JP2016105084A - Food inspection device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a food inspection device capable of accurately determining a foreign matter by reducing a noise due to the optical non-uniformity of an object to be inspected.SOLUTION: A food inspection device 100 includes: a light source 1 for irradiating an object to be inspected with inspection light; an image receiver 2 for receiving light transmitted through the object to be inspected to generate image data; a first polarizer 4 arranged on an optical path between the light source 1 and the object to be inspected for linearly polarizing the inspection light; a second polarizer 5 arranged on an optical path between the object to be inspected and the image receiver 2 for linearly polarizing transmission light; and polarization direction control means capable of changing a relative angle made by the transmission axis of the first polarizer 4 and the transmission axis of the second polarizer 5. The food inspection device 100 is configured to image a first image having a small relative angle and a second image having a large relative angle by using the image receiver 2, and to detect a foreign matter existing in the object to be inspected by using the first image and the second image.SELECTED DRAWING: Figure 1

Description

  The present invention relates to a food inspection apparatus for inspecting agricultural products, processed agricultural products, and other foods. In particular, the inspection object is irradiated with light, and transmitted light including internal information of the inspection object is received and imaged for inspection. The present invention relates to a food inspection apparatus that detects foreign matter present in a target.

  In food inspection equipment that irradiates inspection objects such as food with near-infrared light and detects foreign matter from the image of the transmitted light, noise generated from optical inhomogeneity of the inspection object is reduced. The device has been devised to sharpen the foreign matter to be detected.

  For example, in Patent Document 1, a first inspection light having a central wavelength in the near infrared region and a second inspection light having a central wavelength different from the first inspection light in the near infrared region are to be inspected. A first light and a second inspection imaged based on the first inspection light, the surface light source for irradiating a certain food, and the first inspection light and the second inspection light that are transmitted through the food. An inspection apparatus is disclosed that includes an imaging mechanism that outputs a second image captured based on light, and a difference image generation unit that generates a difference image between the first image and the second image.

  Patent Document 2 discloses, as a method for accurately detecting non-destructive foreign matter mixed in a plant or plant processed product, based on a hyperspectral image obtained by imaging the plant or plant processed product, A foreign matter detection method for a plant or a processed plant product that detects foreign matter mixed in an inspection object, the irradiation step of irradiating measurement light in the near-infrared region to the inspection target, and the irradiation step An imaging step that captures the inspection object by receiving scattered light from the inspection object by light to obtain a hyperspectral image, and between two different pixels included in the hyperspectral image obtained in the imaging step And an analysis step of detecting a foreign substance mixed in the inspection object based on the spectrum shape in the above.

JP-A-2014-44070 JP 2013-164338 A

  The foods, plants, processed plant products, and the like to be inspected in Patent Documents 1 and 2 described above are not optically homogeneous and have irregularities. Along with the difference, the optical inhomogeneity of the inspection object and the influence of unevenness appear as noise.

  Both the apparatus described in Patent Document 1 and the method described in Patent Document 2 discriminate between the inspection object and the foreign object by removing the influence of noise by detecting the difference between the absorption wavelengths of the inspection object and the foreign object. To do.

  However, if there is no clear difference in absorption wavelength between the inspection object and the foreign object, noise cannot be sufficiently removed.

  The present invention has been made in view of the above-described problems, and even when the difference in the rate of change in absorbance or transmittance with respect to the change in the inspection wavelength between the inspection object and the foreign object is not clear, the optical nonuniformity of the inspection object is not clear. An object of the present invention is to provide a food inspection apparatus that can reduce noise caused by property and the like and accurately determine foreign substances.

In order to achieve the above object, a food inspection apparatus according to the present invention comprises:
A light source for irradiating the inspection object with inspection light;
A receiver that receives the transmitted light and scattered light transmitted through the inspection object and generates image data;
A first polarizer disposed on an optical path between the light source and the inspection object to linearly polarize the inspection light;
A second polarizer disposed on an optical path between the inspection object and the receiver, and linearly polarizes the transmitted light;
Polarization direction control means capable of changing a relative angle between the transmission axis of the first polarizer and the transmission axis of the second polarizer,
The first image having a small relative angle and the second image having a large relative angle are picked up using the receiver, and foreign matter present in the inspection object is detected using the first image and the second image. To do.

Alternatively, the food inspection apparatus according to the present invention is
A light source for irradiating the inspection object with inspection light;
A receiver that receives the transmitted light and scattered light transmitted through the inspection object and generates image data;
A first polarizer disposed on an optical path between the light source and the inspection object and converting the inspection light into circularly polarized light;
A second polarizer disposed on an optical path between the inspection object and the receiver, and converts circularly polarized light into linearly polarized light and transmits the second polarizer;
A polarization direction control unit capable of switching a rotation direction of circularly polarized light converted by the first polarizer or a rotation direction of circularly polarized light transmitted by the second polarizer,
A first direction in which the rotation direction of the circularly polarized light converted by the first polarizer and the rotation direction of the circularly polarized light transmitted by the second polarizer are the same direction as the first image using the receiver is opposite to the first image. Two images are picked up, and foreign substances present in the inspection object are detected using the first image and the second image.

  In the food inspection apparatus according to the present invention, it is preferable that the light source has a different wavelength of inspection light to be irradiated when the first image is captured and when the second image is captured.

  In addition, a difference in light transmittance between the inspection object that does not include foreign matter and the inspection target that includes foreign matter at the wavelength of the inspection light emitted when the first image is captured is irradiated when the second image is captured. It is preferable that the difference in the transmittance at the wavelength of the inspection light is large.

  Further, it is preferable to have density adjusting means for performing density adjustment for increasing the contrast of the second image and / or decreasing the contrast of the first image.

  In addition, it is preferable to include an image processing unit that generates a difference image between the first image and the second image on which the density adjustment by the density adjustment unit is performed.

  Furthermore, it is preferable to have image recognition means for determining the presence or absence of foreign matter in the inspection object based on the difference image and / or specifying the position of the foreign matter.

  In the food inspection apparatus according to the present invention, noise due to foreign matter shadows and optical non-uniformity of the inspection object appears relatively clearly in the first image. On the other hand, in the second image, the noise due to the shadow of the foreign matter and the optical non-uniformity of the inspection object is visually unclear, but the noise component due to the same cause as the noise included in the first image (hereinafter referred to as “common noise”). ")").

  Therefore, if the density value of the second image is used, common noise is canceled from the first image, and an image with reduced noise is obtained. By performing foreign object determination on this image, erroneous determination can be prevented.

  Furthermore, if the wavelength of the inspection light to be irradiated is changed between the first image and the second image, the difference in light transmittance depending on the wavelength of the inspection object that does not include the foreign material and the inspection object that includes the foreign material is used. , The SN ratio can be further improved.

  Similarly, by performing density adjustment that increases the density contrast of the second image and / or decreases the contrast of the first image, the density change rate due to common noise included in the first image and the second image is approximated. be able to.

  Therefore, if the difference between the first image and the second image is obtained after the density adjustment, the common noise included in the first image can be canceled, and difference image data that can easily distinguish foreign substances and noise components can be generated. Since only foreign matters appear clearly in the image data from which noise components have been eliminated or reduced, the presence or absence of foreign matters can be accurately determined, and the location of foreign matters can be identified.

It is a figure which shows the basic composition of the food inspection apparatus which concerns on Embodiment 1 of this invention. FIG. 3 is a block diagram illustrating a configuration of a control system of the food inspection apparatus according to the first embodiment. It is FIG. (1) explaining the form of density adjustment and a noise reduction process. It is FIG. (2) explaining the form of density adjustment and a noise reduction process. It is FIG. (3) explaining the form of density adjustment and a noise reduction process. It is FIG. (4) explaining the form of density adjustment and a noise reduction process. It is FIG. (5) explaining the form of density adjustment and a noise reduction process. 3 is a flowchart showing a flow of foreign object inspection processing in the first embodiment. 3 is a drawing-substituting photograph showing the results of Experimental Example 1. FIG. 6 is a drawing substitute photograph (part 1) showing a result of Experimental Example 2; 10 is a drawing substitute photograph (part 2) showing a result of Experimental Example 2. 6 is a drawing substitute photograph showing the results of Comparative Example 1. FIG. 10 is a drawing-substituting photograph showing the results of Experimental Example 3. 10 is a drawing-substituting photograph showing the results of Experimental Example 4. 10 is a drawing-substituting photograph showing the results of Experimental Example 5. 10 is a drawing substitute photograph showing the results of Experimental Example 6. 10 is a drawing-substituting photograph showing the results of Experimental Example 7. 10 is a drawing-substituting photograph showing the results of Experimental Example 8. It is a figure which shows the basic composition of the food inspection apparatus which concerns on Embodiment 2 of this invention. 10 is a drawing substitute photograph (part 1) showing a result of Experimental Example 9; 10 is a drawing substitute photograph (part 2) showing a result of Experimental Example 9; 6 is a drawing substitute photograph showing the results of Comparative Example 2. FIG. It is a drawing substitute photograph which shows the result of Experimental example 10 in Embodiment 3 of this invention. It is a top view which shows the structure of the light source of the food inspection apparatus which concerns on Embodiment 4 of this invention. 10 is a drawing substitute photograph (part 1) showing a result of Experimental Example 11. 14 is a drawing substitute photograph (part 2) showing a result of Experimental Example 11. It is a graph which shows the relationship between the transmittance | permeability of a plum meat containing a ume meat single-piece | unit and a foreign material, and a wavelength. 10 is a drawing substitute photograph (part 1) showing a result of Experimental Example 12; 10 is a drawing substitute photograph (part 2) showing a result of Experimental Example 12; It is a graph which shows the relationship between the transmittance | permeability of a Chinese cabbage kimchi, and a foreign material (rubber glove), and a wavelength.

Hereinafter, a food inspection apparatus according to an embodiment of the present invention will be described with reference to the drawings.
<Embodiment 1>
FIG. 1 shows a basic configuration of a food inspection apparatus 100 according to Embodiment 1 of the present invention. FIG. 2 shows the configuration of the control system of the food inspection apparatus 100.
(Configuration and function of food inspection equipment)
The food inspection apparatus 100 includes a light source 1, an image receiver 2, a transparent plate 3, first and second polarizers 4, 5, a linear motor 6, a rail 7, and a controller 8 (see FIG. 2).

  The light source 1 irradiates the inspection object with inspection light. A light source such as an incandescent lamp or a halogen lamp may be used as long as it includes a wavelength band that easily transmits the inspection object. In the present embodiment, an LED panel in which a number of LEDs (light emitting diodes) are arranged on a substrate is used as the light source 1.

  The light source 1 is arranged in a direction to irradiate light on the inspection object placed on the transparent plate 3. The wavelength of the light source 1 should just be a wavelength which is easy to permeate | transmit a test subject, and can be suitably selected in the range of near-infrared light of wavelength 1200nm to near-ultraviolet light of wavelength 200nm according to the kind and thickness of a test subject. Preferably, light is easily transmitted through the inspection object in the range of visible light of 380 nm to near infrared light of 1100 nm. Also, in this area, the system can be easily constructed because it is within the wavelength range of the Si-type image sensor of a general CCD sensor or CMOS sensor.

  More preferably, it is 600 nm to 1100 nm. In this region, the observed absorption band is very weak, and the transmission is excellent. Moreover, at 600 nm or more, it becomes relatively less susceptible to scattering. In particular, when agricultural products, processed products thereof, other foods, and the like are used as inspection objects, it is preferable to select near-infrared light having a water absorption of 600 nm to 900 nm.

  The image receiver 2 is arranged in a direction to receive light irradiated from the light source 1 and transmitted through the inspection object placed on the transparent plate 3, and has a lens on the front cylindrical portion so that the inspection object is in focus. Is attached. As the receiver 2, a digital camera equipped with an image sensor such as a CCD or CMOS can be used, and it is preferable that the receiver 2 has good sensitivity in the wavelength region of the light source 1. The number of pixels can be selected according to the size of the inspection object and the foreign matter and the purpose of the inspection. The resolution is preferably 256 gradations (8 bits) or more.

  Between the light source 1 and the transparent plate 3, the 1st polarizer 4 is arrange | positioned in parallel with the light source 1 and the transparent plate 3, and the light irradiated from the light source 1 is polarized in the y direction in a figure ( That is, the transmission axis is in the y direction). In the figure, the stripe pattern of the polarizer is attached to make it easy to understand the polarization direction, and is not actually visible with the naked eye.

  For the polarizer, use a known polarizing optical element such as a polarizing plate obtained by pasting a polarizing film on a substrate such as a glass plate or a thermoplastic resin film, or a polarizing plate obtained by applying a dichroic dye or the like to a similar substrate. Can do. When the light source 1 is near-infrared light, a polarizing film for visible light cannot provide sufficient polarization performance, so a wire grid type polarizing film (for example, WGF (trademark) manufactured by Asahi Kasei E-Materials). ) Is preferably used.

  Further, a second polarizer 5 is disposed between the transparent plate 3 and the image receiver 2 in parallel with the light source 1, the transparent plate 3, and the first polarizer 4. The second polarizer 5 includes a y-direction polarization unit 51 that polarizes light transmitted through a measurement object in the y-direction in the drawing and an x-direction polarization unit 52 that polarizes light in the x-direction adjacent to each other. The frame 53 is integrated.

  Similarly to the first polarizer 4, the y-direction polarizing part 51 and the x-direction polarizing part 52 are provided on a polarizing plate in which a polarizing film is pasted on a glass plate or a thermoplastic resin film, or on a glass plate or a thermoplastic resin film. A known polarizing optical element such as a polarizing plate coated with a chromatic dye or a wire grid type polarizing plate can be used, and it preferably has the same polarization characteristics as the first polarizer 4.

  The linear motor 6 and the rail 7 constitute a polarization direction control means, and the second polarizer 5 has a position where the image receiver 2 can receive the light transmitted through the y-direction polarization section 51 by the polarization direction control means ( Hereinafter, it is movable between a “parallel position” and a position where the light transmitted through the x-direction polarization unit 52 can be received by the receiver 2 (hereinafter referred to as “orthogonal position”).

  That is, the frame 53 is coupled to the linear motor 6, and the linear motor 6 reciprocates in the direction indicated by the arrow along the rail 7 in accordance with a command from the polarizer driving unit 82 (see FIG. 2) of the controller 8. it can. The range of the reciprocating motion is such that one end is a parallel position and the other end is an orthogonal position.

  Accordingly, when the linear motor 6 is driven and the second polarizer 5 is positioned on the right side, the image receiver 2 receives the polarized light that has passed through the y-direction polarization unit 51 of the second polarizer 5. The relative angle between the transmission axis of the first polarizer 4 and the transmission axis of the second polarizer is 0 °.

  On the other hand, when the second polarizer 5 is positioned on the left side, the image receiver 2 receives the polarized light that has passed through the x-direction polarization unit 52 of the second polarizer 5, and thus the first polarizer 4. The relative angle between the transmission axis of the second polarizer and the transmission axis of the second polarizer is 90 °.

  Therefore, the receiver 2 has the same inspection object when the relative angle between the transmission axis of the first polarizer 4 and the transmission axis of the second polarizer 5 is small (0 °) and large (90 °) transmitted light and scattered light can be received and imaged. A controller 8 (see FIG. 2), which will be described later, is used to control the operation of the linear motor 6 and the timing at which the receiver 2 captures two types of images.

  In the following description, an image captured by the receiver 2 when the relative angle between the transmission axis of the first polarizer 4 and the transmission axis of the second polarizer 5 is small is referred to as “parallel image”, and the relative angle is An image captured when it is large is called an “orthogonal image”.

  The food inspection apparatus 100 shown in FIG. 1 is preferably shielded by a casing (not shown) except for the controller 8 in order to prevent the influence of external light. Further, if the inner surface of the housing and the surfaces of the linear motor 6 and the rail 7 are painted with a paint that prevents reflection of light, it is easy to prevent the influence of reflected light on imaging.

  The light source 1, the image receiver 2, the first polarizer 4, the transparent plate 3, and the rail 7 are fixed to the housing directly or by an indirect method using attachment parts.

  Next, the controller 8 will be described. The controller 8 controls the operations of the linear motor 6 and the image receiver 2 and processes image data captured by the image receiver 2. As shown in FIG. 2, the controller 8 includes a main control unit 81, a polarizer driving unit 82, an image receiver control unit 83, an image data storage unit 84, and an image processing unit 85.

  The main control unit 81 includes a known controller such as an industrial computer or a programmable controller, and performs sequence control on the entire apparatus.

  The polarizer driving unit 82 moves the linear motor 6 in the right direction or the left direction at a set speed in accordance with a command from the main control unit 81. For positioning, for example, a limit switch (not shown) that is turned on when the linear motor 6 is at a predetermined stop position is attached to the rail 7, and the linear motor is detected when the polarizer driving unit 82 detects a signal of the limit switch. 6 is stopped.

  The image receiver control unit 83 controls the image capturing timing of the image receiver 2 according to a command from the main control unit 81, acquires the captured image data from the image receiver 2, and further stores the image data in the image data storage unit 84. The image storage unit 84 further has an area for storing image data after density adjustment, difference image data, and image data obtained by erasing noise components with a threshold value.

  The image data stored in the image storage unit 84 is used for image processing calculation in the image processing unit 85. The image processing unit 85 includes a density adjustment unit 86, a difference calculation unit 87, and a noise elimination unit 88.

  The density adjusting unit 86 reads the orthogonal image captured by the receiver 2 and stored in the image data storage unit 84, and performs linear density conversion. Appropriate values for constants necessary for conversion are determined in advance by experiments. The difference calculation unit 87 generates a difference image between the orthogonal image and the parallel image whose density has been converted by the density adjustment unit 86. Next, the noise erasure unit 88 applies a density threshold value to the difference image to separate the foreign object from the noise, and generates image data in which the foreign object is clearly visible. Furthermore, you may add known noise processing, such as shrinkage | contraction, expansion | swelling, or the morphological transformation combining these, smoothing processing, etc. as needed.

If the image generated by the noise erasing unit 88 is output to the monitor, the noise is reduced, and the inspector can visually observe the image in which the foreign matter is clearly seen. Therefore, the presence or absence of the foreign matter and the foreign matter are easily present. If so, you can check its position. Moreover, noise can be reduced by providing an image recognition means (foreign matter detection means) that automatically determines the presence or absence of foreign matter by applying an image recognition algorithm to the image after noise elimination and identifies the position of the foreign matter. As a result, foreign matter can be detected based on a clear image, and automatic detection with high accuracy becomes possible.
(Principle of foreign matter inspection)
Next, the principle of foreign matter inspection by the food inspection apparatus 100 will be described. First, the characteristics of the captured parallel image and orthogonal image will be described.

  In the parallel image, since the transmission axis of the first polarizer 4 (polarization direction incident on the measurement target) and the transmission axis of the second polarizer 5 are the same direction, scattering of the light incident on the inspection target is performed. It contains a lot of light components that are not disturbed by polarization. That is, the polarized light incident on the inspection object is scattered by the internal structure of the inspection object and the polarization is disturbed, but a part of the incident light passes through the inspection object without being scattered.

  Accordingly, if the transmission axes of the first polarizer 4 and the second polarizer 5 are made parallel, a part of the scattered light whose polarization is disturbed can be blocked by the second polarizer 5, so that it reaches the receiver 2. The transmitted light can receive a large amount of transmitted light without scattering. Light that is transmitted without being scattered contains a lot of internal information such as shadows of foreign matter and noise caused by optical non-uniformity inside the inspection object, so the parallel image has a clear shape and size. It becomes an image.

  Here, the optical non-uniformity is an optical difference between living tissues such as pericarp and fruit, an optical difference of each component of the mixture, etc., and due to a difference in reflection and absorption of light, This is what causes density unevenness in a captured parallel image.

  On the other hand, in the orthogonal image, the transmission axis of the second polarizer 5 is set in the direction of blocking the transmitted light without being scattered, so that an image having a high ratio of scattered light can be obtained. In the orthogonal image, a shadow of a foreign matter or a fine shadow due to noise hardly appears or is an unclear image.

  Thus, although it is the parallel image and orthogonal image which are visually different, it is estimated that both images contain the noise by the same cause. The thickness of the inspection object is presumed to have the same tendency on the noise of both images. This is the density adjustment described below, and after calculating the difference between the parallel image and the orthogonal image, canceling noise by setting a threshold value for the density value of the difference image, regardless of the thickness of the inspection object with the same setting, This is a result of an experiment in which a clear image of a foreign object with reduced noise can be obtained.

  As described above, noise affected by the thickness of the inspection object appears in both the parallel image and the orthogonal image, but the density change rate (contrast) is small in the orthogonal image. The food inspection apparatus according to the present invention performs density adjustment as preprocessing for removing noise affected by thickness from a parallel image, and specifically increases the contrast of an orthogonal image or The contrast is lowered.

  As a method for increasing the contrast of the orthogonal image, for example, the density change rate may be increased by multiplying the density value of each pixel of the orthogonal image by a constant larger than 1, but the average value of the density of the entire orthogonal image may be increased. Therefore, it is preferable to use linear density conversion or the like that can control the average value separately.

  The above-described density adjustment is performed on the orthogonal image. However, any density adjustment that approximates the density change rate of noise affected by the thickness of the inspection object included in the parallel image and the orthogonal image may be used. You may perform density adjustment which reduces contrast with respect to an image. Further, the density of both the parallel image and the orthogonal image may be adjusted.

  If a difference image between a parallel image and an orthogonal image after density adjustment is generated, noise affected by the thickness of the inspection object can be canceled. Therefore, it is easy to distinguish a foreign object from noise by providing a threshold value for the density of the difference image. .

  Here, the difference image is an image having, as the density value of the corresponding pixel, a value obtained by subtracting the density value of the corresponding pixel of the orthogonal image (or parallel image) from the density value of each pixel of the parallel image (or orthogonal image). It is.

  Note that the density value type of an image is usually an unsigned integer type. However, if the difference is taken, the density value becomes a negative value, and an error may occur in the calculation or the obtained result may not be obtained. In this case, if the variable is converted into a signed integer type or floating point type by the processing program, the calculated result can be obtained without causing an error. In order to save the calculation result as image data, for example, for a calculation result having a negative value, an unsigned integer type is added again by adding a constant having a minimum density value of 0 or more to the density value of all pixels. Convert to

  In order to generate a clear image of only foreign matter, in the difference image, only pixels having a density value higher (or lower) than a predetermined threshold value are replaced with, for example, density value 255 (white), and other pixels are replaced. If it is replaced with 0 (black), a binarized image in which only foreign matter is clear can be obtained.

  Concentration adjustment and threshold value vary depending on conditions such as the type of inspection object and foreign matter, and the wavelength of the light source. Therefore, if an appropriate value is obtained through experimentation and set, the inspection object varies in thickness under certain conditions. A clear image for detecting foreign matter can be obtained.

  Further, processing is performed in which the density value of the orthogonal image whose density is appropriately adjusted is set as a threshold value, the density value of the pixel whose density value of the parallel image is higher than the density value of the orthogonal image is set to 0, and the density value of the low pixel is set to 255. Even if it goes, a clear image of only foreign matters can be obtained.

  The density adjustment and noise reduction processing described above will be specifically described with reference to the drawings. There are four types of density adjustment and noise reduction processing. However, it is not necessarily limited to these forms.

  A first form of density adjustment and noise reduction processing will be described with reference to FIG. FIG. 3A is a graph (raw waveform) representing the density values of pixels in an arbitrary row (horizontal direction) of the parallel image and the orthogonal image captured by the receiver. In the figure, the vertical axis represents density values, and the horizontal axis represents pixel coordinates. Note that the vertical coordinates of the pixels shown in the graph are the same for the parallel image and the orthogonal image. In the figure, the solid line represents the density value of the parallel image, and the broken line represents the density value of the orthogonal image. A circle indicates a portion where the density value of the parallel image has fallen due to the influence of foreign matter.

  In the density value of the parallel image, fine noise based on the optical non-uniformity of the inspection object appears in addition to the influence of the foreign matter, but hardly appears in the density value of the orthogonal image. On the other hand, both images show noise with a relatively large period and in phase.

  The influence of the two foreign matters in the raw waveform of the parallel image is noise with a large period as described above, so that the density level thereof is different, and it is difficult to set a threshold value for separating the foreign matters from fine noise. Therefore, in the present invention, two foreign objects and fine noise can be separated by determining a threshold value based on the density value of an orthogonal image having a large period of noise.

  FIG. 3B is a graph showing the density value when the density adjustment is performed by multiplying the density value of the orthogonal image by one or more constants. The density value of the parallel image is the same as that in FIG. By multiplying the density value of the orthogonal image by one or more density values, the amplitude of the noise having a large period is amplified to the same extent as that of the parallel image.

  FIGS. 3C and 3D show the results of obtaining the difference between the density values of the orthogonal image and the parallel image after the density adjustment. 3C is a difference result obtained by subtracting the density value of the orthogonal image from the density value of the parallel image, and FIG. 3D is a difference result obtained by subtracting the density value of the parallel image from the density value of the orthogonal image. is there. In the figure, a constant value threshold is indicated by a broken line. By taking the difference, it can be seen that the drop in density due to the influence of two foreign substances is at the same level, and fine noise and foreign substances can be separated with one threshold.

  When the waveforms in FIGS. 3C and 3D are binarized with a threshold value, the binarized images shown in FIGS. 4A and 4B are obtained, noise is eliminated, and only the foreign matters become clear. It becomes an image.

  Next, a second embodiment using linear density conversion for density adjustment will be described with reference to FIG. FIG. 5A is the same raw waveform as the waveform shown in FIG. FIG. 5B shows the result of using linear density conversion for density adjustment of an orthogonal image. In the linear density conversion, unlike the method of multiplying by a constant, the average value can be controlled independently while increasing the amplitude of noise. Therefore, as shown in FIG. 5B, density adjustment can be performed by controlling so that fine noise in the parallel image does not appear at a density lower than the density value of the orthogonal image.

  In this state, if the difference obtained by subtracting the density value of the parallel image from the density value of the orthogonal image is obtained, the result shown in FIG. 5C is obtained. If the threshold value is set to 0, foreign matter and noise can be separated. Thereafter, when the difference result is subjected to a process of replacing all density values less than 0 with 0, noise is eliminated and an image with clear foreign matters is obtained as shown in FIG.

  As a modification of the second embodiment, if density processing is performed by linear density conversion and then difference processing with addition of saturation calculation processing is performed, the state shown in FIG. The state shown in FIG. This is the third form.

  Here, in the saturation calculation process, when the density value takes a value outside the image range (0 to 255 in 8 bits) by performing image processing, it is automatically replaced with the minimum value or the maximum value of the range. It is processing. In many commercially available image processing software, a saturation calculation process is added to an image processing command, and when a difference calculation is performed, all density values less than 0 are replaced with 0.

  Next, a fourth embodiment of density adjustment and noise reduction processing will be described with reference to FIG. The raw waveform in FIG. 6 (a) and FIG. 6 (b) after linear density conversion are the same as FIG. 5 (a) and FIG. 5 (b).

  In the fourth mode, the density value of the orthogonal image is directly applied as a threshold without taking the overall difference. That is, the density value of a pixel in which the density value of the parallel image is larger than the density value of the orthogonal image is set to 0, and the density value of the parallel image is changed from the density value of the orthogonal image to the pixel whose density value of the parallel image is equal to or less than the density value of the orthogonal image. The difference obtained by subtracting is set as the density value.

  Then, the result of eliminating the noise is obtained as shown in FIG. This figure is the same as FIG. However, in the state shown in these figures, although noise is eliminated, the brightness (density value) of the foreign matter is low, so it is preferable to perform a process of increasing the brightness. When the density value is multiplied by a constant, the brightness of the foreign matter is increased as shown in FIG. 6D, and when the image is projected on the monitor, it is easy to visually discriminate.

  As described above, the noise can be eliminated from the parallel image based on the density value of the orthogonal image that has been subjected to the density adjustment for increasing the contrast. At this time, there are a method of binarizing and a method of separating the foreign matter and noise by a threshold value and a method of leaving the foreign matter as a gray scale (difference between parallel and orthogonal).

  With either method, it is possible to separate the noise from the foreign object and eliminate the noise. However, since the noise may not be completely removed, the noise is further reduced by performing known noise processing such as morphological transformation and smoothing processing. Can be erased. In this case, the remaining noise is easier to process than the binarization process in which the foreign matter is left in grayscale.

  The above process will be described with reference to FIG. In the case of binarization, the density of the remaining noise is the same as that of a foreign object (255 in the first embodiment), and the size is large as shown in FIG. On the other hand, in the method of leaving the foreign matter in gray scale, the noise density is low as shown in FIGS. 7B and 7C, and the size becomes smaller as the density becomes higher. Therefore, noise can be easily eliminated by erosion (expansion following contraction) processing.

  Further, in the method of leaving the foreign matter in gray scale, the noise remains dark even if the brightness adjustment is performed by multiplying the density value of each pixel by a constant as shown in FIG. This effect is further enhanced when gamma correction with γ <1 is applied to the brightness adjustment.

  The density adjustment method is not limited to adjustment of captured image data. That is, it is possible to obtain the same effect as the density adjustment for the image data by controlling the amount of light received by the receiver 2 during imaging. For example, if the amount of light at the time of orthogonal image capturing is increased from the amount of light at the time of parallel image capturing, the density change rate of noise affected by the thickness of the inspection object can be increased to approximate the parallel image.

  As a method for controlling the amount of light received by the image receiver 2, there are a method for controlling the output of the light source 1 and a method for controlling the shutter speed of the image receiver. In order to control the output of the light source 1, for example, a light source control unit may be provided in the controller 8 to control the value of a current flowing through the LED of the light source. The shutter speed can be controlled, for example, by sending a command for switching the shutter speed from the receiver control unit 83 to the receiver 2 before capturing a parallel image and before capturing an orthogonal image.

For the parallel image and the orthogonal image, the current value or shutter speed to be passed at the time of each image capture can be adjusted by adjusting the density of the image data and applying a numerical value obtained by conducting an experiment in advance as in the case of determining the linear density conversion constant. good.
(Processing flow in foreign object inspection)
Next, the flow of foreign matter inspection processing in this embodiment will be described with reference to the flowchart of FIG.

  First, the polarizer driving unit 82 confirms the position of the second polarizer 5 (step S1), and when it is at a position other than the right end (parallel position), the linear motor 6 is moved to the right (step S2). .

  The movement in the right direction is repeated until it is detected that the second polarizer 5 is at the right end (step S2), and the linear motor 6 is stopped at the detected stage (step S3).

Next, the receiver control unit 83 instructs the receiver 2 to take a first image (parallel image) (step S4), and the image data of the first image is stored in the image data storage unit 84. (Step S5)
Subsequently, the polarizer driving unit 82 confirms the position of the second polarizer 5 (step S6). Since the second polarizer 5 is at a position other than the left end (orthogonal position), the linear motor 6 is moved to the left (step S7).

  In step S6, the process of step S7 is repeated until it is detected that the second polarizer 5 is at the left end, and the linear motor 6 is stopped at the detected stage (step S8).

  Next, the receiver control unit 83 instructs the receiver 2 to capture a second image (orthogonal image) (step S9), and subsequently stores the image data of the second image in the image data storage unit 84 (step S9). S10).

  Subsequently, the process proceeds to the processing in the image processing unit 85, and the density adjustment unit 86 adjusts the density value of the second image (orthogonal image) (step S11). As long as the density adjustment is a process for increasing the density change rate (contrast) of the second image, any of the above-described forms may be employed. For example, a method of multiplying the density value of the second image by a constant larger than 1 may be used, but it is preferable to perform linear density conversion that can control the average density value.

  Next, based on the density value of the second image whose density has been adjusted in step S11, processing for reducing noise in the first image is performed (step S12). As a process for reducing the noise of the first image based on the density value of the second image, for example, the difference calculation unit 87 generates a difference image between the second image after the linear density conversion and the first image, and eliminates the noise. The unit 88 applies a threshold value to the difference image to separate the foreign object and the noise, and generates an image in which the foreign object is clearly seen with the noise reduced. In step S12, it is also possible to obtain an image with reduced noise directly from the first image using the density value of the second image as a threshold value.

  Finally, image recognition is executed based on the image data after noise reduction to determine the presence or absence of foreign matter (step S13), the determination result is output, and the series of processing ends.

  For the image recognition, for example, a method of recognizing a foreign object with the size of a binarized and labeled area can be employed. Also, a known recognition method can be selected according to the inspection object, foreign matter, and required inspection accuracy, such as pattern recognition from which feature points are extracted.

  For example, the determination result is output when a flag is set when the image recognition program recognizes that a foreign object is present in the image, and when the flag is set, a warning is displayed on the monitor or an alarm device is activated. The output means such as Further, the inspection line can be automatically stopped and recorded in the quality database.

Further, the position of the foreign matter may be calculated from the coordinates of the image and output. In addition, when the types of foreign matters are recognized separately based on the feature points of a plurality of types of foreign matters, it is easy to identify the cause of foreign matter contamination in the previous process, which is useful for improvement.
(Experimental example 1)
Next, the results of experiments using the food inspection apparatus 100 according to the present embodiment will be described.

  In this experiment, finely chopped plum meat was used as an inspection object. This plum meat is in an optically non-uniform state with a paste-like pulp and a fiber-containing skin. On the transparent plate 3, a black rubber plate having a rectangular (18 mm × 22 mm) window at the center was placed at a position where the light receiving axis of the image receiver 2 passes through the center of the window.

  On the transparent plate 3 in the window of the rubber plate, the red larvae of red snapper, a piece of rubber glove and an aluminum foil piece were placed as foreign matters, and the plum meat was placed thereon so that the thickness was 5 mm. . The black rubber plate is provided to shield light emitted from the light source 1 and reaching the image receiver 2 without passing through the inspection object.

  The light source 1 is an LED panel in which LEDs with a wavelength of 850 nm (ROHM SIR-568ST3F) are arranged on a 6.53 m x 7.6 cm substrate so that the surface density is 3 per square centimeter. A current of 0.1 A was applied to light up.

  The receiver 2 uses a WATEC monochrome CCD camera (WAT-910HX), a TAMRON IR (infrared) compatible varifocal lens (12VM412ASIR), and focuses on the examination object. The shutter speed was fixed at 1 msec.

  The first polarizer 4 and the y-direction polarization part 51 and the x-direction polarization part 52 of the second polarizer 5 are all glass plates with a wire grid polarization film (WGF-HTU) made by Asahi Kasei E-Materials. did.

  The positional relationship of each component is 50 mm between the light source 1 and the first polarizer 4, 50 mm between the first polarizer 4 and the transparent plate 3, and between the transparent plate 3 and the second polarizer 5. Is 25 mm, and the distance between the second polarizer 5 and the lens attached to the receiver 2 is 20 mm, and all the components are arranged on a straight optical path from bottom to top.

  FIG. 9A is a photograph of a parallel image (relative angle 0 °) captured by setting the second polarizer 5 at a position where the receiver 2 can receive light through the y-direction polarization unit 51. On the other hand, FIG. 9B is a photograph of an orthogonal image (relative angle 90 °) captured by setting the second polarizer 5 at a position where the receiver 2 can receive light through the x-direction polarization unit 52.

  The density adjustment of 1.4 times the density value of all the pixels of the orthogonal image shown in FIG. 9B is performed, and the density value of the parallel image shown in FIG. 9A is subtracted from the density value of the orthogonal image after the density adjustment. A difference image was generated. FIG. 9C is a photograph of an image obtained by adjusting the density by increasing the density value of all the pixels of the difference image by five times.

  Note that when the difference value is executed, if the density value is 0 or less, the process of replacing the value with 0 is performed, so the process of erasing noise with the threshold value is simultaneously performed. The same applies to the subsequent difference calculation.

  Further, density adjustment by linear density conversion was performed on the same orthogonal image (FIG. 9B), and similarly, a difference image from the parallel image (FIG. 9A) was generated. In the linear density conversion, the density value A to the original image is set so that the density value lower limit A of the original image corresponds to the density value 0 of the generated image and the density value upper limit B of the original image corresponds to the density value 255 of the generated image. The density value in the B section is converted linearly. Linear density conversion was performed with the density value lower limit A being 10 and the upper limit B being 150. FIG. 9D is a photograph of an image obtained by adjusting the density by increasing the density value of all the pixels of the difference image 10 times.

9 (c) and 9 (d), the noise due to the influence of plum skin and fiber in the parallel image was greatly reduced. However, in the photo (Fig. 9 (c)) in which the density was adjusted by multiplying by a constant, a little ume peel noise remained on the left side, but it disappeared in the photo (Fig. 9 (d)) after the linear density conversion. .
(Experimental example 2)
Next, the effect was confirmed when the thickness of the test object was changed, with the red larvae of the red spotted larvae as foreign substances and the same plum meat as in Experimental Example 1 as the test object.

  An LED panel in which an LED (OSRAM, SFH4550) with a wavelength of 850 nm is arranged on a 6.53 m x 7.6 cm substrate so that the surface density is 3 per square centimeter is the light source 1, and the current value applied to each LED is The imaging conditions were the same as those in Experimental Example 1, except that 0.1 A was set, the shutter speed was 4 msec, and the density adjustment after the difference calculation was 20 times.

  10 (a) and 10 (b) show a parallel image and an orthogonal image with a thickness of 2 mm, and FIGS. 10 (c) and 10 (d) show a parallel image and an orthogonal image with a thickness of 5 mm. Further, FIGS. 11A and 11B show the results of generating a difference image by the same method as in Experimental Example 1 by performing linear density conversion with the density value lower limit A being 25 and the upper limit B being 170. FIG. FIG. 11A shows a thickness of 2 mm and FIG. 11B shows a thickness of 5 mm.

11 (c) and 11 (d) show the results of generating a difference image using the same method as in Experimental Example 1 by performing linear density conversion with the density value lower limit A being 30 and the upper limit B being 165. FIG. FIG. 11C shows a thickness of 2 mm, and FIG. 11D shows a thickness of 5 mm. In both cases, the noise is almost gone, and only the foreign objects are clear.
(Comparative Example 1)
As a comparative example, a difference calculation was performed between the same parallel image as in experimental example 2 and an image in which all pixels were filled with a constant density value. Using the density value of the filled image as a threshold value, the same effect is obtained as eliminating noise by painting a pixel having a higher density value of the parallel image with one color. Processing was performed in the same manner as in Experimental Example 2 except that the density value of the filled image was as shown below and density conversion was not performed.

  FIGS. 12A and 12B show difference images generated by setting the density value of the filled image as 155. FIG. FIG. 12A shows a thickness of 2 mm, and FIG. 12B shows a thickness of 5 mm. FIGS. 12C and 12D show difference images generated with the density value of the filled image as 160. FIG. FIG. 12C shows a thickness of 2 mm, and FIG. 12D shows a thickness of 5 mm.

  As shown in FIG. 12A, when the density value of the filled image is 155, the foreign matter is unclear when the thickness is 2 mm. On the other hand, the foreign matter is clear at a thickness of 5 mm, but some noise remains.

On the other hand, when the density value of the filled image is 160, foreign matters become clear when the thickness is 2 mm, but noise increases when the thickness is 5 mm. Therefore, it can be seen that the thickness is 2 mm and 5 mm, the foreign matter is clear, and the condition in which noise does not remain cannot be found. From this result, it can be seen that the effectiveness of the experimental example 2 in which the difference from the orthogonal image is obtained, as compared with the comparative example 1 in which the threshold is simply applied to the parallel image.
(Experimental example 3)
The effect of the present invention was confirmed by adjusting the contrast between the parallel image and the orthogonal image by changing the current value of the light amount of the light source 1 instead of adjusting the density of the orthogonal image. An arithmetic process was performed without changing the current value, and the density adjustment after the difference calculation was performed 10 times except that the image was taken under the same conditions as in Experimental Example 2 to generate a difference image. . The current value was set to 0.06 A for the parallel image and 0.1 A for the orthogonal image while checking the difference image.

13A shows a parallel image, FIG. 13B shows an orthogonal image, and FIG. 13C shows a difference image. The noise disappeared and a clear image of foreign matter was obtained as in the case of density conversion on the orthogonal image.
(Experimental example 4)
The effect of the present invention was confirmed by adjusting the contrast between the parallel image and the orthogonal image by changing the shutter speed of the received light amount of the receiver 2 instead of adjusting the density of the orthogonal image. An arithmetic process was performed without changing the shutter speed, and the density adjustment after the difference calculation was performed 10 times, and the image was taken under the same conditions as in Experimental Example 2 to generate a difference image. . The shutter speed was set to 1 msec for the parallel image and 2 msec for the orthogonal image while checking the difference image.

14A shows a parallel image, FIG. 14B shows an orthogonal image, and FIG. 14C shows a difference image. The noise disappeared and a clear image of foreign matter was obtained as in the case of density conversion on the orthogonal image.
(Experimental example 5)
The inspection object was changed from plum meat to strawberry jam, and the effect of the present invention was confirmed using a piece of rubber glove as a foreign object. In this experimental example, the change of the shutter speed and the linear density conversion for the orthogonal image are used together for the density adjustment. The thickness of the jam was 4 mm. The shutter speed was set to 0.01 msec for the parallel image and 0.2 msec for the orthogonal image, the amount of light received by the receiver 2 was adjusted, and linear density conversion with the lower limit A4 and the upper limit B212 was performed on the orthogonal image. The other conditions are the same as in Experimental Example 2.

FIG. 15A shows a parallel image, FIG. 15B shows an orthogonal image, and FIG. 15C shows a difference image. In the difference image, the noise due to the fibers in the parallel image disappears, but the strawberry seeds are clearly visible along with the foreign matter. However, since the seeds are small, it is possible to acquire an image in which only foreign matters are left later by noise processing such as morphological conversion and smoothing processing. If attention is paid to the seeds, the present invention can be applied to confirm the presence / absence of the seeds and the state of dispersion in addition to the foreign substance inspection.
(Experimental example 6)
The test object was changed from plum meat to jelly containing plum pulp, and the effect of the present invention was confirmed using a piece of rubber glove as a foreign object. In this experimental example, the change of the shutter speed and the linear density conversion for the orthogonal image are used together for the density adjustment. The thickness of the jelly was 4 mm. The shutter speed was set to 0.1 msec for the parallel image and 0.5 msec for the orthogonal image, and the amount of light received by the receiver 2 was adjusted, and linear density conversion with the lower limit A8 and the upper limit B173 was performed on the orthogonal image. The other conditions are the same as in Experimental Example 2.

FIG. 16A shows a parallel image, FIG. 16B shows an orthogonal image, and FIG. 16C shows a difference image. In the difference image, the noise due to the fiber in the parallel image disappeared, and a clear image of the foreign matter was obtained.
(Experimental example 7)
The test object was changed from plum meat to Chinese cabbage kimchi, and the effect of the present invention was confirmed using a piece of rubber glove as a foreign object. The thickness of Chinese cabbage kimchi was 4 mm. The shutter speed was set to 0.5 msec, the amount of light received by the image receiver 2 was adjusted, and linear density conversion was performed on the orthogonal image with a lower limit A13 and an upper limit B196. Further, the density adjustment after the difference calculation is set to 20 times. The other conditions are the same as in Experimental Example 2.

FIG. 17A shows a parallel image, FIG. 17B shows an orthogonal image, and FIG. 17C shows a difference image. Noise disappeared and a clear image of foreign matter was obtained.
(Experimental example 8)
The effect of the present invention was confirmed by changing the light source 1 of Experimental Example 2 from near infrared to visible light. The light source was a white LED (Monotaro 24LED round light). The shutter speed was 333 msec, and the density adjustment conditions were linear density conversion (lower limit A was 10 and upper limit B was 191). Moreover, the thickness of plum meat was 4 mm, and the other conditions were the same as in Experimental Example 2.

  FIG. 18A shows a parallel image, FIG. 18B shows an orthogonal image, and FIG. 18C shows a difference image. The noise disappeared and a clear image of the foreign matter was obtained as when the light source was near infrared.

  As described above, in the food inspection apparatus according to the present invention, noise due to the shadow of the foreign matter and the optical non-uniformity of the inspection object appears relatively clearly in the first image. On the other hand, in the second image, the noise due to the shadow of the foreign matter and the optical non-uniformity of the inspection object is visually unclear, but the noise component due to the same cause as the noise included in the first image (hereinafter referred to as “common noise”). ")").

  Therefore, if the density value of the second image is used, common noise is canceled from the first image, and an image with reduced noise is obtained. By performing foreign object determination on this image, erroneous determination can be prevented.

  Further, since the apparatus has density adjusting means for performing density adjustment for increasing the density contrast of the second image and / or reducing the contrast of the first image, the rate of change in density due to common noise included in the first image and the second image Can be approximated.

Therefore, if the difference between the first image and the second image is obtained after the density adjustment, the common noise included in the first image can be canceled, and difference image data that can easily distinguish foreign substances and noise components can be generated. As a result, since only foreign matters appear clearly in the image data from which noise components have been eliminated or reduced, the presence or absence of foreign matters can be accurately determined, and the location of foreign matters can be identified.
<Embodiment 2>
FIG. 19 shows a basic configuration of food inspection apparatus 100 according to Embodiment 2 of the present invention. In the figure, the same members as those shown in FIG. The controller is the same as the controller 8 shown in FIG. 2 except that the control target of the polarizer driving unit 82 is the stepping motor 9.

  In the present embodiment, instead of the y-direction polarizer 51, the x-direction polarizer 52, the frame 53, the linear motor 6 and the rail 7 constituting the second polarizer 5 of the first embodiment shown in FIG. A circular polarizer 54, a pulley-equipped frame 55, a stepping motor 9, a support plate 10, a motor pulley 11 and a pulley belt 12 are used.

  The pulley-equipped frame 55 is rotatably engaged with the hole portion of the support plate 10, and flanges are provided at the upper and lower portions so as not to fall off the support plate 10. The circular polarizer 54 is fixed on the outer periphery of the pulley-equipped frame 55 so that the direction of the transmission axis can be changed as the pulley-equipped frame 55 rotates.

  Further, a case of the stepping motor 9 is fixed to the support plate 10, and a motor pulley 11 is attached to the rotating shaft of the stepping motor 9. The pulley-equipped frame 55 is provided with a pulley groove on the outer periphery, and is engaged by the motor pulley 11 and the pulley belt 12 so that power from the stepping motor 9 is transmitted.

  Therefore, if the rotation of the stepping motor 9 is controlled, the circular polarizer 54 can be rotated, and the relative angle between the transmission axis of the first polarizer 4 and the transmission axis of the second polarizer 5 can be changed. Can do.

By adopting the configuration shown in FIG. 19, the mechanical operation in the polarization direction control can be changed from peristaltic to rotational motion, so that vibration can be reduced. In addition, since the relative angle of the circular polarizer changes from 0 ° to 90 ° every rotation of 180 °, parallel images and orthogonal images can be captured by rotating in one direction.
(Experimental example 9)
The results of experiments using the food inspection apparatus according to the present embodiment will be described. In the first embodiment, an image captured when the relative angle between the first polarizer and the second polarizer is 0 ° is treated as a parallel image, and an image captured when the relative angle is 90 ° is treated as an orthogonal image. If the relative angle of the parallel image is smaller than the relative angle of the orthogonal image, the present invention is not limited to this and the effects of the present invention can be achieved. Therefore, in order to clarify the scope of the present invention, the effective range of the relative angle between the parallel image and the orthogonal image was examined.

  FIG. 20 shows a photograph taken by changing the relative angle between the first polarizer 4 and the second polarizer 5 every 10 °. The imaging conditions were the same as in Experimental Example 2 described above, and the thickness of plum meat was 5 mm. FIG. 21 is a difference image generated by combining the images of FIG. As before, the density value that is 0 or less was replaced with 0 in the difference calculation. The lower limit value A and the upper limit value B of the linear density conversion are as described in the upper part of each photograph.

  The upper photograph in FIG. 21 is a difference image when the relative angle is changed by 10 ° for both the parallel image and the orthogonal image. Even when the parallel image has a relative angle of 20 ° and the orthogonal image has a relative angle of 70 °, a difference image in which the noise is eliminated and the foreign matter becomes clear is obtained. Further, when the parallel image has a relative angle of 30 ° and the orthogonal image has a relative angle of 60 °, the sharpness of the foreign matter slightly decreases, but the sharpness can be sufficiently used for subsequent image recognition.

  As shown in the middle of FIG. 21, when the relative angle of the orthogonal image is fixed at 90 °, a clear foreign object image without noise can be obtained even if the relative angle of the parallel image is 50 °. In this case, the upper limit of the relative angle of parallel images that can be used for image recognition is 70 °.

As shown in the lower part of FIG. 21, when the relative angle of the parallel image is fixed at 0 °, a clear foreign object image without noise can be obtained even if the relative angle of the orthogonal image is set to 70 °. In this case, the lower limit of the relative angle of the orthogonal image that can be used for image recognition is 50 °.
(Comparative Example 2)
FIG. 22 is a difference image in a range in which the effect of the present invention is not obtained as a result of examining the range of the relative angle between the parallel image and the orthogonal image as in Experimental Example 3. If the difference between the relative angles of the parallel image and the orthogonal image is less than 20 °, the noise cannot be erased and the foreign matter cannot be visually recognized. The same applies when the relative angle of the orthogonal image is less than 50 °.

From the above results, the relative angle of the parallel image must be 70 ° or less, the relative angle of the orthogonal image must be 50 ° or more, and the relative angle of the orthogonal image must be 20 ° or more larger than the relative angle of the parallel image. . Preferably, when the relative angle of the parallel image is 50 ° or less, the relative angle of the orthogonal image is 70 ° or more, and the relative angle of the orthogonal image is increased by 50 ° or more with respect to the relative angle of the parallel image, the noise is eliminated, Since an image with clearer foreign matters can be obtained, the accuracy of subsequent image recognition is improved.
<Embodiment 3>
Although Embodiments 1 and 2 both use linearly polarized light, circularly polarized light can also be used in the food inspection apparatus according to the present invention. In the present embodiment, the first polarizer 4 and the second polarizer 5 of the first embodiment are configured by circularly polarizing plates.

  The first polarizer 4 converts light into clockwise circularly polarized light. Although the second polarizer 5 is not shown, the y-direction polarization unit 51 is replaced with a right-handed rotation transmission unit that converts clockwise circularly polarized light into linearly polarized light and transmits it, and an x-direction polarization unit 52 is counterclockwise. It is replaced with a counterclockwise transmission part that converts circularly polarized light into linearly polarized light and transmits it.

  The polarization direction control means switches between a normal transmission position where light from the inspection object can be received through the right rotation transmission section and a reverse transmission position where light can be received through the left rotation transmission section. Further, the image receiver 2 captures the first image at the forward transmission position and the second image at the reverse transmission position, and stores the image data in the image data storage unit 84.

  Similar to the parallel image in the first embodiment, the first image is a clear image of foreign matter and noise. In addition, the second image is an image in which the shadows of foreign matter and noise are unclear as in the orthogonal image.

As in the first embodiment, if the density adjustment by the density adjustment unit and the difference image are acquired and the noise is eliminated by the threshold value, the foreign object can be detected with high accuracy.
(Experimental example 10)
The results of experiments using the food inspection apparatus according to the present embodiment will be described. The effect was confirmed by replacing the polarizer of Experimental Example 7 with a circularly polarizing plate. The first polarizer 4 is a counterclockwise circular polarizer (TCPL manufactured by Biei Imaging Co., Ltd.), and the second polarizer 5 for the first image is also a counterclockwise circularly polarizing plate. In other words, the receiver 2 is arranged so that the light transmitted through the first polarizer 4 can be received in a state where the plum meat is not installed.

  The second polarizer 5 for the second image uses a clockwise circularly polarizing plate (TCPR manufactured by Biei Imaging Co., Ltd.) and the first polarized light in a state where the plum meat to be inspected is not installed. The light was transmitted through the child 4 so that the receiver could not receive the light. Other conditions are the same as in Experimental Example 8. In the density adjustment, linear density conversion (lower limit A is 10 and upper limit B is 191) is performed on the second image, and in the difference calculation, the density value of the first image is subtracted from the density value of the second image.

FIG. 23A shows the first image, FIG. 23B shows the second image, and FIG. 23C shows the difference image. Compared with the linearly polarized light of Experimental Example 7 shown in FIG. 18, the effect is reduced, but even circularly polarized light has the effect of reducing noise and clarifying foreign matters.
<Embodiment 4>
In the first to third embodiments, a light source that irradiates inspection light having the same wavelength in the first image (parallel image) and the second image is used. In the present embodiment, the first image (parallel) is used. A light source capable of irradiating inspection light having different wavelengths in the second image and the second image is used.

  In each of the above-described embodiments, the light source has been described on the assumption that the first image and the second image are irradiated with inspection light having the same wavelength, but the first image and the second image have different wavelengths. If the light source which irradiates inspection light is used, the food inspection apparatus which was further excellent in the noise reduction effect can be provided.

  That is, at the time of imaging the first image, the inspection light in the wavelength region (first wavelength) in which the difference between the light transmittance with respect to the inspection object that does not include the foreign material and the light transmittance with respect to the inspection object that includes the foreign material is relatively large. If it irradiates, the 1st image with a clear shadow of a foreign material will be obtained. On the other hand, when the second image is picked up, if the inspection light in the wavelength region (second wavelength) in which the difference in transmittance is small compared to the first wavelength is irradiated, the shadow of the foreign object is picked up thinner in the second image. The As a result, the S / N ratio is improved in the difference image between the first image and the second image, and foreign matter and noise can be more easily discriminated.

  FIG. 24 shows the configuration of the light source 1 used in the present embodiment. In the present embodiment, an LED panel in which two types of LEDs are arranged in a staggered manner is used as the light source 1 in order to selectively irradiate inspection lights having different wavelengths. In the figure, the first wavelength LED that is lit when the first image is captured is indicated by a white circle, and the second wavelength LED that is lit when the second image is captured is indicated by a black circle.

The lighting switching of the first wavelength and the second wavelength is performed when the LED of the first wavelength is turned on, for example, by switching the power supply in synchronization with the linear motor stop command in steps S3 and S8 in the flowchart of FIG. An image can be taken and a second image can be taken when the second wavelength LED is lit.
(Experimental example 11)
The results of experiments using the food inspection apparatus according to the present embodiment will be described.

  In this experiment, finely chopped plum meat was used as an inspection object. This plum meat is in an optically non-uniform state with a paste-like pulp and a fiber-containing skin. On the transparent plate 3, a light shielding plate having a rectangular (90 mm × 140 mm) window at the center was placed at a position where the light receiving axis of the image receiver 2 passes through the center of the window.

  On the transparent plate 3 in the window portion of the light shielding plate, red larvae of red scallop were placed as a foreign substance, and the plum meat was placed thereon so as to have a thickness of 3 mm. The light shielding plate is provided to shield light emitted from the light source 1 and reaching the image receiver 2 without passing through the inspection object.

  As the light source 1, an LED panel in which LEDs with a wavelength of 850 nm (OSH SHF4550) and LEDs with a wavelength of 940 nm (OSRAM SFH4545) are arranged in a zigzag pattern with the dimensions shown in FIG. The current value when the LED was lit was 7.87 mA at a wavelength of 850 nm and 13.44 mA at a wavelength of 940 nm.

  In this experimental example, the lighting of the LED can be controlled in order to examine the wavelength of the light source when capturing parallel images and orthogonal images. That is, an experiment was performed using an experimental apparatus that can selectively light one of the two wavelengths of LEDs when the first image and the second image are captured.

  A Baumer monochrome CMOS camera (HXC40-NIR) was used for the receiver 2 and a near-infrared lens (NIR-KW25) made by Ad Science was attached to focus on the inspection object. The shutter speed was fixed at 10msec. The first polarizer 4 and the second polarizer 5 are the same as in Experimental Example 1.

  The positional relationship of each component is 65 mm between the light source 1 and the first polarizer 4, 0 mm between the first polarizer 4 and the transparent plate 3 (that is, a polarizing film is attached to the transparent plate), and the transparent plate 3. Between the second polarizer 5 and the second polarizer 5 is 590 mm, and between the second polarizer 5 and the lens attached to the receiver 2 is 15 mm, and all components are arranged on a straight optical path from bottom to top. It is.

  FIG. 25A is a parallel image captured by turning on an LED having a wavelength of 850 nm of the light source 1, and FIG. 25B is a photograph of an orthogonal image captured by turning on an LED having a wavelength of 850 nm. FIG. 4C is a parallel image captured by lighting an LED having a wavelength of 940 nm, and FIG. 4D is a photograph of an orthogonal image captured by lighting an LED having a wavelength of 940 nm. A white circle in the parallel image indicates the position of the foreign object.

  FIGS. 26A to 26D are difference images obtained by performing density adjustment by linear density conversion on the orthogonal image and taking the difference from the parallel image by the same method as in Experimental Example 1. FIG. The figure (a) is a combination of a parallel image (FIG. 25 (a)) taken by lighting an LED with a wavelength of 850 nm and an orthogonal image (figure 25 (b)) taken with an LED having a wavelength of 850 nm turned on. (Hereinafter referred to as “parallel 850−orthogonal 850”).

  FIGS. 26B, 26C, and 26D are difference images according to combinations of parallel 940-orthogonal 940, parallel 940-orthogonal 850, and parallel 850-orthogonal 940, respectively.

  In the difference image of FIG. 26, the density calculation is performed by adjusting the density value lower limit A and the upper limit B of the linear density conversion of the orthogonal image so that the noise is minimized, and then the density values of all the pixels are uniformly multiplied by 40 times. This is an image obtained by adjusting the density. Table 1 shows the linear density conversion conditions.

  When the difference images in FIG. 26 are compared, in the cases (a) and (b) where the wavelength of the light source is the same between the parallel image capturing and the orthogonal image capturing, the same level of noise due to the flesh remains. On the other hand, in (c), which is a parallel 940-orthogonal 850 combination in which different wavelengths are combined, a lot of noise remains with respect to (a) and (b). On the other hand, in (d), which is a combination of parallel 850 and orthogonal 940, the noise is reduced compared to (a) and (b).

  From this result, it can be seen that the use of light sources having different wavelengths during parallel image capture and orthogonal image capture affects the amount of noise remaining in the difference image. Furthermore, when the combination of imaging and wavelength is switched, the influence on the amount of noise is reversed, and when one shows a noise reduction effect, the other worsens.

  This phenomenon will be described with reference to FIG. FIG. 27 shows the relationship between the transmittance and wavelength of plum meat alone (indicated by a broken line) and plum meat containing a foreign object (indicated by a solid line). As the measurement cell, a quartz cell having a thickness of 5 mm (Short optical path length cell 200-34449 manufactured by Shimadzu Corporation) was used.

  As a sample of ume meat, a sample in which only ume meat was filled in a measurement cell was used. On the other hand, the sample of plum meat containing foreign substances was prepared as follows. That is, red larvae were used as foreign substances, 5 were arranged on the light source side of the measurement cell, and the gaps other than the larvae were filled with plum meat. For the transmittance measurement, an ultraviolet / visible / near infrared spectrophotometer UV-3600 manufactured by Shimadzu Corporation was used, and a wavelength range of 500 nm to 1000 nm was measured at a pitch of 0.5 nm.

  As is clear from FIG. 27, the transmittance difference between ume meat alone and ume meat containing foreign matter is large in the wavelength region near 850 nm, but this transmittance difference is small in the wavelength region near 940 nm. Therefore, when imaging is performed by turning on an LED having a wavelength of 850 nm, a parallel image in which the density difference between plum meat and foreign matter is relatively large and the shadow of the foreign matter is clear is obtained.

  On the other hand, when the LED with a wavelength of 940 nm was turned on and picked up, the difference in concentration between plum meat and foreign matter became relatively small, and the decrease in density due to foreign matter became smaller (that is, the shadow of the foreign matter became thinner). An orthogonal image can be obtained.

  Here, in the food inspection apparatus according to the present invention, by generating a difference image between the parallel image and the orthogonal image, the noise component included in both images is canceled. However, it is preferable that the shadow of the foreign object becomes thinner in the direct image.

Therefore, by using light sources of different wavelengths for parallel image capture and orthogonal image capture, the shadow of a foreign object can be made clearer in the former and thinner in the latter, so that a differential image with a good SN ratio can be obtained. it can.
(Experimental example 12)
The same experiment as in Experiment Example 11 was performed by changing the test object from plum meat to Chinese cabbage kimchi and the foreign object from red larvae to broken pieces of rubber gloves.

  A piece of rubber glove was placed on the transparent plate 3, and Chinese cabbage kimchi was placed thereon. In that case, it was made to cover the whole window part on a transparent board, paying attention so that Kimchi may not overlap as much as possible. The experimental conditions were the same as in Experimental Example 11 except that the numerical values in Table 2 were applied to the linear density conversion conditions.

  FIG. 28 shows a parallel image and an orthogonal image captured by the receiver 2. FIG. 28A is a parallel image captured by lighting an LED having a wavelength of 850 nm, and FIG. 28B is a photograph of an orthogonal image captured by lighting an LED having a wavelength of 850 nm. FIG. 6C is a parallel image captured by turning on an LED having a wavelength of 940 nm, and FIG. 9D is a photograph of an orthogonal image captured by turning on an LED having a wavelength of 940 nm. A white circle in the parallel image indicates the position of the foreign object.

  FIG. 29 shows a difference image in this experimental example. FIGS. 29A to 29D are differential images according to combinations of parallel 850-orthogonal 850, parallel 940-orthogonal 940, parallel 940-orthogonal 850, and parallel 850-orthogonal 940, respectively.

  Also in this experiment example, as in experiment example 11, in FIG. 29A and FIG. 29B which are difference images related to the same wavelength combination, in FIG. A lot of noise remains. On the other hand, in the same figure (d) which is the combination of parallel 850-orthogonal 940, the tendency for noise to decrease is seen.

  FIG. 30 shows the relationship between the transmittance and wavelength of Chinese cabbage kimchi and foreign matter (rubber pieces of rubber gloves). The Chinese cabbage kimchi sample was prepared by putting a Chinese cabbage kimchi cut into a size that can be placed in the spectrophotometer sample holder into a transparent polyethylene bag (Unipack A-4) without using a measurement cell. On the other hand, the foreign material sample was similarly used without using a measurement cell, and was placed in a spectrophotometer with five sheets of rubber gloves cut in layers. Others were measured under the same conditions as in Example 11.

  FIG. 30 is a graph comparing a single object to be inspected (displayed by a broken line) and a single foreign object (displayed by a solid line), unlike FIG. From FIG. 30, it can be seen that the transmittance from a wavelength of 850 nm to 940 nm tends to decrease in the Chinese cabbage kimchi, which is the object to be inspected, but tends to increase in the foreign rubber gloves. This indicates that the difference between the transmittance of Chinese cabbage kimchi alone and the transmittance of Chinese cabbage kimchi with fragments of rubber gloves adhering as foreign matter is larger than the difference at the wavelength of 940 nm. .

  The results of FIGS. 27 and 30 support the mechanism by which the S / N ratio is improved by the food inspection apparatus according to the present embodiment.

  As described above, the food inspection apparatus according to the present embodiment can be widely applied to food in general, and is particularly suitable for detecting foreign substances in agricultural products and processed agricultural products. In selecting the respective wavelengths at the time of capturing the parallel image and the orthogonal image, it is possible to design as a dedicated light source as long as the light transmission characteristics of the inspection object and the foreign object are known.

  However, even if the light transmission characteristics are unknown, an optimal wavelength can be determined using an existing light source capable of selecting a plurality of wavelengths. That is, if the difference in transmittance between an inspection object that does not include foreign matter in one wavelength range of the light source and the inspection target that contains foreign matter differ from the difference in transmittance in other wavelength ranges, these light wavelengths are parallel to these wavelength ranges. In addition, by confirming the combination with orthogonality through experiments, the optimum wavelength can be easily selected.

  In the present embodiment, an LED panel in which two types of LEDs are arranged in a staggered pattern is used as the light source 1. However, in order to irradiate inspection light in different wavelength ranges during the first image capturing and the second image capturing. In addition, a mechanism for selectively arranging two LED panels having different wavelengths at the irradiation position may be added to the configuration of FIG. Alternatively, a method may be adopted in which two LED panels are fixed and the optical path is switched by a movable mirror.

  As described in Embodiments 1 to 4, the food inspection apparatus according to the present invention can target agricultural products, processed agricultural products, and other foods. In addition, the present invention is not limited to this, and if the inspection object has optically non-uniform characteristics, an effect of reducing noise based on the non-uniformity is exhibited under a certain condition.

  In order to easily confirm whether the target inspection object can be applied to the food inspection apparatus according to the present invention, a foreign object is charged in the lowermost layer (light source side) of the inspection object and orthogonal to the parallel image. What is necessary is just to image. If the density of the foreign material portion is reduced in the parallel image, and there is no decrease in the density of the foreign material portion in the orthogonal image, or if the density drop is not clear compared to the parallel image, the food inspection apparatus according to the present invention is used. Applicable.

  The maximum value of the thickness of the inspection object varies depending on conditions such as the transparency of the inspection object, the wavelength of the inspection light, and the type and size of the foreign matter to be detected. Can be determined.

  In addition, since all the polarization direction control means in the above-mentioned embodiment are machine control, it takes time to switch the polarization direction. As an alternative, electric polarization direction control means can also be employed. That is, the polarization direction of linearly polarized light can be changed by 90 ° by controlling the voltage of the liquid crystal panel from which the polarizing plate on the emission side is eliminated as the first polarizer. In this case, the second polarizer can be used while being fixed.

  Incidentally, the TN liquid crystal panel has a polarizing plate attached to the incident side of the liquid crystal cell, and makes incident light linearly polarized light. In the liquid crystal cell, two alignment films having grooves orthogonal to each other are arranged above and below. The liquid crystal molecules sandwiched between them are twisted by 90 ° between the alignment films and aligned. Accordingly, the polarization direction of the polarized light that has passed through the polarizing plate is changed by 90 ° along the twist of the alignment of the liquid crystal molecules while passing between the alignment films.

  Here, when a voltage is applied between the alignment films, the liquid crystal molecules change their alignment in the normal direction of the alignment film, and thus emit while maintaining the direction polarized by the polarizing plate. In a liquid crystal panel, a polarizing plate is usually bonded to the emission side of the liquid crystal cell, and the structure allows the transmission and blocking of polarized light to be controlled by controlling the polarization direction described above.

  In the present invention, since the purpose is to control the polarization direction, a liquid crystal panel from which the output side polarizing plate is removed is used. No mechanism is required for controlling the polarization direction, and rapid control is possible, which contributes to an increase in inspection speed.

  Further, if the transparent plate on which the inspection object is placed is changed to a transparent conveyor belt, the inspection can be continuously performed on the conveyor. Even when mechanical polarization direction control means is used, the inspection speed can be improved by performing inspection on the conveyor using a transparent conveyor belt.

  For example, in Embodiment 1, in order to translate the second polarizer 5 for polarization direction control, the motion of the conveyor is intermittent, and a parallel image and an orthogonal image are not captured while the conveyor is stopped. Image image processing requires image position correction. However, in the second embodiment, since the second polarizer is a rotary type, if an image is taken while rotating at a constant speed, the imaging interval between both images can be shortened, and the interval between images can be reduced without stopping the conveyor. Misalignment does not increase. Therefore, image processing can be performed without correcting the position of the image.

  Further, the relative angle between the transmission axis of the first polarizer 4 and the transmission axis of the second polarizer 5 has been described as 0 ° for the parallel image and 90 ° for the orthogonal image, but the relative angle of the orthogonal image is If it is larger than the relative angle of the parallel image, the effect of the present invention is exhibited under a certain condition.

  That is, if the relative angle of the parallel image is 70 ° or less, the relative angle of the orthogonal image is 50 ° or more, and the relative angle of the orthogonal image is 20 ° or more larger than the relative angle of the parallel image, the noise is reduced and the foreign matter is reduced. You can get an image with.

  Preferably, when the relative angle of the parallel image is 50 ° or less, the relative angle of the orthogonal image is 70 ° or more, and the relative angle of the orthogonal image is increased by 50 ° or more with respect to the relative angle of the parallel image, the noise is eliminated. Since a clearer image with foreign matter is obtained, the accuracy of subsequent image recognition is improved.

DESCRIPTION OF SYMBOLS 1 Light source 2 Receiver 3 Transparent plate 4, 5 Polarizer 6 Linear motor 7 Rail 8 Controller 9 Stepping motor 10 Support plate 11 Motor pulley 12 Pulley belt 51 Y direction polarizing part 52 X direction polarizing part 53, 55 Frame 54 Circular polarizer 55 Frame with pulley 81 Main control unit 82 Polarizer drive unit 83 Receiver control unit 84 Image data storage unit 85 Image processing unit 86 Density adjustment unit 87 Difference calculation unit 88 Noise elimination unit 100 Food inspection device

Claims (7)

A light source for irradiating the inspection object with inspection light;
A receiver that receives the transmitted light and scattered light transmitted through the inspection object and generates image data;
A first polarizer disposed on an optical path between the light source and the inspection object to linearly polarize the inspection light;
A second polarizer disposed on an optical path between the inspection object and the receiver, and linearly polarizes the transmitted light;
Polarization direction control means capable of changing a relative angle between the transmission axis of the first polarizer and the transmission axis of the second polarizer,
The first image having a small relative angle and the second image having a large relative angle are picked up using the receiver, and foreign matter present in the inspection object is detected using the first image and the second image. A food inspection apparatus characterized by: A light source for irradiating the inspection object with inspection light;
A receiver that receives the transmitted light and scattered light transmitted through the inspection object and generates image data;
A first polarizer disposed on an optical path between the light source and the inspection object and converting the inspection light into circularly polarized light;
A second polarizer disposed on an optical path between the inspection object and the receiver, and converts circularly polarized light into linearly polarized light and transmits the second polarizer;
A polarization direction control unit capable of switching a rotation direction of circularly polarized light converted by the first polarizer or a rotation direction of circularly polarized light transmitted by the second polarizer,
A first direction in which the rotation direction of the circularly polarized light converted by the first polarizer and the rotation direction of the circularly polarized light transmitted by the second polarizer are the same direction as the first image using the receiver is opposite to the first image. A food inspection apparatus characterized by capturing two images and detecting foreign matter present in the inspection object using the first image and the second image.   The food inspection apparatus according to claim 1, wherein the light source has a different wavelength of inspection light to be irradiated when the first image is captured and when the second image is captured.   The inspection light emitted when the second image is taken is the difference in light transmittance between the inspection object that does not contain foreign matter and the inspection subject that contains foreign matter at the wavelength of the inspection light emitted when the first image is taken. The food inspection apparatus according to claim 3, wherein the food inspection apparatus is large relative to the difference in transmittance at a wavelength of.   The food inspection apparatus according to any one of claims 1 to 4, further comprising density adjusting means for performing density adjustment for increasing the contrast of the second image and / or decreasing the contrast of the first image.   The food inspection apparatus according to claim 5, further comprising an image processing unit that generates a difference image between the first image and the second image that have undergone density adjustment by the density adjustment unit.   The food inspection apparatus according to claim 6, further comprising an image recognition unit that determines presence / absence of a foreign substance in the inspection object based on the difference image and / or specifies a position of the foreign substance.