Method for detecting a defect on a surface of a tire

Abstract

A method for detecting a defect on a surface of a tire includes automated steps of: calculating values of a gradient of a plurality of texture parameters from an image of the surface of the tire, determining an image of the gradient, and thresholding of the image of the gradient to obtain a thresholded image.

Claims

1. A method for detecting a defect on a surface of a tire, the method comprising: automatically calculating, using a computer, values of a gradient of a plurality of texture parameters from an image of the surface of the tire; automatically determining, using the computer, an image of the gradient; automatically thresholding, using the computer, the image of the gradient to obtain a thresholded image; and displaying the thresholded image, wherein the plurality of texture parameters comprises at least five texture parameters.

2. The method according to claim 1, wherein the image of the surface of the tire is an image of an inner surface of the tire.

3. The method according to claim 1, further comprising using a laser system to automatically acquire the image of the surface of the tire by laser triangulation.

4. The method according to claim 1, wherein the computer calculates the values of the gradient in a selection of pixels of a part of the image of the surface of the tire.

5. The method according to claim 1, further comprising automatically calculating, using the computer, a grey-level gradient in the image of the surface of the tire.

6. The method according to claim 1, wherein the thresholding is performed by the computer using a hysteresis process.

7. The method according to claim 1, wherein the plurality of texture parameters consists of five texture parameters.

8. The method according to claim 1, wherein the plurality of texture parameters comprises Haralick parameters.

9. The method according to claim 1, wherein the plurality of texture parameters comprises a local binary pattern.

10. The method according to claim 1, wherein the plurality of texture parameters comprises 25 texture parameters.

11. A computer-readable storage medium storing a program that when executed by a computer causes the computer to perform a method for detecting a defect on a surface of a tire, wherein the method includes: automatically calculating values of a gradient of a plurality of texture parameters from an image of the surface of the tire; automatically determining an image of the gradient; automatically thresholding the image of the gradient to obtain a thresholded image; and displaying the thresholded image, wherein the plurality of texture parameters comprises at least five texture parameters.

12. A method for detecting a defect on a surface of a tire, the method comprising downloading to a computer, via a telecommunications network, coded instructions that when executed by the computer cause the computer to: automatically calculate values of a gradient of a plurality of texture parameters from an image of the surface of the tire, automatically determine an image of the gradient, automatically perform a thresholding process on the image of the gradient to obtain a thresholded image, and display the thresholded image, wherein the plurality of texture parameters comprises at least five texture parameters.

13. An installation for detecting a defect on a surface of a tire, the installation comprising a computer programmed to: automatically calculate values of a gradient of a plurality of texture parameters from an image of the surface of the tire, automatically determine an image of the gradient, automatically perform a thresholding process on the image of the gradient to obtain a thresholded image, and display the thresholded image, wherein the plurality of texture parameters comprises at least five texture parameters.

14. The installation according to claim 13, further comprising a laser system that automatically acquires the image of the surface of the tire.

Description

BRIEF DESCRIPTION OF THE DRAWINGS

(1) An embodiment of the invention will now be described with the aid of the attached drawings, in which:

(2) FIG. 1 is a schematic view of an installation for detecting a defect on a surface of a tire according to the invention, and

(3) FIG. 2 is a diagram showing the execution of the steps of the method for detecting a defect according to the invention.

DETAILED DESCRIPTION OF EMBODIMENTS OF THE INVENTION

(4) FIG. 1 shows an installation 2 according to the invention. This installation 2 is intended to detect defects on a surface of a tire 4. It is assumed here that the surface in question is an inner surface of the tire 4, but provision may be made for it to be another surface of the tire 4.

(5) The installation 2 comprises automated means 6 adapted for acquiring an image of the surface of the tire 4. The automated acquisition means 6 acquire the image of the surface by laser triangulation, and for this purpose they comprise, notably, a laser source 8 and a sensor 10. Since laser triangulation is a known acquisition method, it will not be detailed further in the following text. Provision may alternatively be made for the image of the surface to be acquired by photometric stereo or by means of what is known as an industrial vision camera.

(6) The installation 2 further comprises automated calculation means 12 connected to the automated acquisition means 6 by a connecting member 14 in such a way that they can exchange data. These automated calculation means 12 comprise, notably, a computer 16. The functions of the automated calculation means 12 will be examined subsequently.

(7) A description will now be given of a method for detecting a defect on the surface of the tire 4 according to the invention, executed by the installation 2, and notably by its automated means 6, 12, using a computer program comprising coded instructions for commanding the execution of the detection method when it is executed by these means. In the present case, the computer program is stored on a storage medium readable by the automated means 6, 12. The computer program is also made available for downloading on a telecommunications network.

(8) With reference to FIG. 2, in step A, the image of the inner surface of the tire 4 is initially acquired by laser triangulation, using the automated acquisition means 6. In this way, an image 18, called the surface image, is obtained. This surface image 18 is then transmitted to the automated calculation means 12 by means of the connecting member 14. This image is in black and white, and comprises grey levels.

(9) Subsequently, in step B, the automated calculation means 12 calculate values of a texture gradient from the surface image 18, by means of the computer 16. A description will now be given of an algorithm for calculating these values.

(10) A texture parameter is initially selected. In the field of image processing, there is a known way of using one of the following parameters: co-occurrence matrices, Haralick parameters, morphological parameters (such as granulometries by opening and closing), and a local binary pattern (commonly referred to as an LBP, the abbreviation for the English term Local Binary Pattern). These parameters all have the common feature of corresponding to spatial statistics on the grey level values of the image. More generally, the calculation of the texture gradient values may be applied to any texture parameter and to any statistic calculated on the image.

(11) In the following text, it will be assumed that the automated calculation means 12 calculate values of a gradient of a plurality of texture parameters in the image of the surface. This number of parameters is denoted n.

(12) Let F.sub.i,j be the vector parameters containing the n parameters calculated over a rectangular window having dimensions W.sub.iW.sub.j centered on a pixel (i,j) of the surface image 18, assumed to be rectangular and to have the dimensions S.sub.iS.sub.j.

(13) This vector F.sub.i,j is calculated for any pixel (i,j) of the surface image 18, except at its edges, that is to say

(14) $i\left[\frac{W_i}{2},S_i-\frac{W_i}{2}\right]\mathrm{and}j\left[\frac{W_j}{2},S_j-\frac{W_j}{2}\right].$
This results in a hypermatrix F with a size of (S.sub.iW.sub.i, S.sub.jW.sub.j, n).

(15) Let d be a metric allowing for the distance between two vectors F.sub.i,j. It should be borne in mind that a metric is a function of .sup.n.sup.n in the set which satisfies the following conditions for all F.sub.i,j, F.sub.k,l, F.sub.m,n in .sup.n:

(16) d(F.sub.i,j, F.sub.k,l)=0 if and only if F.sub.i,j=F.sub.k,l

(17) d(F.sub.i,j, F.sub.k,l)=d(F.sub.k,l, F.sub.i,j) (symmetry)

(18) d(F.sub.i,j, F.sub.m,n)d(F.sub.i,j, F.sub.k,l)+d(F.sub.k,l, F.sub.m,n) (triangle inequality).

(19) The Mahalanobis distance and the .sup.2 distance between histograms are examples of metrics. Any metric may be selected. However, certain metrics are more suitable for certain selections of texture parameters. By way of non-limiting examples of embodiment, the following combinations of parameters and metrics yield satisfactory results:

(20) Haralick parameters (vector comprising n=5 texture parameters) in combination with the Euclidean distance,

(21) a local binary pattern (vector comprising n=25 texture parameters) in combination with the .sup.2 distance between histograms, and

(22) a histogram of grey levels created after a sequence of morphological transformations.

(23) The value of the texture gradient at a pixel (i,j) of the surface image 18 for a given translation vector {right arrow over (t)}=(a, b) is then the distance, in the direction of the selected metric, between two points of F with a distance of {right arrow over (t)} between them, divided by the norm of {right arrow over (t)}:

(24) $G_{i,j}^{\stackrel{.fwdarw.}{t}=\left(a,b\right)}=\frac{d\left(F_{i-\frac{a}{2},j-\frac{b}{2}},F_{i+\frac{a}{2},j+\frac{b}{2}}\right)}{\sqrt{a^2+b^2}}$

(25) In practice, the automated calculation means 12 calculate the values of two texture gradients, namely a vertical texture gradient {right arrow over (t.sub.vert)}=(t.sub.vert, 0) and a horizontal texture gradient {right arrow over (t.sub.hor)}=(0, t.sub.hor). Their norm is then deduced:

(26) $.Math.G_{i,j}^{t_{\mathrm{vert}},t_{\mathrm{hor}}}.Math.=\sqrt{{\left(G_{i,j}^{\stackrel{.fwdarw.}{t_{\mathrm{vert}}}=\left(t_{\mathrm{vert}},0\right)}\right)}^2+{\left(G_{i,j}^{\stackrel{.fwdarw.}{t_{\mathrm{hor}}}=\left(0,t_{\mathrm{hor}}\right)}\right)}^2}$

(27) The distance values t.sub.vert and t.sub.hor are selected so that the texture parameter calculation windows do not overlap, i.e. t.sub.vert=W.sub.i and t.sub.hor=W.sub.j.

(28) To optimize the calculation time, the calculation of the gradient values may be carried out, not for each pixel of the original image, but by intervals of step.sub.i pixels vertically and step.sub.j horizontally. In other words, the automated calculation means 12 calculate the values of the texture gradient in a selection of pixels of a part of the image. This is equivalent to performing a decimation on the hypermatrix F. Let {circumflex over (F)} be the sub-sample of F:

(29) ${\hat{F}}_{i,j}=F_{i{\mathrm{step}}_i,j{\mathrm{step}}_j}$$\mathrm{Then},\text{.Math.}{\hat{G}}_{i,j}^{\stackrel{.fwdarw.}{t}=\left(a,b\right)}=\frac{d\left({\hat{F}}_{i-\frac{a}{2{\mathrm{step}}_i},j-\frac{b}{2{\mathrm{step}}_j}},{\hat{F}}_{i+\frac{a}{2{\mathrm{step}}_i},j+\frac{b}{2{\mathrm{step}}_j}}\right)}{\sqrt{a^2+b^2}}$

(30) By way of a non-limiting exemplary embodiment, the parameters may be selected as follows:

(31) W.sub.i=W.sub.j=40 pixels,

(32) t.sub.vert=t.sub.hor=40 pixels, et

(33) step.sub.i=step.sub.i=5 pixels.

(34) Here, a unitary window size W.sub.i=W.sub.j=1 pixel is selected, and the automated calculation means 12 calculate a grey level gradient in the surface image 18 so that G is what is commonly called the gradient of the surface image 18.

(35) The automated calculation means 12 then determine an image of the gradient 20, using the gradient values calculated according to the algorithm described above.

(36) Finally, in step C, the automated calculation means 12 perform a thresholding of the image of the gradient 20, in order to obtain a thresholded image 22. The thresholding is performed by hysteresis. The thresholded image 22 may take the form of a black and white image, also called a binary image. Consequently, this image comprises only two types of pixel, namely black pixels and white pixels.

(37) The automated means 6, 12 of the installation 2 are therefore adapted for: calculating values of a texture gradient from an image of the surface of the tire 4, determining the image of the gradient 20, and thresholding the image of the gradient, in order to obtain the thresholded image 22.

(38) The thresholded image 22 is then submitted to an operator so that he can view the location of any defects on the inner surface of the tire 4. In the thresholded image 22 shown in FIG. 2, the white area correspond to the regions of the inner surface of the tire 4 with a high gradient, characteristic of the presence of defects, and the black areas correspond to the regions of the surface with a low gradient.

(39) On the basis of the characteristics of any defects, such as their size and number, the operator may decide if action needs to be taken in relation to the tire 4, such as the rejection or destruction of the tire. It may also be specified that the automated calculation means 12 themselves should view the thresholded image 22, and decide whether or not to take action in relation to the tire 4.

(40) Clearly, numerous modifications may be made to the invention without departing from the scope thereof.

(41) Provision may be made to use the installation for detecting a defect on a surface of an article other than a tire.