Method for detecting glass-ceramic materials - Patents.com

Abstract

The present invention relates to an automated method (32) for detecting glass-ceramic material in cullet, comprising the following steps: detecting (33) glass-ceramic material in cullet (2), during which fragments deemed to be glass-ceramic material are identified in the cullet; acquiring a digital image resulting from the previous step of detecting glass-ceramic material, wherein the image contains, in particular, at least one group of pixels corresponding to the fragments deemed to be glass-ceramic material; and colorimetrically processing the image, during which at least one group of pixels of the image corresponding to the fragments deemed to be glass-ceramic material is processed by a colorimetric image processing module (7) according to the RGB model. The invention also claims a system for carrying out such a method (32), and a glass manufacturing facility equipped with a system of this kind.

Description

[Technical Field]

[0001] The present invention relates to an automated method for detecting glass-ceramic type materials in a sample of glass fragments or cullet. The present invention is in the field of the industry that manufactures glass-based products. [Background technology]

[0002] Glass fragments, or cullet samples, are used to manufacture various glass products. For example, glass fibers, primarily obtained from cullet, are commonly used in the field of insulation. Cullet may also be used in the manufacture of bottles and other glass containers.

[0003] Fiberglass is produced by a process that involves first heating cullet in a glass furnace to a temperature sufficient to melt the glass, i.e., about 1500° C. The molten glass is then directed into a fiberizing plate-type centrifugal device to produce fibers, which are sized in their path to a conveyor where they are then dried, cured, and shaped.

[0004] Due to user sorting errors, cullet may contain not only glass but also other materials, including glass-ceramic materials, which have properties different from those of glass and can cause significant problems in manufacturing processes for cullet-based products by damaging machinery and / or causing defects in the products.

[0005] For example, in the production of glass fibers presented above, glass-ceramic materials with a melting point of about 1700 ° C are not melted in a glass furnace, where glass is melted at 1500 ° C. When the molten material leaving the furnace is introduced into an internal centrifugal device of the fiberizing plate type to obtain glass fibers, fragments of glass-ceramic material contained in the molten glass clog the holes in the fiberizing plate, especially those with a diameter of less than 1 mm, and the entire production line must then be stopped.

[0006] When cullet-based bottles and containers are manufactured, fragments of glass-ceramic material present in the cullet can cause localized brittleness and / or aesthetic problems in the resulting product.

[0007] There are various systems or methods for detecting glass-ceramic materials in cullet in various ways. When cullet fragments are detected as glass-ceramic fragments, they are removed from the cullet by a device for discharging the given cullet fragments. The detection accuracy of this type of system or method is not perfect, and sometimes cullet fragments that are not glass-ceramic fragments are detected as glass-ceramic fragments due to similarities in their structure or visual appearance. These fragments, sometimes called false positives, are then removed from the cullet even though their properties, such as their melting point, are very suitable for producing, for example, glass fiber, hollow glass, or flat glass. Therefore, these fragments are unnecessarily rejected, which results in waste of raw materials on an industrial scale. Currently, there is no solution for improving the reliability of detecting glass-ceramic materials in cullet. Summary of the Invention [Problem to be solved by the invention]

[0008] The present invention makes it possible to correct errors that may result in the rejection of glass fragments that are not glass-ceramic fragments. [Means for solving the problem]

[0009] It consists of an automated method process for detecting glass-ceramic type materials in cullet, characterized in that it comprises the following steps: Detecting glass-ceramic material in the cullet, during which fragments are identified in the cullet that are deemed to be glass-ceramic material. Acquiring a digital image resulting from the aforementioned step of detecting glass-ceramic material, in particular the image including at least one group of pixels corresponding to the fragments identified as being glass-ceramic material. A step of colorimetrically processing the image, during which at least one group of pixels of the image corresponding to the fragments deemed to be glass-ceramic material is processed by a colorimetric image processing module according to the RGB model.

[0010] According to the invention, this colorimetric processing step makes it possible to check the quality of the detection of the glass-ceramic material detection step and to check whether the fragments considered to be glass-ceramic material indeed correspond to glass-ceramic material and not to another type of glass.

[0011] The step of detecting the glass-ceramic material in the cullet can be carried out by any method, it can be a colorimetric method, a shading method or other method, during this detection step a part of the cullet is considered to be a fragment which may be glass-ceramic material, and the rest of the method is aimed at confirming this further identification.

[0012] Thus, the step of detecting the glass-ceramic material allows for identifying portions of the cullet deemed to be fragments of the glass-ceramic material, and the step of acquiring an image allows for acquiring an image in which one or more pixels are identified as corresponding to the deemed fragments. In particular, the image of the cullet can be acquired by a camera and associated with information regarding the location of the pixels corresponding to the deemed fragments of the glass-ceramic material. Alternatively, when the step of detecting the glass-ceramic material already includes acquiring an image, acquiring a digital image identifying the deemed fragments of the glass-ceramic material can be included in the step of processing the image to form an outline of pixels corresponding to the deemed fragments of the glass-ceramic material.

[0013] In all cases, according to the present invention, it is important that the digital image is sent to a colorimetric processing module according to the RGB model, along with the identification of pixels or groups of pixels in the cullet that correspond to the presence of fragments that are considered to be glass-ceramic material.

[0014] A pixel is the basic unit used to define the quality of a digital image. In other words, it corresponds to a precise point in the image. The number of pixels in an image depends on the resolution of the image acquisition device selected and incorporated into the method. As specified above, "pixels corresponding to fragments deemed to be glass-ceramic material" is understood to mean that in the processed image, at least one region has been identified that corresponds to a fragment of glass-ceramic material in the cullet.

[0015] However, as described above, the pieces identified as glass-ceramic material may not actually be glass-ceramic material pieces. The third step of the method allows for this verification. Therefore, the digital image obtained by identifying the pixels corresponding to the pieces identified as glass-ceramic material is reprocessed by a colorimetric image processing module using the RGB model. The RGB model is a color definition model whose acronym stands for red, green, and blue. The RGB model is based on three data values ​​of brightness that define each color shade in the visible range. Each color is defined by data corresponding to the brightness of each of the three primary colors used to define this color: red data R, green data G, and blue data B. Each RGB data value ranges from 0 to 255. The progression of this step of the method is explained in more detail below.

[0016] As has been explained, according to the present invention, the colorimetric image processing module according to the RGB model converts at least one pixel of the image obtained from the detection of glass-ceramic material fragments into RGB data. The colorimetric image processing module according to the RGB model can also convert the entire processed image into RGB data.

[0017] According to one feature of the present invention, during the colorimetric image processing step, the colorimetric image processing module according to the RGB model processes only pixels corresponding to fragments identified as glass-ceramic material. The method according to the present invention is understood to enable checking the previous detection of glass-ceramic material fragments in the cullet, in the sense that it enables verification that fragments identified as glass-ceramic material are indeed fragments of glass-ceramic material. The method according to the present invention does not require a second detection step, if necessary, to detect fragments of glass-ceramic material that have evaded the detection system during the previous step. Therefore, in this context, the method is advantageously provided such that the colorimetric image processing module according to the RGB model converts into RGB data only those parts of the image that require checking the previous detection according to the present invention, i.e., verifying that the identified fragments are indeed glass-ceramic material. Focusing this inspection step only on pixels corresponding to fragments of glass-ceramic material therefore limits the duration of further colorimetric processing according to the RGB model.

[0018] According to one aspect of the present invention, the colorimetric image processing module according to the RGB model processes pixels or groups of pixels converted into RGB data by considering only the red R data and blue B data of the RGB model. As previously described, the RGB model corresponds to a triplet of data. However, the inventors have shown through various tests that only the R data and the B data are of interest during the verification of an image containing a pixel corresponding to a fragment considered to be a glass-ceramic material. This allows the colorimetric image processing module according to the RGB model to be configured to intentionally omit the calculation of the green data G and to calculate only the red R data and blue B data of the RGB model, thus calculating only two data instead of three, thereby increasing the speed of execution and therefore the efficiency of the method.

[0019] According to one feature of the present invention, the colorimetric image processing module according to the RGB model calculates the ratio B / R of blue data B to red data R for each pixel or group of pixels corresponding to cullet fragments deemed to be fragments of glass-ceramic material. The ratio B / R corresponds to the ratio of blue data to red data found in the acquired image or in cullet fragments identified as fragments of glass-ceramic material. The ratio of blue data to red data, as will be explained later, makes it possible to distinguish between glass-ceramic material and fragments that are not glass-ceramic fragments but that were deemed to be glass-ceramic fragments by an earlier detection step in the cullet (also called false positives). In view of what has been explained, it is understood that, in order to avoid calculating too many ratios later, after the analysis of the image during the first processing step, it is advisable to proceed with the conversion to RGB data only of pixels of the processed image that correspond to fragments deemed to be fragments of glass-ceramic material.

[0020] According to one aspect of the present invention, the RGB model colorimetric image processing module compares the previously calculated ratio B / R of blue data B to red data R against a threshold value. The inventors have found that by comparing this ratio with a carefully selected threshold value, glass-ceramic materials can be distinguished from false positives.

[0021] According to one aspect of the present invention, the threshold value is 0.5, which was determined by the inventors because it is believed that cullet fragments for pixels or groups of pixels having a ratio of blue data B to red data R greater than the 0.5 threshold value can be reliably identified as glass-ceramic fragments.

[0022] Among the different types of glass detected as positive, glass-ceramic material is the only type of glass other than clear glass that has a B / R ratio greater than the threshold of 0.5. Conventional clear glass also has a B / R ratio greater than 0.5 because its B / R ratio is close to 1, but conventional clear glass cannot be considered as a fragment of glass-ceramic material during the first step of the method. Therefore, glass-ceramic material is the only type of glass that meets the following two conditions: positive during the first step of the method and exhibits a B / R ratio greater than 0.5.

[0023] By analogy, it can be inferred that cullet fragments for pixels or groups of pixels having a ratio B / R of blue data B to red data R below a threshold value of 0.5 are no longer considered to be fragments of glass-ceramic material, but rather are considered false positives, and therefore do not need to be removed from the cullet.

[0024] As an example, fragments of glass used for wine or champagne bottles have optical properties close to those of glass-ceramic material, which may be considered to be fragments of glass-ceramic material during the first step of the method, resulting in a false positive. However, fragments of glass that may be considered false positives, such as fragments of glass used for wine or champagne bottles, have a ratio B / R between 0.07 and 0.23. Calculating the ratio B / R of pixels or groups of pixels corresponding to these fragments and then comparing it with a threshold of 0.5 makes it possible to eliminate any doubt about their true nature.

[0025] According to one feature of the invention, the step of detecting the glass-ceramic material is carried out by colorimetry, the term "colorimetry" referring to any kind of processing module that makes it possible to obtain, in numbered or coded data, the correspondence of the color shades appearing in the acquired image.

[0026] According to one feature of the invention, the step of detecting the glass-ceramic material is performed by colorimetric image processing according to the HSV model.

[0027] As a non-limiting example, the step of detecting the glass-ceramic material may be a method of detecting the glass-ceramic material by colorimetric processing according to an HSV model, which is performed on an image of the cullet captured by an image capture device, such as a camera.

[0028] The HSV color space (TSV in French) is an acronym that stands for Hue, Saturation, and Value. The data for each of these three terms makes it possible to specify any shade of color in the visible range.

[0029] "Hue" can be represented by a circle and is expressed by a data range of 0 to 360 degrees. Each angle represents a hue according to the table below. [Table 1]

[0030] "Saturation" is a term that ranges from 0 to 1 and reflects the idea of ​​the amount of color. Saturations closer to 0 tend to be lighter, while saturations closer to 1 result in more saturated colors.

[0031] Lightness is also a representation between 0 and 1, reflecting the idea of ​​brightness or luminance. Any data with a lightness equal to 0 is associated with the color black. The closer the lightness is to 1, the brighter the associated color is.

[0032] All different HSV data can be represented by a cone of rotation, in which all shades of visible color can be converted to an HSV representation, all corresponding to points on the area of ​​this cone of rotation. Hue is the circumference of the cone, saturation is the radius of the cone, and lightness is the height of the cone.

[0033] The HSV model colorimetric image processing module processes the image acquired by the image acquisition device by considering only one data set of the HSV model, more specifically only the data H of this HSV model. As previously described, the HSV model corresponds to a set of three data sets. However, the inventors have shown through various tests that only the hue H is the data of interest during the analysis of cullet by the present detection method. This allows the detection method to be configured to calculate only the data for the hue H of the HSV model, thus calculating only one data set instead of three, thus increasing the speed of execution and therefore the efficiency of the method.

[0034] Colorimetric processing according to the HSV model is performed by determining a single datum for each pixel or group of pixels of the image being analyzed, followed by comparison with a range of data.

[0035] The accuracy of the processing can vary depending on the requirements and / or size of the cullet. Therefore, to ensure better accuracy, image processing can be performed on each pixel of the image captured by the image acquisition device.

[0036] Processing can also be done by groups of pixels to process sets of pixels with less precision but faster processing speeds. The use of one or the other of the analysis modes depends on factors such as the size of the cullet fragments, the number of fragments, or the risk of fragments stacking on top of each other in cullet deposited on a reflective surface.

[0037] The colorimetric analysis process using the HSV model thresholds the HSV hue data H between approximately 50° and 70°, with S=1 and V=0.5. The colorimetric analysis process using the HSV model allows the cullet fragments to be considered as glass-ceramic fragments. After the light source is activated and directs its radiation toward the surface of the cullet, the reflected light is captured by an image capture device, and this same image is subjected to colorimetric analysis using the HSV model. The processed image is analyzed, and the cullet fragments may or may not exhibit a specific color. In practice, depending on the lighting conditions during the execution of the detection method, the glass-ceramic fragments will exhibit a specific color in the image captured and processed by the HSV model, which distinguishes the glass-ceramic fragments from the rest of the cullet. Theoretically, during the execution of the detection method on the cullet, the glass-ceramic fragments will exhibit their own unique color.

[0038] The thresholding process for the colorimetric analysis is based on the color reflected by the glass-ceramic material when exposed to light emitted simultaneously by two types of light sources. Due to the presence of metal oxides in its composition, the glass-ceramic material absorbs some of the ultraviolet light it receives. These metal oxides absorb ultraviolet light, while the corresponding crystals scatter blue light; this is why glass-ceramic materials are the only type of glass that theoretically exhibits a yellow color when cullet is analyzed during the detection method, hence the reason for this threshold selection. Therefore, after the colorimetric analysis process, the calculation module can be configured to consider any pixel or group of pixels with a hue H between 50° and 70° to be glass-ceramic material.

[0039] The present invention also covers a system for carrying out the detection method described above, comprising a module for detecting glass-ceramic materials and a colorimetric image processing module according to the RGB model. Like the method, the system for carrying out the method comprises a module for detecting glass-ceramic materials, which can be of any nature. The colorimetric image processing module according to the RGB model performs the step of verifying the positively detected fragments in the same manner as described above, regardless of the nature of the module for detecting glass-ceramic materials.

[0040] The present invention also relates to an installation for producing glass fibers, hollow glass or flat glass, comprising at least one glass furnace and a forming station, in which cullet is poured into the glass furnace to obtain molten glass intended to supply the forming station, the installation comprising a system for implementing the detection method described above, the implementation system being located on the path of the cullet towards the glass furnace.

[0041] Other characteristics and advantages of the invention will become apparent from both the following description and from several exemplary embodiments, given for purposes of illustration and not limitation with reference to the accompanying schematic drawings. [Brief explanation of the drawings]

[0042] [Figure 1] 1 is a schematic diagram of a detection system according to one embodiment of the present invention. [Figure 2] 1 is a schematic diagram of the behavior of the illumination of the light source of the system on a piece of ordinary glass and on a piece of glass-ceramic material. [Figure 3] FIG. 2 is a schematic diagram of the detection method performed by the system described above. [Figure 4] 1 is a graph showing the percentage of light transmission for different types of glass as a function of the wavelength of the light received. DETAILED DESCRIPTION OF THE INVENTION

[0043] As a non-exhaustive example, the detection system taken as an example here comprises a colorimetric processing module according to the HSV model, which detects glass-ceramic materials.

[0044] 1 represents a detection system 1 according to the invention. This system 1 comprises two types of light sources, a first type corresponding to a white light source 3 and a second type corresponding to a monochromatic ultraviolet light source 4. The white light source(s) 3 and the ultraviolet light source(s) 4 are mounted on a structure 13 which supports the light sources and comprises means for supplying electrical energy. An image acquisition device 5 protrudes above the structure 13. In the example shown, the image acquisition device 5 is supported by mounting means 36, but it may also be directly integrated into the structure 13.

[0045] The light beam emitted by the white light source 3 is directly filtered by a polarizer 16 to limit the light intensity and avoid stray reflections, and the light beam can then be processed. The light beam emitted by the monochromatic ultraviolet light source 4 is in turn filtered by a bandpass filter 15 centered at 365 nm to obtain purely monochromatic ultraviolet illumination. In the example shown, the light source and filter arrangement are in the same position, on opposite sides of the structure 13, to obtain a uniform overall illumination.

[0046] The white light source 3 and the monochromatic ultraviolet light source 4 emit light rays 12 in the direction of an emission area 11. The emission area 11 corresponds to the capture range of the image acquisition device 5. The light rays 12 emitted by the white light source 3 and the monochromatic ultraviolet light source 4 are projected onto a reflective surface 10 on which the cullet 2 is placed.

[0047] In the example shown, the reflective surface 10 is arranged on a conveyor 8 moving in a direction 9. As shown in Figure 2, the light rays emitted by the white light source 3 and by the monochromatic ultraviolet light source 4 have a behavior that may change when they reach the emission zone where the cullets 2 are placed.

[0048] Light rays 14 are reflected by surface 10 towards image capture device 5. These light rays 14 are filtered by analyzer 18, which is crossed with polarizer 16. The combination of polarizer 16 and analyzer 18 is intended to limit stray reflections caused by light emitted by white light source 3, which can make image processing difficult to interpret. Image capture device 5 captures an image of cullet 2 by light rays 14 reflected by reflective surface 10.

[0049] After the image is acquired, it is processed by an image processing device 6 electrically connected to the image acquisition device 5. The image processing device 6 comprises a module 17 for detecting glass-ceramic materials. In the non-exhaustive example shown here, the HSV model colorimetric image processing module is configured to analyze the pixels of the image pixel by pixel or pixel by pixel group and convert these pixels into HSV data. The HSV model colorimetric image processing module is configured to determine hue data H for each pixel or pixel group with fixed saturation S and lightness V data, where these fixed data are equal to S=1 and V=0.5.

[0050] Each determined hue data H is then compared to at least one threshold data. For example, the hue data H is compared to a minimum threshold of 50° and a maximum threshold of 70°. In other words, if a pixel or set of pixels has HSV data equal to 50°≦H≦70°, where S=1 and V=0.5, this pixel or set of pixels is considered to correspond to a piece of glass-ceramic material. This information is communicated to a user of the system 1 so that the user can manually intervene to remove the glass-ceramic piece from the cullet, or to an automated device that allows the piece of glass-ceramic material to be ejected in a targeted manner.

[0051] Alternatively, further colorimetric processing can be performed. The HSV data of pixels corresponding to the fragments identified as glass-ceramic material by the HSV-model colorimetric image processing module are then converted into RGB data by an RGB-model colorimetric image processing module 7 included in the image processing device 6, which calculates the ratio B / R of the blue data B to the red data R of the collected RGB data in order to detect potential false positives resulting from the processing performed by the module 17 for detecting glass-ceramic material and improve the accuracy of the detection system 1. According to that embodiment, the RGB-model colorimetric processing module 7 can also convert the entire image acquired by the image acquisition device 5. Furthermore, the image is directly transmitted to the RGB-model colorimetric processing module 7.

[0052] Figure 2 is a schematic diagram of the projection of light from a light source onto two different types of glass. For schematic simplicity, only two light rays from each light source are shown; in reality, the light source will emit in multiple directions, e.g., at an emission angle of 90°. Furthermore, again for simplicity, the refraction of light rays onto the glass fragments is not shown.

[0053] 2 shows a white light source 3 and a monochromatic ultraviolet light source 4, each emitting its respective light radiation. The white light source 3 emits a white light ray 26, shown by a solid line, and the monochromatic ultraviolet light source 4 emits an ultraviolet light ray 25 having a wavelength of 365 nm, shown by a dotted line. The two light rays intersect in an emission zone 11, where a cullet placed on a reflective surface 10 is present. Here, ordinary glass shards 23 and glass-ceramic shards 24 are present in the cullet.

[0054] Ordinary glass fragment 23 allows all types of light to pass through, so white light 26 and ultraviolet light 25 pass through ordinary glass fragment 23, are reflected by reflective surface 10, and pass again through the structure of ordinary glass fragment 23 towards image capture device 5.

[0055] The glass-ceramic shard 24 exhibits different optical properties than the ordinary glass shard 23. As will be explained in more detail with reference to FIG. 4, glass-ceramic materials have the property of absorbing much of the ultraviolet light. Therefore, not all of the ultraviolet light 25 passes through the structure of the glass-ceramic shard 24, but rather is largely absorbed. Generally, ultraviolet light having a wavelength of 365 nm is hardly reflected back toward the image capture device 5 when its trajectory passes through a shard of glass-ceramic material. As with the ordinary glass shard 23, the white light 26 passes through the glass-ceramic shard 24 and is reflected back toward the image capture device 5.

[0056] Theoretically, the image capture device 5 receives all the white light 26 from the white light source 3 and the ultraviolet light 25 from the monochromatic ultraviolet light source 4, excluding the ultraviolet light 25 that has been significantly absorbed by one or more glass-ceramic fragments 24. This dual illumination and the partial attenuation of the ultraviolet light by the glass-ceramic material results in a possible colorimetric analysis of the captured image, since the glass-ceramic material, unlike the rest of the cullet, exhibits a yellow hue.

[0057] FIG. 3 is a schematic diagram of the detection method performed by the system described above.

[0058] The step 27 of scrolling the conveyor at a given speed involves moving one or more cullets onto their reflective surface or directly onto the reflective mat of the conveyor.

[0059] The conveyor scroll 27 results in the step of placing the cullet in the radiation area 29. Simultaneously, the step of emitting light radiation from a light source 28 is initiated to illuminate the cullet present in the radiation area. The combination of the step of placing the cullet in the radiation area 29 and the step of emitting light radiation from a light source 28 leads to the step of capturing an image by an image capture device 30.

[0060] Once the image is captured, the detection method 32 begins. The captured image is subjected to a glass-ceramic material detection step 33, which is a colorimetric processing of the image, for example, according to the HSV model. The glass-ceramic material detection step 33 analyzes the image acquired during the image acquisition step 30 by applying, for each pixel or group of pixels, a threshold value of H between 50° and 70°, such that 50°≦H≦70° for the specified data S and V. If no pixel or group of pixels falls within this threshold, the detection method 32 ends and a new cycle begins again with a new caret scrolling on the conveyor.

[0061] If one or more pixels, or one or more groups of pixels, fall within this threshold of H, then theoretically the shards corresponding to these pixels or groups of pixels are considered to be glass-ceramic shards. The method therefore continues and ends with step 34, where the image is colorimetrically processed according to the RGB model.

[0062] For this purpose, it is suitable to apply a colorimetric image processing module 7 of the RGB model to the processed digital image acquired from the previous step of detecting glass-ceramic material, in particular to a digital image containing at least one group of pixels corresponding to fragments considered to be glass-ceramic material. In the described example, in which the detection step is performed by colorimetric analysis of the image acquired from the cullet, the image is processed by combining the acquired image with a digital marking of pixels or groups of pixels corresponding to what can be considered to be glass-ceramic material. In examples not described but included in the present invention, in which the detection of glass-ceramic material is performed by other methods, a digital image of the cullet is acquired at this stage and processed to identify what can be considered to be glass-ceramic material on the image.

[0063] During the step 34 of colorimetrically processing an image according to the RGB model, several successive substeps follow one another. First, the image is processed according to the RGB model; this can be either the entire image or only the pixels deemed to be glass-ceramic material, i.e., the pixels corresponding to the thresholding of the data H of the HSV model. This is substep 36, which converts pixels or groups of pixels of the image into RGB data. For each of these pixels, red data R and blue data B are collected, with the values ​​ranging from 0 to 255. The colorimetric image processing module according to the RGB model then calculates the ratio B / R in substep 37, which corresponds to the blue data B of a pixel or group of pixels divided by the red data R of the same pixel or group of pixels. Finally, the ratio obtained for each pixel or group of pixels is compared with a threshold value of 0.5 during substep 38 of comparing the ratios. The B / R threshold is set to be greater than 0.5. In other words, if the ratio B / R of one or more pixels or groups of one or more pixels is greater than this threshold of 0.5, the corresponding fragment is confirmed to be a glass-ceramic fragment, and if this same ratio is less than the threshold of 0.5, the corresponding fragment is identified as a false positive and not a glass-ceramic fragment.

[0064] All of this information is then transmitted to a tracking calculation step 31. During the tracking calculation step 31, target fragments are targeted, i.e., fragments that are deemed to be glass-ceramic fragments according to the step 33 of detecting the glass-ceramic material of the cullet and that have been confirmed as glass-ceramic fragments during the step 34 of colorimetrically processing the image according to the RGB model. These target fragments are tracked by taking into account the speed of the cullet movement, which depends on the speed of the conveyor. Target fragments are understood to be targeted in that, from their known position at time t, the system can accurately determine their position at time t+Δt.

[0065] Once these target fragments are targeted during the tracking calculation step 31, the next step is to remove the glass ceramic fragments 35, which are shown in dotted lines in the figure because they are external to the detection system. The glass ceramic fragment removal step 35 can be performed by a blower device located downstream of the detection system, along the conveyor. Thanks to the tracking calculation step 31, the blower device receives position information from the calculation module and is activated at the location where the glass ceramic fragments are located and at the appropriate time. The glass ceramic fragments are then ejected from the cullet.

[0066] 4 is a graph showing the percentage of light transmission through different types of glass as a function of the wavelength of the emitted light. Four curves corresponding to four different types of glass can be seen on this graph: curve 19 corresponding to ordinary, i.e., most common, glass, curve 20 corresponding to a glass-ceramic material, curve 21 corresponding to a bottle glass commonly used to make wine bottles, and curve 22 corresponding to a champagne glass commonly used to make champagne bottles. On the abscissa of the graph, the region below 400 nm corresponds to the ultraviolet range, while the region above 400 nm corresponds to the visible range.

[0067] Curve 19 for ordinary glass and curve 20 for glass-ceramic material have similar behavior, i.e., a strong increase in the percentage of light transmission until a level of 85-90% transmission is reached. The main difference between these two curves is that curve 19 for ordinary glass increases sharply at shorter wavelengths than curve 20 for glass-ceramic material. Therefore, ordinary glass has a significantly higher transmittance in the ultraviolet range than glass-ceramic material. This difference is the basis for the use of a monochromatic ultraviolet light source with a wavelength of 365 nm, because at this value, the percentage of light transmission for ordinary glass is greater than 80% (point P1 in FIG. 4), while the percentage of light transmission for glass-ceramic material is less than 20% (point P2 in FIG. 4). Thus, the image capture device captures 365 nm ultraviolet light and the entire visible spectrum of white light (e.g., 550 nm) as they pass through a piece of ordinary glass, but not all of the ultraviolet light striking the glass-ceramic piece, because it is largely absorbed by the glass-ceramic material. The illumination conditions of the detection system are such that the glass-ceramic material, through its optical properties, exhibits a yellowish hue corresponding to the selected HSV threshold data, i.e., 50°≦H≦70°. The color hue corresponding to the glass-ceramic material depends on several factors, particularly the light source and the type of image capture device.

[0068] As explained above, this yellow color is explained by the presence of metal oxides in the composition of the glass-ceramic material. The other two curves, namely, the bottle glass curve 21 and the champagne glass curve 22, also behave similarly to each other. These are two types of glass that transmit variable light waves and barely exceed 50% light transmission (point P3 in Figure 4). With regard to ultraviolet light, the bottle glass and the champagne glass have a similar light transmission rate to that of glass-ceramic materials, especially at 365 nm, which corresponds to the irradiation wavelength of a monochromatic ultraviolet light source. Therefore, like glass-ceramic materials, the bottle glass and the champagne glass absorb the ultraviolet light emitted by a monochromatic ultraviolet light source. Furthermore, the wavelength at which these two types of glass transmit light most effectively is approximately 550-570 nm. In the visible spectrum, this wavelength range corresponds to yellow-green.

[0069] In summary, bottle glass and champagne glass have substantially the same UV absorption characteristics as glass-ceramic materials, and their best light transmission ratio corresponds to a yellow-green hue that is very similar to the threshold hue of glass-ceramic materials when the image acquired by the image acquisition device is processed by the colorimetric image processing module using the HSV model. Therefore, bottle glass and champagne glass are two types of glass that are prone to false positives, i.e., they are easily deemed to be fragments of glass-ceramic material when they are not fragments of glass-ceramic material during analysis by the colorimetric image processing module using the HSV model. The advantage of the RGB model colorimetric image processing module becomes fully apparent when the ratio B / R of bottle glass and champagne glass is less than 0.5, which can deny that they are fragments of glass-ceramic material.

Claims

1. A method (32) for the automated detection of glass-ceramic type materials among cullets (2), comprising: detecting (33) the glass-ceramic material in the cullet (2), during which fragments considered to be glass-ceramic material are identified in said cullet (33); acquiring a digital image resulting from the aforementioned step of detecting glass-ceramic material, in particular said image comprising at least one group of pixels corresponding to the fragments identified as being glass-ceramic material; a colorimetric image processing step (34) of said image, during which at least one group of pixels of said image corresponding to the fragments considered to be glass-ceramic material is processed by a colorimetric image processing module (7) according to the RGB model; A method of detection (32), comprising:

2. 2. The detection method (32) according to claim 1, wherein during the colorimetric image processing step (34), the RGB model colorimetric image processing module (7) processes only those pixels corresponding to fragments that are considered to be glass-ceramic material.

3. 3. The detection method (32) according to claim 1 or 2, wherein the RGB model colorimetric image processing module processes the pixel or group of pixels converted into RGB data by considering only red data R and blue data B of the RGB model.

4. 4. The detection method (32) according to any one of claims 1 to 3, wherein the RGB model colorimetric image processing module (7) calculates the ratio B / R of blue data B to red data R for each pixel or group of pixels corresponding to cullet fragments considered to be fragments of glass-ceramic material.

5. 5. The detection method (32) according to claim 4, wherein the RGB model colorimetric image processing module (7) compares the ratio B / R of the blue data B to the red data R calculated previously with a threshold value.

6. 6. The detection method (32) of claim 5, wherein the threshold is 0.

5.

7. The detection method (32) according to any one of claims 1 to 6, wherein the step (33) of detecting the glass-ceramic material is carried out by colorimetry.

8. 8. The method (32) of claim 7, wherein the step (33) of detecting the glass-ceramic material is performed by colorimetric image processing according to the HSV model.

9. 9. A system (1) for carrying out the detection method (32) according to any one of claims 1 to 8, comprising a module (17) for detecting glass-ceramic materials and a colorimetric image processing module (7) according to the RGB model.

10. 10. An installation for producing glass fibers, hollow glass or flat glass, comprising at least one glass furnace and a forming station, in which cullet is poured into the glass furnace to obtain molten glass intended to supply the forming station, the installation comprising a system (1) for performing the detection method (32) according to claim 9, the execution system (1) being located in the path of the cullet towards the glass furnace.

Images