JP5495017B2 - Tooth contact surface detection device - Google Patents

Description

  The present invention relates to a technical field for evaluating the tooth contact of a gear, and more particularly to a technology for evaluating a tooth contact surface of a gear by image processing.

  Conventionally, after applying a paint mainly composed of pigment such as Komyotan on the tooth surface including the tooth contact surface (engagement surface), rotating the gear, visually checking the paint peeling, The quality of the tooth contact was determined accordingly. However, in this conventional example, since it is a visual inspection, it is difficult to quantify the quality determination of the tooth contact surface.

Patent Document 1 proposes a method of extracting a tooth contact surface by photographing a tooth contact surface with a black and white camera and binarizing a grayscale image with a threshold value. In this conventional example, the threshold value is calculated and binarized based on the pixel values on the tooth surface including the tooth contact surface and not including the tooth tip (paragraphs 0048 to 0049, FIG. 8 and FIG. 8). FIG. 9).
Japanese Patent Laid-Open No. 2004-26883 discloses four levels of gradation according to the intensity of tooth contact by using RGB images captured by a color camera as grayscale images for each RGB, binarizing each image and superimposing the binarized images. A method for obtaining an image has been proposed (upper right column to lower left column on page 3, FIG. 3A). In this method, processing under rough natural light is possible (the upper left column on page 4). However, the determination of the quality of the tooth contact surface itself is performed manually (the lower right column on page 3). To page 4, upper left column, FIG. 2).
In Patent Document 3, the RGB image is converted into a grayscale image for each RGB, and the G component and the B component are added and inverted, and added to the R component, so that the subtle adhesion state of Komyotan is clearly luminosity. A technique for obtaining a gray image that appears as a difference has been proposed (paragraphs 0016 to 0018, FIG. 2). In particular, the brightness difference between the lap step portion and the tooth contact surface is made clearer (paragraph 0018, FIG. 6).

Japanese Patent Laid-Open No. 10-096621 Japanese Patent Laid-Open No. 3-115829 JP 2004-220244 JP

With the method described in Patent Document 1, it is impossible to make a determination based on the color of a paint such as Komyotan, and it is not possible to satisfactorily determine the degree of rubbing of Komyotan due to tooth contact.
In the method described in Patent Document 2, an image reflecting the color of paint such as Komyotan can be generated, but the determination of the quality of the tooth contact cannot be automated.
In the method described in Patent Document 3, even if the color of a paint such as Komyotan changes, the determination is performed by the same method using each component of RGB. Accuracy will be reduced. In particular, it cannot cope with variations in the measurement environment, such as variations in the amount of light applied, such as Komyotan, changes in color over time after application, changes in the color of the paint due to changes in the type of illumination and illuminance conditions, etc. .

As described above, in the above conventional example, a paint containing a pigment such as Komyotan is applied to the tooth surface, and the amount and color of the paint is changed by meshing and rubbing the tooth contact surface. An attempt is made to automatically determine whether the tooth contact is good or not based on the change.
[Problem 1] However, in the above conventional example, if there is variation in the measurement environment, there is a disadvantage that the change in the color of the paint due to rubbing of the tooth contact surface cannot be accurately identified. For example, when there is a change in the amount of paint applied, or the color of the paint itself or its color, there is an inconvenience that the color change due to tooth contact cannot be well captured.
[Problem 2] Furthermore, in the above-mentioned conventional example, the coating amount of the paint such as Komyotan, the color change with the lapse of time after application, the change of the illumination type, the change of natural light, the influence of disturbance light, etc. When the color changes, the measurement accuracy becomes unstable, and there is a disadvantage that the measurement conditions cannot be made uniform and the measurement accuracy cannot be stabilized.
That is, no image processing technique that is resistant to noise and can be determined only by a captured image is not disclosed in the above-mentioned documents.

  [Object of the Invention] An object of the present invention is to accurately identify the tooth contact surface even if the measurement environment changes.

  [Focus Point] The inventor of the present invention has the saturation (S) indicating the color tone (the vividness of the color specified by the hue) when the lightness (L) of the image changes due to the change in the coating amount, illumination conditions, and the like. ) Will change. And it came to the idea that the said subject could be solved by devising the image processing according to the difference in hue and stabilizing the lightness as preprocessing for specifying the tooth contact surface.

[Problem Solving Means 1] A first group of the present invention corresponding to Example 1 is a camera that generates a color image by capturing a color image of a tooth surface coated with a paint, and the tooth surface based on the color image. And a tooth contact surface detecting unit for detecting a tooth contact surface.
Then, the tooth contact surface detection unit is configured to specify a color feature value that is a feature value of the color of the tooth surface from the color image, and the color feature value for each color value of the tooth surface. A color difference data generation process for generating color difference data by calculating a color difference of the color difference, and calculating a threshold value from a histogram of the color difference data and binarizing the color difference data to identify the tooth contact surface And a surface identification process.
Thereby, the said subject 1 was solved.

[Problem Solving Means 2] In the second group of the present invention corresponding to the second and third embodiments, as the problem solving means 2, in addition to the configuration of the first group, the color space of the color image is converted to change the color image. A color coordinate conversion unit that calculates brightness, a brightness correction processing unit that corrects the brightness or brightness of the tooth surface so that the brightness of the color image is within a predetermined brightness range, and the brightness-corrected color It has a configuration in which a tooth contact surface detection unit that detects a tooth contact surface on the tooth surface based on an image is provided.
Thereby, the said subject 2 was solved.

  The present invention interprets the meaning of the terms described in each claim in consideration of the description of the present specification and the drawings, and certifies the invention according to each claim. There are the following advantageous effects in relation to

[Functional Effect 1 of Invention] In the tooth contact surface detection apparatus of Problem Solving Means 1, the color feature value specifying process specifies a color feature value to be a feature value of the color of the tooth surface from the color image, and generates color difference data. Since the process generates color difference data by calculating the color difference between each color value of the tooth surface and the color feature value, color difference data close to human color vision based on the color feature value according to the degree of tooth contact In addition, the tooth contact surface identification process calculates a threshold value from the color difference data histogram and binarizes the color difference data to identify the tooth contact surface. The binarized data representing the position and size can be calculated satisfactorily, whereby the tooth contact surface due to the application of the paint such as Komyotan and the tooth contact can be detected stably.
In particular, since the tooth contact surface identification process identifies the tooth contact surface based on the histogram of the color difference data, even if there is a change in the measurement environment, the color generated on the tooth contact surface as a color difference from the reference color feature value. Since the change can be captured, even if there is a change in the paint or the color of the paint, the tooth contact surface can be identified with the same accuracy as the inspection by human color vision.

  [Advantageous Effect 2 of Invention] In the tooth contact surface detection device of Problem Solving Means 2, the color coordinate conversion unit calculates the lightness, and the lightness correction processing unit directs the lightness of the color image within a predetermined lightness range. In order to correct the lightness L of the tooth surface, the hue and saturation can be stably set to favorable conditions. For this reason, in order to determine the tooth contact surface regardless of the illumination conditions and the amount of paint applied The conditions can be homogenized. Then, by making the lightness or luminance of the tooth surface portion constant by lightness correction, the sensitivity of hue and saturation can be increased, conditions that are easy to quantify can be secured, and the measurement accuracy of the tooth contact surface can be stabilized.

It is a block diagram which shows the structural example of one Embodiment of this invention. (Examples 1 to 3) It is a block diagram which shows the structural example of Example 1 of this invention. Example 1 FIGS. 3A to 3C are explanatory views showing an example of the measurement tooth. (Examples 1 to 3) It is a block diagram which shows the structural example of Example 2 of this invention. (Example 2) It is a graph which shows the example of a relationship between a brightness | luminance and saturation. (Examples 2 and 3) It is a graph which shows the example of a relationship between a brightness | luminance and a hue. (Examples 2 and 3) It is a block diagram which shows the structural example of Example 3 of this invention. (Example 3) FIG. 10 is a block diagram illustrating a configuration example of hardware resources according to a third embodiment. (Example 3) 10 is a flowchart illustrating an example of processing steps of Example 3. (Example 3) FIG. 10A is a diagram showing an example of an original image having a brightness of 57, and FIG. 10B is a diagram showing an example of extracting a tooth contact surface from the original image. (Examples 1 to 3) It is explanatory drawing which shows an example of the histogram of the original image shown to FIG. 10 (A). (Examples 1 to 3) It is explanatory drawing which shows an example of the binarized image which binarized the color difference data of the original image shown to FIG. 10 (A). (Examples 1 to 3) 13A is a diagram showing an example of an original image with a brightness of 40, FIG. 13B is a diagram showing an example of extracting a tooth contact surface from the original image, and FIG. 13C is a diagram after correction by the brightness. It is a figure which shows an example which extracted the tooth contact surface from the correction | amendment image. (Examples 1 to 3) It is explanatory drawing which shows an example of the histogram of the original image shown to FIG. 13 (A). (Examples 1 to 3) It is explanatory drawing which shows an example of the binarized image which binarized the color difference data of the original image shown to FIG. 13 (A). (Examples 1 to 3) FIG. 16A is a diagram showing an example of an original image having a brightness of 67, and FIG. 16B is a diagram showing an example of extracting a tooth contact surface from the original image. (Examples 1 to 3) It is explanatory drawing which shows an example of the histogram of the original image shown to FIG. (Examples 1 to 3) FIG. 17 is an explanatory diagram illustrating an example of a binarized image obtained by binarizing the color difference data of the original image illustrated in FIG. (Examples 1 to 3)

  Three embodiments are disclosed as modes for carrying out the invention. Example 1 is a tooth contact surface detection device having a tooth contact surface detection unit 14, Example 2 is a tooth contact surface detection device having a lightness correction processing unit 28, and Example 3 includes a tooth contact surface detection unit 14 and a tooth contact surface detection unit 14. This is a tooth contact surface detection device having both the brightness correction processing unit 28. Embodiments including Examples 1 to 3 are referred to as embodiments.

  The detection target of the present embodiment is the tooth contact surface 46 of various gears and is not limited to a specific gear, and various tooth contact surfaces 46 of various gears can be detected. If the tooth surface 42 of the gear is metal, it can be detected better, and in particular, it can also be used for inspections such as whether or not the degree of tooth contact on the many tooth surfaces 42 is uniform. .

Referring to FIG. 1, the tooth contact surface detection device detects a tooth surface 42 of a final gear (ring gear 96) of a differential (difference, differential device) used for a drive wheel of a four-wheel vehicle. The example shown is a spiral bevel gear.
The tooth contact surface detection device includes a camera 12 that captures an image of the tooth surface 42 to which the paint 40 is applied, and an illumination device 13 that illuminates the tooth surface 42, and according to the embodiment, the tooth contact surface detection unit. 14 and the brightness correction processing unit 28 or both.

  The differential device is a mechanism that changes the rotational speed of the left and right drive wheels according to the radius of rotation of the four-wheeled vehicle when turning right or left, and a drive input 91 such as an FR propeller shaft or an FF transmission. And the left and right drive shafts 92. The differential device includes a ring gear 96 that meshes with a drive pinion 90 that serves as a drive input 91, a differential case 94 that rotates integrally with the ring gear 96, and a differential that rotates by rotating the differential case 94 according to the drive input 91. A pinion gear and a side gear (not shown) that rotate for each of the left and right drive shafts 92 (axle shafts) in the case 94 are provided. The left and right side gears rotate according to the resistance given to the left and right wheels by rotation while revolving, and transmit the rotational force to the drive shaft 92.

  The differential case 94 is rotatably supported by the gear carrier 98. The ring gear 96 of the differential case 94 has a large number of tooth surfaces 42. As the tooth surface 42 of the ring gear 96 engages with the drive pinion 90, the drive input 91 is transmitted to the drive shaft 92 and the drive wheel. For this reason, the quality of the meshing surface (tooth surface 42) of the ring gear 96 is important for performance stability.

  The tooth contact surface detection device of the present embodiment detects, for example, such a tooth contact surface 46 on the tooth surface 42 of the ring gear 96, and the area of the tooth contact surface 46, the position of the center of gravity, or a predetermined area. And the difference from the center of gravity position, the degree of tooth contact, and the like are calculated. In addition to the tooth surface 42 of a normal gear, the tooth contact surface 46 can be detected even if the tooth surface 42 has a complicated shape such as a spiral bevel gear.

  In this embodiment, after the coating material 40 is applied to the tooth surface 42, the ring gear 96 and the drive pinion 90 are engaged to rub the coating material 40 on the tooth contact surface 46, and the color of the coating material 40 discolored by this rubbing (striking). The state is identified by image processing. As the coating material 40, a powder having a fine particle size made of an inorganic pigment having a high chroma is preferable. For example, a light pigment which is a red pigment can be used. Komyotan is also said to be a red lead, and it has been mainly used in previous inspections with lead tetroxide as the main component. Further, instead of lead, other pigments having a fine particle size may be used.

  Human color vision does not have the same resolution for all colors, but has high discrimination power for colors ranging from red to orange, so if you use the red pigment, Mitsumetan, you can visually inspect subtle color changes. Can be identified with higher resolution.

  In this embodiment, the state of discoloration of the paint 40 due to the hit is captured by the camera 12, and the color image 52 is digitally processed to detect the tooth contact surface 46. In the conventional example, the color system of RGB (red R, green G, blue B) is simply used as a premise to digitally process the color image 52. However, in this embodiment, the color difference δ or the simple color difference is used. To calculate Δ, a color system other than RGB is also used.

  Expressing color quantitatively is called color specification, and the system for color specification is called color specification system. As the color system, there are a color system and a developed color system that expresses the color of an object, and a color mixture system that expresses a color based on the same color by color mixture. The mixed color system includes an additive color mixture useful for specifying light source light such as a monitor and a subtractive color mixture useful for specifying object light such as printed matter. RGB is used in additive color mixture, and CMYK (cyan, yellow, magenta, black) is frequently used in subtractive color mixture.

  As the color developing system, there is a Munsell color system in which color tables are arranged based on human senses. The Munsell color system has lightness (Munsell value V) on the vertical axis, hue (Munsell Hue H) in the circumferential direction, saturation (Munsell chroma C) in the radial direction, and the color “hue lightness / saturation”. Represented by For example, red in the JIS conventional color name is expressed as “5R 4/14”. In the Munsell color system, for each hue value, the color is expressed by the lightness that specifies brightness and the saturation that specifies the vividness of the color. Can be close to the senses. However, since the color table is used, it is difficult to quantify the color between the color tables.

  In the mixed color system, various colors can be expressed by combinations of stimulus values such as RGB. This combination of primary stimulus values is called a color matching function. If the color matching function is used, an arbitrary color can be accurately identified using the original stimulus value. However, in the RGB color system, a negative color mixture occurs with respect to R, which makes handling difficult. Therefore, an original stimulus value XYZ that does not cause a negative color mixture is determined, and the XYZ color system is defined. The XYZ color system and the xy chromaticity diagram are often used for color quantification. In addition, handling of colors by combinations of respective stimulus values in these various color systems is also referred to as a color space.

  In the XYZ color system, the distance in the color space does not correspond to the color difference caused by human senses, and the color difference obtained as the distance in the xy chromaticity diagram is not perceptually equal. An ellipse showing the range of the same color as the central color is called an Aquadam ellipse. When this Aquadam ellipse is superimposed on the xy chromaticity diagram, the size of the ellipse varies depending on the coordinates of the xy chromaticity diagram.

  Unlike the color matching function, this color difference is non-linear and it is difficult to define a uniform color space. However, as a uniform color space including lightness, XYZ is used as an original stimulus value, and L * a * b * color space and L * u * v * color space have been proposed. L * represents lightness, and a * b * and u * v * represent color perception attributes composed of hue and saturation. The L * a * b * color space is constructed to correspond to the Munsell color system, and the L * u * v * color space corresponds well to the Aquadam ellipse. In this uniform color space, the distance in the space can be a color difference that is close to a human sense.

  In computer graphics, the HSL color system and the HSB color system are used, although they are not the color systems in the above-described chromatics. This color model is similar to the Munsell color system, and expresses colors with three axes. In the HSB color system, hue (hue H), saturation (saturation S), and luminance (brightness B) are three axes. In the HSL color system, brightness (lightness L) is used instead of luminance B. The hue represents a color system at an angle of 360 degrees, and has a hue circle similar to the Munsell color system. Saturation is represented by a value from 0% to 100%, where 0% is an achromatic color and 100% is the most vivid color in its hue and lightness. Lightness indicates the level of brightness, 0% is black, and 100% is the brightest in its hue and saturation.

  The color image 52 photographed in RGB can be converted into the HSL color system using a predetermined formula. Further, the HSL color system and the HSB color system are similar to the Munsell color system, and are characterized in that the distance in the space of the HSL color system can be easily handled as a color difference.

<1 tooth contact surface detection device: change correspondence>
<1.1 Color difference data>
Example 1 of this embodiment is disclosed. In the first embodiment, information processing based on the color difference data 56 is performed in order to accurately identify the tooth contact surface 46.

  The tooth contact surface detection device includes a camera 12 and a tooth contact surface detection unit 14 as main elements, and the tooth contact surface detection unit 14 generates a tooth on the tooth surface 42 based on the color image 52. The contact surface 46 is detected. For this purpose, the tooth contact surface detection unit 14 includes a color feature value specifying process 16, a color difference data generating process 18, and a tooth contact surface identifying process 20.

  In the example illustrated in FIG. 2, the tooth contact surface detection device 14 further includes a lighting device 13. Then, the paint 40 is applied to the tooth surface 42 which is the measurement object, and the color of the paint 40 changes on the tooth contact surface 46 of the tooth surface 42 due to the contact of the gear.

The camera 12 images the tooth surface 42 to which the paint 40 is applied in color to generate a color image 52. The color image 52 is, for example, data in the RGB color system, and has a red R value, a green G value, and a blue B value for each pixel.
In a preferred example, the illumination device 13 emits illumination light of a predetermined spectral component with a predetermined brightness, the camera 12 receives reflected light reflected by the tooth surface 42, and a color image. 52 is good. By using a halogen lamp, a xenon lamp, or the like as the illumination device 13, the color of the tooth surface 42 can be accurately set to the color image 52 when the spectral component of the illumination light is white light close to natural light. When a fluorescent lamp or the like is used, correction according to the spectrum of the fluorescent lamp may be performed depending on the color of the paint 40.

  The color feature value specifying process 16 of the tooth contact surface detection unit 14 specifies a color feature value 54 that is a color feature value of the tooth surface 42 from the color image 52. The color feature value 54 is a color that serves as a reference for calculating the color difference data 56, and may be obtained from the captured color image 52. In the present embodiment, the color of the paint 40 is bright and vivid at the stage where the paint 40 such as Komyotan is applied, and the portion where the tooth contact occurs is dark and dark. The color feature value 54 may be a feature value that makes it easy to detect the color difference δ of the paint 40 according to the presence or absence of the tooth contact. The color feature value 54 may be, for example, the color of the darkest part or the lightest part of the surface to which the paint 40 is applied. The color of the darkest part is the color of the pixel having the maximum lightness L and saturation S, and the color of the brightest part is the color of the pixel having the minimum lightness L and saturation S. it can. Since the brightest color tends to change depending on the lighting environment or the like, it is desirable to set the dark color to the color feature value 54 in order to stabilize the color feature value 54 and the color difference δ.

The color difference data generation process 18 generates color difference data 56 by calculating a color difference δ with respect to the color feature value 54 for each color value of the tooth surface 42. The color difference δ can be approximated by the distance between the coordinate values of two colors when the color value in the color space by the color system is used as the coordinate value. If the color space of the color image 52 is an L * a * b * color space or an L * u * v * color space, a color difference δ close to human color vision can be handled. Also, in the HSB color space and the HSL (LHS) color space, the distance in the color space can be treated as a color difference δ that is relatively close to human color vision. Data for which the color difference δ is calculated for all color values of the tooth surface 42 is referred to as color difference data 56. When the color difference data 56 based on the color difference δ is used, data close to human color vision can be quantified, so that it is easy to match with the conventional visual inspection method.
Further, the color difference data 56 represents the color difference as a numerical value, and is one-channel data. When visualized, the color difference data 56 can be a gray scale (grayscale image).

  The tooth contact surface identification process 20 calculates a threshold value 62 from the histogram 58 of the color difference data 56. The color difference data 56 is data indicating the distance from the color feature value 54. When the color feature value 54 is a dark color, the distance to the dark and dark portion of the tooth surface 42 is a small numerical value that is bright and vivid. The distance to is far and large. This one-channel data can obtain a histogram 58, and it is possible to separate a dark and dark portion where a tooth contact occurs and a bright and bright portion where there is no tooth contact. For example, when the discriminant analysis method is used, it is possible to obtain the threshold value 62 between the dark part to be extracted and the background with the bright part as the background.

  The tooth contact surface identification process 20 uses the threshold value 62 to binarize the color difference data 56 to identify the tooth contact surface 46. When the color difference data 56 is binarized using the threshold value 62 obtained from the histogram 58, binarized data 60 in which only the tooth contact surface 46 is extracted can be obtained. By using this binarized data 60, the position, size (area) and shape characteristics of the tooth contact surface 46 can be detected by image processing.

FIG. 3A shows an example of a blue B channel image when the color image 52 is RGB, and FIG. 3B shows an example of a red R channel in gray scale. FIG. 3C shows an example of the measurement result of the tooth surface 42 of the measurement tooth 50.
Referring to FIG. 3, the ring gear 96 has a tooth surface 42 that is a side surface of the gear and a tooth tip 44 that is an upper portion of the gear. The tooth surface 42 includes a tooth contact surface 46 where tooth contact has occurred and a non-tooth contact surface 48 which is a portion other than the tooth contact surface 46 of the tooth surface 42. The tooth contact is one time or multiple times. Further, since the force applied to the tooth surface 42 at the time of tooth contact is not uniform, the tooth contact strength varies depending on the position of the tooth surface 48. However, as a result of the tooth contact test, it is possible to classify the tooth contact surface 46 where the tooth contact occurred and the non-tooth contact surface 48 including a portion where there was a slight tooth contact.

  In the example shown in FIGS. 3A and 3B, pattern matching using the feature of the shape of the tooth tip 44 is used to specify the measurement tooth 50. The color image 52 is picked up while rotating the ring gear 96, and the tooth surface 42 sandwiched between the upper and lower tooth tips 44 is extracted as the measurement tooth 50. As shown in FIG. 3C, when the tooth contact surface detection unit 14 extracts the measurement tooth 50, the tooth contact surface is detected by the color feature value specifying process 16, the color difference data generation process 18, and the tooth contact surface identification process 20. 46 is extracted.

  Next, a specific calculation method for the color difference data 56 will be described. In a preferred example, the color image 52 is represented by the HSL (LHS) color system, and each color value of the color image 52 is represented by lightness L, hue H, and saturation S. Then, the color difference data generation processing 18 calculates the color difference δ by the color difference calculation formula of the following equation (1) when the color value of the color feature value 54 is lightness Lo, hue Ho, and saturation So. In the following equation (1), since the distance in the HSL color space is the color difference δ, it is possible to generate color difference data 56 close to human color perception compared to the color difference due to the distance in the RGB color space. Then, the color difference data 56 close to human color vision using Equation (1) is highly related to the tooth contact intensity, and therefore can be detected with high accuracy.

  In another preferred example, when the color difference data generation processing 18 sets the color value of the color feature value 54 as lightness Lo, hue Ho, and saturation So, and coefficients as α, β, and γ, the following formula ( The simple color difference Δ is calculated using the simple color difference calculation formula 2). In the calculation of the simple color difference Δ, the simple color difference Δ can be obtained only by subtraction, multiplication and addition of numerical values, so that the color difference data 56 can be calculated at high speed. Then, by adjusting the coefficients α, β and γ, simple color difference data 56 having no practical problem can be calculated according to the color of the paint 40. Further, in the example using Komyotan, even if α = β = γ = 1, it can be detected with a certain accuracy. When adjusting the coefficients α, β, and γ, the coefficient values may be optimized using the least-squares method or the like with the color difference group calculated by Equation (1) as a target value.

1.1 Effect of Color Difference Data As described above, the color feature value specifying process 16 specifies the color feature value 54 to be the color feature value of the tooth surface 42 from the color image 52, and the color difference data generating process 18 Since the color difference data 56 is generated by calculating the color difference δ with respect to the color feature value 54 for each color value of the tooth surface 42, it is close to human color vision based on the color feature value 54 corresponding to the degree of tooth contact. The color difference data 56 can be generated, and the tooth contact surface identification processing 20 calculates a threshold value 62 from the histogram 58 of the color difference data 56 and binarizes the color difference data 56 to thereby generate the tooth contact surface. In order to identify 46, the tooth-contact surface 46 by application | coating of the coating materials 40, such as Komichitan, and tooth contact, can be detected favorably.

  In addition, by using the color difference data 56, it is possible to obtain data that satisfactorily reflects the color change due to the tooth contact, and to binarize from the histogram 58 of the color difference data 56. 48 can be separated and the tooth contact surface 46 can be detected stably.

In the calculation of the color difference δ, if the HSL color space is used and the distance in the color space is the color difference δ, quantification close to human color vision is possible, and consistency with the conventional visual inspection is maintained. Stable detection by image processing can be performed while utilizing conventional know-how of inspection.
Further, in the example of calculating the simple color difference Δ, the calculation is simplified and the tooth contact surface 46 can be detected by high-speed processing.

<2 tooth contact surface detector: homogenization>
<2.1 brightness correction>
Next, Example 2 of this embodiment is disclosed. In the second embodiment, brightness correction is performed with a predetermined brightness range 68 as a target in order to make the measurement conditions uniform.

  The tooth contact surface detection device includes a camera 12, a color coordinate conversion unit 26, a lightness correction processing unit 28, and a tooth contact surface detection unit 14 as main elements. In a preferred example, the camera 12 may include a gain adjustment unit 32, and the brightness correction processing unit 28 may include a gain adjustment processing 34.

  In the example shown in FIG. 4, the camera 12 images the tooth surface 42 to which the paint 40 is applied in color and generates a color image 52. A color coordinate conversion unit 26 in the image processing unit 24 converts the color system (color space) of the color image 52 to calculate the brightness of the color image 52. The color space before conversion includes an RGB color space, and the color space after conversion includes an HSL color space and an HSB color space. Further, as a color space after conversion, an L * a * b * color space or an L * u * v * color space may be employed. The lightness L is named luminance B depending on the color space. The lightness L is a value having no tint different from hue and saturation, and is a value that gives brightness from white to black.

  The lightness correction processing unit 28 corrects the lightness L or luminance of the tooth surface 42 so that the lightness L of the color image 52 falls within a predetermined lightness range 68. The brightness range 68 is a slightly brighter range from an intermediate value of the brightness L (for example, 50 (%) if 0 (%) to 100 (%)), for example, a range of 50 to 60 in the brightness L of the HSL color space. And good. When correcting the lightness in the HSL space, the lightness of each pixel is multiplied by (55 / L) to correct the central value of the lightness L to the vicinity of 55, whereby the lightness L is set to a predetermined lightness range. 68 so that it is within 68. When the brightness is corrected in the RGB color space, the luminance value of each RGB channel is preferably multiplied by (55 / L). This magnification is a dimensionless ratio of the numerator and denominator in lightness (a ratio expressed as a normal number or percentage). The numerator is the center value of the brightness range 68, and the denominator is the brightness of the color image 52. This ratio is called a brightness correction ratio. The brightness correction ratio plays a role of correcting brightness values brighter than the center value (for example, 55) darker and brightness values darker than the center value brighter.

  The tooth contact surface detection unit 14 detects a tooth contact surface 46 on the tooth surface 42 based on the color image 52 whose brightness has been corrected. When the brightness is corrected, the deviation between the hue H and the saturation S is small, and the detection according to the color of the tooth contact surface 46 can be performed satisfactorily.

FIG. 5 shows the correlation between the lightness L and the saturation S. In this example, the measurement tooth 50 was photographed in Examples 1 to 3 while changing the illumination conditions and the aperture of the camera 12, and the overall lightness L and saturation S of the tooth contact surface 46 were calculated. As shown in FIG. 5, the saturation S is maximized in the vicinity of lightness L = 60. As the saturation S increases, the color difference δ between the tooth contact surface 46 and the non-tooth contact surface 48 increases, so that the tooth contact surface 46 can be easily identified.
When the lightness exceeds L = 50, identification becomes easy. Furthermore, when the lightness exceeds L = 60, color skipping (a part of the RGB channel is over-range) occurs, the saturation S decreases, and the tooth contact surface 46 is difficult to identify.

  FIG. 6 shows a correlation between the lightness L and the hue H. And it is a relatively constant value under the condition of lightness L = 50 to 60. On the other hand, when the lightness L is out of this condition, the hue H changes, and the actual color and the color in the color image 52 are different.

  As shown in FIG. 5 and FIG. 6, the condition of lightness L = 50 to 60 is optimal, the lightness range 68 is preferably in the range of 50 to 60, and the median is 55. Therefore, the lightness correction processing unit 28 may correct the lightness so that the lightness L of the color image 52 becomes the median value L = 55 (A5).

2.1 Effects of Lightness Correction As described above, the color coordinate conversion unit 26 calculates the lightness L of the color image 52, and the lightness correction processing unit 28 sets the lightness L of the color image 52 within a predetermined lightness range 68. Since the lightness L of the tooth surface 42 is corrected toward the surface, the hue and saturation can be stably set to favorable conditions. For this reason, the tooth contact surface regardless of the illumination conditions and the coating amount of the paint 40 The measurement accuracy can be stabilized by homogenizing the conditions for determining 46.

<2.2 Gain adjustment>
Referring to FIG. 4 again, the camera 12 includes a gain adjustment unit 32 that adjusts the gain of imaging. The lightness correction processing unit 28 includes a gain adjustment process 34 that controls the gain adjustment when the lightness L of the color image 52 is outside the lightness range 68.
The gain adjustment unit 32 is, for example, a gain adjustment function of the image input board 74 of the camera 12, and may be an automatic adjustment of the light amount of the lighting device 13 or an automatic adjustment of the squeezing of the camera 12.

  In the gain adjustment process 34, for example, an average value or median value of the lightness L of the entire color image 52 or the tooth surface 42 is obtained, and this lightness is larger than a predetermined lightness value (for example, 60). Or, when the maximum lightness among the lightness L of the entire color image 52 or the tooth surface 42 is larger than a predetermined lightness value (for example, 65), the image is determined as a defective color image 52 and the gain is readjusted. It is good to control.

The threshold 62 for determining the quality of the color image 52 may be determined using the correlation data shown in FIGS. 4 and 5 according to the measurement object, the hue of the paint 40, the maximum saturation, and the like. With this gain adjustment, re-photographing can be controlled when an overflow occurs in any of the RGB channels and a uniform image cannot be obtained by the brightness correction processing.
Further, if the color image 52 re-photographed after gain adjustment is further subjected to brightness correction processing toward the brightness range 68, detection under a more uniform constant condition can be performed.

2.2 Effect of Gain Adjustment As described above, when the lightness L of the color image 52 is outside the lightness range 68, the gain adjustment processing 34 controls the gain adjustment, and the gain adjustment unit 32 controls the camera 12 according to this control. If the color image 52 is too bright and overflows, the color reproducibility deteriorates, and the brightness correction cannot restore the homogeneity, the gain is adjusted again by adjusting the imaging gain of the lighting device 13 or the lighting device 13. The range of brightness of the color image 52 can be set to a certain range by re-photographing with the adjustment. And by this gain adjustment, the determination of the tooth contact with low accuracy can be excluded in advance, and the detection accuracy can be stabilized even if the lighting environment such as the amount of solar radiation changes.
As described above, even if there is a change in the photographing condition or the like, the adverse effect of the overflow can be excluded in advance, and the color image 52 can be homogenized. Therefore, the detection of the tooth contact surface 46 can be quantified with higher accuracy than the determination by a skilled person.

<3 tooth contact surface detection device: corresponding to change and homogenization>
<3.1 Lightness and color difference>
Furthermore, Example 3 is disclosed. The third embodiment is intended to make maximum use of the hue and saturation information of the color image 52 in order to detect the size and position of the tooth contact accurately and stably.

  The tooth contact surface detection device includes a camera 12, a color coordinate conversion unit 26, a lightness correction processing unit 28, and a tooth contact surface detection unit 14 as main elements. The tooth contact surface detection unit 14 includes a color feature value specifying process 16, a color difference data generating process 18, and a tooth contact surface identifying process 20. In a preferred example, a gain adjustment unit 32 and a gain adjustment process 34 may be provided.

  In the example illustrated in FIG. 7, the camera 12 captures a color image of the tooth surface 42 to which the paint 40 is applied to generate a color image 52. Then, the color coordinate conversion unit 26 converts the color space of the color image 52 and calculates the lightness L of the color image 52. The lightness correction processing unit 28 corrects the lightness L or luminance of the tooth surface 42 so that the lightness L of the color image 52 is within a predetermined lightness range 68. For example, when the characteristic brightness value is L1, correction is performed by multiplying the brightness L of each pixel of the color image 52 by (55 / L1) disclosed in the second embodiment. Such lightness correction makes it possible to make the color reproducibility in the color image 52 uniform even if there is a change in the photographing environment, the coating amount of the paint 40, or the like.

  Then, the tooth contact surface detection unit 14 detects a tooth contact surface 46 on the tooth surface 42 based on the color image 52 whose brightness has been corrected. At this time, the color feature value specifying process 16 specifies a color feature value 54 to be a color feature value of the tooth surface 42 from the color image 52. For example, the color feature value 54 may be the darkest portion of the paint 40 disclosed in the first embodiment (the point at which the sum of lightness L and saturation S is maximized). Then, the color difference data generation process 18 generates color difference data 56 by calculating a color difference δ with respect to the color feature value 54 for each color value of the tooth surface 42. The color difference δ can be obtained as a distance in the color space using the above-described equation (1). A simple color difference Δ according to the equation (2) may be used.

  The tooth contact surface identification process 20 calculates the threshold value 62 from the histogram 58 of the color difference data 56 and binarizes the color difference data 56 to identify the tooth contact surface 46. The image processing unit 24 performs image processing such as pattern matching for extracting the tooth surface 42 to be measured. For example, the image processing unit 24 differentiates the color image 52 to extract an edge in the image, and compares the edge with a predetermined pattern to identify a target edge in the image.

  FIG. 8 shows a configuration example of hardware resources of the tooth contact surface detection device. In the example illustrated in FIG. 8, an image input board 74 that adjusts an image captured by the camera 12 and inputs the image to the CPU 72, a CPU 72 that performs calculation, a main storage device 76 that provides a storage space for the CPU 72, and a nonvolatile hard disk And the like, and a tooth surface 42 to be measured.

  The image input board 74 functions as the gain adjustment unit 32 that adjusts the gain of the color image 52. Further, the color image 52 may have a function of converting the color system (color space). The CPU 72 functions as the lightness correction processing unit 28 and the tooth contact surface detection unit 14 illustrated in FIG. 7 by executing a previously introduced tooth contact surface detection program. Further, the CPU 72 may be configured to function as the color coordinate conversion unit 26. The main storage device 76 is a primary storage device of the CPU 72 and temporarily stores programs and data. The auxiliary storage device 78 is, for example, a hard disk, and includes program data for tooth contact surface detection, a center value of a predetermined brightness range 68, pattern data used for pattern matching by image processing, and the center of gravity of the tooth contact surface 46. An allowable range for the position and area is stored. The drive mechanism 80 generates tooth contact by rotating a gear (for example, a ring gear 96) to which the paint 40 is applied. Further, the drive mechanism 80 rotates the gear when positioning the relative position of the tooth surface 42 with respect to the camera 12.

  FIG. 9 shows an example of the measurement flow of the third embodiment. The tooth contact surface detecting device is activated (step S1), and the applied paint 40, for example, light alum, is detected (step S2). Next, the drive mechanism 80 is controlled to rotate the ring gear 96 to perform tooth matching (step S3).

  After the tooth contact is made by the tooth alignment, the tooth surface 42 is positioned in order to photograph the tooth surface 42. That is, first, the image processing unit 24 detects the position of the tooth tip 44 by pattern matching (step S4), and confirms whether or not the position of the tooth tip 44 is within an allowable range (step S5). If it is not within the allowable range, the drive mechanism 80 is controlled to rotate the ring gear 96 and position it again. If it is within the allowable range, the camera 12 is controlled to photograph the tooth surface 42 (step S6, imaging step).

  Next, the color image 52 is decomposed into RGB channels (step S7), and coordinates are converted into the HSL color space (LHS color space) using the RGB values, and the lightness L, saturation S, and hue H are obtained. Calculate (step S8, color coordinate conversion step).

  When identifying the difference in color, it is known that the sensitivity around brightness L = 50 is high. When the lightness L is low, it is difficult to identify the color, and when it is too high, the hue H may be shifted due to color skipping (when any of the RGB channels exceeds the range). Therefore, the gains of the camera 12 and the image input board 74 are automatically corrected so that the party lightness L is in the vicinity of 55 (55-5), and re-photographing is performed. That is, the gain adjustment process 34 determines the quality of the color image 52 based on the value of the lightness L in the HSL color space, and re-photographs the image when it is defective.

  Specifically, when the lightness L of the color image 52 is outside the predetermined lightness range 68, the gain is adjusted as a defect of the color image 52 (step S10), and photographing is performed again (step S6). The brightness range 68 is, for example, 50 to 60. As the characteristic lightness L1 of the color image 52, the maximum value, average value, median value, minimum value, etc. of the lightness in the color image 52 can be used.

  Next, the brightness is corrected on the software toward the brightness L = 55 by multiplying the brightness L of the HSL or the brightness of each RGB channel by (55 / L1) (step S11, brightness correction process). By making the standard of the lightness L constant, it is possible to reduce the influence of light and shade due to the difference in the illumination conditions and the amount of light applied, and the color image 52 can be made uniform.

  Then, the contour data of the tooth surface 42 is read from the auxiliary storage device 78 (step S13), and the tooth surface 42 to be measured is detected by pattern matching (step S13). By detecting the tooth surface 42, it is possible to extract only data of the tooth contact surface 46 and the non-tooth contact surface 48 in the tooth surface 42. Subsequently, the color difference data 56 is calculated. Specifically, first, the darkest part of Komyotan (the point where the sum of L and S is the maximum) is selected as the color feature value 54, set to the reference point 0, the lightness reference value Lo, the hue reference value Ho, and the coloring. The degree reference value So is calculated (step S14, color feature value specifying step).

  Next, the color difference data generation process 18 calculates the color difference δ from the values of brightness L, hue H, and saturation S of each pixel of the color image 52 according to the above-described equation (1) (step S15, color difference data generation step). ). Further, the simple color difference Δ may be calculated by the equation (2). The color difference δ or the simple color difference Δ is converted into 256 gradations and converted into a grayscale monochrome image (grayscale), and an image with a clearer color difference is calculated by making the lightness L = 55 constant. Can be easily identified. By making the lightness L condition substantially constant, the discrimination between the tooth contact surface 46 and the non-tooth contact surface 48 can be quantitatively reproduced.

  Subsequently, the tooth contact surface identification processing 20 extracts candidates for the tooth contact surface 46 using a binarization technique based on the discriminant analysis method based on the color difference data 56 that is a grayscale monochrome image (step S17, tooth contact). Surface identification step). When there are multiple region candidates, the same processing is performed on the upper teeth and lower teeth, and the region existing at the same position is set as the tooth contact surface 46. When the tooth contact surface 46 is detected, the gravity center position and area of the tooth contact surface 46 are calculated (step S18), and the predetermined gravity center position and area allowable values are read from the auxiliary storage device 78 (step S19). By comparing, the quality of the tooth contact surface 46 is determined (step S20).

  The centroid position can be obtained from the average value of the x coordinate values and the average value of the y coordinate values of each pixel included in the tooth contact surface 46, and the allowable value of the centroid position may be a coordinate value within a certain range. it can. The area of the tooth contact surface 46 can be the number of pixels (pixels) included in the tooth contact surface 46, and the allowable value of the area can be in the range of the number of pixels. The size of each pixel of the color image 52 and the color difference data 56 and the size in the real space can be associated with each other by a scale value. If the scale value is measured in advance, the position of the center of gravity, the area, and their allowable values Can be the length and width in real space.

  The process shown in FIG. 9 can be realized by the CPU 72 executing a tooth contact surface detection program.

  FIG. 10A is a diagram showing an example of an original image with a lightness L = 57, and FIG. 10B is a diagram showing an example of extracting a tooth contact surface 46 from the original image. In the example shown in FIG. 10, the brightness is 57 and it is within the predetermined brightness range 68, so that brightness correction is not required. As shown in FIG. 10A, the tooth surface 42 and the tooth tip 44 of the ring gear 96 are photographed, and one of the teeth is the measurement tooth 50. When the tooth contact surface 46 is extracted from the original image shown in FIG. 10A, an example shown in FIG. 10B is obtained.

  FIG. 11 is an explanatory diagram showing an example of the histogram 58 of the original image having the lightness L = 57 shown in FIG. 10A, and FIG. 12 shows binary data of the color difference data 56 of the original image shown in FIG. It is explanatory drawing which shows an example of the digitized binarized image. When the data of the tooth surface 42 of the measurement tooth 50 is extracted and the color difference data 56 is converted into a gray scale, two peaks suitable for discriminant analysis can be obtained as shown in FIG. That is, the color difference data 56 is divided into a background portion (a group of gradation values belonging to the non-tooth-contact surface 48) and a characteristic portion (a group of gradation values belonging to the tooth-contact surface 46). If the value (value 145 in the example shown in FIG. 11) is binarized with the threshold value 62, only the characteristic portion (tooth contact surface 46) can be extracted.

  That is, as shown in FIGS. 11 and 12, in the case of lightness 57, the histogram 58 of the color difference data 56 is clearly separated from the background portion and the feature portion by the threshold value 62 (value 145). When the color difference data 56 is binarized with the value 62, the tooth contact surface 46 can be detected satisfactorily as shown in FIG. If the binarized data 60 is used, the position of the center of gravity and the area of the tooth contact surface 46 can be calculated, and the values also correspond well with the actual tooth contact surface 46.

  FIG. 13A is a diagram illustrating an example of an original image having the lightness L = 40 for the same object as that in FIG. 10A, and the brightness is low and lightness correction is required. FIG. 13B is a diagram showing an example in which the tooth contact surface 46 is extracted from the original image, and the tooth contact surface 46 of the measurement tooth 50 is detected to be larger. When the lightness is low as in the original image of lightness L = 40, the identified region tends to be wider than the actual tooth contact surface 46 as compared with the optimal FIG. 10B. FIG. 13C is a diagram showing an example in which the tooth contact surface 46 is extracted from the color difference data 56 after the luminance value is corrected by the lightness L, and the center of gravity position and area of the tooth contact surface 46 are well extracted. The correction shown in FIG. 13C is a result of luminance correction in which the RGB channels of the color image 52 are multiplied by a constant, and an area close to FIG. 10B can be detected.

  FIG. 14 is an explanatory diagram showing an example of the histogram 58 of the original image shown in FIG. 13A, and FIG. 15 shows an example of a binarized image obtained by binarizing the color difference data 56 shown in FIG. Since the lightness 40 and the dark color image 52 are used, the tooth contact surface 46 that is a characteristic portion is extracted larger than the actual size. On the other hand, when the brightness is corrected, as shown in FIG. 13C, the tooth contact surface 46 can be detected in substantially the same manner as the case shown in FIG.

  FIG. 16A is a diagram illustrating an example of an original image having a brightness of 67, and FIG. 16B is a diagram illustrating an example of extracting a tooth contact surface 46 from the original image. In the example shown in FIG. 16, the brightness L = 67, and the brightness is too high to correct the brightness. FIG. 17 is an explanatory diagram showing an example of the histogram 58 of the original image shown in FIG. 16B, and FIG. 18 shows binarization obtained by binarizing the color difference data 56 of the original image shown in FIG. It is explanatory drawing which shows an example of an image. As shown in FIG. 16A, when the lightness is too high, the tooth contact surface 46 becomes unclear due to the R channel over-range. Then, the peak of the histogram 58 disappears, and it is difficult to identify the tooth contact surface 46. For this reason, correction by lightness correction is impossible, and it is preferable to perform re-photographing with the gain of the image input board 74 lowered.

3.1 Effect of Lightness and Color Difference As described above, the RGB color image 52 is converted into the HSL space, lightness correction is performed so that the lightness L falls within the predetermined lightness range 68, and then the color difference data 56 is converted into the color difference data 56. Since the tooth contact surface 46 is extracted based on the color difference data 56 calculated, it is possible to reduce the influence of the illumination illuminance condition and the intensity of light and darkness, and to increase the sensitivity with which the color difference can be identified. Thus, the tooth contact surface can be accurately identified, and in particular, the area and the center of gravity can be stably measured.

DESCRIPTION OF SYMBOLS 12 Camera 13 Lighting apparatus 14 Tooth contact surface detection part 16 Color feature value specific process 18 Color difference data generation process 20 Tooth contact surface identification process 24 Image processing part 26 Color coordinate conversion part 28 Lightness correction process part 32 Gain adjustment part 34 Gain adjustment process 40 Paint 42 Tooth surface 44 Tooth tip 46 Tooth contact surface 48 Non-tooth contact surface 50 Measuring tooth 52 Color image 54 Color feature value 56 Color difference data 58 Histogram 60 Binary data 62 Threshold value 68 Brightness range 72 CPU
74 Image input board 76 Main storage device 78 Auxiliary storage device 80 Drive mechanism 90 Drive pinion 91 Drive input 92 Drive shaft 94 Differential case 96 Ring gear 98 Gear carrier δ Color difference Δ Simple color difference L Lightness H Hue S Saturation

Claims (6)

A camera that captures a color image of a tooth surface to which a paint has been applied;
A tooth contact surface detection unit that detects a tooth contact surface at the tooth surface based on the color image;
This tooth contact surface detector
A color feature value specifying process for specifying a color feature value to be a feature value of the tooth surface color from the color image;
Color difference data generation processing for generating color difference data by calculating a color difference with the color feature value for each color value of the tooth surface,
A tooth contact surface identification process for calculating a threshold value from a histogram of the color difference data and binarizing the color difference data to identify the tooth contact surface;
A tooth contact surface detecting device. When the color difference data generation processing sets each color value of the color image as lightness L, hue H, saturation S, and color values of the color feature values as lightness Lo, hue Ho, saturation S0, the following formula (1) )
The tooth contact surface detecting device according to claim 1, wherein the color difference is calculated by: In the color difference data generation process, each color value of the color image is set to lightness L, hue H, and saturation S, and the color value of the color feature value 54 is set to lightness Lo, hue Ho, and saturation S0, and the coefficient is α, When β and γ, the following formula (2)
The tooth contact surface detecting device according to claim 1, wherein a simple color difference is calculated by: A camera that captures a color image of a tooth surface to which a paint has been applied;
A color coordinate conversion unit that converts the color space of the color image and calculates the brightness of the color image;
A lightness correction processing unit that corrects the lightness or luminance of the tooth surface toward the lightness of the color image within a predetermined lightness range;
A tooth-contact surface detection unit that detects a tooth-contact surface at the tooth surface based on the color image whose brightness has been corrected,
This tooth contact surface detector
A color feature value specifying process for specifying a color feature value to be a feature value of the tooth surface color from the color image;
Color difference data generation processing for generating color difference data by calculating a color difference with the color feature value for each color value of the tooth surface,
A tooth contact surface identification process for calculating a threshold value from a histogram of the color difference data and binarizing the color difference data to identify the tooth contact surface;
A tooth contact surface detecting device. An imaging process for generating a color image by imaging a tooth surface to which a paint is applied in color;
A color coordinate conversion step of converting the color space of the color image and calculating the brightness of the color image;
A lightness correction step of correcting the lightness or brightness of the tooth surface toward the lightness of the color image within a predetermined lightness range;
A color feature value specifying step for specifying a color feature value to be a feature value of the color of the tooth surface from the color image whose brightness has been corrected;
A color difference data generation step for generating color difference data by calculating a color difference with the color feature value for each color value of the tooth surface;
A tooth contact surface identifying step for calculating a threshold value from a histogram of the color difference data and binarizing the color difference data to identify the tooth contact surface;
A tooth contact surface detection method comprising:   A tooth contact surface detection program for executing the method according to claim 5 using a CPU.