JP5712392B2 - Device for determining foreign matter contamination in an object - Google Patents

Description

  The present invention relates to a foreign matter contamination determination device for determining whether or not a foreign matter containing a specific component other than a component constituting an object is mixed in an object such as food, and more particularly, in the near infrared region. The present invention relates to a foreign matter contamination determination device capable of inspecting an object nondestructively using the light of.

2. Description of the Related Art Conventionally, as a foreign matter contamination determination device in this type of object, for example, one described in Patent Document 1 (Japanese Patent No. 4220285) is known.
As shown in FIG. 40, the food as the object Ma to be examined is irradiated with white light in the near-infrared region from the light source unit 100, and the reflected light and the absorbance spectrum obtained thereby are measured. A second-order differential process is performed on the spectrum to select a wavelength band that shows a different second-order differential spectrum between the object and the foreign object, and a spectral image obtained by the light of the selected wavelength band is selected, for example, by a CCD camera or the like. The foreign substance Fa mixed in the object Ma is detected by performing a second derivative process on the absorbance spectrum of the selected wavelength band acquired by the imaging unit 101 and creating a second derivative spectral image.

  Further, as in this prior art, the object Ma to be examined is irradiated with near-infrared light, and an image obtained by imaging the object Ma is processed using the reflected light and absorbance spectrum obtained thereby. For example, Patent Document 2 (Japanese Patent Application Laid-Open No. 2001-99783), Patent Document 3 (Japanese Patent Application Laid-Open No. 2006-177890), and Patent Document 4 (Japanese Patent Application Laid-Open No. 2006-177890) are disclosed. 2009-168747), Patent Document 5 (Japanese Patent Laid-Open No. 2002-131239), and the like.

Japanese Patent No. 4220285 JP 2001-99783 A JP 2006-177890 A JP 2009-168747 A JP 2002-131239 A

  However, in this conventional foreign matter contamination determination device Sa, an image obtained by capturing an image of the object Ma is subjected to image processing, and it is determined whether or not the foreign material Fa is mixed in the object Ma based on this image. The foreign object Fa adhering to the surface of the object can be detected, but when the foreign object Fa is mixed in the object Ma or when the components constituting the object Ma have a form similar to the foreign object Fa If the size of the foreign object Fa is small, the foreign object Fa may not be specified from the image, and there is a problem that the accuracy of determining the contamination of the object Ma is poor.

  The present invention has been made in view of such problems, and it is possible to reliably select a specific component other than the components constituting the object to be examined so that a foreign object can be reliably identified. It is an object of the present invention to provide a foreign matter contamination determination method and a foreign matter contamination determination device in an object that can reliably determine whether foreign matter is mixed or not and improve the determination accuracy.

  In order to achieve such an object, the foreign matter contamination determination apparatus according to the present invention determines whether a foreign matter containing a specific component other than the component constituting the object is mixed in the object. In the foreign matter contamination determination apparatus, a light source unit that irradiates the object to be examined with light in the near infrared region, a light receiving unit that receives reflected or transmitted light from the object, and a light receiving unit that receives the light received by the light receiving unit A control unit for determining the presence or absence of a foreign substance in the object to be examined based on absorbance, and the control unit is previously irradiated with a sample object with a foreign substance and a sample object without a foreign substance from the sample object. A discriminant storage function for storing a discriminant relating to the attributed wavelength caused by the specific component calculated by statistical analysis of the second derivative spectrum in the absorbance with respect to the wavelength of the reflected or transmitted near infrared region; And a foreign matter discriminating function for discriminating the presence or absence of foreign matter in the object from the calculation result calculated from the absorbance of the light received by the light receiving unit and the discriminant, and the light source unit is arranged in the vicinity of a predetermined wavelength range. An infrared region of light is configured to irradiate, and a light splitting unit that splits the light received by the light receiving unit is provided. The foreign matter discrimination function of the control unit is based on the absorbance of the light split by the light splitting unit. It is configured to determine the presence or absence of foreign matter inside.

  Thus, when determining whether or not a foreign object containing a specific component other than the component constituting the object is mixed in the object to be examined using the apparatus of the present invention, the sample object with the foreign object is Near-infrared rays are irradiated from the light source part to the sample object without foreign matter, the reflected light or transmitted light from the sample object is received by the light receiving part, and the light received by the light receiving part is spectrally separated by the spectroscopic part. A discriminant related to the assigned wavelength due to the specific component is calculated by statistical analysis of the second derivative spectrum, and the calculated discriminant is stored in the discriminant storage function of the control unit, and the object to be examined The near-infrared ray is irradiated from the light source unit, the reflected light or transmitted light from the object to be examined is received by the light receiving unit, the light received by the light receiving unit is dispersed by the spectroscopic unit, and the absorbance of this light is measured, These absorbances and discriminants To determine the presence or absence of foreign matter in an object by calculating the calculated result by the foreign matter discrimination function.

  In this case, since the discriminant relating to the belonging wavelength due to the specific component other than the component constituting the object to be examined is calculated by the second derivative spectrum in the absorbance of the reflected light or transmitted light received from the sample object, Components that make up the object to be examined even if they are mixed inside the object, or if the components that make up the object have a form similar to that of a foreign object, or if the size of the foreign object is small Certain components other than can be selected with certainty. And since the calculation result is calculated from the discriminant relating to the attribute wavelength attributed to the specific component and the absorbance of the light received from the object to be examined, it becomes possible to reliably identify the foreign matter containing the specific component, Therefore, it is possible to reliably determine whether or not foreign matter is mixed in the object, so that the determination accuracy can be improved.

  In addition, the light is divided in each wavelength region by the spectroscopic unit. In the present invention, reflected light or transmitted light from the object to be inspected is received by the light receiving unit, and the received light is split. Since it is a spectrum, the processing speed in the control unit is faster than in the so-called pre-spectrometry, in which the light from the light source unit is split before irradiating the object to be inspected. Can do.

In the determination by the control unit, the object is food and the foreign material is animal hair as necessary. Animal hair as used herein is a concept including human hair such as human hair and eyebrows.
Thereby, since food can be inspected nondestructively and it can be discriminate | determined whether the animal hair is mixed in food, a sanitary food can be produced.

If necessary, the specific component is cystine.
The main component of human body hair is keratin protein, which is composed of 80% to 90%. The amino acid composition of this keratin protein contains 14% to 18% of cystine in human body hair, and this cystine may or may not be present in general food materials and proteins. It is about a trace. Therefore, by detecting this cystine, it is possible to determine whether or not human body hair is mixed in food.

In constructing the discriminant, the discriminant stored by the control unit is selected from a plurality of different discriminating methods as necessary.
Specific examples include discriminant analysis method, multiple regression analysis method, PLS analysis method, neural network backpropagation method, SVM method and the like. Since an optimum discrimination method corresponding to the object or foreign object to be examined can be selected from a plurality of discrimination methods, the discrimination accuracy can be improved.

Further, if necessary, the control unit uses two or more discriminants from the plurality of discriminants and performs a calculation based on the corresponding discriminants, and at least one of the discrimination results determines whether a foreign object is present. In this case, it is determined that foreign matter is mixed.
As a result, even if foreign matter is not mixed in the object based on the result of discrimination by one discriminant, if foreign matter is found mixed in the object by the result of discrimination by another discriminant, Therefore, it is possible to provide an object free from foreign matter more reliably.

  Specifically, as one of the discriminating methods in constructing the discriminant, the discriminant is configured according to the discriminant analysis method by secondarily differentiating the absorbance spectrum of the light received from the sample object as necessary.

  In this discriminant analysis method, the discriminant is calculated according to the formula satisfying the relationship of the following general formula (A-1) using the absorbances of the first to n wavelengths with which the discriminating results are improved as necessary. It is said.

  In general formula (A-1), zi is the discrimination score of each group, λ is the wavelength, A is the absorbance, A1 (λ1) is the absorbance at the first wavelength (λ1), and A2 (λ2) is the second wavelength (λ2). An (λn) is the absorbance at the nth wavelength (λn), a0, a1, a2,... Are the coefficients that determine the discriminant score zi, measured in a sufficiently large population It is determined by solving equation (1-3) so that the following equation (1-1) takes the maximum value from the absorbance and the discrimination score.

In Formula (1-1), F (a1, a2) represents that zi is a function determined by coefficients a1 and a2, Sb represents a variation between groups, St represents a discrimination score for the entire group and its average Is the sum of squares of the difference, that is, the variation of the entire group.
In Equation (1-2), St is the total sum of squares of the difference between the discrimination score of the entire group and its average, and Sw is the sum of the sum of squares of the discrimination scores within the group and the average difference thereof. , Sb is the variation between groups as the sum of the sum of squares of the average difference between the discrimination score in each group and the discrimination score of the entire group.

  In addition, this discriminant analysis method divides the above discriminant into several groups based on the discriminant score for sample objects with known foreign matter in advance, if necessary, and averages each group or between each group. Calculate the Mahalanobis's general distance from the value and the center of gravity position, measure the absorbance of the subject, and determine which group is closest to the known group and the subject's Mahalanobis's general distance from the absorbance. May be used as a variable, and may be calculated using an equation that satisfies the relationship of the following general formula (A-2).

  In general formula (A-2), D is the Mahalanobis general distance in group i, λ is the wavelength, A is the absorbance, A1 (λ1) is the absorbance at the first wavelength (λ1), and A2 (λ2) is the second wavelength (λ2). ),..., An (λn) is the absorbance at the nth wavelength (λn), s11... Snn is the dispersion covariance matrix between each group, and these D1 and D2 are calculated. Which group belongs to the D1 or D2 of the known group is determined from the absorbance measured for the subject.

Further, as one of the discriminating methods in the construction of discriminant, if necessary, the discriminant is calculated by multiple regression analysis by secondarily differentiating the absorbance spectrum of light received from the sample object after MSC processing It is said.
The feature of the foreign matter contamination determination method according to the present invention is to find the wavelength range of the belonging wavelength caused by a specific component of the foreign matter mixed in the object such as food, and use the wavelength range to detect the foreign matter mixed in the object. It is a point to determine whether or not. That is, by a multiple regression analysis with the absorbance spectrum of light received from a sample object with foreign matter, first, a first wavelength having a high correlation coefficient is obtained, and then second to n wavelengths having a high correlation coefficient are obtained.

  Specifically, the above discriminant is configured by a formula satisfying the relationship of the following general formula (B) with the absorbances of the first to n wavelengths having a high correlation coefficient as variables.

  In selecting the first wavelength (λ1) to the nth wavelength (λn) in the general formula (B), first, the absorbance of the sample object artificially mixed with the foreign substance and the sample object not mixed with the foreign substance The near-infrared wavelength region of the first wavelength (λ1) in which the correlation coefficient attributed to the foreign matter obtained by discriminant analysis with respect to the absorbance is 0.800 or more is selected, and then the first wavelength (λ1) By the discriminant analysis between the near-infrared wavelength region and the wavelength range of 1100 nm to 2200 nm, the correlation coefficient is higher than the correlation coefficient of the near-infrared wavelength region of the first wavelength (λ1). By selecting the wavelength range of the second wavelength (λ2), and then by the multiple regression analysis of the near-infrared wavelength range of the first wavelength (λ1) and the second wavelength (λ2) and the wavelength range of 1100 nm to 2200 nm Belonging to the above foreign matter A wavelength range of the third wavelength (λ3) that is a correlation coefficient greater than or equal to the correlation coefficient of the near-infrared wavelength range of the first wavelength (λ1) and the second wavelength (λ2) is selected, and thus the first wavelength (Λ1) to the (n-1) wavelength (λn-1) near infrared wavelength range and 1100 nm to 2200 nm wavelength range is assigned to the foreign matter by multiple regression analysis of the first wavelength (λ1) The wavelength range of the nth wavelength (λn) that is equal to or greater than the correlation coefficient of the near infrared wavelength range is selected.

  In addition, as one of the discriminating methods in the construction of discriminant, if necessary, the discriminant is subjected to second-order differentiation after the MSC process on the absorbance spectrum of light received from the sample object, and PLS (Partial least square projection to Lent. The structure is calculated by the (Structure) analysis method.

  Further, as one of the discriminating methods in the construction of discriminant, if necessary, the discriminant is subjected to second-order differentiation after the MSC process for the absorbance spectrum of light received from the sample object (back propagation of the neural network method). The configuration is calculated by the back propagation method.

  Furthermore, as one of the discriminating methods in constructing the discriminant, if necessary, the discriminant is subjected to second derivative after the MSC process for the absorbance spectrum of the light received from the sample object, and the SVM (Support Vector Machine) method. The configuration is calculated as follows.

  In each of these discriminants, a statistically dominant wavelength is selected from wavelengths in the near infrared region in the range of 1100 nm to 2500 nm as necessary.

  In particular, the wavelength in each discriminant described in the discriminant analysis method, the multiple regression analysis method, the back propagation method of the neural network method, and the SVM method is the first wavelength of 1320 to 1460 nm, 1530 to 1542 nm, and 1770 to 1776 nm. The second wavelength is selected from the range of 1144 to 1158 nm, 1252 to 1276 nm, 1320 to 1338 nm, 1346 to 1384 nm, 1406 to 1434 nm, 1470 to 1510 nm, 1598 to 1650 nm, and the third wavelength. 1100 to 1138 nm, 1230 to 1324 nm, 1352 to 1378 nm, 1394 to 1428 nm, 1480 to 1502 nm, 1540 to 1580 nm, 1610 to 1656 nm, 1664 to 1696 nm, 1742 to 17 0 nm, 1760-1776 nm, 1816-1850 nm, 1880-1890 nm, 2112-2120 nm, 2322-2340 nm, and the fourth wavelength is 1100-1122 nm, 1146-11168 nm, 1250-1282 nm, 1350-1382 nm, 1558-1576 nm , 1676 to 1696 nm, 1740 to 1752 nm, 1808 to 1866 nm, 2124 to 2140 nm, 2258 to 2282 nm, 2322 to 2340 nm, and the fifth wavelength is 1100 to 1142 nm, 1152 to 1210 nm, 1248 to 1326 nm, 1394 to 1420 nm, Select from the range of 1590 to 1640 nm, 1880 to 1884 nm, 2400 to 2420 nm, and select one from the wavelength range of each of these ranges That is a combination that does not exhibit multicollinearity.

Preferably, the first wavelength is selected from the range of 2158 to 2186 nm, the second wavelength is selected from the range of 1606 to 1640 nm, the third wavelength is selected from the range of 1250 to 1298 nm, and the fourth wavelength is selected from 1682 to 1694 nm. The fifth wavelength is selected from a range of 1278 to 1322 nm, and a combination of wavelengths in each of these ranges.
More preferably, the first wavelength is 2172 nm, the second wavelength is 1618 nm, the third wavelength is 1260 nm, the fourth wavelength is 1692 nm, and the fifth wavelength is 1134 nm.

  In addition, the first wavelength is selected from a range of 1354 to 1386 nm, 1518 to 1540 nm, 1578 to 1612 nm, 1692 to 1694 nm, 1734 to 1744 nm, 1810 to 1870 nm, 2048 to 2070 nm, and 2150 to 2202 nm as another combination as necessary. The second wavelength is selected from the range of 1352-1386 nm, 1566-1618 nm, 1682-1694 nm, 1732-1748 nm, 1810-1872 nm, 2048-2086 nm, 2152-2166 nm, 2190-2198 nm, 2252-2266 nm, 2322-2334 nm. The third wavelength is selected from the range of 1782 to 1808 nm and 2448 to 2454 nm, and the fourth wavelength is 1222 to 1228 nm, 1276 to 1286 nm, 1362 380 nm, 1572 to 1606 nm, 1726 to 1740 nm, 1810 to 1844 nm, 2042 to 2088 nm, 2150 to 2166 nm, 2258 to 2272 nm, and the fifth wavelength is 1372 to 1392 nm, 1524 to 1548 nm, 1834 to 1878 nm, 2190 to 2214 nm These are selected from the ranges of the above, and one of the wavelength ranges in each of these ranges is selected as a combination that does not exhibit multicollinearity.

Preferably, the first wavelength is selected from the range of 2154 to 2196 nm, the second wavelength is selected from the range of 1736 to 1744 nm, the third wavelength is selected from the range of 1798 to 1806 nm, and the fourth wavelength is from 2046 to 2074 nm. Select and select the fifth wavelength from the range of 1382 to 1392 nm, which is a combination of wavelengths in each of these ranges.
More preferably, 2162 nm as the first wavelength, 1740 nm as the second wavelength, 1804 nm as the third wavelength, 2054 nm as the fourth wavelength, and 1388 nm as the fifth wavelength are combinations of these wavelengths.

  In each of the above discriminants, after selection of the first wavelength, the wavelength range that does not exhibit multicollinearity is derived up to the 16th wavelength.

Further, in the foreign matter contamination determination device of the present invention, the object to be examined is placed and driven as necessary, and the placed object is irradiated with light from the light source unit. In addition, the apparatus includes a transport unit that sequentially transports reflected light or transmitted light from the object to an inspection position where the light receiving unit can receive the light.
As a result, a large number of objects to be examined can be easily and sequentially inspected, so that workability can be improved.

Furthermore, if necessary, the light source unit is configured to output light through an optical fiber and include a waveguide of the optical fiber.
Thereby, it is possible to reliably irradiate the object to be examined simultaneously with a large number of wavelength regions included in the light in the near infrared region irradiated from the light source unit.

Furthermore, if necessary, the light receiving unit includes a plurality of light receiving elements that receive reflected light or transmitted light from the object to be examined. In this case, it is effective to configure the light receiving element with an optical fiber.
Thereby, it is possible to reliably receive a large number of wavelength regions included in the reflected light or transmitted light from the object to be examined.

Moreover, you may comprise the said control part provided with the remote operation function as needed.
Thereby, since the inspection of the object to be examined can be performed regardless of the place, the versatility of the apparatus can be improved.

According to the foreign matter contamination determination device of the present invention, the attribute wavelength caused by a specific component other than the component constituting the object to be examined by the second derivative spectrum in the absorbance of the reflected light or transmitted light received from the sample object When the foreign object is mixed inside the object, the component constituting the object has a form similar to the foreign object, or the size of the foreign object is small. In addition, a specific component other than the components constituting the object to be examined can be reliably selected. And since the calculation result is calculated from the discriminant relating to the attribute wavelength attributed to the specific component and the absorbance of the light received from the object to be examined, it becomes possible to reliably identify the foreign matter containing the specific component, Therefore, it is possible to reliably determine whether or not foreign matter is mixed in the object, so that the determination accuracy can be improved.
Also, since it is so-called post-spectrometry, which splits reflected light or transmitted light from the object to be examined, compared to so-called pre-spectroscopy, which divides the light from the light source unit and then irradiates the object to be examined. The processing speed in the control unit is increased, so that the determination can be performed quickly and accurately.

It is a perspective view which shows the foreign material contamination determination apparatus in the object which concerns on embodiment of this invention. It is the schematic which shows the foreign material contamination determination apparatus in the object which concerns on embodiment of this invention. It is a block diagram which shows the structure of the control part in the foreign material mixing determination apparatus in the object which concerns on embodiment of this invention. It is a flowchart which shows the control flow in the foreign material contamination determination apparatus in the object which concerns on embodiment of this invention. It is a control flow which concerns on embodiment of this invention, Comprising: It is a flowchart which shows the control flow of a sample measurement and a discrimination routine. It is a control flow which concerns on embodiment of this invention, Comprising: It is a flowchart which shows the control flow of a foreign material discrimination | determination routine. It is a figure which shows the precision of the discriminant of discriminant analysis concerning the experiment example (1-1) of this invention. It is a figure which shows the discrimination | determination result in a discriminant analysis in connection with the experiment example (1-1) of this invention. It is a figure which shows the discrimination rate in a discriminant analysis in connection with the experiment example (1-1) of the present invention. It is a figure which shows the precision of the discriminant of a discriminant analysis in connection with the experiment example (1-2) of this invention. It is a figure which shows the discrimination | determination result in discriminant analysis in connection with the experiment example (1-2) of this invention. It is a figure which shows the discrimination rate in a discriminant analysis in connection with the experiment example (1-2) of this invention. It is a figure which concerns on the experiment example (2-1) of this invention, and shows the absorbance second derivative spectrum of cystine and hair. It is a figure which concerns on the experiment example (2-1) of this invention, and shows the precision of the discriminant of a multiple regression analysis. It is a figure which shows the discrimination | determination result in a multiple regression analysis in connection with the experiment example (2-1) of this invention. It is a figure which shows the discrimination rate of a multiple regression analysis in connection with the experiment example (2-1) of this invention. It is a figure which concerns on the experiment example (2-2) of this invention, and shows the precision of the discriminant of multiple regression analysis. It is a figure which shows the discrimination result in a multiple regression analysis in connection with the experiment example (2-2) of this invention. It is a figure which shows the discrimination rate of a multiple regression analysis in connection with the experiment example (2-2) of this invention. It is a figure which shows the precision of the discriminant of a PLS method in connection with Experimental example (3) of this invention. It is a figure which shows the discrimination result by PLS method in connection with the experiment example (3) of this invention. It is a figure which shows the discrimination rate of PLS method in connection with the experiment example (3) of the present invention. It is a figure which shows the precision of the discriminant of the back propagation method of a neural network method in connection with the experiment example (4-1) of the present invention. It is a figure which shows the discrimination | determination result in the back propagation method of a neural network method in connection with the experiment example (4-1) of this invention. It is a figure which shows the discrimination rate of the back propagation method of a neural network method in connection with the experiment example (4-1) of the present invention. It is a figure which shows the precision of the discriminant of the back propagation method of a neural network method in connection with the experiment example (4-2) of this invention. It is a figure which shows the discrimination | determination result in the back propagation method of a neural network method in connection with the experiment example (4-2) of this invention. It is a figure which shows the discrimination rate of the back propagation method of a neural network method in connection with the experiment example (4-2) of the present invention. It is a figure which shows the learning result using the SVM method in connection with the experiment example (5-1) of the present invention. It is a figure which shows the discrimination | determination learning result by the selection wavelength in a SVM method in connection with the experiment example (5-1) of this invention. It is a figure which shows the discrimination | determination result using the discriminant by SVM method in connection with the experiment example (5-2) of this invention. It is a figure which shows the discrimination | determination result using the discriminant by SVM method in connection with the experiment example (5-3) of this invention. It is a figure which shows the discrimination | determination result at the time of combining the discriminant of the back propagation method of a discriminant analysis method, a multiple regression analysis method, and a neural network method according to the experiment example (6-1) of this invention. It is a figure which shows the discrimination rate at the time of combining the discriminant of the back propagation method of the discriminant analysis method, the multiple regression analysis method, and the neural network method according to the experiment example (6-1) of the present invention. It is a figure which shows the discrimination | determination result at the time of combining the discriminant of the back propagation method of a discriminant analysis method, a multiple regression analysis method, a PLS method, and a neural network method in connection with the experiment example (6-2) of the present invention. It is a figure which shows the discrimination rate at the time of combining the discriminant of the back-propagation method of a discriminant analysis method, a multiple regression analysis method, a PLS method, and a neural network method in connection with the experiment example (6-2) of the present invention. The figure which shows the discrimination result at the time of combining the discriminant of the discriminant analysis method, the multiple regression analysis method, the PLS method, the back-propagation method of the neural network method, and the SVM method according to the experiment example (6-3) of the present invention. is there. The figure which shows the discrimination rate at the time of combining the discriminant of the discriminant analysis method, the multiple regression analysis method, the PLS method, the backpropagation method of the neural network method, and the SVM method according to the experiment example (6-3) of the present invention. is there. It is a figure which shows the foreign material contamination determination apparatus in the object which concerns on another embodiment of this invention. It is a figure which shows an example of the conventional foreign material mixing determination apparatus in an object.

  Hereinafter, a foreign matter contamination determination apparatus according to an embodiment of the present invention will be described with reference to the accompanying drawings.

  The foreign matter contamination determination apparatus according to the embodiment of the present invention determines whether or not human body hair as a foreign matter containing a specific component not present in food is mixed in the food as the object to be examined. It is to be determined. Here, the food includes any food material, processed food, cooked food, etc. provided for food. Here, the human body hair includes hair, eyebrows, eyelashes and the like.

  Here, the specific component is cystine. The main component of human body hair is keratin protein, which is composed of 80% to 90%. The amino acid composition of this keratin protein contains 14% to 18% of cystine in human body hair, and this cystine may or may not be present in general food materials and proteins. It is about a trace. Therefore, by detecting this cystine, it is possible to determine whether or not human hair is mixed in the food.

  As shown in FIGS. 1 to 3, the foreign substance contamination determination device S according to the embodiment of the present invention includes a machine base 1 and an object M to be examined provided on the machine base 1. The transport unit 10 that transports the driven and mounted object M, the light source unit 20 that irradiates the object M to be examined with light in the near infrared region, and the reflected or transmitted light from the object M is received. The light receiving unit 30 includes a control unit 40 that determines the presence or absence of a foreign substance F in the object M to be examined based on the absorbance of light received by the light receiving unit 30.

  The transport unit 10 is driven by mounting the object M to be examined, and the mounted object M is irradiated with light from the light source unit 20 and receives reflected or transmitted light from the object M as a light receiving unit. 30 is sequentially conveyed to an inspection position where light can be received, and is constituted by a belt conveyor 10a.

The light source unit 20 is placed on the partition plate 2 provided on the machine base 1 above the transport unit 10 and is configured to output white light by an optical fiber and include an optical fiber waveguide. A hole (not shown) is formed in the partition plate 2, and light from the light source unit 20 is irradiated to the object M to be examined through this hole. In addition, although not shown, the partition plate 2 is formed with a plurality of punch holes for releasing heat generated by the operation of the light source unit 20 to the outside.
The light source unit 20 incorporates a shutter (not shown), and a reference measurement reference white plate is installed on the shutter.

The light receiving unit 30 is placed on the partition plate 2 provided on the machine base 1 above the transport unit 10 and is provided in an equiangular relationship to receive reflected light or transmitted light from the object M to be examined. A plurality of light receiving elements 31 are provided. A hole (not shown) is formed in the partition plate 2, and reflected light or transmitted light from the object M to be examined is received through the hole. In addition, although not shown, the partition plate 2 is formed with a plurality of punch holes for releasing the heat generated by the operation of the light receiving unit 30 to the outside. The light receiving unit 30 has a built-in shutter (not shown), and a reference measurement reference white plate is installed on the shutter.
A spectroscopic unit 50 that splits and extracts the light received by the light receiving unit 30 for each wavelength region is provided below the transport unit 10 and on the mounting plate 3 provided on the machine base 1. The spectroscopic unit 50 includes an acousto-optic element (not shown) that splits the light transmitted from the light receiving unit 30 into single wavelength light.
The light receiving element 31 is composed of an optical fiber, connected to the spectroscopic unit 50 in series or in parallel, and connected to an electric circuit in the control unit 40 via the control cable 51 from the spectroscopic unit 50 to perform signal processing.

  The entire signal processing is performed as follows. Diffuse reflected light is detected by each light receiving element 31 and sent to the spectroscopic unit 50 by the optical fiber of the light receiving unit 30. The spectroscopic unit 50 divides the light for each wavelength range and converts the light into an electric signal based on the intensity of the light and extracts it. Then, it connects with the electric circuit in the control part 40 via the control cable 51. FIG. As shown in FIG. 3, the electrical signal converted from the light receiving element 31 through the spectroscopic unit 50 is transmitted to the control unit 40, and integrated control with a discriminant storage function or the like through the communication unit 41 provided in the control unit 40. It is sent to the arithmetic processing unit 42 to determine the presence or absence of foreign matter F.

  The control unit 40 statistically analyzes the second derivative spectrum in the absorbance with respect to the wavelength in the near-infrared region irradiated and reflected or transmitted from the sample object M with foreign matter and the sample object M without foreign matter in advance. A discriminant storage function for storing a discriminant relating to the attributed wavelength caused by the specific component calculated by the analysis, and a calculation result calculated from the absorbance and discriminant of the light received by the light receiving unit 30, in the object M A foreign matter discrimination function for discriminating the presence or absence of the foreign matter F is provided.

  As shown in FIG. 3, the control unit 40 includes a communication unit 41, a discriminant storage function, a foreign matter determination function, a processing history storage function, an Internet communication function, a light quantity control function, and a result display function. And a motor control circuit 43 that controls the drive motor of the light source unit 20, a spectral control circuit 44 that controls the acoustooptic device, and an operation command unit 45.

The discriminant storage function in the total control arithmetic processing unit 42 of the control unit 40 includes a discriminant analysis storage function, a regression expression storage function, and a neuro discrimination function.
The discriminant analysis storage function is a function for storing discriminants calculated by the discriminant analysis method, the regression expression storage function is a function for storing discriminants calculated by the multiple regression analysis method, and the neuro discriminant function is a neural network. This is a function for storing discriminants calculated by a method (two types of back propagation (BP) and support vector machine (SVM) in this embodiment).

  In addition, the foreign substance discrimination function in the overall control calculation processing unit 42 of the control unit 40 is equipped with several types of discriminant analysis methods, various results and parameters are calculated, and a calculation for discriminating an optimum result from these numerical values ( Calculation) function.

  Furthermore, the processing history storage function in the total control arithmetic processing unit 42 of the control unit 40 is a function for accumulating the measured number of samples, accumulating the determination results OK and NG, and accumulating and storing them in time series. is there.

  Furthermore, the Internet communication function in the general control arithmetic processing unit 42 of the control unit 40 is used for grasping the discrimination status and system status when used in a remote place, and the version of the application (discrimination method). It is a function for performing maintenance such as up by remote control. Although this function is not used in the present embodiment, by using this Internet communication function, the inspection of the object M to be inspected can be performed regardless of the location, thereby improving the versatility of the apparatus S. be able to.

  Further, the light amount control function in the total control calculation processing unit 42 of the control unit 40 is able to measure at an appropriate level of the light detection sensor in the spectroscopic unit 50 when performing lamp replacement or measuring an object M having a large difference in gloss. This is a function that adjusts the light amount of the light source and the sensitivity of the sensor in the spectroscopic unit 50 so that appropriate measurement can be performed at all times.

  Furthermore, the result display function in the overall control calculation processing unit 42 of the control unit 40 is a function for displaying the determination result on the display unit 46 described later.

In FIG. 1, reference numeral 46 denotes a display unit made up of a CRT or the like provided in the control unit 40. The data is displayed on the display unit 46. The display of the display unit 46 is operated by a screen operation unit (not shown), and can be switched and displayed as appropriate, such as an input setting screen, a discrimination result display screen, and an error display screen. An animation or the like may be displayed during measurement. The discrimination result may be displayed on the LCD panel. Further, the determination result may be output as a sound. Further, an external data output interface may be provided.
In FIG. 1, reference numeral 47 denotes a setting panel provided in the control unit 40. The setting panel 47 is used for checking the lamp lighting cumulative time, canceling the error, checking the date / time setting, checking the operation of the spectroscopic unit, shifting to and releasing the manual operation of each mechanism, and the like.

Further, in FIG. 1, reference numeral 48 denotes a control button, which includes a start button and a stop button. When the start button is pressed, the discrimination program is executed, and when the stop button is pressed, the discrimination program is stopped. If the start button is pressed again while the discrimination program is stopped, the program is resumed.
Further, in FIG. 1, reference numeral 4 is a main power supply switch for supplying power to the apparatus S, reference numeral 49 is a control unit power supply for supplying power to the control unit 40 and the light source unit 20, and reference numeral 52. Is a power supply for a spectroscopic unit that supplies power to the spectroscopic unit 50. The spectroscopic power supply 52 uses a linear power supply with less ripple noise in order to remove noise.

In FIG. 1, reference numeral 5 denotes an emergency stop button for stopping the program and operation of the device S in an emergency, and reference numeral 6 denotes a control unit 40, a light source unit 20, a light receiving unit 30, a control unit power supply 49, and the like. This is a heat exhaust fan for releasing heat generated by operation to the outside of the machine base 1. Although stray light and disturbance light may enter from the exhaust heat fan 6, these lights are shielded by a plurality of punch holes provided in the partition plate 2.
Further, in FIG. 1, reference numeral 7 denotes a control cable wiring for accommodating a control cable 51 for signal transmission to the light source unit 20, the light receiving unit 30, the spectroscopic unit 50, and the control unit 40 so that the control cable 51 can be taken in and out of the machine base 1. Reference numeral 8 denotes a sample detection sensor that detects the object M conveyed by the conveyance unit 10. Two sample detection sensors 8 are provided, whereby the conveyance speed can be checked based on the detection interval, and a shutter opening instruction or sample measurement can be performed at an appropriate timing. Further, even when the conveyance speed is changed due to an operator's mistake, the speed is automatically recognized, so that it is possible to cope with it.

  Then, the combination of the discriminant stored by the discriminant storage function in the overall control arithmetic processing unit 42 of the control unit 40 and the wavelength of the selected near infrared ray is determined as follows.

  Specifically, using the apparatus S, the sample object M with foreign matter and the sample object M without foreign matter are irradiated with near infrared rays in advance, and the reflected light or transmitted light from the sample object M is received and received. A discriminant related to the attribute wavelength attributed to a specific component is calculated by statistical analysis of the second derivative spectrum of light absorbance.

  The discriminant stored by the control unit 40 is selected from a plurality of different discriminating methods. In this embodiment, five discriminant methods are selected: discriminant analysis method, multiple regression analysis method, PLS analysis method, neural network method back-propagation method, and SVM method. Hereinafter, each analysis method will be described in detail.

<Discriminant analysis method (1)>
The discriminant calculated by discriminating and analyzing the absorbance spectrum of the light received from the sample object M by the second order differentiation method is the following general formula (A− 1).

  In general formula (A-1), zi is the discrimination score of each group, λ is the wavelength, A is the absorbance, A1 (λ1) is the absorbance at the first wavelength (λ1), and A2 (λ2) is the second wavelength (λ2). An (λn) is the absorbance at the nth wavelength (λn), a0, a1, a2,... Are the coefficients that determine the discriminant score zi, measured in a sufficiently large population It is determined by solving equation (1-3) so that the following equation (1-1) takes the maximum value from the absorbance and the discrimination score.

In Formula (1-1), F (a1, a2) represents that zi is a function determined by coefficients a1 and a2, Sb represents a variation between groups, St represents a discrimination score for the entire group and its average Is the sum of squares of the difference, that is, the variation of the entire group.
In Equation (1-2), St is the total sum of squares of the difference between the discrimination score of the entire group and its average, and Sw is the sum of the sum of squares of the discrimination scores within the group and the average difference thereof. , Sb is the variation between groups as the sum of the sum of squares of the average difference between the discrimination score in each group and the discrimination score of the entire group.

<Discriminant analysis method (2)>
The discriminant calculated by this discriminant analysis method can also be calculated by another discriminant method. Specifically, the discriminant is divided into several groups on the basis of the discrimination score for the sample object M in which the presence or absence of the foreign substance F is known in advance, and the Mahalanobis from the average value or the center of gravity position within each group or between each group is calculated. Calculate the generalized distance, measure the absorbance of the subject, and determine from the absorbance the known group and the generalized distance of the subject's Mahalanobis, using the absorbance of the 1st to nth wavelengths as a variable. It is calculated by an equation that satisfies the relationship (A-2).

  In general formula (A-2), D is the Mahalanobis general distance in group i, λ is the wavelength, A is the absorbance, A1 (λ1) is the absorbance at the first wavelength (λ1), and A2 (λ2) is the second wavelength (λ2). ),..., An (λn) is the absorbance at the nth wavelength (λn), s11... Snn is the dispersion covariance matrix between each group, and these D1 and D2 are calculated. Which group belongs to the D1 or D2 of the known group is determined from the absorbance measured for the subject.

<Multiple regression analysis method>
Also, the discriminant calculated by the multiple regression analysis after second-order differentiation of the absorbance spectrum of the light received from the sample object M by MSC processing uses the first to n-th wavelength absorbances having a high correlation coefficient as variables. General formula (B).

  In selecting the first wavelength (λ1) to the nth wavelength (λn) in the general formula (B), first, the absorbance of the sample object M in which the foreign substance F is artificially mixed and the sample object in which the foreign substance F is not mixed The near-infrared wavelength region of the first wavelength (λ1) in which the correlation coefficient attributed to the foreign substance F obtained by discriminant analysis with the absorbance of M is 0.800 or more is selected, and then the first wavelength (λ1 ) Of the near-infrared wavelength range and the wavelength range in the range of 1100 nm to 2200 nm are attributed to the foreign matter F and have a higher correlation coefficient than the correlation coefficient of the near-infrared wavelength range of the first wavelength (λ1). The wavelength range of the second wavelength (λ2) is selected, and then the foreign matter is obtained by multiple regression analysis of the near-infrared wavelength range of the first wavelength (λ1) and the second wavelength (λ2) and the wavelength range of 1100 nm to 2200 nm. The first wavelength (λ1 And a wavelength region of the third wavelength (λ3) that is equal to or greater than the correlation coefficient of the near-infrared wavelength region of the second wavelength (λ2), and thus the first wavelength (λ1) to (n -1) Phase relationship between the near-infrared wavelength range of the first wavelength (λ1) attributed to the foreign substance F by multiple regression analysis of the near-infrared wavelength range of the wavelength (λn-1) and the wavelength range of 1100 nm to 2200 nm A wavelength region of the nth wavelength (λn) that is greater than or equal to the number is selected.

<PLS method>
In addition, a discriminant calculated using a PLS (Partial Least Square Projection to Lent Structure) analysis method after second-order differentiation of the absorbance spectrum of light received from the sample object M after MSC processing is used.

<Back propagation method of neural network method>
Furthermore, the discriminant calculated by the back propagation method of the neural network method using the second derivative of the absorbance spectrum of the light received from the sample object M after MSC processing is used.

<SVM method>
Furthermore, the discriminant calculated using the SVM (Support Vector Machine) method by secondarily differentiating the absorbance spectrum of the light received from the sample object M after MSC processing is used.

  In each of these discriminants, a statistically dominant wavelength is selected from wavelengths in the near infrared region in the range of 1100 nm to 2500 nm.

  Specifically, the following wavelengths are used for the discriminant analysis method, the multiple regression analysis method, the back propagation method of the neural network method, and the SVM method. That is, the wavelength in each discriminant is selected from the range of the first wavelength from 1320 to 1460 nm, from 1530 to 1542 nm, from 1770 to 1776 nm, from 2154 to 2206 nm, and the second wavelength from 1144 to 1158 nm, from 1252 to 1276 nm, from 1320 to 1338 nm, 1346 to 1384 nm, 1406 to 1434 nm, 1470 to 1510 nm, 1598 to 1650 nm, and the third wavelength is 1100 to 1138 nm, 1230 to 1324 nm, 1352 to 1378 nm, 1394 to 1428 nm, 1480 to 1502 nm, 1540 to 1580 nm, 1610 -1656nm, 1664-1696nm, 1742-1750nm, 1760-1776nm, 1816-1850nm, 1880-1890nm, 2112-2120nm The second wavelength is selected from the range of 2322 to 2340 nm, and the fourth wavelength is 1100 to 1122 nm, 1146 to 1168 nm, 1250 to 1282 nm, 1350 to 1382 nm, 1558 to 1576 nm, 1676 to 1696 nm, 1740 to 1852 nm, 1808 to 1866 nm, 2124 to 2140 nm, 2258 Select from the range of ˜2282 nm, 2322 to 2340 nm, and select the fifth wavelength from the range of 1100 to 1142 nm, 1152 to 1210 nm, 1248 to 1326 nm, 1394 to 1420 nm, 1590 to 1640 nm, 1880 to 1884 nm, 2400 to 2420 nm, and these These are combinations that do not exhibit multicollinearity, in which one is selected from each wavelength range.

Preferably, the first wavelength is selected from the range of 2158 to 2186 nm, the second wavelength is selected from the range of 1606 to 1640 nm, the third wavelength is selected from the range of 1250 to 1298 nm, and the fourth wavelength is selected from 1682 to 1694 nm. The fifth wavelength is selected from a range of 1278 to 1322 nm, and a combination of wavelengths in each of these ranges.
More preferably, the first wavelength is 2172 nm, the second wavelength is 1618 nm, the third wavelength is 1260 nm, the fourth wavelength is 1692 nm, and the fifth wavelength is 1134 nm.

  As another combination, the first wavelength is selected from the range of 1354 to 1386 nm, 1518 to 1540 nm, 1578 to 1612 nm, 1692 to 1694 nm, 1734 to 1744 nm, 1810 to 1870 nm, 2048 to 2070 nm, 2150 to 2202 nm, and the second wavelength is selected 1352 to 1386 nm, 1566 to 1618 nm, 1682 to 1694 nm, 1732 to 1748 nm, 1810 to 1872 nm, 2048 to 2086 nm, 2152 to 2166 nm, 2190 to 2198 nm, 2252 to 2266 nm, 2322 to 2334 nm, and the third wavelength to 1782 -1808 nm, 2448-2454 nm, and the fourth wavelength is 1222-1228 nm, 1276-1286 nm, 1362-1380 nm, 15 2 to 1606 nm, 1726 to 1740 nm, 1810 to 1844 nm, 2042 to 2088 nm, 2150 to 2166 nm, 2258 to 2272 nm, and the fifth wavelength ranges from 1372 to 1392 nm, 1524 to 1548 nm, 1834 to 1878 nm, 2190 to 2214 nm And a combination that does not exhibit multicollinearity, in which one is selected from the wavelength ranges of these ranges.

Preferably, the first wavelength is selected from the range of 2154 to 2196 nm, the second wavelength is selected from the range of 1736 to 1744 nm, the third wavelength is selected from the range of 1798 to 1806 nm, and the fourth wavelength is from 2046 to 2074 nm. Select and select the fifth wavelength from the range of 1382 to 1392 nm, which is a combination of wavelengths in each of these ranges.
More preferably, 2162 nm as the first wavelength, 1740 nm as the second wavelength, 1804 nm as the third wavelength, 2054 nm as the fourth wavelength, and 1388 nm as the fifth wavelength are combinations of these wavelengths.

Therefore, when it is determined whether or not the foreign object F is mixed in the object M to be examined using the foreign object mixing determination device S in the object according to this embodiment, it is as follows.
In the total control calculation processing unit 42 of the control unit 40, a combination of the discriminant stored by the discriminant storage function and the wavelength of the selected near infrared ray is set. This will be described with reference to the flowcharts shown in FIGS.

  As shown in FIG. 4, the main power supply switch 4 is pressed to supply electricity to the control unit power supply 49, the spectroscopic unit power supply 52, and the like. When the main power supply switch 4 is turned on, the control unit 40 is activated to check each mechanism such as the spectroscopic unit 50, the sensor, and the lamp (1-1). At this time, a screen indicating that the system is being initialized is displayed on the display unit 46, and the indicator lamp is lit yellow (1-2). After completion of the check operation of each mechanism, a reference waveform serving as a reference for calculating absorbance is measured (1-3) and stored in a storage device inside the control unit 40 (1-4). After that, the main screen is displayed and the indicator lights green (1-8). In the meantime, if there is a setting value call operation (1-5) from the setting panel 47, a confirmation screen for the lamp lighting accumulated time, date time, etc. is displayed (1-6). In order to perform discrimination measurement, the setting panel 47 is operated to return to the main screen (1-7).

  The object M to be examined is placed on the transport unit 10, and the start switch of the control button 48 is pressed when the indicator light on the main screen is green (1-9). When the start switch is pressed, sample measurement (measurement of the object M to be examined) and a discrimination routine are executed (1-10).

  In this sample measurement (measurement of the object M to be inspected) and the discrimination routine, as shown in FIG. 5, first, the control unit 40 instructs the operation of the belt conveyor 10a of the transport unit 10 (2-1). The object M to be examined is transported. The object M to be inspected is detected by the input of the sample detection sensor 8 to detect the presence or absence of the object M to be inspected (2-2), and the shutter before the light source of the light source unit 20 and the light receiving unit 30 is instructed to open. (2-3). When the shutter is opened, the object M to be inspected is irradiated with white light, the reflected light or the transmitted light is received from the detection unit in the light receiving unit 30, transmitted to the spectroscopic unit 50 through the optical fiber, and the spectroscopic unit 50. Then, the intensity of the near infrared wavelength is extracted by converting the voltage for each wavelength range and transmitted to the control unit 40 by the control cable 51 (2-4). The intensity spectrum of the reflected light is obtained, and the presence / absence analysis of foreign matter F is performed by the foreign matter discrimination routine (2-5). When the detection of the reflected light or transmitted light is completed, the shutter is closed by a shutter closing instruction (2-6).

  In the foreign matter determination routine, as shown in FIG. 6, the noise removal processing is performed on the reflection intensity (reflection spectrum) transmitted from the spectroscopic unit 50 (3-1, 3-2), and the reference waveform stored at the time of initialization is stored. Reading (3-3) and calculating the near-infrared absorbance of the measurement sample (object M to be examined) (3-4). A standardization process is performed on the absorbance (3-5), a second derivative process is performed (3-6), and a centering and scaling process is performed on the second derivative spectrum (3-7). Using this spectrum, discriminant analysis (3-8), multiple regression analysis (3-9), PLS analysis (3-10), and neural network analysis back-propagation (3- 11) Each discrimination score and parameter are calculated by the method of the support vector machine (3-12). Using the discrimination score by each discrimination method, the optimum combination comparison (standardized comparison) is performed (3-13), and the final discrimination result is calculated (3-14).

  When the sample measurement (measurement of the object M to be examined) is completed, the process returns to FIG. 4 to display the discrimination result and the cumulative total of the discrimination results so far (1-13). In the meantime, when an error occurs or an emergency stop button is pressed in each mechanism (1-11), an error occurrence screen is displayed and the indicator light turns red (1-12). To return, it is necessary to check the mechanism of the lamp, spectroscopic unit 50, sensor, etc. The discrimination result and the reflection spectrum at the time of sample measurement (measurement of the object M to be examined) are stored in the storage device S in the control unit 40. This flow is repeated until the stop button of the control button is pressed (1-15).

Next, experimental examples will be described.
(Experimental example 1)
First, it is determined whether or not the foreign matter F is mixed in the object M using the discriminant calculated by the discriminant analysis method and the absorbance of the light received from the object M to be examined.

(Experimental Example 1-1)
In this experiment, confectionery such as cake, bun, and chocolate is used as a population for creating discriminants, and hair is artificially mixed in the surface and inside (intermediate part, deep part) of these confections and their reflections. Light was detected, absorbance was calculated, standardization processing was performed, secondary differentiation processing, scaling and centering were performed, and a discriminant was obtained for the spectrum.

  The discriminant used two wavelengths of 2054 nm, 2162 nm, and 1740 nm, which are considered to have absorption in the related N—H bond and S—H bond, according to the near-infrared absorption band of cystine. The discriminant at this time has coefficients of a0 = 0, a1 = 0.215, a2 = 2.164, a3 = 0.272, a4 = −1.492, a5 = −0.508, The wavelength λ1 = 2205 nm, the second wavelength λ2 = 2162 nm, the third wavelength λ3 = 1740 nm, the fourth wavelength λ4 = 1604 nm, and the fifth wavelength λ5 = 1388 nm. As a result, as shown in FIGS. 7 to 9, the discrimination rate was 98% for the presence of foreign matter and 100% for the absence of foreign matter.

(Experimental example 1-2)
In this experiment, confectionery such as cake, bun, and chocolate is used as a population for creating discriminants, and hair is artificially mixed in the surface and inside (intermediate part, deep part) of these confections and their reflections. Light was detected, absorbance was calculated, standardization processing was performed, secondary differentiation processing, scaling and centering were performed, and a discriminant was obtained for the spectrum. At that time, in order to further improve the accuracy, the Mahalanobis general distance with each group was also examined.

  As a discriminant, wavelength selection was performed by using a stepwise variable selection method with an input F value of 2.0 and a removal F value of 2.0 as a reference value for increasing the stepwise variable. The discriminant at this time has the following coefficients: a0 = 0, a1 = −0.525, a2 = 1.064, a3 = 1.021, a4 = −1.191, a5 = 1.649. The wavelength λ1 = 1134 nm, the second wavelength λ2 = 1260 nm, the third wavelength λ3 = 1492 nm, the fourth wavelength λ4 = 1618 nm, and the fifth wavelength λ5 = 2172 nm. As a result, as shown in FIGS. 10 to 12, the discrimination rate was 100%.

(Experimental example 2)
Next, it is determined whether or not the foreign matter F is mixed in the object M using the discriminant calculated by the multiple regression analysis method and the absorbance of the light received from the object M to be examined.

(Experimental example 2-1)
In this experiment, cakes, buns, chocolates, and other confectionery were used, and the surface and interior (intermediate part, deep part) of these confectionery were artificially mixed with hair, irradiated with white light, and the reflected light was detected. Then, the absorbance was calculated and standardized to obtain a second derivative spectrum. Focusing on the second derivative spectrum, paying attention to cystine which is a component of hair, the same absorption band as shown in FIG. 13 and the absorption bands of N—H and S—H bonds are selected from 2054 nm, 2162 nm and 1740 nm, The fourth wavelength and the fifth wavelength are combined, and as shown in FIGS. 14 to 16, a discrimination formula of 98% for foreign matter discrimination and 92% discrimination for foreign matter is constructed. The coefficients of the discriminant at this time are K0 = 0.381, K1 = −0.093, K2 = −0.092, K3 = −0.117, K4 = 0.634, K5 = 0.219, One wavelength λ1 = 2205 nm, second wavelength λ2 = 2162 nm, third wavelength λ3 = 1740 nm, fourth wavelength λ4 = 1604 nm, and fifth wavelength λ5 = 1388 nm.

(Experimental example 2-2)
In this experiment, cakes, buns, chocolates, and other confectionery were used, and the surface and interior (intermediate part, deep part) of these confectionery were artificially mixed with hair, irradiated with white light, and the reflected light was detected. Then, the absorbance was calculated and standardized to obtain a second derivative spectrum. The regression equation was constructed by performing multiple regression analysis on the second derivative spectrum at the selected wavelength in discriminant analysis. As shown in FIG. 17 to FIG. 19, a discriminant of 100% for foreign matter discrimination and 96% for foreign matter discrimination was constructed. The coefficients of the discriminant at this time are K0 = 0.382, K1 = 0.216, K2 = −0.660, K3 = −0.422, K4 = 0.490, K5 = −0.991, One wavelength λ1 = 1134 nm, second wavelength λ2 = 1260 nm, third wavelength λ3 = 1492 nm, fourth wavelength λ4 = 1618 nm, and fifth wavelength λ5 = 2172 nm.

(Experimental example 3)
Next, it is determined whether or not the foreign substance F is mixed in the object M using the discriminant calculated by the PLS analysis method and the absorbance of the light received from the object M to be examined.

  In this experiment, cakes, buns, chocolates, and other confectionery were used, and the surface and interior (intermediate part, deep part) of these confectionery were artificially mixed with hair, irradiated with white light, and the reflected light was detected. Then, the absorbance is calculated, the standardization process is performed, the second derivative process is performed, and the spectrum subjected to the scaling and centering process is applied to the PLS by applying the GA analysis method around the selected wavelength used in the discriminant analysis. A discriminant was constructed by the analytical method. The coefficients of the discriminant at this time are a0 = 0, a1 = −0.526, a2 = −0.366, a3 = −0.810, a4 = 0.571, a5 = −0.402, a5 = 0 967, a5 = −1.168, first wavelength λ1 = 1178 nm, second wavelength λ2 = 1252 nm, third wavelength λ3 = 1276 nm, fourth wavelength λ4 = 1364 nm, fifth wavelength λ5 = 1556 nm, fifth wavelength λ5 = 1614 nm and the fifth wavelength λ5 = 2164 nm. The results are shown in FIGS. As a result, as shown in FIG. 22, a calibration curve having a foreign matter presence / absence discrimination rate of 100% was constructed.

(Experimental example 4)
Next, it is determined whether or not the foreign matter F is mixed in the object M using the discriminant calculated by the back propagation method of the neural network method and the absorbance of the light received from the object M to be examined. .

(Experimental example 4-1)
In this experiment, cakes, buns, chocolates, and other confectionery were used, and the surface and interior (intermediate part, deep part) of these confectionery were artificially mixed with hair, irradiated with white light, and the reflected light was detected. The absorbance is calculated, the standardization process is performed, the second derivative process is performed, and the spectrum obtained by scaling and centering the spectrum is the absorption band of N—H and S—H, which is the absorption band of cystine. Based on the wavelengths 2054nm, 2162nm, and 1740nm, the 4th and 5th wavelengths are added, the input layer 5, the intermediate layer 2, the output layer 1, and the model is trained and trained under the condition of 8000 educations. Formula. In the discriminant, as shown in FIGS. 23 to 25, the discrimination rate is 100%.

(Experimental example 4-2)
In this experiment, cakes, buns, chocolates, and other confectionery were used, and the surface and interior (intermediate part, deep part) of these confectionery were artificially mixed with hair, irradiated with white light, and the reflected light was detected. The absorbance is calculated, the standardization process is performed, the second derivative process is further performed, and the spectrum subjected to the scaling and centering process is performed on the spectrum using the selected wavelength used in the discriminant analysis, the input layer 5, the intermediate layer 2, Modeling was performed under the condition of the output layer 1 and the number of educations of 8,000 times, and this was used as a discriminant. In the discriminant, as shown in FIGS. 26 to 28, the discrimination rate is 100%.

(Experimental example 5)
Next, it is determined whether or not the foreign matter F is mixed in the object M using the discriminant calculated by the support vector machine (SVM) method and the absorbance of the light received from the object M to be examined.

(Experimental example 5-1)
In this experiment, absorbance data in the wavelength range of 1100 to 2500 nm was obtained using a total of 84 food items, including hair contaminated food with foreign matter (39 cases) and non-contaminated hair non-contaminated food (45 cases). The SVM program of the RBF function of the C-SVM classification method or Nu-SVC classification method using a function was constructed with LabVIEW (V8.5), and the data of known foods with or without foreign matter F were learned. As a result, as shown in FIG. 29, an SVM classification function having a discrimination rate of 91.67% could be obtained.

  Then, using this data, learning was performed by the SVM method using the sigmoid function in the c-svc classification method. As a result, as shown in FIG. 30, a discrimination rate of 100% was obtained. That is, it can be said that an extremely effective SVM function for detecting hair was obtained.

(Experimental example 5-2)
This experiment learned absorbance data in the wavelength range of 1100 to 2500 nm using a total of 39 food items including hair-contaminated food with foreign matter (13 cases) and non-contaminated hair-free food (26 cases). This was performed using the SVM function. As a result, as shown in FIG. 31, the discrimination rate was 87.18%. From this result, although the discrimination rate tends to be low in the hair non-contaminated food group, the discrimination rate is good in the hair contaminated food group, and it is possible to discriminate the presence or absence of hair contamination in the SVM method. I can say that.

(Experimental Example 5-3)
This experiment uses SVM selection learning function in the case of wavelength selection, using a total of 39 food items including hair-contaminated food with foreign matter (13 cases) and hair-non-contaminated food without foreign matter (26 cases). Carried out. The results are shown in FIG. As a result, the discrimination rate was 97.4% in 38 cases out of 39 cases.

  From the above experimental results of the SVM method, it was found that hair detection can be easily performed by creating hair detection learning by the SVM method. That is, it became clear that the SVM method is an extremely effective means as one of hair detection techniques. In particular, the method based on the selected wavelength is very good, but it can be inferred that any wavelength selection method can be used, for example, GA (genetic algorithm) or stepwise method can be used to obtain good results. Is. This experimental example is not limited to the wavelength used in the SVM method.

(Experimental example 6)
Next, using the discriminant consisting of several discriminant methods among the discriminant analysis method, the multiple regression analysis method, the PLS analysis method, the back propagation method of the neural network method, and the SVM method, Whether or not the foreign matter F is mixed in the object M is determined using the absorbance of the light received from the object M.

(Experimental example 6-1)
In this experiment, several kinds of cakes, several kinds of buns, and several kinds of cream puffs were used, and several kinds of hair were artificially mixed on the surface and inside (intermediate part, deep part) of these confectionery. Irradiate white light, detect its reflected light, calculate its absorbance, perform standardization, and calculate the second derivative spectrum. This is a backprop of discriminant analysis method, multiple regression analysis method, neural network method. Each discriminant of the gation method was used. The results are shown in FIG. 33 and FIG. As a result, the discrimination rate was 100%.

(Experimental example 6-2)
In this experiment, several kinds of cakes, several kinds of buns, and several kinds of cream puffs were used, and several kinds of hair were artificially mixed on the surface and inside (intermediate part, deep part) of these confectionery. Irradiate white light, detect its reflected light, calculate its absorbance, perform standardization to calculate the second derivative spectrum, discriminant analysis method, multiple regression analysis method, PLS method, neural network method Each discriminant of the backpropagation method was used. The results are shown in FIGS. 35 and 36. As a result, the discrimination rate was 100%.

(Experimental example 6-3)
In this experiment, several types of cakes, several types of buns, and several types of cream puffs were used, and three types of hair were artificially mixed inside the food, such as in the middle or deep of each food, and foreign matter discrimination analysis was performed. As a result, the calculation result shown in FIG. 37 was obtained. The total result is shown in FIG. As a result, a satisfactory result with a discrimination rate of 100% was obtained.

  From the above results, when each discriminant is combined, the discrimination rate becomes 100%. Therefore, even if the foreign matter F is mixed in the food surface or inside, the final judgment is made by the combination by each discrimination method. It can be said that foreign matter contamination can be determined with higher accuracy.

FIG. 39 shows the foreign matter contamination determination apparatus S proposed in the development process of the present invention, and is disclosed for reference. In this apparatus S, unlike the above-described embodiment, the spectroscopic unit 50 is incorporated in the light source unit 20. That is, the light emitted from the light source unit 20 is dispersed for each wavelength range by the acoustooptic element of the spectroscopic unit 50 through the spectroscopic unit 50, and the dispersed light is irradiated to the object M to be examined. The reflected light or transmitted light from the object M is received by the light receiving unit 30 for each wavelength range.
Other configurations are the same as those in the above embodiment.

  In the above embodiment, foreign matter contamination determination in an object is performed by combining five different determination methods: discriminant analysis method, multiple regression analysis method, PLS analysis method, neural network backpropagation method, and SVM method. However, the present invention is not necessarily limited to this, and only one type of determination method may be used, or several types of methods may be combined, and may be appropriately changed.

S Foreign matter mixing discrimination device M Object F Foreign matter 1 Machine base 2 Partition plate 3 Placement plate 4 Main power supply switch 5 Emergency stop button 6 Heat exhaust fan 7 Control cable wiring port 8 Sample detection sensor 10 Conveying section 10a Belt conveyor 20 light source unit 30 light receiving unit 31 light receiving element 40 control unit 41 communication unit 42 total control calculation processing unit 43 motor control circuit 44 spectral control circuit 45 operation command unit 46 display unit 47 setting panel 48 control button 49 control unit power supply 50 spectral unit 51 Control cable 52 Power supply for spectrometer

Claims (6)

In the foreign matter contamination determination device in the object for determining whether or not foreign matter that is animal hair containing cystine as a specific component other than the component constituting the object is mixed in the object that is food ,
A light source unit that irradiates light in the near-infrared region to the object to be examined, a light receiving unit that receives reflected or transmitted light from the object, and a light receiving unit that receives the light received by the light receiving unit. A control unit for determining the presence or absence of foreign matter in the object to be inspected,
The control unit is preliminarily analyzed by statistical analysis of the second derivative spectrum in the absorbance with respect to the wavelength in the near-infrared region irradiated and reflected from the sample object with and without the foreign object. A discriminant storage function for storing a discriminant relating to the assigned wavelength caused by the calculated specific component, and a foreign matter in the object from a calculation result calculated from the absorbance of the light received by the light receiving unit and the discriminant. With a foreign matter discrimination function to determine the presence or absence of
The light source unit is configured to irradiate light in the near infrared region of a predetermined wavelength range,
A spectroscopic unit that splits the light received by the light receiving unit is provided,
The foreign matter discrimination function of the control unit is configured to determine the presence or absence of foreign matter in an object based on the absorbance of light dispersed by the spectroscopic unit ,
The discriminant stored in the control unit is calculated by discriminating and analyzing the absorbance spectrum of the light received from the sample object by second-order differentiation. An apparatus for determining foreign matter contamination in an object, which is calculated by an expression satisfying a relationship of the following general formula (A-1) using absorbance as a variable:

In general formula (A-1), zi is the discrimination score of each group, λ is the wavelength, A is the absorbance, A1 (λ1) is the absorbance at the first wavelength (λ1), and A2 (λ2) is the second wavelength (λ2). An (λn) is the absorbance at the nth wavelength (λn), a0, a1, a2,... Are the coefficients that determine the discriminant score zi, measured in a sufficiently large population It is determined by solving equation (1-3) so that the following equation (1-1) takes the maximum value from the absorbance and the discrimination score.

In Formula (1-1), F (a1, a2) represents that zi is a function determined by coefficients a1 and a2, Sb represents a variation between groups, St represents a discrimination score for the entire group and its average Is the sum of squares of the difference, that is, the variation of the entire group.
In Equation (1-2), St is the total sum of squares of the difference between the discrimination score of the entire group and its average, and Sw is the sum of the sum of squares of the discrimination scores within the group and the average difference thereof. , Sb is the variation between groups as the sum of the sum of squares of the average difference between the discrimination score in each group and the discrimination score of the entire group.
For the wavelength in the discriminant, the first wavelength is selected from the range of 1320 to 1460 nm, 1530 to 1542 nm, 1770 to 1776 nm, and 2154 to 2206 nm, and the second wavelength is 1144 to 1158 nm, 1252 to 1276 nm, 1320 to 1338 nm, 1346 to 1384 nm, 1406 to 1434 nm, 1470 to 1510 nm, 1598 to 1650 nm, and the third wavelength is 1100 to 1138 nm, 1230 to 1324 nm, 1352 to 1378 nm, 1394 to 1428 nm, 1480 to 1502 nm, 1540 to 1580 nm, 1610 to 1656 nm 1664 to 1696 nm, 1742 to 1750 nm, 1760 to 1776 nm, 1816 to 1850 nm, 1880 to 1890 nm, 2112 to 2120 nm, 2 Select from the range of 22 to 2340 nm, and the fourth wavelength is 1100 to 1122 nm, 1146 to 1168 nm, 1250 to 1282 nm, 1350 to 1382 nm, 1558 to 1576 nm, 1676 to 1696 nm, 1740 to 1852 nm, 1808 to 1866 nm, 2124 to 2140 nm, 2258 Select from the range of ˜2282 nm, 2322 to 2340 nm, and select the fifth wavelength from the range of 1100 to 1142 nm, 1152 to 1210 nm, 1248 to 1326 nm, 1394 to 1420 nm, 1590 to 1640 nm, 1880 to 1884 nm, 2400 to 2420 nm, and these A combination that does not exhibit multi-collinearity is selected from one of the wavelength ranges in the above. The wavelength in the discriminant is different from the first wavelength in the range of 1354 to 1386 nm, 1518 to 1540 nm, 1578 to 1612 nm, 1692 to 1694 nm, 1734 to 1744 nm, 1810 to 1870 nm, 2048 to 2070 nm, and 2150 to 2202 nm. And select the second wavelength from the range of 1352-1386 nm, 1566-1618 nm, 1682-1694 nm, 1732-1748 nm, 1810-1872 nm, 2048-2086 nm, 2152-2166 nm, 2190-2198 nm, 2252-2266 nm, 2322-2334 nm The third wavelength is selected from the range of 1782 to 1808 nm and 2448 to 2454 nm, and the fourth wavelength is 1222 to 1228 nm, 1276 to 1286 nm, 62 to 1380 nm, 1572 to 1606 nm, 1726 to 1740 nm, 1810 to 1844 nm, 2042 to 2088 nm, 2150 to 2166 nm, 2258 to 2272 nm, and the fifth wavelength is 1372 to 1392 nm, 1524 to 1548 nm, 1834 to 1878 nm, 2190 2. The foreign matter contamination determination apparatus according to claim 1, wherein the combination is selected from a range of ˜2214 nm and does not exhibit multi-collinearity in which one is selected from a wavelength range of each of these ranges. An object to be inspected is placed and driven, and the placed object is irradiated with light from the light source unit, and reflected or transmitted light from the object can be received by the light receiving unit. 3. The foreign matter contamination determination device in an object according to claim 1 , further comprising a conveyance unit that sequentially conveys the position. 4. The foreign matter contamination determination device according to claim 1 , wherein the light source unit is configured to output light through an optical fiber and include a waveguide of the optical fiber. 5. The foreign matter mixed in the object according to claim 1, wherein the light receiving unit includes a plurality of light receiving elements that receive reflected light or transmitted light from the object to be examined. Discriminator. 6. The foreign matter contamination determination device in an object according to claim 5 , wherein the light receiving element is constituted by an optical fiber.

Images