JP2021117091A - Optical granule discrimination device - Google Patents

Abstract

To provide an optical granule discrimination device capable of accurately determining quality of granules even if good and defective granules are similar in color and shape.SOLUTION: An inspection unit for optically inspecting granules being conveyed by conveying means comprises a visible light source, near-infrared light source, visible light detection unit, and near-infrared light detection unit. A determination unit plots one wavelength components out of red (R), green (G), and blue (B) light detected by the visible light detection unit from a plurality of good samples and defective samples and a plurality of near-infrared components detected by the near-infrared detection unit to create a three-dimensional optical correlation diagram and set thresholds.SELECTED DRAWING: Figure 15

Description

The present invention relates to an optical granularity discriminating device for discriminating whether a granular material such as cereals, beans, seeds, etc. is a good product or a defective product based on an optical inspection.

For grains such as cereals, it is necessary to sort and separate foreign substances such as rice husks and stones and defective products from non-defective products such as brown rice. Therefore, various devices have been proposed for discriminating between good and bad granules by optical inspection and eliminating defective products. For example, in Patent Document 1, the particles transferred by the transfer means are irradiated with light, and the wavelength components of R (red), G (green), and B (blue) light transmitted or reflected by the particles are ternary. An optical granular material sorter that plots on the original color space, determines whether the product is a good product or a defective product from the three-dimensional color distribution of the granular material, and eliminates only the defective product is disclosed.

Japanese Patent No. 6037125

When discrimination is performed using the three wavelength components of light of red (R), green (G), and blue (B) with an optical granular material sorter as described in Patent Document 1, a good product and a defective product are obtained. If the colors and shapes of the lights are similar, for example, black stones are mixed in the black beans, or harmful substances such as black mycelium are mixed in the seeds of black sunflowers, as shown in FIG. As shown, the set of good products and the set of defective products overlapped, and it was not possible to determine whether the product was good or bad.

Therefore, the problem to be solved by the present invention is to provide an optical granular material discriminating device capable of discriminating good or bad with high accuracy even if a good product and a defective product of the granular material have similar colors and shapes. To do.

The invention according to claim 1 of the present application includes an inspection unit that optically inspects the granules transferred by the transfer means, and whether the granules are good or defective based on the optical inspection by the inspection unit. In an optical granular matter discriminating device provided with a determination unit for discriminating between, the inspection unit includes a visible light source that irradiates the granular material with visible light and a near-infrared light source that irradiates the granular material with near-infrared light. And the visible light detection unit that detects visible light transmitted through the granules or visible light reflected from the granules, and near-infrared light transmitted through the granules or near-infrared light reflected from the granules. It has at least a near-infrared light detection unit for detection, and the determination unit has red (R) and green (G) detected by the visible light detection unit for a plurality of non-defective sample and a plurality of defective samples. , One wavelength component of the blue (B) light and a plurality of near-infrared light components detected by the near-infrared light detection unit are plotted in a three-dimensional space to create a three-dimensional optical correlation diagram. It is an optical granular matter discriminating device characterized by setting a threshold value.

In the invention according to claim 2 of the present application, the determination unit creates a Mahalanobis distance boundary surface and an Euclidean distance boundary surface for partitioning a non-defective product aggregate and a defective product aggregate on the three-dimensional optical correlation diagram, and the Mahalanobis distance boundary surface. A two-dimensional plane perpendicular to the intersection of the surface and the Euclidean distance boundary surface is created, and an inertial equivalent ellipse is applied to the defective aggregate on the two-dimensional plane to create a closed region, and a threshold value is created in the closed region. The optical granular material discriminating device according to claim 1.

In the invention according to claim 3 of the present application, the determination unit determines red (R), green (G), and blue (B) detected by the visible light detection unit with respect to a plurality of non-defective sample samples and a plurality of defective sample samples. Of the light, one wavelength component and two near-infrared light components obtained by combining a plurality of types detected by the near-infrared light detection unit are plotted in a three-dimensional space to create a plurality of types of three-dimensional optical correlation diagram. The optical granular material discriminating device according to claim 1 or 2.

The invention according to claim 4 of the present application claims that the near-infrared light component detected by the near-infrared light detection unit is a near-infrared light component within the contour of the granular material detected by the visible light detection unit. The optical granular material discriminating device according to 1 to 3.

In the invention according to claim 5, when the determination unit sets the threshold value in the closed region, it is parallel to the minor axis of the inertial equivalent ellipse and parallel to the two straight lines passing through both end points of the major axis and the major axis. Any of claims 1 to 4 for creating an extrinsic rectangle composed of two straight lines passing through both end points of the short axis and creating two straight lines connecting the center of gravity of the non-defective product aggregate and the end points of the extrinsic rectangle in the major axis direction. It is an optical granular matter discriminating apparatus described in the ellipse.

In the invention according to claim 6, the determination unit uses the Maharanobis distance boundary surface that minimizes the Maharanobis distance as the first plane when setting the threshold value in the closed region, and uses the second plane. As a third plane, a plane connecting the center of gravity of the non-defective product assembly and one end of the circumscribing rectangle in the long axis direction is used, and as a third plane, a plane connecting the center of gravity of the non-defective product assembly and the other end of the circumscribing rectangle in the long axis direction is used. The long side of one side of the circumscribing rectangle is used as the fourth plane, the long side of the other side of the circumscribing rectangle is used as the fifth plane, and the side far from the good product aggregate is used as the sixth plane. The optical granular material discriminating device according to any one of claims 1 to 5, which utilizes one short side of the circumscribing rectangle.

The invention according to claim 7 of the present application includes a display and input means capable of input by an operator based on the display contents on the display, and the display includes red (R), green (G), and blue ( A plurality of the three-dimensional optical correlation diagrams of the arbitrary one wavelength component and any two near-infrared light components of the light of B) can be displayed, and the input means can be operated by the operator. The optical granular material discriminating device according to any one of claims 1 to 6, wherein the three-dimensional optical correlation diagram displayed on the display can be selected based on the above.

In the invention according to claim 8, the inspection unit includes a first inspection unit located on the front side of the granular material to be transferred and a second inspection unit located on the rear side of the granular material. The optical granular material according to any one of claims 1 to 7, wherein the first inspection unit and the second inspection unit include the visible light detection unit and the near-infrared light detection unit, respectively. It is a discrimination device.
The inspection unit includes a first inspection unit located on the front side of the granular material to be transferred and a second inspection unit located on the rear side of the granular material, and includes the first inspection unit and the said inspection unit. The second inspection unit is the optical granular material discriminating device according to any one of claims 1 to 7, further comprising the visible light detection unit and the near-infrared light detection unit, respectively.

The invention according to claim 9 of the present application is the optical granularity discriminating apparatus according to any one of claims 1 to 8, wherein the granular material transferred by the transfer means is a seed or a grain.

The invention according to claim 10 of the present application includes an inspection unit that optically inspects the granules transferred by the transfer means, and whether the granules are good or defective based on the optical inspection by the inspection unit. In an optical granular matter discriminating device provided with a determination unit for discriminating between, the inspection unit includes a visible light source that irradiates the granular material with visible light and a near-infrared light source that irradiates the granular material with near-infrared light. And the visible light detection unit that detects visible light transmitted through the granules or visible light reflected from the granules, and near-infrared light transmitted through the granules or near-infrared light reflected from the granules. It has at least a near-infrared light detection unit for detection, and the determination unit has red (R) and green (G) detected by the visible light detection unit for a plurality of non-defective sample and a plurality of defective samples. , Blue (B) and a plurality of near-infrared light components detected by the near-infrared light detection unit are used as parameters for multivariate analysis, and based on the result of the multivariate analysis, the visible light detection is performed. Optical granularity discrimination characterized by plotting the wavelength component detected by the unit and the near-infrared light component detected by the near-infrared light detection unit to create a three-dimensional optical correlation diagram and setting a threshold value. It is a device.

According to the inventions according to claims 1 and 2, one wavelength component of the red (R), green (G), and blue (B) lights detected by the visible light detection unit and the near-infrared light detection unit. Based on the plurality of near-infrared light components detected by, a three-dimensional optical correlation diagram was created, a closed region was created in the defective aggregate, and a threshold value was set in the closed region. This makes it possible to accurately discriminate granules that could not be discriminated by the three wavelength components of red (R), green (G), and blue (B) conventionally detected by the visible light detection unit. became.

According to the invention of claim 3, one wavelength component of the red (R), green (G), and blue (B) lights detected by the visible light detection unit and the near-infrared light detection unit detect it. Based on the two different near-infrared light components, a three-dimensional optical correlation diagram of a plurality of types (for example, G.850 nmNIR / 1550 nmNIR, R / 1200 nmNIR / 1550NIR, etc.) is created. As a result, a plurality of three-dimensional optical correlation diagrams can be compared relative to each other, and the wavelength component of visible light most suitable for discriminating granules can be selectively configured.

According to the invention of claim 4, in order to accurately detect the near-infrared light component in the granular material to be discriminated, the inside of the contour of the granular material is matched with the contour of the granular material detected by the visible light detection unit. I am trying to detect the near-infrared light component of. This makes it possible to discriminate between good and bad granules with high accuracy.

According to the inventions according to claims 5 and 6, when setting a threshold value in the closed region, the Maharanobis distance boundary surface that minimizes the Maharanobis distance is used as the first plane, and a good product set is used as the second plane. A plane connecting the center of gravity of the body and one end of the circumscribing rectangle in the long axis direction is used, and a plane connecting the center of gravity of the non-defective assembly and the other end of the circumscribing rectangle in the long axis direction is used as the third plane. The long side of one side of the circumscribing rectangle is used as the fourth plane, the long side of the other side of the circumscribing rectangle is used as the fifth plane, and the extrinsic side far from the non-defective aggregate is used as the sixth plane. By using the short side on one side of the rectangle, it is possible to discriminate between good and bad with high accuracy.

According to the invention of claim 7, it is possible to select and display the three-dimensional optical correlation diagram displayed on the display based on the operation of the operator. As a result, when selecting one wavelength component from the red (R), green (G), and blue (B) lights, the operator selects the wavelength component having the most significant difference (high correlation). This makes it possible to discriminate between good and bad with high accuracy.

According to the invention of claim 8, a first inspection unit and a second inspection unit are provided before and after the transferred granular material, and the inspection unit can obtain a visible light image and a near-infrared light image, respectively. For example, when the determination result of a defective product is output by at least one inspection unit, it can be determined as a defective product, and the accuracy of determining whether the product is good or defective can be improved.

According to the invention of claim 9, it is possible to discriminate between good and bad seeds other than the black sunflower seeds shown in the embodiment and grains such as rice with high accuracy.

According to the invention of claim 10, the three wavelength components of red (R), green (G), and blue (B) detected by the visible light detection unit with respect to the plurality of non-defective sample and the plurality of defective samples, and Multivariate analysis is performed using multiple near-infrared light components detected by the near-infrared light detector as parameters, and based on the results of the multivariate analysis, the wavelength components detected by the visible light detector and the near-infrared light detector The near-infrared light component detected by is plotted to create a three-dimensional optical correlation diagram, and the threshold value is set. As a result, the correlation between the three wavelength components of red (R), green (G), and blue (B) detected by the visible light detection unit and the near-infrared light components such as 850 nm, 1200 nm, and 1550 nm is derived. Therefore, it is possible to create an optimum three-dimensional optical correlation diagram for discriminating between a non-defective product and a defective product and set a threshold value.

It is a schematic cross-sectional view of the optical granular matter discriminating apparatus which shows 1st Embodiment of this invention. It is a schematic block diagram of the optical granular matter discriminating apparatus which concerns on 1st Embodiment of this invention. It is a detailed block diagram of the determination part which concerns on 1st Embodiment of this invention. Images of good and defective samples of granular material according to the first embodiment of the present invention are shown, and in order from the top, visible light images captured by the CCD cameras 13a and 13b, and the first wavelength NIR camera 14 The near-infrared light images taken by -1a and 14-1b and the near-infrared light images taken by the first wavelength NIR cameras 14-1a and 14-1b are fitted to the contours of the granules obtained by the CCD cameras 13a and 13b. Object recognition in which the near-infrared light image after the embedded object recognition and the near-infrared light image by the second wavelength NIR cameras 14-2a and 14-2b are embedded in the contour of the granular material obtained by the CCD cameras 13a and 13b. It is a later near-infrared light image. It is a flow chart which showed the processing procedure of a signal processing part. It is a flow chart which showed the calculation procedure of the threshold value. It is a figure which showed the calculation formula related to the threshold value calculation. 6 is a three-dimensional optical correlation diagram of G, a first wavelength NIR, and a second wavelength NIR in a three-dimensional space of a non-defective sample and a defective sample according to the first embodiment of the present invention. It is a G / first wavelength NIR / second wavelength NIR correlation diagram in the optimum two-dimensional plane of a non-defective product sample and a defective product sample according to the first embodiment of the present invention. It is a figure when an inertial equivalent ellipse is applied to the defective aggregate of the optimum two-dimensional plane. It is the figure when the circumscribed rectangle is applied to the inertial equivalent ellipse. It is a figure when two straight lines connecting the both end points in the long axis direction of the circumscribed rectangle of the inertial equivalent ellipse and the center of gravity of a good product aggregate are created. It is a figure at the time of making six planes when forming a closed region surrounding an inertial equivalent ellipse. It is a figure explaining the exception in the drawing of a two-dimensional plane. It is a figure when the calculation of the threshold value created by 6 planes surrounding an inertial equivalent ellipse is made to correspond with 3 sensitivity levels of minimum (MIN), intermediate (MID) and maximum (MAX). This is three-dimensional color distribution data related to a conventional optical granular material sorter. It is a typical visible light image and near infrared light image of the granular matter to be discriminated in another embodiment.

Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings.

[First Embodiment]
The first embodiment of the present invention will be described with reference to the drawings. The optical granule discriminating device 1 in the present embodiment is a black sunflower seed that conforms to a standard that is a non-defective product from among the granules to be discriminated based on the threshold value determined based on the optical inspection. It is a device that discriminates between defective products (harmful substances such as mycelium, stones, non-standard products, etc.) mixed in black sunflower seeds.

FIG. 1 shows a schematic cross-sectional view of the optical granular material discriminating device 1. As shown in the figure, the optical granular matter discriminating device 1 of the present embodiment has an inclined chute 3 which is a means for transferring granular matter, a storage tank 4 for storing granular matter, and a storage tank 4 in the machine body 2. A vibration feeder 5 that conveys particles to the upper end of the chute 3, an inspection unit 6 that optically inspects the particles that fall from the chute 3, an ejector nozzle 7 provided below the inspection unit 6, and an ejector nozzle. A non-defective product collection unit 8 that receives the non-defective product that has fallen without receiving the blown air of the ejector nozzle 7 and a defective product collecting unit 9 that collects the defective product that has been blown off by the blown air of the ejector nozzle 7 are installed and configured. Further, on the front side of the machine body 2, at least a door 10 that can be opened and closed for performing various maintenance and a display 11 that also serves as a touch panel (input means) and a monitor for operation are installed.

The inspection unit 6 that performs an optical inspection on the granular material is housed inside the front box 12a and the rear box 12b installed with the falling locus a of the granular material in between. Inside the front box 12a, there are a CCD camera 13a which is a visible light detection unit, a first wavelength NIR camera 14-1a and a second wavelength NIR camera 14-2a which are near infrared light detection units, and a visible light source 15a. A near-infrared light source 16a and a background 17a of a CCD camera 13b, which will be described later, are accommodated. Further, in the rear box 12b, a CCD camera 13b which is a visible light detection unit, a first wavelength NIR camera 14-1b and a second wavelength NIR camera 14-2b which are near infrared light detection units, and a visible light source 15b And the near-infrared light source 16b and the background 17b of the CCD camera 13a are accommodated. The first wavelength NIR cameras 14-1a and 14-1b of the present embodiment can detect near-infrared light of 850 nm and take an image, and the second wavelength NIR cameras 14-2a, 14-2b is capable of detecting near-infrared light having a wavelength of 1550 nm and taking an image.

Translucent portions 18a and 18b made of transparent plate-shaped members are provided on the sides of the front box 12a and the rear box 12b facing the falling locus a of the granular material. Then, the background 17a in the front box 12a faces the CCD camera 13b in the rear box 12b via the translucent portions 18a and 18b, and the background 17b in the rear box 12b is the CCD camera in the front box 12a. It is arranged so as to face 13a.

FIG. 2 shows a schematic block diagram of the optical granular material discriminating device 1 of the present embodiment. As shown, the CCD cameras 13a and 13b for detecting the visible light component reflected from the granules or transmitted through the granules, and the first near-infrared light component for detecting the reflected or transmitted near-infrared light component from the granules. The wavelength NIR cameras 14-1a and 14-1b and the second wavelength NIR cameras 14-2a and 14-2b are connected to the determination unit 19, and the signal processing unit 20 that processes the image data in the determination unit 19 is electrically connected. Is connected.

Further, the signal processing unit 20 is connected to the CPU / memory unit 21 so as to be capable of bidirectional communication. The CPU / memory unit 21 can store the image processed by the signal processing unit 20 and perform arithmetic processing described later to calculate a threshold value for discriminating granular matter. Further, the signal processing unit 20 is connected to the ejector nozzle 7 via the ejector drive circuit 22, and the CPU / memory unit 21 is connected to the display 11 having an input means such as a touch panel.

FIG. 3 shows a detailed block diagram of the determination unit 19 described above. As shown, the signal processing unit 20 receives image data captured by the CCD cameras 13a and 13b, the first wavelength NIR cameras 14-1a and 14-1b, and the second wavelength NIR cameras 14-2a and 14-2b. The image data acquisition unit 23 that temporarily stores the image data, the threshold data storage unit 24 that stores the threshold data calculated by the CPU / memory unit 21, and the image processing unit 25 that binarizes the image data are imaged. It is provided with a good / bad discriminating unit 26 for discriminating whether the granular material is a good product or a defective product. Then, a signal from the good / bad discriminating unit 26 is sent to the ejector drive circuit 22, and the valve of the ejector nozzle 7 is opened / closed.

As shown in FIG. 3, the CPU / memory unit 21 sets a threshold value based on the image data storage unit 27 that stores the image data from the image data acquisition unit 23 and the image data stored in the image data storage unit 27. At least a threshold calculation unit 28 for calculation and a signal transmission / reception unit 29 for receiving an operation signal from the input means of the display 11 and outputting image data to the display 11 are provided.

Hereinafter, the procedure for discriminating granular matter by the optical granular material discriminating device 1 of the present embodiment will be described. FIG. 5 shows a flow chart showing a processing procedure of the signal processing unit 20. In FIGS. 5, in steps 101 to 103, a non-defective product sample and a defective product (including foreign matter) sample prepared in advance by the operator are flowed through the chute 3, respectively, and three-dimensionally relating to the non-defective product and the defective product (including foreign matter). This is a non-defective / defective pattern learning step in which the determination unit 19 learns the optical correlation data.

In FIG. 5, steps 104 to 108 are threshold value calculation steps for automatically calculating a threshold value that is a boundary between a non-defective product pattern and a defective product pattern. Step 109 is a threshold value determination step that automatically adjusts the threshold value calculated in the threshold value calculation step.

(Good product pattern / defective product pattern learning process)
First, in step 101, the non-defective sample prepared in advance is flowed to the chute 3, and the non-defective sample dropped from the chute 3 is used as the CCD cameras 13a and 13b, the first wavelength NIR cameras 14-1a and 14-1b, and the second wavelength NIR camera 14-. Images are taken at 2a and 14-2b. The image data of the captured non-defective sample is stored in the image data storage unit 27 via the image data acquisition unit 23, and is displayed on the display 11.

Next, in the same manner as in the case of the non-defective sample, the defective sample is flowed to the chute 3, and the defective sample is used as the CCD cameras 13a and 13b and the first wavelength NIR cameras 14-1a and 14-1b and the second wavelength NIR camera 14-. Images are taken at 2a and 14-2b. The image data of the photographed defective sample is stored in the image data storage unit 27 via the image data acquisition unit 23 and displayed on the display 11. Up to this point, it is not the actual sorting work, but the preparatory work for determining the threshold value described later. In the non-defective product pattern / defective product pattern learning step, since the non-defective product sample and the defective product sample are selected in advance, the ejector nozzle 7 is not operated.

Next, in step 102, the operator visually confirms the image of each sample displayed on the display 11 again, and specifies what should be a non-defective product and what should be a defective product (including foreign matter) by an input operation. .. In addition, FIG. 4 shows images of a good sample and a defective sample of granular material, and in order from the top, visible light images captured by the CCD cameras 13a and 13b, the first wavelength NIR camera 14-1a, and the like. An object in which the near-infrared light images taken by 14-1b and the near-infrared light images taken by the first wavelength NIR cameras 14-1a and 14-1b are embedded in the contours of the granules obtained by the CCD cameras 13a and 13b. Near-infrared light image after recognition, near after object recognition in which the near-infrared light image by the second wavelength NIR cameras 14-2a and 14-2b is fitted to the contour of the granular material obtained by the CCD cameras 13a and 13b. The infrared light images are shown in order from the top. That is, the visible light images taken by the CCD cameras 13a and 13b have a clear outline of the granules, but the near-infrared light images taken by each NIR camera have an unclear outline of the granules. Therefore, by fitting the near-infrared light image obtained by each NIR camera into the contour of the granular material obtained by the CCD cameras 13a and 13b, the near-infrared light image after object recognition is displayed. There is.

Here, when the near-infrared light images captured by the NIR cameras are fitted into the contours of the granules obtained by the CCD cameras 13a and 13b, the visible light images of the granules obtained by the CCD cameras 13a and 13b are used. If there is a deviation in the near-infrared light image of the granular material obtained by each NIR camera, the displaced portion may be mistakenly recognized as a defective portion, which may lead to poor discrimination. Therefore, object recognition is performed from the contour of the granular material obtained from the visible light image, and the CCD cameras 13a, 13b and each NIR camera are used so that the near-infrared light image is not displaced when the near-infrared light image is superimposed on the contour of the granular material. It is preferable to adjust the orientation and position.

Further, in the present embodiment, as shown in FIG. 4, the first wavelength NIR cameras 14-1a and 14-1b provide a near-infrared optical image having a wavelength component of 850 nm and the second wavelength NIR cameras 14-2a and 14-. A near-infrared light image having a wavelength component of 1550 nm is imaged by 2b. That is, when it is difficult to determine the quality from the CCD image and the 850 nm near-infrared light image as in the defective products No. 1 and No. 2 in FIG. 4, the 1550 nm near-infrared light image is not a good product. Since it differs from a non-defective product, it is possible to accurately discriminate the quality of the discriminating object based on this.

Next, the process proceeds to step 103, and one wavelength component of the red (R), green (G), and blue (B) light wavelengths is added to the image data of a large number of non-defective and defective samples. Two near-infrared light components (hereinafter, sometimes referred to as "NIR") are plotted in a three-dimensional space to create a three-dimensional optical correlation diagram as shown in FIG. As a result, as shown in FIG. 16, the problem that a non-defective product and a defective product cannot be effectively distinguished by the wavelength components of red (R), green (G), and blue (B) light is effectively solved. It becomes possible to solve it. In this embodiment, as shown in FIG. 8, a good sample and a defective sample are plotted in the three-dimensional space of each axis of G, the first wavelength NIR, and the second wavelength NIR, and the first The wavelength component of the wavelength NIR is 850 nm, and the wavelength component of the second wavelength NIR is 1550 nm. In addition, which of the wavelength components of red (R), green (G), and blue (B) light is to be selected is determined by confirming the plotting mode in the three-dimensional space on the display 11. The operator selects a wavelength component having a significant difference (highly correlated) by the input means. Of course, the determination unit 19 may automatically select the item without the operator selecting the item.

(Threshold calculation process)
Proceeding to step 104, the non-defective product aggregate formed by the dots related to the non-defective product (black dots in FIG. 8) and the defective product aggregate formed by the dots related to the defective product (gray dots in FIG. 8) are roughly classified. In step 105, a statistic of multivariate data for each aggregate of non-defective product aggregate / defective product aggregate is calculated.

The calculation of this statistic may be performed by the calculation of the center of gravity vehicle or the variance-covariance matrix. For example, the calculation formula of the center of gravity vector is represented by the formula (1) in FIG. The arithmetic expression of the variance-covariance matrix is represented by the equation (2) of FIG.

Next, the Mahalanobis square distance from each center of gravity vector of the non-defective product aggregate and the defective product aggregate is obtained. Here, the Mahalanobis square distance is a function of the value of the multivariate data, and the calculation formula of the Mahalanobis square distance is expressed by the formula (3) of FIG.

Next, in step 106, the boundary surface between each aggregate is obtained. When determining this interface, the values of the multivariate data are classified into the aggregates that minimize the Mahalanobis square distance, and the aggregates that belong to all the values of the multivariate data in the multivariate space are determined. Then, the boundary surface represented by the reference numeral m in FIG. 8 is determined.

Next, in step 107, the Euclidean distance at which the distance between the centers of gravity of the non-defective product aggregate and the defective product aggregate is the longest is selected, and the boundary surface having a wide threshold effective range is searched. At this time, if the center of gravity vector of the non-defective product aggregate is P (Xp1, Xp2, Xp3, ... Xpn) and the center of gravity vector of the defective product aggregate is Q (Xq1, Xq2, Xq3, ... Xqn), then The Euclidean square distance is expressed by the equation (4) in FIG.

Next, in step 107, the boundary surface between the aggregates is obtained. When determining this boundary surface, the values of the multivariate data are classified into the aggregate having the maximum Euclidean square distance, and the boundary surface indicated by the symbol u in FIG. 8 is determined.

The equation of the plane m of the boundary surface that minimizes the Mahalanobis distance is expressed by the equation (5) of FIG. 7, and the equation of the plane u of the boundary surface that maximizes the Euclidean distance is the equation (6) of FIG. If it is represented by, two characteristic planes m and u as shown in FIG. 8 can be obtained. Then, in step 108, the three-dimensional optical correlation diagram of FIG. 8 is rotated so that the line-of-sight direction (line-of-sight vector) coincides with a position where these two different planes m and u intersect and appear as a line segment. As a result, the optimum threshold value obtained by reducing the dimension from the three-dimensional space shown in FIG. 8 to the two-dimensional plane as shown in FIG. 9 is obtained. Therefore, it is possible to provide an optical granular material discriminating device 1 that greatly simplifies signal processing and is easy for an operator to handle.

The line segment L (see FIG. 8 and the like) at which the plane m of the equation (5) of FIG. 7 and the plane u of the equation (6) of FIG. 7 intersect can be obtained by the equation (7) of FIG. .. Then, when the direction vector e of the line of intersection is obtained by the outer product calculation of the normal vectors of the two planes m and u, the equation (8) in FIG. 7 is obtained. Then, the point P through which the line of intersection L passes is given by the equation (9) in FIG. Once the line of intersection L is obtained as described above, it can be converted into a G / first wavelength NIR / second wavelength NIR correlation diagram (see FIG. 9) in an optimum two-dimensional plane in which the viewpoint is placed on the line of intersection L. It becomes.

(Threshold calculation process)
Next, in step 109, the discrimination threshold between the non-defective product and the defective product is automatically calculated based on the line of intersection L on the two-dimensional plane of FIG. Therefore, the details of the threshold value calculation process will be described below based on the flow chart of FIG.

The flow chart of FIG. 6 is a flow chart showing in detail the threshold value calculation step of step 109 of FIG. First, in step 201, an inertial equivalent ellipse is applied to the defective assembly indicated by the gray dots in FIG. 9 (see FIG. 10). Here, the inertial equivalent ellipse is a feature quantity representing an ellipse equivalent to a quadratic moment around the center of gravity, which is substantially equal to the defective product aggregate, and it is possible to grasp the characteristics of how the defective product aggregate spreads. In practice, the length of the major axis is a multiple of the standard deviation (a positive integer multiple) and the length of the minor axis is a multiple of the standard deviation (in order to make the defective region sufficiently larger than the distribution of the defective aggregate). Create an inertial equivalent ellipse as a positive integer multiple). This is an empirical value and changes depending on the type of granules, so it is preferable to be able to change it freely.

Then, in step 202, the center of gravity G of the inertial equivalent ellipse and the inclination angle Θ in the major axis V direction are obtained, and then in step 203, the distance of the major axis V and the distance of the minor axis W are calculated (see FIG. 10). ).

In step 204, in the inertial equivalent ellipse, two straight lines parallel to the short axis and passing through both end points of the long axis and two straight lines parallel to the long axis and passing through both end points of the short axis are drawn. That is, the circumscribing rectangle of the inertial equivalent ellipse is created by the four straight lines (see FIG. 11). This circumscribing rectangle serves as a tentative reference when creating automatic sensitivity.

Next, in step 205, the center of gravity on the non-defective product aggregate side is calculated. This is obtained by calculating a simple average of all non-defective product data (see FIG. 11).

Then, the following processing is performed in order to obtain the relationship between the non-defective product aggregate and the defective product aggregate. That is, in step 206, the center of gravity on the non-defective product aggregate side obtained in step 205 and the both end points of the circumscribing rectangle on the defective product aggregate side obtained in step 204 are connected in the long axis direction, and two straight lines (FIG. 12, reference numerals (FIG. 12, reference numerals) 2) and (3) straight line) are created.

From the above, six planes forming a closed region surrounding the inertial equivalent ellipse fitted to the defective aggregate are created. That is, as shown in FIG. 13, as six planes surrounding the inertial equivalent ellipse, the first plane is the boundary plane (1) that minimizes the Maharanobis distance, and the second plane is the center of gravity on the good product aggregate side and the defective product set. The plane connecting one end of the circumscribing rectangle on the body side in the long axis direction (2), the third plane is the plane connecting the center of gravity on the good product aggregate side and the other end of the circumscribing rectangle on the defective product assembly side in the long axis direction (3). , The fourth plane is the long side (4) on one side of the circumscribing rectangle, the fifth plane is the long side (5) on the other side of the circumscribing rectangle, and the sixth plane is the circumscribing rectangle on the side far from the good product aggregate. It is the short side (6) on one side.

The six planes (FIG. 13, reference numerals (1) to (6)) forming the closed region are obtained by drawing such as creating an circumscribing rectangle, but the drawing is not limited to this. In order to improve efficiency by replacing complicated arithmetic processing such as being created by the above with a simple array reference processing, it is possible to replace it with a look-up table (LUT) in advance and store it in a memory or the like.

As an exception to the drawing of the plane in paragraph 0044, the angle γ (see FIG. 14) formed by the straight line between the centers of gravity of each aggregate and the boundary surface (1), and the major axis and the boundary surface (1). When both angles ω (see FIG. 14) are larger than 45 °, the fourth plane and the fifth plane are on the short side, and the sixth plane is on the long side.

Next, in step 208 of FIG. 6, the sensitivity is adjusted. As the sensitivity level, the range is represented by a numerical level of 0 to 100. That is, when the sensitivity level is 0, the sensitivity is the minimum sensitivity (MIN), and it is not possible to distinguish between a non-defective product and a defective product, and the non-defective product is mixed with the defective product at the time of sorting, and the sensitivity is low. When the sensitivity level is 50, the sensitivity is medium (MID), and a good product and a defective product can be accurately discriminated from each other. When the sensitivity level is 100, the maximum sensitivity (MAX) is obtained, and a good product and a defective product can be discriminated with extremely high accuracy. However, the good product is also selected together with the defective product and the yield is poor.

Further, as shown in FIG. 15, the calculation of the threshold value created by the six planes surrounding the inertial equivalent ellipse described above, and the three sensitivity levels of the above-mentioned minimum sensitivity (MIN), medium sensitivity (MID), and maximum sensitivity (MAX). The correspondence with is as follows.

The minimum sensitivity (MIN) is the sixth plane (6), the medium sensitivity (MID) is the first plane (1), and the maximum sensitivity (MAX) is the second plane (2) and the third plane (3). It is the same as the seventh plane (7) formed by a straight line that divides the angle formed by 3) into two equal parts and a straight line that is perpendicular to the straight line. That is, the sensitivity level is set in each of the sixth plane (6), the first plane (1), and the seventh plane (7) in FIG. 15, for example, the sensitivity level is medium sensitivity (MID). ), The area above the first plane (1) on FIG. 15 is a non-defective product region, and the region below is a defective product region. It is determined by the good / bad determination unit 26.

Three thresholds created by the first plane, the sixth plane, and the seventh plane as described above, and three sensitivity levels consisting of the minimum sensitivity (MIN), the medium sensitivity (MID), and the maximum sensitivity (MAX). The correspondence of is automatically created by touching the sensitivity creation button (not shown) arranged on the display 11. Further, since the threshold value is calculated by reducing the dimension from the three-dimensional space of FIG. 8 to the two-dimensional plane of FIG. 9, signal processing can be greatly simplified.

The threshold value determined by the threshold value calculation unit 28 based on the flow as described above is stored in the threshold value data storage unit 24 of the signal processing unit 20. Subsequently, the actual discrimination operation is performed, and the granular material to be discriminated is flowed from the storage tank 4 to the chute 3 in a state where the ejector nozzle 7 can be driven. When the granules dropped from the chute 3 reach the inspection unit 6, the granules are transferred to the CCD cameras 13a, 13b, the first wavelength NIR cameras 14-1a, 14-1b, and the second wavelength NIR cameras 14-2a, 14-. 2b takes an image.

The good / bad discriminating unit 26 reads a threshold value from the threshold data storage unit 24, and based on this threshold value, the CCD cameras 13a, 13b, the first wavelength NIR cameras 14-1a, 14-1b, and the second wavelength NIR camera 14 From the image data captured by -2a and 14-2b, it is determined whether the granular material is a good product or a defective product by using the wavelength component of green (G) and the two near-infrared light components. Of course, in the case of a defective product having a clear color difference as compared with a non-defective product, good / defective can be discriminated only by the wavelength component of green (G).

Even if the good / bad discriminating unit 26 determines that the granular material is a good product, the ejector drive circuit 22 does not open the valve of the ejector nozzle 7, and the granular material goes to the non-defective product collecting unit 8. And fall naturally. When the good / bad discriminating unit 26 determines that the granular material is defective reaches the ejector nozzle 7, the ejector drive circuit 22 opens the valve of the ejector nozzle 7, and the granular material falls due to the blown air from the ejector nozzle 7. It is blown away from the defective product collection unit 9 and falls.

In the first embodiment, the inspection unit 6 that performs the optical inspection has a front box 12a (first inspection unit) on the front side and a rear box 12b (first inspection unit) on the rear side with the falling locus a of the granular material in between. 2 inspection units), each of which houses a CCD camera that is a visible light detection unit, and a first wavelength NIR camera and a second wavelength NIR camera that are near-infrared light detection units. Therefore, two visible light images, two first-wavelength near-infrared light images, and two second-wavelength near-infrared light images are acquired for one granular material. With this configuration, for example, if it is determined that either the front side or the rear side is defective, it can be collected by the ejector nozzle 7 to the defective product collecting unit 9, and the determination is highly accurate. Can be done.

[Other Embodiments]
The first embodiment of the examples of the optical granular material discriminating device of the present invention has been described above. However, the present invention is not necessarily limited to the above-described embodiment, and includes, for example, the following modifications.

For example, in the above-mentioned first embodiment, the granules to be discriminated are black sunflower seeds, but other granules may be used, and for example, rice can be discriminated. Specifically, as shown in FIG. 17, based on the threshold value determined based on the optical inspection, white rice and sardine are regarded as non-defective products from among the granular substances to be discriminated, and colored rice and lightly burnt. It is possible to discriminate rice and foreign matter as defective products. For example, as shown in the figure, "white rice", "shirata", and "foreign matter" cannot be distinguished from white rice by a visible light image, but "shirata" can be distinguished from white rice by a near-infrared light image of 850 nm. It is possible, and further, "foreign matter" can be distinguished from white rice by a near-infrared light image of 1550 nm. Then, a three-dimensional optical correlation diagram is created based on each of the captured near-infrared light images and visible light images, and the threshold value is obtained in the same manner as in the above-described embodiment to accurately discriminate between non-defective products and defective products. It becomes possible to do. In the other embodiment shown in the figure, "Shirata" is regarded as a non-defective product, but "Shirata" can be regarded as a defective product according to the setting of the operator, and in relation to the yield, it is regarded as a non-defective product. It is possible to freely select defective products.

In addition, the objects to be discriminated are not limited to the above-mentioned black sunflower seeds and rice, but also grains such as wheat, beans and nuts, resin pieces such as pellets and beads, and fine articles such as pharmaceuticals, ores and silas. The optical granular material discriminating device of the present invention can be effectively applied when sorting raw materials composed of other granular materials into non-defective products and defective products and eliminating foreign substances mixed in the raw materials. be.

In the first embodiment described above, the discrimination target is a black sunflower seed, and the visible light of green (G), the reflection component of near-infrared light having a wavelength of 850 nm, and the reflection component of near-infrared light having a wavelength of 1550 nm are detected. Therefore, the three-dimensional optical correlation diagram was created, and it is possible to select the wavelengths of visible light and near-infrared light in which the most significant difference appears, depending on the type of discrimination target. When selecting the wavelengths of visible light and near-infrared light, the operator may visually select the three-dimensional optical correlation diagram displayed on the display 11, and the determination unit 19 automatically selects the wavelength. It may be configured.

More specifically, NIR cameras that capture near-infrared light images of 1200 nm, which are other near-infrared light components, may be added (including those capable of capturing multiple wavelengths with one NIR camera). , It is possible to acquire near-infrared light images of a plurality of wavelengths according to the type of discrimination target. In this case, multivariate analysis is performed including the wavelength components of red (R), green (G), and blue (B) and the wavelength components of near-infrared light of 850 nm, 1200 nm, and 1550 nm, and among the wavelength components, Since it is possible to create a three-dimensional optical correlation diagram based on the wavelength component with the highest correlation, it is possible to set appropriate thresholds for various types of discrimination objects and perform highly accurate discrimination. ..

Further, in the first embodiment, a CCD camera capable of detecting three wavelength components of red (R), green (G), and blue (B) is used as a visible light detection unit, but only a specific wavelength is detected. It is also possible to use a possible visible light detection unit. Further, in the first embodiment, the first wavelength NIR cameras 14-1a and 14-1b and the second wavelength NIR cameras 14-2a and 14-2b are provided according to the wavelength of the near infrared light. A camera capable of capturing a plurality of types of wavelength components with one NIR camera may be used.

Further, in the first embodiment, the object is recognized from the contour of the granular material obtained from the visible light image, and the CCD camera 13a is prevented from shifting when the near-infrared light image is superimposed on the contour of the granular material. , 13b and the first wavelength NIR cameras 14-1a and 14-1b and the second wavelength NIR cameras 14-2a and 14-2b are configured to be adjusted, but the method is not necessarily limited to such a method. For example, when a deviation is observed in the near-infrared light image displayed on the display 11 after recognition of an object, the deviation can be corrected by manually or automatically correcting the position of the image.

Further, in the first embodiment, the CCD camera, which is a visible light detection unit, and the NIR camera, which is a near-infrared light detection unit, provided in the front box 12a and the rear box 12b, respectively, are used for one granular object. , Two visible light images and two near-infrared light images were acquired, and high-precision discrimination was performed. However, the configuration is not necessarily limited to this, and for example, only one of the front box 12a and the rear box 12b includes a CCD camera that is a visible light detection unit and a near-infrared light detection unit. A NIR camera may be provided. In yet another modification, one of the front box 12a and the rear box 12b, the CCD camera and the NIR camera, can be provided as a spare in case of failure.

1 Optical granularity discriminator 2 Machine 3 Shoot 4 Storage tank 5 Vibration feeder 6 Inspection unit 7 Ejector nozzle 8 Good product collection unit 9 Defective product collection unit 10 Door 11 Display 12a Front box 12b Rear box 13a, 13b CCD camera 14-1a , 14-1b 1st wavelength NIR camera 14-2a, 14-2b 2nd wavelength NIR camera 15a, 15b Visible light source 16a, 16b Near infrared light source 17a, 17b Background 18a, 18b Transmissive unit 19 Judgment unit 20 Signal processing Unit 21 CPU and memory unit 22 Ejector drive circuit 23 Image data acquisition unit 24 Threshold data storage unit 25 Image processing unit 26 Good / bad discrimination unit 27 Image data storage unit 28 Threshold calculation unit 29 Signal transmission / reception unit

Claims (10)

It is provided with an inspection unit that optically inspects the granules transferred by the transfer means, and a determination unit that determines whether the granules are non-defective or defective based on the optical inspection by the inspection unit. In the optical granularity discriminator
The inspection unit reflected visible light that irradiates the granules with visible light, a near-infrared light source that irradiates the granules with near-infrared light, and visible light transmitted through the granules or reflected from the granules. It has at least a visible light detection unit that detects visible light and a near-infrared light detection unit that detects near-infrared light transmitted through the granules or near-infrared light reflected from the granules.
The determination unit
One wavelength component of the red (R), green (G), and blue (B) light detected by the visible light detection unit and the near infrared light for a plurality of non-defective sample and a plurality of defective samples. An optical granularity discriminating device characterized in that a plurality of near-infrared light components detected by the light detection unit are plotted in a three-dimensional space to create a three-dimensional optical correlation diagram and a threshold value is set. The determination unit
The Mahalanobis distance boundary surface and the Euclidean distance boundary surface that separate the non-defective product aggregate and the defective product aggregate are created on the three-dimensional optical correlation diagram.
Create a two-dimensional plane perpendicular to the line of intersection between the Mahalanobis distance boundary surface and the Euclidean distance boundary surface.
The optical granular material discriminating device according to claim 1, wherein a closed region is created by applying an inertial equivalent ellipse to the defective aggregate on the two-dimensional plane, and a threshold value is set in the closed region. The determination unit
One wavelength component of the red (R), green (G), and blue (B) light detected by the visible light detection unit and the near infrared light for a plurality of non-defective sample and a plurality of defective samples. The optical granularity discrimination according to claim 1 or 2, wherein two near-infrared light components obtained by a combination of a plurality of types detected by the light detection unit are plotted in a three-dimensional space to create a plurality of types of three-dimensional optical correlation diagrams. Device. The optical granules according to claims 1 to 3, wherein the near-infrared light component detected by the near-infrared light detection unit is a near-infrared light component within the contour of the granules detected by the visible light detection unit. Object discrimination device. The determination unit
When setting the threshold value in the closed region, an circumscribing rectangle consisting of two straight lines parallel to the minor axis of the inertial equivalent ellipse and passing through both end points of the major axis and two straight lines parallel to the major axis and passing through both end points of the minor axis. The optical granular material discriminating device according to any one of claims 1 to 4, wherein the two straight lines connecting the center of gravity of the non-defective product aggregate and both end points of the circumscribing ellipse in the major axis direction are created. The determination unit
When setting the threshold value in the closed region, the Maharanobis distance boundary surface that minimizes the Maharanobis distance is used as the first plane, and the center of gravity of the good product aggregate and the long axis direction of the circumscribing rectangle are used as the second plane. As a third plane, a plane connecting the center of gravity of the non-defective assembly and the other end of the circumscribing rectangle in the long axis direction is used as a third plane, and one side of the circumscribing rectangle is used as a fourth plane. The long side of the extrinsic rectangle is used as the fifth plane, the long side of the other side of the extrinsic rectangle is used as the fifth plane, and the short side of one side of the extrinsic rectangle far from the non-defective aggregate is used as the sixth plane. The optical granular material discriminating device according to any one of 1 to 5. A display and an input means capable of input by an operator based on the display contents on the display are provided.
The display has a plurality of the three-dimensional optical correlation diagrams of red (R), green (G), and blue (B) light having an arbitrary wavelength component and two arbitrary near-infrared light components. Can be displayed,
The optical granular material discriminating device according to any one of claims 1 to 6, wherein the input means can select the three-dimensional optical correlation diagram displayed on the display based on the operation of the operator. The inspection unit includes a first inspection unit located on the front side of the granular material to be transferred and a second inspection unit located on the rear side of the granular material.
The optical granular material discrimination according to any one of claims 1 to 7, wherein the first inspection unit and the second inspection unit include the visible light detection unit and the near infrared light detection unit, respectively. Device. The optical granularity discriminating device according to any one of claims 1 to 8, wherein the granular material transferred by the transfer means is a seed or a grain. It is provided with an inspection unit that optically inspects the granules transferred by the transfer means, and a determination unit that determines whether the granules are non-defective or defective based on the optical inspection by the inspection unit. In the optical granularity discriminator
The inspection unit reflected visible light that irradiates the granules with visible light, a near-infrared light source that irradiates the granules with near-infrared light, and visible light transmitted through the granules or reflected from the granules. It has at least a visible light detection unit that detects visible light and a near-infrared light detection unit that detects near-infrared light transmitted through the granules or near-infrared light reflected from the granules.
The determination unit
The three wavelength components of red (R), green (G), and blue (B) detected by the visible light detector and the near-infrared light detector detect a plurality of non-defective samples and a plurality of defective samples. Multivariate analysis using multiple near-infrared light components as parameters
Based on the result of the multivariate analysis, the wavelength component detected by the visible light detection unit and the near infrared light component detected by the near infrared light detection unit are plotted to create a three-dimensional optical correlation diagram. An optical granularity discriminating device characterized by setting a threshold value.

Images