JP2019074324A - Inspection method and device - Google Patents

Abstract

To provide an inspection method and device capable of definitely distinguishing ingredients or the like similar in color.SOLUTION: An inspection object is imaged by a first imaging portion 30 which captures an image of a first wavelength and a second imaging portion 40 which captures an image of a second wavelength. An inspection image is created which is composed of a normalized spectral index of each of pixels computed from a spectral intensity of each of the pixels of the first captured image of the inspection object captured by the first imaging portion 30 and a spectral intensity of each of the pixels of the second captured image of the inspection object captured by the second imaging portion 40. A state of the inspection object is inspected on the basis of the inspection image.SELECTED DRAWING: Figure 1

Description

  The present invention relates to an inspection method and apparatus, and more particularly to an inspection method and apparatus for inspection using a color camera.

  Instant food products such as cup noodles are obtained by adding ingredients called kaken to main ingredients such as dry noodles and green noodles and putting them in a food container such as a cup-shaped heat-resistant and water-resistant container.

  In instant food products such as cup noodles, the amount of main ingredients such as dried noodles and green noodles, and the type and number of ingredients called "flaming" are strictly defined. Therefore, it is necessary to inspect the packaging condition before shipping the instant food. In the inspection of the filling state of such instant food, the packaging state is imaged by a color camera, and the types and the number of main materials and ingredients are identified from the colors of the main materials and ingredients.

  However, depending on the type of main material and ingredients, it may be difficult to identify the color camera image. As an apparatus for extracting only a specific image from images captured by a color camera, an optical system including a plurality of imaging means having a photoelectric conversion function, and a lens for distributing the same optical image to the plurality of imaging means Between a plurality of optical band pass filters having different transmission wavelength bands provided between an imaging means and a lens, and between video signals obtained by photoelectrically converting images of different wavelengths by a plurality of imaging means An object extraction type TV camera is known that includes an image operation unit that calculates the level difference of each pixel corresponding to the same spatial coordinate and outputs a video signal having a value proportional to its absolute value (Patent Document 1 1).

Japanese Patent Application Laid-Open No. 2002-077897

  However, conventional object extraction type TV cameras have difficulty in clearly identifying ingredients of similar colors.

  An object of the present invention is to provide an inspection method and apparatus capable of clearly identifying ingredients of similar colors and the like.

  In the inspection method according to the present invention, the inspection target is imaged by a first imaging unit that captures an image of a first wavelength, and a second imaging unit that captures an image of a second wavelength, and the first imaging unit Calculated from the spectral intensity of each pixel of the first captured image of the inspection object captured by the second imaging unit and the spectral intensity of each pixel of the second captured image of the inspection object captured by the second imaging unit An inspection method comprising: generating an inspection image composed of a normalized spectral index of each pixel; and inspecting the state of the inspection object based on the inspection image.

  In the inspection method described above, the spectral intensity for the first wavelength and the first inclination of the spectrum due to the spectral intensity of the second wavelength with respect to the first portion included in the inspection object are included in the inspection object. The first wavelength and the second wavelength are selected to be different from the spectral intensity for the first wavelength and the second inclination of the spectrum due to the spectral intensity of the second wavelength for the part 2; The first portion and the second portion included in the inspection target may be identified based on the difference in the normalized spectral index of each pixel of the inspection image.

  In the inspection method described above, spectral intensity of the first wavelength and spectral intensity of the second wavelength with respect to the first portion and the third portion other than the second portion included in the inspection object The first wavelength and the second wavelength may be selected such that a third slope is different from the first slope and / or the second slope.

  In the inspection method described above, the inspection object is cup noodles, the first part included in the inspection object is a roller char, and the first part included in the inspection object is powder. The first wavelength may be 685 nm and the second wavelength may be 670 nm.

  An inspection apparatus according to the present invention includes a first imaging unit configured to capture an image of a first wavelength, a second imaging unit configured to capture an image of a second wavelength, and an inspection captured by the first imaging unit. Normalization of each pixel calculated from the spectral intensity of each pixel of the first captured image of interest and the spectral intensity of each pixel of the second captured image of the inspection object captured by the second imaging unit The inspection image generating apparatus is characterized by generating an inspection image constituted by a spectral index, and an inspection unit which inspects a state of the inspection object based on the inspection image.

  In the inspection apparatus described above, the spectral intensity for the first wavelength and the first inclination of the spectrum by the spectral intensity of the second wavelength with respect to the first portion included in the inspection object are included in the inspection object. The first wavelength and the second wavelength are selected to be different from the spectral intensity for the first wavelength and the second inclination of the spectrum due to the spectral intensity for the second wavelength with respect to the two parts. The inspection unit may identify the first portion and the second portion included in the inspection target based on the difference in the normalized spectral index of each pixel of the inspection image.

  In the inspection apparatus described above, the spectrum intensity of the first wavelength and the spectrum intensity of the second wavelength with respect to the first portion and the third portion other than the second portion included in the inspection object The first wavelength and the second wavelength may be selected such that a third slope is different from the first slope and / or the second slope.

  In the inspection apparatus described above, the inspection object is cup noodles, the first part included in the inspection object is a roller char, and the first part included in the inspection object is powder. The first wavelength may be 685 nm and the second wavelength may be 670 nm.

  As described above, according to the present invention, the inspection target is imaged by the first imaging unit that captures an image of the first wavelength, and the second imaging unit that captures an image of the second wavelength, and the first Calculated from the spectral intensity of each pixel of the first captured image of the inspection object captured by the imaging unit and the spectral intensity of each pixel of the second captured image of the inspection object captured by the second imaging unit Since the inspection image composed of the normalized spectral index of each pixel is generated and the condition of the inspection object is inspected based on the inspection image, ingredients and the like having similar colors can be clearly identified. .

FIG. 1 is a view showing an inspection apparatus according to an embodiment of the present invention. FIG. 2 is a diagram showing the details of the camera of the inspection apparatus according to an embodiment of the present invention. FIG. 3 is an explanatory view (part 1) of the detection principle of the inspection apparatus according to the embodiment of the present invention. FIG. 4 is an explanatory view (part 2) of the detection principle of the inspection apparatus according to the embodiment of the present invention. FIG. 5 is an explanatory view (part 3) of the detection principle of the inspection apparatus according to the embodiment of the present invention. FIG. 6 is an explanatory view (part 1) of program processing of the inspection of the inspection apparatus according to the embodiment of the present invention. FIG. 7 is an explanatory view (part 2) of program processing of the inspection of the inspection apparatus according to the embodiment of the present invention. FIG. 8 is a flowchart of program processing of the inspection of the inspection apparatus according to an embodiment of the present invention. FIG. 9 is a diagram showing an NDSI inspection image in the program processing of the inspection of the inspection apparatus according to the embodiment of the present invention. FIG. 10 is an explanatory diagram of a process of masking an NDSI image using a raw image and a luminance correction image in a program process of an examination of an examination apparatus according to an embodiment of the present invention. FIG. 11 is an explanatory diagram of a process of determining the area of the roller / char in the program process of the inspection of the inspection apparatus according to the embodiment of the present invention. FIG. 12 is an explanatory diagram of a process of determining the area of the curry-like powder in the program process of the inspection of the inspection apparatus according to the embodiment of the present invention. FIG. 13 is a view showing a determination image in the program processing of the inspection of the inspection apparatus according to the embodiment of the present invention.

[One embodiment]
An inspection apparatus according to an embodiment of the present invention will be described using FIGS. 1 to 8.

(Subject to inspection)
The inspection apparatus of the present embodiment is an apparatus for inspecting whether the contents of cup noodles are filled in a defined state. In cup noodles, ingredients such as dried noodles and raw type noodles are stored in a cup of a packaging container, as well as ingredients called kaen. As well as the amount of main material, the type of ingredients, the number of each ingredient, the size, and the like are strictly defined as being accommodated in one cup.

  For example, in the case of cup noodles of curry flavor, dry noodles as a main ingredient are stored in a cup of a packaging container, and kolo char, potatoes, carrots, green onions, etc. are stored as ingredients called kaen. Koro Char is a new ingredient that has come to replace Dice Mince (Mystery Meat).

  In the final step of producing cup noodles, the inspection apparatus according to the present embodiment captures an image of the state before all the materials are stored in the cup and the opening at the upper end of the cup is sealed by the lid material from above The image is analyzed to check if the main material and ingredients are stored as specified.

  If it is judged by the inspection apparatus of the present embodiment that the main material and the ingredients are stored as prescribed, the opening at the upper end of the cup is sealed with a lid to complete a product as a cup noodle. If it is judged by the inspection device of the present embodiment that the main material and the ingredients are not accommodated as specified, the cup is rejected and does not become a product.

(Inspection device)
The inspection apparatus of the present embodiment will be described with reference to FIG.

  The inspection apparatus 20 of the present embodiment is provided with a work table 22 on which a work (cup noodle) 10 is placed. The workpiece (cup noodle) 10 to be inspected is conveyed by a belt conveyor or the like (not shown) and passes a predetermined position of the workpiece table 22.

  An illumination umbrella 24 for illuminating the work (cup noodle) 10 uniformly from above is provided above the work table 22. A strobe light bulb (not shown) is provided in the lighting umbrella 24, and a strobe power supply 26 is connected to the strobe light bulb (not shown).

  A beam splitter 28 is provided on the illumination umbrella 24. The beam splitter 28 splits the light from the work (cup noodle) 10 into two.

  A first camera 30 is provided in the direction in which light from the work (cup noodles) 10 is split in two by the beam splitter 28. The first camera 30 is provided with a first imaging element 32, and further, a first band pass filter 34 for transmitting light of a specific first wavelength, and a first imaging element 32 for transmitting the transmitted light. And a first lens 36 for collecting light.

  A second camera 40 is provided in the orthogonal direction of the light from the work (cup noodle) 10 divided into two by the beam splitter 28. The second camera 40 is provided with a second imaging element 42, and further, a second band pass filter 44 for transmitting light of a specific second wavelength, and a second imaging element 42 for transmitting the transmitted light. And a second lens 46 for collecting light.

  The light of the first wavelength set by the first camera 30 and the light of the second wavelength set by the second camera 40 will be described later.

  The inspection apparatus 20 further includes an image memory 50 for storing captured images of the first camera 30 and the second camera 40, and an operation unit 52 for performing an operation described later on the captured image stored in the image memory 50, A strobe power supply 26, a first camera 30, a second camera 40, an image memory 50, and a control unit 54 for controlling the operation of the arithmetic unit 52 are provided.

  The image memory 50 stores an image captured by the first imaging element 32 of the first camera 30 and an image captured by the second imaging element 42 of the second camera 40 based on a control signal from the control unit 54. Do.

  The calculation unit 52 performs calculations to be described later using various captured images of the first camera 30 stored in the image memory 50 and various captured images of the second camera 40.

  The control unit 54 controls the strobe power supply 26, the first camera 30, the second camera 40, the image memory 50, and the calculation unit 52 to operate in synchronization with each other.

  The image memory 50, the operation unit 52, and the control unit 54 function as an inspection unit that inspects the workpiece (cup noodle) 10 that is an inspection object.

(Camera of inspection device)
The details of the first camera 30 and the second camera 40 of the inspection apparatus 20 of the present embodiment will be described with reference to FIG.

  The first camera 30 and the fourth camera 40 are cameras called multi-spectral cameras or hyper-spectral cameras capable of imaging with resolution of wavelengths.

  As shown in FIG. 2A, electromagnetic waves are called radio waves, microwaves, infrared rays, visible light, ultraviolet rays, X-rays, and gamma rays depending on their wavelengths. The wavelength range of 800 nm to 400 nm is an electromagnetic wave that is called visible light and can be viewed by people.

  As shown in FIG. 2B, the multispectral camera or hyperspectral camera has an ability to separate and acquire an electromagnetic wave from infrared rays to ultraviolet rays, for example, an electromagnetic wave with a wavelength of 350 nm to 1100 nm, to 151.

  The inspection apparatus 20 of the present embodiment uses a captured image of light of wavelength 685 nm (FIG. 2 (c1)) and a captured image of light of wavelength 670 nm (FIG. 2 (c2)). The reason for selecting these wavelengths will be described later.

  With the first band-pass filter 34 in the first camera 30 as a band filter that transmits the wavelength 685 nm, in the first camera 30, the first image sensor 32 acquires a captured image of light having a wavelength of 685 nm.

  The second band pass filter 44 in the second camera 40 is a band filter that transmits a wavelength of 670 nm, and in the second camera 40, a second image sensor 42 acquires a captured image of light with a wavelength of 670 nm.

(Detection principle)
The detection principle of the inspection apparatus 20 of the present embodiment will be described using FIGS. 3 to 5.

  The inspection principle will be described by way of an example of a curry-flavored cup noodle to be inspected by the inspection apparatus 20 of the present embodiment.

  FIG. 3A shows a captured image of a curry-flavored cup noodle. In curry-flavored cup noodles, kolo char is used as an ingredient for meat. Koro Char is a new meat ingredient that has come to replace Diceminch. In cup noodles with curry taste, it is important to check the presence or absence and number of rolls and chars.

  However, as shown in FIG. 3 (a), Koro Char is brown similar to the color of the curry-flavored powder sprinkled on the top of the cup noodles, and can be identified even with a multi-spectral camera capable of identifying 151 colors. It is difficult to do.

  The curry-flavored flour is mainly composed of plant-derived components, whereas Koro-char is a meat ingredient and contains almost no plant-derived components. The inventors of the present application paid attention to this point and intensively studied a method of discriminating between curry-flavored powder and kolo char. As a result, it was found that there is a difference between the spectrum of curry-flavored powder and the spectrum of kolo char due to the presence or absence of plant-derived components.

  FIG. 3 (b) shows the spectrum of curry-flavored powder and kolo char. The spectrum of curry-like powder and the spectrum of kolo char have almost the same spectral shape over the wavelength range of 400 nm to 1000 nm. However, the spectral intensity of the curry-like powder is reduced in the wavelength range of 600 nm to 700 nm, and more specifically, in the wavelength range of 650 nm to 670 nm. This is attributable to the plant-derived component.

  FIG. 3 (c) shows spectra of curry-flavored powder and kolo char in the wavelength range of 620 nm to 700 nm. As a result of examining the spectrum shown in FIG. 3C, the inventors of the present application have conceived the idea of focusing on the spectral intensities at a wavelength of 670 nm (λ2) and a wavelength of 685 nm (λ1).

  In the wavelength range of 620 nm to 700 nm, the spectrum of curry-flavored powder and the spectrum of colo char tend to increase in spectral intensity as the wavelength increases. However, while the spectral intensity of kolo char is gradually increased, the spectral intensity of curry-flavored powder is low at around a wavelength of 670 nm due to the plant-derived component.

  The spectral intensity of the curry powder is low at around 670 nm, so the spectral intensity of the 670 nm (λ2) spectral intensity of the colo char and the spectral intensity of the 685 nm (λ1) spectrum are relatively loose, and curry The inclination of the spectrum due to the spectral intensity at a wavelength of 670 nm (λ2) of the taste powder and the spectral intensity at a wavelength of 685 nm (λ1) is relatively sharp. In the present embodiment, koro char and curry powder are identified based on the difference in the slope of the spectrum.

  In addition, since the spectral intensity of curry-like powder is low at around 670 nm, as shown in FIG. 4A, the spectral intensity of the wavelength 650 (λ3) of Coro char and the spectrum of the wavelength 670 nm (λ2) The inclination of the spectrum due to the intensity is relatively steep, and the inclination of the spectrum due to the spectral intensity of the wavelength 650 nm (λ3) of the curry powder and the spectral intensity of the wavelength 670 nm (λ2) is relatively gentle. It is conceivable to distinguish Koro char and curry powder based on the difference in the slope of the spectrum.

  However, discrimination based on the difference between the spectral intensity of the wavelength 650 (λ3) and the inclination of the spectrum of the wavelength 670 nm (λ2) has a possibility of false detection.

  Other ingredients and cups themselves are also captured in the captured image of the cup noodle in addition to the flour of the kolo char and curry flavor. As shown in FIG. 4 (b), the spectral intensity of the other ingredients is at a relative wavelength of 650 nm (.lambda.3) and 670 nm (.lambda.2). The inclination of the spectrum by the spectral intensity of 650 nm (λ3) and the spectral intensity of wavelength 670 nm (λ2) is also relatively gentle. It is difficult to distinguish the inclination of the spectrum of potatoes and cups from the inclination of the spectrum based on the spectral intensity of the wavelength 650 nm (λ3) of the curry powder and the spectral intensity of the wavelength 670 nm (λ2). There is a possibility that the wavelength of 670 nm (λ2) and the wavelength of 685 nm (λ1) are more desirable.

  The principle of detecting the difference in the slope of the spectrum will be described with reference to FIG.

  Assuming that the spectral intensity of each pixel of the captured image of wavelength 685 nm (λ1) is H1 and the spectral intensity of each pixel of the captured image of wavelength 670 nm (λ2) is H2, the normalized spectral index (NDSI) is It becomes a formula.

NDSI = (H1-H2) x C / (H1 + H2)
However, C is a constant.

  This NDSI shows the inclination of the spectrum by the spectral intensity of each pixel of a captured image, and the spectral intensity of wavelength 670 nm ((lambda) 2).

  FIG. 5 (a) is an image at a spectral intensity H1 of a wavelength 685 nm (λ1) of each pixel of the captured image, and FIG. 5 (b) is an image of a spectral intensity H2 at a wavelength 670 nm (λ2) of each pixel of the captured image FIG. 5C is an image of NDSI calculated from the spectral intensity H1 at a wavelength of 685 nm (λ1) and the spectral intensity H2 at a wavelength of 670 nm (λ2) of each pixel of the captured image. The image of NDSI shown in FIG. 5C shows the inclination of the spectrum due to the spectral intensity of each pixel of the captured image and the spectral intensity at a wavelength of 670 nm (λ 2).

  Experimentally grasp the range D1 of the NDSI value of the relatively gentle slope of the spectrum by the spectral intensity of wavelength 670 nm (λ 2) and the spectral intensity of wavelength 685 nm (λ 1) of Koro Char, and the value of NDSI is the range If it is a value of D1, it is judged as an image of Koro Char.

  Similarly, the spectrum intensity of the curry flavor powder at wavelength 670 nm (λ 2) and the spectrum intensity at wavelength 685 nm (λ 1) of the relatively steep slope of the spectrum NDSI value range D2 of the value of NDSI is experimentally grasped. If the value is the value of the range D2, it is judged as a curry-like powder image.

  Specifically, assuming that the constant C is 1500, the range D1 of the NDSI values of the relatively gentle slopes of the spectrum by the spectral intensity of the wavelength 670 nm (λ 2) of the coro char and the spectral intensity of the wavelength 685 nm (λ 1) , The spectrum intensity of the curry-flavored powder at a wavelength of 670 nm (λ2) and the spectrum intensity of the spectrum at a wavelength of 685 nm (λ1). , Values in the range of about 128-204.

  In the image shown in FIG. 5D, the inclination of the spectrum of each pixel is identified by the predetermined range D1 of the NDSI value of the roller char described above and the predetermined range D2 of the NDSI value of the curry powder. is there. If noise is removed from the image shown in FIG. 5 (d), the image shown in FIG. 5 (e) is obtained. Thereby, the image of Koro char and the image of the curry-like powder in the captured image of the cup noodle are visually identified.

(Program processing of examination)
The program processing of the inspection of the inspection apparatus 20 of the present embodiment will be described using FIGS. 6 to 8.

  Before the inspection of the workpiece (cup noodle) 10 which is the inspection object, a preparatory work is performed to correct the difference in luminance of the captured image by the first camera 30 and the second camera 40 (step S01).

  In the first camera 30, the first imaging element 32 acquires a captured image of light having a wavelength of 685 nm. In the second camera 40, the second imaging element 42 acquires a captured image of light having a wavelength of 670 nm.

  A calibration white board 60 as shown in FIG. 7A is placed on the work table 22, and the calibration white board 60 is imaged by the first camera 30 and the second camera 40 of the inspection apparatus 20. The white calibration plate 60 is a calibration standard plate for a ceramic spectrophotometer, and has a luminance ratio of 90% or more over substantially the entire visible light region. White board images 61 and 62 which are captured images of the white board for calibration are acquired by the first camera 30 and the second camera 40 and stored in the image memory 50.

  In addition, preparation work is performed to correct the difference between the coordinates of the captured image by the first camera 30 and the second camera 40 (step S02). The calibration chess board 70 as shown in FIG. 7B is placed on the work table 22, and the calibration chess board 70 is imaged by the first camera 30 and the second camera 40 of the inspection apparatus 20. The calibration chess board 70 is a calibration standard board for a ceramic spectrophotometer, is a black and white chess board pattern, and a black portion has a luminance ratio of 20% or less over substantially the entire visible light region. The white portion has a luminance rate of 90% or more. Chessboard images 71 and 72 which are captured images of the calibration chessboard 70 are acquired by the first camera 30 and the second camera 40 and stored in the image memory 50.

  Next, in order to perform coordinate correction processing to match the coordinates of the luminance correction image 92 by the second camera 40 with the coordinates of the luminance correction image 91 by the first camera 30, the arithmetic unit 52 performs the correction processing by the first camera 30. The calibration data 73 for coordinate correction is generated using the chessboard image 71 and the chessboard image 72 by the second camera 40 (step S03). The calibration data 73 is stored in the image memory 50.

  Note that calibration data may be generated to perform coordinate correction processing in which the coordinates of the luminance correction image 91 by the first camera 30 are aligned with the coordinates of the luminance correction image 92 by the second camera 40.

  Subsequently, the inspection on the workpiece (cup noodle) 10 to be inspected is started. The workpiece (cup noodle) 10 to be inspected is transported by a transport mechanism (not shown) such as a belt conveyor, and sequentially passes a predetermined position of the workpiece table 22. When it is detected by a sensor (not shown) that the work (cup noodle) 10 has reached a predetermined position on the work base 22, the transport mechanism (not shown) outputs a trigger signal.

  Control unit 54 detects the output of the trigger signal from the transport mechanism (not shown) (step S04).

  When detecting the trigger signal, the control unit 54 controls the strobe power supply 26 to make the strobe light bulb (not shown) in the lighting umbrella 24 emit light, and at the same time control the first camera 30 and the second camera 40 Then, the work (cup noodle) 10 is imaged (step S05). Thereby, the raw image 81 which is a captured image of the work (cup noodle) 10 from the first camera 30 and the raw image 82 which is a captured image of the work (cup noodle) 10 from the second camera 40 are obtained It is stored in the memory 50.

  Next, using the white board images 61 and 62, chessboard images 71 and 72, and raw images 81 and 82 stored in the image memory 50, the operation unit 52 performs an operation based on the above-described detection principle to obtain a judgment image 120. Generate

  First, the brightness correction processing for generating the brightness correction images 91 and 92 from the white plate images 61 and 62 and the raw images 81 and 82 is executed (step S06). From the value R1 of each pixel of the raw image 81 of the first camera 30 and the value W1 of each pixel of the white plate image 61 of the first camera 30 by the calculation unit 52, Calculate the value B1.

B1 = R1 ÷ W1 × C
However, C is a constant, for example, 100.

  From the value R2 of each pixel of the raw image 82 of the second camera 40 and the value W2 of each pixel of the white plate image 62 of the second camera 40 by the calculation unit 52, Calculate the value B2.

B2 = R2 ÷ W2 × C
However, C is a constant, for example, 100.

  In the raw images 81 and 82, for example, there is a fluctuation in brightness that the central part is bright and the outer peripheral part is dark, but the degree of the fluctuation and the place of the fluctuation are between the first camera 30 and the second camera 40. not the same. The luminance correction step corrects the fluctuation due to the fluctuation of the brightness in each captured image and corrects the difference in the fluctuation of the brightness by the first camera 30 and the second camera 30.

  The calculated luminance corrected images 91 and 92 are stored in the image memory 50.

  Next, coordinate correction processing for correcting coordinate values between the luminance correction image 91 and the luminance correction image 92 stored in the image memory 50 is executed (step S07). In order to perform coordinate correction processing to match the coordinates of the luminance correction image 92 by the second camera 40 with the coordinates of the luminance correction image 91 by the first camera 30, the second stored in the image memory 50 by the arithmetic unit 52. The calibration data 73 is applied to the luminance correction image 92 by the camera 40 to generate a coordinate correction image 102 of the luminance correction image 92 by the second camera 40. The coordinate correction image 102 is stored in the image memory 50.

  Next, using the luminance correction image 91 by the first camera 30 and the coordinate correction image 102 by the second camera 40, the Normalized Difference Spectral Index (NDSI) of each pixel is calculated, and the NDSI is calculated. An inspection image 110 is generated (step S08).

  Assuming that the spectral intensity of each pixel of the luminance-corrected image 91 by the first camera 30 is H1 and the spectral intensity of each pixel of the coordinate-corrected image 102 by the second camera 40 is H2, Normalized Difference Spectral Index: NDSI) becomes the following equation.

NDSI = (H1-H2) x C / (H1 + H2)
However, C is a constant and is 1500 in this embodiment.

  The NDSI inspection image 110 is stored in the image memory 50.

  Next, a discrimination process of discriminating the collo char and the curry-like powder from the NDSI inspection image 110 stored in the image memory 50 is started (step S09).

  For example, based on whether the value of each pixel of the NDSI inspection image 110 is within the predetermined range D1 for the roller char or the predetermined range D2 for the curry flavor (c) powder, each pixel is Determine whether it is char or curry-like powder, and generate a judgment image 120 in which the portion of colo char and the portion of curry-like powder are color-coded (step S10).

  The details of the process of step S10 will be described with reference to FIGS.

  A judgment image 120 in which the portion of the roller char and the portion of the powder are colored simply based on whether the value of each pixel of the NDSI inspection image 110 is within the predetermined range D1 or within the predetermined range D2. In some cases, background portions outside of the cup noodles of the raw images 81 and 82, an image of the cup noodle cup itself, and an image of other ingredients, for example, an image of a potato may be colored.

  In the present embodiment, in the raw image 81 of the first camera 30, each pixel having an intensity equal to or less than a predetermined threshold value is erased and is not detected. This can prevent the background portion of the cup noodle outside from being colored in red or blue.

  Further, in the present embodiment, in the luminance correction image 91 of the first camera 30, each pixel having an intensity equal to or higher than a predetermined threshold value is erased and is not detected. This can prevent the image of the cup noodle itself and the image of other ingredients such as potatoes from being colored in red or blue.

  FIG. 9 shows an NDSI examination image. FIG. 9 (a1) shows the NDSI inspection image 110, and FIG. 9 (a2) is a graph showing the NDSI value along the line AA 'of the NDSI inspection image 110. As shown in FIG.

  As shown in FIG. 9, the NDSI of each pixel of the NDSI image 110 has a value of 0 to 250. Based on whether the NDSI value of each pixel of the NDSI inspection image 110 is within the predetermined range D1 or within the predetermined range D2, the portion of the roller char and the portion of the powder are determined.

  FIG. 10 is an explanatory diagram of a process of masking the NDSI image 110 using the raw image 81 of the first camera 30 and the luminance correction image 91 of the first camera 30.

  FIG. 10A1 is a view showing a raw image 81, and FIG. 10A2 is a graph showing a mask along line A-A 'of the raw image 81. As shown in FIG. Each pixel of the raw image 81 of FIG. 10 (a1) has a value in the range of 0 to 255 based on the brightness. The mask in FIG. 10A2 is a mask for removing a region in which the value of each pixel of the raw image 81 is 40 or less. An area in which the value of each pixel of the raw image 81 is 40 or less is “0”, and an area in which the value of each pixel is larger than 40 is “1”. Areas of "0" are masked and areas of "1" are not masked.

  FIG. 10 (b1) is a view showing the luminance correction image 91, and FIG. 10 (b2) is a graph showing a mask along the line AA 'of the luminance correction image 91. As shown in FIG. Each pixel of the luminance correction image 91 in FIG. 10B1 has a value in the range of 0 to 255 based on the brightness. The mask in FIG. 10B is a mask for removing an area in which the value of each pixel of the luminance correction image 91 is 153 or more. An area in which the value of each pixel of the luminance correction image 91 is 153 or more is “0”, and an area in which the value of each pixel is smaller than 153 is “1”. Areas of "0" are masked and areas of "1" are not masked.

  10 (c1) and 10 (c2) show a state in which the NDSI inspection image 110 shown in FIG. 9 is masked by the mask of FIG. 10 (a2) and the mask of FIG. 10 (b2). FIG. 10 (c 1) shows the masked NDSI inspection image 110, and FIG. 10 (c 2) is a graph showing the NDSI value along the line AA ′ of the NDSI inspection image 110.

  From the original NDSI inspection image 110 shown in FIG. 9 in the NDSI inspection image 110 shown in FIG. 10 (c1) (c2), other components such as a background portion outside the cup noodles, an image of the cup noodle cup itself, a potato It can be seen that the part of the material image has been removed.

  In the present embodiment, based on whether the value of each pixel of the NDSI inspection image 110 shown in FIGS. 10C1 and 10C2 is within the predetermined range D1 or within the predetermined range D2, Determine the part of the and the part of the powder.

  FIG. 11 is an explanatory diagram of a process of determining the area of the roller char from the NDSI inspection image 110 shown in FIGS. 10 (c1) and 10 (c2).

  FIG. 11 (a1) shows the masked NDSI inspection image 110, and FIG. 11 (a2) is a graph showing the NDSI value along the line AA 'of the NDSI inspection image 110.

  When the range D1 (range of 0 to 123) of the NDSI value is extracted from the masked NDSI inspection image 110 and the extraction result is binarized, it becomes as shown in FIG. 11 (b1) (b2). FIG. 11 (b 1) shows the binarized NDSI inspection image 110, and FIG. 11 (b 2) is a graph showing the NDSI value along the line AA ′ of the NDSI inspection image 110.

  When the binarized NDSI inspection image 110 shown in FIG. 11 (b1) (b2) is subjected to expansion processing to replace the pixel “1” with the maximum luminance of the neighboring pixels, FIG. It becomes as shown in c2). FIG. 11 (c 1) shows the expanded NDSI inspection image 110, and FIG. 11 (c 2) is a graph showing the NDSI value along the line AA ′ of the NDSI inspection image 110. Details of the expansion processing will be described later.

  When the masked NDSI inspection image 110 is extracted and binarized by extracting the range D1 (range of 0 to 123) of the NDSI value and further dilation processing, as shown in FIG. 11 (c1) (c2), Area is correctly extracted.

  FIG. 12 is an explanatory diagram of a process of determining the region of the curry-flavored powder from the NDSI inspection image 110 shown in (c1) and (c2) in FIG.

  In the process of FIG. 12, the area of the roller char extracted in the process of FIG. 11 is removed from the NDSI inspection image 110 shown in FIGS. 10 (c1) and 10 (c2). 12 (a1) and 12 (a2) show the NDSI inspection image 110 from which the area of the roller char has been removed. FIG. 12 (a1) shows the NDSI inspection image 110 from which the area of the roller char has been removed, and FIG. 12 (a2) is a graph showing the NDSI value along the line A-A 'of the NDSI inspection image 110. .

  When the range D2 (range of 128 to 204) of the NDSI value is extracted from the NDSI inspection image 110 of FIG. 12 (a1) (a2) and the extraction result is binarized, it is shown in FIG. 12 (b1) (b2) It will be shown. FIG. 12 (b 1) shows the binarized NDSI inspection image 110, and FIG. 12 (b 2) is a graph showing the NDSI value along the line A-A ′ of the NDSI inspection image 110.

  With regard to the binarized NDSI inspection image 110 shown in FIGS. 12 (b1) and 12 (b2), expansion processing to replace the pixel of “1” with the maximum luminance of the neighboring pixel and replacing with the minimum luminance of the neighboring pixel When the contraction process is performed, it is as shown in FIG. 12 (c1) (c2). FIG. 12 (c 1) shows the expanded and contracted NDSI inspection image 110, and FIG. 12 (c 2) is a graph showing the NDSI value along line A-A ′ of the NDSI inspection image 110.

  As described above, expansion processing is performed on the binarized NDSI inspection image 110 shown in FIGS. 11 (b1) and (b2), and the binarized NDSI inspection image shown in FIGS. 12 (b1) and (b2) The expansion process and the contraction process are performed on 110. Here, the details of the expansion process and the contraction process will be described.

  In expansion processing and contraction processing, processing is generally performed on a binarized black-and-white image. The process of replacing white pixels as white as one pixel at the periphery of the pixel of interest is referred to as dilation process, and the process of replacing black pixels as black as one pixel at the periphery as black is referred to as erosion process. When processing is performed on a gray image instead of a binarized image as in the present embodiment, in the case of dilation processing, the maximum luminance value in the vicinity of the target pixel is replaced with the luminance value of the target pixel. In the case of contraction processing, the gray image is processed by replacing it with the minimum luminance value. The expansion process and the contraction process may be performed alone, but the expansion process and the contraction process may be repeated to perform the process. By using expansion processing and contraction processing, inspection of defects such as chipping, breakage, dust, and dust can be performed, and characters and lines of a thin linear pattern can be extracted.

  If the NDSI inspection image 110 shown in FIG. 12 (a1) (a2) is binarized by extracting the range D2 (range 128 to 204) of the NDSI value, and further expansion processing and contraction processing, FIG. 12 (c1) (c1) As shown in c2), the area of curry-like powder is correctly extracted.

  FIG. 13 shows a judgment image obtained by superposing the extraction result of the region of koro char shown in FIG. 11 (c1) (c2) and the extraction result of the region of curry taste powder shown in FIG. 12 (c1) (c2) FIG. FIG. 13 (a1) shows a judgment image 120 on which the extraction results are superimposed, and FIG. 13 (a2) is a graph showing the extraction results along the line AA 'of the judgment image 120. FIG. The filled area is the area of the roll, and the hatched area is the area of the curry-like powder.

  Thus, when the process of step S10 is completed, next, based on the determination image 120 stored in the image memory 50, it is determined whether or not the ingredients of the work (cup noodle) 10 are in accordance with a predetermined rule. (Step S11). This determination may be made by displaying the determination image 120 on a monitor (not shown), and the inspector may make the determination visually or by counting the number of pixels.

  When the inspector judges visually, it is visually judged whether the filled area and the hatched area have a predetermined area or not.

  When judging by counting the number of pixels, the number of pixels of the area of the roller image in the judgment image 120, that is, the filled area is counted, and it is judged whether or not it is less than a predetermined value. Further, the number of pixels of the curry-like powder region in the determination image 120, that is, the hatched region is counted, and it is determined whether or not the number is equal to or less than a predetermined value.

  As a result of the determination in step S11, if it is determined that the flour of curry or curry taste is not stored as prescribed in the work (cup noodle) 10, the work (cup noodle) 10 is rejected (step S12) ), Shift to the detection process of the next work (cup noodle) 10.

  If, as a result of the determination in step S10, it is determined that the flour of curry or curry taste is stored as prescribed in the work (cup noodle) 10, the work (cup noodle) 10 is not rejected next, without being rejected. It moves to the detection processing of work (cup noodle) 10 of.

[Modified embodiment]
Not limited to the above embodiment, various modifications are possible. The inspection object, the material to be inspected, the numerical values of the various ranges, the various dimensions, and the materials of the respective members in the above-described embodiment are merely examples, and the present invention is not limited thereto.

10 ... Work (cup noodles)
20 ... inspection device 22 ... work stand 24 ... umbrella for illumination 26 ... strobe power supply 28 ... beam splitter 30 ... first camera 32 ... first imaging element 34 ... first band bus filter 36 ... first lens 40 ... Second camera 42 Second imaging element 44 Second band-pass filter 46 Second lens 50 Image memory 52 Operation unit 54 Control unit 60 White board 61 for calibration 60 White board image 70 Calibration chess board 71, 72 ... Chess board image 73 ... Calibration data 81, 82 ... Raw image 91, 92 ... Brightness correction image 102 ... Coordinate correction image 110 ... NDSI inspection image 120 ... Determination image

Claims (8)

The inspection target is imaged by a first imaging unit that captures an image of a first wavelength, and a second imaging unit that captures an image of a second wavelength,
The spectral intensity of each pixel of the first captured image of the inspection object captured by the first imaging unit, and the spectrum of each pixel of the second captured image of the inspection object captured by the second imaging unit Generate an inspection image consisting of the normalized spectral index of each pixel computed from the intensity,
An inspection method comprising: inspecting a state of the inspection target based on the inspection image. In the inspection method according to claim 1,
The spectral intensity for the first wavelength with respect to the first portion contained in the inspection object and the first inclination of the spectrum due to the spectral intensity of the second wavelength are the second for the second portion contained in the inspection object The first wavelength and the second wavelength are selected to be different from a spectral intensity for one wavelength and a second slope of the spectrum due to the spectral intensity of the second wavelength,
An inspection method comprising: identifying the first portion and the second portion included in the inspection target based on the difference in the normalized spectral index of each pixel of the inspection image. In the inspection method according to claim 2,
The third gradient of the spectrum by the spectral intensity for the first wavelength and the spectral intensity of the second wavelength with respect to the third portion other than the first portion and the second portion included in the inspection object, The first wavelength and the second wavelength are selected to be different from the first inclination and / or the second inclination. In the inspection method according to any one of claims 1 to 3,
The inspection object is cup noodles,
The first part included in the inspection object is a roller char,
The first part included in the inspection object is powder,
The first wavelength is 685 nm,
The second wavelength is 670 nm. A first imaging unit for capturing an image of a first wavelength;
A second imaging unit for capturing an image of a second wavelength;
The spectral intensity of each pixel of the first captured image of the inspection object captured by the first imaging unit, and the spectrum of each pixel of the second captured image of the inspection object captured by the second imaging unit An inspection unit configured to generate an inspection image composed of a normalized spectral index of each pixel calculated from the intensity, and to inspect the state of the inspection target based on the inspection image; . In the inspection apparatus according to claim 5,
The spectral intensity for the first wavelength with respect to the first portion contained in the inspection object and the first inclination of the spectrum due to the spectral intensity of the second wavelength are the second for the second portion contained in the inspection object The first and second wavelengths are selected to be different from a spectral intensity for one wavelength and a second slope of the spectrum due to the spectral intensity of the second wavelength,
The inspection unit identifies the first portion and the second portion included in the inspection target based on the difference in the normalized spectral index of each pixel of the inspection image. apparatus. In the inspection apparatus according to claim 6,
The third gradient of the spectrum by the spectral intensity for the first wavelength and the spectral intensity of the second wavelength with respect to the third portion other than the first portion and the second portion included in the inspection object, The inspection apparatus, wherein the first wavelength and the second wavelength are selected to be different from the first inclination and / or the second inclination. The inspection apparatus according to any one of claims 5 to 7.
The inspection object is cup noodles,
The first part included in the inspection object is a roller char,
The first part included in the inspection object is powder,
The first wavelength is 685 nm,
The second wavelength is 670 nm.