• 专利附录
  • 专利说明
  • 权利要求
  • 类似技术
  • 同族专利
  • 引用文献
  • 法律状态
  • 优先权
    • 专利摘要:
    The method according to the invention serves to determine surface damage or superficial changes of objects to be inspected. According to the invention, it is ensured that an accurate and numerically stable, robust and simple determination of such surface defects is made possible.
    公开号:AT513126A4
    申请号:T50307/2012
    申请日:2012-08-01
    公开日:2014-02-15
    发明作者:Daniel Soukup；Reinhold Huber-Moerk；Konrad Mayer
    申请人:Ait Austrian Inst Technology；
    IPC主号:
    专利说明:

    1
    The invention relates to a method for testing a coherent image area for defects in the surface area of an object imaged in the image area according to the preamble of independent claim 1, in particular concave surface areas or surface defects or depressions. Furthermore, the invention relates to a data carrier according to claim 12.
    Basically, there is the possibility in the art of superficially applying to objects, e.g. Metal profiles, arranged surface defects, in particular pores or cracks, seen by an image processing unit is directed in the form of a line sensor on the surface of the article and the object is moved relative to the line sensor, so that a surface image of the surface of the article is created. In order to be able to better recognize surfaces in this case, the prior art further provides that two light sources, which are inclined in the direction of travel of the object, irradiate the object area of the light source in different colors and from different angles. As a result, different patterns result in the area of imaging of surface defects, with essentially parallel or mutually associated color areas in the image, each having the color of the light of the light source.
    The prior art is shown in detail in FIGS. 1 to 7. FIG. 1 shows an elongate article G to be imaged, in the present case a railway rail, which is displaced in a direction of travel z, which corresponds to the longitudinal extension direction z of the object G, with respect to a line sensor 2. Line images of the object G are taken with the line sensor 2. The receiving area of the object G lies in a plane xy, which is aligned normal to the longitudinal extension direction z of the object G. The area of the surface of the object G is located in the receiving area of the line sensor 2.
    In addition, in the apparatus shown in FIG. 1, two light sources 31, 32 arranged in fixed position relative to the line sensor 2 are provided.
    The first of the two light sources 31 emits radiation in the wavelength range of a first image channel R and is arranged in front of the line sensor 2 in the longitudinal direction z in the direction of movement. The first light source 31 emits a light plane that intersects with the plane of the receiving area 20 in the area of the object G.
    The second of the two light sources 32 emits radiation in the wavelength range of a second image channel B and is arranged in the longitudinal direction z after the line sensor 2. The second light source 32 emits a light plane that intersects with the plane of the receiving area 20 in the area of the object G. 2
    The line sensor 2 has a number of pixel sensors for separate detection of the light in one of the two light sources 31, 32 for one of the two image channels R, B. With each of the pixel sensors, an intensity value for each of the two image channels R, B can thus be determined separately. The pixel sensors thus measure or determine how large the proportion of the respective light of the light sources 31, 32 reflected thereon is.
    By joining the line images recorded with the line sensor 2, a surface image D is created. The area image D has two color channels R, B, wherein the color of the first image channel R corresponds to the color or the preferred wavelength range of the first light source 31 and the color of the second image channel B to the preferred wavelength range of the second light source 32. In the present case, a red image channel R and a blue color channel B are used. in Fig. 2, a detail of Fig. 1 is shown. The two beams 33, 34 of the light sources 31, 32 impinge on the object G whose surface is normal to the line of sight 23 of the line sensor 2 in the region imaged on the line sensor 2. The emitted from the first light source 31 beam 34 is deflected by the object G, wherein the reflected beam 36 is deflected approximately in the direction of the second light source 32.
    The emitted from the second light source 32 beam 33 is deflected by the object G, wherein the reflected beam 35 is deflected approximately in the direction of the first light source 31.
    FIG. 3 shows a detail from FIG. 1, wherein the object region g1 of the object G is inclined in the direction of the longitudinal extension direction z. The outgoing from the first light source 31 light beam 34 is imaged due to the tilt of the surface of the object in the range gt on the line sensor 2, the reflected beam 36 is thus on the line of sight 23 of the line sensor 2. The outgoing from the second light source 32 beam 33 is in the area g ,, in which the line of sight 23 of the line sensor 2 strikes the object G, the reflected light beam 35 in this direction is deflected sideways.
    FIG. 4 shows a detail from FIG. 1, wherein the object region g2 of the object G is inclined counter to the longitudinal direction z. The outgoing light beam 33 from the second light source 32 is imaged on the line sensor 2 due to the tilting of the object in the region g2, the reflected beam 35 thus lies on the line of sight 23 of the line sensor 2. The beam 34 coming from the second light source 31 becomes in the range g2, in which the line of sight 23 of the line sensor 2 meets the object G, reflects, the relevant reflected light beam 36 goes off to the side. 3
    FIG. 5 shows the first or red image channel of the digital image produced by the line sensor, in which the front flank of a shallow depression, viewed in the direction of the object to be inspected, is shown. FIG. 6 shows the second image channel, in the present case the blue image channel, of the digital image B, in which the rear flank of a shallow depression is shown in the object direction of the object to be inspected. FIG. 7 shows this by the two in FIG 5 and Fig. 6 illustrated image channels R, B created digital image D.
    The above-described method known from the prior art serves to determine surface damage or superficial changes of objects to be inspected. An essential problem of such methods is that an accurate and numerically stable, robust and simple determination of such surface defects is very difficult.
    Thus, the object of the invention, in a method of the type mentioned above to avoid the disadvantages mentioned above and to provide a method that does not have these disadvantages.
    The invention solves this problem with a method of the type mentioned above with the characterizing features of independent claim 1.
    According to the invention, in a method of inspecting a contiguous image area within a two-dimensional digital image, the presence of surface defects in the surface area of an object imaged on the image area within the digital image, wherein the imaged object has a longitudinal extent whose longitudinal direction coincides with the first coordinate directions of the two-dimensional digital image , and wherein the respective digital image has at least two image channels, so that each pixel is assigned a first and a second intensity value, it is provided that the following steps (a) to (d) are carried out for each pixel with image coordinates within the image region (a) for the two intensity values of the image channels for the respective pixel, a deviation value is determined by means of a deviation function Δ (·, ·) according to Δ (Ρ) = Δ (β (Ρ), B (P)), the value d the more the first intensity value exceeds the second intensity value, the more the first intensity value exceeds the second intensity value, (b) in the same way the number of possible results of the deviation functions evaluated in step (a) into a number of 4 for each pixel of the image area
    Intervals, the individual intervals being indexed by the increasing interval mean value, and it is determined in which of the intervals the deviation value determined for the pixel in step (a) falls and this interval is assigned to the respective pixel, (c) for all pixels in each case a number of mutually different displacement values is predetermined, and a respective number of further pixels is determined for the respective pixel whose coordinate extending in the direction of the longitudinal extent is greater by one of the predetermined displacement values than the corresponding coordinate of the respective pixel and their respective others Coordinate of the corresponding coordinate of the respective pixel is equal, m = k + d, η = I, (d) for each of the respective pixel associated further pixel (i) each have a further deviation value according to A '(Pm) = A' (R (Pm), B (Pm)), the value of the further deviation the more the second intensity value exceeds the first intensity value, and (ii) the number of possible results of the further deviation function Δ 'evaluated in step (d, i) in a number of others in the same way for each further pixel Intervals, the individual intervals being indexed by the increasing interval mean value, and it is determined in which of the further intervals the further deviation value determined for the further pixel in step (d, i) falls, and (iii) one assigned to the respective pixel Interval, the further interval ascertained in step (d, ii) and the respective displacement value of uniquely addressed memory area C (l, P, d) initialized before the execution of the step is increased by a predetermined value, so that after passing through all the pixels for each combination each interval defined in step b), one in step (d, ii and a shift value are each provided with a count, and (e) a Cooccurrence matrix is then determined from the memory area C for each of the shift times, the individual elements of the cooccurrence matrices being determined according to the following rule:
    Cd = [Cij.dli Cij, d = C (lj, lj, d) (f) the individual elements of the Cooccurrence matrices are normalized, and an averaged Cooccurrence matrix C is formed as the mean of the normalized Cooccurrence matrices, and (g) from the elements of the averaged Cooccurrence matrix £ those of their elements £ y
    5, whose respective index exceeds a predetermined minimum value i0, j0, as a result of the accumulation a result value u is obtained which is compared with a threshold value t and (h) that in the case where t > u is considered a surface defect in the respective surface area imaged on the image area.
    In order to allow an advantageous discrimination of blue and red color components in pixels, it can be provided that the deviation function Δ (-, -) is defined as follows: R {P) - B (P), if R (P) -B ( P) > 0 Δ {Ρ) = A (R (P), B (P »= 0, otherwise.
    For the same purpose, it can advantageously be provided that the further deviation function is determined by the deviation function as follows: A'tP,) = A '(R (Pi)> B (P,)) = Δ (Β {Ρ,), RfPO).
    In order to reduce the numerical effort and to obtain results in a reasonable amount of time, it can be provided that a number of, in particular equal, intervals in the range of values of the deviation function is selected which overlap-free and all possible values of the deviation function lie within the value range of the deviation function , and / or - that a number of, in particular equally large, further intervals in the value range of the further deviation function is selected which overlap-free and all possible values of the further deviation function covering the value range of the further deviation function covering, and / or - that the number of intervals and the number of further intervals is chosen to be the same.
    In order to be able to detect surface defects of different extent in the direction of expansion of the object to be inspected, it can be provided that integer values in a range between 1 and 10 are selected as shift values, and / or a number of three to five shift values is selected and / or - that a number of immediately consecutive integer values is selected as the shift values whose smallest value is between 1 and 3 and / or whose maximum value is between 5 and 10 and / or whose number is between 3 and 8.
    A specific adaptation to certain types of surface defects envisages: - that the minimum value be chosen such that the intervals whose assigned index is less than the minimum value or the minimum value are at least 5%,
    6, in particular at least 10%, of the entire range of values of the deviation function, and / or - that the further minimum value is chosen so that the further intervals whose associated index is smaller than the further minimum value or the further minimum value, a share of at least 5 %, in particular at least 10%, of the entire value range of the further deviation function.
    In order to avoid the detection of false positives, it can be provided that for a further pixel a value in the memory area is increased only when the further pixel lies within the image area.
    In order to be able to judge large and small aberrations according to the same scales as well as to achieve an image resolution independent of the image resolution, it can be provided that the individual cooccurrence matrices are normalized by dividing each of the values of the respective cooccurrence matrix by the total sum of the individual elements is divided by the respective Cooccurrence matrix.
    In order to be able to detect surface defects with different extents in the direction of expansion of the object to be inspected, it can be provided that the average cooccurence matrix is produced by element-wise arithmetic mean-value formation of the cooccurrence matrices, in particular according to the following rule: C = 1 / N * SUM ( d = 1 ... N) Cd where N equals the number of cooccurrence matrices.
    Advantageously, to detect an object it may be provided that, prior to the determination of surface defects, the digital image is captured as follows: a) that the object is picked up by a line sensor having a pickup area in a predetermined plane normal to the longitudinal direction of the object; the object is located in the receiving area of the line sensor, b) that at least two light sources fixed relative to the line sensor are used, wherein the first light source for emitting radiation in the wavelength range of the first image channel is arranged in front of the line sensor in the longitudinal direction and the second light source for emitting a Radiation in the Welienlängenbereich of the second image channel in the longitudinal direction of extension is arranged behind the line sensor and the respective light of both light sources is irradiated in the region of the object, in the receiving area c) that the line sensor each has a number of pixel sensors for separately detecting the light of one of the two light sources for each one of the two image channels, with each of the pixel sensors separately an intensity value of the reflected and reflected from each of the two light sources Light is measured, 7 d) that with each pixel sensor in each case a first intensity value is determined, which corresponds to the light emitted from the first light source and reflected on the pixel sensor, and a second intensity value is determined, the one emitted by the second light source and on the E) that the line sensor and the light sources are fixed relative to one another and are moved in the longitudinal direction of the object to be inspected relative thereto, and f) that a number of line images are taken with the line sensor in accordance with their recording sequence e be assembled into a two-dimensional digital image.
    An advantageous embodiment of the invention, which improves on the above-mentioned and illustrated in FIGS. 1 to 7 prior art, is shown below with reference to FIGS. 8 to 13.
    Fig. 8 shows a digital image of the subject in which a number of image areas have been detected having features that could represent surface defects. Fig. 9 shows one of the image areas shown in Fig. 8 in detail. Fig. 10 shows the judgment of an image area having different shift values. Fig. 11 shows a preferred embodiment for determining deviation values by a deviation function. FIG. 12 shows the meaning of the displacement value on the basis of a surface defect located on the object G. Fig. 13 shows a number of coocurrence matrices for different displacement values. Fig. 14 shows a mean Cooccurrence matrix as well as the determination of whether there is a surface defect.
    As shown in Fig. 8, the digital image D has a plurality of image areas L1 (..., L5, which are basically considered to be surface defects F and checked for the presence of surface defects F. For all of these image areas L, .. In the evaluation of the individual image regions L1t..., L5, it is generally determined whether the brightness values of the individual pixels P of the image region L 1... U are of a predetermined background value The background value may be determined in a variety of ways known in the art, in particular the background value may be determined by the formation of a moving average, and the image areas Lt ..... L5 are those
    Areas in which the intensity values of the individual pixels P deviate by a predetermined threshold from the respective background value. In the present exemplary embodiment, the image areas Lt,..., Ls are each shown as rectangles for the sake of clarity. In principle, however, it is possible and customary for the individual to 8
    Error to be examined image areas U, L5 have a very different shape, each of the image areas Lu ..., L5 each having a contiguous set of pixels. Furthermore, the longitudinal direction z of the object G is shown in FIG. in Flg. 9, the individual pixels of the image area L - are shown schematically in more detail. For each of the pixels P of the image area Li shown in FIG. 9, the following steps (a) to (e) are sequentially performed. Each of the pixels P, as shown in Fig. 10, each two image coordinates k, I assigned. In this case, the coordinate I corresponds to the respective position of the sensor pixel of the line sensor, which has determined the respective brightness value. The second image coordinate k of the respective pixel P of the digital image D corresponds in each case to the recording time of the line. Measured on the object G, the coordinate direction of the image coordinate k extends in the longitudinal extension direction z or in the direction of advancement of the object G relative to the line sensor 2.
    In preparation, the two intensity values R (P) and B (P) of the image channels for the respective pixel are respectively determined with each sensor pixel of the line sensor.
    In the first step (a), a deviation value Δ (Ρ) = A (R (P), B (P)) is determined separately for each pixel on the basis of the respective intensity value in the first and second image channels R, B. In this case, a variable is found which indicates how much the intensity values R (P) of the respective pixel P in the first or red image channel R are different from the intensity values B (P) in the second or blue image channel B. For this purpose, a deviation function Δ is used, which in the present exemplary embodiment of the invention is illustrated in more detail in FIG. 12. If the intensity value R (P) of the pixel P in the first or red color channel R is greater than the intensity value B (P) in the blue color channel B, then the value of the deviation function A (R (P), B (P) corresponds to the value of Difference of the two intensity values R (P) -B (P) for the respective pixel P. However, if the second intensity value B (P) is greater than the first intensity value R {P) of the pixel, then the deviation value determined by evaluation of the mapping function Δ has Δ (P) is the value 0.
    The deviation function Δ (.,.) Is defined in this particular embodiment of the invention as follows: {R (P) -B (P) if R (P) -B (P)>. 0 Δ (Ρ) = A (R (P), B (P)) = {0, otherwise g
    This course of the deviation function Δ is shown schematically in FIG. 11 (solid line). In principle, any deviation functions Δ can be used whose value is greater in each case the more the first intensity value R {P) in the respective pixel P exceeds the second intensity value B (P) in the respective pixel P. The thus determined deviation value Δ {Ρ) is assigned to the respective pixel P.
    The determination of the deviation value Δ (Ρ) takes place in order to determine whether in the respective pixel the component or the intensity value of the first image channel substantially exceeds the intensity value of the second image channel and thus a superficial unevenness g1 as shown in Fig. 3, is present. Pixels P, which for this reason have a significantly higher proportion in the first or red image channel R, are thus regarded as images of a surface area gi of the object G which drops in the direction of travel z. Such pixels are shown in Figs. 9 and 10 by the symbol X.
    Pixels P whose intensity value B (P) in the second or blue image channel B is greater than their first intensity value R (P), on the other hand, show a surface area g2 which drops counter to the direction of travel z and are indicated by the symbol 0 in FIGS. 9 and 10 shown.
    In a second step (b), a number of intervals h, ..., I ^ are set within the value range of the deviation function Δ {.,. In the present embodiment, each intensity value R (P), B (P) of a pixel P is in a range between 0 and 255. Consequently, the value range of the deviation function Δ (Ρ) is also between 0 and 255. This range of values will be described below in eight different , equally large, non-overlapping and the entire value range filling intervals h ..... I ™ * divided. The individual intervals are indexed according to increasing interval average. Each interval U,..., Imax thus has an interval width of 32, the first interval li ranging between 0 and 31, the second interval l2 ranging between 32 and 63,..., And finally the last interval Imax in the range between 224, ..., 255 lies. For each of the pixels P it is now checked in which of the intervals h,..., Imax the deviation value Δ (Ρ) determined for the pixel is obtained, which was determined as the result of the deviation function Δ evaluated in step (a). This interval will henceforth be referred to as l (P).
    Alternatively, of course, it is also possible to specify different interval widths for adaptation to specific types of errors. For example, it may also be provided that the intervals with a lower interval mean value have a smaller interval width than the intervals with a larger interval mean value.
    In a third step (c), a number of different
    Shift values di ..... dmax specified. In the present case, the
    Displacement values d1 = 1, d2 = 2, ..., d6 = dmax = 6 specified. 10
    The meaning of the shift value d is shown in FIG. 11, which shows the picking up of five object lines of the object G at five consecutive pickup times t = 1... T = 5.
    At the first recording time t = 1, the object G has on its surface an edge of a surface defect in the region imaged on the line sensor 2 with an inclination counter to the direction of travel z, resulting in a situation as shown in FIG. The second color value exceeds the first color value, since only the light from the first light source 31 falls on the line sensor 2.
    At the last recording time t = 5, the object G has on its surface an edge of a surface defect, in particular a pore, in the region imaged on the line sensor 2 with an inclination in the direction of travel z, resulting in a situation as shown in FIG. 4. The second color value exceeds the second color value.
    Due to the composition of the digital image D by individual line images, the created pixels of the two inclined surface regions g 1, g 2 have a spacing of d = 4 pixels in the longitudinal direction z of the object G. By selecting different displacement values di, find large surface defects to be detected. The width of the respective surface defect corresponds in each case to d times the travel which the object G travels between two successive acquisition times. A number of further pixels P: ..... Pmax are respectively determined for the respective pixel P, wherein the coordinate m of the further pixel Pi,..., P ^ extending in the direction of the longitudinal extension z is in each case one of the predetermined shift values di , ..., dma) < is greater than the corresponding coordinate k extending in the longitudinal direction z of the respective pixel P. The other coordinate n of the further pixel P 1 ..... Pmax corresponds to the respective coordinate I of the respective pixel P. Consequently, m = k + d, η = I. Thus, for the first further pixel Pi, the coordinate m of the first further pixel Pi extending in the longitudinal extension direction z differs by the predetermined first displacement value d1 = 1 from the respective coordinate k of the pixel P extending in the longitudinal extension direction z, thus m = k + di = k + 1. The two other coordinates η, l of the pixel P and of the first further pixel Pi are the same size. Accordingly, for the second further pixel P2, the coordinate m of the second further pixel P2 extending in the longitudinal direction z differs by the predetermined first displacement value d2 = 2 from the respective coordinate k of the pixel P extending in the longitudinal direction z, thus m = k + d2 = k + 2. The other two coordinates η, I of the pixel P and the second further pixel P2 are the same size. For each of the other pixels Pi, ..., Pmax associated with the respective pixel P, the following steps (d, i), (d, ii) and (d, iü) are carried out. 11 For the respective further pixel P ..... Pmax in step (d, i) another
    Deviation value Δ '(Ρ,), Δ' (Ρ2), ..., Δ '(Ρη) ..... A' {Pmax) according to A '(Pm) = Δ ^ Ρ ™), B {Pm) ). The thus determined value of the further deviation function Δ 'is greater, the more the second intensity value B (Pm) exceeds the first intensity value R (Pm). An example of such a further deviation function is shown in FIG. 12 as a dashed line. This further deviation function is set as follows: {B (Pm) -R (Pm) when B (Pm) -R (Pm)>. 0 A '(Pm) = A' (R (Pm), B (Pm)) = {0, else.
    The further deviation function Δ 'in this particular exemplary embodiment of the invention is defined as follows: Δ' (Ρ,) = A '{R (P,), B (P,)) = Δ ^ Ρ,), R (P,) ).
    The value range of the further deviation function Δ ', which in the present exemplary embodiment corresponds to the value range of the deviation function Δ, is subdivided into a number of further intervals I /,... Advantageously, the subdivision of the value range of the further deviation function Δ 'takes place analogously to the subdivision of the value range of the deviation function Δ. In the present exemplary embodiment, the value range of the further deviation function, which comprises the range of natural numbers between 0 and 255, is subdivided into 8 further value ranges 1 /,..., Lax ', which in each case encompass the entire value range of the further deviation function Δ' and one have the same interval width of 32 each. Each of the intervals W ..... Imax 'is indexed according to increasing interval mean value.
    In step {d, ii), for all further pixels Pi,... Pmax and the values of the deviation function Δ 'associated therewith are determined, in which of the further intervals I /, l2' ..... Imax 'assigned to them fall.
    As shown in Fig. 13, a memory area C is used, which is uniquely addressable by specifying an interval l (P), a further interval l '(Pm) and a shift value d. In the present exemplary embodiment, this memory area C has a total of 8 × 8 × 6 numerical memory areas Ci, j, d, which are initialized to a predetermined value, in the present case 0, at the beginning of the execution of the method. In the present exemplary embodiment, the memory area C is arranged as a three-dimensional memory field which can be addressed via three indices i, j, d. The first index i corresponds to the index of the respective interval l (P) in which the deviation value Δ (Ρ) associated with the respective pixel P is located. The second index j corresponds to the index of the further interval l '(Pm) within which the further 12 determined in step (d, i)
    Deviation value A '(Pm) is. The third index d corresponds to the respective displacement value d by which the first coordinates k, m of the pixel P and of the further pixel Pm situated in the longitudinal direction z differ from one another.
    In step (d, iii), a storage area cy, d is now uniquely addressed on the basis of the interval I assigned to the respective pixel P or its index i, the further interval Γ assigned to the further pixel Pm or its index j and the respective shift value d , The respective memory area ci (| id addressed thereby is increased by a predetermined value, in the present case by the value 1. After passing through all the pixels P, for each addressed memory location Ci, j, d of the memory area C or for each combination i, j, dj of each interval I defined in step b, each further interval Γ specified in step (d, ii) and each shift value d are each a count indicating how many times such a specified combination [i, j, d ] was found during the completed steps (a) to (d).
    The individual two-dimensional memory areas..., Cmax of the memory area C, which represent two-dimensional memory areas with the same displacement value d in each case, are referred to below as cooccurrence matrices Cd. The elements Ci, j, d of the coocurrence matrices Cd are determined according to the following rule:
    Cd - [Cjj, d] j Cjtjtd = C (lj, lj, d)
    The individual elements of the Cooccurrence matrices Cd are normalized in the present embodiment such that the sum of all entries within the respective Cooccurrence matrix results in a total of 1. Scaling can also be done in other ways, such as giving the sum of the squares of the individual entries of the Cooccurrence matrix Cd a total of 1. The coocurrence matrix Ci, for which d = di = 1 holds, is in Flg. 13 shown in more detail.
    Subsequently, an average coocurrence matrix C, which is formed as the mean value of the normalized cooccurrence matrices C, is determined from the individual ones of the cooccurrence matrices shown in FIG. The individual entries Cg of the averaged Cooccurrence matrix Q are determined as the mean value of the entries c 1 of the memory area, wherein an averaging takes place via the third index d.
    As shown in FIG. 14, those elements whose respective index i, j exceeds a predetermined minimum value i0, jo are accumulated from the individual elements of the determined cooccurrence matrix C. In the present case, the minimum values i0 = j0 = 1 are set. As a result of the accumulation, a result value u is determined, which is compared with a predetermined threshold value t. If the result value u the 13
    Threshold t, it is concluded that there is a surface error F in the respective selected image area l_i ..... L5.
    权利要求:
    Claims (12)
    [1]
    Claims 1. A method of inspecting a contiguous image area (L) within a two-dimensional digital image (D) for the presence of surface defects (F) in the surface area of an object (G) imaged on the image area (L) within the digital image (D) wherein the imaged article (G) has a longitudinal extent whose longitudinal direction (z) coincides with the first coordinate directions of the two-dimensional digital image (D), and wherein the respective digital image (D) has at least two image channels (R, B) such that each pixel (P) in each case a first and a second intensity value (R (P), B (P)> is assigned, characterized in that for each pixel (P) with image coordinates (k; I) within the image area (L), the following Steps (a) to (d) are carried out, namely that (a) for the two intensity values (R (P), B (P)) of the image channels (R, B) for the respective pixel (P) a deviation w ert (Δ (Ρ)) is determined by means of a deviation function Δ (·, -) according to Δ {Ρ) = A (R (P), B (P)), the value of the deviation function (Δ) being greater, je more the first intensity value &lt; R (P)) exceeds the second intensity value (B (P)), (b) respectively for all pixels (P) of the image area in the same manner, the number of possible outcomes of the deviation functions evaluated in step (a) is divided into a number of intervals (I ,, l2, ..., Imax), wherein the individual intervals are indexed by increasing interval mean value, and it is determined in which of the intervals for the pixel (P) in step (a) determined deviation value (Δ (Ρ)) and this interval (l {P)) is assigned to the respective pixel (P), (c) respectively for each pixel (P) a number of mutually different shift values <di, d2, ..., d ™ ") is given, and for the respective pixel (P) in each case a number of further Pi xeln {P: ..... Ρ ™ χ) is determined, in the direction of the longitudinal extent (z) extending coordinate (m) in each case by one of the predetermined displacement values (d ,, d2, ..., dmax) is greater than the corresponding coordinate (k) of the respective pixel (P) and their respective other coordinate (s) is equal to the corresponding coordinate (I) of the respective pixel (P), m = k + d, η = I, <d) each of the further pixels (P 1..., Pma *) (i) assigned to the respective pixel (P) has in each case a further deviation value <Δ '(Ρ,), Δ' (Ρ 2) ..... A '( Pm) · - A '(Pmax)) is determined according to A' (Pm) = A '(R {Pm), B (Pm)), the value of the further deviation function (Δ') being larger the more second intensity value (B (P)) exceeds the first intensity value (R (P)) and 2 (ii) respectively for all further pixels (P) in the same way the number of possible outcomes of the further evaluated in step (d, i) Deviation function Δ 'in a number of other In tervals (1 /, l2 '..... Ima **) is divided, whereby the individual intervals are indexed by increasing interval mean value, and it is determined in which of the further intervals (l' (Pm)) that for the further pixel (Pm) further deviation value (A '(Pm)) determined in step (d, i) falls and (iii) an interval (I) assigned to the respective pixel, the further interval (Γ) determined in step (d, ii) ) and the respective offset value (d) of uniquely addressed memory area C {1, Γ, d) initialized before step <a) is increased by a predetermined value, such that after passing through all the pixels (P) for each combination (I , Γ, d) each interval defined in step b), a further interval (1 /, l2 ', ..., W) determined in step (d, ii) and a shift value (d) each have a count, and (e) subsequently from the storage area C for each of the shift values (d) respectively a coocurrence matrix (C, ..... CmEU!) is determined, the individual elements (Cj, j> d) of the coocurrence matrices (Cd) being determined according to the following rule: Cd = [Cj, j, d] j Cij, d = C (lit lj, d) (f) the individual elements of the Cooccurrence matrices are normalized, and an averaged Cooccurrence matrix C is formed as the mean of the normalized Cooccurrence matrices, and (g) from the Elements of the averaged Cooccurrence matrix C those of their elements Oy are accumulated whose respective index (i, j) exceeds a predetermined minimum value i0, jo, as a result of the accumulation a result value u is obtained, which is compared with a threshold value t and h) that in the event that t &gt; u is considered a surface defect (F) in the respective surface area imaged on the image area (L).
    [2]
    A method according to claim 1, characterized in that the deviation function Δ (-, -) is set as follows: {R {P) - B (P), when R (P) -B (P) &gt; 0 Δ (Ρ) = A (R (P), B (P)) = {0, else.
    [3]
    3. The method according to claim 1 or 2, characterized in that the further deviation function <Δ '(·, ·)) is determined by the deviation function (Δ (-, -)) as follows: 3 Δ' (Ρ,) = A '(R (Pt), Β (Ρ,)) = ΔΐΕΚΡΟ, RfP,)).
    [4]
    4. The method according to any one of the preceding claims, characterized in that - a number of, in particular equally large, intervals (h, .... Imax) is selected in the value range of the deviation function, the overlapping and all possible values of the deviation function covering in the range of values lie the deviation function, and / or - that a number of, in particular equally large, further intervals (l'i, ..., I'max) is selected in the value range of the further deviation function, the overlapping and covering all possible values of the further deviation function lie in the value range of the further deviation function, and / or - that the number of intervals (I, ..... Imax) and the number of further intervals (P1 (..., I'max) is selected to be the same.
    [5]
    5. Method according to one of the preceding claims, characterized in that - integer values in a range between 1 and 10 are selected as shift values (d), and / or - a number of three to five shift values (d) is selected and / or - that a number of immediately consecutive integer values whose smallest value is between 1 and 3 and / or whose maximum value is between 5 and 10 and / or whose number is between 3 and 8 is selected as shift values (d).
    [6]
    6. The method according to any one of the preceding claims, characterized in that - the minimum value (i0) is chosen such that the intervals (U ..... Ii0) whose associated index (i) is less than the minimum value or the minimum value (i0) equals a proportion of at least 5%, in particular at least 10%, of the entire range of values of the deviation function, and / or - that the further minimum value (jo) is chosen such that the further intervals (l. ., I'jo) whose assigned index (j) is less than the further minimum value or the further minimum value (j0) equals a proportion of at least 5%, in particular at least 10%, of the entire value range of the further deviation function.
    [7]
    7. The method according to any one of the preceding claims, characterized in that for a further pixel (Pi, ..., Pmax) only one value in the memory area (C) is increased when the further pixel (P1, .... P ™ ") lies within the image area (L).
    [8]
    8. Method according to one of the preceding claims, characterized in that the individual cooccurrence matrices (Cd) are normalized by dividing each of the values of the respective cooccurrence matrix (Cd) by the total sum of the individual elements of the respective cooccurrence matrix (Cd ) is divided.
    [9]
    9. The method according to any one of the preceding claims, characterized in that the average Cooccurence matrix (C) by elementary arithmetic averaging of the Cooccurrence matrices (Cd) is created, in particular according to the following rule: C = 1 / N * SUM (i = 1 ... N) Cd where N equals the number of cooccurrence matrices (Cd).
    [10]
    10. The method according to any one of the preceding claims, characterized in that prior to the determination of surface defects (F), the digital image (D) is recorded as follows: a) that the object (G) is recorded with a line sensor (2) whose receiving area ( 26) lies in a predetermined plane (xy) oriented normal to the longitudinal direction (z) of the article (G), the article (G) being in the receiving region (20) of the line sensor (2), b) at least two the first light source (31) for emitting radiation in the wavelength range of the first image channel (R) in the longitudinal direction (z) in front of the line sensor (2) is arranged, and the second light source (32) for emitting radiation in the wavelength range of the second image channel (B) in the longitudinal direction (z) behind the line sensor (2) is arranged and the respective light of both light sources (31, 32) is irradiated in the region of the object (G) which lies in the receiving region of the line sensor (2), c) that the line sensor (2) each comprise a number of pixel sensors for separate detection of the light of one of the two light sources (31, 32) for each one of the two image channels (R, B), with each of the pixel sensors separately an intensity value (R, B) of the light emitted from each of the two light sources (31, 32) and reflected on it is measured, d) that with each pixel sensor in each case a first intensity value (R) is determined, which corresponds to the output from the first light source (31) and reflected on the pixel sensor light, and a second intensity value (B) is determined, which is the e) that the line sensor (2) and the light sources (31, 32) are fixed in relation to one another and in the longitudinal direction of extension (32) emitted and reflected on the pixel sensor (32) z) the object to be inspected (G) are moved relative thereto, and f) that a number of line images are picked up with the row sensor (2) and assembled into a two-dimensional digital image (D) according to their recording order.
    [11]
    11. The method according to any one of the preceding claims, characterized in that individual blobs are detected from the digital image by means of the blob analysis and the detected blobs are used as coherent image areas (L).
    [12]
    12. data carrier on which a computer program for carrying out a method according to one of the preceding claims is stored.
    类似技术:
    公开号 | 公开日 | 专利标题
    DE102012104282A1|2013-11-21|Method and device for inspecting surfaces of a tested object
    DE10020893B4|2018-07-05|Method for optical shape detection of objects
    DE2260090B2|1978-12-21|Photoelectric device for determining the roughness or smoothness of diffuse-scattering surfaces
    WO2016005571A1|2016-01-14|Determining the position of an object in the beam path of an optical device
    EP2693403B1|2015-02-11|Method for testing a contiguous section of an image for errors in the surface area of an object
    EP2619525B1|2015-07-22|Method for optically scanning an edge in or on a surface region
    WO2007079934A2|2007-07-19|Method and system for the optical inspection of a periodic structure
    DE102013212495A1|2014-12-31|Method and device for inspecting a contoured surface, in particular the underbody of a motor vehicle
    DE102011012729A1|2012-09-06|Optical test method using intensity curve
    DE102015121673A1|2017-06-14|shape investigation
    DE20317095U1|2004-03-11|Cast metal part surface error inspection unit has imaging of reflected light with processing algorithm grouping pixels above threshold for error examination and contour construction
    DE102013114687A1|2015-06-25|Method and device for the three-dimensional optical measurement of objects with a topometric measuring method and computer program for this purpose
    DE10301094B4|2015-09-03|Device for measuring the distance of distance points to a camera
    DE102016008744A1|2017-03-02|Image measuring apparatus, control program of such apparatus and non-volatile recording medium on which the control program is recorded
    EP2913632B1|2019-09-11|Method for measuring an object by means of X-ray fluoresence
    DE102018122816A1|2020-03-19|Method and device for determining a property of an object
    DE102019208114A1|2020-12-10|Device for 3D measurement of object coordinates
    EP3462164A1|2019-04-03|Assembly and method of inspecting moved plate-shaped objects
    DE102017106764A1|2017-10-26|TEST DEVICE, STORAGE MEDIUM AND PROGRAM
    EP3049757A1|2016-08-03|Chassis measurement under ambient light
    EP3048589A1|2016-07-27|Inspection of ovi features
    EP2902963B1|2016-10-05|Method for producing an image of an object
    DE102011083621B4|2018-03-22|Method and device for manufacturing control of a collimator
    DE102020209849A1|2022-02-10|Method for determining an optical crosstalk of a lidar sensor and lidar sensor
    DE102019100615A1|2020-07-16|Method and device for detecting a coating on a surface
    同族专利:
    公开号 | 公开日
    EP2693403A1|2014-02-05|
    AT513126B1|2014-02-15|
    EP2693403B1|2015-02-11|
    引用文献:
    公开号 | 申请日 | 公开日 | 申请人 | 专利标题
    DE19511534A1|1995-03-29|1996-10-02|Fraunhofer Ges Forschung|Detecting 3=D fault locations with automatic monitoring of specimen surfaces using camera|
    US6166393A|1997-08-22|2000-12-26|Fraunhofer-Gesellschaft Zur Forderung Der Angewandten|Method and apparatus for automatic inspection of moving surfaces|
    EP2463175A2|2010-12-13|2012-06-13|AIT Austrian Institute of Technology GmbH|Method for locating surface defects|
    AT516207B1|2014-08-20|2016-06-15|Ait Austrian Inst Technology|Method of detecting elevations on specular surfaces|
    JP6249241B2|2015-03-30|2017-12-20|Ｊｆｅスチール株式会社|Metal plate surface defect detection method|
    CN107709977B|2015-06-25|2020-05-08|杰富意钢铁株式会社|Surface defect detection device and surface defect detection method|
    KR20180009791A|2015-06-25|2018-01-29|제이에프이 스틸 가부시키가이샤|Surface flaw detection method, surface flaw detection device, and manufacturing method for steel material|
    法律状态:
    2019-04-15| MM01| Lapse because of not paying annual fees|Effective date: 20180801 |
    优先权:
    申请号 | 申请日 | 专利标题
    ATA50307/2012A|AT513126B1|2012-08-01|2012-08-01|Co-occurrence matrix|ATA50307/2012A| AT513126B1|2012-08-01|2012-08-01|Co-occurrence matrix|
    EP13455004.5A| EP2693403B1|2012-08-01|2013-05-28|Method for testing a contiguous section of an image for errors in the surface area of an object|
    [返回顶部]