JP2019124518A - Tablet inspection method and tablet inspection device - Google Patents

Abstract

To provide a tablet inspection method with which it is possible to suppress misdetection of a defect.SOLUTION: A tablet inspection method for inspecting the appearance of a tablet comprises steps (a) to (c). In the step (a), a tablet is photographed, and captured images are generated, including a plurality of images in each of which the appearance of the tablet seen from a plurality of imaging directions is shown. In the step (b), an inspection process for detecting a defect candidate of the tablet is executed on the captured images. In the step (c), when a defect candidate common in two or more images out of the plurality of images is detected, by the inspection process, in a first region of the tablet imaged in the plurality of images, it is determined that there is occurrence of a defect in the tablet.SELECTED DRAWING: Figure 12

Description

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

  Conventionally, a tablet inspection apparatus for inspecting the appearance of a tablet has been proposed (for example, Patent Document 1). In patent document 1, the inspection apparatus is provided with the conveyor which conveys a tablet, and the imaging device for imaging the tablet in conveyance. The imaging device captures an image so that the upper surface of the tablet viewed from the upper side and the side surface of the tablet viewed from four sides are included in one captured image.

  Specifically, the imaging device includes one camera and four prisms. The camera is disposed on the upper side of the tablet, and light from the upper surface of the tablet is imaged on the imaging surface of the camera through a lens or the like. Also, the prisms are arranged in the four sides of the tablet and reflect light from the side of the tablet to the camera. The light is also imaged on the imaging surface of the camera via a lens or the like. Thereby, the camera can image the external appearance of the tablet seen from five directions. The inspection apparatus performs an appearance inspection of a tablet by performing image processing on a captured image captured by a camera.

JP, 2012-123009, A

  Foreign matter such as dust may enter the inside of the imaging device and adhere to, for example, a lens or the like. In this case, even if the tablet does not actually have a defect, the foreign matter may overlap with the tablet in the captured image. In this case, the foreign matter is erroneously detected as a defect.

  In addition, some of the plurality of light receiving elements constituting the imaging surface of the camera may fail. The pixel value of the pixel corresponding to the malfunctioning light receiving element is, for example, zero without indicating a normal value. When this pixel is located inside the tablet in the captured image, the pixel is erroneously detected as a defect even if the tablet does not actually have a defect.

  Then, an object of this invention is to provide the tablet inspection method and tablet inspection apparatus which can suppress the misdetection of a defect.

  A first aspect of the tablet inspection method is a tablet inspection method for inspecting the appearance of a tablet, which is an image of a tablet, and a captured image including a plurality of images in which the appearance of the tablet viewed from a plurality of imaging directions is captured. Step (a) of generating, Step (b) of performing inspection processing on the captured image for detecting a defect candidate of the tablet, and the first region of the tablet reflected in the plurality of images by the inspection processing And (c) determining that the tablet has a defect when a common defect candidate is detected in two or more of the plurality of images.

  A second aspect of the tablet inspection method is the tablet inspection method according to the first aspect, wherein the second region of the tablet is formed only on n (n is an integer of 2 or more) of the plurality of images. A step of determining that the tablet has a defect when a defect candidate common to two or more of the n images is detected in the second region by the inspection process and d).

  A third aspect of the tablet inspection method is the tablet inspection method according to the first or second aspect, wherein the third region of the tablet is shown in only one of the plurality of images, and the inspection The method further includes the step (f) of determining that the tablet has a defect when a defect candidate is detected in the third region of the one image by the processing.

  A fourth aspect of the tablet inspection method is the tablet inspection method according to any one of the first to third aspects, wherein a mask area not to be inspected in the inspection process is set for each of the plurality of images, The fourth region of the tablet is copied to an inspection target region other than the mask region in only m (an integer of 2 or more) images among the plurality of images, and in the step (c), the inspection process is performed. When a common defect candidate is detected in the fourth area among two or more of the m images, it is determined that the tablet has a defect.

  A fifth aspect of the tablet inspection method is the tablet inspection method according to the fourth aspect, in each of the plurality of images, a pixel value is equal to or more than a predetermined value than other regions due to specular reflection of light to the tablet An area including an area to be raised is set as the mask area.

  A sixth aspect of the tablet inspection method is the tablet inspection method according to the fourth or fifth aspect, wherein the tablet has a first main surface, which is the first region, and a first surface facing the first main surface. An end having two main surfaces and a side surface connecting the peripheral edge of the first main surface and the peripheral edge of the second main surface, and located on both sides of a side surface area occupied by the side surface of the tablet in each of the plurality of images. An area is set to the mask area.

  A seventh aspect of the tablet inspection method is the tablet inspection method according to any one of the first to sixth aspects, wherein the tablet has a first main surface which is the first region, and the first main surface. And a side surface connecting the periphery of the first main surface and the periphery of the second main surface, wherein the step (b) includes, in each of the plurality of images, the tablet Identifying a main surface edge corresponding to the outline of the main surface area indicated by the first main surface, and a function predetermined as the shape of the outline of the main surface area in each of the plurality of images (B2) of obtaining an approximate line of the main surface edge as the contour of the main surface region based on the step (b3) of detecting a defect candidate in each of the plurality of images; In c), the plurality of images of the position of each pixel with respect to the main surface area Based on the correspondence relationship between each other, the detected defect candidates in the step (b3) to determine whether common to the two or more images.

  An eighth aspect of the tablet inspection method is the tablet inspection method according to the seventh aspect, wherein the first main surface of the tablet has a substantially circular shape in plan view, and the function shows an ellipse It is a function and the approximation line is an ellipse.

  A ninth aspect of the tablet inspection method is the tablet inspection method according to the seventh or eighth aspect, wherein the tablet has a substantially disc shape, and the side surface of the tablet in each of the plurality of images. The side surface area occupied by and the main surface area constitute a tablet area, and when each of the plurality of images is referred to as an inspection image, the step (b1) performs an edge detection process on the inspection image. A step (b11) of performing the edge image generation, and a step (b12) of specifying a tablet edge corresponding to the contour of the tablet area in the inspection image from the edge image; and the first main surface in the inspection image A part of the periphery of the tablet that is a part of the outline of the tablet area, and a main surface outer edge and a side surface outer edge respectively corresponding to the periphery of the second main surface in the inspection image from the tablet edge And (b13) in the edge image, searching for pixels in a search area at a predetermined distance from the side outer edge along the predetermined direction, and at the boundary between the main area and the side area. A step (b14) of identifying a corresponding boundary edge, and a step (b14) of identifying one set of the main surface outer edge and the boundary edge as the main surface edge.

  A tenth aspect of the tablet inspection method is the tablet inspection method according to the ninth aspect, wherein, in the step (b14), the pixel in the search area of the edge image, which corresponds to the pixel The boundary edge is identified by searching for a pixel whose pixel value of the pixel of the inspection image is larger than a predetermined threshold.

  An eleventh aspect of the tablet inspection method is the tablet inspection method according to the ninth or tenth aspect, wherein, in the step (b14), in the search area, the side outer edge is moved from the side outer edge toward the main surface outer edge The pixel is searched to identify the boundary edge.

  A twelfth aspect of the tablet inspection apparatus is the tablet inspection apparatus for inspecting the appearance of a tablet, which is an image of a tablet, and a captured image including a plurality of images in which the appearance of the tablet viewed from a plurality of imaging directions is captured. The image processing unit includes an imaging unit to generate and an image processing unit, and the image processing unit performs an inspection process for detecting a defect candidate of a tablet on the captured image, and the inspection process generates the plurality of images. In the first region of the tablet to be photographed, when a common defect candidate is detected in two or more of the plurality of images, it is determined that the tablet has a defect.

  According to the first aspect of the tablet inspection method and the twelfth aspect of the tablet inspection apparatus, it is determined that the tablet has a defect when a common defect candidate is detected in two or more images. It is possible to suppress false reports at the time of defect detection.

  According to the second aspect of the tablet inspection method, it is possible to further suppress the rare false alarm.

  According to the third aspect of the tablet inspection method, defects in the third area can be detected.

  According to the fourth aspect of the tablet inspection method, defects in the fourth area can be appropriately detected.

  According to the fifth aspect of the tablet inspection method, it is possible to avoid false detection of the light intensity on the tablet as a defect.

  According to the sixth aspect of the tablet inspection method, since the end area of the side area where the detection of defects is difficult is set as the mask area, the influence on the defect detection due to the false detection or defect detection of the defect candidate in the end area is avoided. it can.

  According to the seventh aspect of the tablet inspection method, since the outline of the main surface area is obtained based on a function determined in advance as the shape of the outline of the main surface area, the accuracy of specifying the main surface area is high. According to this, it is possible to improve the accuracy of the correspondence between the two or more images of the pixels in the main surface area, and to improve the determination accuracy of the step (c).

  According to the eighth aspect of the tablet inspection method, since the appropriate function is adopted according to the shape of the tablet, the main surface can be identified with high accuracy.

  According to the ninth aspect of the tablet inspection method, the main surface edge can be appropriately obtained.

  According to the tenth aspect of the tablet inspection method, the main surface area can be identified with higher accuracy in each of the plurality of images.

  According to the eleventh aspect of the tablet inspection method, even if a dividing line is formed on the main surface of the tablet, it is possible to avoid erroneous detection of an edge corresponding to the dividing line as a boundary edge.

It is a figure showing roughly an example of composition of a tablet inspection device. It is a perspective view which shows an example of a tablet. It is a figure showing roughly an example of an internal configuration of an imaging head. It is a figure showing roughly an example of an internal configuration of an imaging head. It is a figure showing roughly an example of an image pick-up picture. It is a flow chart which shows an example of operation of a tablet inspection device. It is a flow chart which shows an example of inspection processing. It is a figure showing roughly an example of a picture. It is a figure which shows roughly an example of the tablet edge corresponding to the outer periphery of a tablet. It is a figure for demonstrating an example of the identification method of the boundary edge corresponding to the boundary between the main surface and side surface of a tablet in an image. It is a flowchart which shows an example of the identification method of a boundary edge. It is a flowchart which shows an example of the defect true / false process with respect to a main surface area | region. It is a figure which shows an example of the correspondence of each pixel between images. It is a flowchart which shows another example of a test | inspection process. It is a flowchart which shows an example of the defect true / false process with respect to a side area | region. It is a figure for demonstrating another example of the identification method of a boundary edge. It is a flowchart which shows another example of the identification method of a boundary edge. It is a figure for demonstrating another example of the identification method of a boundary edge. It is a figure which shows roughly another example of a captured image. It is a flowchart which shows another example of a test | inspection process.

  Hereinafter, embodiments will be described in detail with reference to the drawings. In the drawings, for the purpose of easy understanding, the dimensions and the numbers of the respective parts are drawn in an exaggerated or simplified manner as necessary. In the drawings, parts having similar configurations and functions are denoted by the same reference numerals, and redundant description will be omitted in the following description. Further, in each drawing, in order to clarify the directional relationship of the components, an XYZ orthogonal coordinate system in which the Z-axis direction is a vertical direction and the XY plane is a horizontal plane is appropriately attached.

First Embodiment
FIG. 1 is a view schematically showing an example of the configuration of a tablet printing apparatus 10. As shown in FIG. The tablet printing apparatus 10 includes a tablet inspection apparatus 1 and a print head 6.

  The tablet inspection apparatus 1 is an apparatus for inspecting the appearance of the tablet 9. FIG. 2 is a perspective view schematically showing an example of the tablet 9. In the example of FIG. 2, the tablet 9 has a substantially disc shape, and includes a pair of main surfaces 9a and 9b and a side surface 9c. For example, a so-called flat lock or a lock with R can be adopted as the disk shape. The main surfaces 9a and 9b face each other, and have substantially the same circular shape in plan view. One and the other of the main surfaces 9a and 9b may be referred to as front and back, respectively, and may also be referred to as top and bottom.

  In the example of FIG. 2, the peripheral edge and side surface 9c of the main surface 9a of the tablet 9 are connected to each other to form a corner, and the peripheral edge and side surface 9c of the main surface 9b are also connected to each other to form a corner . The diameter of the tablet 9 may be set to, for example, about 5 mm to about 10 numbers, and the thickness thereof may be set to, for example, about 5 mm or less.

  As exemplified in FIG. 2, a dividing line 91 may be formed on the major surface 9 a of the tablet 9. The dividing line 91 is a groove extending straight through the center of the main surface 9a from end to end of the main surface 9a. The same dividing line may be formed on the main surface 9b.

  Referring to FIG. 1, tablet inspection apparatus 1 includes hopper 2, transport drum 3, transport unit 4, imaging head 5, sorting unit 7 and control unit 8.

  The hopper 2 is a loading unit for loading the tablet 9 into the tablet printing apparatus 10. The hopper 2 is provided above the ceiling of a housing (not shown) of the tablet printing apparatus 10. The tablets 9 fed from the hopper 2 are guided to the transport drum 3. The elements other than the hopper 2 are provided inside the housing of the tablet printing apparatus 10.

  The transport drum 3 has a substantially cylindrical shape, and the central axis of the transport drum 3 is disposed along the Y-axis direction. The transport drum 3 is rotated counterclockwise on the paper surface of FIG. The rotation drive motor is controlled by, for example, the control unit 8.

  A plurality of suction holes (not shown) are formed along the circumferential direction on the outer peripheral surface of the transport drum 3. Each of the plurality of suction holes is in communication with a suction mechanism (not shown) provided inside the conveyance drum 3. The suction mechanism is controlled by, for example, the control unit 8. By operating the suction mechanism, negative pressure lower than atmospheric pressure can be applied to each of the plurality of adsorption holes. Thus, each suction hole of the transport drum 3 can hold one tablet 9 by suction.

  The transport unit 4 is disposed below the transport drum 3. The tablets 9 adsorbed and held on the outer peripheral surface of the transport drum 3 move in the circumferential direction as the transport drum 3 rotates. When the tablet 9 moves to the lower side of the transport drum 3, the suction mechanism releases the adsorption on the tablet 9, whereby the tablet 9 falls and is delivered to the transport unit 4.

  The transport unit 4 transports the tablets 9. In the example of FIG. 1, the transport unit 4 is a belt conveyor, and includes a transport belt 41 and a pair of pulleys 42. The pair of pulleys 42 is disposed, for example, at an interval in the X-axis direction, and the central axis of itself is disposed along the Y-axis direction. Each of the pair of pulleys 42 rotates around its own central axis.

  The transport belt 41 is stretched around a pair of pulleys 42. As at least one of the pair of pulleys 42 is rotationally driven by a drive motor (not shown), the transport belt 41 travels in the direction shown by the arrow in FIG. 1. The drive motor is controlled by, for example, the control unit 8.

  Also on the outer peripheral surface of the conveyance belt 41, a plurality of suction holes (not shown) are formed along the circumferential direction. Each of the plurality of suction holes is in communication with a suction mechanism (not shown) provided inside the conveyance belt 41. The suction mechanism is controlled by, for example, the control unit 8. By operating the suction mechanism, negative pressure lower than atmospheric pressure can be applied to each of the plurality of adsorption holes. Thus, each suction hole of the conveyance belt 41 can hold one tablet 9 by suction.

  As the transport belt 41 travels while suction-holding the tablets 9, the tablets 9 are transported in the direction away from the transport drum 3 along the X-axis direction.

  The imaging head 5 is disposed on the downstream side of the transport drum 3 midway along the transport path of the tablets 9 by the transport portion 4 and at a position facing the transport portion 4. The imaging area of the imaging head 5 includes a part of the transport belt 41. When the tablet 9 moves in the imaging area, the imaging head 5 images the tablet 9 to generate a captured image. The imaging head 5 outputs this captured image to the control unit 8. An example of a specific internal configuration of the imaging head 5 will be described in detail later.

  The control unit 8 inspects the appearance of the tablet 9 by performing image processing on the input captured image. The controller 8 determines that the tablet 9 is defective if the appearance of the tablet 9 is defective, and determines that the tablet 9 is good if the appearance of the tablet 9 is not defective. Examples of the defects include impurities attached to the tablet 9 or defects such as defects in shape (e.g., chipping) of the tablet 9. An example of specific image processing by the control unit 8 will be described in detail later.

  The print head 6 is disposed on the downstream side of the imaging head 5 above the transport belt 41 while the tablets 9 are being transported. The print head 6 is controlled by, for example, the control unit 8 and performs a printing process on the tablet 9. The print head 6 includes a plurality of discharge nozzles (not shown), and discharges ink droplets from the discharge nozzles by an inkjet method. The inkjet method may be a piezo method in which a piezoelectric element (piezoelectric element) is deformed by applying a voltage to discharge ink droplets, or a heater is energized to heat the ink, whereby the ink droplets are discharged. It may be a thermal method of discharging the ink. In the present embodiment, in order to print the tablets 9, as the ink, an edible ink manufactured using a material recognized by the Food Sanitation Law is used.

  In the example of FIG. 1, since the print head 6 is located on the downstream side of the transport path from the imaging head 5, the control unit 8 may determine the necessity of the printing process according to the inspection result of the tablet 9. . More specifically, when the tablet 9 is determined to be non-defective in the appearance inspection, the control unit 8 controls the print head 6 to print the tablet 9 so that the tablet 9 is defective. If it is determined, the print head 6 may be controlled so as not to print the tablet 9. According to this, unnecessary print processing can be avoided.

  The sorting unit 7 sorts the tablets 9 in accordance with the result of the appearance inspection. For example, the sorting unit 7 has a non-defective box and a defective item box. These boxes have a box-like shape opened upward. The non-defective box stores the tablets 9 determined to be non-defective, and the defective box stores the tablets 9 determined to be non-defective. For example, these boxes are arranged side by side along the X-axis direction on the lower side of the transport belt 41. When the tablet 9 judged to be non-defective is positioned immediately above the opening of the non-defective box, the control unit 8 controls the suction mechanism in the transport belt 41 to release the adsorption on the tablet 9. Thereby, the tablet 9 falls into the inside of the non-defective item box and is accommodated therein. The tablets 9 determined to be defective are similarly stored inside the defective box.

  As described above, the control unit 8 controls various components, and performs an appearance inspection of the tablet 9 by performing image processing on a captured image input from the imaging head 5.

  The control unit 8 is an electronic circuit device, and may have, for example, a data processing device and a storage medium. The data processing device may be, for example, an arithmetic processing device such as a CPU (Central Processor Unit). The storage unit may have a non-transitory storage medium (for example, a ROM (Read Only Memory) or a hard disk) and a temporary storage medium (for example, a RAM (Random Access Memory)). The non-temporary storage medium may store, for example, a program that defines a process performed by the control unit 8. The processing unit executes this program, whereby the control unit 8 can execute the processing defined in the program. Of course, part or all of the processing performed by the control unit 8 may be performed by hardware.

<Imaging head>
The imaging head 5 captures an image of the tablet 9 being transported from a plurality of imaging directions. The tablets 9 may be adsorbed and held by the conveyance belt 41 with the main surface 9b facing the conveyance belt 41 side, and may be adsorbed and held by the conveyance belt 41 with the main surface 9a facing the conveyance belt 41. Sometimes. In the following, for convenience of explanation, it is assumed that the tablet 9 is held on the transport belt 41 with the main surface 9 b facing the transport belt 41 side. Further, in the state where the tablets 9 are adsorbed and held by the conveyance belt 41, at least a part of the side surface 9 c on the main surface 9 a side is exposed from the conveyance belt 41.

  FIG. 3 and FIG. 4 are diagrams schematically showing an example of the internal configuration of the imaging head 5. For example, the imaging head 5 includes an inspection camera 51, a lens group 52, a mirror 53, and a pyramidal mirror 54. The inspection camera 51 is an image sensor such as, for example, a charge coupled device (CCD) image sensor or a complementary metal oxide semiconductor (CMOS) image sensor. The inspection camera 51 is disposed in a position in which the imaging surface thereof faces the conveyance belt 41 at a position where it faces a part of the conveyance path of the tablet 9 in the Z-axis direction.

  The mirror 53 is provided to guide light from the major surface 9 a and the side surface 9 c of the tablet 9 to the pyramidal mirror 54 described later. In describing the position of the mirror 53, for convenience, it is assumed that the tablet 9 is stopped at a position facing the inspection camera 51 in the Z-axis direction. Referring to FIG. 3, mirror 53 is located between inspection camera 51 and tablet 9 in the Z-axis direction, and is disposed outside of tablet 9 in plan view (that is, viewed along the Z-axis direction). It is done. Also, the mirror 53 is disposed to face each surface of the pyramidal mirror 54.

  The pyramidal mirror 54 guides the light from the major surface 9 a and the side surface 9 c of the tablet 9 reflected by the mirror 53 to the imaging surface of the inspection camera 51 via the lens group 52. The pyramidal mirror 54 is configured of four mirrors, and the pyramidal mirror 54 is disposed to face each of the mirrors 53.

  A part of the light reflected or scattered by the main surface 9a and the side surface 9c of the tablet 9 proceeds obliquely upward to one of the mirrors 53, is reflected by the mirror 53, and the light reflected by the mirror 53 is pyramidal type After being reflected by the mirror 54, an image is formed on the imaging surface of the inspection camera 51 through the lens group 52. In other words, the angles of the reflecting surfaces of the mirror 53 and the pyramidal mirror 54 with respect to the ground are adjusted so that light traveling obliquely upward from the major surface 9 a and the side surface 9 c of the tablet 9 can be reflected toward the imaging surface of the inspection camera 51 It is done.

  Thereby, the inspection camera 51 can image the appearance of the tablet 9 viewed from the mirror 53. That is, the inspection camera 51 can substantially image the tablet 9 in an oblique direction, and the captured image includes both the main surface 9 a and the side surface 9 c of the tablet 9. In FIG. 3, an example of the light path is shown by a broken line.

  The imaging head 5 includes a plurality of (four in FIG. 4) mirrors 53. For example, two mirrors 53 are spaced apart in the X-axis direction, and the remaining two mirrors 53 are spaced apart in the Y-axis direction. According to this, the tablet 9 is surrounded by the mirror 53 in four directions in plan view. Further, pyramidal mirrors 54 are disposed at the centers of the mirrors 53 at intervals. The light reflected by each mirror 53 is further reflected by the pyramidal mirror 54, and is imaged on the imaging surface of the inspection camera 51 through the lens group 52. Specifically, the light reflected by the four mirrors 53 is imaged on different areas of the imaging surface of the inspection camera 51. Thereby, the imaging head 5 images the tablet 9 from four directions, and generates a captured image including a plurality of images in which the appearance of the tablet 9 viewed from the four directions is captured.

  FIG. 5 is a diagram schematically illustrating an example of the captured image IM1. The captured image IM1 includes four images IM11 to IM14 that respectively capture the appearance of the tablet 9 viewed from the four imaging directions, and in any of these, the main surface 9a and the side surface 9c of the tablet 9 are captured. There is. Here, since the tablet 9 is imaged from four directions, the side surface 9c of the tablet 9 can be imaged over the entire circumference.

  In the example of FIG. 5, the entire area of the main surface 9 a of the tablet 9 is shown in any of the four images IM <b> 11 to IM <b> 14 (corresponding to inspection images). Therefore, the main surface 9a is a common surface common to the images IM11 to IM14. On the other hand, on the side surface 9 c of the tablet 9, regions corresponding to the imaging direction are respectively captured in the images IM 11 to IM 14. Further, in the example of FIG. 5, defect candidates d1 and d2 appear in the captured image IM1. The defect candidate mentioned here is a candidate that may be a defect of the tablet 9.

  In the example of FIG. 5, since the defect candidate d1 is commonly shown in any of the images IM11 to IM14, it can be said that the defect is generated on the tablet 9. On the other hand, the defect candidate d2 appears in only one image IM12, and does not appear in the other images IM11, IM13, and IM14. The defect candidate d2 is not a defect generated in the tablet 9, but can be said to be generated due to the imaging head 5. For example, the defect candidate d2 is considered to indicate an abnormality (for example, attached matter) of the lens group 52, the mirror 53 or the pyramidal mirror 54, or an abnormality of the imaging surface of the inspection camera 51. Therefore, it is not preferable to detect the defect candidate d2 as a defect of the tablet 9.

  The imaging head 5 may have an illumination light source (not shown). The illumination light source has, for example, a ring shape, and irradiates the tablet 9 with light. Thereby, the brightness of the tablet 9 in the captured image IM1 can be improved.

  In the examples of FIGS. 3 and 4, although the light path is bent by the mirror 53 and the pyramidal mirror 54, this is not necessarily the case. Another optical element such as a prism may be employed as an element for bending the light path.

<Inspection>
The control unit 8 performs image processing on the captured image IM1 input from the imaging head 5, and performs an appearance inspection of the captured tablet 9. Thus, the control unit 8 functions as an image processing unit.

  FIG. 6 is a flowchart showing an example of the operation in the tablet inspection apparatus 1. First, in step S1, the imaging head 5 captures an image of the tablet 9 during conveyance from a plurality of imaging directions, and generates a captured image IM1. The imaging head 5 outputs the captured image IM1 to the control unit 8.

  Next, in step S2, the control unit 8 performs an inspection process (image process) on the captured image IM1. That is, the control unit 8 performs an inspection process on the images IM11 to IM14.

  Although a specific example of this inspection process will be described later, here first, the outline will be briefly described. For example, the control unit 8 performs identification processing and candidate detection processing as inspection processing. The identification process is a process of identifying a tablet area occupied by the tablet 9 in each of the images IM11 to IM14. Thereby, the position and shape of the tablet area are specified in each of the images IM11 to IM14.

  The candidate detection process is a process of detecting a defect candidate in each of the images IM11 to IM14. The candidate detection process includes, for example, a process of determining whether or not the pixel is a pixel indicating a defect candidate based on the pixel value of each pixel of each of the images IM11 to IM14, and the defect candidate is detected as a determination result .

  The identification process identifies the tablet area of the tablet 9 in each of the images IM11 to IM14, and the candidate detection process identifies the position of the defect candidate in each of the images IM11 to IM14. Each of the images IM11 to IM14 is identified. That is, the position of the defect candidate on the tablet 9 is specified for each of the images IM11 to IM14. Therefore, it can be said that this inspection process is a process of detecting the position of the defect candidate with respect to the tablet area.

  Next, in step S3, the control unit 8 performs defect authenticity processing based on the result of inspection processing of the images IM11 to IM14. The defect authenticity process is a process of determining whether the defect candidate detected by the inspection process indicates a defect of the tablet 9, that is, the process of determining the authenticity of the defect candidate. Specifically, in the region (for example, the main surface 9a) of the tablet 9 that appears in any of the images IM11 to IM14 by the inspection processing, the control unit 8 selects a common defect candidate (for example, the defect candidate d1) ) Is detected, it is determined that the tablet 9 is defective. On the other hand, when a defect candidate (for example, defect candidate d2) is detected in only one of the regions of images IM11 to IM14 by inspection processing, control unit 8 generates the defect candidate on tablet 9 Judge not a defect.

  Although a specific example of this defect authenticity processing will be described later, here first, the outline will be briefly described. For example, the control unit 8 obtains a geometrical correspondence of each pixel in the tablet area in the images IM11 to IM14. That is, the control unit 8 obtains the correspondence of pixels that capture the same point on the tablet 9. Next, the control unit 8 determines whether the defect candidate is common to the images IM11 to IM14 based on the correspondence relationship. Specifically, the positions of defect candidates are compared based on the correspondence relationship. The control unit 8 determines that the tablet 9 has a defect when the positions correspond to each other in the images IM11 to IM14, and when the positions do not correspond to each other, the defect candidate is not a defect generated in the tablet 9 I will judge.

  In addition, even if a common defect candidate is not detected in all four images IM11 to IM14, when a common defect candidate is detected in two or more images, the possibility that the defect candidate is a defect of tablet 9 is excluded I can not finish it. Therefore, when a defect candidate common to at least two of the images IM11 to IM14 is detected, the control unit 8 may determine that the tablet 9 has a defect. According to this, it can be avoided that the tablet 9 having a possibility of defect is judged as non-defective.

  As described above, according to the tablet inspection apparatus 1, unless a common defect candidate is detected in at least two of the images IM11 to IM14, the defect candidate is not regarded as a defect. Therefore, false reports at the time of defect detection can be suppressed.

<Specific example of inspection process>
FIG. 7 is a flowchart showing a specific example of inspection processing. Here, an example of inspection processing for one image IM12 will be described as a representative. The inspection process of the other images IM11, IM13, IM14 is the same.

  In order to describe the inspection process for the image IM12, first, each area in the image IM12 is defined. FIG. 8 schematically shows an example of the image IM12. Here, a tablet area TR, a background area BR, a main area Ra and a side area Rc are introduced. The tablet area TR is an area occupied by the tablet 9 in the image IM12. The background area BR is an area other than the tablet area TR in the image IM12. The main surface area Ra is an area occupied by the main surface 9a of the tablet 9 in the image IM12, and the side surface area Rc is an area occupied by the side surface 9c of the tablet 9 in the image IM12. The main surface area Ra and the side surface area Rc constitute a tablet area TR. Each area is similarly defined for the images IM11, IM13, IM14. Hereinafter, in order to distinguish the main surface 9a and the side surface 9c of the tablet 9 in the image IM12 from the main surface 9a and the side surface 9c of the actual tablet 9, they are referred to as a main surface 9aa and a side surface 9ca, respectively.

  The control unit 8 executes steps S21 to S26 as the identification process. Here, the main surface area Ra and the side surface area Rc that constitute the tablet area TR are specified. First, in step S21, the control unit 8 performs edge detection processing on the image IM12 to generate an edge image. However, here, the control unit 8 deletes the background area BR in the image IM12, and performs edge detection processing on the image IM12 after the deletion.

  In order to delete the background area BR, the control unit 8 first distinguishes between the tablet area TR and the background area BR of the image IM12. For example, the control unit 8 uses a binarization process on the image IM12 to distinguish between the tablet area TR and the background area BR. Since the background area BR includes the conveyance belt 41, if the contrast between the actual color of the conveyance belt 41 and the color of the tablets 9 is increased in order to accurately distinguish the tablet area TR and the background area BR. Good. For example, when the tablets 9 are white-based, the transport belt 41 is made black. After specifying the background area BR, the control unit 8 deletes the background area BR in the image IM12. For example, the control unit 8 deletes the background area BR by setting the pixel values of all the pixels in the background area BR to zero.

  Next, the control unit 8 performs edge detection processing on the image IM12 from which the background area BR has been deleted, and generates an edge image. Although the specific processing method of the edge detection processing need not be particularly limited, an example thereof will be briefly described. For example, the control unit 8 performs edge strength processing on the image IM12 after removal to generate an edge strength image. The edge strength processing includes, for example, calculation processing of the difference in pixel value between the respective pixels, whereby an area with a large difference in pixel value between the pixels in the image IM12 becomes an edge strength image. In the edge strength image, the control unit 8 determines for each pixel whether the pixel value is larger than a predetermined edge threshold. The edge threshold may be set in advance and stored in the storage medium of the control unit 8, for example. The control unit 8 detects a pixel having a pixel value larger than the edge threshold to generate an edge image.

  Since this edge image is generated based on the image IM12 from which the background area BR has been removed, an edge in the background area BR (for example, the unevenness of the transport belt 41) is not detected in the edge image. That is, the edge (hereinafter referred to as a tablet edge) corresponding to the contour of the tablet region TR is positioned at the outermost periphery of the edge group in the edge image.

  Next, in step S22, the control unit 8 specifies the tablet edge corresponding to the contour of the tablet region TR as follows. That is, the control unit 8 specifies the edge located at the outermost periphery among the edge group in the edge image as the tablet edge P0 (see FIG. 9).

  FIG. 9 is a view schematically showing an example of the tablet edge P0. In the example of FIG. 9, the tablet edge P0 is schematically shown divided into a plurality of edges. The tablet edge P0 is composed of a side outer edge P1, a main surface outer edge P2, and a pair of side ridge line edges P4. The side outer edge P1 is an edge corresponding to a portion of the periphery of the major surface 9b of the tablet 9 that is reflected in the image IM12. Since the major surface 9b has a circular shape, the side outer edge P1 ideally has a semi-elliptical shape. The term "semi-elliptical shape" as used herein refers to a shape obtained by dividing an ellipse into two at its major axis.

  The main surface outer edge P2 is an edge corresponding to a portion of the periphery of the main surface 9aa of the tablet 9 in the image IM12 which is not in contact with the side surface 9ca. It can be said that this main surface outer edge P2 is an edge corresponding to a part of the contour of the tablet region TR in the periphery of the main surface 9aa of the tablet 9. Since the major surface 9a has a circular shape, the major surface outer edge P2 ideally has a semi-elliptical shape. More specifically, the side outer edge P1 and the main surface outer edge P2 have a semi-elliptical shape that bulges to the opposite side.

  The pair of side edge line edges P4 ideally extends in a straight line, and connects both ends of the main surface outer edge P2 to both ends of the side surface outer edge P1. The pair of side edge line edges P4 is an edge corresponding to a part of the outline of the side surface 9ca of the tablet 9 in the image IM12. The pair of side edge line edges P4 extend substantially in parallel, and the extending direction thereof is predetermined by the arrangement of the internal configuration of the imaging head 5. Here, it is assumed that the side edge line edge P4 extends in a direction intersecting at approximately 45 degrees with respect to the lateral direction in the image IM12.

  In addition, a pair of side ridgeline edge P4 is not completely perfect parallel in fact. The pair of side edge line edges P4 are slightly inclined so as to be closer to each other as the side outer edge P1 side is approached. In the following, for the sake of simplicity, it is assumed that the pair of side ridge edges P4 are parallel. If it thinks more strictly, what is necessary is just to grasp the extending direction of side edge line edge P4 described below as the direction represented by the bisector of the extending direction of a pair of side edge line edge P4.

  In the example of FIG. 9, the boundary edge P3 corresponding to the boundary between the main surface 9aa and the side surface 9ca of the tablet 9 in the image IM12 and the dividing line edge P5 corresponding to the dividing line 91 are indicated by broken lines. It can be said that the boundary edge P3 is an edge corresponding to the boundary between the main surface region Ra and the side surface region Rc. The boundary edge P3 ideally has a semi-elliptical shape that bulges on the opposite side to the major surface outer edge P2. One set of the main surface outer edge P2 and the boundary edge P3 corresponds to the periphery of the main surface 9aa of the tablet 9, and ideally exhibits an elliptical shape. Hereinafter, one set of the major surface outer edge P2 and the boundary edge P3 is also referred to as a major surface edge P23.

  Referring again to FIG. 7, in step S23, the control unit 8 extracts the side outer edge P1 and the main surface outer edge P2 from the tablet edge P0 identified in step S22. For example, the control unit 8 identifies an edge of the tablet edge P0 that extends along the predetermined direction D1 (= the extension direction of the side edge P4) as the side edge P4. Then, the control unit 8 identifies one of the two edges obtained by removing the side edge line edge P4 from the tablet edge P0 on the predetermined side (upper right side in the figure) as the side outer edge P1, and the other as the main surface outer edge P2. Identify

  Next, in step S24, the control unit 8 specifies the boundary edge P3 in the edge image. For example, in consideration of the similarity between the shape of the boundary edge P3 and the shape of the side outer edge P1, the boundary edge P3 is specified as described below.

  First, the geometrical relationship between the boundary edge P3 and the side outer edge P1 will be described. The boundary edge P3 and the side outer edge P1 have a semi-elliptical shape bulging on the same side. Here, since the tablet 9 is not so thick, it can be considered that the boundary edge P3 and the side outer edge P1 have substantially the same shape. That is, the boundary edge P3 exists in a region in which the side surface outer edge P1 is moved in parallel along the predetermined direction D1 toward the main surface outer edge P2 by the length of the side surface ridge line edge P4. Since the length of the side ridge line edge P4 is predetermined according to the thickness of the tablet 9 and the arrangement of the internal configuration of the imaging head 5, if the side outer edge P1 is specified, the region where the boundary edge P3 exists is estimated. be able to.

  Therefore, the control unit 8 searches for a boundary edge P3 by searching for pixels in a search area R1 (see also FIG. 10) separated from the side surface outer edge P1 by a predetermined distance toward the main surface outer edge P2 along the predetermined direction D1. Identify. FIG. 10 is a view schematically showing an example of the search area R1. In explaining an example of the search area R1, a center line L0 of the search area R1 and lines L1 to L4 forming an outline of the search area R1 are introduced.

  The center line L0 is a line obtained by moving the side outer edge P1 by the length of the side edge P4 along the predetermined direction D1 toward the main surface outer edge P2. In the example of FIG. 10, although the center line L0 and the boundary edge P3 are schematically shown by one curve, they may be different in practice.

  The lines L1 and L2 are lines obtained by moving the center line L0 along the predetermined direction D1 to the opposite side by a predetermined width. In FIG. 10, the line L1 is closer to the side outer edge P1 than the line L2. The lines L3 and L4 extend along the predetermined direction D1 and connect both ends of the lines L1 and L2, respectively. The search area R1 is an area surrounded by one set of these lines L1 to L4.

  In FIG. 10, part of the dividing line edge P5 is also shown. Hereinafter, for convenience of description, a pixel group arranged in the search area R1 along the predetermined direction D1 will be referred to as a “row”.

  FIG. 11 is a flowchart illustrating an example of the search process of the control unit 8. First, in step S51, the control unit 8 initializes the values N and M to one. The value N indicates the number of the row in the search area R1, and the value M indicates the number of the pixel belonging to the row. The first pixel is a pixel on the line L1, and the Mth pixel is a pixel next to the (M-1) th pixel in each row.

  Next, in step S52, the control unit 8 determines whether or not the Nth row and Mth pixel in the search area R1 of the edge image indicates an edge. If a negative determination is made, in step S53, the control unit 8 adds 1 to the value M to update the value M, and executes step S52 again using the updated value M. That is, if a pixel indicating an edge (hereinafter also referred to as an edge pixel) can not be detected, the same determination is made for the next pixel.

  When an affirmative determination is made in step S52, in step S54, the control unit 8 recognizes the pixel as a component of the boundary edge P3. Next, in step S55, the control unit 8 adds 1 to the value N to update the value N, and initializes the value M to 1. Next, at step S56, control unit 8 determines whether value N is larger than reference value Nref. The reference value Nref is the total number of rows included in the search area R1. That is, the control unit 8 determines whether all the rows have been searched. If it is determined that all the lines have not been searched, the control unit 8 executes step S52 again. If it is determined that all the lines have been searched, the control unit 8 ends the search processing.

  As described above, in each row in the search area R1, the control unit 8 sequentially selects pixels from the line L1 and specifies the edge pixel detected first as a component of the boundary edge P3. In the example of FIG. 10, the search direction of the pixel is schematically shown by a solid arrow in the search area R1, and the end point of the arrow indicates a component of the boundary edge P3.

  The control unit 8 specifies, as the main surface edge P23, a pair of the main surface outer edge P2 specified in step S23 and the boundary edge P3 specified in step S24.

  As described above, according to the present embodiment, the main surface edge P23 can be appropriately and easily obtained.

  Further, according to the above-described search processing, the pixel on the dividing line edge P5 is not identified as a component of the boundary edge P3. Because the dividing line edge P5 is located on the main surface outer edge P2 side with respect to the boundary edge P3, the control unit 8 moves the search area R1 from the line L1 side to the line L2 side (that is, from the side surface outer edge P1 to the main surface This is because the search is made to the outer edge P2 side). That is, according to this search processing, the control unit 8 can specify the component of the boundary edge P3 before searching for the pixel of the dividing line edge P5. That is, false detection of the dividing line edge P5 as the boundary edge P3 can be avoided.

  By the way, although the main surface edge P23 corresponds to the periphery of the main surface 9aa of the tablet 9 in the image IM12, the main surface edge P23 actually detected does not necessarily exhibit an ideal elliptical shape. For example, the light irradiated in the vicinity of the peripheral edge of the main surface 9a of the tablet 9 has unevenness, or the light reflected regularly from a part in the vicinity of the peripheral edge is imaged on the inspection camera 51 and a part of pixel values May have a very large value compared to other pixel values. Alternatively, a defect may occur in the vicinity of the peripheral edge. In such a case, the major surface edge P23 may be different from the ideal elliptical shape.

  Therefore, in step S25, the control unit 8 calculates an approximate line (ellipse) approximate to the major surface edge P23 as the contour of the major surface area Ra based on the function of the ellipse. More generally, the control unit 8 calculates an approximate line of the major surface edge P23 as the contour of the major surface region Ra based on a reference function predetermined as the shape of the contour of the major surface region Ra in the image IM12. Do. The reference function here is a function that represents the shape of the outline of the main surface area Ra, and is a function whose position and size are variables. In the case of a function of an ellipse, for example, the length of the major axis, the length of the minor axis, the extending direction of the major axis, and the center can be adopted as variables. The control unit 8 calculates an ellipse E1 (more specifically, the length of the major axis, the length of the minor axis, and the extending direction and center of the major axis) approximating the major surface edge P23 by the least square method, for example. The ellipse E1 may be, for example, an expression shown in the coordinate system of the captured image IM1 (or the image IM12).

  As described above, since the approximate line of the principal surface edge P23 is calculated as the contour of the principal surface region Ra based on the function indicating the contour of the principal surface region Ra, the contour of the principal surface region Ra is specified with higher accuracy. can do. As a result, the position of the defect candidate with respect to the main surface area Ra can be specified with high accuracy. Moreover, in the above-mentioned specific example, the tablet 9 has a substantially disk shape. That is, the main surface 9a of the tablet 9 has a substantially circular shape. Therefore, in the image IM12, the periphery of the major surface 9aa ideally has an elliptical shape. Corresponding to this, a function of an ellipse is used as a function indicating the contour of the main surface area Ra. Therefore, the outline of main surface area Ra can be specified appropriately.

  Hereinafter, among the two half ellipses obtained by dividing the ellipse E1 by its long axis, the half ellipse on the side 9ca side is also referred to as the half ellipse El. The semi-ellipse E11 corresponds to an approximation line of the boundary edge P3.

  Next, in step S26, the control unit 8 specifies the contour of the side surface region Rc. For example, first, the control unit 8 calculates an approximation line (semi-ellipse) approximating the side outer edge P1 based on the ellipse E1. The control unit 8 calculates the ellipse E2 along the side outer edge P1 by using the Levenberg-Marquardt (LM) method while moving the ellipse E1 along the predetermined direction D1. The ellipse E2 corresponds to the periphery of the major surface 9b of the tablet 9. Therefore, the control unit 8 divides this ellipse E2 by its major axis, and obtains a semi-elliptic E21 which is farther from the boundary edge P3 among the obtained two semi-elliptics as an approximation line of the side outer edge P1.

  When calculating the ellipse E2, the control unit 8 may reduce the ellipse E1 at a predetermined rate. In a perspective view such as the image IM12, the periphery of the major surface 9b of the tablet 9 is smaller than the periphery of the major surface 9a of the tablet 9 at a predetermined ratio. This predetermined ratio is preset according to the thickness of the tablet 9.

  The control unit 8 does not necessarily calculate the ellipse E2 based on the ellipse E1, and does not need to calculate the semi-elliptic E21 from the ellipse E2. For example, the control unit 8 may calculate a half ellipse E21 using the LM method while obtaining the half ellipse E11 from the ellipse E1 and moving the half ellipse E11 along the predetermined direction D1. In addition, the semi-ellipse E11 may be reduced at a predetermined ratio in calculating the semi-ellipse E21.

  The control unit 8 calculates a pair of straight lines connecting the both ends of the semi-ellipse E11 and E21, and grasps the semi-ellipse E11 and E21 and the pair of straight lines as the outline of the side region Rc occupied by the side surface 9ca of the tablet 9. .

  Since the approximation line (half ellipses E11 and E21) approximating the boundary edge P3 and the side outer edge P1 in this way is adopted as a part of the outline of the side region Rc, the side region Rc is specified with higher accuracy. Can.

  Next, in step S27, the control unit 8 performs candidate detection processing on the image IM12. Hereinafter, a specific example of the candidate detection process will be described.

<First candidate detection processing (edge strength processing)>
At the boundary between the area occupied by the defect candidate d1 and d2 (see FIG. 8) and the area around the defect candidate d1 and d2 in the image IM12, the difference in pixel value between the pixels becomes large. Therefore, the control unit 8 may detect a defect candidate as follows. That is, the control unit 8 performs edge strength processing on the image IM12 from which the background area BR is deleted, and generates an edge strength image. If an edge strength image is already generated during generation of an edge image, the edge strength image may be used.

  The control unit 8 determines, for each pixel, whether the pixel value of the edge intensity image is larger than the threshold Th1, and when an affirmative determination is made, determines that the pixel indicates the contour of the defect candidate. . Thereby, defect candidates d1 and d2 can be detected. For example, the threshold value Th1 may be set in advance and stored in a storage medium of the control unit 8.

<Second candidate detection process>
The control unit 8 may perform a second candidate detection process described below in place of the above-described first candidate detection process or together with the first candidate detection process.

  Now, if the defect candidate d1, d2 is, for example, a black colored deposit, the pixel value of the region corresponding to the defect candidate d1, d2 is smaller than the pixel value of the other region. Therefore, the control unit 8 may detect a defect as follows. That is, the control unit 8 determines for each pixel whether the pixel value in the tablet region TR of the image IM12 is smaller than the threshold Th2, and when an affirmative determination is made, the pixel indicates a defect candidate. It is determined that For example, the threshold value Th2 may be set in advance and stored in a storage medium of the control unit 8.

  Hereinafter, a pixel indicating a defect candidate (or an outline thereof) is also referred to as a candidate pixel.

<Specific example of defect authenticity processing>
FIG. 12 is a flowchart showing an example of the defect authenticity process. Here, an example in which it is determined whether or not a defect candidate in the main surface area Ra indicates a defect of the tablet 9 will be described. In step S31, the control unit 8 determines whether at least one defect candidate is common to two or more of the images IM11 to IM14. Specifically, the control unit 8 determines whether the defect candidate is common based on the correspondence between the images IM11 to IM14 of the position of each pixel with respect to the main surface area Ra. Therefore, the control unit 8 first obtains this correspondence. That is, the control unit 8 obtains the correspondence relationship of the pixels indicating the same point on the major surface 9 a of the tablet 9 in the images IM11 to IM14.

  Such correspondence can be obtained, for example, as follows. For example, the control unit 8 generates bird's eye view images IM41 to IM44 based on the images IM11 to IM14, respectively. The bird's-eye view image referred to here is an image obtained by viewing the tablet 9 along the Z-axis direction. For example, the control unit 8 calculates the imaging direction of the image IM11 based on the ratio of the major axis A and the minor axis B of the ellipse E1 in the image IM11. The imaging direction can be represented by an angle θ (sin θ = B / A) formed by the imaging direction and the horizontal plane. Note that the angle θ does not necessarily have to be calculated, and may be set in advance according to the arrangement of the internal configuration of the imaging head 5.

  The control unit 8 performs image conversion on the image IM11 based on the angle θ to generate a bird's-eye view image IM41. The shape of the outline of the main surface 9a in this bird's-eye view image IM41 is a circle whose diameter is the long axis of the ellipse E1. Similarly, the control unit 8 generates bird's-eye view images IM42 to IM44 based on the images IM12 to IM14. Thereby, in the bird's-eye view images IM41 to IM44, the appearance of the tablet 9 seen from the Z-axis direction is captured. FIG. 13 is a diagram schematically showing an example of bird's-eye view images IM41 to IM44. Hereinafter, the main surface 9a of the tablet 9 in the bird's-eye view images IM41 to IM44 is called a main surface 9ab in order to be distinguished from the main surface 9a of the actual tablet 9.

  When the sizes of the circles F1 indicating the peripheral edge of the main surface 9ab of the tablet 9 are largely different from each other in the bird's-eye view images IM41 to IM44, the control unit 8 determines that the difference between the sizes of these circles F1 is larger than a predetermined value. The bird's eye view images IM41 to IM44 may be enlarged or reduced so as to be smaller. In such bird's-eye view images IM41 to IM44, pixels at the same position in the circle F1 indicate the same point on the main surface 9a of the tablet 9. That is, FIG. 13 substantially shows the correspondence of the respective pixels between the main surface regions Ra of the images IM11 to IM14.

  The control unit 8 determines whether or not the defect candidate for the main surface area Ra matches in two or more bird eye images of the bird eye images IM41 to IM44. In the present application, "coincidence" does not necessarily mean perfect coincidence, and includes a state in which the difference is smaller than a predetermined degree. If the defect candidates match each other in two or more bird eye images, the control unit 8 determines that the defect candidates (for example, defect candidate d1) are common in the two or more images, and the tablet 9 (more Specifically, it is determined that a defect has occurred on the main surface 9a).

  The match determination of the defect candidate may be performed as follows. For example, the control unit 8 determines whether or not the positions (positions with respect to the peripheral area Ra) of the plurality of candidate pixels constituting the defect candidate match each other in two or more images. For example, in each of the images IM11 to IM14, the defect candidate d1 is assumed to be configured by candidate pixels Q1 [0] to Q1 [10]. Control unit 8 determines whether or not the positions of candidate images Q1 [n] (n: 0 to 10) with respect to main surface region Ra coincide with each other in two or more of images IM11 to IM14. The control unit 8 may determine that the defect candidates match when these match. According to this, it is possible to judge the match of the defect candidate based on the position and the shape of the defect candidate. Of course, all the candidate pixels constituting the defect candidate do not have to match each other between two or more images, and some candidate pixels may have a mismatch. The more the number of candidate pixels that allow the mismatch, the looser the condition on the shape.

  For such coincidence determination, the control unit 8 is required to specify candidate pixels constituting each defect candidate in each of the images IM11 to IM14. This can be done, for example, as follows. That is, for example, in each of the images IM11 to IM14, the control unit 8 calculates the distance between candidate pixels and extracts candidate pixel groups having a small distance between each other as the same defect candidate. Thus, for example, candidate pixels Q1 [0] to Q1 [10] belonging to defect candidate d1 in each of images IM11 to IM14 are identified, and candidate pixels Q2 [0] to Q2 [10] belonging to defect candidate d2 in image IM12. Is identified.

  Here, the positions of the candidate pixels Q1 [n] with respect to the main surface area Ra coincide with one another in the plurality of images IM11 to IM14. Therefore, the control unit 8 determines that the defect candidate d1 is a defect of the tablet 9, and determines that the tablet 9 has a defect. On the other hand, since the positions of the candidate pixels Q2 [n] with respect to the main surface area Ra do not match among the plurality of images IM11 to IM14, the control unit 8 determines that the defect candidate d2 is not a defect of the tablet 9.

  If a negative determination is made in step S31, the control unit 8 determines in step S33 that no defect has occurred in the tablet 9 (more specifically, the major surface 9a).

  In addition, when it is not necessary to detect the number of defects, and it is sufficient to simply detect the presence or absence of the defects of the tablet 9, it is not necessary to make the determination for all the defect candidates. If it is determined that one defect candidate is shared by two or more pixels, the determination on another defect candidate may be omitted. On the other hand, when detecting the number of defects generated in the tablet 9, the determination may be performed for all defect candidates.

  In the above-described example, since the outline of the main surface area Ra is obtained based on a function determined in advance as the shape of the outline of the main surface area Ra, the specification accuracy of the main surface area Ra is high as described above. According to this, it is possible to improve the accuracy of the correspondence between the images of the positions of the respective pixels in the main surface region Ra, and to improve the accuracy in the determination of the defect true or false.

<Side area>
In the above-described example, the appearance inspection on the major surface 9 a of the tablet 9 has been described. Here, an appearance inspection on the side surface 9 c of the tablet 9 will be described. On the side surface 9 c of the tablet 9, regions corresponding to the imaging direction appear in the images IM 11 to IM 14, respectively. For example, referring to FIG. 5, region 9c1 of side surface 9c of tablet 9 is shown in images IM11 to IM13 and not shown in image IM14. The area 9c1 has a width substantially equal to the area obtained by dividing the side surface 9c of the tablet 9 in the circumferential direction, for example, into six equal parts. In the example of FIG. 5, the area 9c1 is located approximately at the center of the side surface 9c in the image IM11 and at the end of the side surface 9c in the images IM12 and IM13. Therefore, the area 9c1 appears wide in the image IM11, and appears very narrow in the images IM12 and IM13. Therefore, the defect in the area 9c1 of the tablet 9 is less likely to be visually recognized in the images IM12 and IM13 than the image IM11, and is not easily detected as a defect candidate.

  Therefore, the area 9c1 at the end of the side face 9c in the images IM12 and IM13 may be excluded from the target of the candidate detection process. In other words, the control unit 8 may set an area on the end side of the side surface area Rc in the images IM12 and IM13 as a mask area to be not inspected. According to this, it is possible to prevent false detection of a defect candidate in the end region or the occurrence of defect detection failure.

  FIG. 14 is a flowchart showing an example of the inspection process. In the example of FIG. 14, step S28 is further executed as compared with FIG. This step S28 is performed between steps S26 and S27. In step S28, the control unit 8 sets a mask area. For example, the control unit 8 sets end regions located at both ends of the side region Rc identified in step S26 as a mask region. The width of the end area (the width along the semi-ellipse E11) may be determined in advance, or the control unit 8 may calculate a predetermined ratio of the width of the side area Rc as the width of the end area.

  The control unit 8 does not detect a pixel in the mask area as a defect candidate in the candidate detection process of step S27. According to this, the area 9c1 of the tablet 9 is subjected to the substantial candidate detection process only in the image IM11. This is because the area 9c1 does not appear in the image IM14, but appears in the mask area in the images IM12 and IM13.

  The control unit 8 performs defect authenticity processing on the area 9c1 as follows. That is, when the defect candidate is detected in the area 9c1 by the candidate detection process for the image IM11, the control unit 8 selects the tablet 9 (more specifically, regardless of the result of the candidate detection process for the other images IM12 to IM14. It is determined that a defect has occurred in the region 9c1) of the side surface 9c. That is, for the area 9c1 where candidate detection processing is performed only with the image IM11, candidate detection processing is not performed with the other images IM12 to IM14. Therefore, if a defect candidate is detected in the image IM11, the defect candidate is the tablet 9 It is regarded as a defect. Thereby, a defect in the region 9c1 of the side surface 9c can be detected.

  Referring to FIG. 5, a region 9c2 adjacent to the region 9c1 of the side surface 9c of the tablet 9 in the circumferential direction is shown in the images IM11 and IM12 and is not shown in the images IM13 and IM14. The area 9c2 has, for example, the same width as an area obtained by dividing the side surface 9c of the tablet 9 in the circumferential direction into twelve equal parts. The area 9c2 is located on the end side of the center of the side surface 9c in the images IM11 and IM12, but is away from the end of the side surface 9c. In the images IM11 and IM12, the area 9c2 is not set as the mask area, and corresponds to the inspection target area.

  The control unit 8 performs defect authenticity processing on the area 9c2 as follows. That is, when the common candidate detection in the area 9c1 is detected by the candidate detection process on the images IM11 and IM12, the control unit 8 selects the tablet 9 (more specifically, regardless of the result of the candidate detection process on the images IM13 and IM14). In fact, it is determined that a defect has occurred in the region 9c2) of the side surface 9c. That is, with respect to the area 9c2 that is captured only in the images IM11 and IM12, candidate detection processing is not performed in the other images IM13 and IM14, so when a common defect candidate is detected in the area 9c2 of the images IM11 and IM12, the defect The candidate is regarded as a defect of the tablet 9. Thereby, a defect in the area 9c2 of the side surface 9c can be detected with high accuracy.

  In the above example, although the areas 9c1 and 9c2 have been described, the same applies to other areas. In short, the control unit 8 sets the first region of the tablet 9 in which the candidate detection process is performed only on one image, and the second region of the tablet 9 in which the candidate detection process is performed in a plurality of images. It is sufficient to carry out defect authenticity processing as in That is, the control unit 8 determines that a defect has occurred in the tablet 9 when a defect candidate is detected in the first region of the one image for the first region, and a plurality of second regions are detected. It is determined that the tablet 9 has a defect when a common defect candidate is detected in the second region of two or more images of the images.

  FIG. 15 is a flowchart showing an example of defect authenticity processing for the side surface area. Compared to FIG. 12, step S34 is further executed, and the conditions for executing step S32 are different. Step S34 is executed when a negative determination is made in step S31. In step S34, the control unit 8 determines whether the area corresponding to at least one defect candidate in the side surface area Rc is not a target of defect detection in all other images (that is, not a target of candidate detection processing). To judge.

  The correspondence between the images IM11 to IM14 of the position of each pixel with respect to the side surface region Rc can be obtained in the same manner as the main surface region Ra. For example, four images of the tablet 9 viewed from the side are generated based on the images IM11 to IM14, respectively. In the four images, the side surface 9c of the tablet 9 seen from the four directions is shown. The correspondence between the four images of the pixels in the side surface region Rc may be set in advance.

  For example, the case where a defect candidate is detected in the area 9c1 of the image IM11 will be described. This area 9c1 is not targeted for defect detection in all the other images IM12 to IM14. For example, the area 9c1 is a mask area in the images IM12 and IM13, and does not originally exist in the image IM14. Therefore, in this case, an affirmative determination is made in step S34. At this time, in step S32, the control unit 8 determines that the tablet 9 (more specifically, the side surface 9c) has a defect. That is, even if the defect candidate is not common to two or more images, if the defect candidate is originally another image outside the object of defect detection, the defect candidate is not naturally detected in the other image . Therefore, in this case, the defect candidate detected in one image is regarded as the defect of the tablet 9.

  If a negative determination is made in step S34, the control unit 8 determines in step S33 that the side surface 9c of the tablet 9 is not defective.

  When it is not necessary to detect the number of defects, and it is sufficient to simply detect the presence or absence of defects in the tablet 9, the determination may not necessarily be performed on all the defect candidates in each of steps S31 and S32. Further, if it is determined that a defect has occurred in the tablet 9 in one of the defect authenticity processing (FIG. 12) for the main surface area (FIG. 12) and the defect authenticity processing for the side surface area (FIG. 15), the other may be omitted.

  In the case of detecting the number of defects generated in the tablet 9, determination may be made for all defect candidates. That is, regardless of the determination result in step S31, step S34 is necessarily performed, and determination on all defect candidates may be performed in each of steps S31 and S32. Further, in this case, both the defect authenticity processing for the main surface area and the defect authenticity processing for the side surface area are performed.

<Boundary edge>
In the above-described example, the edge pixel first detected in the search in each row of the search area R1 is grasped as a component of the boundary edge P3. In this case, for example, if a defect occurs in the boundary between the main surface 9aa and the side surface 9ca of the tablet 9 in the image IM12, the calculation accuracy of the ellipse E1 may be reduced. The details will be described below.

  FIG. 16 is a view schematically showing an example of an edge in an edge image when a crack is generated in the boundary between the major surface 9 aa and the side surface 9 ca of the tablet 9. As illustrated in FIG. 16, an edge P ′ is detected along the outline of the chipping generated in the tablet 9. In the example of FIG. 16, the edge P 'is schematically shown by a thick line. According to the above-described search processing, pixels on the edge on the line L1 side among the edge P 'are grasped as components of the boundary edge P3. In the example of FIG. 16, some of the pixels grasped as components of the boundary edge P3 are schematically indicated by black circles. That is, the component of the boundary edge P3 is biased to the edge Pc 'on the line L1 side of the edge P'. Therefore, the semi-ellipse E11 approximating the boundary edge P3 is likely to deviate from the periphery of the major surface 9aa of the tablet 9. That is, the identification accuracy of the main surface area Ra may be reduced.

  Here, it is intended to further improve the identification accuracy of the main surface region Ra. Specifically, paying attention to the distribution of pixel values of the defect area in the image IM12, the constituent elements of the boundary edge P3 are dispersed into an edge Pc ′ of the edge P ′ and an edge Pa ′ on the line L2 side. , It is intended to identify the boundary edge P3.

  The chipped surface of the tablet 9 is rougher than the surface roughness of the major surface 9a and the side surface 9c, and the luminance component in the defect area in the image IM12 is dispersed. Therefore, the luminance component also varies in the outline (edge P ') of the defect area. Therefore, the pixel values on the edge Pa 'in the image IM12 also vary within a certain range, and the pixel values on the edge Pc' also vary within the same range.

  Therefore, the control unit 8 searches for pixels that are within the search area R1 of the edge image, and for which the pixel value in the corresponding image IM12 is larger than a predetermined threshold (hereinafter referred to as a brightness threshold). The brightness threshold is set in advance in accordance with the brightness of the major surface 9 aa of the tablet 9. For example, the brightness threshold is preset to a value within the above range. Such brightness threshold can be set, for example, by simulation or experiment. The brightness threshold may be stored in, for example, a storage medium of the control unit 8. Hereinafter, an example of a specific operation will be described.

  FIG. 17 is a flowchart showing an example of a search process by the control unit 8. Compared to FIG. 11, the control unit 8 further executes step S57. This step S57 is executed when an affirmative determination is made in step S52. In step S57, the control unit 8 determines whether the pixel value of the Nth row and Mth pixel in the search area R1 of the image IM12 is larger than the brightness threshold.

  When a negative determination is made in step S57, the control unit 8 executes step S53, and when a positive determination is made, the control unit 8 executes step S54.

  According to this, even if it is an edge pixel in the search area R1, if the pixel value of the pixel of the image IM12 corresponding to the edge pixel is smaller than the brightness threshold, the edge pixel is the boundary edge P3. If the pixel value of the pixel of the image IM12 is larger than the brightness threshold, the edge image is recognized as a component of the boundary edge P3.

  As a result, pixels grasped as components of the boundary edge P3 are dispersed on the edge P '. FIG. 18 is a view schematically showing an example of an edge in an edge image when a chipping occurs in the boundary between the major surface 9 aa and the side surface 9 ca of the tablet 9. Also in the example of FIG. 18, some of the pixels grasped as the boundary edge P3 are schematically shown by black circles. As illustrated in FIG. 18, the pixels are dispersed at the edges Pa 'and Pc' of the edge P '. In other words, the brightness threshold is set so that the pixel grasped as the boundary edge P3 is not biased to only one side at the edge P '. As a more specific example, the brightness (for example, the average value, median value, maximum value or minimum value of pixel values (or brightness values) of the main surface 9 aa of the tablet 9) applies to the side surface 9 ca of the tablet 9. When the brightness is higher than the brightness, the brightness of the main surface 9 aa is adopted as the brightness threshold, and when the brightness of the main surface 9 aa of the tablet 9 is lower than the brightness of the side surface 9 ca of the tablet 9, the brightness The brightness of the side 9 ca is adopted as the threshold.

  A semi-elliptic E11 approximating such boundary edge P3 is a semi-elliptic closer to the periphery of the major surface 9aa of the tablet 9. As a result, the main surface region Ra can be identified with higher accuracy.

<Center line of search area>
In the above example, the center line L0 of the search area R1 is described as a line obtained by moving the side outer edge P1, but the side outer edge P1 may be enlarged at a predetermined magnification during this movement. Strictly speaking, the boundary edge P3 is larger than the side outer edge P1 at a predetermined magnification.

<Mask area>
In the above example, the control unit 8 sets the end region of the side surface region Rc as the mask region in each of the images IM11 to IM14. However, other regions may be set as mask regions. Here, for the preferable description, a sugar-coated tablet is adopted as the shape of the tablet 9 and described. The tablet 9 has a flat shape in which spheres are reduced in a uniaxial direction, and the tablet 9 has a circular shape when viewed perpendicularly to the widest surface. The shape of the tablet 9 is not limited to this.

  FIG. 19 schematically shows an example of a captured image IM1 '. The captured image IM1 'is an image obtained by imaging the tablet 9 of a sugar-coated tablet by the imaging head 5. Images IM11 'to IM14' of the captured image IM1 'show the appearance of the tablet 9 as viewed from different imaging directions. The contour of the tablet 9 in each of the images IM11 'to IM14' ideally has an elliptical shape.

  In the example of FIG. 19, the correspondence of each point on the surface of the tablet 9 in the images IM11 'to IM14' is indicated by a two-dot chain line. Such a correspondence relationship may be set in advance in association with, for example, a function indicating the outline of the tablet 9, and may be stored in the storage medium of the control unit 8.

  In the example of FIG. 19, mask areas MR1 to MR4 are set in the images IM11 'to IM14'. The mask regions MR1 to MR4 are regions in which the pixel values can be much higher than those in the other regions. Such a region is generated by specular reflection of light emitted from the illumination light source by the tablet 9. That is, when the specularly reflected light is imaged on a part of the imaging surface of the inspection camera 51 through the mirror 53, the pyramidal mirror 54 and the lens group 52, an area corresponding to the part of the imaging surface The pixel value of the pixel in the regular reflection region is much higher than the other pixel values. Therefore, the mask area is set to include this regular reflection area. Thereby, it can be avoided that the intensity of light on the tablet 9 is erroneously detected as a defect.

  Here, since a ring-shaped light source is assumed as the illumination light source, in the example of FIG. 19, the mask regions MR1 to MR4 are shown as elliptical ring regions. The mask area MR1 may be set in advance in association with a function indicating the outline of the tablet 9, for example. Alternatively, for example, the control unit 8 may specify an area where the pixel value of the pixel in the image IM11 'is larger than the mask threshold, and set the area as the mask area MR1. The same applies to mask regions MR2 to MR4. The mask threshold may be set in advance and stored in a storage medium of the control unit 8, for example.

  Since tablets 9 are imaged from a plurality of imaging directions, the specular reflection areas corresponding to the respective imaging directions are different from each other on the surface of the tablet 9. Of course, there are cases where portions of the specular reflection areas corresponding to the two imaging directions overlap on the surface of the tablet 9. However, three or more specular reflection areas rarely overlap each other in the same area on the surface of the tablet 9. That is, when the mask areas MR1 to MR4 are projected on the surface of the tablet 9, although two of the mask areas MR1 to MR4 overlap on the surface of the tablet 9, it is rare that three overlap in the same area. It is even less likely that four will overlap in the same area. If the mask regions MR1 to MR4 overlap in the same region, the regions are included in the mask regions MR1 to MR4 in the images IM11 'to IM14', so inspection is performed on any of the images IM11 'to IM14'. It does not become an object. But such a possibility is very low.

  Here, it is assumed that the four mask regions MR1 to MR4 do not overlap in the same region. In this case, the common surface of the tablet 9 that appears in any of the images IM11 'to IM14' is divided into the following four types of regions. That is, the first area is an area located in the inspection target area in all of the images IM11 ′ to IM14 ′, and the second area is located in the inspection target area in only three of the images IM11 ′ to IM14 ′. It is an area, and the third area is an area located in the inspection target area in only two of the images IM11 ′ to IM14 ′, and the fourth area is the inspection target area in only one of the images IM11 ′ to IM14 ′ It is an area located inside.

<Inspection processing>
Next, inspection processing for the sugar-coated tablet 9 will be described. Unlike the tablet 9 of FIG. 2, this tablet 9 does not have a clear boundary between the main surface and the side surface. Therefore, since it is difficult for the control unit 8 to identify the main surface area and the side surface area, the position of the defect candidate is grasped by the position with respect to the tablet area. Therefore, the control unit 8 specifies the contour of the tablet area in the identification process.

  FIG. 20 is a flowchart illustrating an example of the inspection process. In step S211, the control unit 8 generates an edge image as in step S21. Next, in step S212, the control unit 8 specifies the tablet edge in the same manner as in step S22. Next, in step S213, an approximate line (ellipse) approximating the tablet edge is specified as the contour of the tablet region based on the function of the ellipse. If it explains more generally, based on a standard function which shows an outline of tablet 9 in IM11-IM14, an approximation line which approximates a tablet edge is specified. Next, in step S214, the control unit 8 sets a mask area, for example, as shown in FIG.

  Next, the control unit 8 performs candidate detection processing on each of the images IM11 'to IM14'. The candidate detection process is as described above. Thereby, in each image IM11 '-IM14', a defect candidate is detected and the position of the defect candidate with respect to a tablet area | region is pinpointed.

  The control unit 8 operates as follows in the defect authenticity process. That is, the control unit 8 causes the tablet 9 to have a defect when a common defect candidate is detected in two or more of the images IM11 ′ to IM14 ′ in the first to third regions. I will judge. The control unit 8 determines that the tablet 9 has a defect when a defect candidate is detected in one of the images IM11 'to IM14' in the fourth region.

  An example of a specific defect authenticity process is similar to the defect authenticity process for the side surface area described with reference to FIG. Specifically, "side area" and "side of tablet" may be read as "tablet area" and "tablet", respectively.

Modified example.
<Principal surface of tablet>
In the above-mentioned example, the tablet 9 conveyed in the posture where the principal surface 9 b faces the conveyance belt 41 side was described as the inspection object. When the tablet 9 is transported in a posture in which the major surface 9 a of the tablet 9 faces the transport belt 41, the tablet inspection apparatus 1 performs an appearance inspection on the major surface 9 b and the side surface 9 c of the tablet 9.

  For example, after the appearance inspection on the tablet 9, the transport posture of the tablet 9 may be reversed, and the appearance inspection on the tablet 9 may be performed in that state. According to this, an appearance inspection can be performed on the entire surface (main surfaces 9a and 9b and side surface 9c) of the tablet 9.

  As mentioned above, although the tablet inspection method and the tablet inspection device were explained in detail, the above-mentioned explanation is illustration in all the aspects, and this indication is not limited to it. The various modifications described above can be combined and applied as long as they do not contradict each other. And, it is understood that many variations not illustrated may be envisaged without departing from the scope of the present disclosure.

1 tablet inspection device 5 imaging unit (imaging head)
8 Image processing unit (control unit)
9 tablets 9a first main surface (main surface)
9b Second main surface (main surface)
9c Side surface d1, d2 defect candidate IM1, IM1 'Image pickup image IM11 to IM14, IM11' to IM14 Image MR1 to MR4 Mask area P0 Tablet edge P1 Side outer edge P2 Main surface outer edge P3 Boundary edge P23 Main surface edge R1 Search area Ra Main area Rc Side area TR Tablet area

Claims (12)

A tablet inspection method for inspecting the appearance of a tablet, comprising
Imaging the tablet to generate a captured image including a plurality of images in which the appearance of the tablet viewed from the plurality of imaging directions is captured;
Performing an inspection process of detecting a defect candidate of a tablet on the captured image;
When a defect candidate is generated in the tablet when a common defect candidate is detected in two or more of the plurality of images in the first region of the tablet captured in the plurality of images by the inspection process A tablet inspection method comprising the step of determining (c). The tablet inspection method according to claim 1, wherein
The second region of the tablet is included in only n (n is an integer of 2 or more) of the plurality of images.
When a defect candidate common to two or more of the n images is detected in the second region by the inspection process, the step (d) of judging that the tablet has a defect is further included. Have a tablet inspection method. The tablet inspection method according to claim 1 or 2, wherein
The third region of the tablet is captured in only one of the plurality of images,
The tablet inspection method further provided with the process (f) of judging that a defect has arisen in the tablet, when a defect candidate is detected in the 3rd field of the one picture by the inspection processing. A tablet inspection method according to any one of claims 1 to 3, wherein
Mask regions not to be inspected in the inspection process are set for each of the plurality of images;
The fourth area of the tablet is captured in an examination area other than the mask area only in m (an integer of 2 or more) of the plurality of images.
In the step (c),
The tablet inspection method which judges that a defect has arisen in the tablet, when a common defect candidate is detected in the 4th field among two or more pictures of the m pieces by the inspection processing. The tablet inspection method according to claim 4, wherein
The tablet inspection method, wherein in each of the plurality of images, a region including a region where a pixel value is higher by a predetermined value or more than other regions due to regular reflection of light to the tablet is set as the mask region. A tablet inspection method according to claim 4 or claim 5, wherein
The tablet has a first main surface which is the first region, a second main surface facing the first main surface, and a side surface connecting the peripheral edge of the first main surface and the peripheral edge of the second main surface. Have
The tablet inspection method according to claim 1, wherein end regions located on both sides of a side surface area occupied by the side surface of the tablet in each of the plurality of images are set in the mask area. The tablet inspection method according to any one of claims 1 to 6,
The tablet has a first main surface which is the first region, a second main surface facing the first main surface, and a side surface connecting the peripheral edge of the first main surface and the peripheral edge of the second main surface. Have
In the step (b),
Identifying a main surface edge corresponding to an outline of a main surface area indicated by the first main surface of the tablet in each of the plurality of images;
In each of the plurality of images, a step (b2) of finding an approximation line of the main surface edge as the contour of the main surface region based on a function predetermined as a shape of the contour of the main surface region;
Detecting a defect candidate in each of the plurality of images (b3);
In the step (c), the defect candidate detected in the step (b3) is common to the two or more images based on the correspondence between the plurality of images of the position of each pixel with respect to the main surface region. A tablet inspection method to determine whether to The tablet inspection method according to claim 7, wherein
The first main surface of the tablet has a substantially circular shape in plan view,
The tablet inspection method, wherein the function is a function indicating an ellipse, and the approximate line is an ellipse. A tablet inspection method according to claim 7 or 8, wherein
The tablet has a substantially disc shape,
In each of the plurality of images, the side surface area occupied by the side surface of the tablet and the main surface area constitute a tablet area,
When each of the plurality of images is referred to as an examination image,
In the step (b1),
Performing an edge detection process on the inspection image to generate an edge image (b11);
Identifying a tablet edge corresponding to the contour of the tablet area in the inspection image from the edge image (b12);
A main surface outer edge and a side surface outer edge respectively corresponding to a part of the periphery of the first main surface which is a part of the contour of the tablet area in the inspection image and the periphery of the second main surface in the inspection image Extracting from the tablet edge (b13),
In the edge image, a pixel in a search area which is separated by a predetermined distance from the side outer edge along a predetermined direction is searched to specify a boundary edge corresponding to a boundary between the main area and the side area. A process (b14),
Identifying a pair of the main surface outer edge and the boundary edge as the main surface edge (b14). The tablet inspection method according to claim 9,
In the step (b14),
The tablet inspection method which is a pixel in the search area of the edge image, and searches for a pixel whose pixel value of the inspection image corresponding to the pixel is larger than a predetermined threshold to specify the boundary edge. The tablet inspection method according to claim 9 or 10, wherein
In the step (b14),
The tablet inspection method which searches the pixel from said side face outer edge toward said main surface outer side edge in said search area, and specifies said border edge. A tablet inspection apparatus for inspecting the appearance of a tablet, comprising:
An imaging unit configured to image a tablet and generate a captured image including a plurality of images in which the appearance of the tablet viewed from the plurality of imaging directions is captured;
And an image processing unit,
The image processing unit
Performing an inspection process for detecting a defect candidate of a tablet on the captured image;
When a defect candidate is generated in the tablet when a common defect candidate is detected in two or more of the plurality of images in the first region of the tablet captured in the plurality of images by the inspection process Tablet inspection device which performs and the process to judge.

Images