analyzing the areas of interest in which the characteristic relief is formed by one or more alphanumeric characters.
• control algorithms comprises at least one specific method for analyzing the areas of interest in which the characteristic relief is formed by streaks.
• the collection of control algorithms comprises at least one specific method for analyzing the areas of interest in which the characteristic relief is smooth.
• the collection of control algorithms comprises at least one specific method for analyzing the areas of interest in which the characteristic relief is formed by a removable wafer.
• the collection of control algorithms comprises at least one specific method for analyzing the areas in which the characteristic relief comprises a vent.
• the collection of control algorithms comprises at least one specific method for analyzing the areas of interest in which the characteristic relief is formed by the tread.
• control algorithms comprises at least one specific method for analyzing the areas in which the characteristic relief comprises a wear indicator.
• control algorithms comprises at least one specific method for analyzing the areas in which the characteristic relief comprises texture.
• the invention finally comprises a computer program comprising code instructions adapted, when executed on a computer, to control the implementation of the steps of the above method.
• Figure 1 schematically shows the coordinate change operation performed during pretreatment of the image of the surface.
• FIG. 2 schematically represents the operation of flattening the image made during the pretreatment of the image of the surface.
• Figure 3 illustrates an angular registration operation performed during pretreatment of the image of the surface.
• FIG. 4 schematically shows a search and a matching of the characteristic points.
• Figures 5 and 6 illustrate the shape and positioning of a mapping on a sidewall of a tire.
• Fig. 7 is a schematic view of a tread formed from the juxtaposition of elements having basic patterns.
• the image of the tire surface may be a two-dimensional image, similar to a photograph, and representative of the appearance of the surface. This image can be a black and white or color image. It reflects the color or the level of gray as well as the shine.
• the acquisition of the topographic image of the flank surface can be performed according to the conventional stereovision technique, in which it is proposed to use two separate cameras dedicated to acquisition of terrain data.
• the two cameras are positioned to take a shot of the surface to be inspected at different angles of view.
• the three-dimensional coordinates of the point of the surface are then calculated by triangulation, determining, after calibration of the cameras, the angles of views different from this point of the surface seen by the two cameras.
• a scattered point cloud is then obtained in the three-dimensional space and representative of the surface of the flank.
• a device capable of making two shots simultaneously in a single rotation of the tire is described by way of example in the publication WO 2009/077534.
• the raw data of the two-dimensional image and the three-dimensional image which together form the image of the tire surface, and which are obtained by the acquisition system, are then transmitted to a processor containing the coded instructions which, when they are executed by said processor, make it possible to implement the different steps of the method.
• a number of image preprocessing operations are generally performed.
• the development of this pretreatment is done by an operator interacting with the processor, using the digital data from the reference image of the reference tire.
• the operator can use data previously input into the processor memory such as CAD design data. associated with the tire being processed, as well as specific algorithms whose processing steps are not within the scope of the present invention.
• This first work makes it possible to determine a set of main characteristics related to the particular dimension of the tire. These main characteristics are intended to be reused by the processor to perform the pretreatment or execute the control algorithms of each of the envelopes to inspect.
• the pretreatment may comprise a first operation which consists in recalibrating the two-dimensional image, so as to correct the shadow zones generated by the variations of illumination during the shooting, or the deformations related to the optical system, and which are specific to the acquisition system used. These shadows or deformations are reproduced at each shot.
• the operator will then determine the compensations to be made and record these compensations in the main characteristics of the dimension, so as to apply the same compensations to all the tires to inspect the same dimension as the reference tire.
• the three-dimensional image is also reworked.
• the three-dimensional coordinates x, y, z of surfaces is analyzed to be expressed in an orthonormal coordinate system OX, OY, OZ in which the axis OZ is substantially coincident with the axis of rotation of the tire. It is then easy to convert the x, y, z coordinates by projection in the plane OXY and to change the Cartesian coordinates x, y in the plane OXY into polar coordinates of type p, ⁇ , as illustrated in FIG.
• the average profile of the curvature of the surface in a radial plane is determined by projecting all the points in the plane formed by the axes OZ and OX ', as illustrated in FIG. 2, which corresponds to a projection in a radial plane.
• the shape of the average radial profile will be given by the shape of the point cloud in this radial plane, from which we can extract a mean curve by averaging the values in a direction OZ.
• the surface obtained by deploying again this average radial profile corresponds substantially to the surface of the tire on which there would be no marking in relief.
• Flattening can also be done by following the profile of the surface according to a determined pattern, for example a line in which the value of p is constant, and by detecting the localized variations of the profile, significant of the embossed markings present on said area. The juxtaposition of these lines also gives a flat surface on which only the relief elements appear.
• this average profile is recorded in the list of main characteristics of the dimension, in order to be reused for the flattening of the tires to be inspected.
• the operator At the next stage of his work, the operator, always interacting with the processor, draws up a map of the surface and determines the limits of the different areas of interest. It then assigns to each of these areas of interest the specific registration and control algorithms best suited to the specific characteristics of these areas, as illustrated in Figure 5. This determination can be automated, for example if the data of CAD design of the mold used to bake the tire, are available.
• the resetting strategy consists in using data of increasing complexity based on information initially derived from a one-dimensional signature of the image, then the three-dimensional image and, if necessary, the three-dimensional image of the tire surface. This strategy makes it possible to graduate the complexity of the data / algorithms used, in parallel with a reduction in the areas concerned:
• the one-dimensional signature is adapted to treat the whole surface
• the three-dimensional images make it possible to treat areas of very homogeneous or little relief interest
• the three-dimensional images make it possible to register areas of interest comprising important reliefs and varied.
• a simple way of proceeding consists in making a one-dimensional signature of the reference image by summing raw data formed by the value of the gray levels of the pixels of a line of the image. two-dimensional image or preferably of the three-dimensional image of this reference tire. We then obtain a 1D curve.
• a second signature is obtained and, by sliding the two signatures over each other so as to match them, a value of the circumferential offset to be obtained is obtained. angularly matching the image of the tire to be inspected angularly with the image of the reference tire.
• a similar operation can be performed by summing the values of the gray levels, raw or filtered, of a column when it is necessary to perform a registration in the radial direction.
• the radial and circumferential signatures 1D, raw or filtered, of the reference tire are then part of the main characteristics.
• this first one-dimensional registration which is generally carried out during the pretreatment phase, makes it possible to project the map on the surface of the tire to be inspected to define the corresponding zones, and to apply to these zones the algorithm of registration and control predefined by the operator and corresponding best to the specificities of the relief of the zone.
• Another more complex registration operation this time using the two-dimensional image of the tire surface, consists in determining in the image of reference of the reference tire, a few easily detectable characteristic points so as to be able to match them with the same reference points of the surface to be inspected of the tire to be inspected as illustrated in FIG. 4.
• the characteristic points may be replaced by the characteristic relief contours, forming characteristic thumbnails, and which are portions of the two-dimensional image of the area of interest considered.
• This mapping operation can be used for the registration of the entire surface but preferably applies within a particular area of interest which may have been previously determined by a one-dimensional angular registration step, in order to allow the use of simple local transformations of the translations type.
• ABCU circumferentially juxtaposed elements
• the wear indicators present on the image of the tread of the tire to be inspected are identified. Then, the wear indicators are grouped by subsets corresponding to the basic pattern comprising wear indicators and the characteristic point of each of these subassemblies is determined. A sequence of distances is then determined by calculating the distances between the characteristic points of each of the subsets of wear indicators identified on the surface of the tread to be inspected, and this sequence of distances is made to coincide with the known sequence. distances between the characteristic points of the basic patterns. Finally, it is projected on the surface to be inspected the shape of the boundary between the elements according to the known sequence of positioning of said elements.
• the sequence of distances between the characteristic points of the basic motifs of the sculpture is one of the main characteristics.
• a registration algorithm of this type is described by way of example in the publication WO 2012 052300.
• the control algorithm of the tread can then be performed by comparing the three-dimensional surface of the three-dimensional image of the tread surface of the tire to be inspected with the image of the three-dimensional surface of the tread of the reference tire.
• the zones of interest Zi Z12, Zi 3 , Zi 4 and Zi 5 of the figure are zones of interest arranged on the sidewall of the tire and containing reliefs similar to drawings for which specific registration and control algorithms have been developed.
• the purpose of these registration and control algorithms is to match, in the most accurate way possible, the three-dimensional surface of the envelope to be inspected and the three-dimensional surface of the reference envelope, so as to be able to determine, by difference , molding anomalies. It is therefore necessary to carry out a controlled deformation of the surface to be inspected to take account of very slight changes in position generated by the release of the stresses at the moment of demolding, as has already been mentioned.
• Another way to proceed is to search, iteratively, for an affine transformation function comprising a homothety whose ratio has an absolute value other than 1.
• This transformation function is applied to the characteristic points of the reference surface, so that the value representing the sum of the distances between each of the characteristic points of the reference surface, transformed using the transformation function, and the points of the surface to be inspected that are matched to them, is minimal.
• the characteristic points chosen by the operator to perform these precise readjustments are integrated in the main characteristics of the reference tire, and stored in the memory of the processor.
• Zones of interest Z 2 , Z 3 , Z 4 are particular zones containing a vent.
• the vents also called teats, are caused by venting devices placed in the mold to promote the flow of occluded gases at the time of molding.
• Each mold has a venting system of its own.
• vents cause shadow areas, and therefore measurement anomalies around the surface where they are implanted.
• a specific algorithm allows to reconstruct the surface, usually smooth, around the foot of the vent.
• the algorithm searches the areas of the surface comprising pixels whose gray level value is below a given threshold, and determines the boundaries of a bounding box. Within the bounding box and for a given line secant of the zone containing pixels whose gray level value is less than a given threshold, each of said pixels is assigned a gray level value equal to the average value of gray level of a set formed by the pixels of a reference segment belonging to said line, and placed close to the zone considered.
• a reference segment is arranged on the side of the main axis of said zone corresponding to the angular sector forming a positive angle with the direction of the shadow area, and assigning the average gray level value of the pixels of the reference segment to all the pixels of the line containing said reference segment and lying between the middle of the reference segment and the intersection of said pixel line with the contour of the foot of the vent.
• the areas of interest Z 6 , Z 7 , Z 8 , Z 9 , Z 0 , Zn are zones containing one or more alphanumeric characters.
• OCR optical character recognition
• zones such as the zone of interest Z 6
• the markings they comprise are formed by removable plates inserted in the mold. These pads are changed daily or weekly, and may vary in alignment with the mold surface. This results in localized variations of relief subject to special tolerancing.
• the undulations of the surface are not precisely described.
• the specific algorithms allowing the analysis of the conformity of these zones of striations are based on the use of the tools of the morphological analysis of the two-dimensional or three-dimensional image of the surface of the tire. These algorithms provide for determining at least one dilation of a basic representation comprising a streak zone so as to obtain an expanded representation, and to determine at least one erosion of the basic representation so as to obtain an eroded representation, and to determine a difference between the dilated representation and the eroded representation so as to obtain a difference representation showing the anomalies.
• These dilations and erosions use structuring elements whose size and orientation must be adjusted according to the size and orientation of the streaks.
• the operator therefore adapts the parameters of the structuring elements to the ridges of each of the striation zones of the reference tire and records these main characteristics in the memory of the processor for each of the zones considered.
• Zone of interest Z19 is a smooth zone in which no relief is supposed to be present.
• the anomalies present on these surfaces may be localized molding defects or hollow variations caused by localized deformations of the carcass reinforcement ply, or else detectable spots because of the highlighting they cause.
• control algorithms particularly adapted to these surfaces have been developed to control compliance.
• a first specific control algorithm is described for example in the publication WO2012 / 156262. This control algorithm is based on the analysis of the three-dimensional gray-scale image of the surface.
• steps are taken in which, using linear structuring elements of successively increasing sizes and oriented in the circumferential direction, a series of morphological openings of the image of the surface of the surface is made. pneumatic. Subtracting from the value of the image obtained after morphological opening with a given structuring element, the value of the image obtained after morphological opening with the structuring element of immediately smaller size so as to obtain a succession of flattened images by difference, by initializing the procedure by subtracting the image obtained using the structuring element of smaller size. Finally, the images flattened by difference are thresholded to obtain binary images, and the set-up union of the values of each of the binary images is done to obtain a final binary image, in which only the marking patterns appear in relief. .
• the disjoint patterns are identified in the binary image of the surface, and the image of the bulge of said surface, in which the relief markings are made, is determined by difference with the starting image of the tire surface. have been removed and where only the localized deformations of the surface mentioned above remain.
• this type of algorithm can also be used for the determination of the average radial profile for performing the flattening during the preprocessing phase.
• Another control algorithm is also based on the analysis of the texture of the three-dimensional image of the surface and also uses the tools of morphological analysis.
• a specific algorithm of this type is described for example in the publication WO 2013/045593. According to this publication, via factorial spaces in which the data are formed by morphological filters and the variables are formed by the multivariate images of the tire surface with previously identified anomalies, the most suitable series of morphological filters are determined. to highlight said anomalies. These series of filters are then applied to the image of the tire to be inspected and, using a classifier, the presence of these anomalies is detected.
• an image of the orientation of the elevation gradients of the surface in which is assigned to each point of the image a a gray level value proportional to the angle formed with a direction given by the projection in the plane of the image of a non-zero standard vector substantially corresponding, at this point, to the gradient vector tangent to the surface and oriented in the direction of the steepest slope.
• a filtered image of the orientations is then determined by transforming the image of the orientation of the elevation gradients with the aid of a digital filter able to select the areas comprising structures similar to a reference image of the orientation. elevation gradients of steeper slope of a blister.
• This reference image of the orientation of steepest elevation gradients of a blister that is in the form of a circumferential grayscale gradient is then considered a major feature of the dimension, and stored in the processor memory.
• the operator can then choose to have the processor execute this specific algorithm in all the zones or only on certain specific zones that are more sensitive to the appearance of these air inclusions.
• first level algorithms Of all the specific control algorithms made available to the operator and stored in the processor memory, some may be considered first level algorithms. These particular algorithms are not very greedy in terms of computation time and make it possible to quickly identify if the zone in which they are implemented includes an element that can give rise to doubt about the presence of a defect. It is then wise to implement these so-called first-level algorithms and to decide, depending on the result obtained, whether or not it is appropriate to launch one of the specific algorithms that is more resource-intensive.
• contour of the reference image present in the area of interest of the reference tire may also prove to be a valuable aid .
• the contour of the images can be obtained, in known manner, from the two-dimensional image or the three-dimensional image of the surface.
• the shape of the contours, the geometric moments, the geometrical measurements, the characteristic points of the reliefs of the reference tire, are considered as main characteristics and recorded in the memory of the processor.
• the set of specific or reduced registration and control algorithms form a collection of algorithms which is also stored in the memory of the processor.
• an enriched base is obtained in which each zone of interest of the tire is associated with one or more algorithms of the tire. calibration and control characteristics for which the choice of the main characteristics is determined.
• the processor then automatically implements the specific registration and control algorithms defined for each zone using the main characteristics assigned to these algorithms and previously stored as explained above.
• the set formed by the image of the reference surface of the reference tire, the main characteristics, the mapping with the optimal principle of registration associated with it, and the choice of registration and control algorithms to implement for each of the areas of interest, is the basic model for optimally performing control of a tire to be inspected from the same mold as the reference tire.
• This basic model must be adapted every time an intervention such as the drilling of a vent or the change of a plate, is performed on the mold.
• this basic model can easily be adapted to serve as a basic model for a reference tire from a different mold of the same dimension.

Landscapes

• Engineering & Computer Science (AREA)
• Computer Vision & Pattern Recognition (AREA)
• Physics & Mathematics (AREA)
• General Physics & Mathematics (AREA)
• Theoretical Computer Science (AREA)
• Quality & Reliability (AREA)
• Length Measuring Devices By Optical Means (AREA)

Applications Claiming Priority (2)

Publications (1)

Family

ID=54707895

Family Applications (1)

Country Status (5)

Families Citing this family (17)

Family Cites Families (17)

• 2015
  • 2015-07-27 FR FR1557142A patent/FR3039684B1/fr not_active Expired - Fee Related
• 2016
  • 2016-07-26 CN CN201680043971.1A patent/CN107851314A/zh active Pending
  • 2016-07-26 US US15/741,378 patent/US10346971B2/en active Active
  • 2016-07-26 EP EP16760124.4A patent/EP3329460A1/de not_active Withdrawn
  • 2016-07-26 WO PCT/FR2016/051941 patent/WO2017017371A1/fr active Application Filing

Also Published As

Similar Documents

Legal Events