Method for quantifying burns generated by grinding a part

Abstract

The invention relates to a method (100) for quantifying burns generated by grinding a part (10), the method comprising at least: a step (110) of acquiring images Ii, i being an integer, of at least one zone (S1) of the part (10) to be observed on a background (12) by way of an apparatus (20) comprising an observation device (22) equipped with an image sensor (24) and a light source (26) able to illuminate the part (10), said acquisition apparatus (20) being configured to acquire images in N grayscale levels k, k = [0...N], N being an integer, each grayscale level being associated with a grayscale value gk, each image Ii comprising a first series of pixels associated with the part (10) and a second series of pixels associated with the background (12), a step (120) of processing the images acquired in the image acquisition step (110), and a step (130) of computing the level of surface burns ρ on the part (10).

Description

[0001] DESCRIPTION

[0002] TITLE: METHOD FOR QUANTIFYING BURNS GENERATED BY GRINDING A PART

[0003] Technical field of the invention

[0004] The invention relates to the technical field of methods for quantifying burns generated by grinding a part, in particular made of titanium or titanium alloy, by digital image analysis.

[0005] Technical background

[0006] There are many known techniques for quantifying burns on a ground part. Grinding is a machining process in which material is removed using a grinding wheel to give a part a desired shape and dimensions, thereby modifying its surface finish. This grinding process is widely used in the aeronautics industry to machine (among other things) titanium or titanium alloy parts. However, the heating generated during grinding can lead to a metallurgical change in the titanium or titanium alloy, commonly referred to as "burning".

[0007] Burn detection on titanium or titanium alloy parts is generally carried out by immersion in a nitro-hydrofluoric bath. On this type of part, grinding typically generates local changes in the material's properties, because titanium is a material with low thermal conductivity. Thus, during the grinding process of a titanium or titanium alloy part, local heating generally occurs, which has the effect of generating a phase change in the material, which is the cause of local changes in the mechanical properties of the part. These local changes in the part's properties are called burns. They are likely to weaken the part and reduce its service life. They are identifiable by the presence of dark spots on its surface after nitro-hydrofluoric etching.

[0008] Avoiding the generation of burns during grinding on titanium or titanium alloy parts is complex and not very industrial. It is therefore necessary to quantify the rate of burned surface in order to deduce the influence on the mechanical properties of the part. The quantification of these burns is essential to confirm that the requirements in terms of accepted anomaly rates are verified, and thus confirm that the condition of the part corresponds to the use to which it is to be put. This quantification is conventionally carried out by a visual analysis of the images. In the methods of quantifying burns known from the prior art, the quantification is carried out by comparison with typical images, which makes it subjective because it depends on the analysis and experience of the operator, as well as the working environment.

[0009] According to a first method well known from the prior art, the operator observes the part by means of an observation device, then compares the surface of the ground part with the typical images in order to conclude on the conformity or not of the part. This technique is however not precise because it is subject to the subjective analysis and experience of the operator. The level of acceptance of the parts can therefore be high due to the operator's lack of experience and / or lack of objectivity in the analysis technique.

[0010] A second image analysis technique was therefore developed to enable more precise quantification of burns generated by grinding the part. It consists of acquiring an image of the ground part in intensity levels on a gray scale and dividing the image into three gray intensity levels. Although this method is more precise than the method described previously, it is also subject to the analysis and experience of the operator. Indeed, it is not possible to use the same gray levels for all the images analyzed because the lighting conditions of the part and the environment in which the part is located fluctuate. Thus, a gray level can, depending on the lighting conditions and the environment of the part, be associated with a gray level characteristic of a burn as well as with a gray level characteristic of an absence of burn.The operator must therefore readjust the gray levels for each image, which makes the process of quantifying burns using this second method non-repeatable.

[0011] In a third method known from the prior art, a first step consists of acquiring a color image, i.e. in intensity levels on an RGB scale, i.e. red-green-blue, of the part with fixed acquisition parameters, a second step consists of defining an analysis zone, a third step consists of performing thresholding and a fourth step consists of filtering the zones associated with anomalies. If this method effectively takes into account burns, it also quantifies the anomalies generated by scratches and all other anomalies present on the surface. This method therefore requires the analysis of the operator with regard to the choice of the filter to be applied during image processing and is therefore not objective.

[0012] The burn quantification techniques previously stated all depend on the operator's level of expertise, and are also highly dependent on the environment in which the part is located. The paper HE BAOFENG et al. “A survey of methods for detecting metallic grinding burn”, Measurement, vol. 134, October 30, 2018 discloses a method for quantifying burns generated by grinding a part.

[0013] The invention aims to overcome at least some of the aforementioned problems and in this regard proposes a method for quantifying burns generated by grinding a part which is repeatable, precise and little, if any, dependent on the level of expertise of the operator.

[0014] Summary of the invention

[0015] The invention proposes for this purpose a method for quantifying the burns generated by grinding a part comprising at least one area to be observed, the method comprising at least:

[0016] - a step of acquiring images of the area to be observed on a background by means of an apparatus comprising an observation device equipped with an image sensor and a light source capable of illuminating the room, said acquisition apparatus being configured to acquire images in N gray levels k, k = [0...N], N integer, each gray level being associated with a gray value gk, each image comprising a first series of pixels associated with the area S1 to be observed and a second series of pixels associated with the background,

[0017] - a step of processing the images acquired during the image acquisition step,

[0018] - a step of calculating the surface burn rate p on the area to be observed, said processing and calculation steps being implemented by computer, the method being characterized in that the image acquisition step comprises:

[0019] - a first sub-step of positioning the part between the image sensor and a screen so that the image sensor, the part, and the screen are aligned in this order, said screen having a matte surface having a gray value corresponding to a reference black go and located opposite the part,

[0020] - a second sub-step of adjusting the exposure time T exp and intensity l e x P of the light source at predetermined values, as well as the magnification level G of the image,

[0021] - a third sub-step in which a gray value gsc is changed from a gray level, called target white, from a maximum gN to a first intermediate gray value gw between the reference black go and the maximum gN without including these values,

[0022] - a fourth sub-step in which a histogram is produced representing the number of pixels Qk, Q integer, as a function of the gray level k and the intensity l is adjusted e x P of the light source so that the gray value g PjC to which is associated the maximum number of pixels Q m ax detected by the acquisition device corresponds to the first intermediate gray value gki defined during the third sub-step, - a fifth sub-step in which the target white gsc is changed from the first intermediate gray value gki to the maximum gN and a first image h is acquired,

[0023] - a sixth sub-step in which the third, fourth and fifth sub-steps are repeated for second and third intermediate gray values ​​gk2, gk3 distinct from each other and distinct from the first intermediate gray value gki so as to obtain a second image I2 and a third image I3 respectively, in that the image processing step comprises a sub-step during which, for each of the first, second and third images h, eï I3, a thresholding of the first series of pixels is carried out so as to identify a third series of pixels associated with the burns, and in that the surface burn rate p on the area to be observed corresponds to the ratio R of an average of the areas of the third series of pixels associated with the burns in the first series of pixels of the first, second and third images h, I2 and I3 to an average of the areas of the pixels of the first series of pixels in the first,second and third images h, I2 and I3.,

[0024] The method according to the invention makes it possible to objectify the quantification of the burns generated by grinding the part since the calculation of the surface burn rate is based on an objective approach during image acquisition. Indeed, the lighting parameters of the part, namely the exposure time and the intensity of the light source, are fixed, so that the impact of the environment and the ambient lighting of the place where the part is located on the reflections and the rendering of the part are considerably reduced. Thus, the phenomenon of variability of the gray levels depending on the parameters of the environment observed in the methods of the prior art is considerably reduced in the method according to the invention. This makes it possible to carry out the acquisition of images on the basis of a controlled environment and therefore makes it possible to improve the repeatability of the image acquisition.

[0025] Furthermore, the acquisition of the first, second and third images at the three fixed intermediate grey values ​​makes it possible to discriminate burns from other anomalies that can be observed on the surface of the part and to retain as a basis for the calculation only the anomalies corresponding to the surface burns generated during the grinding of the part. The image acquisition step of the method according to the invention therefore sets up a precise protocol which substantially improves the reliability of the quantification of the surface burn rate. The fact of using the same intermediate grey values ​​for each part analyzed makes the method according to the invention objective because it no longer depends on the assessment and experience of the operator, unlike what is currently implemented in the methods of the prior art. In addition, the method according to the invention can be implemented with standard means.Indeed, the method uses an observation device, an image sensor, a light source and a computer. In addition, the image acquisition protocol implemented in the method according to the invention only uses common means of image acquisition and visualization software used in devices on the market since, notably, it essentially requires the creation of a histogram. It is therefore simple and quick to implement.

[0026] According to various characteristics of the invention which may be taken together or separately: the number of gray levels N is equal to 256, the maximum gN being equal to 255, the first gki, second gk2 and third gk3 intermediate gray values ​​being between 110 and 140; the first intermediate gray value gw is equal to 110, the second intermediate gray value g><2 is equal to 130 and the third intermediate gray value gk3 is equal to 140; the second processing step comprises a first masking sub-step in which: o a mask corresponding to a binary image is generated in which the pixels of the areas of interest of the first series of pixels have a first binary value equal to 1 and the rest of the pixels have a second binary value equal to 0, o the mask is applied to the first, second and third images h, I2 and I3 so as to obtain a second series of images HM, I2M and l 3Mcomprising only the areas of interest of the first series of pixels; the second processing step comprises, following the first masking sub-step, a second contrast adjustment sub-step by applying the CLAHE method to the second series of HM, I2M and l images 3M ; the thresholding sub-step is performed by applying the Otsu method; the thresholding sub-step is implemented after the second contrast adjustment sub-step during the second adjustment sub-step of the image acquisition step, the image sharpness, the image dimensions and / or the resolution of the image sensor are further adjusted; the observation device is a binocular, the image sensor is a CCD camera and the light source is a ring of LEDs; the part is a titanium or titanium alloy part machined by grinding. Brief description of the figures

[0027] Other objects, characteristics and advantages of the invention will appear more clearly in the description which follows, made with reference to the appended figures, in which:

[0028] - figure 1 is a schematic view illustrating the different stages of a method according to the invention,

[0029] - figure 2 is a schematic view illustrating the different steps of a method according to an exemplary embodiment of the invention,

[0030] - figure 3 is a schematic view illustrating the different images obtained after sub-steps of the processing step according to an example implementation of the invention,

[0031] - figure 4 is a schematic view of an installation intended to acquire images according to the acquisition step of the method according to the invention.

[0032] Detailed description of the invention

[0033] The invention relates to a method 100 for quantifying burns generated by grinding a part 10.

[0034] Definitions - general comments

[0035] Throughout the description of the present application, the term grinding designates the machining process during which local heating is generated during the removal of material by means of a grinding wheel on the part in order to give it a desired shape and dimensions. This process is likely to generate a change in the surface condition of the part which, depending on the material from which the part is made, is likely to cause local modifications in the hardness of the part. In the context of the invention, the part 10 is a part whose surface condition has been modified following a grinding process.

[0036] The part 10 for which the burns generated by grinding are to be quantified may be any part, for example an aircraft part. This part defined in a three-dimensional space is not limited by its shape, its dimensions, its color, etc. This part 10 comprises an external surface S and at least one area S1 to be observed. The area S1 to be observed may be a specific area of ​​the part 10 where it is desirable to measure the burn rate. It may also be the entire part of the part 10 that is visible from an observation device 22 and which will be better described below. An exemplary embodiment is illustrated in FIG. 4. In the context of the invention, the part 10 is preferably made of titanium or a titanium alloy. In the case of a titanium alloy, the part 10 is mainly composed of titanium and may include other metals such as aluminum, vanadium, zirconium, molybdenum, chromium, etc.

[0037] As stated in the preamble to this description, titanium is a material with low thermal conductivity. The thermal conductivity of titanium is more precisely 20 W.rrr 1 .K' 1 , at 20°C, which is more than 11 times lower than the conductivity of aluminum, for example. Titanium alloys, for their part, have a thermal conductivity generally between 10 and 30 W.m' 1 .K' 1, at 20°C. Due to their low thermal conductivities, titanium and titanium alloys respond to local heating by a phase change which is the cause of local changes in the hardness of the part. These local changes in the properties of the part 10, called "burns", are likely to weaken the part 10 and reduce its service life. They are identifiable by the presence of dark spots on the surface S of the part. The method according to the invention proposes to objectively quantify the rate of these burns.

[0038] The grinding of titanium or titanium alloy parts is carried out by means of a rotating grinding wheel removing material by abrasion. Conventionally, this grinding operation is carried out under lubrication. The surface of the raw part, i.e. the part before grinding, may be previously prepared mechanically by various methods known to those skilled in the art, such as for example milling or turning, or prepared chemically by various methods known to those skilled in the art, such as for example cleaning, degreasing, in order to avoid degradation of the properties of the finished part 10, i.e. the part after grinding, and a significant modification of its physicochemical properties.The use of organic solvents makes it possible to solubilize the majority of greases and oils present on the surface of the part, and therefore to degrease it, which promotes the good holding of the part during the grinding process.

[0039] The revelation of the burns after rectification is carried out by immersing the part in a bath of nitro-hydrofluoric acid. The surface of the part must be previously chemically prepared by various processes known to those skilled in the art, such as for example cleaning, degreasing. In the remainder of this description, part 10 designates the part after the implementation of rectification and etching with nitro-hydrofluoric acid.

[0040] Of course, as follows from the above, the invention can be implemented on any part other than a titanium or titanium alloy part that has undergone grinding. Indeed, the method can be applied to any part whose surface condition has been modified by grinding and which has burns. It should nevertheless be specified that the composition of the chemical bath for revealing the burns after grinding can be adapted according to the material from which the part is made. For example, grinding burns on case-hardened steels are revealed by Nital etching (nitric acid and alcohol).

[0041] At this point, it should be noted that in Figures 1 and 2, the steps associated with dotted line rectangles are optional within the scope of the invention.

[0042] Description of the general embodiment

[0043] With reference to Figure 1 and as already indicated previously, the invention relates to a method 100 for quantifying the burns generated by grinding a part 10. The method 100 comprises the following steps:

[0044] - a step 110 of image acquisition / ,, i integer, of the zone S1 to be observed on a background 12,

[0045] - a step 120 of processing the images acquired during the image acquisition step 110,

[0046] - a step 130 of calculating the surface burn rate p on the zone S1 to be observed.

[0047] With reference to Figure 4, the image acquisition step 110 is implemented by means of an apparatus 20 comprising the observation device 22. The observation device 22 is advantageously a binocular magnifier. The binocular magnifier makes it possible to carry out the observation and the acquisition of images. In this regard, the binocular magnifier 22 is provided with an image sensor 24 and a light source 26 capable of illuminating the part 10. The binocular magnifier 22 makes it possible to adapt the magnification, the sharpness, the size of the acquisition zone for a constant image acquisition. The acquisition zone corresponds at least to the zone S1 of the part 10 as well as a part of the environment of the part 10 likely to be, partially or entirely, captured by the image sensor 24.

[0048] The image sensor 24 is preferably a CCD camera, namely a camera using photographic sensors based on a charge coupled device. The CCD camera 24 has an acquisition frequency suitable for observing the burns present on the surface S of the part 10. A standard CCD camera capable of acquiring a few images per second is sufficient for implementing the method according to the invention. A camera capable of acquiring several thousand images per second, although more efficient, does not confer more advantages than a standard camera.

[0049] The light source 26 used to illuminate the part 10 and, more generally, the acquisition zone may be a ring of light-emitting diodes or a ring of LEDs (Light Emitting Diodes in English). It allows the brightness and reflections of the part 10 to be precisely controlled, and thus reduces the impact of the light coming from the environment in which the part 10 is located. The lighting may be carried out continuously or in synchronization with the image sensor 24. When the lighting is used continuously, this allows the frequency of the ring of LEDs to be decorrelated from that of the CCD camera 24.

[0050] Let us now return to Figure 1. Prior to the actual acquisition step 110, a calibration step 105 can be implemented. This calibration step can comprise the calibration of the parameters of the CCD cameras 24. For example, the calibration of the parameters of the CCD camera can be carried out by a white balance.

[0051] The acquisition device 20 is configured to acquire images in N gray levels k, k = [0... A / ], N integer, each gray level k being associated with a gray value gk. The gray levels k are advantageously distributed between a minimum gray value, noted go and called reference black, and a maximum gray value, noted gN and called maximum. The reference black go is so called because it corresponds to the gray level associated with the darkest hue. Conversely, the maximum gN corresponds to the gray level k associated with an absence of hue or white.

[0052] In the context of the invention, the gray values ​​gk located between the reference black go and the maximum gN are called intermediate gray values. According to an advantageous embodiment, the number of gray levels N is equal to 256, which makes it possible to carry out image processing by associating a byte with each gray level k. In this case, the gray value associated with the reference black go is equal to 0 while the gray value associated with the maximum g is equal to 255.

[0053] The image acquisition step 110 according to the invention is described in more detail in the following sections.

[0054] According to a first aspect of the method 100 according to the invention, the image acquisition step 110 comprises a first sub-step 111 of positioning the part 10 between the image sensor 24 and a screen 30 of the apparatus 20 so that the image sensor 24, the part 10, and the screen 30 are aligned in this order. Thus, from the point of view of the image sensor 24, the screen 30 is positioned in the background of the part 10, so that in practice it is identifiable in the background 12 of the images.

[0055] In this regard, in the context of the invention, the screen 30 has a matte surface 31 having a gray value corresponding to the reference black go and located opposite the part 10. The screen 30 is therefore not glossy and, in this case, it absorbs at least 70% of the visible light. Due to its matte tone, the screen 30 reflects very little of the incident light, in particular emitted by the light source 26, as will be seen below. The orientation of the surface 31, namely opposite the part 10, and its characteristics, in this case its matte tone and its gray value go, contribute to better control of the variability of the gray levels in the acquisition zone, and incidentally, in the analysis zone. This also guarantees colorimetric neutrality of the background 12 of the images acquired by means of the image sensor 24.In addition, since the surface 31 has a gray value corresponding to the reference black go, it is easy to distinguish and differentiate from the part 10 and the burns in the image obtained at the end of the image acquisition step 110, which facilitates subsequent processing during the image processing step 120.

[0056] According to a second aspect of the method 100 according to the invention, the step 110 of acquiring images comprises a second sub-step 112 of adjusting the exposure time T exp and intensity l exp of the light source 26 to predetermined values, as well as the magnification level G of the image.

[0057] The exposure time T exp and the intensity l expof the light source 26 are determined prior to the image acquisition step 110 and applied during the implementation of this step 110, so that the lighting conditions of the room 10 remain stable regardless of the illumination provided by daylight, i.e. the intensity of the daylight. This contributes to better control of the variability of the gray levels in the acquisition area, and incidentally, in the analysis area. The impact of daylight is also considerably limited thanks to the predetermined values ​​of T exp and the exp . In this regard, it can be noted that the method 100 can advantageously be implemented in a room or hall which is not exposed to daylight.

[0058] The magnification G of the image can be adjusted in order to maximize the presence of the area S1 to be observed in the acquired images and therefore that the background 12 constitutes only a tiny part of the image. Adjusting the magnification G makes it possible to adjust the content of the analysis area. The analysis area corresponds to the part of the acquisition area which is actually captured by the image sensor 24 and which, consequently, becomes an image. Although each image illustrates the area S1 to be observed, it necessarily includes a part of the background 12 representing the background surrounding the part 10. Each image therefore includes a first series of pixels associated with the part 10, in particular with the area S1 to be observed, and a second series of pixels associated with the background 12.

[0059] Preferably, each image comprises at least 60% of pixels from the first series of pixels (associated with the area S1 to be observed) and less than 40% of pixels from the second series of pixels (associated with the background 12). Even more preferably, each image comprises at least 80% of pixels from the first series of pixels and less than 20% of the second series of pixels. This makes it possible to better see the area S1 to be observed and makes it possible to better identify the burns and other anomalies present on the surface S of the part 10. According to a particular implementation, during the second sub-step 112 of adjustment of the image acquisition step 110, it is also advantageous to adjust the sharpness of the image / ,, the dimensions of the image and / or the resolution of the image sensor 24. These parameters make it possible to further improve the quality of the images obtained and make it possible to facilitate the image processing step 120 subsequent to the image acquisition step 110.

[0060] According to a third aspect of the method 100 according to the invention, the image acquisition step 110 comprises a third sub-step 113 during which a gray value gsc is changed from a gray level, called target white, from a maximum N to a first intermediate gray value gki between the reference black go and the maximum g / v without including these values. The target white is so called because it is the target gray level for which it is desired to maximize the number of pixels of the image to be acquired having this gray level in proportion to the total number of pixels contained in the image to be acquired. In other words, it is the gray level / crossing out the gray value gk with which a maximum number of pixels Qmax is to be associated. This makes it possible to avoid over-saturation of the image of the room 10 and to highlight the most interesting pixels.This will be better described with reference to the description relating to the following sub-steps of the image acquisition step.

[0061] In addition to the observation device 22 and the screen 30, the apparatus 20 also advantageously comprises a computer 40 for implementing the third sub-step 113. The target white gsc, initially associated with the maximum gray value gw, is modified so as to be associated with a gray value between the reference black go and the maximum g^ without including go and gN, called the first intermediate gray value gki. The third sub-step 113 is therefore an assignment step in which a new gray value is assigned to the target white. Preferably, the first intermediate gray value gki is between 110 and 140 for a gray level dynamic of 8 bits. Generally, this value gki will be between 40% and 60% of the average value of the maximum gray level following the encoding of the image.In the example shown here, the grayscale dynamic range is 8 bits and the maximum grayscale is 255, the intermediate grayscale value gki is for example between 110 and 140. We will come back to this later.

[0062] According to a fourth aspect of the method 100 according to the invention, the image acquisition step 110 further comprises a fourth sub-step 114 during which a histogram is produced representing the number of pixels Qk, Q integer, as a function of the gray level k and the intensity l is adjusted e x P of the light source 24 so that the gray value g p / c to which is associated the maximum number of pixels Q maxdetected by the acquisition device 20 corresponds to the first intermediate gray value gki defined during the third sub-step 113. The production of the histogram is also advantageously implemented by the computer 40. As indicated, the histogram is a function representing the number of pixels Q / < as a function of the gray level k. The histogram advantageously expresses in real time the proportion of the different gray levels k in the analysis zone, this analysis zone becoming an image as soon as it has been captured by means of the image sensor 24. The histogram can be plotted to facilitate the identification of the gray level and the gray value g P corresponding ic to which is associated the maximum number of pixels Q max but this is not mandatory. This step can be carried out by digital analysis using a suitable digital processing tool, for example the software associated with the 24 image sensor or the computer.

[0063] The operations carried out during the fourth sub-step 114 may involve a succession of observation and adjustment steps on the part of the operator. Indeed, the operator must first adjust the intensity of the light source 26 according to the difference observed between the actual gray value at which the peak g is located PiC of the histogram and the first intermediate gray value gki = gsc at which the peak g should be located p / c of the histogram. If the actual gray value at which the peak g is located PiC of the histogram is less than the first intermediate gray value gki at which the peak g should be located PiC of the histogram, the operator increases the intensity of the light source 26. On the contrary, if the actual gray value at which the peak g is located p / cof the histogram is greater than the first intermediate gray value gki at which the peak g should be located p / c of the histogram, the operator decreases the intensity of the light source 26. As long as g PjC is different from gki , the operator repeats the observation and adjustment steps.

[0064] The operator is therefore constrained by the first intermediate gray value gki = gsc to be reached, so that this step does not depend on the subjective assessment of the operator but is, on the contrary, carried out according to a well-defined protocol. Once the histogram highlights a superposition of the real gray value at which the peak g is located p / c and the first intermediate gray value gki, the fourth substep 114 is completed. This means that the maximum number of pixels Q maxdetected by the acquisition device 20 corresponds to the first intermediate gray value gki. At the end of the fourth sub-step 114, the room 10 is therefore illuminated by the light source 26 with an illumination Ei defined by the first intermediate gray value gki ■ The illumination Ei obtained is therefore set to the gray values ​​actually measured by the image sensor 24.

[0065] According to a fifth aspect of the method 100 according to the invention, the image acquisition step 110 further comprises a fifth sub-step 115 during which the target white gæ is changed from the first intermediate gray value gki to the maximum g / v. The fifth sub-step 115 thus performs the reverse of what was done during the third sub-step 113. Like the third sub-step 113, the fifth sub-step 115 can be implemented by the computer 40. The target white gsc, associated with the first intermediate gray value gki during the third sub-step 113, is modified so as to be associated with the maximum g^ The fifth sub-step 115 is therefore a reassignment step in which the target white is reassigned the gray value gN corresponding to the maximum. As mentioned previously, according to a preferred embodiment, this maximum is equal to 255.By setting the full dynamic range of the image to 8 bits, as we set the E1 lighting to gki, all the pixels in the room are correctly exposed (in the middle of the dynamic range). This allows for independence from ambient lighting.

[0066] Still during the fifth sub-step 115, once the reassignment has been carried out, a first image h of the part 10 is acquired. At this step of the method, the part 10 is illuminated by the light source 26 so that its illumination Ei corresponds to that defined by the first intermediate gray value gki ■

[0067] According to a sixth aspect of the method 100 according to the invention, the image acquisition step 110 comprises a sixth sub-step 116 during which the third 113, fourth 114 and fifth 115 sub-steps are repeated for second and third intermediate gray values ​​gk2, gk3 distinct from each other and distinct from the first intermediate gray value gw so as to obtain a second image I2 and a third image respectively. The sixth sub-step 116 is therefore actually composed of six sub-steps since the entire process carried out during the third 113, fourth 114 and fifth 115 sub-steps is implemented for each of the second and third intermediate gray values ​​gk2, gk3. At this stage, it should be recalled that the first, second and third images h, I2 and I3 do not necessarily illustrate the entire part 10 but at least the area S1 to be observed.

[0068] At the end of the fourth sub-step 114 implemented for the acquisition of the second image I2, the room 10 is illuminated by the light source 26 with an illumination E2 defined by the second intermediate gray value gk2. Preferably, the second intermediate gray value g><2 is between 110 and 140 while being distinct from the first intermediate gray value gki- Similarly, at the end of the fourth sub-step 114 implemented for the acquisition of the third image I3, the room 10 is illuminated by the light source 26 with an illumination E3 defined by the third intermediate gray value gk3. Preferably, the third intermediate gray value gk3 is between 110 and 140 while being distinct from the first intermediate gray value gki and the second intermediate gray value gk2.The first, second and third images 11, 12 and 13 are therefore obtained by a well-defined protocol which depends neither on the experience of the operator nor on the environment in which the part 10 is located, which objectifies the method 100 of quantifying the burns generated during the grinding of the part 10.

[0069] Incidentally, the step 120 of processing the first, second and third images h, I2 and I3 is no longer impacted by subjective input data but, on the contrary, by objective input data. The burns present on the part can therefore be distinguished from the other anomalies present on the surface S of the part 10, and therefore identified, which allows their quantification. The subsequent step 130 of calculating the surface burn rate p on the zone S1 of the can therefore also be implemented from reliable and objective input data.

[0070] According to a seventh aspect of the method 100 according to the invention, the image processing step 120 comprises a sub-step 123 during which, for each of the first, second and third images h, eï I3, a thresholding of the first series of pixels is carried out so as to identify a third series of pixels associated with the burns. The thresholding makes it possible to discriminate the pixels associated with the burns, i.e. the third series of pixels, from the other pixels of the first series of pixels, which makes it possible to better distinguish the burns of the surface S of the sample and other potential anomalies located on the surface of the sample.

[0071] According to a preferred implementation, this sub-step 123 is implemented by Otsu thresholding. The first, second and third images h, I2 and I3 previously mentioned are then transformed into binary images. The images are called binary because for a given gray level ki., all the pixels of the image having a gray value gci lower than the gray value of this gray level ki. are brought back to a first gray value, while, in parallel, all the pixels of the image having a gray value goe higher than the gray value of this gray level ki. are brought back to a second given gray value. Thus, the images are binarized. The selected gray level ki. makes it possible to highlight the burns by thresholding effect.In this regard and preferably, the first gray value is the reference black go while the second gray value is the maximum gN, which allows the burns to be distinguished very clearly (Figure 3, right column, fourth image from the top of the figure).

[0072] Otsu's method allows to automatically obtain a threshold to separate pixels into two classes. In this regard, a first class Ci of gray levels k is defined, with k between 0 and / ., ki_ being an arbitrarily chosen reference gray level and satisfying the condition 0 < ki_ N, and a second class C2 of gray levels k with k between ki_ and N. From there, it is convenient to assign very different gray values ​​to the pixels associated with the gray levels of the first class Ci and to the pixels associated with the gray levels of the second class C2, a gray value being assigned to each class. Otsu's method does not require parameter adjustment (automatic thresholding) and contributes to obtaining objective results.

[0073] As an alternative to the Otsu method, it is also possible to implement the watershed method to perform the 123 thresholding of the first series of pixels of the first, second and third images h, I2 and I3 and thus identify a third series of pixels associated with the burns. According to another alternative, the 123 thresholding sub-step can be implemented by the region growing method. That being said, the Otsu method has the advantage of not requiring parameter adjustment and is simpler to implement than the two aforementioned methods.

[0074] According to an eighth aspect of the method 100 according to the invention, the surface burn rate p on the zone S1 to be observed corresponds to the ratio R of an average of the areas of the third series of pixels associated with the burns in the first series of pixels of the first, second and third images h, eï on an average of the areas of the pixels of the first series of pixels in the first, second and third images h, I2 and I3.

[0075] The calculation of the surface burn rate p therefore results from the data of the three images h, and I3 acquired during the method 100 according to the invention. The surface burn rate p is therefore calculated on the basis of objectively established data and originating from the analysis of three images of the part 10 under well-defined lighting conditions. It is therefore possible to calculate and compare with each other the surface burn rates p of different zones S1 to be observed because the step 110 of acquiring images of the method 100 according to the invention is repeatable.

[0076] Particularly advantageously, the first intermediate gray value gki is equal to 110, the second intermediate gray value g^2 is equal to 130 and the third intermediate gray value is equal to 140, the number of gray levels N being equal to 256. These intermediate gray values ​​are both weakly dispersed around a central value between the reference black go and the maximum gN while being sufficiently different from each other so that the first, second and third images h, I2 and I3 are sufficiently contrasted with each other. Gray values ​​between 110 and 140 correspond to halftones to which it is easier to match g PjC during the adjustment carried out during the fourth sub-step 114 of the image acquisition step 110.

[0077] Furthermore, illumination of the part in gray values ​​gk too close to the reference black go or too close to the maximum gN would not make it possible to distinguish the burns on the part during the fifth sub-step 115. Indeed, under these lighting conditions, the part 10 would be, depending on the case, either too dark or too bright for the contrast between the burns and the rest of the part 10 to be sufficient for the image processing step 120 to be able to be implemented easily.

[0078] With reference to Figure 2, according to a particular implementation, the second processing step 120 comprises a first masking sub-step 121 in which a mask 14 is generated 1210 (Figure 3, image in the left column) corresponding to a binary image in which the pixels of the areas of interest of the first series of pixels have a first binary value equal to one and the rest of the pixels have a second binary value equal to zero, then the mask 14 is applied 1211 to the first, second and third images h, I2 and I3 so as to obtain a second series of images HM, I2M and l 3M including only the first set of pixels.

[0079] The first masking sub-step 121 makes it possible to target the area S1 to be observed in the image (Figure 3, right-hand column, first image from the top of the figure) and to extract it from the background 12 or, similarly, to target the background 12 of the image and to extract the area S1 to be observed. In either case, the first masking sub-step 121 makes it possible to separate the first series of pixels associated with the part 10 from the second series of pixels associated with the background 12. Once this separation is carried out, the image processing can continue on the first series of pixels, then the identification and quantification of the burns can be carried out as mentioned in relation to the seventh aspect of the invention. The first masking sub-step 121 therefore makes it possible to retain only the areas of interest for the quantification of the burns (Figure 3, right-hand column, second image from the top of the figure).

[0080] As previously stated, mask 14 is a binary image in which the pixels of the areas of interest in the first set of pixels have a first binary value equal to 1 and the rest of the pixels have a second binary value equal to 0. In other words, the mask is composed of 0s and 1s.

[0081] The mask 14 is generated by determining a threshold gray value gu close to the gray value go, and therefore to the reference black, in order to isolate the second series of pixels, corresponding to the pixels representing the background 12, to keep only the first series of pixels, corresponding to the pixels representing the area to be observed. Preferably, the threshold gray value gu used for the mask 14 is between 1 and 10 for a gray level dynamic of 8 bits.

[0082] Then, during step 1211, this gray threshold is applied to the first, second and third images h, I2 and I3, which makes it possible to generate, respectively, fourth, fifth and sixth images IIM, I2M and l 3M of the same size where all pixels with gray values ​​gk below the threshold are assigned the binary value 1 while the gray values ​​gk of the rest of the pixels are set to 0 (example of mask 14, the reverse would be possible). Then, we invert mask 14 (binary image) so that pixels at 1 go to 0 and represent the background, while pixels at 0 go to 1 and represent the room.

[0083] The images IIM, >2M and ISM are grayscale images obtained from the first, second and third grayscale images h, I2 and I3 and the binary mask 14. Since the first grayscale image h and the mask 14 have the same size, the pixels of the first image h superimposed / added to pixels of the mask at 1 are kept in the new image I4 and the pixels of the acquisition image superimposed / added to pixels of the mask at 0 are not kept and are set to 0 by the binary logic operators. The same applies to the second and third images I2 and I3. The image 1211 contains only pixels other than black composing the part to be viewed. These operations make it possible to separate the first series of pixels from the second series of pixels.

[0084] In this regard, although the threshold gray value gu is close to the gray value go, it is possible during step 1211 that some pixels set to the value 0 should in reality be set to the value 1 and vice versa. This can occur in particular when the sharpness of the image is not sufficient to properly identify the contours of the different zones. In particular, this concerns the pixels located in the immediate vicinity of the first series of pixels. It is therefore advantageous to define the threshold gray value gu by analyzing the pixels located near the first series of pixels, preferably at a distance not exceeding more than 5%, preferably not more than 3% of an outline of the first series of pixels. This is a tolerance.

[0085] According to a particular implementation illustrated in figure 2, the second processing step 120 comprises, following the first masking sub-step 121, a second contrast adjustment sub-step 122 aimed at processing the second series of images LM, I2M and l 3M so as to better distinguish the burns from other possible anomalies of the surface S of the part 10 and of the surface S itself. In other words, the second contrast adjustment sub-step 122 aims to highlight the contours of the burns in the images, and more precisely the contours of the burns in the first series of pixels (Figure 3, right column, third image from the top of the figure).

[0086] In this regard, a threshold pixel number Qæuii is defined 1221, with Qseuii < Qmax, then the histogram obtained during the fourth sub-step 114 of the image acquisition step is clipped 1222 from the threshold pixel number Qseuii for each gray level k having a pixel number Qk greater than the threshold pixel number Qæuii. Following this operation, a number of clipped pixels Qs is obtained which is then redistributed 1223 homogeneously over all the gray levels k so as to obtain a clipped histogram. This method, known as contrast limited adaptive histogram equalization (CLAHE), makes it possible to improve the contrast of an image without amplifying the noise in the homogeneous regions of the image. Other contrast enhancement methods could be used.

[0087] The CLAHE method differs from ordinary histogram equalization in that the adaptive method calculates multiple histograms, each corresponding to a distinct section of the image, and uses them to redistribute the brightness values ​​of the image. It can thus improve local contrast and enhance edge definition in each region of an image.

[0088] Preferably, prior to step 1221, each IIM, >2M and hM image may be subdivided into several thumbnails, and depending on the number of thumbnails, an appropriate maximum slope limiting the amplification of the contrast may be defined. The maximum slope is a parameter for preventing overamplification of noise in relatively homogeneous regions of an image. It limits the stretching of the contrast in the intensity transfer function. Very high maximum slope values ​​allow the histogram equalization to obtain maximum local contrast. The value 1 gives the original image. For example, if the number of gray levels of the histogram established in the fourth sub-step 114 of the image acquisition step is equal to 256, as in a previously mentioned example, the thumbnails may advantageously have a dimension of 40x40 pixels. In this case, a maximum slope of two may preferably be selected.As previously mentioned, other contrast enhancement methods can be implemented to adjust 122 the contrast.

[0089] At the end of this contrast adjustment step 122, each of the fourth, fifth and sixth IIM, >2M and ISM images therefore presents an improved contrast, which makes it possible to better define the outline of the burns on the surface S of the part 10, to identify and distinguish these burns from other anomalies which may possibly be present on the surface (Figure 3, right-hand column, third image from the top of the figure).

[0090] As illustrated in Figure 3, the thresholding sub-step 123 can advantageously be implemented following the second contrast adjustment sub-step 122, which makes it possible to have a more precise thresholding 123. This being the case, when neither the first masking sub-step 121 nor the second contrast adjustment sub-step 122 are implemented, the thresholding sub-step 123 is implemented after the sixth sub-step 116 of the image acquisition step.

[0091] As follows from the previous sections, the image acquisition step 110 is computer-assisted, that is to say that some of its sub-steps are implemented by the computer 40. The steps 120 of image processing and 130 of calculating the surface burn rate p on the part 10 are steps that can be fully implemented by the computer. Preferably, the computer 40 is the same as that used for the image acquisition step 110. Preferably, also, the computer 40 is equipped with a processor configured to implement the different steps and sub-steps of the method 100 implemented by computer.

[0092] In this regard, data processing software may be installed on the processor to automate the processing of the data in real time. What is important in the context of the present invention is that the processor is able to process the image data received from the image sensor 24.

[0093] The implementations shown in the figures cited are only possible examples, in no way limiting, of the invention which on the contrary encompasses design variants within the reach of those skilled in the art.

Claims

CLAIMS 1. Method (100) for quantifying burns generated by grinding a part (10) comprising an area (S1) to be observed, the method comprising at least: - a step (110) of acquiring images / ,, i integer, of the area (S1) to be observed on a background (12) by means of an apparatus (20) comprising an observation device (22) provided with an image sensor (24) and a light source (26) capable of illuminating the part (10), said acquisition apparatus (20) being configured to acquire images in A / gray levels k, k = [0... A / ], A / integer, each gray level being associated with a gray value gk, each image / / comprising a first series of pixels associated with the part (10) and a second series of pixels associated with the background (12), - a step (120) of processing the images acquired during the image acquisition step (110), - a step (130) of calculating the surface burn rate p on the area (S1) to be observed, said processing steps (120) and (130) of calculation being implemented by computer, the method (100) being characterized in that the image acquisition step (110) comprises: - a first sub-step (111) of positioning the part (10) between the image sensor (24) and a screen (30) so that the image sensor (24), the part (10), and the screen (30) are aligned in this order, said screen (30) having a matte surface (31) having a gray value corresponding to a reference black go and located opposite the part (10), - a second sub-step (112) for adjusting the exposure time T exp and the intensity lexp of the light source (26) at predetermined values, as well as the magnification level G of the image, - a third sub-step (113) in which a gray value gæ of a gray level, called target white, is changed from a maximum g to a first intermediate gray value gi between the reference black go and the maximum g without including these values, - a fourth sub-step (114) in which a histogram is produced representing the number of pixels Qk, Q integer, as a function of the gray level k and the intensity l is adjusted exp of the light source (24) so that the gray value g PiC to which is associated the maximum number of pixels Q max detected by the acquisition device corresponds to the first intermediate gray value gw defined during the third sub-step (113), - a fifth sub-step (115) in which the target blank gæ is passed from the first intermediate gray value gki to maximum gN and we acquire a first image h, - a sixth sub-step (116) in which the third (113), fourth (114) and fifth (115) sub-steps are repeated for second and third intermediate gray values gk2, gk3 distinct from each other and distinct from the first intermediate gray value gw so as to obtain a second image I2 and a third image respectively, in that the image processing step (120) comprises a sub-step (123) during which, for each of the first, second and third images h, I2 and , a thresholding of the pixels of the first series of pixels is carried out so as to identify a third series of pixels associated with the burns, and in that the surface burn rate p on the zone (S1) to be observed corresponds to the ratio R of an average of the areas of the third series of pixels associated with the burns in the first series of pixels of the first, second and third images h,I2 and I3 on an average of the pixel areas of the first series of pixels in the first, second and third images h, I2 and I3., 2. Method (100) according to claim 1, wherein the number of gray levels N is equal to 256, the maximum gN being equal to 255, the first gki, second gk2 and third gk3 intermediate gray values being between 110 and 140.

3. Method (100) according to claim 2, wherein the first intermediate gray value gki is equal to 120, the second intermediate gray value gk2 is equal to 125 and the third intermediate gray value gk3 is equal to 130.

4. Method (100) according to any one of the preceding claims, in which the second processing step (120) comprises a first masking sub-step (121) in which: - a mask (14) corresponding to a binary image is generated (1210) in which the pixels of the areas of interest of the first series of pixels have a first binary value equal to 1 and the rest of the pixels have a second binary value equal to 0, then - the mask (14) is applied (1211) to the first, second and third images h, I2 and I3 so as to obtain a second series of images HM, I2M and l 3M including only the first set of pixels.

5. Method (100) according to the preceding claim, in which the second processing step (120) comprises, following the first masking sub-step (121), a second contrast adjustment sub-step (122) by applying the CLAHE method to the second series of HM, I2M and I2M images. 3M .

6. Method (100) according to the preceding claim, in which the thresholding sub-step (123) is carried out by applying the Otsu method.

7. Method (100) according to any one of the preceding claims, in which during the second sub-step (112) of adjusting the step (110) of acquiring images, the sharpness of the image is further adjusted. the image dimensions and / or the resolution of the image sensor (24).

8. Method (100) according to any one of the preceding claims, wherein the observation device (22) is a binocular magnifier, the image sensor (24) is a CCD camera and the light source (26) is a ring of LEDs.

9. Method (100) according to any one of the preceding claims, wherein the part (10) is a titanium or titanium alloy part treated by grinding.

Images