JP2016137474A - Inspection method and inspection device for hollow fiber module - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide an inspection method and inspection device for a hollow fiber module capable of accurately inspecting the hollow fiber module product that has been developed to achieve a lower presence density for improvement in performance thereof, for that a conventional inspection process has incorrectly detected a side surface part of the hollow fiber as a defect at an inspection for a picked-up image of an end of the hollow fiber module.SOLUTION: In an inspection method for a hollow fiber module performed on the basis of a picked-up image of an end of the hallow fiber module, after noise including a side surface part of the hollow fiber is removed from the picked-up image, the inspection is performed for the image from that noise including the side surface part of the hollow fiber has been removed.SELECTED DRAWING: Figure 1

Description

  The present invention relates to a method and apparatus for measuring and inspecting the shape of a hollow fiber module.

  Hollow fiber mainly refers to the form of a macaroni-like membrane. Since the membrane area per unit volume can be widened, it is often used as a filter for water purifiers and wastewater treatment. The type of film used and the form of utilization vary depending on the application. Some of them are used as filters for hemodialysis.

  Hemodialysis is a method of artificially filtering waste products in blood such as urea and creatinine. Specifically, filtration is carried out by passing blood through a pipe-shaped product in which hollow fibers are bundled and flowing a dialysate around it. A bundle of these hollow fibers is called a hollow fiber module.

  The hollow fiber module used for hemodialysis must be able to secure the expected blood flow rate for the purpose of filtering blood, and the bundled hollow fibers are flat or crushed. There must be no more than this number. Therefore, the hollow fiber module is an area in which quality control at the time of manufacture is very important.

  Patent Document 1 discloses an apparatus that takes in an end face of a hollow fiber module as image data with an imaging device and inspects the number of hollow fibers that are flattened by image processing. In order to increase the filtration efficiency, the hollow fiber module uses a thin hollow fiber having an outer diameter of about 300 μm, and bundles 10,000 or more hollow fibers into one module in order to secure the amount to be filtered. Therefore, it is not an easy task to artificially count how many hollow fibers are in a flat state by looking at the end face of the hollow fiber module. Then, the method for performing automatically using image processing was proposed.

  Further, in Patent Document 2, in the observation of the end face of the hollow fiber module, for the problem that it is difficult to obtain a contrast between the resin-hardened portion and the hollow fiber by imaging with white illumination, the observation is performed using light in the ultraviolet region. An invention to perform is disclosed.

JP 2003-67724 A JP 2006-64501 A

  With the improvement in performance of hollow fiber modules and the diversification of products, hollow fiber modules with low yarn density have been manufactured. In the hollow fiber module having a low yarn density, when the end surface is inspected, the distance between the yarns is large, and therefore the possibility that the hollow fiber side surface is imaged is increased. Therefore, even when the techniques described in Patent Documents 1 and 2 are used, when the hollow fiber side surface is imaged, it is erroneously recognized and detected as a flat state or a collapsed state in the defect determination process, and inspection performance and production There was a problem of causing a decrease in sex.

  The present invention has been conceived in view of the above problems.

That is, the present invention is as follows.
(1) a step of binarizing a captured image of the end face of the hollow fiber module with a first threshold value to obtain image data A;
Binarizing the captured image with one or more thresholds different from the first threshold to obtain one or more image data X;
Measuring the feature quantity of the image data A, and separating the image data C from which the normal thread candidates are extracted based on the measurement results and the image data D from which the defect candidates are extracted;
Creating image data E from the image data C;
Removing the image data E from the image data D to obtain the image data F;
Obtaining image data G obtained by combining the image data C and the image data F;
Calculating any two of the image data A and the one or more image data X to obtain the image data H;
Calculating the image data G and the image data H to obtain the image data I;
Measuring a feature amount of the image data I;
A method for inspecting a hollow fiber module.
(2) One of the image data X is image data B obtained by binarizing the captured image with a second threshold value that is larger than the first threshold value,
The method for inspecting a hollow fiber module according to (1), wherein the step of acquiring the image data H is a step of acquiring the image data H by calculating the image data A and the image data B.
(3) One of the image data X is image data J acquired by binarizing the captured image with a third threshold smaller than the first threshold,
The method for inspecting a hollow fiber module according to (1), wherein the step of acquiring the image data H is a step of acquiring the image data H by calculating the image data A and the image data J.
(4) One of the image data X is image data B obtained by binarizing the captured image with a second threshold value that is larger than the first threshold value, and the other one of the image data X is , Image data J obtained by binarizing the captured image using a third threshold smaller than the first threshold,
The hollow fiber module inspection method according to (1), wherein the step of acquiring the image data H is a step of acquiring the image data H by calculating the image data B and the image data J.
(5) The operation in the step of obtaining the image data H is an operation of subtracting image data having a larger threshold used for binarization from image data having a smaller threshold used for binarization. Inspection method of the hollow fiber module in any one of said (1)-(4).
(6) The method for inspecting a hollow fiber module according to (1), wherein the step of measuring the feature amount of the image data I includes a step of measuring the roundness of the hollow fiber.
(7) In a hollow fiber module inspection apparatus comprising means for irradiating light to an end face of a hollow fiber module, means for imaging reflected light from the end face, and image processing means,
The image processing means includes
The captured image obtained by the imaging means is binarized with a first threshold to obtain image data A, and the captured image is binarized with one or more thresholds different from the first threshold. A binarization processing unit for acquiring one or more image data X;
An image data classification unit that measures the feature amount of the image data A, and classifies the image data C from which normal thread candidates are extracted and the image data D from which defect candidates are extracted based on the measurement results;
A mask image creation unit for creating image data E from the image data C;
A mask image removing unit that removes the image data E from the image data D to obtain image data F;
A whole-thread image creation unit that obtains image data G by combining the image data C and the image data F;
An image data calculation unit (1) that calculates any two of the image data A and the one or more image data X to obtain image data H;
An image data calculation unit (2) for calculating the image data G and the image data H to obtain image data I;
An inspection apparatus for a hollow fiber module, including a calculation unit that measures a feature amount of the image data I.

  Even when a hollow fiber module with a low yarn density is imaged and the side surface of the hollow fiber is imaged, it is possible to suppress a decrease in inspection performance by removing the side surface of the hollow fiber from the inspection target.

It is a figure which shows the structure of the inspection apparatus of the hollow fiber module of this invention. It is a schematic diagram of the captured image obtained by imaging the end surface of a hollow fiber module with an imaging device. It is the image which expanded a part of FIG. It is a figure which shows the flow of a process of an image processing apparatus. It is the image which expanded a part of captured image. It is a figure which shows the result of binarization processing with the 1st threshold value. (Image data A) It is a figure which shows the result of the binarization process by the 2nd threshold value. (Image data B) This is a result of subtracting image data B from image data A. (Image data H) It is a figure for demonstrating when obtaining a short diameter. It is the figure which collected the data of the hollow part with high roundness. (Image data C) It is an image including a hollow part with low roundness. (Image data D) It is a figure which shows the production method of a distance mask. It is the image which created the distance mask. (Image data E) This is a result of using a distance mask for the image data D (image data F). It is the image which extracted all the hollow parts. (Image data G) This is a result of subtracting image data H from image data G. (Image data I) It is the figure which showed the determination flow of the hollow fiber module. It is a figure which shows the flow of a process of the image processing apparatus in 2nd Embodiment. This is the result of binarizing the captured image with the third threshold. (Image data J) It is a figure which shows the flow of a process of the image processing apparatus in 3rd Embodiment.

  DESCRIPTION OF EMBODIMENTS Embodiments of a hollow fiber module inspection apparatus according to the present invention will be described with reference to the drawings.

  In FIG. 1, the block diagram of the hollow fiber module test | inspection apparatus 10 which is a suitable form of the test | inspection apparatus of the hollow fiber module of this invention is shown. Hereinafter, the hollow fiber module to be inspected is simply referred to as “module M”. The hollow fiber module inspection device 10 includes a hollow fiber module support device 101, a light irradiation device 102, an imaging device 103, an image processing device 104, an operation command device 105, a display device 106, and transfer devices 107 and 108.

  In the hollow fiber module inspection apparatus of the present invention, it is preferable to shorten the inspection time by simultaneously imaging both end faces of the module M. Therefore, the hollow fiber module inspection device 10 has two systems of light irradiation devices 102, imaging devices 103, and image processing devices 104.

  The imaging apparatus 103 is indicated by reference numerals 1031 and 1032, and the image processing apparatus 104 is indicated by reference numerals 1041 and 1042.

  Moreover, the light irradiation apparatus 102 consists of the light source A and the light irradiation part B, and one was shown by code | symbol 102A1, 102B1, and the other was shown by code | symbol 102A2, 102B2.

  The module support device 101 includes a stage 101A and a support base 101B. The support base 101B mounts the module M, and the height and position can be changed back and forth and up and down and right and left with reference to the center of the end face of the module M. Thus, the inclination of the module M can be adjusted so that the center of the end face of the module M is parallel to the optical center axis of the lens 103L of the imaging device 103.

  The imaging device 103 includes an imaging device 1031, an imaging device 1032, a lens 103L1 and a lens 103L2. In addition, the imaging devices 1031 and 1032 are also provided with a mechanism that can change the height and position in front and back, up and down, left and right with respect to the center of the end face of the module M, and the center of the module M and the centers of the imaging devices 1031 and 1032 are aligned. It can be used for that. Thereby, it can match | combine with an optical center axis | shaft also for every kind from which the outer diameter of the module M and the length of a cylinder differ.

  The support base 101B is installed on the stage 101A in a movable form. This is because the module M for inspection is received from the transfer device 107 and passed to the transfer device 108 after the inspection.

  From the viewpoint of the module M, the transfer is performed in the following order.

  First, in the manufacturing process of the module M, the module M before the inspection is transferred by the transfer device 107 from the pre-inspection process. Next, it is placed on the support base 101B that has moved on the stage 101A. The support base 101 </ b> B moves on the stage 101 </ b> A to the front of the imaging device 103. Module M performs alignment such as up / down / left / right before imaging, and imaging and inspection are performed. After completion of the inspection, the support base 101B moves to the position of the transfer device 108. The module M is transferred to the next process by the transfer device 108.

  The transfer devices 107 and 108 are controlled by transfer commands Trc7 and Trc8 output from an operation command device 105 described later.

  The light irradiation device 102 includes a light source 102A and a light irradiation unit 102B. The light source 102A adjusts the amount of light according to the irradiation command Acc from the image processing apparatus 104.

  The image processing device 104 includes an image processing device 1041 and an image processing device 1042.

  The light source 102 </ b> A is used to irradiate the end face with light when imaging the end face of the module M. The light source 102A is preferably capable of irradiating light including a wavelength in the ultraviolet region in order to capture a high contrast between the hollow fiber to be inspected and the resin portion.

  The light irradiation unit 102B is a part that irradiates the end surface of the module M with light from the light source 102A. For this purpose, LED lighting or the like can be suitably used. In addition, you may use an optical fiber from a light source to a light irradiation part. In addition, imaging may be performed by irradiating the end face of the hollow fiber module using a direct-light LED in which a light source and a light irradiation unit are integrated.

  Moreover, since the hollow fiber module inspection apparatus 10 has two systems of imaging devices, it also has two systems of light irradiation devices.

  The imaging device 103 is for obtaining image data by imaging the end face of the module M mounted on the support device 101. Although there is no particular limitation on the configuration, it is desirable to have a photoelectric converter such as a CCD or a CMOS. An optical system such as a lens, a lens barrel, a diaphragm, and a shutter for image formation on a CCD or CMOS may be appropriately determined.

  The photoelectric converter can be either a line type or an area type. The hollow fiber module inspection device 10 has two imaging devices, and thus has two imaging devices 1031 and 1032 on the left and right.

  The image processing apparatus 104 starts an inspection in response to an inspection start command Fic. Since the hollow fiber module inspection apparatus 10 has two image processing apparatuses 104, the hollow fiber module inspection apparatus 10 includes two left and right image processing apparatuses 1041 and an image processing apparatus 1042. Specifically, the number of hollow fibers that are flattened or crushed among the individual hollow fibers that can be seen on the end face of the module M is obtained by performing image processing on the image data captured by the imaging device 103. The image processing apparatus 104 includes an MPU (Micro Processor Unit) and a memory, and processes image data using a processing program stored in the memory. Therefore, most of the processing in the image processing apparatus 104 is software processing. However, a part of the processing may be combined in hardware to form an LSI. Details of the configuration and processing contents of the image processing apparatus 104 will be described later.

  The operation command device 105 controls the processing flow of the entire inspection device 10. A human may look at the result of each process and instruct the next process, or the operation commanding apparatus 105 may control the process based on a predetermined procedure and criteria. The operation command device 105 is connected to the transfer devices 107 and 108 and the image processing devices 1041 and 1042. Further, although not shown, it may be connected to the support base 101B of the support apparatus 101 so that it can move on the stage 101A.

  In such a connection state, the operation command device 105 can output transfer commands Trc7 and Trc8 to the transfer device 107 and the transfer device 108, respectively. Further, the operation command device 105 can transmit an inspection start command Fic1 to the image processing device 1041 and similarly transmit an inspection start command Fic2 to the image processing device 1042.

  Then, the operation command device 105 receives the result data Dout from the image processing device 104. Then, from Dout1 and 2 which are the result data from the image processing apparatuses 1041 and 1042, it is possible to determine whether the module M to be inspected is a non-defective product or a defective product and output inspection data.

  The operation command device 105 is also connected to the display device 106, and can display the inspection results of individual modules, the obtained image data, the inspection results of the module M processed so far, and the like. Further, it may be connected to another computer so that data such as inspection results can be transferred.

  Next, the operation of the hollow fiber module inspection apparatus 10 will be described. Note that although the operation subject is described as being performed by the operation command device 105, the present invention is not limited to this.

  When the hollow fiber module inspection device 10 is started, the support table 101B is first moved to a predetermined position by the transfer command Trc7 of the operation command device 105, and the transfer device 107 places the uninspected module M on the support table 101B.

  Next, the support base 101 </ b> B moves on the stage 101 </ b> A and is positioned in front of the imaging device 104. Then, the imaging light is applied to the end face of the module M by the irradiation command Acc, and the inspection is started by the inspection start command Fic. Specifically, after imaging and obtaining image data, the image data is processed and an inspection result is output. The inspection result is transmitted from the image processing device 104 to the operation command device 105 as Dout. The contents of the inspection result Dout include the result of how many hollow fibers visible on the end face of the module M are flat or crushed.

  The operation command device 105 receives the inspection results Dout1 and Dout2 from the two image processing devices 104, and makes a final determination from these results. The final determination is a determination as to whether the module M to be inspected is a good product or a defective product.

  When the inspection is completed, the support base 101B is moved to a predetermined position. Then, a transfer command Trc8 is issued, and the inspected module M is transferred to the next step by the transfer device 108. Finally, it is determined whether or not to inspect the next module M. If it is continued, the process returns to the beginning. Otherwise, the process ends.

  FIG. 2 schematically shows a captured image obtained by imaging the end face of the module M with the imaging device 104. FIG. 2 includes a case a for storing a yarn bundle, a number of hollow fibers b and a resin portion c existing inside the case a.

  FIG. 3 is an enlarged view of a part of FIG. 2, and the hollow fiber b is mainly a ring-shaped and thick-walled portion b1 that is imaged with a high brightness, and a brightness lower than that of the thick-walled portion b1 therein. It is comprised from the hollow part b2 imaged. In addition, a gap c1 that is a region surrounded by a plurality of hollow fibers b, and a hollow fiber side surface b3 that corresponds to a side surface part of the hollow fiber b are included.

  In this image, since it is effective to determine a defect by the shape of the hollow portion b2 as described in Patent Document 1, the shape of the hollow portion b2 is an inspection target. However, the brightness of the hollow fiber side surface b3 is a value similar to the brightness of the hollow portion b2, and it has been a problem that the hollow fiber side surface b3 is determined to be the hollow portion b2.

  Next, the inspection method for the hollow fiber module of the present invention will be described.

In the inspection method of the hollow fiber module according to the present invention, there is a step of binarizing the captured image of the end surface of the hollow fiber module with a first threshold value to obtain image data A,
However, there is a separate step of acquiring image data X from the captured image by binarizing with one or more thresholds different from the first threshold, and even if there is only one image data X May be,
For the image data A, there is a step of measuring the feature amount and separating the image data C from which the normal yarn candidates are extracted based on the measurement results and the image data D from which the defect candidates are extracted,
Creating image data E by expanding and contracting the image data C;
Removing the image data E from the image data D which is a defect candidate to obtain the image data F;
Obtaining image data G obtained by combining the image data C and the image data F;
Calculating any two of the image data A and the one or more image data X to obtain the image data H;
Calculating the image data G and the image data H to obtain the image data I;
Measuring a feature amount of the image data I.

  The two pieces of image data used for the calculation in the “step of calculating any two of the image data A and the one or more image data X to obtain the image data H” are the image data A and the image data A combination of data X or a combination of two image data X binarized with two different thresholds may be used.

  Moreover, in this process, what can extract characteristically the noise containing the hollow fiber side surface b3 is preferably used for a calculation. For the calculation, calculation for extracting a difference between two image data is preferably used, and specific examples thereof include subtraction or exclusive OR between two image data. Of these, it is a simpler operation, and therefore, it is possible to perform the operation at a higher speed, so it is preferable to perform subtraction between the two image data. Therefore, the calculation in the step of acquiring the image data H is preferably an operation of subtracting image data having a larger threshold value used for binarization from image data having a smaller threshold value used for binarization. .

  The characteristic amount in the present invention represents a measured value or a calculated value used for separating normal yarn from defects. Specific examples thereof include the area of the cluster of white pixels (= the number of pixels), the major axis, the minor axis, the ratio of the major axis to the minor axis, and the like. Among the feature quantities, in the present inspection content, the defect is identified by using a plurality of feature quantities among these, but the ideal hollow fiber shape is a perfect circle, and the actual hollow fiber In order to evaluate the distortion of the shape of the hollow part, it is preferable to use roundness, which is an index representing how close to a perfect circle. The method for obtaining the roundness is as described later. Therefore, in the hollow fiber module inspection method of the present invention, it is preferable that the step of measuring the feature amount of the image data I includes a step of measuring the roundness of the image data I.

  It is preferable that the inspection method of the hollow fiber module of the present invention is implemented by any one of the following first to third embodiments.

<First Embodiment>
A first embodiment will be described.

  In the first embodiment, one of the image data X is image data B obtained by binarizing the first captured image with a second threshold value that is larger than the first threshold value.

  That is, the image data H is obtained by calculating the image data A and the image data B.

  FIG. 4 shows an image processing flow in the image processing apparatus 104 of the hollow fiber module inspection apparatus 10 in the first embodiment. 5, 6, 7, 8, 10, 11, 13, 14, 15, and 16 show actual captured images and image data corresponding to the image processing flow. Is.

  First, the imaging device 104 acquires a captured image as shown in FIG. 5 (S201). At this time, if the captured image includes luminance unevenness derived from the light source or noise derived from the imaging device 104, noise removal processing is performed (S202). The noise removal process may be omitted if noise is not included in the captured image, and the method may determine what process is appropriately performed depending on the cause of the noise.

  Next, binarization processing is performed on the captured image (S203). The inspection method of the hollow fiber module of the present invention includes a step of obtaining image data A by binarizing the captured image with a first threshold value. The binarization process determines the magnitude of each pixel data in the image data from a preset threshold value (here, the first threshold value), and determines whether it is a black pixel (0) or a white pixel (1). This is a rewriting process. In this case, binarization is executed with a predetermined threshold value with the resin portion c as the background as a black pixel (0) and the thick portion b1 as the foreground as white (1). The first threshold value is selected based on the fact that the shape of the hollow fiber membrane can be extracted with the highest accuracy in the image data A.

  Moreover, the evaluation method of inspection accuracy was carried out by inspecting 100 images of the end face of the hollow fiber module including true defects and false detections by both the conventional method and the method of the present invention. It can be evaluated by calculating and comparing “how much the number of false detections is reduced by comparison”.

  FIG. 6 shows image data A after the binarization process is performed on the captured image of FIG. If the image data A obtained by binarization with the first threshold contains a lot of noise, a process for removing noise at the outer peripheral portion is executed in the image data A (S204).

  The inspection method of the hollow fiber module of the present invention includes a step of binarizing the captured image with one or more thresholds different from the first threshold to obtain one or more image data X. Here, in the first embodiment, one of the image data X is image data B acquired by binarizing the captured image with a second threshold value that is larger than the first threshold value (S301). .

  FIG. 7 shows image data B obtained by binarization with a second threshold value that is larger than the first threshold value. Since the hollow fiber side surface b3 has a lower luminance value than the thick part b1, when using a second threshold value larger than the first threshold value, white pixels on the hollow fiber side surface b3 compared to the thick part b1 Will be less.

  Next, in the first embodiment, in the step of obtaining the image data H, the image data A and the image data B are calculated to obtain the image data H (S302). FIG. 8 shows the image data H acquired by the calculation. The calculation method in this step is not particularly limited as long as it is a calculation that can characteristically extract noise including the hollow fiber side surface b3. For the calculation, calculation for extracting the difference between the image data A and the image data B is preferably used. Specific examples thereof include subtraction of the image data B from the image data A, exclusive OR of the image data A and the image data B. Etc. Of these, the calculation is simpler, and therefore it is possible to perform the calculation at a higher speed. Therefore, it is preferable to subtract the image data B from the image data A.

  FIG. 8 shows image data H obtained by subtracting image data B from image data A.

  Since the image data H may include a part other than noise such as a part of the thick part b1 or the hollow part b2, the image data H is characterized with respect to the foreground of the image (a cluster of white pixels) as necessary. A quantity measurement is performed, and parts other than noise are deleted (S303).

  On the other hand, in the method for inspecting a hollow fiber module of the present invention, the feature amount of the image data A is measured, and the image data C obtained by extracting normal yarn candidates and the image data D obtained by extracting defect candidates based on the measurement result. And a step of separating into two.

  Specifically, only the black pixel block (hollow portion b2) surrounded by the white pixel block (thick portion b1) is extracted from the image data A as white pixels by reversing black and white. The feature amount is measured (S205). The definition and specific example of the feature amount are as described above. Also, in this step, the defect is identified using a plurality of feature amounts. However, the ideal hollow fiber hollow part has a perfect circle shape, and the distortion of the actual hollow fiber hollow part shape is evaluated. It is preferable to use roundness, which is an index representing how close to a perfect circle is. Of the feature quantities measured for each cluster of white pixels corresponding to the extracted hollow portion b2, roundness will be described as an example.

  In the present invention, the roundness is a value calculated by calculating a minor axis with respect to a mass of white pixel images as a target and taking a ratio with a design value of the hollow portion b2.

Next, a method for measuring the minor axis will be described. FIG. 9 is a schematic diagram of a method for measuring the tangent line D when the diameter of the particle is measured every unit angle Δθ = 45 degrees in the orthogonal coordinate system with the center of gravity coordinates of the hollow fiber as the center. Among the measured values measured for each unit angle,
Minimum diameter: Dmin
Minor axis: DS
, DS is approximated as shown in equation (1).

DS = Dmin (1)
From the above, after obtaining the minor axis DS for each piece of data of the cluster of white pixels, the roundness ratio (%) is obtained as follows with the design value of the hollow portion b2 as DR.

Cratio (%) = DS / DR × 100 (2)
Next, the image data A is classified into image data C and image data D according to each roundness from the result of the characteristic amount measurement. As will be described later, mask image data for removing the gap c1 is created by this separation processing, so only those with high roundness are selected (S206). Specifically, the image data C extracted by collecting only hollow portions b2 (called normal yarn candidates) having a roundness of 80% or more is output. A cluster of white pixels having a very high roundness can be regarded as the hollow portion b2 incorporated in the module M. By this processing, the hollow portion b2 of the hollow fiber b where the hollow portion b2 is close to a perfect circle is extracted from the image data A to obtain the image data C. FIG. 10 shows the image data C.

  Further, the remaining part of the image data C extracted from the image data A, that is, the part of the image data A having a roundness of less than 80% is collected and output as the defect candidate image data D. FIG. 11 shows the image data D.

  The image data A includes a gap c1 that is surrounded by a plurality of hollow fibers b and has an independent shape. Since these gaps c1 are surrounded by the thick part b1, they are extracted as a cluster of white pixels by the hollow extraction process (S205). Since this gap portion may resemble the shape of the hollow portion b2 having a low roundness, if it is erroneously determined as a hollow fiber b having a low roundness, the number of defects existing in the module M increases. Therefore, a normal hollow fiber module M that is not actually a defective product is erroneously determined to be a defective product.

  Therefore, in the method for inspecting a hollow fiber module according to the present invention, the image data C is generated by performing expansion / contraction processing on the image data C, and the image data E is removed from the image data D to obtain the image data F. The process of carrying out is included. In this step, first, image data E called a distance mask is created by performing an expansion / contraction process to be described later on the image data C, and then the image data E is removed from the image data D to obtain the image data D from the image data D. A process of removing the gap c1 between the hollow fibers is performed (S222). This processing is performed according to the following principle.

  FIG. 12 is a diagram for explaining a mask created based on the normal yarn extracted from the image data C. FIG. In FIG. 12, a thick part b1 and a hollow part b2 of the flat hollow fiber are observed. When the thickness of the hollow fiber is D1, the distance between the hollow portions b2 of the adjacent hollow fibers is D2, and the distance between the hollow portions b2 of the normal hollow fibers sandwiching the crushed hollow fibers is D3, D1, D2 , D3 has the following relationship.

D2 ≧ 2 × D1 (3)
D3 ≧ 4 × D1 (4)
Equations (3) and (4) indicate a range where the flat hollow portion b2 exists. In other words, the gap c1 includes a portion that does not satisfy the expressions (3) and (4). Therefore, noise that does not satisfy the above equations (3) and (4) is removed as noise. The processing procedure will be described below.

  First, an expansion / contraction process is performed on the image data C from which the normal yarn has been extracted (S231). The expansion process is a process in which individual pixel data is focused on the binarized image data, and when the pixel data is 1, all pixels around the pixel are set to 1. The process is a process of expanding a pixel having a value of 0.

  Here, the expansion process is performed for a distance corresponding to four times D1, and the contraction process is performed for a distance corresponding to twice D1. The difference between the gap part c1 and the flat hollow part b2 is that the adjacent hollow part b2 having a high roundness is more than twice the film thickness D1, or the hollow part b2 having a high roundness is between D1. It was the point whether it was more than 4 times away. Therefore, when the expansion process is performed for a distance of four times the film thickness from the hollow part b2 having a high roundness, the hollow parts b2 overlap each other. All of the gap c1 that is within a distance of 4 times the film thickness from the center of the hollow part b2 having a high roundness is completely filled with this expansion process. However, the flat hollow part b2 cannot be removed by this process.

  Next, when the film is contracted by twice the film thickness D1, the hollow part b2 having a high roundness becomes an image expanded by the film thickness. Since the expansion / contraction is not completely reversible, in the contracted data, the gap c1 does not exactly match the original part, and partially overlaps the expansion / contraction data of the hollow part b2 having a high roundness.

  The gap c1 can be deleted from the image data D by using the image data that has been subjected to the expansion and contraction processing. The result of expanding and contracting the image data C obtained by extracting the hollow portion having a high roundness becomes a filter called a distance mask. In practice, this processing is performed on the image data C, and the image data E shown in FIG. 13 is output. The image data E and the image data D from which defect candidates are extracted are overlapped, and when even a little overlaps, a process of removing the portion as the gap c1 is performed (S222). Image data F (FIG. 14) from which the gap has been removed can be obtained by this processing.

  The inspection method of the hollow fiber module of the present invention includes a step of obtaining image data G obtained by combining the image data C and the image data F. Specifically, the logical sum of the image data F and the image data C is taken (S207), and the image data G (FIG. 15) that can include the normal yarn candidate hollow portion b2, the defect candidate hollow portion b2, the hollow fiber side surface b3, and the like. )

  The inspection method of the hollow fiber module of the present invention includes a step of obtaining the image data I by calculating the image data G and the image data H. Specifically, the calculation is performed using the image data G and the image data H obtained by extracting the noise of the hollow fiber side surface b3, the case a, etc. obtained by the above-described calculation to obtain the image data I ( S208). The calculation in this step may be any calculation that can remove noises such as the hollow fiber side surface b3 and the case a. For the calculation, a calculation for extracting the difference between the image data G and the image data H is preferably used. Specific examples thereof include subtraction of the image data H from the image data G, exclusive OR of the image data G and the image data H. In the image G, for example, an operation in which only a portion indicated by a white pixel in the image data H is filled with a black pixel can be used. Of these, the calculation is simpler, and therefore it is possible to perform the calculation at higher speed. Therefore, it is preferable to subtract the image data H from the image data G. Image data I obtained by subtracting image data H from image data G is shown in FIG.

  The inspection method of the hollow fiber module of the present invention includes a step of measuring a feature amount of the image data I. The step of measuring the feature amount (S209) is as described above. When the roundness is measured as the feature amount of the image data I, the obtained roundness measurement data of the hollow portion b2 is sorted according to a threshold value set in advance. For example, those having a roundness of 70% or less can be classified as “flat defects”, and those having a roundness of 20% or less can be classified as “crushed defects”. Finally, a portion where the “crush defect” and “flat defect” are present in the captured image is marked and output as an inspection result image, and defect determination is performed (S210).

  The inspection and the inspection data acquisition processing are completed by the above processing.

  The subsequent processing will be described by taking the case of the above-described hollow fiber module inspection apparatus 10 as an example.

  After the process of obtaining the inspection data is completed, the defect and processed data are sent to the operation command device 105. The operation command device 105 may project these data on the display device 106.

  Next, determination processing performed by the operation command device 105 based on these results will be described. The operation command device 105 receives Dout1 and Dout2 from the two image processing devices 104. Then, a final determination is made based on the results from these two systems.

  FIG. 17 shows a detailed flow of the determination process. When the operation commanding device 105 receives Dout1 and Dout2, respectively, the operation commanding device 105 acquires the number of flat defects and crushing defects. Here, the flat defect is Nh, and the crushing defect is Nt. Also, “1” is added to the result obtained from the image processing apparatus 1041, and “2” is added to the result obtained from the image processing apparatus 1042.

  The module M to be inspected is inspected at both ends, the crushing defect Nt is obtained by Nt1 + Nt2, and the flat defect Nh is obtained by Nh1 + Nh2. Therefore, a non-defective product or a defective product can be determined by setting appropriate threshold values for these defects. Here, a threshold value Tt is set for the number of crushing defects, and a threshold value Th is set for a flat defect. Then, if the crushing defect Nt and the flat defect Nh are larger than the threshold values Tt and Th, respectively, the defect is determined, and if not, the non-defective product is determined. Finally, the lot number, Dout1 and 2, and the determination result of the module M to be inspected are transferred to a predetermined location.

  The inspection of the module M of the present invention is completed by the above processing flow. Returning to FIG. 1, the module M that has been inspected is transferred from the support base 101 </ b> B to the next process by the transfer device 108 as described above.

  As described above, the processing for extracting the image data of the hollow fiber side surface b3 in the inspection of the end face of the module M is implemented, and the true defect is obtained by performing the inspection based on the image obtained by removing the hollow fiber side surface b3 from the captured image. It is possible to accurately extract the flat hollow fiber b.

<Second Embodiment>
Next, a second embodiment will be described.

  In the second embodiment, one of the image data X is image data J obtained by binarizing the first captured image with a third threshold smaller than the first threshold.

  The step of acquiring the image data H is a hollow fiber module inspection method in which the image data A and the image data J are calculated to acquire the image data H.

  Also in the second embodiment, the imaging device shown in FIG. 1 can be used to capture an image of the end face of the module M. However, after acquiring a captured image, it differs in that the inspection process flow shown in FIG. 18 is used.

  In the inspection processing flow of FIG. 18, the normal inspection processing steps (S201 to S207) are the same as the processing performed in FIG. 4 which is the inspection processing flow of the first embodiment, but noise such as the side surface b3 of the hollow fiber is removed. The extraction method is different.

  Specifically, when binarizing the captured image, binarization is executed using a third threshold value smaller than the first threshold value (S304), and image data J (FIG. 19) is created. That is, in the second embodiment, one of the image data X is image data J obtained by binarizing the captured image with a third threshold value smaller than the first threshold value. When a threshold value smaller than the first threshold value is used, there is a feature that it is easy to extract noise such as the hollow fiber side surface b3 included in the captured image. For this reason, the image data J cannot be directly used for the subsequent inspection process, but by performing the calculation of the image data A and the image data J, the image data in which noise including the hollow fiber side surface b3 is characteristically extracted. Can be acquired (S305). That is, in the second embodiment, the step of obtaining the image data H is a step of obtaining the image data H by calculating the image data A and the image data J. The calculation in this step is not particularly limited as long as it is a calculation that can characteristically extract noise including the hollow fiber side surface b3 as in the case of the first embodiment. For the calculation, calculation for extracting a difference between two image data is preferably used. Specific examples thereof include subtraction of image data A from image data J, exclusive OR of image data A and image data J, and the like. Can be mentioned. Of these, the calculation is simpler, and therefore it is possible to perform the calculation at a higher speed. Therefore, it is preferable to subtract the image data A from the image data J.

  Since the image data H may include a part other than noise, such as a part of the thick part b1, a feature amount measurement is performed on the foreground of the image (a cluster of white pixels) as necessary. (S306), the parts other than noise such as the thick part b1 and the hollow part b2 are erased. The subsequent processing is the same as the processing shown in FIG.

  Since the density of the hollow fiber b and the density of the hollow fiber side surface b3 are different depending on the production type and production conditions, the second embodiment can provide more noise removal effect than the first embodiment. There is a case. Moreover, the effect of noise removal can be enhanced by setting the third threshold value smaller than the first threshold value. However, if the third threshold value is too small, noise that should not be removed may be removed. Therefore, the third threshold value needs to be set appropriately.

<Third Embodiment>
Next, a third embodiment will be described.

  In the third embodiment, one of the image data X is image data B obtained by binarizing a captured image with a second threshold value larger than the first threshold value, and the other one of the image data X One is image data J obtained by binarizing the captured image using a third threshold smaller than the first threshold.

  The image data H is obtained by calculating the image data B and the image data J.

  Also in the third embodiment, the imaging device shown in FIG. 1 can be used to capture an image of the end face of the module M. However, it is different in that the inspection processing flow shown in FIG. 20 is used for the captured image (S201).

  In the flow in FIG. 20, the normal inspection process (S201 to S207) is the same as the process in FIG. 4 which is the inspection process flow of the first embodiment, but noise such as the hollow fiber side surface b3 is extracted. The method is different.

  Specifically, when binarizing the captured image, binarization is performed with a second threshold value that is larger than the first threshold value (S301), and the image data B is created, and at the same time, a third threshold value that is smaller than the first threshold value is used. Then, the captured image is binarized (S304), and image data J is created. That is, in the third embodiment, one of the image data X is image data B obtained by binarizing the captured image with a second threshold value that is larger than the first threshold value, and the image Another one of the data X is image data J obtained by binarizing the captured image using a third threshold smaller than the first threshold.

  Next, the image data H is obtained by performing the calculation of the image data B and the image data J (S305). That is, in the third embodiment, the step of obtaining the image data H is a step of obtaining the image data H by calculating the image data B and the image data J. The calculation in this step is not particularly limited as long as it is a calculation that can characteristically extract noise including the hollow fiber side surface b3 as in the case of the first embodiment. For the calculation, calculation for extracting the difference between the image data B and the image data J is preferably used. Specific examples thereof include subtraction of the image data B from the image data J, exclusive OR of the image data B and the image data J. Etc. Of these, the calculation is simpler, and therefore, the calculation can be performed at a higher speed. Therefore, the image data B is subtracted from the image data J.

  Since the image data H may include a part other than noise, such as a part of the thick part b1, a feature amount measurement is performed on the foreground of the image (a cluster of white pixels) as necessary. (S306), the parts other than noise such as the thick part b1 and the hollow part b2 are erased.

  Further, the image data H having a high noise ratio such as the hollow fiber side surface b3 is subtracted from the image data B including a low ratio including noise on the hollow fiber side surface b3 and the like to create the image data H (S307), thereby creating the hollow fiber side surface b3. Etc. may be easier to extract. If necessary, a feature amount measurement is performed on the image data H in order to select noise to be removed (S308), and the thick part b1 of the hollow fiber and the hollow part b2 of the hollow fiber are deleted. The subsequent processing is the same as the processing shown in FIG.

  In addition, since the density of the hollow fiber b and the density of the hollow fiber side surface b3 differ depending on the production type and production conditions, in some cases, all the noises cannot be completely removed in the first and second embodiments. In that case, the effect of noise removal can be enhanced by selecting the third threshold value to be larger and the third threshold value to be smaller by using the third embodiment. If the difference is too large, noise that should not be removed may be removed, so the second threshold and the third threshold need to be set appropriately.

<Inspection equipment for hollow fiber module>
An inspection apparatus for a hollow fiber module according to the present invention is an inspection apparatus for a hollow fiber module comprising means for irradiating light to an end face of the hollow fiber module, means for imaging reflected light from the end face, and image processing means. Yes,
The image processing means includes
The captured image obtained by the imaging means is binarized with a first threshold to obtain image data A, and the captured image is binarized with one or more thresholds different from the first threshold. A binarization processing unit for acquiring one or more image data X;
An image data classification unit that measures the feature amount of the image data A, and classifies the image data C from which normal thread candidates are extracted and the image data D from which defect candidates are extracted based on the measurement results;
A mask image creation unit for creating image data E from the image data C;
A mask image removing unit that removes the image data E from the image data D to obtain image data F;
A whole-thread image creation unit that obtains image data G by combining the image data C and the image data F;
An image data calculation unit (1) that calculates any two of the image data A and the one or more image data X to obtain image data H;
An image data calculation unit (2) for calculating the image data G and the image data H to obtain image data I;
It is an inspection device for a hollow fiber module, including a calculation unit that measures a feature amount of the image data I.

  In the hollow fiber module inspection device of the present invention, the binarization processing unit binarizes the captured image obtained by the imaging unit with a plurality of threshold values. For example, the binarization process is performed on the captured image shown in FIG. 5 with the first threshold value, and the image data A shown in FIG. 6 is acquired. Further, binarization processing is performed with one or more threshold values different from the first threshold value, and one or more image data X is acquired. Specifically, binarization processing is performed with a second threshold value that is larger than the first threshold value, and image data B shown in FIG. 7 is acquired. Further, binarization processing is performed with a third threshold value smaller than the first threshold value, and image data J shown in FIG. 19 is acquired.

  The image data sorting unit measures the feature amount of the image data A shown in FIG. Specifically, only the black pixel block (hollow portion b2) surrounded by the white pixel block (thick portion b1) is extracted from the image data A as white pixels by reversing black and white. Perform feature measurement. From the measurement result, the image data C (FIG. 10) from which normal thread candidates are extracted and the image data D (FIG. 11) from which defect candidates are extracted are classified.

The mask image creation unit creates image data E (FIG. 13) called a distance mask by performing expansion and contraction processing on the image data C shown in FIG.
The mask image removing unit removes the image data E shown in FIG. 13 from the image data D shown in FIG. 11, thereby removing the gap c1 between the hollow fibers from the image data D. The image shown in FIG. Data F is acquired.

  The whole thread image creation unit obtains the image data G (FIG. 15) by combining the image data C shown in FIG. 10 and the image data F shown in FIG. Specifically, the logical sum of the image data F and the image data C is taken to obtain the image data G (FIG. 15) from which all the hollow portions b2 have been extracted.

  The image data calculation unit (1) calculates any two of the image data A shown in FIG. 6 and one or more image data X to obtain the image data H (FIG. 8). When the second threshold value in the binarization processing unit is larger than the first threshold value, the image data B shown in FIG. 7 corresponds to the image data X, and the third threshold value in the binarization processing unit is the first threshold value. If it is smaller than the threshold value, the image data J shown in FIG.

  The image data calculation unit (2) calculates the image data G shown in FIG. 10 and the image data H shown in FIG. 8, and acquires the image data I (FIG. 16). Specifically, the image data G shown in FIG. 10 and the image data H obtained by extracting the noise of the hollow fiber side surface b3, case a, etc. obtained by the calculation of the image data calculation unit (1) described above (FIG. 8). Is used to obtain image data I (FIG. 16).

  The calculation unit measures the feature amount of the image data I.

  The present invention can be used for automatic defect inspection of hollow fiber modules.

DESCRIPTION OF SYMBOLS 10 Hollow fiber module inspection apparatus 101 Module support apparatus 102 Light irradiation apparatus (102A1, 102B1, 102A2, 102B2)
103 Imaging devices (1031, 1032, 103L1, 103L2)
104 Image processing device (1041, 1042)
105 Operation command device 106 Display device 107, 108 Transfer device M Hollow fiber module a Case b Hollow fiber c Resin part b1 Hollow fiber thick part b2 Hollow fiber hollow part b3 Hollow fiber side face c1 Resin part to hollow fiber What is imaged as an enclosed gap

Claims (7)

A step of binarizing a captured image of an end face of the hollow fiber module with a first threshold value to obtain image data A;
Binarizing the captured image with one or more thresholds different from the first threshold to obtain one or more image data X;
Measuring the feature quantity of the image data A, and separating the image data C from which the normal thread candidates are extracted based on the measurement results and the image data D from which the defect candidates are extracted;
Creating image data E by expanding and contracting the image data C;
Removing the image data E from the image data D to obtain the image data F;
Obtaining image data G obtained by combining the image data C and the image data F;
Calculating any two of the image data A and the one or more image data X to obtain the image data H;
Calculating the image data G and the image data H to obtain the image data I;
Measuring a feature amount of the image data I;
A method for inspecting a hollow fiber module. One of the image data X is image data B obtained by binarizing the captured image with a second threshold value that is greater than the first threshold value,
The method for inspecting a hollow fiber module according to claim 1, wherein the step of acquiring the image data H is a step of calculating the image data A and the image data B to acquire the image data H. One of the image data X is image data J acquired by binarizing the captured image with a third threshold smaller than the first threshold,
The method for inspecting a hollow fiber module according to claim 1, wherein the step of acquiring the image data H is a step of calculating the image data A and the image data J to acquire the image data H. One of the image data X is image data B obtained by binarizing the captured image with a second threshold value that is greater than the first threshold value, and the other one of the image data X is the image capturing device. Image data J obtained by binarizing an image using a third threshold smaller than the first threshold,
The hollow fiber module inspection method according to claim 1, wherein the step of acquiring the image data H is a step of calculating the image data B and the image data J to acquire the image data H.   The calculation in the step of acquiring the image data H is an operation of subtracting image data having a larger threshold used for binarization from image data having a smaller threshold used for binarization. The inspection method of the hollow fiber module as described in any one of -4.   The method for inspecting a hollow fiber module according to any one of claims 1 to 5, wherein the step of measuring the feature amount of the image data I includes a step of measuring the roundness of the hollow fiber. In a hollow fiber module inspection apparatus comprising means for irradiating light to an end face of a hollow fiber module, means for imaging reflected light from the end face, and image processing means,
The image processing means includes
The captured image obtained by the imaging means is binarized with a first threshold to obtain image data A, and the captured image is binarized with one or more thresholds different from the first threshold. A binarization processing unit for acquiring one or more image data X;
An image data classification unit that measures the feature amount of the image data A, and classifies the image data C from which normal thread candidates are extracted and the image data D from which defect candidates are extracted based on the measurement results;
A mask image creating unit that creates image data E by expanding and contracting the image data C;
A mask image removing unit that removes the image data E from the image data D to obtain image data F;
A whole-thread image creation unit that obtains image data G by combining the image data C and the image data F;
An image data calculation unit (1) that calculates any two of the image data A and the one or more image data X to obtain image data H;
An image data calculation unit (2) for calculating the image data G and the image data H to obtain image data I;
An inspection apparatus for a hollow fiber module, including a calculation unit that measures a feature amount of the image data I.