JP2015040818A - Method and apparatus for grain classification - Google Patents

Abstract

PROBLEM TO BE SOLVED: To further efficiently classify grains using near infrared light.SOLUTION: A grain classification method and a grain classification apparatus 100 include: receiving diffuse reflection light from an examination object 3 consisting of grains in each of pixels arranged two-dimensionally and acquiring a spectrum; then determining a group to which the examination object 3 belongs for each individual on the basis of the spectrum. This configuration allows imaging of a plurality of individuals at one time and classification thereof for each individual on the basis of the imaging result. Consequently, efficient classification of the grains using near infrared light becomes possible.

Description

  The present invention relates to a grain classification method and a grain classification apparatus.

  Conventionally, it has been studied to obtain a near-infrared light spectrum from cereals by irradiating near-infrared light to a crop to be inspected, and to evaluate the quality of the grain based on the spectrum. For example, Patent Document 1 discloses a method for nondestructively obtaining quality information of fruits and vegetables based on a spectral spectrum obtained by irradiating an object with near infrared light as a measurement light beam.

Japanese Patent Laid-Open No. 9-297053

  However, if the apparatus described in Patent Document 1 is used for the total inspection of inspection objects, the work load increases. In particular, when each test object is small, such as a grain, for example, setting the test objects one by one in the apparatus and measuring them greatly increases the amount of work.

  The present invention has been made in view of the above, and an object of the present invention is to provide a grain classification method and a grain classification apparatus that can more efficiently classify grains using near-infrared light.

  In order to achieve the above object, a grain classification method according to the present invention is a grain classification method for classifying a grain, which is an inspection object, to belong to which group among a plurality of groups. On the other hand, by irradiating near-infrared light, the transmitted light or diffuse reflected light from the inspection object is received for each of the two-dimensionally arranged pixels, and an imaging step for acquiring a spectrum, based on the spectrum, And a determination step of determining a group belonging to each individual for the inspection object.

  The grain classification apparatus according to the present invention is a grain classification apparatus that classifies a grain as an inspection object to belong to which group among a plurality of groups, and the near-infrared ray is applied to the inspection object. A light source for irradiating light and transmitted light or diffuse reflected light from the inspection object emitted by irradiation of near-infrared light from the light source are received for each two-dimensionally arranged pixel to obtain a spectrum. An imaging unit, and a determination unit that determines a group belonging to each individual of the inspection target based on the spectrum obtained in the imaging unit.

  In the grain classification method and grain classification apparatus described above, after the diffusely reflected light or transmitted light from a test object made of grain is received at a two-dimensionally arranged pixel and a spectrum is acquired, the test object is acquired based on the spectrum. A group belonging to each individual is determined. By having such a configuration, for example, imaging of a plurality of individuals can be performed at once, and classification can be performed for each individual based on the imaging result. Therefore, it is possible to more efficiently classify grains using near infrared light.

  Here, it is preferable that the near-infrared light includes at least part of light in the wavelength range of 1800 to 2200 nm. Grain classification can be more suitably performed by including light in the above-described wavelength range in the near-infrared light that irradiates the inspection object.

  In addition, as a configuration that effectively exhibits the above-described operation, for example, in the determination step, the spectrum obtained for each pixel obtained in the imaging step is acquired in a pixel obtained by imaging a specific individual among the inspection object. In addition, there is a mode in which an average spectrum relating to the specific individual is obtained by averaging the spectrum, and a group to which the specific individual belongs is determined based on the shape of the average spectrum.

  In addition, as another configuration that effectively exhibits the above-described operation, for example, in the determination step, each pixel obtained by imaging a specific individual among the inspection object in the spectrum for each pixel obtained in the imaging step. The group to which the spectrum acquired in (1) belongs is determined, and the group to which the specific individual belongs is determined based on the determination result of the group to which the spectrum belongs.

  ADVANTAGE OF THE INVENTION According to this invention, the grain classification method and grain classification apparatus which can perform the classification of the grain using near-infrared light more efficiently are provided.

It is a figure explaining the composition of the grain classification device concerning this embodiment. It is a figure explaining a hyperspectral image. It is a figure explaining the classification | category method of a grain. It is a figure which shows the spectrum data (averaged data) for every individual. It is a figure which shows the result of having performed SNV conversion with respect to each spectrum data shown in FIG. It is a figure which shows the result of having performed the secondary differentiation about each spectrum data shown in FIG. It is a figure which expands and shows the wavelength range of 1800 nm-2200 nm among the spectrum data shown in FIG. It is a figure which shows the result of having performed PCA using the spectrum data after the SNV conversion shown in FIG. It is a figure which shows the result of having performed PCA using the spectrum data after the SNV conversion shown in FIG. It is a figure which shows the result of having performed PCA using the spectrum data after the secondary differentiation shown in FIG.

DESCRIPTION OF EMBODIMENTS Hereinafter, embodiments for carrying out the present invention will be described in detail with reference to the accompanying drawings. In the description of the drawings, the same elements are denoted by the same reference numerals, and redundant description is omitted.
(Structure of grain classification device)

  A grain sorting apparatus 100 according to the present invention will be described with reference to FIG. The grain classification apparatus 100 according to the present embodiment is an apparatus whose main purpose is to classify grains as the inspection object 3 distributedly placed on the belt conveyor 2 into a plurality of groups.

  Examples of the grains that are the inspection object 3 of the grain sorting apparatus 100 include wheat, rice, and corn. The classification by the grain classification device 100 includes, for example, classifying a mixture of a plurality of types of grains for each type, or classifying a specific type of grains with some index. In particular, the grain classification apparatus 100 according to the present embodiment enables classification with an index that can be classified into different groups even if the appearance is similar. Examples of such indicators include varieties, brands, and production areas. Further, for example, classification based on an index affected by the amount of various components, such as deliciousness and safety, is possible. It is also possible to classify defective products from non-defective products.

  The grain sorting apparatus 100 is preferably used for in-line inspection of all inspection objects to be conveyed. In the following embodiment, a case where the inspection object 3 is rice and the variety (brand) is classified using the grain classification apparatus 100 will be described. Therefore, it is preferable to arrange | position the test object 3 on the belt conveyor 2 so that individuals may not overlap, and it is still more preferable to arrange | position individuals separately.

  The grain classification apparatus 100 measures the spectrum of diffuse reflection light obtained by irradiating the inspection object 3 with measurement light that is near infrared light, and inspects the inspection object 3 based on the spectrum. For this reason, the grain classification device 100 includes a light source 10, an imaging unit 20, and an analysis unit 30 (determination unit).

  The light source 10 irradiates measurement light which is near infrared light toward a predetermined irradiation area A1 on the belt conveyor 2. The wavelength range of the measurement light emitted by the light source 10 is appropriately selected depending on the inspection object 3. Specifically, as the measurement light, near infrared light having a wavelength range of 800 nm to 2500 nm is preferably used, and light having a wavelength range of 1000 nm to 2300 nm is particularly preferably used. Further, when the inspection object 3 is a grain, since the fluctuation of the peak indicating the characteristics of the individual is confirmed in the wavelength band of 1800 to 2200 nm, the light in the above wavelength band may be included in the measurement light. preferable. For example, a halogen lamp is preferably used as the light source 10.

  The irradiation region A1 is a partial region of the surface (mounting surface 2b) of the belt conveyor 2 on which the inspection object 3 is mounted. The irradiation area A1 extends in the width direction (x-axis direction) perpendicular to the traveling direction 2a (y-axis direction in FIG. 1) of the placement surface 2b, and covers from one end to the other end of the placement surface 2b. It is an area.

  In the present embodiment, the configuration in which the light emitted from the light source 10 is directly applied to the irradiation area A1 will be described. However, the light from the light source 10 is shaped on a line via, for example, a cylindrical lens. It may be configured to emit light.

  The near infrared light L1 output from the light source 10 is diffusely reflected by the inspection object 3 placed on the irradiation area A1. A part of the light enters the imaging unit 20 as diffusely reflected light L2.

The imaging unit 20 has a function as a hyperspectral sensor that acquires a hyperspectral image. Here, the hyperspectral image in this embodiment is demonstrated using FIG. FIG. 2 is a diagram for explaining the outline of the hyperspectral image. As shown in FIG. 2, the hyperspectral image is an image configured by _N pixels P _{1 to} P _N. Specifically it shows the two pixel P _n and P _m as them example in FIG. Each of the pixels P _n and P _m includes spectral information S _n and S _m including a plurality of intensity data. And the intensity data is data indicating a spectral intensity at a particular wavelength (or wavelength band), 2, 15 intensity data has been retained as Supetoru information S _n and S _m, superposition of these It is shown in the state. As described above, the hyperspectral image H is characterized by having a plurality of intensity data for each pixel constituting the image, so that a three-dimensional image having both a two-dimensional element as an image and an element as spectral data. Configuration data. In the present embodiment, the hyperspectral image H refers to an image composed of pixels having intensity data in at least five wavelength bands per pixel.

In FIG. 2, the inspection object 3 is also shown. That, P _n in FIG. 2 is a pixel of the captured inspection object 3, P _m is the pixel of the captured background (e.g., belt conveyor and the conveying container). Thus, the imaging unit 20 acquires not only the inspection object 3 but also an image obtained by imaging the background. Therefore, in the process related to the classification of the inspection object 3, the pixel that captured the background is specified, and the spectrum data acquired by the pixel is distinguished from the spectrum data acquired by the pixel that captured the inspection object 3. It is necessary to handle.

  Returning to FIG. 1, the imaging unit 20 according to the present embodiment includes a camera lens 24, a slit 21, a spectroscope 22, and a light receiving unit 23. The imaging unit 20 has a visual field region 20s (imaging region) extending in a direction (x-axis direction) perpendicular to the traveling direction 2a of the belt conveyor 2. The visual field region 20s of the imaging unit 20 is a linear region included in the irradiation region A1 of the placement surface 2b, and is a region in which the diffuse reflected light L2 that has passed through the slit 21 forms an image on the light receiving unit 23.

  The slit 21 is provided with an opening in a direction parallel to the extending direction (x-axis direction) of the irradiation region A1. The diffuse reflected light L <b> 2 that has entered the slit 21 of the imaging unit 20 enters the spectroscope 22.

  The spectroscope 22 splits the diffusely reflected light L2 in the longitudinal direction of the slit 21, that is, the direction perpendicular to the extending direction of the irradiation area A1 (y-axis direction). The light split by the spectroscope 22 is received by the light receiving unit 23.

  The light receiving unit 23 includes a light receiving surface in which a plurality of light receiving elements are two-dimensionally arranged, and each light receiving element receives light. As a result, the light receiving unit 23 receives light of each wavelength of the diffusely reflected light L2 reflected at each position along the width direction (x-axis direction) on the belt conveyor 2. Each light receiving element outputs a signal corresponding to the intensity of received light as information on a two-dimensional planar point composed of a position and a wavelength. A signal output from the light receiving element of the light receiving unit 23 is transmitted from the imaging unit 20 to the analyzing unit 30 as spectral data for each pixel related to the hyperspectral image.

  The analysis unit 30 obtains the reflected light spectrum of the diffuse reflected light L2 from the input signal, and performs an inspection based on the obtained spectrum. The result of analysis by the analysis unit 30 is notified to the operator of the grain classification device 100 by outputting the result to, for example, a monitor connected to the analysis unit 30 or a printer.

The analysis unit 30 includes a CPU (Central Processing Unit), a RAM (Random Access Memory) and a ROM (Read Only Memory) that are main storage devices, a communication module that performs communication with other devices such as a detection unit, The computer is configured with hardware such as an auxiliary storage device such as a hard disk. And the function as the analysis part 30 is exhibited because these components operate | move.
(Cereal classification method)

  Next, a classification method using the grain classification apparatus 100 will be described with reference to FIG. As shown in FIG. 3, first, the near-infrared light L <b> 1 is output from the light source 10 to the irradiation area A <b> 1 including the inspection object 3 placed and moved on the belt conveyor 2, thereby inspecting the object 3. By receiving the diffuse reflection light L2 including the light diffusely reflected by the imaging unit 20, the visual field region 20s (imaging region) is imaged (S01: imaging step). Thereby, the spectrum data for each pixel obtained in the imaging unit 20 is sent from the imaging unit 20 to the analysis unit 30 and acquired in the analysis unit 30 (S02). The imaging unit 20 continuously captures images corresponding to the movement of the belt conveyor 2 and accumulates them in the analysis unit 30, so that in the analysis unit 30, each pixel is arranged in a two-dimensional hyperspectrum. Obtained as an image.

  Next, the analysis unit 30 identifies a pixel that images the inspection object 3 among these pixels, that is, a pixel to be analyzed (measurement target pixel) (S03). First, as a method for determining whether or not a pixel is a measurement target pixel, for example, the following method can be cited. Specifically, there is a method of determining whether or not data of a specific wavelength is within a predetermined threshold in the spectral data of each pixel. This is effective when the inspection object 3 includes a specific component when the background and the inspection object 3 are compared. Further, by performing four arithmetic operations on the reflected light intensity data at the first wavelength and the reflected light intensity data at the second wavelength different from that, a score relating to the pixel is calculated, and the calculated score is calculated in advance. There is a method of determining whether or not the pixel is a measurement target pixel depending on whether or not it is within the determined threshold value. By these methods, it is determined whether or not the pixel is an image of the inspection object 3.

  In addition, it is preferable to determine beforehand the index used for determination whether it is a measurement object pixel, such as the threshold value used for extracting a target pixel, so that a target pixel can be suitably determined. Moreover, the calculation and determination criteria used for extracting the target pixel may be used in combination with some of the above-described plural types of determination methods. Even when the same determination method is used, a plurality of calculations may be performed, or a plurality of threshold values may be used for the determination. In addition, after performing processing such as normalization, smoothing, and differentiation on the spectrum data for each pixel, an operation for determining whether the pixel is a target pixel may be performed. Moreover, although the spectrum data detected in the imaging part 20 are spectrum data of diffuse reflection light, after calculating an absorbance spectrum based on this, subsequent evaluation may be performed using the absorbance spectrum.

  Spectral data related to the pixel specified as the measurement target is used for classification for each individual by one of the following two methods.

  First, as a first method, among the spectrum data for each pixel obtained in the imaging step, a specific individual is obtained by averaging the spectrum data acquired in the pixel in which the specific individual is imaged among the inspection objects. There is a method of determining a group to which a specific individual belongs based on the shape of the averaged spectrum after obtaining the averaged spectrum. As a second method, among the spectrum data for each pixel obtained in the imaging step, a group to which the spectrum acquired in each pixel obtained by imaging a specific individual among inspection objects belongs is determined, and the spectrum There is a method of determining a group to which a specific individual belongs based on the determination result of the group to which the data belongs.

  In the first method, first, a measurement target pixel that captures the same individual is identified from pixels that are identified as having captured the inspection object 3, and averaged data is created (S04). As a method for specifying the measurement target pixels that image the same individual, for example, when the inspection objects 3 are measured apart from each other, a background is provided between pixels that are imaged from different individuals. Since it is considered that captured pixels are included, pixels that are determined as measurement target pixels are determined as pixels obtained by capturing the same individual for pixels determined to be measurement target pixels that are adjacent to each other. A method is mentioned. In addition, when the size of the inspection object 3 and the arrangement of the inspection object 3 on the belt conveyor 2 are known in advance, the measurement target pixel obtained by imaging the same individual is combined with the information. Also good. Thereafter, averaged data is created for the spectral data of the pixels specified as the measurement target pixels obtained by imaging the same individual. Thereby, one averaged data is created for one individual.

  Next, the classification of individuals is determined using this averaged data (S05). As a method of determining the classification, first, in order to distinguish a plurality of different types of grains, spectrum data is obtained for a plurality of grains whose varieties are known, and based on this, the grains are classified based on the varieties. A method of classification is obtained in advance. Based on this information, it is determined what kind of individual the individual is using, using the spectrum data relating to the individual grains of unknown varieties. Examples of methods for classifying the spectrum data related to grains into a plurality of types include multivariate regression analysis such as Principal Component Analysis (PCA), PLS (Partial Least Squares) regression analysis, and support vector machine. : SVM). Then, by outputting the analysis result (S06), the classification for each individual is completed.

  Next, in the second method, compared to the first method, it is determined to which group the inspection object imaged by the pixel belongs based on the spectral data for each pixel (S14), and thereafter The determination result related to the measurement target pixel that images the same individual among the pixels that are identified as having imaged the inspection object 3 is identified (result obtained in S14) This is a method for determining to which group the individual belongs (S15). First, as a method for classifying the spectrum data obtained at each pixel, the same method as the classification of the averaged data in the first method is used. Thereafter, after the classification is performed for each pixel, the measurement target pixel in which the same individual is imaged is specified as in the first method. Here, in the case where the same individual is captured by a plurality of pixels, it should have a determination result for each of the plurality of pixels, so based on the plurality of determination results, The determination result is calculated. Then, by outputting the analysis result (S06), the classification for each individual is completed.

  Here, as described above, the fact that the grains can be classified into a plurality of groups by the principal component analysis will be described using examples with reference to FIGS.

  First, a diffuse reflection spectrum was obtained using the above-mentioned grain classification apparatus 100 for 10 rice seeds (Hitomebore, Koshihikari, Sasanishiki, Hanaechizen, Kaguramochi) for each 10 grains. The wavelength range from which the spectrum data was acquired was 1000 nm to 2500 nm. For the obtained spectrum data, first, after converting the diffuse reflection spectrum into an absorbance spectrum, the measurement target pixel was specified based on the absorbance spectrum of each pixel. And the spectrum data of the measurement object pixel which concerns on the same individual among the specified measurement object pixels were averaged. FIG. 4 shows the spectral data corresponding to the averaged data for the spectral data of each pixel as a result. Furthermore, FIG. 5 shows a result of performing standard normal variant (SNV) conversion on each of the spectral data shown in FIG. According to the results shown in FIG. 5, it was found that, for example, in the region of wavelength 1600 nm or less or wavelength 1900 nm or more, only Kaguramochi shows a different spectrum shape compared to other brands. Of the five brands, Kaguramochi is glutinous rice, while the other brands are sticky rice. Therefore, it was found that it is possible to distinguish between glutinous rice and glutinous rice by performing SNV conversion on the spectral data obtained in each pixel.

  Next, FIG. 6 shows a result of second-order differentiation of the spectrum data shown in FIG. FIG. 7 shows an enlarged spectrum obtained by extracting only the wavelength range of 1800 nm to 2200 nm from the secondary differential spectrum shown in FIG. In the second-order differential spectrum shown in FIGS. 6 and 7, it was confirmed that the peak was shifted in Kaguramochi alone compared with other brands in the vicinity of the wavelength of 1900 nm. Furthermore, it was confirmed that a peak shift occurred between the “Kagamomochi, Koshihikari and Hanaechizen” group and the “Hitomebore, Sasanishiki” group in the vicinity of a wavelength of 2000 nm.

  Next, PCA was performed using the spectrum data after SNV conversion shown in FIG. As a result, as shown in FIG. 8, it was confirmed that it was possible to discriminate between glutinous rice (Hitomebore, Koshihikari, Sasanishiki, Hanaechizen) and glutinous rice (Kagamochi).

  Similarly, when analysis was performed using different components using PCA using the spectrum data after SNV conversion shown in FIG. 5, as shown in FIG. 9, “Hanaetizen”, “Koshihikari, Kaguramochi”, It was confirmed that discrimination could be made into 3 groups of “Himomebore, Sasanishiki”.

  Furthermore, when PCA was performed using the spectrum data obtained by performing the second order differentiation shown in FIG. 6, all five brands could be classified as shown in FIG.

  From the above results, it was confirmed by PCA that all 5 brands could be identified. As a result, it is confirmed that the determination of the brand, that is, the classification, is possible based on the spectrum obtained even for the inspection object whose brand is unknown, by performing the PCA using the conditions for obtaining the results of FIG. It was done.

  In the case of classifying grains according to other indicators, (1) Deriving analysis conditions for classification (groups belonging to a known group and using a plurality of grains belonging to different groups, (2) to classify grains by including the stage of measuring and analyzing the test object whose group is unknown based on the derived analysis conditions. it can.

  As described above, in the grain classification method and the grain classification apparatus 100 according to the present embodiment, after each of the two-dimensionally arranged pixels receives diffuse reflected light from the inspection object 3 made of grain and acquires a spectrum. Based on the spectrum, the group to which the inspection object 3 belongs is determined for each individual. By having such a configuration, for example, imaging of a plurality of individuals can be performed at once, and classification can be performed for each individual based on the imaging result. Therefore, it is possible to more efficiently classify grains using near infrared light.

  Further, when the inspection object 3 is a grain, the near-infrared light emitted from the light source 10 includes at least part of light in the wavelength range of 1800 to 2200 nm, thereby further classifying the grain. Can be done well. This is because, as described in the above embodiment, since the spectrum shape of the grain tends to change in the above wavelength range, it is preferable to classify the inspection object using the change in the spectrum shape. is there.

  As a grain classification method, as a first method, in the determination step, the spectrum acquired in a pixel obtained by imaging a specific individual in the inspection object 3 among the spectra for each pixel obtained in the imaging step. Can be obtained by determining the group to which the specific individual belongs based on the shape of the averaged spectrum. Further, as a second method, in the determination step, based on each of the spectra acquired in each pixel obtained by imaging a specific individual among the inspection objects, among the spectra for each pixel obtained in the imaging step, The group to which the inspection target 3 belongs is determined, and the group to which the specific individual belongs is determined based on the determination result of the group to which the inspection target 3 belongs based on the spectrum acquired in each pixel obtained by imaging the specific individual. A method is mentioned. In any of the methods, since the group to which the individual belongs is determined using the spectrum of one or more pixels obtained by imaging a specific individual, it is possible to classify the grains with high accuracy. Further, in the case of the first method, since an averaged spectrum is created in advance and classification is performed on one spectrum (averaged spectrum) for each individual, the classification system is compared with the second method. Since the amount of data handled at the time of analysis is reduced while maintaining, the processing can be performed more efficiently.

  As mentioned above, although embodiment of this invention has been described, this invention is not limited to embodiment mentioned above, A various change is possible in the range which does not deviate from the summary.

  For example, in the above embodiment, the case where a spectrum acquired in the wavelength range of near-infrared light is used is described, but a spectrum acquired in another wavelength range can also be used.

  Moreover, in the said embodiment, although it has set it as the structure which images the diffuse reflection spectrum of the test target object 3 in the imaging part 20, it arrange | positions the light source 10 in the position facing the imaging part 20 via the belt conveyor 2, for example. The transmission light spectrum of the inspection object 3 may be acquired by the imaging unit 20.

  DESCRIPTION OF SYMBOLS 100 ... Measurement system, 3 ... Test object, 10 ... Light source part, 20 ... Imaging part, 21 ... Slit, 22 ... Spectroscope, 23 ... Light-receiving part, 30 ... Analysis part.

Claims (5)

A grain classification method for classifying a grain to be inspected into a group among a plurality of groups,
By irradiating the inspection object with near-infrared light, the transmitted light or diffuse reflection light from the inspection object is received for each of the two-dimensionally arranged pixels, and an imaging step for acquiring a spectrum;
A determination step of determining a group belonging to each individual for the inspection object based on the spectrum;
Grain classification method including.   The grain classification method according to claim 1, wherein the near-infrared light includes at least part of light in a wavelength range of 1800 to 2200 nm. In the determination step,
Of the spectrum for each pixel obtained in the imaging step, an averaged spectrum relating to the specific individual is obtained by averaging the spectrum acquired in a pixel obtained by imaging the specific individual among the inspection objects. Seeking
The grain classification method according to claim 1 or 2, wherein a group to which the specific individual belongs is determined based on a shape of the averaged spectrum. In the determination step,
Among the spectra for each pixel obtained in the imaging step, determine a group to which the spectrum acquired in each pixel obtained by imaging a specific individual among the inspection objects belongs,
The grain classification method according to claim 1 or 2, wherein a group to which the specific individual belongs is determined based on a determination result of the group to which the spectrum belongs. A grain classification device for classifying a grain to be inspected into a group among a plurality of groups,
A light source for irradiating the inspection object with near-infrared light;
An imaging unit that receives the transmitted light or diffusely reflected light from the inspection object emitted by irradiation of near infrared light from the light source for each of the two-dimensionally arranged pixels, and acquires a spectrum;
Based on the spectrum obtained in the imaging unit, a determination unit that determines a group belonging to each individual for the inspection object;
Grain sorting equipment including.