MXPA99008907A - Method for examining an unwinding strip surface by pre-classification of detected surface defect - Google Patents

Abstract

The invention concerns a method for examining an unwinding strip surface consisting in the following steps:forming with imaging means at least a digital image of at least one strip (10) surface;filtering said at least one digital image to detect surface defects and processing said at least one filtered digital image to identify the type of surface defect corresponding to each detected defect. Previous to the digital image processing, characterising the defects by determining for each of them the value of predetermined parameters characteristic of surface defects and carrying out said defect pre-classification, based on predetermined values of said parameters, according to a set of predefined classes, said processing step being carried out on each class.

Description

INSPECTION PROCEDURE OF THE SURFACE OF A DEVELOPABLE BAND BY THE PRIOR CLASSIFICATION OF A DEFECT SURFACE DETECTED Field of the Invention The present invention relates to a method of inspecting the surface of an uncoiling strip, in particular of a laminated sheet that is unrolled at a high speed, as well as to an installation that allows to put into operation such an inspection procedure.

Background of the Invention In surface inspection systems of the conventional type, in which the surfaces of an uncoiling web are controlled automatically, the inspection of the surface is carried out by forming at least one numerical image of at least one of the faces of the web. The band, consisting of a set of successive lines of image elements each assigned to a numerical value, filters said at least one numerical image for the detection of irregularities of the surface Ref.31394 by the detection of relative variations of said values. numerical, and the irregularities of the surface are treated for the identification of the type of surface defect that correspond to each irregularity detected. According to this inspection technique, the treatment of surface irregularities is generally carried out by identifying the defect, among a set of defects that may appear on the inspected surface, which corresponds to each irregularity. Thus, the analysis of the irregularities is carried out identically, whatever the nature of the irregularities detected. Accordingly, this type of surface inspection system has relatively low processing speeds, in particular because of the nature of the treatment stage, which requires a large number of relatively long and complex calculations. The object of the invention is to remedy these drawbacks and "to provide a surface inspection method that allows a prior selection of detected surface irregularities, whatever the nature of the inspected surface.

It thus has as its object a method for inspecting the surface of an unrollable strip of the aforementioned type, comprising the steps consisting of: forming, with the aid of means of taking views, at least one numerical image of at least one of the faces of the band, consisting of a set of successive lines of the elements of the images, each assigned to a numerical value; - filtering said at least one numerical image for the detection of surface irregularities, by detecting the relative variations to said numerical values; and treating said at least one filtered numerical image for the identification of the type of surface defect corresponding to each irregularity detected; characterized in that, prior to the treatment step of said at least one numerical image, a general characterization of the irregularities is carried out determining for each of them the value of the predetermined parameters characteristic of the surface defects and a previous classification is made of said irregularities, based on the determined values of said parameters, according to a set of predefined classes, said treatment step is carried out on each class. The irregularities detected are previously classified according to a set of classes on each of which the treatment of the image is made, it is known that the latter is accelerated considerably with the help of this stage of previous roughing. In addition, this previous classification allows to reduce the rate of recognition errors and thus improve the quality of the identification. The method according to the invention can also have one or more of the following characteristics: each predetermined parameter that appears or resembles a general reference point axis in a space whose dimensions correspond to said parameters, prior to said previous classification, each irregularity is represented in said space by a point whose coordinates are the values of said parameters, and said previous classification is made by the identification of the region to which each point belongs, and by the assignment of the irregularity corresponding to the corresponding class of said region; a second mode of characterization of the irregularities is determined for each predefined class whose number of characteristic parameters is lower than the number of characteristic parameters of the general characterization and, after the previous classification stage, the value of the detected values is determined for each detected irregularity. characteristic parameters of the second mode of specific characterization of said class to which the irregularity belongs, from the values of the characteristic parameters of the general characterization; a simplified mark of representation of the irregularities is determined for each region whose number of axes is lower than the number of axes of general reference point and, after the previous classification stage, a change stage is carried out for each irregularity represented. mark of said general reference point towards said simplified mark specific to the region to which the irregularity belongs; the stage of treatment of the irregularities takes a first stage of identification of the defect that corresponds to each irregularity, between a set of types of specific defects of the class to which this irregularity belongs and a second stage of classification of said defect identified with object of the confirmation and precision of the classification resulting from said first stage of classification; - the procedure has a stage of qualifying the types of defects according to a first type of defects identified in a certain and / or precise manner and a second type of defects identified in an uncertain and / or inaccurate manner and in which said second stage of Classification is carried out only on the defects of the type classified as uncertain and / or imprecise; the procedure also includes a stage of regrouping the identified defects using a set of predefined criteria, especially geometric and / or topographical criteria; the method further comprises the steps of counting the number of identified defects of the same type per unit length, and comparing said number of defects of each type with a predetermined threshold value representative of the minimum number of defects from which said defects they are capable of presenting a periodic characteristic, in order to detect periodic defects; subsequently to the step of determining the value of said parameters, and prior to said previous classification step, a specific classification of the irregularities is carried out according to a set of elementary classes, and the population of said elementary classes is analyzed in view of the detection of periodic faults, subsequent to the filtering step, in response to a detection of an image element of an irregularity, a storage area of the lines of successive image elements supplied by the view-taking means is delimited in a memory. and having at least one image element corresponding to at least one irregularity, each storage area is segmented into suspect areas each having at least one surface irregularity, the suspect areas are matched with the successive storage areas and that correspond to the same irregularity and compares the total number of lines d e the image elements of the suspicious areas matched with a detection threshold of the large length defect, and, in the event that it exceeds said threshold, the processing step of said at least one numerical image filtered only on one of said suspicious zones appeared or paired, the result of the treatment is attributed to the other suspicious zones that appeared. Another object of the invention is an installation for inspecting the surface of an unrollable strip for putting into operation a method such as that defined above., characterized in that it carries means for taking views of at least one of the faces of the band. A memory for storing at least one image of the band in the form of lines and columns of elements of the image each associated to a numerical value, a filtering circuit of said at least one numerical image for the detection of the irregularities of the surface of the band, for the detection of the relative variations of said numerical values, and a signal processing unit connected to the filtering circuit and carrying means of calculating the values of the characteristic parameters of the defects of the surfaces, these means for classifying irregularities detected according to a set of predefined classes from the values of said parameters and the means of identifying each irregularity among a set of types of defects susceptible of 'correspond to said irregularity.

Other features and advantages will be apparent from the following description, given only by way of example and made with reference to the accompanying drawings on which: - Figure 1 is a general outline of an embodiment of a surface inspection facility; according to the invention; Figure 2 represents a part of an image supplied by the view-taking means of the installation of Figure 1 and stored in the memory; Figures 3a and 3e depict different images of the surface of a band during a step of dividing the images; Figure 4 is a flow chart illustrating the general operation of the installation of Figure 1; Figure 5 is a flowchart showing the different stages of treatments of the filtered numerical images; Figures 6a and ßb are diagrams showing, according to the length and width of the surface defects, the different kinds of defects, respectively for a semi-finished product (DKP) and for a galvanized product; and Figure 7 is a flow chart showing the stages of a program for analyzing the detected surface defects.

Detailed description of the invention The installation shown in FIG. 1 is intended for detecting the defect of the surface of a strip 10 in the unrolled course at high speed, for example a laminated sheet coming out of a rolling line. The surfaces of the sheet 10 are inspected by means of a sight-taking apparatus 12 which is supplied to a filtering stage 14 of the numerical images of the surface of the strip. In the illustrated embodiment, the installation carries a single viewing device 12 fixed on one of the surfaces of the strip, but of course, the installation can be equipped with two viewing apparatuses adapted to form images of each surface of the strip. the strip 10. The viewing apparatus 12 can be constituted by any type of apparatus suitable for the contemplated use, whose field width is substantially equal to the width of the inspection zone of the strip 10, such an inspection zone can be constituted by the entire width of the band. The same can also be constituted by one or several matrix cameras that supply images of defined length, considering the unrolled senses of the band, either by a camera or several linear cameras that supply images of infinite length. In the case where a matrix or linear camera is not sufficient to cover the entire width of the band's inspection area, several cameras are used distributed over the width of the band. With reference to Figure 2, the view-taking apparatus 12 forms the lines i of M elements of the image Ii 3, or pixels, addressable, for a marking of the pixels according to the length of the band 10, by the line n ° iy, according to the width, by the column n ° j of elements of the image, each element of the image is associated with a numerical value representative of a level of gray. The lines of elements of the image are stored in a memory 18 of the filtering state under the command of a management circuit 20. According to a first example, the view-taking apparatus is constituted by a linear camera that supplies the memory 18, 10,000 lines of 2048 pixels per second, these lines are stored in the memory in successive directions. According to another example, the view making apparatus is constituted by two matrix chambers distributed over the width of the band to cover the width of the band and adapted to take 10 images / s. Each image supplied by a single camera is constituted by 1024 lines of 1024 pixels, supplied to the memory 18. Thus, the system of taking of views permanently dispatches lines of elements of the image, each element of the image is associated with a numerical value which represents a level of gray. It is known that it is cadenced by the line if it is a linear camera and cadence by a group of lines if it is a matrix camera. Referring again to Figure 1, it is observed that the filtering stage 14 also carries a filtering circuit 21 constituted by an image processing operator that ensures the detection of the relative variations of the numerical values of the elements of the image or the pixels for the detection of the irregularities of the surface. Preferably, the filtering circuit is constituted by a contour detection circuit, for example a detector of the "Prewitt" type, which detects the variations of levels of the gray between the elements of the image located in the proximity of one another, which allows to detect zones of the sheet 10 that present irregularities of the surface. As shown in Figure 1, the output of the filtering circuit 14 is connected to a signal processing unit 22 that carries a first step 24 of segmentation of the numerical images in areas of image elements that delimit each, a irregularity of the surface detected by the filtering step 14, and a second stage 26 of processing the signals, constituted by a calculation circuit 28 associated with a corresponding memory 30, in which the processing algorithms for the recognition and processing are stored. the identification of the defect of the surface, for each zone that presents an irregularity of the surface. The installation shown on FIG. 1 is also provided with a display device 32 connected to the output of the treatment unit 22 whose input is connected to an output of the calculation circuit 28 and which allows the visualization of the detected surface defects , associated with the information regarding the type of defect and the parameters representative of the severity of these defects, as will be described in detail below. The description of the operation of the installation to be described frequently is going to be made with reference to Figures 2 to 7. On Figure 3a, a part of the sheet 10 that has a set of irregularities of the surface is represented. as 34. The field of the view-taking apparatus 12 preferably covers the entire width of the band 10. With reference to Figure 4, in the course of a first step 36, the view-taking apparatus takes the successive lines of the image elements of the surface of the strip 10, these picture elements are stored, in the memory 18, associated with a value of the gray level. In the course of this first step 36 of taking views, the management circuit 20 performs, if necessary, a fusion of the images supplied by the view-taking apparatus 12, regrouping the successive pixels, on the one hand in the sense of the width of the band 10 in the case where several cameras are used to cover the entire width of the inspection area, to obtain in the memory 18 an image whose width corresponds to that of the inspected area, and on the other hand, in the sense of the length of the band 10, in the case where the view taking apparatus 12 uses one or more matrix cameras, merging the groups of pixel lines supplied in turn. The image, stored in the memory 18, later called "gross image", is constituted by a set of image elements I ±, 3, i that designates the direction of the line in the memory, which varies from 1 to N, and j that designates the number of an element of the image of each line and that varies from 1 to M, M is equal for example to 2048, each element of the image is associated with a numerical value of the gray value. It is going to be noted that the N value depends on the capacity of the memory. This capacity must be adapted for the memorization of a sufficient number of lines with respect to the subsequent treatment to be performed. For example, for the storage of an image that corresponds to a length of 15 m of sheet metal with a number of lines of elements of the image equal to 1024 / m, N is preferably equal to 15360 lines. When the capacity of the memory is saturated, the lines that arrive successively are memorized in place of the lines of pixels older and previously memorized and treated normally.

When the memory 18 is saturated and the oldest pixel lines have not been addressed, a saturation alarm is issued to indicate that a zone of the band will not be inspected. In this case, the non-inspected area is marked on the band, for the identification and storage in a file of successive non-memorized lines, for the purpose, for example, of a statistical analysis of the non-inspected portions of the band. However, taking into account the average speed of unwinding of the band and the average density of irregularities of the surface to be identified for a given band type, an average calculation power required corresponding to an average processing speed beyond can be determined. of which there is no risk, in practice, of erasing untreated lines. Preferably, the treatment modules are thus dimensioned so that the instantaneous processing speed is greater than this average speed. Thus, in addition to its role of fusion of the images, the memory 18 ensures a plug paper that allows to withstand the variations, and in particular the increases, of the treatment load due to an increase in the density of the irregularities of the surface .

In the course of the following step 38, a binary image representing the contour lines of the irregularities of the surface is associated with each image stored in the memory 18. To do this, in the course of this step, the successive lines "of the elements of the raw image are filtered by the filtering circuit 21, constituted as mentioned above for example by a bidirectional Pretit filter of the classical type, which its function is to detect the variations of the gray levels of the elements of the gross image which translate the existence of irregularities of the surface in view of the determination of its inscribed contour, in the associated binary image In the embodiment described, it is considered that the filter used is a Prewitt filter, but of course, any other type of filter adapted for the contemplated use can also be used.The Prewitt filter ensures a reference point of the contour position of the irregularity of the surface by detecting, on each line of a raw image, the elements of the image that may belong to a contour line of the irregularity, these elements of the images are designated later "suspicious pixels". The filter used here allocates a numerical value "1" to each element of the binary image associated with each pixel suspect of the raw image supplied by the view-taking apparatus 12, the other pixels of the binary image are maintained at 0 This filtering step 38 thus makes it possible to form in the memory 18 a binary image consisting of a set of binary image elements Bi (), each of which is assigned a binary value equal to 1 for a pixel belonging to a contour of an irregularity and equal to a null value for a pixel that does not belong to a contour of a surface irregularity In the course of the next step 40, the binary image stored in the memory 18 is treated with the help of a classic connectivity operator that applies a mask to this image to force * the numerical value "1" of the pixels of the binary image that have a null value and located between two relatively suspicious image elements e, in order to obtain and define continuous lines for each detected contour.

After having undergone this treatment, the raw and binary images are cleaned to eliminate the defects or spots delimited by a contour whose surface is lower than a certain threshold, for example 3x3 pixels. A binary image is then obtained, superimposed on the raw image supplied by the view-taking apparatus 12, and which shows the contours delimiting the irregularities of the surface detected in the raw image. The binary image and the raw image are then ready for processing. In the course of the next stage 42, the management circuit 20 successively analyzes each line of the memorized binary image, for the deton of the binary elements of value "1", that is to say susp. Since a suspicious pixel is detd, the management circuit 20 marks the corresponding line number, opens a predetermined storage area in the form of a window in memory 18 (step 44) from this line number and keeps this advantage open while the management circuit det the suspicious pixels in the following lines. This window, designated immediately by the "suspicious window", thus contains the suspicious pixels, that is to say, susceptible of belonging to an irregularity of the surface. The management circuit 20 closes the suspicious window at the moment when no suspicious pixel is detd anymore in a predetermined number of successive lines of the binary image, registering the number of the last line in which a suspicious pixel has been marked. The suspicious window thus defined in memory 18 represents a segment of the raw image, associated with a segment of the corresponding binary image, and containing at least one irregularity of the surface to be identified and recognized. In particular, the window, opened at the time of step 44, is kept open while the number of the last successive lines of elements of the image stored in said window that do not contain the susppixel, does not exceed a predetermined threshold number. of the successive binary lines, this threshold is at least equal to 1. Thus, in the course of the following step 45, the number of the successive lines of the elements of the image that do not contain the susppixel is compared to this number of threshold and, in the case of equality, the suspicious window is closed (stage 46). On the other hand, at the time of step 47, the number of lines recorded in the open window is compared with a predetermined threshold called "detection of the large-length window" or of "detection of a longitudinal defect". This predetermined threshold corresponds to the predetermined maximum capacity of the storage areas in the memory 18. If the number of registered lines is higher than this threshold, the window is closed (step 48) and it is decided, at the time of stage 50 next, that the window is a window called "suspicious of large length", which contains an irregularity of the surface whose number of lines of elements of the image is greater than the detection threshold of the longitudinal defect. It will also be noted that, in the described embodiment, the suspicious windows are opened successively. The surface inspection procedure is followed by the phases of division of the suspicious windows stored in the memory 18 in areas called "suspicious zones" each presenting an irregularity of the surface, using either the component corresponding to the image gross, or the component that corresponds to the binary image. To do this, step 24 performs, in the course of steps 58 to 64 below, a calculation, with the help of appropriate means, for example of software means, of the accumulation profile of the numerical values or of the binary values respectively for each raw image or each binary image, on the one hand in the longitudinal direction and, on the other hand, in the direction of the width, on the projection of the numerical values or the binary values according to two perpendicular axes and on the positioning of the thresholds of the profiles in such a way that the suspicious zones that each one incorporates are delimited, an irregularity of the surface. Of course, the calculation of this profile can be carried out from the associated numerical values of the gross image or from the binary values of the image stored after the treatment, in what follows of the description, it will be considered that the treatment of the image is made from the binary image. This calculation operation is initiated by a phase of segmentation of each suspicious window of the suspicious band that encompasses the irregularities, each band is then segmented into one or several suspicious areas. First, in the course of step 58, step 24 calculates, with the help of a 24-a calculation circuit (Figure 1), the sum of the binary values of each line of the suspicious window to obtain, on the columns M, a first transverse profile, in the direction of the width of the band. The curve shown on Figure 3b is thus obtained. In the course of the following step 60, this profile is presented at the entrance of a focus or framing circuit 24-b, to be framed so as not to separate the elements of the image of an irregularity located in proximity to each other. The framing circuit 24-b can be constituted by any type of appropriate filter, such as a filter RIF of finite impressive response, or RII of infinite impresional response, but is preferably constituted by a filter of the sliding window type that allows to supply a framed profile r (x) whose values are determined according to the following relationship: K r (x) = SF (xi) x Q (i) (1) i = -K in which K designates the width of the sliding window, F (x-i) designates the value of the column (x-i) of the profile to be framed, Q designates the coefficient of the sliding window filter, chosen for example equal to 1, and X designates the number of the column of the profile framed. The profile thus framed is next provided with a threshold with the aid of a threshold providing circuit 24-c, in the course of the next step 62, by comparison with a threshold value for detecting the irregularities. The framed profile is thus obtained and provided with a threshold represented on Figure 3c that delimits the suspicious bands, represented with the help of dotted lines on Figure 3a, which includes each one, one or more defects of the surface. As mentioned above, the suspicious bands thus defined are segmented immediately into the suspicious zones that each presents, an irregularity of the surface. To do this, in the course of step 64 below, steps 58, 60 and 62 are performed again and applied independently to each line of elements of the image of each suspect band, so as to obtain an accumulation profile of the binary values in the longitudinal direction, as shown on Figure 3d. This longitudinal profile is framed immediately and provided with a threshold, as previously, to obtain the image represented on Figure 3e in which the suspicious areas are defined, such as 66, which delimit each one, a defect of the detected surface, each Defect can of course lead to several objects or segments of irregularity. Each suspicious zone thus defined contains a segment of the gross image and the segment of the corresponding binary image. Preferably, the suspect areas 66 thus delimited are further presented at the input of a second calculation circuit 24-d, connected to the output of the positioning circuit of the threshold 24-c, by means of which the defects of small dimensions are eliminated . To do this, in the course of step 68 below each suspicious area of the binary image is treated independently with the aid of a classical tagging algorithm in order to delimit the constituent objects of a surface defect, each object it is defined by a set of elements of the image suspected of contact with each other.

The surface of each object is calculated immediately, as well as the average surface of the objects that belong to the same suspicious area. Objects of small dimensions are removed from the treatment. To do this, it is decided to eliminate the objects whose individual surface is less than a predetermined percentage of the calculated average surface. Thus, at the output of the calculation circuit 24-d, the suspicious areas each containing a defect are obtained, of which the small objects have been eliminated. These suspicious areas thus cleaned are then stored in the memory 30 of the calculation circuit 28 for the purpose of being treated, as will be described in detail below with reference to the Figure 5. It should be noted that the 24-a, frame 24-b, and 24-c threshold and 24-d calculation circuits are classic type circuits. They will not be described in detail in the following. In the case where a suspicious window has been classified as a suspicious window of large length in the course of the preceding stage 50, the stage of treatment of the images is preceded by a phase of elimination of the treatment of certain suspicious areas, which allows reduce the load of the calculation circuit 28. For this purpose, since it is detected (stages 47, 48 and 50) a suspicious window of large length and that is divided into suspicious areas as described above, is marked in the course of the step 70 following at least one suspicious area of this window whose lower line of elements of the image belong to that of said window. This suspicious area thus marked is then classified as "suspicious zone divided or cut down". The suspicious window following a suspicious window of large length is described as a "suspicious window of prolongation". It is known that a suspicious window of extension can be equally large. After the division, as described above, of a suspicious window of prolongation in the suspicious areas, the at least one suspicious area of this window is marked whose upper line of elements of the image belongs to that of the window, this suspicious zone it is then classified as a "suspicious zone divided or cut up" or "suspect area of prolongation" (stage 71).

The suspicious areas "cut out at the bottom" of the large-length window and those "trimmed at the top" are matched to the suspicious window of prolongation (stage 72). In the course of the following step 73, it is determined whether the suspect window of extension is itself of large length. If this is the case, at least one suspicious area of this window is marked whose lower line of elements of the image belong to that of said window, this suspicious area is then qualified as "suspicious area cut down from below" and it performs the same treatment of the recomposition of this suspicious zone with the suspicious zones "cut off from the top" of the next window, extension call (stage 74). As the appearance or association of the suspicious areas cut from one window to the next, the length of each defect is updated. In the course of step 75 below, the treatment unit 22 compares the length of each defect with the length of a large window length, i.e. with the detection threshold of the long defect mentioned above. Since this length exceeds that of a window of large length, the defect is qualified as being a long defect (step 76) and a "long defect group" defined by an area of the memory of the treatment stage is opened in which place all the successive cut and associated suspicious zones that constitute in effect a threshold and the same defect called "long defect". They are eliminated immediately after the treatment of the image, all the suspected areas of prolongation that belong to the groups of the "long defect", so, in each group of "long defect", the treatment of the image is done only on the first suspicious zone ("cut down") ) and, to simplify the treatment, the result of this treatment is assigned to all suspected areas of prolongation of the same "long defect" group. As the appearance or association of the suspicious areas cut from one window to the next, updating the length of each defect associated with the suspicious areas that correspond from one window to the next, can be seen in the course of stage 75 that this defect is not a long defect. The segmentation of one of such defects can not take place on more than two successive windows, but it will be classified as a long defect.

In this case, a storage area is opened in the memory 30 in the form of a suspicious area called "recomposition" in which the two suspicious areas cut from the same defect are placed, conveniently joined and centered, the size of said window it is adapted to frame such a defect as in the case of suspicious non-trimmed zones (step 77). Suspicious areas of recomposition are treated immediately like all other suspicious areas. The segmentation phase of the gross and binary images in the suspect areas to be treated is now finished, then the treatment of each suspicious zone delimited in stages 58 to 68 is carried out, with the exception of the suspected areas of prolongation of the group of "long defect". The treatment of each suspect zone will now be described with reference to Figures 5 and 7. This treatment begins with a step 78 for calculating the defect identification parameters, generally qualified as the stage for extracting the parameters. In a manner known per se, the nature of the parameters capable of characterizing the defects or irregularities of the surface of the band to be inspected, and necessary to recognize and identify them in a precise and reliable manner, is determined. The method of calculating these parameters is also determined, especially as a function of the values of the elements of the image, of the gross or binary image of a suspicious area containing said defect or said irregularity of the surface. In a classical manner, among these parameters one generally finds the length, the width and the surface of a surface irregularity in a suspicious area, the average intensity of the gray levels of the elements of the gross image inside the defect , the type of deviation of these levels of gray ... The number of parameters necessary for an accurate and reliable recognition, designated immediately as P, can be very high and reaches for example 65. The nature and mode of calculation of the parameters of the defects, are now defined for a type of band to be inspected, proceeding to the calculation of the P parameters for each suspicious zone.

Each suspicious zone or irregularity can thus be represented by a point in a P-dimensional space. This high number P of parameters is a disadvantage, with respect to time and means of treatment of recognition of suspicious areas. To avoid, or at least limit, this disadvantage, a roughing step 80 is carried out which makes it possible to considerably simplify the treatment of each suspicious area by classifying the irregularities according to a set of roughing classes. This roughing step, which constitutes a prior classification of irregularities, according to a set of predefined classes, allows us to divide the general problem of analysis of irregularities into a set of simpler problems to deal with. In particular, within each roughing class, a set of elementary classes or families of defects is defined, whose number is limited. In order to be able to operate the roughing step, it is necessary to have foreseen a preliminary phase for defining the roughing classes and, possibly, from its associated simplified mark, generally before the operation of the method according to the invention.

This preliminary phase is specific to a type of band that is going to be inspected. As an example of the previous phase that ends in the definition of the roughing classes, we proceed by learning in the following way. A surface inspection is carried out, as described above, up to this stage of the procedure, of a sufficient number of samples of the same type of band to have a sufficiently large population and representative of the suspicious areas, each irregularity of which is represented by a point in the P-dimensional space mentioned above. According to the method known elsewhere on the factorial correspondence analysis, it is marked how these points are grouped in clouds in this space. It is considered then that each region of the space that delimits a cloud allows defining a typology of defects, and the defects of the same cloud have elements in common and will be able to be represented eventually in a simplified mark appropriate for this cloud or for this typology . To define the axes of a simplified mark suitable for a typology or for a given cloud, the main axes of inertia of this cloud can be used, whose positions and directions can be calculated in a manner known per se. Thus, all defects of the same class can be represented in the same simplified mark in a space whose dimension is less than P, that is to say that all defects of the same class can be characterized by a reduced number of parameters, lower than P Using traditional mathematical methods, trademark change matrices are established that allow moving from a representation of the defects in a P-dimensional space to a presentation of the same defect in a simplified mark of reduced dimensions. Thus, in this preliminary phase intended to prepare the roughing, typologies or "roughing classes" of the defects have been defined and a simplified mark of representation of the defect, appropriate for each kind of roughing. According to a specific example, these roughing classes can be defined from the length (L) or the width (1) of the irregularities; with reference to Figures 6a and 6b, for example 5 and 6 roughing classes are defined, respectively for a sheet "DKP" and for a galvanized sheet, namely a class of small defects (pt), a class of fine defects and short (fe), a class of fine and long defects (fl), a class of medium and short defects (me), a class of medium and long defects (ml) and a class of wide defects (the); a simplified representation mark is associated to each class. After the step of extracting the parameters, step 80 of pre-sorting or roughing, properly so called, can now be started. To do this, each defect or irregularity of the surface of the suspicious zone is distributed in the different roughing classes previously defined, depending on the value of the P parameters of a defect and the characteristics that define these classes. This previous distribution of the defects in roughing classes makes it possible to considerably simplify the recognition of the defects, making this recognition on each sort of roughing. In a variant, all the defects of the same class are represented in the simplified mark associated to this class, using the matrix of change of the mark of this class, applied to the parameters P. It is then led to a simplified characterization of all the defects, by a reduced number of parameters, which limits the amount of calculations to be made at the time of recognition. The subsequent step 82 of the treatment consists in recognizing and identifying the defects of each sort of roughing. The identification and recognition treatment is specific to each type of roughing and is generally defined prior to the function of the types of defects that are likely to be found in each class. This identification and recognition treatment may consist of a classification based, for example, on the so-called "Coulomb spheres" method. Other known methods can also be used, such as the discriminant analysis methods, the decision tree method or the method that passes through the determination of "K" plus the next neighbor. According to the method of the Coulomb spheres, the typologies of the defect, specific to a given roughing class, are represented by spheres, markable, in their position and size, in the simplified space associated with this class. Each sphere corresponds to a defect type and / or a defect identification name.

Thus, in order to recognize and identify a defect of a given roughing class, at the time of step 83, any sphere belonging to the defect is marked and the identification name associated with this sphere is assigned (step 84). Advantageously, this recognition and identification operation can be carried out very quickly because, the number of spheres and the number of parameters is reduced because in the preceding step of roughing, the classification calculations can be made on a reduced number of criteria . In the particular case where, within a given roughing class, a defect that does not belong to any sphere would be found, the name of the nearest sphere identification is attributed to it. Thus, at the end of step 84 of assigning a name and identifying the defect to each irregularity, all irregularities are identified as corresponding to a particular type of defect. The following stage 86 consists of carrying out a second classification using a second stage of classification of the calculation circuit 28, starting from a reduced number of classes, for the purpose, for example, of forming the result provided for the first stage of classification and of disappearing certain uncertainties that could appear in the identification of certain defects, or finally, for example, to differentiate in a narrower typology of the defects of the same type that could be decided not to differentiate at the level of the first stage of classification, in the absence of sufficient classification operations at this level. In order to be able to put into operation this second stage 86 of classification, it is necessary to have foreseen a preliminary qualification phase of each elementary class. In this preliminary phase, the statistical treatments of validation or non-validation of the classification carried out for the identification of the defects are carried out, using the procedure that is going to be described, so that the elementary classes that contain the most errors of classification of the defect are marked. These elementary classes, of reduced number, which contain the largest number of classification errors, are classified as "elementary classes of uncertain identification", the others, which contain a smaller number of classification errors, are qualified as "elementary classes of uncertain identification". true identification. " The second classification, put into operation in stage 86, is carried out only on the defects or irregularities classified in the elementary classes of uncertain identification. The second classification stage uses, for example, one of the classification methods mentioned above. It is adapted for example to validate or not the belonging of these defects to these classes of uncertain identification. In the case of non-validation, the defect is then considered as not being a defect and is eliminated from the treatment. It can also be adapted to distribute the defects of certain elementary classes of uncertain identification in the classes of precise identification, predefined according to a narrower typology. It is to be noted that this supplementary classification carries a very small number of defect types and can be carried out very quickly. At the exit of these stages 80 to 86, each defect is identified and recognized, that is assigned to an elementary class.

The processing phase of the images is achieved by a stage 88 of data fusion in the course of which certain defects' are regrouped, using previously defined criteria, especially taking into account the geometry and the topology of the defects (for example: distance of the effects between them, identical position above and below the band, proximity of the edge of the band, ...). This fusion phase allows us to remedy certain imperfections that may appear at the moment of recognition of the defects and the resolution of any particular problem of confusion, without returning the reason for the results already confirmed. The decision to regroup the defects is made after the confrontation of the information that comes from the close vicinity of an object to be recognized, from the order of position for example, of other devices for taking views (for example set to the other side of the band), or of the data related to the treatment of the band (nature of the band, point of detection, ...). In particular, it is decided to regroup among them the defects for which an ambiguity about the name subsists, as well as the defects of the same nature.

On the other hand, the defects showing particular proximity relationships, for example, defects situated in the vicinity, on the same face of the strip or on an opposite face, as well as defects located in the same longitudinal alignment or on the other side, are grouped together. cross. Thus, for example, in the case of a galvanized sheet, a defect of the "grained drag" type results in a multitude of surface irregularities located in the vicinity of the edge of the sheet. In this case, the identification of the defect is not totally reliable. In fact, each of these irregularities can be recognized as belonging to a "granular drag", or be recognized individually as a defect of another type, especially an "exfoliation", or a "swelling". In this particular case, the irregularities located in the vicinity of the edge of the sheet are fused and aligned with each other and they are identified as belonging to a defect of the type of "granular drag". In the same way, according to another example, in the course of this fusion step, the defects located in the same position, on the upper and lower faces of the sheet and are given an identical name, are grouped together. In the course of this melting step and as described above, with reference to step 76 of Figure 4, the long defects, cut off at the time of the opening of the suspicious windows, are also grouped, assigning, as mentioned above. , the name of the defect of the suspected area of large width to the defects of the suspected areas of prolongation of the same group. In the course of this fusion step, the population of the elementary class of the defect is also analyzed on a given length of the band, that is to say the number of defects per unit of length that present the same identification. This population is then compared to a predetermined threshold, called the threshold of presumption of the periodic defect. This threshold is determined for the same given length of the band. When the population of an elementary class exceeds this threshold, it is considered that these defects of this class probably represent a periodic character.

To validate this character, a classical method of detecting periodic defects can be used. For example, the histogram of the distance between each defect of this class is plotted and, if this histogram shows a periodicity (fundamental or harmonic), a specific group of "periodic defect" is opened in the memory and the periodic defects of this class are regrouped in this same group. According to a variant, this stage of detection and regrouping of periodic defects can be carried out after the extraction of the parameters but before identification and recognition, even before roughing or pre-sorting. This variant then assumes a specific classification treatment, relatively summary because it is based on the characterization of the defects according to a high number P of parameters and, for the detection of the periodic defects, the population of the defined elementary classes is then analyzed in this specific classification. This variant has the advantage of setting a result that does not depend on the operation of the recognition modules (roughing and sorting downstream).

After having detected, recognized and possibly regrouped the defects corresponding to the detected irregularities, the subsequent phase of the inspection procedure consists of analyzing the defects in order to determine the severity, to allow the determination of the malfunction of the band. This phase will now be described with reference to Figure 7. Previously, before putting the procedure into operation, for each class or each type of defect, depending on different possible intrinsic nocices of the defect type, a set is defined of subclasses, each subclass is associated with a possible intrinsic noxiousness of the defect type. Each subclass can be assigned eventually with an intrinsic gravity coefficient. It is known that each irregularity of the surface is in this state, identified and thus characterized by the characteristic parameters, in particular by a reduced number of parameters. At the time of the first stage 90 of this phase of analysis of the defects, the defects, assimilated in a merger group in the preceding stage, are assimilated to a single defect called "fusion defect". For this purpose, for these grouped defects, the parameters that characterize the fusion defect are calculated by the linear combination of the values of the parameters that characterize each defect or irregularity of the fusion group. From the values of the parameters characterizing the non-regrouped defects and the melting defects, at the time of step 92 below, a supplementary classification of these defects is carried out according to the set of subclasses appropriate for each type of defect. This supplementary classification can be carried out according to the same type of methods as those used at the time of recognition of the defects. This supplementary classification ends in a result independent of the subsequent uses of the sheet. At the end of this supplementary classification, an "intrinsic defect profile" of the band can be defined for a given list of the population of each subclass of "severity" of each type or "elementary class" of the defect, this population is reported to one unit of band length; this profile can be represented for example in the form of histograms of the population of each subclass, placed side by side in a predetermined order (subclasses after subclasses, classes after classes). In parallel, for a given use of the band, it can be defined following the same formalism (for example: histograms in the same order) a "permissible malfunction profile", namely, for each subclass of "severity" of each possible defect type, a maximum allowable population for this given use (always reported to the same unit of band length). This "allowable malfunction profile" is not defined "once and for all" for a given use; the same can vary in its function, for example, the evolution of the memory of the loads of this use. The intrinsic defect profile of the inspected strip is then compared, in step 94, with respect to the admissible defective profile of desired use of said band. Thus, in the course of step 94, if it is found that the intrinsic defective profile of the inspected band re-enters (or is contained) in the admissible defective profile of the intended use of this band, this band is considered as acceptable or validated for this use (stage 96). If this is not the case, this inspected band is considered as unacceptable or "defective" against this use (stage 98). In order to avoid putting this inspected strip into waste, the use in the allowable malfunction profile of which the intrinsic malfunction profile of this inspected strip re-enters (or is contained) is then investigated., and this band is assigned to this other use. It is known in fact that a plate having a predetermined number of defects of a given severity and of a particular type can not be defective for one use, but can be defective for another use. For example, a sheet having a scratch is defective if it is not rolled at the time of a further processing step but is considered to be non-defective if it is re-rolled, the scratches are then crushed or debased. The decisive advantage of this method of assessing the malfunction of a band by the measurement of an intrinsic defective profile is that this measurement is independent of the subsequent use and downstream of the band, and of the evolution that refers to the criteria that have to be satisfied for this use. Advantageously, the intrinsic malfunction profiles of the inspected strips can, conversely, serve to follow the evolution and eventual derivatives of the manufacturing processes of these bands, according to for example the manufacturing campaigns; it is thus possible, for example, to mark possible derivatives of the behavior of the rolling chain upstream. The intrinsic defective profiles of the inspected bands can also serve to identify the derivatives on the inspection system thereof. According to a simplified variant of the defect analysis procedure, a coefficient whose value is a function of the estimated severity for a given use can be assigned to each subclass of "severity" of the defect types, and define the defective profile of a band by the sum of the populations of each one of the subclasses multiplied by the corresponding coefficient. To validate this use, it is then simply verified that the result obtained, namely said sum, does not exceed a predetermined value defined for this use. Other simplified variants, based on the use of the coefficients, can be contemplated.

It is noted that in relation to this date the best method known by the applicant to carry out the aforementioned invention, is that which is clear from the present description of the invention.

Having described the invention as above, property is claimed as contained in the following

Claims (13)

1. A method for inspecting the surface of an unrollable strip, for the detection of a surface defect, comprising the steps consisting of: forming, with the help of the means for taking the views, at least one numerical image of at least one of the faces of the band, consisting of a set of successive lines of elements of the images, each assigned a numerical value; filtering said at least one numerical image for the detection of surface irregularities, by detecting the relative variations of said numerical values, and treating said at least one filtered numerical image for the identification of the defect type of the surface corresponding to each irregularity detected; characterized in that, prior to the treatment step of said at least one numerical image, a general characterization of the irregularities is carried out determining for each one of them the value of the predetermined parameters characteristic of the surface defects and a previous classification is made of said irregularities, based on the determined values of said parameters, according to a set of predefined classes, the treatment stage is carried out on each class.

2. The method according to claim 1, characterized in that each predetermined parameter resembles a general reference point axis in a space whose dimensions correspond to said parameters, delimiting in said space corresponding regions each, to one of said pre-defined classes, previously to said previous classification, each irregularity is represented in said space by a point whose coordinates are the values of said parameters, and the previous classification is made by the identification of the region to which each point belongs, and by the assignment of the irregularity corresponding to the corresponding class of said region.

3. The method according to claim 1, characterized in that a second mode of characterization of the irregularities is determined for each predefined class whose number of characteristic parameters is lower than the number of characteristic parameters of the general characterization and, subsequently to the previous classification stage , it is determined for each detected irregularity the value of the characteristic parameters of the second specific characterization mode of said class to which the irregularity belongs, from the values of the characteristic parameters of the general characterization.

4. The method according to claim 2, characterized in that a simplified marking of irregularities is determined for each region whose number of axes is less than the number of axes of the general reference point and, after the previous classification stage, performs for each irregularity represented, a step of changing the mark of said general reference point towards said simplified mark specific to the region to which the irregularity belongs.

5. The method according to any of claims 1 to 4, characterized in that the stage of treatment of irregularities leads to a first stage of identification of the defect that corresponds to each irregularity, among a set of types of defects specific to the class to which said irregularity belongs and a second stage of classification of said identified defect for the purpose of confirmation and of the precision of the classification resulting from said first stage of classification.

6. The method according to claim 5, characterized in that it carries a step of qualifying the types of defects identified according to a first type of defects identified in a certain and / or accurate manner and a second type of defects identified in an uncertain manner and / or imprecise and because said second stage of classification is carried out only on the defects of the type classified as uncertain and / or imprecise.

7. The method according to any of claims 1 to 6, characterized in that it also carries a stage of regrouping the identified defects using a set of predefined criteria, especially geometric and / or topographic criteria.

8. The method according to any of claims 1 to 7, characterized in that it also carries the steps of counting the number of identified defects of the same type per unit length, and comparison of said number of defects of each type with a value of predetermined threshold representative of the minimum number of defects from which said defects are susceptible of presenting a periodic character, for the purpose of detecting periodic defects.

9. The method according to any of claims 1 to 7, characterized in that, after the step of determining the value of said parameters, and prior to said previous classification, a specific classification of the irregularities is carried out according to a set of elementary classes , and the population of said elementary classes is analyzed in order to detect periodic defects.

10. The method according to any one of claims 1 to 9, characterized in that after the filtering step, in response to a detection of an image element of an irregularity, a memory area is delimited for storage of the lines of the elements of the image supplied successively by the means of taking views and carrying at least one element of the image corresponding to at least one irregularity, each storage area is segmented into suspicious areas each having at least one irregularity of the surface, the suspicious areas of the successive storage areas are matched and correspond to the same irregularity and the total number of lines of the image elements of the suspicious areas matched is compared with a detection threshold of the defect of large length, and in the case of exceeding said threshold, the treatment step of said at least one im Agen filtered only on one of these paired suspicious zones, the result of the treatment is assigned to the other paired suspicious zones.

11. An installation for inspecting the surface of an unrollable strip for putting into operation a method according to any of the preceding claims, characterized in that it carries means for taking views of at least one of the faces of the strip, a memory for storing at least one image of the band in the form of lines and columns of elements of the image, each associated with a numerical value, a filtering circuit of said at least one numerical image for the detection of the irregularities of the surface of the band, by the detection of the relative variations of said numerical values, and a signal processing unit connected to said filtering circuit and carrying means for calculating the values of the parameters characteristic of the surface defects, means of classifying detected irregularities according to a predefined set of classes s from the values of said parameters and means of identifying each irregularity among a set of types of defects capable of corresponding to said irregularity.

12. The installation according to claim 11, characterized in that the treatment unit carries in addition to the second calculation means, for each irregularity detected, second characteristic parameters, whose number is lower than said characteristic parameters of the surface defects, with the object of the classification of irregularities detected according to a set of corresponding classes.

13. An installation according to one of claims 11 and 12, characterized in that the treatment unit also has second means for classifying the irregularities in order to confirm the classification carried out by the first means of classifying the irregularities.