JP2020003358A - Steel material component identification device and program therefor - Google Patents

Abstract

To provide a device capable of identifying steel material components by performing image processing to a spark generated by a spark test of steel.SOLUTION: A steel material component identification device includes: a humidity detection section; a luminance conversion processing section for converting an image of a spark into a gray scale spark image; a binarization processing section for performing binarization with a predetermined threshold for each of pixels of the gray scale spark image; a short straight line matching processing section for matching each of a plurality of templates showing a short straight line which is equivalent to a plurality of angles with the binarized spark image and for storing types and positions of the matched templates; a bursting section extraction processing section for, when the number of the matched templates is equal to or more than a predetermined number in a range of an arbitrary spark image, recognizing it as a bursting section of the spark and extracting the bursting section, and for counting the number of bursting as the total number of the bursting sections in the entire spark image and the number of short straight lines as the total number of the matched templates in the entire spark image; and a steel material identification processing section for identifying steel materials on the basis of bursting density obtained by dividing the number of bursting by the number of short straight lines, and humidity.SELECTED DRAWING: Figure 2

Description

  The present invention relates to a steel component identification device and a program therefor.

  In the steel manufacturing process, a spark test is widely used to discriminate and eliminate different materials. The spark test is a test for estimating steel type or discriminating dissimilar materials by grinding steel ingots, billets, steel materials and other steel products using a grinder and observing the characteristics of the generated sparks. , JIS G 0566 (for example, see Non-Patent Document 1).

  FIG. 14 is a diagram showing the shapes and names of steel sparks. As shown in the figure, the spark is divided into "root", "center", and "tip" from its position, and the shape and density of streamlines and ruptures change in each part of the spark.

  Conventionally, the spark test has been performed as a sensory test by visual inspection by an inspector with experience in the steel material inspection process, etc., but the judgment result varies due to individual differences and environmental fluctuations, and the appropriate test result It was difficult to obtain. In addition, since the result of the sensory test cannot be recorded as a necessity of the sensory test, the test technique is largely based on experience or handing down, and it is difficult to evaluate the technical improvement. It is said that the error of the carbon component weight ratio [C] value by a person is about 0.20 to 0.50%.

  As a technique for automatically performing such a spark test by a device without visual observation, there is a device that identifies a steel material component by performing image processing on a spark generated by a steel spark test (for example, see Patent Document 1). ).

JP 2012-247206 A

Japanese Industrial Standards, JIS G 0566 Steel Spark Testing Method, Japanese Standards Association

  However, in the method of Patent Document 1, the obtained burst density may vary. This is considered to be due to some kind of on-site environment.

  The present invention has been made in order to solve such a conventional problem, and in a method of performing image processing of a spark generated by a spark test of steel to identify components of a steel material, the influence of a variation in burst density is reduced. It is to provide a device and a method that are possible.

  The present invention provides a humidity detection unit that detects the humidity of the surrounding environment, a brightness conversion processing unit that converts an image of a spark generated by grinding a steel material into a grayscale spark image, and a pixel of the grayscale spark image. A binarization processing unit for performing binarization at a predetermined threshold value, and a plurality of templates representing short straight lines corresponding to a plurality of angles with respect to the binarized spark image, and a type of the matched template And a short straight line matching processing unit that stores the position, and when the number of matched templates is equal to or more than a predetermined number in the range of an arbitrary spark image, the rupture portion is regarded as a rupture portion of the spark, and the rupture portion is extracted. The number of ruptures, which is the total number of ruptures, and the rupture part extraction processing unit that counts the number of short straight lines, which is the total number of matched templates of the entire spark image, A rupture densities, a humidity and, steel component identifying apparatus and a steel identification processor identifies a steel based on.

  Further, the present invention provides a computer, a humidity capture unit for acquiring the humidity of the surrounding environment, a brightness conversion processing unit for converting a spark image generated by grinding a steel material into a grayscale spark image, and a grayscale Binarization processing means for performing binarization at a predetermined threshold value for each pixel of the spark image, and for the binarized spark image, matching a plurality of templates representing short lines corresponding to a plurality of angles, A short straight line matching processing means for storing the type and position of the matched template; and extracting a rupture portion by regarding the rupture portion of the spark as a rupture portion of the spark when the number of the matched templates is equal to or more than a predetermined number in the range of any spark image Rupture number extraction, which counts the number of ruptures, which is the total number of ruptures in the entire spark image, and the number of short straight lines, which is the total number of the matched templates in the entire spark image. Processing means, a rupturable density obtained by dividing the number of rupture in a short number of straight lines, a steel component identification program to function as a steel material identification processing means for identifying a steel based on the humidity.

  ADVANTAGE OF THE INVENTION According to the steel material component identification apparatus and its method of this invention, it becomes possible to recognize the kind of steel material accurately by correcting the burst density with the humidity of the surrounding environment.

It is a figure showing composition of a steel material constituent discriminating device of this embodiment. It is a flowchart which shows the steel material component identification routine of this embodiment. It is a figure which shows the spark image processed by the steel component identification device of this embodiment. FIG. 4 is a diagram illustrating cross-binarization according to the present embodiment. It is a figure showing the template showing the short straight line used by the steel material constituent identification device of this embodiment. It is a figure showing the method of short straight line matching processing of this embodiment. It is a figure showing the range which extracts a rupture part of this embodiment. It is a figure which shows the relationship between burst density and humidity. It is a figure showing composition of a steel material constituent discriminating device of this embodiment. It is a flowchart which shows the steel material component identification routine of this embodiment. It is a figure showing the method of short straight line matching processing of this embodiment. It is a figure showing the method of the short straight line matching processing using a plurality of templates of this embodiment. It is a figure showing the range which extracts a rupture part of this embodiment. It is a figure which shows the shape and the name of a steel spark.

  Hereinafter, a steel component identification device and a method thereof according to a first embodiment of the present invention will be described in detail with reference to the drawings.

  FIG. 1 is a block diagram showing the configuration of the steel component identification device according to the first embodiment of the present invention. The steel material component identification device according to the present embodiment includes a camera 11 that captures an image of a spark generated when a steel B to be inspected comes into contact with a grinder C, a humidity detector 12 that measures the humidity of the surrounding environment, and an image captured by the camera 11. And a computer 13 for quantifying the spark based on the detected spark image and the humidity measured by the humidity detector 12. The computer 13 includes a CPU 131 that performs arithmetic processing, a RAM 132 that is a work area for data, and a ROM 133 that stores a control program of the CPU 131. The steel material component identification device of the present embodiment configured as described above quantifies the spark of the steel material by performing the following processing.

  FIG. 2 is a flowchart illustrating a steel component identification routine executed by the CPU 131 of the computer 13. The steel material component identification device according to the present embodiment includes a humidity acquisition processing unit, an image acquisition processing unit, a luminance conversion processing unit, a binarization processing unit, and a CPU 131 executing a control program stored in a ROM 133. , A thin line processing unit, a short straight line matching processing unit, a rupture part extraction processing unit, and a steel material identification processing unit.

  Hereinafter, the detailed processing of each of steps S11 to S18 will be described.

(Humidity intake)
In step S11, the CPU 131 of the computer 13 loads the humidity of the surrounding environment measured by the humidity detector 12 into the RAM 132, and proceeds to step S12.

(Image capture)
In step S12, the CPU 131 of the computer 13 loads the spark image captured by the camera 11 into the RAM 132 as a color image, and proceeds to step S13.

  In order to quantitatively classify steel materials, it is preferable to always generate sparks under the same conditions so that sparks of the same type of steel do not differ. In the present embodiment, a mechanism for pressing the steel material B against the grinder C by a mechanical element (not shown) such as a constant load spring or a tension spring is employed. By using mechanical elements such as a constant load spring and a tension spring, it is possible to press the steel material at a constant position and at a constant angle with a constant force.

  Since the shooting environment is a special environment for shooting a bright spark in an environment where the surroundings are dim, it is preferable to use a camera that can manually set the shutter speed, aperture, and the like as the camera 11. Further, since the steel material B is cut into the grinder C, there is a problem that the contact area between the steel material B and the grinder C increases, the pressing pressure decreases, and the generation of spark changes with time. In the camera 11 of the present embodiment, a camera capable of performing high-speed continuous shooting (for example, 60 fps) in a short time is used in order to reduce the influence of a time change of the occurrence of a spark.

  Similarly, it is preferable to shorten the photographing time (for example, 1.0 second or less) in order to avoid a change in spark due to the passage of time. Further, the streamline length and the number of bursts of the spark that can be confirmed in the image change depending on the shutter speed. Therefore, it is preferable to select a shutter speed suitable for analysis (for example, 1 / 160s).

(Brightness conversion)
In step S13, the CPU 131 converts the brightness of each pixel of the spark image to convert the image into a grayscale image having only the brightness value, stores the image in the RAM 132, and proceeds to step S14.

  FIG. 3A is an example of a grayscale image of a spark image. In step S2, the color spark image obtained in step S1 is converted to grayscale. In the present embodiment, the RGB values of each pixel in the spark image are converted to a gray scale value Y by a predetermined conversion formula. For example, Y = R × 0.299 + G × 0.587 + B × 0.114 is used as the conversion formula.

(Binarization processing)
In step S14, the CPU 131 executes a binarization process of the spark image, stores the image in the RAM 132, and proceeds to step S15.

  In the present embodiment, the binarization processing is performed at a predetermined threshold for each pixel to be subjected to the binarization processing. In the present embodiment, cross binarization suitable for the case where the extraction target is a linear shape such as a spark is used.

  FIG. 4 is a diagram illustrating the cross binarization used in step S14 of the present embodiment. The cross-binarization method subtracts the brightness value of the center pixel of the cross from the average brightness value of the pixels other than the center pixel in the cross, and if the value is greater than a set threshold, sets the center pixel of the cross to black ( FIG. 4 (a)), otherwise it is white (FIG. 4 (b)). Furthermore, in order to make the spark a black pixel, the black and white output of the cross-binarized image is inverted. FIG. 3B shows an example of an image obtained by performing cross-binarization and inverting black and white.

(Thinning processing)
In step S15, the CPU 131 performs a spark image thinning process, stores the image in the RAM 132, and proceeds to step S16.

  FIG. 3D shows an example of an image subjected to the thinning processing. The thinning process is a process of extracting a core line which is the center line of a spark. In the present embodiment, thinning processing is performed on the image (FIGS. 3B and 3C) that has been subjected to the cross-binarization processing in step S14. In the thinning process according to the present embodiment, black pixels are cut from the outside so that the thickness of the spark becomes 1 pixel, and no further cut is performed when the line width becomes 1 pixel. Then, the end points are saved, and the connectivity of the figures is saved. By this processing, the core line (center line) of the spark is obtained, and the characteristics of the streamline shape of the spark are simply represented.

(Short straight line matching)
In step S16, the CPU 131 performs a short straight line matching process on the spark image, stores the type and position of the matched template in the RAM 132, and proceeds to step S17.

  As shown in FIG. 5, in the steel material component identification device of the present embodiment, templates T1 to T28 representing short straight lines corresponding to angles in 28 directions (every 6.5 degrees) are prepared in advance and stored in the RAM 132.

  FIG. 6 is a diagram illustrating a method of the short straight line matching process. Scanning of pattern numbers (T1 to T28) is performed on the thin line groups a to d in the thin line image I5, and a pattern template indicating an angle is matched. When P (for example, 70) [%] or more is matched with a certain template, the pattern number is recorded. In the figure, five patterns of pattern templates T14, T22, T4, T8, and T26 are respectively matched to portions a, b, c, d, and e in the thinned image I5.

(Rupture extraction process)
In step S17, the CPU 131 extracts a burst portion of the spark from the spark image and counts the number of burst sparks. Then, the number of rupture portions of the entire image and the total number of short straight lines existing in each rupture portion are stored in the RAM 132, and the process proceeds to step S18.

  The rupture of the spark has a morphological feature that the spark flies in various directions. Therefore, it is possible to extract the rupture portion by checking the direction in which the spark flies from the angle of the short straight line. In the present embodiment, a case where there are, for example, four or more types of short straight lines having different angles is regarded as a rupture portion.

  As shown in FIG. 7, the total of the types of pattern numbers is obtained for an arbitrary spark image range (a × b [pixel]) (12 × 12) in the image I6, and the total value is m (this time) 4) If more than one type, the range is defined as the range in which rupture exists. In the figure, since four types of patterns T14, T22, T26 and T4 are included, it can be recognized as a rupture. Then, the total number of ruptures is obtained by counting the number of ruptures in the entire image. At the same time, the total number of short straight lines existing in the entire image is counted from the total number of matched templates.

(Steel type discrimination processing)
In step S18, the CPU 131 calculates the burst density of the spark, estimates the amount of carbon in the steel from the burst density of the spark and the humidity, and recognizes the type of steel.

  As a result of the evaluation by the present inventors, it has been found that even when the amount of carbon contained in the steel changes, the number of bursts is almost equal. Therefore, it is difficult to determine the carbon content only from the number of bursting sparks. On the other hand, as a result of the evaluation by the present inventors, it was also found that, when the amount of carbon contained in the steel was small, the amount of spark increased conversely.

Therefore, in this embodiment, an evaluation criterion called “burst density” is defined. In the present embodiment, the burst density is represented by the following equation, and the numerical value of the burst density confirms the amount of carbon in the steel material.
Burst density (D) = number of bursts (E) / number of short straight lines (L)

  Further, since the form of the spark is three-dimensional and accompanied by a dynamic and unstable form change such as “generation”, “growth”, “burst” and “extinction”, the base of the spark shown in FIG. There is a difference in characteristics such as brightness and density of the spark at each part of the part and the tip part. As described in Patent Literature 1, when feature values are simultaneously extracted from the entirety of a single spark image at a specific point in time, it may be difficult to perform imaging or image processing while ensuring effective accuracy of component analysis.

  Therefore, in the present embodiment, the burst densities are calculated from a plurality of spark images, respectively, and the average burst density is used to confirm the carbon content in the steel material. This makes it possible to reduce the prediction error as compared with the case where the carbon amount is confirmed only by the burst density obtained from one spark image.

  As an image acquisition method, when acquiring a plurality of spark images, acquiring at predetermined time intervals, or acquiring a plurality of spark images continuously after a predetermined time has elapsed after the occurrence of a spark, Such a method may be used. In the present embodiment, the images to be evaluated are evaluated by averaging 10 consecutive images of one steel material four times.

  FIG. 8 is a diagram showing the relationship between the burst density and the humidity according to the type of steel having a predetermined carbon content. As shown in FIG. 8A, in SCR420, SCR440, SCM420, SCM440, SNCM420, and SNCM439 having a predetermined carbon content, the burst density and the humidity have a substantially linear correlation according to the steel type. It becomes. The burst density decreases with increasing humidity. In this way, a calibration curve corresponding to a steel type having a predetermined carbon content can be created. As shown in FIG. 8B, in S25C, S35C, S45C, and S55C having a predetermined carbon content, the burst density and the humidity have a substantially linear correlation according to the type of steel. The burst density decreases with increasing humidity. In this way, a calibration curve corresponding to a steel type having a predetermined carbon content can be created. Therefore, the steel type can be specified from the burst density, the humidity, and the calibration curve corresponding to the steel type. In FIG. 8, linear approximation is used, but another approximation method such as curve approximation may be used.

  In the present embodiment, the steel type is directly specified from the burst density, the humidity, and the calibration curve corresponding to the steel type, but is not limited thereto. For example, by an experiment, the correlation between the burst density and the humidity and the carbon content in the steel was determined, and a calibration curve corresponding to the carbon content was created.From the calibration curve corresponding to the burst density and the humidity and the carbon content, The carbon content may be specified, and a steel type corresponding to the carbon content may be specified.

  In the present embodiment, one or more calibration curve data is stored in the RAM 132 as a database, and a relationship between the humidity and the burst density in a predetermined steel material, which is measured in advance, is stored as a database. The obtained ambient humidity is applied with reference to one or more calibration curve data stored in the RAM 132, and a steel material is identified from the corresponding one of the one or more calibration curve data.

(2nd Embodiment)
Hereinafter, a steel component identification device and a method thereof according to a second embodiment of the present invention will be described in detail with reference to the drawings. Note that the description of the parts having the same configuration and processing as the steel material component identification device and the method thereof according to the first embodiment will be appropriately omitted.

  FIG. 9 is a block diagram illustrating a configuration of the steel component identification device according to the embodiment of the present invention. The steel material component identification device according to the present embodiment includes a camera 21 that captures an image of a spark generated when the steel B to be inspected comes into contact with the grinder C, a humidity detector 22 that measures the humidity of the surrounding environment, and an image captured by the camera 21. A computer 23 for quantifying sparks based on the detected spark image and the humidity measured by the humidity detection unit 22. The computer 23 includes a CPU 231 that performs arithmetic processing, a RAM 232 that is a work area for data, and a ROM 233 that stores a control program of the CPU 231. The steel material component identification device of the present embodiment configured as described above quantifies the spark of the steel material by performing the following processing.

  FIG. 10 is a flowchart illustrating a steel component identification routine executed by the CPU 231 of the computer 23. The steel material component identification device according to the present embodiment includes a humidity capture processing unit, an image capture processing unit, a luminance conversion processing unit, a binarization processing unit by executing a control program stored in the ROM 233 by the CPU 231. , A short straight line matching processing unit, a rupture part extraction processing unit, and a steel material identification processing unit.

  Hereinafter, the detailed processing of each of steps S21 to S27 will be described.

(Humidity intake)
In step S21, the CPU 231 of the computer 23 loads the humidity of the surrounding environment measured by the humidity detector 22 into the RAM 232, and proceeds to step S22.

(Image capture)
In step S22, the CPU 231 of the computer 23 loads the spark image captured by the camera 21 into the RAM 232 as a color image, and proceeds to step S23.

(Brightness conversion)
In step S23, the CPU 231 converts the brightness of each pixel of the spark image to convert the image into a grayscale image having only the brightness value, stores the image in the RAM 232, and proceeds to step S24.

  FIG. 3A is an example of a grayscale image of a spark image. FIG. 3A shows that the spark image includes a spark line (core line) and a burst portion. In step S23, the color spark image obtained in step S22 is converted to grayscale. In the present embodiment, the RGB values of each pixel in the spark image are converted to a gray scale value Y by a predetermined conversion formula. For example, Y = R × 0.299 + G × 0.587 + B × 0.114 is used as the conversion formula.

(Binarization processing)
In step S24, the CPU 231 executes a spark image binarization process, stores the image in the RAM 232, and proceeds to step S25.

  In the present embodiment, the binarization processing is performed at a predetermined threshold for each pixel to be subjected to the binarization processing. By performing binarization, spark lines (core lines) become clearer. In the present embodiment, cross binarization suitable for the case where the extraction target is a linear shape such as a spark is used.

  FIG. 4 is a diagram illustrating the cross binarization used in step S3 of the present embodiment. The cross-binarization method subtracts the brightness value of the center pixel of the cross from the average brightness value of the pixels other than the center pixel in the cross, and if the value is greater than a set threshold, sets the center pixel of the cross to black ( FIG. 4 (a)), otherwise it is white (FIG. 4 (b)). Furthermore, in order to make the spark a black pixel, the black and white output of the cross-binarized image is inverted. FIG. 3B shows an example of an image obtained by performing cross-binarization and inverting black and white.

(Short straight line matching)
In step S25, the CPU 231 performs a short straight line matching process on the spark image, stores the type (pattern number) and position of the matched template in the RAM 232, and proceeds to step S26.

  As shown in FIG. 5, in the steel material component identification device of the present embodiment, templates T1 to T28 representing short straight lines corresponding to angles in 28 directions (every 6.5 degrees) are prepared in advance and stored in the RAM 22. The templates T1 to T28 of this embodiment include short straight lines each having a thickness of 1 pixel and a length of 10 pixels.

  FIG. 11 is a diagram illustrating a method of the short straight line matching process. The binarized image I5 in FIG. 11 illustrates a burst position of a specific spark in the spark image in order to explain step S25. In step S25 of the present embodiment, short straight line matching is performed on the core lines of other sparks present in the spark image.

  Scanning of pattern numbers (T1 to T28) is performed on the skeleton groups a to d in the binarized image I5, and a pattern template indicating an angle is matched. In step S4 of the present embodiment, the matching using the pattern template is performed only for the angle of the line, and is not performed for the length of the line. FIG. 11 is a diagram illustrating some examples of a matching method using a pattern template. Matching is also performed on a core line in which no pattern template is matched in the figure.

  FIG. 12 is a diagram illustrating a method of a short straight line matching process using a plurality of templates according to the present embodiment. Using the short straight line images of the templates T1 to T28 shown in FIG. 5, the spark information is extracted by fitting to the core line A of the spark image. When there are a plurality of short straight lines to be matched (for example, T1, T3, and T5 in FIG. 5), the short straight lines of each template are extended, and a better match is selected. In FIG. 12, only the template T5, which is a short straight line extension, does not protrude from the width of the core line A and stays within the core line A, so that the template T5 is selected.

  The extension length of the short straight line of each template may be determined according to the width of the core wire of the spark, and is preferably 0.5 times or more the width of the core wire. If the length is shorter than that, even if the templates in all directions are arranged, they may remain within the core wire and may not protrude even one. The extended length is more preferably at least one time, more preferably at least two to three times the width of the core wire.

  As shown in FIG. 12, the short line of each template is extended by arranging an arbitrary point within the spark core line, selecting a short straight line (for example, T1) protruding from the core line, and using this as a reference line. By arranging a short straight line longer than the reference line (T1) on another short straight line (for example, T3, T5) which does not protrude, the other short straight line may be extended based on the short straight line. In addition, the extension of the short straight line may extend an existing short straight line, replace the short straight line with a longer straight line, or overlap a longer straight line.

  In the present embodiment, spark recognition is performed using short straight line matching (template matching). In the conventional short straight line matching, an accurate inclination cannot be taken when recognizing a core line having a certain width, which may cause erroneous recognition. In the short straight line matching according to the present embodiment, when a plurality of templates having different angles match, the short straight line is virtually extended, and a more matching one is selected. By doing so, a more accurate core wire angle can be obtained.

  Then, the pattern number of the matched template is recorded. In FIG. 11, five patterns of pattern templates T14, T22, T4, T8 and T26 are respectively matched to the a, b, c, d and e portions in the binarized image I5 by the method shown in FIG. ing.

(Rupture extraction process)
In step S26, the CPU 231 extracts a burst portion of the spark from the spark image, and counts the number of burst sparks. Then, the number of rupture portions of the entire image and the total number of short straight lines existing in each rupture portion are stored in the RAM 232, and the process proceeds to step S27.

  The rupture of the spark has a morphological feature that the spark flies in various directions. Therefore, it is possible to extract the rupture portion by checking the direction in which the spark flies from the angle of the short straight line. In the present embodiment, a case where there are, for example, four or more types of short straight lines having different angles is regarded as a rupture portion.

  An image I6 in FIG. 13 illustrates a core line at a burst position of a specific spark in the spark image in order to explain step S26. As shown in FIG. 13, for the range of an arbitrary spark image (a × b [pixel]) (12 × 12 in this case) in the image I6, the total of the types of the pattern numbers scanned is obtained, and the total value If there are m or more (4 in this case) types, that range is defined as the range in which rupture exists.

  Here, the range of the “arbitrary” spark image is set because the spark pattern included in the selected image changes depending on the magnification, size, thickness of the core line, and the like of the image (FIG. 3 ( b)). As means for selecting the range of the “arbitrary” spark image, it is also possible to select the number of pixels from the image as shown in FIG. 13 or to select a plurality of pixels in advance and perform the matching with FIG. . The above operation is a means for dividing an image as a preliminary step for improving the prediction accuracy by determining a spark image from the entire image. Further, in the range of one selected spark image, the number of bursts changes depending on the movement of the range itself up and down, left and right, the size of the range, and the thickness of the core wire. It does not cause an extreme change in the number of bursts.

  In FIG. 13, since four types of patterns T14, T22, T26, and T4 are included, it can be recognized as a rupture portion. Then, the total number of ruptures is obtained by counting the number of ruptures in the entire image. At the same time, the total number of short straight lines existing in the entire spark image is counted from the total number of templates matched using the pattern template.

(Steel type discrimination processing)
In step S27, the CPU 231 calculates the burst density of the spark, estimates the amount of carbon in the steel from the burst density and the humidity of the spark, and recognizes the type of steel.

  In the present embodiment, one or more calibration curve data, in which the relationship between humidity and burst density in a predetermined steel material measured in advance is used as a calibration curve, is stored in the RAM 232 as a database, and the CPU 231 calculates the calculated burst density and the measured burst density. The obtained ambient humidity is applied with reference to one or more calibration curve data stored in the RAM 232, and a steel material is identified from the corresponding one of the one or more calibration curve data.

  Although the embodiment has been described above, the embodiment is presented as an example and is not intended to limit the scope of the invention. The new embodiment can be implemented in other various forms, and various omissions, replacements, and changes can be made without departing from the spirit of the invention. This embodiment and its modifications are included in the scope and gist of the invention, and are also included in the invention described in the claims and equivalents thereof.

11: Camera 12: Humidity detector 13: Computer 131: CPU
132: RAM
133: ROM

Claims (11)

A humidity detector for detecting the humidity of the surrounding environment;
A brightness conversion processing unit that converts an image of a spark generated by grinding a steel material into a grayscale spark image,
A binarization processing unit that performs binarization at a predetermined threshold for each pixel of the grayscale spark image,
For the binarized spark image, each of a plurality of templates representing short lines corresponding to a plurality of angles is matched, and a short line matching processing unit that stores the type and position of the matched template,
When the matched template is equal to or more than a predetermined number in the range of any spark image, a burst portion is extracted as a burst portion of the spark, and the number of bursts is the total number of the burst portions of the entire spark image, and A burst part extraction processing unit that counts the number of short straight lines that is the total number of the matched templates of the entire spark image,
A burst density obtained by dividing the burst number by the short straight line number, and the humidity, a steel material identification processing unit that identifies a steel material based on the humidity.
A steel material component identification device comprising: The steel material identification processing unit,
The steel material component identification device according to claim 1, wherein a carbon content of the steel material is identified based on the burst density and the humidity, and the steel material is identified from the carbon content.   The binarization processing unit, when the value obtained by subtracting the luminance value of the center pixel of the cross from the average luminance value of the pixels other than the center pixel in the cross is larger than the set threshold, sets the center pixel of the cross to black, 3. The steel material component identification device according to claim 1, wherein a cross-binarization process of whitening otherwise is performed. 4.   4. The thinning processing unit according to claim 1, further comprising a thinning processing unit configured to remove black pixels of the spark image binarized by the binarization processing unit from the outside to have a predetermined pixel width. 5. Steel component identification device.   The steel material component identification device according to any one of claims 1 to 4, wherein the short straight line matching processing unit records a template in which the spark image matches a predetermined percentage or more.   The steel material component according to any one of claims 1 to 5, wherein the steel material identification processing unit calculates each of the burst densities from a plurality of spark images, and identifies the steel material from an average burst density obtained by averaging the burst densities. Identification device. The brightness conversion processing unit,
Convert the image of the spark including the rupture part generated by grinding the steel material and the core wire of a predetermined width into a grayscale spark image,
The short straight line matching processing unit,
A plurality of templates each representing a short line corresponding to a plurality of angles with respect to a core line having a predetermined width of the binarized spark image are respectively matched by extending the short lines of the plurality of templates, and the extended short lines Memorizes the type and position of the template that fits in the core,
The steel material component identification device according to claim 1, wherein:   The steel material component identification device according to claim 7, wherein the short straight line matching processing unit determines an extension length of a short straight line of the plurality of templates according to a width of the core wire.   The short straight line matching processing unit arranges an arbitrary point in the core line, selects a short straight line that protrudes from the core line from among the short lines of the plurality of templates, sets this as a reference line, and protrudes from the core line. The steel material component identification device according to claim 7, wherein the other short straight line is extended by arranging a short straight line longer than the reference line on another short straight line that is not present.   The steel material component identification device according to claim 9, wherein the extension of the short straight line extends an existing short straight line, replaces the short straight line with a longer straight line, or overlaps a longer straight line. Computer
Humidity acquisition means for acquiring the humidity of the surrounding environment;
Brightness conversion processing means for converting a spark image generated by grinding steel material into a grayscale spark image,
Binarization processing means for performing binarization at a predetermined threshold for each pixel of the grayscale spark image,
For the binarized spark image, each of a plurality of templates representing short lines corresponding to a plurality of angles is matched, and a short line matching processing unit that stores the type and position of the matched template,
When the matched template is equal to or more than a predetermined number in the range of any spark image, a burst portion is extracted as a burst portion of the spark, and the number of bursts is the total number of the burst portions of the entire spark image, and Rupture portion extraction processing means for counting the number of short straight lines that is the total number of the matched templates of the entire spark image,
Burst density obtained by dividing the number of bursts by the number of short straight lines, and the humidity, steel material identification processing means for identifying a steel material based on,
A steel component identification program characterized by functioning.