JP2004101339A - Visual inspection apparatus of round rod body - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a visual inspection apparatus capable of easily, reliably detecting a defect such as a flaw and dirt on the surface of a cigarette. <P>SOLUTION: The visual inspection apparatus comprises: a rotary mechanism for allowing the cigarette to rotate on its axis at a prescribed position; and a line sensor 9 that images the circumferential surface of the round rod body rotating on its axis by the rotary mechanism along the axial direction for each line for obtaining an appearance image over the entire circumference of the round rod body. Then, the visual inspection apparatus forms a misregister image by giving prescribed misalignment to an image to be inspected that is obtained by the line sensor 9 and is subjected to visual inspection, finds the misregister image between the deviation image and the image to be inspected (an image processing section 12), and judges each pixel level in the differential image for extracting a visual defect site in the round rod body by a host computer 13. <P>COPYRIGHT: (C)2004,JPO

Description

[0001]
TECHNICAL FIELD OF THE INVENTION
BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a round bar body appearance inspection apparatus that can easily and reliably detect defects such as scratches and dirt on the surface of a round bar body made of cigarettes or the like.
[0002]
[Related background art]
In producing cigarettes, it is important to maintain the production quality sufficiently high. One of the items for evaluating the production quality of a cigarette is that there is no scratch or dirt on the outer peripheral surface. In order to detect such a defect in appearance, the outer peripheral surface of the cigarette is exclusively imaged by a camera, and the appearance image (inspection target image) and a previously obtained non-defective (standard) image (standard image) are obtained. This is performed by executing image processing such as comparison. (For example, see Patent Document 1)
[0003]
[Patent Document 1]
JP-A-5-180777
[0004]
[Problems to be solved by the invention]
However, when the outer peripheral surface of a cigarette is imaged to obtain an appearance image thereof, the image signal level itself may fluctuate due to fluctuations of an illumination light source or the like. Moreover, the scratches and stains present on the outer peripheral surface of the cigarette often resemble the color components of the base. Therefore, there is a problem that the image processing itself becomes complicated, and it is difficult to sufficiently increase the inspection accuracy. Furthermore, since it takes time to perform image processing until the presence or absence of a defect is determined from one inspection target image, there is a problem in continuously inspecting the appearance of many cigarettes.
[0005]
The present invention has been made in view of such circumstances, and its purpose is to easily and reliably detect defects such as scratches and dirt on the surface of a round bar made of cigarettes, and It is an object of the present invention to provide a round rod appearance inspection apparatus suitable for speeding up the inspection processing.
[0006]
[Means for Solving the Problems]
In order to achieve the above-described object, the round bar appearance inspection apparatus according to the present invention is suitable for detecting a scratch or dirt on the outer peripheral surface of a cigarette or the like from its surface image and eliminating defective products. hand,
For example, a rotating mechanism for rotating a round bar made of a cigarette at a predetermined position, and an image of the circumferential surface of the round bar rotated by the rotating mechanism taken one line at a time along the axial direction of the round bar. A line sensor that obtains an appearance image over the entire circumference of the
In particular, a predetermined positional deviation is given to the inspection target image obtained by the line sensor and subjected to the appearance inspection to form a deviation image, and a difference image between the deviation image and the inspection target image is obtained (difference image formation). Means), a pixel level of the difference image is determined, and an appearance defect portion of the round bar is extracted (defect portion determination means).
[0007]
That is, the present invention obtains, as a difference image, a difference between an inspection target image obtained by imaging the outer peripheral surface of the round bar and a displacement image obtained by shifting (shifting) the inspection target image by a predetermined number of pixels. A pixel portion having a relative difference with respect to a base pixel signal component of the inspection target image is reliably extracted. In particular, the image signal component of the base of the inspection target image is canceled, and a defective portion such as a scratch or dirt is regarded as a region of a pixel whose image signal component is relatively changed from the base, thereby causing illumination fluctuation or the like. A defect portion such as a scratch or dirt can be detected without being concerned with the overall level change of the inspection target image caused by the defect.
[0008]
Further, in addition to the above-described functions, the appearance inspection apparatus for a round bar according to the present invention further includes a known image showing characters / graphics present on the surface of the round bar and an appearance image obtained by the line sensor. Pattern matching means for pattern-matching the pixels forming the characters / graphics included in the appearance image, and pixels forming the characters / graphics detected by the pattern matching means as background pixel components. And a pre-processing means for forming an inspection target image to be subjected to a visual inspection by filling in the image.
[0009]
Incidentally, the known image is composed of a plurality of partial known images obtained by dividing the existence area of the character / graphic on the surface of the round bar into a plurality of regions in the circumferential direction of the round bar, and the pattern The matching means is configured to match an image of an area where the character / figure may exist in the appearance image with the plurality of partial known images to obtain a degree of matching between the images.
[0010]
At this time, it is preferable that the inspection target image is obtained as an appearance image over a range larger than the entire circumference of the round bar, for example, over 1.2 to 1.5 rotations. In this way, even if the cut-out position of the appearance image (image to be inspected) is not defined in the circumferential direction of the round bar, in combination with the above-described division of the known image in the circumferential direction, Since the appearance image can be surely subjected to the pattern matching processing over the entire circumference of the round bar, it is possible to reliably detect the above-described character / figure existence area in the appearance image.
[0011]
BEST MODE FOR CARRYING OUT THE INVENTION
Hereinafter, an apparatus for inspecting the appearance of a round bar body according to an embodiment of the present invention will be described by taking an example of an apparatus for inspecting the appearance of a cigarette and eliminating defective products.
FIG. 1 is a view schematically showing the overall schematic configuration of a cigarette appearance inspection apparatus. Reference numeral 1 denotes a cigarette transport mechanism for aligning a plurality of cigarettes side by side and transporting them one by one continuously at predetermined intervals. Department. The cigarette transport mechanism unit 1 transports a plurality of cigarettes put into the hopper 2 in parallel with each other at predetermined intervals by aligning their axial directions, and transports the cigarettes one by one from the hopper drum 3. And two-stage transfer drums 5a and 5b for transferring the produced cigarette to a rolling drum 4 forming a part of a cigarette rotating mechanism. The cigarette sent from the rolling drum 4 is further conveyed to the outlet (discharge chute) via the transfer drums 5c, 5d, 5e, and 5f in order.
[0012]
Incidentally, each of the drums 3, 5a, 5b, to 5f is provided with a plurality of holding grooves (not shown) for adsorbing and holding cigarettes along the axial direction of the drum circumferential surface at equal intervals in the circumferential direction. It consists of things. These holding grooves suck and hold cigarettes by being suction-driven at predetermined rotation positions for each of the drums 3, 5a, 5b, to 5f. Then, when each of the drums 3, 5a, 5b, to 5f having sucked the cigarette rotates by a predetermined angle, the suction system of the holding groove is opened, whereby the holding of the cigarette by the holding groove is released. By the rotation of each of the drums 3, 5a, 5b, to 5f and the suction control of each of the holding grooves, the cigarette is sequentially conveyed by a predetermined distance along the outer peripheral surface of each of the drums 3, 5a, 5b, to 5f. You.
[0013]
The drums 3, 5a, 5b, to 5f are rotationally driven in synchronization with each other under the control of the machine control unit 6. Further, as the drums 3, 5a, 5b, to 5f rotate, the suction of the cigarettes by the holding grooves is respectively controlled. Through these series of controls, cigarettes are conveyed one by one at a predetermined timing and subjected to the appearance inspection described below.
[0014]
The above-described rolling drum 4 forms a cigarette rotating mechanism that forms a pair with a kick roller 7 that is provided in parallel with the peripheral surface of the rolling drum 4 and that holds the cigarette at a predetermined position and rotates. That is, the cigarette rotating mechanism includes a rolling drum 4 that is driven to rotate in the transport direction of the cigarette C and a kick that is driven to rotate in a direction opposite to the rolling drum 4, as shown in an enlarged manner in FIG. And a roller 7. The rolling drum 4 and the kick roller 7 have their outer peripheral surfaces provided in parallel with a gap slightly smaller than the diameter of the cigarette C.
[0015]
In particular, on the peripheral surface of the rolling drum 4, a first holding groove 4a and a second holding groove 4b for holding the cigarette C are provided alternately and parallel at a predetermined distance in the circumferential direction. Between the first holding groove 4a and the second holding groove 4b is a rolling surface 4c for rotating the cigarette C on its own axis. The peripheral surface of the kick roller 7 is a rolling surface 7a for holding the cigarette C between itself and the rolling surface 4c of the rolling drum 4 and rotating the cigarette C on its own. An escape groove 7b for releasing the holding of the cigarette C with the drum 4 is provided.
[0016]
The rolling drum 4 and the kick roller 7 having such a peripheral shape are rotationally driven at the same peripheral speed in synchronization with each other. Then, the cigarette C sucked and held by the first holding groove 4a of the rolling drum 4 and transported to a position facing the kick roller 7 is moved to the first holding groove at the front end edge of the rolling surface 7a of the kick roller 7. It is dispensed from 4a and inserted between the rolling surfaces 4c and 7a. With the rotation of the rolling drum 4 and the kick roller 7, the cigarette C sandwiched between the rolling surfaces 4c and 7a is rotated at the sandwiching position. The rotation of the cigarette C is performed over, for example, 1.2 to 1.5 turns by rolling over the entire area of the rolling surfaces 4c and 7a. The cigarette C that has rolled over the entire area of each rolling surface 4c, 7a falls into the second holding groove 4b and at the same time falls into the relief groove 7b, thereby being released from the above-described trapping between the rolling surfaces 4c, 7a. Are further transported as the rolling drum 4 rotates.
[0017]
On the other hand, as described above, the cigarette C is held at a predetermined position, and for example, the cigarette C is rotated over a period of 1.2 to 1.5 turns. Illumination light sources 8a and 8b for illuminating the outer peripheral surface of the cigarette C and a line sensor 9 for imaging the peripheral surface of the cigarette C line by line along its axial direction are provided. The line sensor 9 is a color sensor that outputs the image signal as an RGB signal. In particular, the line sensor 9 is driven at a predetermined timing in synchronization with the rotation of the rolling drum 4, so that the line sensor 9 sequentially moves one line along the axial direction along the circumferential surface of the cigarette C in synchronization with the rotation of the cigarette C. The outer appearance image is obtained over the entire circumference of the cigarette C, and in particular, the outer appearance image over 1.2 to 1.5 rotations is obtained.
[0018]
The appearance image of the cigarette C captured by the line sensor 9 in this manner is given to the appearance inspection unit 10 mainly composed of an image processor described below, and the appearance inspection is performed.
The appearance inspection unit 10 is provided in parallel with the first image processing unit 11 for inputting the appearance image of the cigarette C captured by the line sensor 9 as shown in FIG. The image processing apparatus includes a plurality of second image processing units 12a, 12b,..., And 12n that perform image processing on the appearance images distributed by the image processing unit 11 and execute the appearance inspection.
[0019]
The first image processing unit 11 basically combines the image signals obtained by the line sensor 9 one by one for one to two lines over the 1.2 to 1.5 turns of the cigarette C. The appearance image over the entire circumference of C is required. Each time an appearance image over the entire circumference of the cigarette C is obtained, the first image processing unit 11 cyclically distributes the appearance image to the plurality of second image processing units 12a, 12b,. are doing. In each of the second image processing units 12a, 12b, to 12n, each time one appearance image (processing target image) of the cigarette C is given from the first image processing unit 11, the appearance image is added to the appearance image. The appearance inspection is performed by performing predetermined image processing described later.
[0020]
That is, in this embodiment, the image in each of the second image processing units 12a, 12b, to 12n is compared with a cycle (inspection cycle) for sequentially obtaining one appearance image of the cigarette C via the line sensor 9. Since the time required for processing (appearance inspection) generally takes a long time, the appearance image is sequentially distributed to the plurality of second image processing units 12a, 12b, to 12n. The second image processing units 12a, 12b, to 12n process the plurality of appearance images obtained at predetermined intervals in parallel by the second image processing units 12a, 12b, to 12n, thereby sequentially capturing the outer peripheral surface of the cigarette C. The appearance inspection of each cigarette C is reliably executed without delaying the transport cycle of C.
[0021]
The appearance inspection results executed in the second image processing units 12a, 12b, to 12n are provided to the host computer 13 in association with the cigarettes C. When an abnormal appearance (defective) is detected, the host computer 13 issues an exclusion command to the machine control unit 6 described above. In synchronization with the transport of the cigarette C via the cigarette transport path constructed by the transfer drums 5c, 5d, 5e and 5f described above, the defective article removing unit 6a incorporated in the cigarette transport path is activated and the corresponding cigarette is activated. (Defective product) C is eliminated.
[0022]
The transport time required for the cigarette C imaged by the line sensor 9 to reach the defective product removing unit 6a is managed in synchronization with the drive of the transport system by the machine control unit 6. Therefore, the second image processing devices 12a, 12b,..., And 12n that perform image processing on the appearance image of the cigarette C in relation to the imaging timing of the cigarette C by the line sensor 9 are managed. Only the cigarette C can be easily and reliably removed.
[0023]
Further, as described above, the inspection result given to the host computer 13 is stored in a main memory (not shown) provided in the host computer 13 together with the appearance image of the cigarette C. The appearance images registered in the main memory and the inspection results are stored in the image file (hard disk) 14 as required, for example, when a predetermined number of appearance inspections are completed. The appearance image and the inspection result stored in the image file 14 are appropriately read out by the image processing unit 11 and analyzed again.
[0024]
Here, the image processing (appearance inspection) on the appearance image of the cigarette C in the above-described appearance inspection unit 10 will be described in more detail. The line sensor 9 is driven at a predetermined timing in synchronization with the rotation of the rolling drum 4 as described above under the control of the synchronization circuit 9a, and sequentially captures the outer peripheral surface of the cigarette C one line at a time along its axial direction. . At this time, the cigarette C is rotated by the rolling drum 4 and the kick roller 7 at the same position for 1.2 to 1.5 turns as described above. The line sensor 9 takes an image of the entire circumference in synchronization with the rotation of the cigarette C. Therefore, the line sensor 9 obtains an image signal composed of 1024 pixels per line along the axial direction of the cigarette C over 300 lines corresponding to 1.2 turns of the cigarette C as shown in FIG. Will be. The image signal composed of [1024 pixels × 300 lines] is subjected to an appearance inspection as an appearance image of the cigarette C.
[0025]
Therefore, as shown in FIG. 5, the appearance inspection unit 10 first captures an image of the outer peripheral surface of the cigarette C using the line sensor 9 over 1.2 turns thereof, and FIG. The appearance image as shown is obtained [Step S1], and the appearance image is distributed to one of the plurality of second image processing units 12a, 12b to 12n as described above [Step S2]. Then, for each of the image processing units 12a, 12b, to 12n, image processing including a plurality of types of items described below is executed.
[0026]
This image processing includes a pattern matching process [Step S3], a vertical difference process [Step S4], a length determination process [Step S5], a diagonal cut determination process [Step S6], and a horizontal difference process [Step S7] which will be described in detail later. , Morphology processing [Step S8], color stain determination processing [Step S9], and shading processing [Step S10]. When an abnormal appearance (defective) of the cigarette C is detected for each of these processes, the contents of the defect (defective item name) are stored as a defect history, and a corresponding defective item counter (not shown) is stored. The number of defects is incremented by step [Steps S11, S12, to S18].
[0027]
When all the appearance inspections based on the plurality of image processes described above are completed, it is determined whether or not the appearance image (inspection target image) should be saved as an image file (step S19). Then, for example, when it is determined that it is necessary to create an image file according to the content of the defect, the appearance image (image to be inspected) is acquired as an image file in the main memory described above [Step S20]. Such a series of image processing is repeatedly executed while determining whether or not the appearance inspection has been completed for all of the plurality of cigarettes C to be processed [Step S21]. Then, when all the processing is completed, the image file obtained in the main memory and its inspection history are written into the hard disk 14 as described above to form an image file [Step S22], and all the processing is completed.
[0028]
Here, the pattern matching process [Step S3] will be described. In the pattern matching process, when a brand of the cigarette C to be inspected is specified, a brand name or an emblem described on the peripheral surface of the cigarette C of the relevant brand. It is determined whether or not the cigarette C whose appearance image has been obtained by using the characters / graphics such as is a cigarette of a brand to be inspected, and excludes any cigarette other than the brand to be inspected.
[0029]
That is, the pattern matching process uses a known image 21 in which characters and graphics such as a brand name and an emblem described on the periphery of the cigarette C are registered in advance as shown in FIG. This is executed by calculating the degree of coincidence between the known image 21 and the appearance image of the cigarette C to be subjected to the appearance inspection (matching processing function 22). When the degree of coincidence is equal to or greater than a predetermined set value, it is determined that the pattern (character / graphic) matches, that is, the cigarette of the brand to be inspected (the coincidence degree determination function). 23).
[0030]
By the way, prior to the appearance inspection, the known image 21 obtains the appearance image of the cigarette C whose brand is clear via the line sensor 9 and, for example, displays the character / graphic under the instruction of the operator who monitors the appearance image. Is obtained by, for example, cutting out the description area of the image into a rectangular shape. In particular, in this embodiment, the above-described known image 21 is divided into a plurality of (for example, three) regions in the circumferential direction of the cigarette C, and partial known images 21a, 21b, and 21c are associated with each of these regions. I'm asking. At the time of the appearance inspection, an image of the partial region 24a in which the brand name, the emblem, and the like are particularly likely to be described in the appearance image 24 of the cigarette C to be inspected, and the partial known images 21a, 21b, and 21c. Are compared with each other, and the degree of coincidence of each of the partial known images 21a, 21b, and 21c with the partial region 24a is obtained.
[0031]
In this way, the degree of coincidence of each of the partial known images 21a, 21b, and 21c with respect to the partial area 24a is obtained, and the degree of coincidence is comprehensively determined, whereby the appearance image 24 of the cigarette C is obtained. Even if it is not possible to determine the position based on the specific position, the pattern matching process for the character or figure is performed with high accuracy. That is, the imaging start position at the time of imaging the outer periphery of the cigarette C is not defined, and as shown in the external appearance image 24b in FIG. Even when the area of the figure is cut off at the upper and lower ends, pattern matching for the character or figure can be performed accurately.
[0032]
If it is confirmed that the cigarette C is of the brand specified by the above-described pattern matching, a filling process is executed accompanying the pattern matching process (a filling process function 25). As shown in the concept of FIG. 7, the filling process replaces the data of the pixel expressing the character / graphic with the data of the surrounding base pixel, thereby obtaining the character / graphic from the appearance image 24 described above. It plays a role in creating an inspection target image 26 in which information is masked. By masking (filling) the pixels representing the characters / graphics, the pixels representing the characters / graphics (the image portions that are intentionally colored and are clearly different from the base) in image processing to be described later as defects such as dirt. A defect such as erroneous detection is prevented beforehand.
[0033]
As described above, the pattern matching process is performed to confirm that the cigarette C is a cigarette C of the brand to be inspected, and the appearance image in which characters and graphics such as the brand name and emblem are filled out is a defect in appearance of the cigarette C as the inspection target image. Is provided for the image processing for detecting. This image processing is first executed from vertical difference processing [Step S4]. This vertical difference processing is one of the characteristic processing in the visual inspection apparatus according to the present invention, and is shifted from the inspection target image 26 by several pixels in the vertical direction as shown in FIG. This is performed by creating an image 27, superimposing the shift image 27 and the inspection target image 26, and obtaining a difference image 28 indicating the difference therebetween. In particular, in this vertical difference processing, the inspection target image 26 is converted into a black and white image (monochrome image) based on its luminance component, and a shifted image of the black and white image is created. Then, the difference between the images when these images are superimposed is obtained as a difference image 28.
[0034]
According to such vertical difference processing, for example, a defective portion 26a such as peeling is present in the wrapping portion of the cigarette C (overlapping portion to be glued) in the cigarette C, and the pixel of the defective portion is located in the peripheral portion (base). Even in the case where the difference is similar to the pixel, the defect can be easily and accurately detected as a pixel portion that is clearly different from other portions in the difference image 28.
[0035]
That is, the outer peripheral surface (base) of the cigarette C generally has a substantially uniform color component. In addition, defective portions such as peeling in the above-described wrapping portion have almost the same color components as the above-mentioned base material, and are often slightly different from the base material. Therefore, even if the signal level of each pixel of the inspection target image 26 is simply discriminated by a predetermined threshold value, it is considerably difficult to clearly identify the difference. Furthermore, when the illumination light sources 8a and 8b that illuminate the surface of the cigarette C have fluctuations and the illumination intensity itself fluctuates, the level of the appearance image also changes accordingly. Therefore, even if the signal level of each pixel of the inspection target image 26 is determined using the predetermined threshold value as described above, it is difficult to clearly distinguish the difference between the above-described defective portion and its base.
[0036]
Specifically, when the inspection target image is given as pixels of 256 gradations, the level of the pixel which should be originally all white (gradation level: 255) is changed to the gradation level with the decrease of the illumination light amount. It may be obtained as a pixel signal of about [200]. Then, when the determination level for detecting a defective portion such as a stain on all white is set to about [210], there is a problem that all of the detection is performed as a defective portion.
[0037]
In this regard, in the vertical difference processing, the inspection target image 26 and a shifted image 27 obtained by shifting the inspection target image by several pixels are overlapped, and a difference between these images 26 and 27 is obtained. Then, as shown schematically in FIG. 9, even if the signal level of the pixel in the defective portion is only slightly different from the signal level of the pixel in the base portion, the base pixel level in the difference image is used as a reference. As a result, it is possible to detect only a defective portion as a portion having a different pixel signal level. Incidentally, in the difference image 28, the base portion having no defective portion is obtained as a pixel whose difference is zero (0). Therefore, even if the overall image level is reduced due to a decrease in the amount of illumination, the dirty portion is detected based on the level of the base. As a result, regardless of the change in the amount of illumination light, it is possible to reliably detect a defective portion having a similar color component to the substrate, such as peeling in the wrapping portion, as a relative difference portion to the substrate by the vertical difference processing. Becomes possible.
[0038]
After the above-described vertical difference processing is performed, a determination is made on the length of the cigarette C from the appearance image (image to be inspected). The determination of the length is performed by detecting left and right edges indicating the appearance image of the cigarette C from the image obtained by the line sensor 9 as shown in the processing concept of FIG. Specifically, by utilizing the fact that the line sensor 9 has 1024 pixels per line as described above, it is possible to determine from what pixel to what pixel the image of the cigarette C is obtained for each line. Each is detected (edge detection 31). Then, the average of the left edge detection pixels and the average of the right edge detection pixels obtained over 300 lines are respectively obtained (averaging process 32), and the number of pixels between these detection edge points is calculated as described above. It is obtained as the length of the cigarette C in the appearance image (inter-edge distance detection processing 33).
[0039]
Thereafter, by comparing the reference pixel number indicating the standard length on the appearance image in the cigarette C of various lengths previously determined with the number of pixels between the detected edge points, the appearance image is obtained. Is determined (pixel number comparing means 34). That is, since the length per pixel on the image captured by the line sensor 9 is determined by the imaging system, the length of the cigarette C is determined according to the number of pixels between the detection edges indicating the length of the cigarette C. Judge (detect). If the detection length (number of pixels) is different from the length (reference number of pixels) of the cigarette C of the brand to be inspected, or if it is out of the allowable error range, this is regarded as defective. judge.
[0040]
When determining the length of the filter tip portion in the cigarette C, for example, according to the character / graphic such as the brand name or emblem obtained in the above-described pattern matching process, the character / graphic in the axial direction of the cigarette C is determined. The position of the center of gravity is determined in advance. Then, as shown in FIG. 11, the number of pixels from the position of the center of gravity to each of the edges may be detected.
[0041]
By determining the number of pixels (length) up to the edges at both ends based on the position of the center of gravity of the character / graphic in this way, for example, the colors of the cigarette C and the filter paper are similar to each other. Even when it is difficult to detect the boundary on the image, it is possible to accurately determine the lengths of the wrapper portion (tobacco portion) and the filter chip portion.
[0042]
Next, the process of determining the end of the cigarette C at an angle is described (step S6). The diagonal cut refers to a state in which the end is cut with a certain inclination θ without being cut at a right angle to the axial direction, as shown in an external view of the cigarette C viewed from a diagonal direction in FIG. . When such a diagonally cut cigarette C is imaged using the above-described line sensor 9, the edge portion of the appearance image is developed in a waveform in the circumferential direction of the cigarette C, for example, as shown in FIG. In particular, since the appearance image is detected over 1.2 turns of the cigarette C as described above, the edge of this waveform always includes the maximum protrusion and the maximum cutout.
[0043]
Therefore, in the diagonal cut determination processing, as described above, the minimum edge position Emin and the maximum edge position Emax are detected when the pixels of the edge portion in the appearance image are detected for each line. Then, the difference (Emax−Emin) between the minimum edge position Emin and the maximum edge position Emax is obtained as the oblique cutting angle θ at the end of the cigarette C. By determining the angle θ of the oblique cutting in this manner, the quality of the cutting of the cigarette C is determined.
[0044]
After the determination of the shape of the cigarette C as described above, the determination of dirt and scratches on the surface of the cigarette C is performed. In this determination processing, since the detection conditions are different in the cigarette C in the wrapping paper portion (tobacco portion), the filter chip portion, and the boundary portion (seam portion), for example, as shown in FIG. The processing is performed while being divided into a plurality of areas in the longitudinal direction of the cigarette C. Here, the description will be made assuming that the area is divided into the three areas described above. However, the area may be divided into four or more areas, and the size of the detection area is arbitrarily set according to the specification of the cigarette C or the like. You may do it.
[0045]
Then, for each of the above detection areas, the stain detection process shown in FIG. 15, the linear scratch determination process shown in FIG. 16, the color stain detection process shown in FIG. 17, and the density detection process shown in FIG. Good. More specifically, in the dirt detection processing, the processing target image (monochrome image) 26 is laterally shifted to create a laterally shifted image [Step S31], and the laterally shifted image and the processing target image 26 are superimposed. Then, the horizontal difference image is obtained [Step S32]. Then, by judging the horizontal difference image in the same manner as the above-described vertical difference image, a pixel having a color component different from that of the base image is detected as dirt [Step S33]. According to such a horizontal difference process, it is possible to easily and accurately detect an area of a color component different from the base (base color) of the surface of the cigarette C based on the base color component. . Moreover, similarly to the above-described vertical difference processing, it is possible to reliably detect an area of a color component different from the base material regardless of a change in the illumination intensity of the illumination light sources 8a and 8b and a change in the illumination color component. .
[0046]
On the other hand, in the linear flaw determination processing, for example, the above-described processing target image is binarized, and the morphology processing for emphasizing an edge component by differentiating the binarized image is performed [Step S41]. Then, the length of the edge emphasized by the morphology processing is detected as the number of pixels [Step S42], and the edge length (number of pixels) is determined [Step S43]. Then, the number of edges longer than a predetermined length is checked, and if the number of edges is larger than the predetermined number, this may be determined as a defective product having a linear scratch.
[0047]
In addition, the stain on the surface of the cigarette C may generally appear as a yellow stain due to the stain of tobacco leaves on a white background forming a wrapping paper, and often has a color component similar to the base material. The image is reconverted into a CMY image signal [Step S51]. Then, it is determined whether or not a unique color component is included for each of the re-converted C (cyan) component, Y (magenta) component, and Y (yellow) component (steps S52, S53, S54).
[0048]
That is, in the related art, only the difference Δ between the color of the detection target and the reference color in the RGB space is regarded as [ΔR + ΔG + ΔB], and it is only determined whether or not this color difference Δ is larger than the determination threshold. . In this regard, in the color stain detection processing in this embodiment, the color difference Δ with respect to the reference color is regarded as ΔC / ΔY, ΔM / ΔY for each color component in the CMY space, and these are individually compared with the determination threshold. .
[0049]
Therefore, the RGB color image signal obtained by the line sensor 9 is converted into the CMY image signal which is a complementary color system in this manner, and the presence of a different color is detected for each color component. For example, it is possible to reliably detect even a stain having a color component similar to that of the base material, such as a pale yellow on a white background, a yellow on a red background, and a dark blue on a blue background. However, instead of such conversion of the image signal to the CMY system, it is also possible to emphasize each of the images of the color components in advance by using each color filter of the YMC system, and to process the image. .
[0050]
Thereafter, in order to detect the overall dirt of the appearance image described above, the appearance image composed of a color image is converted into a black and white image as shown in FIG. 18 [Step S61], and the black and white image is discriminated by a predetermined determination threshold. Binarization is performed [Step S62]. Then, the size of the region determined as a black pixel by this binarization (the size of a united region of black pixels) is determined [Step S63], and a region having a region having a size equal to or larger than a predetermined value is greatly stained. It may be determined that the product is defective. At this time, the black-and-white image may be discriminated by two levels of thresholds, and the size of a pixel that is not included in a predetermined threshold range may be determined.
[0051]
Thus, by sequentially performing the above-described image processing, it is possible to reliably detect various defects appearing in the appearance of the cigarette C. Moreover, the appearance inspection of the cigarette C can be easily performed without being affected by a change in the lighting environment of the cigarette C due to the lighting light sources 8a and 8b. Further, as described above, since the image processing (appearance inspection) is performed using the plurality of image processing units 12a, 12b, to 12n in parallel, even when the appearance inspection is performed while the cigarette C is transported at a high speed, all of the cigarettes C are surely inspected. Can be inspected.
[0052]
By the way, the above-described image processing is a procedure for automatically performing the appearance inspection on the continuously transported cigarette C. However, it is also possible to perform this manually (manually). In this case, the inspection may be performed according to the inspection procedure shown in FIG. In this manual inspection process, for example, using the first image processing unit 11 described above, one appearance image read using the line sensor 9 or an appearance image stored as an image file on the hard disk 14 is selectively used. [Step S71]. Then, a manual (manual) inspection switch is turned on [Step S72], and pattern matching processing [Step S73], vertical difference processing [Step S74], length determination processing [Step S72] are performed as in the case of the automatic inspection described above. S75], diagonal cut determination processing [step S76], horizontal difference processing [step S77], morphology processing [step S78], color stain determination processing [step S79], and shading processing [step S80] are sequentially performed.
[0053]
When it is determined by the pattern matching process that the cigarette is not the brand (type) to be inspected, the partial known images 21a, 21b, and 21c that have obtained the determination result and the inspection target image 24 (partial area 24a) Are displayed (step S81), and the process ends. If a defective portion is detected in the vertical difference processing, the detected defective portion is colored and clearly indicated so as to be easily visible [Step S82]. Further, if a mismatch with the standard length is detected by the length judgment processing, the difference (dimension) of the length is displayed as data [step S83], and the end of the cigarette C is subjected to the oblique cutting judgment processing. If a diagonal cut is detected, the width of the diagonal cut (angle of the diagonal cut) is displayed [Step S84].
[0054]
If the image processing is further advanced and defects are detected in the horizontal difference processing, the morphology processing, the color stain detection processing, and the shading processing, respectively, those defective detection parts are colored and the defective parts are clearly visible. [Steps S85, S86, S87, S88]. In this case, it is preferable that the color for coloring the defect detection part is different from each other in accordance with the content (type) of the defect detection. With this configuration, the content of the defect can be easily grasped from the difference in the color of the image to be inspected, so that the cause of the defect can be effectively used.
[0055]
Further, as described above, if the inspection result of the cigarette C subjected to the appearance inspection is filed as an inspection history in the hard disk 14 as the number of inspections, the number of defective products, the number of detected defects by item, etc., the inspection history file is used in the cigarette manufacturing process. It is also easy to find the cause of defective products. In particular, by monitoring the number of cigarettes that are continuously rejected as defectives, it can be regarded that there is a problem in the fact that a large number of defectives are generated in the cigarette manufacturing apparatus itself. It is also possible to determine that some trouble has occurred in the illumination system for obtaining an image or the imaging system using the line sensor 9. Therefore, it is possible to make an effect of self-diagnosing the function of the appearance inspection apparatus itself.
[0056]
Note that the present invention is not limited to the above-described embodiment. For example, the order of the execution procedure of the above-described image processing can be changed according to the important items of the visual inspection, and it is not necessary to execute all of them. Further, here, the peripheral surface of the cigarette C is imaged over the entire periphery thereof using the color line sensor 9, but an image is input using a monochrome line sensor and color filters such as Y, M, and C. Of course, it is possible. Although the appearance inspection of cigarettes has been described here as an example, the invention can be similarly applied to the appearance inspection of a round rod such as a filter chip and a heat insulator for piping. In addition, the present invention can be variously modified and implemented without departing from the gist thereof.
[0057]
【The invention's effect】
As described above, according to the present invention, it is possible to easily and reliably detect defects such as scratches and dirt on the surface of the round bar without being affected by the fluctuation of the illumination light source. In addition, a defective portion of a color component similar to the substrate can be detected by a simple image processing, and a great effect can be obtained in practical use such that the processing can be speeded up easily.
[Brief description of the drawings]
FIG. 1 is a diagram showing an overall configuration of a visual inspection apparatus according to an embodiment of the present invention.
FIG. 2 is a diagram showing a schematic configuration of a main part of a rotation mechanism for rotating a cigarette.
FIG. 3 is a diagram illustrating a schematic configuration of a main part of a visual inspection unit in the visual inspection device.
FIG. 4 is a diagram showing an example of an appearance image of a cigarette captured by a Latin sensor.
FIG. 5 is a diagram showing a schematic processing procedure in a visual inspection unit.
FIG. 6 is a diagram schematically showing the concept of a pattern matching process.
FIG. 7 is a diagram schematically illustrating the concept of a filling process.
FIG. 8 is a diagram schematically illustrating the concept of a difference process.
FIG. 9 is a diagram showing an example of a signal level of a difference image obtained from an original image and a shift image.
FIG. 10 is a diagram schematically illustrating the concept of a length detection process.
FIG. 11 is a view showing the concept of distance detection from the position of the center of gravity of a character or figure in the length detection processing.
FIG. 12 is a perspective view showing an obliquely cut end of the cigarette end;
FIG. 13 is a diagram schematically illustrating the concept of a diagonal cut detection process.
FIG. 14 is a diagram showing an example of area division setting in the dirt detection processing.
FIG. 15 is a diagram showing a schematic procedure of a dirt detection process.
FIG. 16 is a diagram showing a schematic procedure of a linear scratch determination process.
FIG. 17 is a diagram showing a schematic procedure of a color stain detection process.
FIG. 18 is a diagram showing a schematic procedure of a density detection process.
FIG. 19 is a diagram showing an example of a procedure of a manual appearance inspection process.
[Explanation of symbols]
4 Rolling roller
7 Kick Roller
8a, 8b Illumination light source
9 Line sensor
10 Appearance inspection department
11 First image processing unit
12a, 12b, to 12n Second image processing unit
13 Host computer
14 Hard disk (image file)

Claims (4)

A rotation mechanism for rotating the round bar body at a predetermined position,
A line sensor that images the peripheral surface of the round bar body that is rotating by the rotation mechanism one line at a time along its axial direction to obtain an external appearance image over the entire circumference of the round bar body;
A difference image forming means for forming a shift image by giving a predetermined positional shift to the inspection target image obtained by the line sensor and subjected to the appearance inspection, and obtaining a difference image between the shift image and the inspection target image; ,
A defect determining means for determining each pixel level of the difference image and extracting a defective portion of the appearance of the round bar body; The visual inspection device for a round bar according to claim 1,
Further, a pixel forming the character / graphic included in the external image by pattern-matching a known image showing the character / graphic present on the surface of the round bar and the external image obtained by the line sensor. Pattern matching means for detecting,
And a pre-processing means for forming an inspection target image to be subjected to an appearance inspection by filling a pixel forming the character / graphic detected by the pattern matching means with a background pixel component thereof. Appearance inspection equipment. The known image is composed of a plurality of partial known images obtained by dividing the area where the character / graphic on the surface of the round bar is divided into a plurality of regions in the circumferential direction of the round bar,
3. The pattern matching unit according to claim 2, wherein an image of an area where the character / graphic in the appearance image is likely to be present is compared with the plurality of partial known images to determine a degree of matching. Apparatus for inspecting the appearance of round rods as described. The appearance inspection device for a round bar according to claim 1 or 2, wherein the round bar is made of cigarette.

Images