JP6909467B2 - Amber Fragment Type and Transparency Estimator, Estimator, and Estimator Program - Google Patents

Description

The present invention relates to a method for estimating the type and transparency of amber fragments, an estimation device, and an estimation program.

Amber is a fossilized resin of a plant that thrived in ancient times. Amber has been used for jewelry for a long time, and in recent years, the use of amber has been diversified due to the development of new processing technology and the discovery of anti-allergic components. In Japan, the area around Kuji City, Iwate Prefecture has long been famous as an amber producing area, and plays a part in the major industries in this area. On the other hand, it is important to discriminate the color and quality of amber depending on the purpose of use, but the current situation is that amber is discriminated by the human eye. Therefore, automatically discriminating the color and quality of amber using a computer contributes to cost reduction and speeding up of processing.

As a technique relating to a method for classifying gemstones, for example, Patent Document 1 describes a method for automating the classification of diamond gemstones, which has characteristics corresponding to absorption by water vapor and diamonds from an infrared absorption sample spectrum of the gemstones. The step of subtracting the properties corresponding to the intrinsic absorption by the lattice; the step of analyzing the sample spectrum to identify the given absorption properties corresponding to the diamond lattice defects; the step of classifying the gemstones according to the presence and / or strength of the given absorption properties. Steps; and methods having steps to store the classification results in a database are disclosed.
Further, as a technique related to amber, for example, Patent Document 2 describes a method for producing recycled amber, which is formed by heating and pressurizing a raw material of powdery amber, and the average bending resistance of the molded product when molded alone. A method for producing recycled amber is disclosed, which comprises using a plurality of raw material lots having different forces and mixing these raw material lots to form a mixture, and molding the mixture.

Special Table 2015-591579 JP-A-2015-193132

However, in Patent Documents 1 and 2, the type and transparency of amber are not estimated, and as far as the present inventors have investigated, there is no existing technique.

Therefore, it is an object of the present invention to provide an estimation method, an estimation device, and an estimation program capable of estimating the type and transparency of amber fragments.

The first aspect of the present invention is an upper color image acquisition step of irradiating an amber fragment resting on a horizontal plane with light from above to acquire an upper color image, and a masking process on the upper color image to obtain an upper mask image. In the upper mask image creation step, the upper mask image analysis step, and the upper mask image analysis step, which analyzes the upper mask image and acquires information on the RGB component and HSV component of the pixels of the amber fragment region in the upper mask image. From the ratio of pixels in which the average value of the obtained R component and the average value of the V component are equal, and the average value of the S component, the outer skin / amber discrimination step for discriminating between the outer skin and the amber, and the outer skin / amber Targeting the pixels of the amber fragment area of the upper mask image in which the amber fragment identified as amber in the discrimination process is captured, the degree of attribution to each amber of the amber type consisting of yellow amber, brown amber and black amber is calculated, and the degree of attribution is calculated. Includes an amber type determination step that discriminates amber fragments to the highest amber type, and an amber transparency estimation step that estimates the transparency of amber fragments determined to be yellow amber in the amber type determination step using fuzzy inference. , Amber fragment type and transparency estimation method.

In the present invention, the "exodermis" means amber fragments having impurities attached to the surface. Further, in the present invention, the "RGB component" means the R component, the G component, and the B component in the RGB color system. Further, in the present invention, the "HSV component" means the H component, the S component, and the V component in the HSV color system.

In the first aspect of the present invention, the upper mask image creation step is a grayscale step of grayscale the upper color image to obtain a grayscale image and a binary step of binarizing the grayscale image to obtain a binary image. It is preferable to include a conversion step, a complement / noise removal step of complementing and removing noise of the binary image, and an inversion processing step of inverting the binary image obtained in the complement / noise removing step.

In the first aspect of the present invention, the amber type determination step uses a predetermined average value of the R components of yellow amber, brown amber and black amber and an average value of the R component obtained in the upper mask image analysis step. Using the R component approximation calculation step that calculates the approximation of the components, the most frequent value of the H component of yellow amber, brown amber and black amber determined in advance, and the most frequent value of the H component obtained in the upper mask image analysis step. The degree of attribution of the pixels of the amber fragment region to yellow amber, brown amber and black amber based on the H component approximation calculation step for calculating the approximation of the H component and the approximation of the R component and the H component. It is preferable to include a step of calculating the degree of attribution for calculating the amber and a step of discriminating the amber fragment to discriminate the type of amber having the highest degree of attribution.

In the first aspect of the present invention, the yellow amber transparency estimation step includes a lower color image acquisition step of irradiating amber fragments determined to be yellow amber in the amber type determination step with light from below to acquire a lower color image. Lower mask image creation process that masks the lower color image to create the lower mask image, and lower mask image analysis that analyzes the lower mask image and acquires information on the RGB components of the pixels in the amber fragment region in the lower mask image. A plurality of predetermined values based on the average value of the G component and the average value of the B component obtained in the step and the lower mask image analysis step, and based on the predetermined average value of the G component and the average value of the B component of the amber. The fuzzy logic product of the conformity calculation process that calculates the conformity of the inference rule consisting of the above rules to each rule using the preamble membership function, and the conformity to each rule and the postpart membership function. Including a drawing step of creating a figure which is an inference result and a transparency calculation step of calculating the transparency of amber fragments as the final inference result by performing non-fuzzy logic using Mamdani's inference method to obtain the center of gravity of the figure. Is preferable.

In the first aspect of the present invention, the antecedent part of the inference rule comprises the following A1 to A4.
A1: The average value of the G component is low and the average value of the B component is low. A2: The average value of the G component is low and the average value of the B component is high. The average value of the B component is low A4: The average value of the G component is high and the average value of the B component is high The latter part of the inference rule consists of the following B1 to B3.
B1: Low transparency B2: High transparency or low transparency B3: High transparency The inference rule may consist of _{the following rules δ 1 to δ 4} .
δ ₁ : If the antecedent part is A1, the consequent part is B1 δ ₂ : If the antecedent part is A2, the consequent part is B2 δ ₃ : If the antecedent part is A3, The consequent part is B2 δ ₄ : If the antecedent part is A4, the consequent part is B3

In the first aspect of the present invention, it is preferable that the illuminance on the amber surface of the light irradiated in the upper color image acquisition step is 500 lpx or more and 2500 lpx or less.

In the first aspect of the present invention, it is preferable that the illuminance on the amber surface of the light irradiated in the lower color image acquisition step is 500 xl or more and 1000 lp or less.

A second aspect of the present invention has a horizontal plane on which the amber shards can be placed, a light source capable of irradiating the amber shards placed on the horizontal plane with light from above, and an imaging means, and has an upper portion of the amber slab. An upper color image acquisition unit that can acquire a color image, an upper mask image creation unit that masks the upper color image to create an upper mask image, and an upper mask image analysis that analyzes the amber fragment region in the upper mask image. The proportion of pixels in which the average value of the R component and the average value of the V component obtained by the upper mask image analysis unit that acquires information on the RGB component and HSV component of the pixel are equal, and the S component From the average value, target the pixels of the amber fragment area of the upper mask image in which the outer skin / amber discriminating unit that discriminates between the outer skin and the amber and the amber fragment that is discriminated as amber by the outer skin / amber discriminating unit are shown. The amber type discriminating unit, which calculates the degree of attribution to each amber of the amber type consisting of yellow amber, brown amber, and black amber, and discriminates the amber fragments to the amber type with the highest attribution, and the amber type discriminating unit, and the amber and amber type discriminating part. It is an amber fragment type and transparency estimation device including a yellow amber transparency estimation unit that estimates the transparency of the determined amber fragment using fuzzy inference.

A third aspect of the present invention is a step of irradiating an amber fragment resting on a horizontal plane with light from above to acquire an upper color image, and a step of applying mask processing to the upper color image to create an upper mask image. And, the step of analyzing the upper mask image and acquiring the information of the RGB component and the HSV component of the pixel of the amber fragment region in the upper mask image, and the average value of the R component and the average of the V component obtained in the upper mask image analysis step. Amber in the upper mask image showing the amber fragments identified as amber in the process of discriminating between the outer skin and amber from the ratio of pixels with the same value and the average value of the S component and the outer skin / amber discrimination step. The process of calculating the degree of attribution to each amber of the amber type consisting of yellow amber, brown amber and black amber for the pixels of the fragment area, and determining the amber fragment to the amber type with the highest degree of attribution, and the amber type discrimination This is an estimation program for the type and transparency of amber fragments to cause a computer to execute a process of estimating the transparency of amber fragments determined to be yellow amber in the process using fuzzy inference.

According to the present invention, it is possible to provide an estimation method, an estimation device, and an estimation program capable of estimating the type and transparency of amber fragments.

It is a flowchart explaining the estimation method S10 which concerns on one Embodiment of the type and transparency estimation method of the amber fragment of this invention. It is a figure which shows typically the state of the upper color image acquisition process S1. It is a flowchart explaining one Embodiment of the upper mask image analysis process S2. It is a flowchart explaining one Embodiment of amber type determination process S5. It is a flowchart explaining one Embodiment of the yellow amber transparency estimation step S6. It is a figure which shows typically the state of the lower color image acquisition process S61. It is a figure which shows an example of the antecedent part membership function used in the goodness-of-fit calculation process S64. It is a figure which shows an example of the consequent part membership function used in the drawing process S65. And adaptability to Rule δ ₁ ~δ ₄ shown in Table 4 is a diagram showing the graphic to be created as a result of inference ε ₁ ~ε ₄ from fuzzy logic product of the consequent membership functions shown in FIG. FIG. 5 is a diagram showing an example in which the center of gravity of the figure shown in FIG. 9 is obtained and the transparency (ε _{0 = 81.9%) of the amber fragment is calculated as the final inference result.} It is a figure explaining the estimation apparatus 100 which concerns on one Embodiment of the type and transparency estimation apparatus of the amber fragment which concerns on the 2nd aspect of this invention.

The above-mentioned actions and gains of the present invention will be clarified from the embodiments for carrying out the invention described below. Hereinafter, embodiments of the present invention will be described with reference to the drawings. The forms shown below are examples of the present invention, and the present invention is not limited to these forms.

1. 1. Method for Estimating Type of Amber Fragment and Transparency FIG. 1 is a flowchart illustrating an estimation method S10 according to an embodiment of a method for estimating the type and transparency of amber fragment according to the first aspect of the present invention. The estimation method S10 will be described below with reference to FIG.
As shown in FIG. 1, the estimation method S10 includes an upper color image acquisition step S1, an upper mask image creation step S2, an upper mask image analysis step S3, an outer skin / amber discrimination step S4, and an amber type discrimination step S5. And the yellow amber transparency estimation step S6.

1.1. Upper color image acquisition process S1
The upper color image acquisition step S1 (hereinafter, may be referred to as “S1”) is a step of irradiating amber crushed pieces resting on a horizontal plane with light from above to acquire an upper color image. FIG. 2 schematically shows the state of S1. As shown in FIG. 2, the light source 2 arranged above the amber crushed piece 1 in the vertical direction (vertical direction on the paper surface) irradiates the amber crushed piece 1 resting on the horizontal plane toward the amber crushed piece 1. Will be done. In this state, an upper color image is acquired by taking a picture with the image pickup means 3 arranged vertically above the amber fragment 1. The member forming the horizontal plane has a horizontal plane, and the material thereof is not particularly limited as long as the amber fragments can be allowed to stand still. The horizontal plane may have translucency, but when the image is taken on a transparent plate, the acquired image features may change due to the reflection of the illumination light on the plate. It is preferably a white surface to be formed. The imaging means 3 is not particularly limited as long as it can acquire an upper color image, and for example, a visible region camera or the like can be used. Further, the photographing is preferably performed in a dark room. The number, position, and direction of light irradiation for irradiating the amber fragments with light are not particularly limited. For example, as shown in FIG. 2, light may be irradiated from an oblique direction with respect to the vertical direction. From the viewpoint of preventing the entire image from becoming dark, the illuminance on the surface of the amber fragment is preferably 500 lpx or more. On the other hand, from the viewpoint of preventing the amount of overexposure from increasing, the illuminance on the surface of the amber crushed piece is preferably 2500 xl or less. The illuminance can be appropriately adjusted by adjusting the output of the LED lighting or using a dimming film.

1.2. Upper mask image creation process S2
The upper mask image creation step S2 (hereinafter, may be referred to as “S2”) is a step of applying a mask process to the upper color image acquired in S1 to create an upper mask image. FIG. 3 shows a flowchart illustrating an embodiment of the upper mask image creating step S2. As shown in FIG. 3, the upper mask image creation step S2 includes a grayscale conversion step S21, a binarization step S22, a complement / noise removal step S23, and an inversion processing step S24, and the upper color image is It is preferable to perform the mask treatment according to the steps S21 to 24.

1.2.1. Grayscale process S21
The grayscale step S21 (hereinafter, may be referred to as “S21”) is a step of grayscale the upper color image acquired in S1 to obtain a grayscale image. The method of grayscale the upper color image is not particularly limited as long as it can be grayscaled. For example, using a pixel of the acquired upper color image (x, y) in S1, RGB component _{C xy (R = P r,} G = P g, B = P b) based on the RGBtoYCC conversion formula The brightness value V _xy can be calculated to create a grayscale image with 256 gradations. The RGB to YCC conversion formula is shown in the following formula (1).

1.2.2. Binarization step S22
The binarization step S22 (hereinafter, may be referred to as “S22”) is a step of binarizing the grayscale image obtained in S21 to obtain a binary image. The method of binarizing the grayscale image is not particularly limited as long as the binarization is possible. _{For example, a pixel (0 to 255) in which the brightness value V xy} (0 to 255) of the pixel (x, y) of the grayscale image obtained in S21 is equal to or greater than the threshold Thr is defined as a white pixel (255), and in other cases, a black pixel. By setting (0), a binary image can be created. The conditions for creating a binary image are shown in Eq. (2) below. The threshold value Thr at the time of creation may be set manually for each image, or may be automatically set using the following equation (3). In equation (3), V _μ is the average value of lightness in the grayscale image, and V _σ is the standard deviation value of lightness in the grayscale image.

1.2.3. Complementary / noise removal process S23
Complementing / noise removing step S23 (hereinafter, may be referred to as “S23”) is a step of complementing and removing noise of the binary image obtained in S22. The method for complementing and removing noise from a binary image is not particularly limited as long as complementation and noise removal are possible. For example, it can be performed by performing an opening process and a closing process on the binary image obtained in S22. Specifically, as the opening process, the shrinkage process is performed and then the expansion process is performed the same number of times, and as the closing process, the expansion process is performed and then the shrinkage process is performed the same number of times. The number of times of each process is changed from 0 to 10 times by one time, and the number of times determined to be optimal for each original image can be appropriately adopted.

1.2.4. Inversion processing step S24
The inversion processing step S24 (hereinafter, may be referred to as “S24”) is a step of performing an inversion process on the binary image after complementation and noise removal obtained in S23. For example, in S22, when a binary image composed of white pixels (255) and black pixels (0) is created as described above, the binary image to which complementation and noise removal are performed by S23 can be obtained. The black pixels are the amber crushed pieces, and the white pixels are the background. Therefore, in S24, by performing an inversion process in which the black pixel (0) is changed to the white pixel (255) and the white pixel (255) is changed to the black pixel (0), the black pixel is the background portion and the white pixel is the amber fragment portion. It is possible to create an upper mask image that is.

1.3. Upper mask image analysis step S3
The upper mask image analysis step S3 (hereinafter, may be referred to as “S3”) analyzes the upper mask image obtained in S2 and may refer to the pixels of the amber fragment region in the upper mask image (hereinafter, referred to as amber pixels). It is a step of acquiring the information of the RGB component and the HSV component of (there is). Hereinafter, a method of acquiring information on RGB components and a method of acquiring information on HSV components will be described in order.

13.1. RGB component information acquisition In S3, in order to acquire the RGB component information, first, the amber pixels existing in the upper mask image are extracted. Next, the R component histogram, the G component histogram, and the B component histogram are calculated based on the RGB components of the extracted amber pixels. Based on the calculated RGB histogram, the average value μ, the variance value σ ² , and the mode value are calculated, and these information are acquired as feature quantities. When the value (gradation value) on the X-axis of the histogram is x and the Y component (frequency) at x is Y (x), the formula for calculating the average value μ is dispersed in the following formula (4). The formula for calculating the value σ ² is shown in the following formula (5).

1.3.2. Acquisition of HSV component information The HSV color system can specify colors that are close to human senses, and is represented by three components: H (hue), S (saturation), and V (brightness). That is, since the HSV color system is easy to handle as a color feature, it is considered to be useful as a color feature in amber. The HSV component can be calculated using a conversion formula from the RGB component. The conversion formula used for the color space conversion from the RGB color system to the HSV color system is shown in the following formula (6).

The RGB component is converted to the HSV component using the above formula (6), and the H component histogram, the S component histogram, and the V component histogram are calculated, respectively. Based on the HSV histogram, the average value μ, the variance value σ ² , and the mode value are calculated, and these information are acquired as feature quantities. Similar to the calculation of the feature amount of the RGB histogram, the average value μ is ^{calculated by using the above formula (4), and the variance value σ 2} is calculated by using the above formula (5). Since the average value μ and the dispersion value σ ² of the H component in the calculation method do not consider that the H component is cyclic data, only the mode value is used as the feature amount in the H component.

1.4. Exodermis / amber discrimination step S4
In the hull / amber discrimination step S4 (hereinafter, may be referred to as “S4”), the average value of the R component and the average value of the V component obtained in S3 are equal (hereinafter, “R = V””. It is a step of discriminating between the outer skin and amber from the ratio of pixels and the average value of the S component.

In S4, the amber and the exodermis can be discriminated by calculating the ratio of the pixels in which R = V, that is, MAX (R, G, B) = R in the amber pixels. Specifically, among the amber pixels in the target image, those having a high ratio of R = V can be discriminated as amber, and those having a low ratio can be discriminated as having a possibility of exodermis. The threshold value for the ratio of R = V can be appropriately set, and is preferably 75%, for example.

Only by discriminating the exodermis using the R component and the V component, amber such as black amber, which can have a low ratio of R = V, is discriminated as the exodermis, and the discrimination accuracy is lowered due to this. Therefore, for the purpose of improving the discrimination rate of each amber and exodermis, it is possible to discriminate the exodermis by setting a threshold value for the S component. Specifically, the average value of the S component of the amber fragment region in the target image can be calculated, and if the calculated average value exceeds the threshold value, it can be determined as amber, and if it is less than the threshold value, it can be determined as an exodermis. The threshold value of the S component can be appropriately set, and is preferably 32.9, for example.

1.5. Amber type determination step S5
The amber type determination step S5 (hereinafter, may be referred to as “S5”) targets the pixels (amber pixels) in the amber fragment region of the upper mask image in which the amber fragments determined to be amber in S4 are captured. This is a step of calculating the degree of attribution to each amber of the amber type consisting of tea amber and black amber, and discriminating the amber fragment to the amber type having the highest degree of attribution. FIG. 4 shows a flowchart illustrating one embodiment of S5. As shown in FIG. 4, the amber type discrimination step S5 preferably includes an R component approximation degree calculation step S51, an H component approximation degree calculation step S52, an attribution degree calculation step S53, and a discrimination processing step S54. ..

1.5.1. R component approximation degree calculation step S51
In the R component approximation degree calculation step S51 (hereinafter, may be referred to as “S51”), a predetermined average value of the R components of yellow amber, brown amber and black amber and an average value of the R component obtained in S3 are obtained. This is a step of calculating the degree of approximation of the R component using the product.
The average value of the R components of the predetermined yellow amber, brown amber and black amber is μ _a , and the average value of the R component of the amber fragment region in the target image is μ _b . From these two average values, the following equation (7) The degree of approximation α (0 to 1.0) is calculated by using.

In the formula (7), A is the average value of the R component (A = μ _a ) determined in advance, B is the average value of the R component of the amber fragment region in the target image (B = μ _b ), and C is the allowable range error. be. For the predetermined average value (A = μ _a ) of the R component, perform the above S1 to S51 for a reliable number of yellow amber, brown amber and black amber shards as a control group (control). It can be a calculated value.

1.5.2. H component approximation degree calculation step S52
In the H component approximation degree calculation step S52 (hereinafter, may be referred to as “S52”), the mode of the H component of yellow amber, tea amber and black amber determined in advance and the mode of the H component obtained in S3 are the most frequent. This is a step of calculating the degree of approximation of the H component using the value.

This is a step of calculating the degree of approximation of the H component using the mode of the H component of yellow amber, brown amber and black amber determined in advance and the mode value of the H component obtained in S3. Predetermined yellow amber, the mode of the H component of tea amber and black amber m _a, the mode of the H component of succinic debris area in the target image and m _b, the following equation from these two most frequent The degree of approximation β (0 to 1.0) is calculated using (8).

In the formula (8), A is the most frequent value of a predetermined H component (A _{= m} a), B is the H component amber debris area of the target image mode value (B _{= m} b), C is acceptable It is a range error. Predetermined H component of the mode (B = m _b) is yellow amber numbers reliable as control group (control), as a target amber debris group tea amber and black amber, perform the above S1~S52 , Can be a calculated value.

1.5.3. Attribution calculation step S53
In the attribution calculation step S53 (hereinafter, may be referred to as “S53”), the pixels of the amber fragment region (amber) are based on the degree of approximation of the R component obtained in S51 and the degree of approximation of the H component obtained in S52. This is a step of calculating the degree of attribution of (pixels) to yellow amber, brown amber, and black amber. Specifically, from the approximation degree α calculated in S51 and the approximation degree β calculated in S52, the degree of attribution γ (0 to 1.0) to yellow amber, brown amber, and black amber is calculated by the following equation (9). It can be calculated using the formula.

1.5.4. Discrimination processing step S54
The discrimination processing step S54 (hereinafter, may be referred to as “S54”) is a step of discriminating amber fragments according to the type of amber having the highest degree of attribution. The degree of attribution to each amber calculated in S53 is compared, and amber fragments are discriminated according to the amber type having the maximum value. For example, when the results shown in Table 1 below are obtained in S53, the type of amber fragment is determined to be yellow amber.

1.6. Yellow amber transparency estimation step S6
The yellow amber transparency estimation step S6 (hereinafter, may be referred to as “S6”) is a step of estimating the transparency of the amber fragment determined to be yellow amber in S5 by using fuzzy inference. Since fuzzy inference is an inference method that takes ambiguity into consideration, it is considered to be useful for estimating the transparency of amber in which the change in gradation is indefinite. FIG. 5 shows a flowchart illustrating one embodiment of S6. As shown in FIG. 5, the yellow amber transparency estimation step S6 includes a lower color image acquisition step S61, a lower mask image creation step S62, a lower mask image analysis step S63, a conformity calculation step S64, and a drawing step S65. It is preferable to include the transparency calculation step S66.

16.1. Lower color image acquisition process S61
The lower color image acquisition step S61 (hereinafter, may be referred to as “S61”) is a step of irradiating the amber fragment determined to be yellow amber in S5 with light from below to acquire a lower color image. FIG. 6 schematically shows the state of S61. As shown in FIG. 6, light is emitted from the light source 2 arranged below the amber crushed piece 1 in the vertical direction (vertical direction of the paper surface) to the amber crushed piece 1 with respect to the amber crushed piece 1 standing on a horizontal plane. Be irradiated. In this state, a lower color image is acquired by taking a picture with the imaging means 3 arranged vertically below the amber fragment 1. The member forming the horizontal plane needs to have a horizontal plane and allow the amber shards to stand still, and to be able to acquire a lower color image of the amber shard 1 from the lower part of the horizontal plane, so that it has translucency. Therefore, it is preferable that the member is transparent or translucent and transmits light. The light source 2 and the imaging means 3 can be used in the same manner as in S1, except that the arrangement is different. From the viewpoint of preventing the entire image from becoming dark, the illuminance on the surface of the amber fragment is preferably 500 lpx or more. On the other hand, from the viewpoint of preventing the amount of overexposure from increasing, it is preferably 1000 lpx or less. In S61, the member forming the horizontal plane has translucency, and the light transmitted through the member is used for photographing, so that overexposure is more likely to occur than in S1. Therefore, the upper limit value of the preferable illuminance in S61 is set lower than the upper limit value of the preferable illuminance in S1.

16.2. Lower mask image creation step S62
The lower mask image creating step S62 (hereinafter, may be referred to as “S62”) is a step of applying mask processing to the lower color image obtained in S61 to create a lower mask image. The method of mask processing applied to the lower color image can be the same as the process applied to the upper color image in S11 to 14 described above, and description thereof will be omitted here.

1.6.3. Lower mask image analysis step S63
The lower mask image analysis step S63 (hereinafter, may be referred to as “S63”) analyzes the lower mask image obtained in S62, and provides information on the RGB component of the pixels (amber pixels) in the amber fragment region in the lower mask image. Is the process of acquiring. The method of acquiring the RGB component information of the amber pixel in the lower mask image can be the same as the method of acquiring the RGB component information of the amber pixel in the upper mask image in S3 described above, and the description thereof is omitted here. ..

1.6.4. Goodness of fit calculation process S64
The goodness-of-fit calculation step S64 (hereinafter, may be referred to as “S64”) is based on the average value of the R component and the average value of the B component obtained in S63, and a predetermined average value of the G component of yellow amber and the average value of the G component. This is a step of calculating the goodness of fit of an inference rule consisting of a plurality of predetermined rules to each rule by using the membership function of the preceding part, based on the average value of the B component.

Table 2 shows an example of the inference rule predetermined in S64.

In S64, first, how much the G component and the B component of the target image belong to the antecedent part 1 and the antecedent part 2 in the inference rule is calculated as a grade, and an ambiguous proposition is quantified. Specifically, in the antecedent part 1, the proposition "the average value of the G component is low" is set to G1, and the proposition "the average value of the G component is high" is set to G2, and the respective grades are calculated. Next, in the antecedent part 2, as in the antecedent part 1, the proposition that "the average value of the B component is low" is set to B1 and the proposition that "the average value of the B component is high" is set to B2.

The antecedent membership function can be, for example, the step function shown in FIG. In FIG. 7, x is the average value of the G component and the B component of the amber fragment region in the target image, A is the average value of the G component and the B component in the predetermined transparent yellow amber, and B is the predetermined opaque value. It is an average value of G component and B component in yellow amber. The average value of G component and B component in the predetermined transparent yellow amber and the average value of G component and B component in the predetermined opaque amber are the reliable number of yellow amber as a control group (control). , Tea amber and black amber crushed pieces of amber can be subjected to the above steps S61 to S63 to obtain a calculated numerical value.

Next, using the fuzzy logical product shown in the following equation (10), the goodness of fit to each rule from the grades of the antecedent part 1 and the antecedent part 2 obtained by the antecedent part membership function δ _i (i = 1 to 4) are calculated.

Table 3 shows an example of calculating the grades of the antecedent part 1 and the antecedent part 2 in S64, and Table 4 shows an example of calculating the degree of conformity with each rule.

1.6.5. Drawing process S65
The drawing process S65 (hereinafter, may be referred to as “S65”) creates a figure which is an inference result from the goodness of fit to each rule obtained in S64 and the fuzzy logical product of the consequent membership function. It is a process to do. FIG. 8 shows an example of the consequent membership function that can be used in this step. _{Further, the inference result ε i} (i = 1 to 4) is obtained from the fuzzy logical product of the goodness of fit to the _{rule δ i} (i = 1 to 4) shown in Table 4 and the consequent part membership function shown in FIG. The figure to be created is shown in FIG.

1.6.6. Transparency calculation step S66
In the transparency calculation step S66 (hereinafter, may be referred to as “S66”), non-fuzzying is performed using the reasoning method of Mamdani, the center of gravity of the figure created in S66 is obtained, and the transparency of the amber fragment is obtained as the final reasoning result. Is the process of calculating.
The equation used in Mamdani's inference method is shown in the following equation (11).

ここで、ε_０は最終推論結果、ｗ_ｊ（ｊ＝１〜３）は重み係数である。式（１１）を用いて最終推論結果として透明度を算出する。図９に示す図形の重心を求め、最終推論結果として琥珀砕片の透明度（ε_０＝８１．９％）を算出した例を図１０に示す。

Here, ε ₀ is the final inference result, and w _j (j = 1-3) is the weighting coefficient. Transparency is calculated as the final inference result using equation (11). FIG. 10 shows an example in which the center of gravity of the figure shown in FIG. 9 is obtained and the transparency (ε _{0 = 81.9%) of the amber fragment is calculated as the final inference result.}

2. Amber Fragment Type and Transparency Estimating Device FIG. 11 is a diagram illustrating an estimation device 100 according to an embodiment of an amber fragment type and transparency estimating device according to a second aspect of the present invention. The estimation device 100 will be described below with reference to FIG.
As shown in FIG. 11, the estimation device 100 includes an upper color image acquisition unit 10, an upper mask image creation unit 20, an upper mask image analysis unit 30, an outer skin / amber discrimination unit 40, and an amber type discrimination unit 50. And the amber transparency estimation unit 60. The estimation device 100 is a device capable of carrying out the estimation method according to the first aspect of the present invention.

2.1. Upper color image acquisition unit 10
The upper color image acquisition unit 10 is a portion where the above S1 is performed, and as illustrated in FIG. 1, light is emitted from above to a horizontal plane on which the amber crushed piece 1 can be placed and an amber crushed piece 1 placed on the horizontal plane. It has a light source 2 capable of irradiating and an imaging means 3, and is a portion capable of acquiring an upper color image of amber fragments. The above S1 is performed in the upper color image acquisition unit 10, and the acquired information on the upper color image is sent to the upper mask image creation unit 20.

2.2. Upper mask image creation unit 20
The upper mask image creation unit 20 is a portion where the above S2 is performed, and based on the information regarding the upper color image sent from the upper color image acquisition unit 10, the upper color image is masked to create an upper mask image. It is a part. The above S2 is performed in the upper mask image creating unit 20, and the information regarding the created upper mask image is sent to the upper mask image analysis unit 30.

As shown in FIG. 11, the upper mask image creation unit 20 has a grayscale unit 21, a binarization unit 22, a complement / noise removal unit 23, and an inversion processing unit 24, and has an upper portion. It is preferable that the color image is masked by the upper mask image creating unit 20 to the inversion processing unit 24.

2.2.1. Grayscale section 21
The grayscale unit 21 is a part where the above S21 is performed, and is a part where the upper color image is grayscaled and a grayscale image is created based on the information about the upper color image sent from the upper color image acquisition unit 10. be. The above S21 is performed in the grayscale unit 21, and the information regarding the created grayscale image is sent to the binarization unit 22.

2.2.2. Binarization unit 22
The binarization unit 22 is a portion where the above S22 is performed, and is a portion for binarizing the grayscale image and creating a binary image based on the information regarding the grayscale image sent from the grayscale conversion unit. The above S22 is performed in the binarization unit 22, and the information regarding the created binary image is sent to the complement / noise removal unit 23.

2.2.3. Complementary / noise removing unit 23
The complement / noise removal unit 23 is a portion where the above S23 is performed, and is a portion where the binary image is complemented and noise is removed based on the information regarding the grace binary image sent from the binarization unit 22. The above S23 is performed in the complement / noise removal unit 23, and the information regarding the binary image after the complement / noise removal is sent to the inversion processing unit 24.

2.2.4. Inversion processing unit 24
The inversion processing unit 24 is a portion where the above S24 is performed, and is inverted to a binary image after completion and noise removal based on the information regarding the binary image after completion and noise removal sent from the complement / noise removal unit 23. This is the part that is processed to create the upper mask image. The above S24 is performed in the inversion processing unit 24, and the information regarding the created upper mask image is sent to the upper mask image analysis unit 30.

2.3. Upper mask image analysis unit 30
The upper mask image analysis unit 30 is a part where the above S3 is performed, analyzes the upper mask image based on the information about the upper mask image sent from the upper mask image creation unit 20, and RGB of the amber pixels in the upper mask image. It is a part for acquiring information on components and HSV components. The above S3 is performed in the upper mask image analysis unit 30, and the information on the RGB component and the HSV component of the amber pixel in the acquired upper mask image is sent to the outer skin / amber discrimination unit 40 and the amber type discrimination unit 50.

2.4. Exodermis / amber discrimination unit 40
The outer skin / amber discrimination unit 40 is a portion where the above S4 is performed, and the average value of the R component is based on the information of the RGB component and the HSV component of the amber pixel in the upper mask image sent from the upper mask image analysis unit 30. It is a part where the outer skin and amber are discriminated from the ratio of pixels in which the average value of the V component and the average value of the V component are equal (R = V) and the average value of the S component. When the above S4 is performed by the outer skin / amber discrimination unit 40 and it is determined to be the outer skin, information indicating that the amber fragment shown in the target image is the outer skin is sent to the estimation result display unit 70, and the amber is displayed. If it is determined, information about the upper mask image is sent to the amber type determination unit 50.

2.5. Amber type discriminating unit 50
The amber type discrimination unit 50 is a part where the above S5 is performed, and based on the information about the upper mask image sent from the outer skin / amber discrimination unit 40, the amber pixels of the upper mask image are targeted as yellow amber and brown amber. It is a part that calculates the degree of attribution to each amber of the amber type consisting of black amber and black amber, and discriminates amber fragments to the amber type with the highest degree of attribution. When the above S5 is performed by the amber type determination unit 50 and it is determined to be brown amber or black amber, information indicating that the amber fragment shown in the target image is brown amber or black amber is sent to the estimation result display unit 70. If it is determined to be yellow amber, information indicating that it is yellow amber is sent to the yellow amber transparency estimation unit 60.

As shown in FIG. 11, the amber type discrimination unit 50 includes an R component approximation degree calculation unit 51, an H component approximation degree calculation unit 52, an attribution degree calculation unit 53, and a discrimination processing unit 54. Is preferable.

2.5.1. R component approximation degree calculation unit 51
The R component approximation degree calculation unit 51 is a portion where the above S51 is performed, and is a predetermined average value of the R components of yellow amber, brown amber, and black amber, and an average value of the R component sent from the upper mask image analysis unit 30. It is a part for calculating the degree of approximation of the R component using. The R component approximation degree calculation unit 51 performs the above S51, and the calculated information on the approximation degree of the R component is sent to the attribution degree calculation unit 53.

25.2. H component approximation degree calculation unit 52
The H component approximation degree calculation unit 52 is a part where the above S52 is performed, and is the mode of the H component of the predetermined yellow amber, brown amber, and black amber, and the mode of the H component sent from the upper mask image analysis unit 30. This is a part for calculating the degree of approximation of the H component using the mode. The above S52 is performed in the H component approximation degree calculation unit 52, and the calculated information on the approximation degree of the H component is sent to the attribution degree calculation unit 53.

2.5.3. Attribution calculation unit 53
The attribution calculation unit 53 is a part where the above S53 is performed, and the information on the approximation degree of the R component sent from the R component approximation degree calculation unit 51 and the approximation degree of the H component sent from the H component approximation degree calculation unit 52 are used. Based on this, it is a part for calculating the degree of attribution of amber pixels to yellow amber, brown amber, and black amber. The attribution degree calculation unit 53 performs the above S53, and the calculated information regarding the attribution degree to each amber is sent to the discrimination processing unit 54.

2.5.4. Discrimination processing unit 54
The discrimination processing unit 54 is a part where the above S54 is performed, and is a part for discriminating amber fragments according to the type of amber having the highest degree of attribution based on the information regarding the degree of attribution to each amber sent from the degree of attribution calculation unit 53. Is. When the determination processing unit 54 performs the above S54 and determines that the amber is brown or black amber, information indicating that the amber fragment shown in the target image is brown or black amber is sent to the estimation result display unit 70. If it is determined to be yellow amber, information indicating that it is yellow amber is sent to the yellow amber transparency estimation unit 60.

2.6. Yellow amber transparency estimation unit 60
The yellow amber transparency estimation unit 60 is a portion where the above S54 is performed, and is a portion for estimating the transparency of the amber fragment determined to be yellow amber by the amber type determination unit 50 by using fuzzy inference. S6 is performed in the yellow amber transparency estimation unit 60, and the transparency of the amber fragment is calculated.

As shown in FIG. 11, the yellow amber transparency estimation unit 60 includes a lower color image acquisition unit 61, a lower mask image creation unit 62, a lower mask image analysis unit 63, a goodness of fit calculation unit 64, and a drawing unit. It is preferable to include 65 and the transparency calculation unit 66.

2.6.1. Lower color image acquisition unit 61
The lower color image acquisition unit 61 is a portion where the above S61 is performed, and as illustrated in FIG. 6, light is emitted from above to a horizontal plane on which the amber crushed piece 1 can be placed and an amber crushed piece 1 placed on the horizontal plane. It has a light source 2 capable of irradiating and an imaging means 3, and is a portion capable of acquiring a lower color image of amber fragments. The lower color image acquisition unit 61 performs the above S61, and the acquired information on the lower color image is sent to the lower mask image creation unit 62.

2.6.2. Lower mask image creation unit 62
The lower mask image creation unit 62 is a portion where the above S62 is performed, and based on the information regarding the lower color image sent from the lower color image acquisition unit 61, the lower color image is masked to create the lower mask image. It is a part. The above S62 is performed in the lower mask image creating unit 62, and the information regarding the created lower mask image is sent to the lower mask image analysis unit 63.

2.6.3. Lower mask image analysis unit 63
The lower mask image analysis unit 63 is a part where the above S63 is performed, analyzes the lower mask image based on the information about the lower mask image sent from the lower mask image creation unit 62, and RGB of the amber pixels in the lower mask image. This is the part where the information of the ingredients is acquired. The lower mask image analysis unit 63 performs the above S63, and the information regarding the RGB component of the amber pixel in the acquired lower mask image is sent to the goodness-of-fit calculation unit 64.

2.6.4. Goodness of fit calculation unit 64
The conformity calculation unit 64 is a part where the above S64 is performed, and the average value of the R component and the average of the B component are based on the information on the RGB component of the amber pixel in the lower mask image sent from the lower mask image analysis unit 63. Based on the value, the degree of conformity to each rule of the inference rule consisting of a plurality of predetermined rules is determined based on the predetermined average value of the G component and the average value of the B component of yellow amber. It is a part calculated using. The goodness-of-fit calculation unit 64 performs the above S64, and the calculated information regarding the goodness of fit to each rule is sent to the drawing unit 65.

2.6.5. Drawing department 65
The drawing unit 65 is a part where the above S65 is performed, and based on the information regarding the goodness of fit to each rule sent from the goodness-of-fit calculation unit 64, the goodness of fit to each rule and the consequent part membership function This is the part that creates the figure that is the inference result from the fuzzy logical product. The drawing unit 65 performs the above S65, and information about the created figure is sent to the transparency calculation unit 66.

2.6.6. Transparency calculation unit 66
The transparency calculation unit 66 is a part where the above S66 is performed, and based on the information about the figure sent from the drawing unit 65, non-fuzzy is performed using the inference method of Mamdani, and the center of gravity of the figure created in S66 is obtained. , This is the part that calculates the transparency of amber fragments as the final inference result. The above S66 is performed in the transparency calculation unit 66, and the information regarding the transparency of the amber fragment calculated as the final inference result is sent to the estimation result display unit 70.

2.7. Estimated result display unit 70
The estimation result display unit 70 receives information on the estimation result sent from the outer skin / amber discrimination unit 40, the amber type discrimination unit 50, or the yellow amber transparency estimation unit 60, and displays the information on the display device based on the information on the estimation result. This is the part that generates the data to be displayed and displays it on the display device. Although FIG. 11 illustrates a form in which the estimation result display unit 70 is provided separately from the estimation device 100, the estimation device 100 may be provided with the estimation result display unit 70.

3. 3. Amber fragment type and transparency estimation program The present invention also has an aspect as an estimation program. That is, the third aspect of the present invention is a program for estimating the type and transparency of amber fragments for causing a computer to execute the above S1 to S6. By causing a processor such as a CPU (central processor) included in the computer to execute the estimation program according to the third aspect of the present invention, it is possible to acquire information on the type and transparency of the amber fragment as an output result. ..

Hereinafter, the present invention will be described in more detail with reference to Examples, but the present invention is not limited to the following specific embodiments.

<Amber type discrimination test>
The estimation method of the present invention is applied to a total of 185 amber fragments consisting of 70 yellow amber foils, 40 brown ambers, 40 black ambers and 35 exodermis, which are visually determined by a conventional method in advance. , It was verified whether it is possible to distinguish the amber type.

(Upper color image acquisition process)
An upper color image of each amber fragment was acquired in a dark room using a visible region camera (Canon EOS 7D 5184 × 3456 pixels; 256 gradations of RGB each). Amber fragments were placed statically on a horizontal plane formed by placing white drawing paper on a table, and a visible region camera was installed at a height of 55 cm from the horizontal plane. Table 7 shows the illuminance at the time of acquiring the upper color image of each amber type.

(Upper mask image creation process)
The upper color image of each of the acquired amber fragments was subjected to grayscale, binarization, complementation / noise removal, and pixel inversion processing to create an upper mask image of each amber fragment. The threshold value in binarization was automatically set using the above equation (3), and the opening process and the closing process were performed 10 times each.

(Upper mask image analysis processing)
The pixels (amber pixels) of the amber fragment region existing in the created upper mask image were extracted. The threshold value of the RGB component is set to 240, and when all the RGB components in the pixels (x, y) are equal to or higher than the threshold value, it is assumed that the pixel loses color information due to overexposure, and it is not subject to analysis like the background part. And said. Based on the RGB components of the extracted amber pixels, the R component histogram, G component histogram, and B component histogram are calculated, and the average value, variance value, and mode value are calculated based on the RGB histogram, and these are the features. Obtained as a quantity. The acquired RGB component is converted to the HSV component, the H component histogram, the S component histogram, and the V component histogram are calculated respectively, and the average value, the variance value, and the mode value are calculated based on the HSV histogram. These were acquired as feature quantities.

(Exodermis / amber discrimination processing)
From the analysis result of the upper mask image of each amber fragment, the ratio of pixels in which R = V, that is, MAX (R, G, B) = R in the amber pixels, and the average value of the S component were calculated. When the threshold value of the ratio of R = V is 75%, the threshold value of the average value of the S component is 32.9, the ratio of R = V is less than 75%, and the average value of the S component is less than 32.9. Was identified as an exodermis, and in other cases it was identified as amber.

(Amber type discrimination)
Based on the R and H components of the amber fragment region in the upper mask image showing the amber fragment identified as amber, the degree of attribution to yellow amber, brown amber and black amber was calculated, and the amber type of the amber fragment was determined. .. Specifically, the degree of approximation of the R component is calculated from the average value of the R component in the amber fragment region in the upper mask image showing the amber fragment determined to be amber and the average value of the R component in each predetermined amber type. , The mode of the H component in the amber fragment region of the target image determined to be amber and the mode of the H component in each predetermined amber type are calculated, and the degree of approximation of the H component is calculated as the degree of approximation of the R component. The degree of attribution was calculated based on the degree of approximation of the H component, and amber fragments were discriminated from the amber having the highest degree of attribution. _{The values of the average value μ a} of the R component and the value of the error tolerance C in each of the predetermined amber types used for calculating the degree of approximation of the R component are determined in advance used for calculating the degree of approximation of the H component in Table 5 below. and the value of the mode value m _a and the error tolerance range C of the H component in each succinic type in Table 6 below, respectively. The discrimination results are shown in Table 7.

[result]
As shown in Table 7, 167 out of 185 amber fragments (90.3%) appearing in 185 images were successfully identified. Especially in yellow amber, 70 out of 70 (100%) were successfully discriminated. On the other hand, in the discrimination of exodermis, 20 out of 35 (57.1%) were lower in discrimination accuracy than yellow amber, brown amber and black amber. It is considered that this is because the ratio of R = V and the average value of the S component became high due to the color information of the exposed amber. As a result of measuring the processing time per target image, the average processing time was about 53 ms.

2. Yellow Amber Transparency Estimation Test The estimation method of the present invention was applied to 70 pieces of yellow amber used in the amber type discrimination test, and the estimation accuracy of the transparency of yellow amber was verified. It should be noted that, of the 70 pieces of amber crushed pieces, 50 pieces are classified as transparent yellow amber and 20 pieces are classified as opaque yellow amber by visual discrimination which is a conventional method in advance.

(Lower color image acquisition process)
The lower color image is acquired in the same way as the upper color image acquisition process, except that the visible region camera used in the upper color image acquisition process in the amber type discrimination test is installed below the horizontal plane and the horizontal plane is formed of a transparent resin plate. did.

(Lower mask image creation process)
The lower color image of each of the acquired amber fragments was subjected to the same processing as the upper mask image creation process in the amber type discrimination test to create a lower mask image.

(Lower mask image analysis processing)
The lower mask image of each of the prepared amber fragments was subjected to the same processing as the upper mask image analysis processing in the amber type discrimination test, and the information of the RGB component of the amber pixel in the lower mask image was acquired.

(Goodness of fit calculation process)
The inference rules shown in Table 2 are adopted, and based on the average value of the R component and the average value of the B component obtained by the lower mask image analysis process, the average value of the G component and the average of the B component of yellow amber are predetermined. The degree of conformity of the inference rule to each rule was calculated based on the value. Average value mu _c of the G component and the B component in the predetermined transparent yellow _amber, and indicates the average value mu _c of the G component and the B component in the predetermined transparent yellow amber in Table 8 below.

(Drawing process)
From the goodness of fit to each rule and the fuzzy logical product of the consequent membership function, a figure which is an inference result as illustrated in FIG. 9 was created.

(Transparency calculation process)
Non-fuzzy was performed using Mamdani's inference method, the center of gravity of the created figure was obtained, and the transparency of the amber fragment was calculated as the final inference result. Those having a calculated transparency of 70% or more were discriminated as transparent yellow amber, and those having a transparency of 70% or more were discriminated as opaque yellow amber. The discrimination results are shown in Table 9.

[result]
As shown in Table 9, the yellow amber transparency estimation method succeeded in discriminating 57 out of 70 target images (81.4%). Especially in the transparent yellow amber, 45 out of 50 sheets (90.0%) were successfully discriminated. Further, the average transparency of transparent yellow amber was 79.4%, and the average transparency of opaque yellow amber was 62.1%, both of which satisfied the evaluation criteria of transparency. On the other hand, in the estimation of the transparency of opaque yellow amber, there was a case where the transparency was erroneously determined to be 70% or more. It is considered that this is because the thickness of the target amber is thin and the light is easily transmitted, so that the values of the G component and the B component are high. As a result of measuring the processing time per target image, the average processing time was about 68 ms.

1 Amber shards 2 Light source 3 Imaging means 10 Upper color image acquisition unit 20 Upper mask image creation unit 21 Grayscale conversion unit 22 Binarization unit 23 Complementary / noise removal unit 24 Inversion processing unit 30 Upper mask image analysis unit 40 Outer skin / Amber discrimination unit 50 Amber type discrimination unit 51 R component approximation degree calculation unit 52 H component approximation degree calculation unit 53 Attribution degree calculation unit 54 Discrimination processing unit 60 Yellow amber transparency estimation unit 61 Lower color image acquisition unit 62 Lower mask image creation unit 63 Lower mask image analysis unit 64 Conformity calculation unit 65 Drawing unit 66 Transparency calculation unit 100 Estimator

Claims (9)

An upper color image acquisition process for acquiring an upper color image by irradiating amber fragments placed on a horizontal surface with light from above,
An upper mask image creation step of applying mask processing to the upper color image to create an upper mask image, and
An upper mask image analysis step of analyzing the upper mask image and acquiring information on RGB components and HSV components of the pixels of the amber fragment region in the upper mask image.
Distinguishing between exodermis and amber from the ratio of pixels in which the average value of the R component and the average value of the V component obtained in the upper mask image analysis step are equal to each other and the average value of the S component. Process and
Each amber type consisting of yellow amber, brown amber, and black amber is targeted at the pixels of the amber fragment region of the upper mask image in which the amber fragment identified as amber in the outer skin / amber discrimination step is captured. An amber type determination step of calculating the degree of attribution and determining the amber fragment as the amber type having the highest degree of attribution,
An amber transparency estimation step of estimating the transparency of the amber fragment determined to be yellow amber in the amber type determination step using fuzzy inference, and a yellow amber transparency estimation step.
A method for estimating the type and transparency of amber fragments, including. The upper mask image creation process
A grayscale step of grayscale the upper color image to obtain a grayscale image, and
A binarization step of binarizing the grayscale image to obtain a binary image, and
Complementing / noise removing steps for complementing and removing noise from the binary image,
Inversion processing step of performing inversion processing on the binary image obtained in the complementation / noise removal step, and
The method for estimating the type and transparency of the amber fragment according to claim 1, which comprises. The amber type determination process
With the R component approximation degree calculation step of calculating the approximation degree of the R component using the predetermined average value of the R component of yellow amber, brown amber and black amber and the average value of the R component obtained in the upper mask image analysis step. ,
H component approximation degree calculation to calculate the approximation degree of H component using the mode value of H component of yellow amber, brown amber and black amber determined in advance and the mode value of H component obtained in the upper mask image analysis step. Process and
A degree of attribution calculation step of calculating the degree of attribution of the pixel of the amber fragment region to yellow amber, brown amber and black amber based on the degree of approximation of the R component and the degree of approximation of the H component.
A discrimination processing step for discriminating the amber fragments according to the type of amber having the highest degree of attribution, and
The method for estimating the type and transparency of amber shards according to claim 1 or 2, which comprises. The yellow amber transparency estimation step
A lower color image acquisition step of irradiating the amber fragment determined to be yellow amber in the amber type determination step with light from below to acquire a lower color image, and a lower color image acquisition step.
A lower mask image creation step of applying mask processing to the lower color image to create a lower mask image,
A lower mask image analysis step of analyzing the lower mask image and acquiring information on the RGB component of the pixel of the amber fragment region in the lower mask image.
Based on the average value of the G component and the average value of the B component obtained in the lower mask image analysis step, a plurality of predetermined rules are set based on the average value of the G component and the average value of the B component of yellow amber. A goodness-of-fit calculation process for calculating the goodness of fit of an inference rule consisting of the above rules using the membership function of the preceding part, and
From the fuzzy logical product of the goodness of fit to each rule and the consequent membership function, the drawing process to create the figure that is the inference result, and
A transparency calculation step of performing non-fuzzying using Mamdani's inference method to obtain the center of gravity of the figure, and calculating the transparency of the amber fragment as the final inference result.
The method for estimating the type and transparency of amber fragments according to any one of claims 1 to 3. The antecedent part of the inference rule consists of the following A1 to A4.
A1: The average value of the G component is low and the average value of the B component is low. A2: The average value of the G component is low and the average value of the B component is high. The average value of the B component is low A4: The average value of the G component is high and the average value of the B component is high.
B1: Low transparency B2: High transparency or low transparency B3: High transparency The inference rule consists of the following rules δ _{1 to δ 4} .
δ ₁ : If the antecedent part is A1, the consequent part is B1 δ ₂ : If the antecedent part is A2, the consequent part is B2 δ ₃ : If the antecedent part is A3, The consequent part is B2 δ ₄ : If the antecedent part is A4, the consequent part is B3. The method for estimating the type and transparency of amber shards according to claim 4. The method for estimating the type and transparency of amber fragments according to any one of claims 1 to 5, wherein the illuminance of the light emitted in the upper color image acquisition step on the amber surface is 500 xl or more and 2500 lp or less. The method for estimating the type and transparency of amber fragments according to any one of claims 4 to 6, wherein the illuminance of the light emitted in the lower color image acquisition step on the amber surface is 500 xl or more and 1000 lp or less. It has a horizontal plane on which the amber shards can be placed, a light source capable of irradiating the amber shards on the horizontal plane with light from above, and an imaging means, and can acquire an upper color image of the amber shards. Upper color image acquisition section and
An upper mask image creation unit that creates an upper mask image by masking the upper color image,
An upper mask image analysis unit that analyzes the upper mask image and acquires information on RGB components and HSV components of pixels in the amber fragment region in the upper mask image.
Distinguishing between exodermis and amber from the ratio of pixels in which the average value of the R component and the average value of the V component are equal to each other obtained by the upper mask image analysis unit and the average value of the S component. Department and
Each amber type consisting of yellow amber, brown amber, and black amber is targeted at the pixels of the amber fragment region of the upper mask image in which the amber fragment determined to be amber by the outer skin / amber discrimination unit is captured. An amber type discriminating unit that calculates the degree of attribution and discriminates the amber fragment into the amber type having the highest degree of attribution,
An amber transparency estimation unit that estimates the transparency of the amber fragment determined to be yellow amber by the amber type determination unit using fuzzy inference, and a yellow amber transparency estimation unit.
A device for estimating the type and transparency of amber fragments. The process of irradiating amber fragments placed on a horizontal surface with light from above and acquiring an upper color image, and
The process of masking the upper color image to create the upper mask image, and
A step of analyzing the upper mask image and acquiring information on the RGB component and the HSV component of the pixel of the amber fragment region in the upper mask image.
A step of discriminating between the exodermis and amber from the ratio of pixels in which the average value of the R component and the average value of the V component obtained in the upper mask image analysis step are equal to each other and the average value of the S component.
Each amber type consisting of yellow amber, brown amber, and black amber is targeted at the pixels of the amber fragment region of the upper mask image in which the amber fragment identified as amber in the outer skin / amber discrimination step is captured. A step of calculating the degree of attribution and determining the amber fragment according to the amber type having the highest degree of attribution, and
A step of estimating the transparency of the amber fragment determined to be yellow amber in the amber type determination step by using fuzzy inference, and a step of estimating.
Amber fragment type and transparency estimation program to run the computer.

Images