JP4765890B2 - Foreign object detection device - Google Patents

Description

  The present invention relates to a foreign object detection device and a foreign object detection method. More specifically, the present invention relates to a foreign object detection device that images a liquid that is an object to be inspected and analyzes the captured image to determine the presence or absence of foreign objects in the liquid. The present invention relates to a foreign object detection method.

Conventionally, various apparatuses and methods for measuring fine particles such as foreign matters contained in a liquid have been developed in order to examine the degree of foreign matter contamination in the liquid. For example, a particle counter is used as such a foreign matter detection device in a liquid. In the particle counter, a liquid to be inspected is poured into a thin transparent tube, and the tube is irradiated from the outside with an illumination light source such as a laser beam. Then, the illumination light is received by a light receiving element arranged to face the illumination light source with the tube interposed therebetween. If foreign matter is mixed in the liquid, the illumination light is blocked or scattered by the foreign matter, and the amount of light reaching the light receiving element is reduced, so by measuring the amount of illumination light and the number of pulses, The particle size and the number of particles in the liquid can be measured.
However, when measuring using a particle counter, since the diameter of the tube through which the liquid passes needs to be made as thin as the size of the light receiving element, only about 20 ml of liquid can flow for 60 seconds, for example. Therefore, there is a problem that the measurement time becomes very long when the amount of the liquid to be measured is large. In addition, when bubbles are included in the liquid, scattering of illumination light occurs even in the bubbles, and there is a possibility that the bubbles are erroneously determined as foreign matters.

  As another method, there is used an apparatus that detects a foreign substance attached to a paper filter by flowing a liquid to be inspected through a paper filter for a certain period of time and then taking a picture after removing the paper filter. In this apparatus, since the liquid to be inspected can be flowed in a large amount at a time by increasing the diameter of the filter, there is an advantage that measurement can be performed in a short time as compared with the particle counter. However, in this apparatus, the paper filter has to be replaced every time measurement is performed, and a very troublesome labor is required. In particular, when this apparatus is used in an inspection process of a manufacturing process in which a plurality of inspection objects are inspected continuously, the paper filter must be frequently replaced, which is very inconvenient.

  On the other hand, an inspection window is provided in the measuring part of the tube through which the liquid to be inspected flows, and the inspection window is irradiated with illumination light, and the light transmitted through the liquid or the light reflected by the liquid is received to remove the liquid. A foreign substance inspection system has been developed that includes an imaging means for imaging and detects a foreign substance in real time by comparing a portion where the level of a video signal captured by the imaging means changes rapidly with a reference level of an inspection object. (See Patent Document 1). Since this foreign matter inspection system uses a line sensor or an area sensor as the image pickup means, the width of the inspection window can be made wider than that of the measurement unit of the particle counter. Therefore, there is an advantage that the measurement time can be shortened. Furthermore, unlike the apparatus using a paper filter, it is not necessary to replace parts every time measurement is performed, and complicated labor is not required.

  However, bubbles are often generated, for example, when liquid is poured into the inspection system. Since the light applied to the bubbles is scattered in the same manner as the foreign matter, when the liquid containing the bubbles is illuminated and imaged, the video signal rapidly decreases at the bubble portion. For this reason, in the foreign matter inspection system described in Patent Document 1, there is a possibility that bubbles included in the liquid to be measured are erroneously detected as foreign matter, similarly to the particle counter.

  Therefore, the number of man-hours required for measurement is small, it is possible to measure in a short time, and even if bubbles are contained in the liquid to be inspected, it is possible to distinguish foreign substances from bubbles and have good accuracy. Development of a foreign object detection device and a foreign object detection method capable of detecting foreign objects with the use of the sensor is desired.

Japanese Patent No. 3048049

In view of the above problems, the present invention can detect foreign objects with good accuracy by distinguishing foreign substances from bubbles even when bubbles are included in the liquid to be inspected. An object is to provide an apparatus and a foreign object detection method.
It is another object of the present invention to provide a foreign object detection device and a foreign object detection method that can be inspected in a short time without the need for parts replacement during the inspection.

According to the first aspect of the present invention, the foreign object detection device according to the present invention illuminates the transmission part (24) visible from the outside provided in the measurement tube (2) through which the liquid to be inspected flows. In synchronization with the light emission of the light source unit (3), the imaging unit (4) that acquires the inspection image obtained by photographing the liquid flowing through the transmission unit (24), and the inspection image is increased by binarizing with a predetermined threshold value A foreign substance candidate area detecting means (54) for detecting a low luminance area as a foreign substance candidate area by dividing into a luminance area and a low luminance area, and feature quantity extraction for extracting at least one feature quantity indicating the circularity of the foreign substance candidate area Means (55) and determination means (56) for determining that the foreign substance candidate region is a foreign object image when at least one feature quantity extracted by the feature quantity extraction means (55) satisfies a predetermined condition. It is characterized by having.
With such a configuration, it is not necessary to replace parts during inspection, and it is not necessary to make the measuring tube (2) as thin as a particle counter, so that inspection can be performed in a short time. Moreover, since it is possible to distinguish and recognize the bubbles and the foreign matters, the foreign matters can be accurately detected. Note that the high luminance region represents a region where the amount of light received by the imaging unit (4) is relatively large when acquiring the inspection image, and the low luminance region is a relative amount of light received by the imaging unit (4). Represents a small area.
The feature amount is a circumferential luminance difference representing the luminance difference near the edge of the foreign substance candidate region, and the feature amount extraction means (55) obtains the centroid position of the foreign substance candidate region, and the width of the foreign matter candidate region from the centroid position. Alternatively, it is preferable to calculate a difference between the maximum value and the minimum value of the pixel values for a plurality of pixels existing at a predetermined distance of ½ or less of the height to obtain a circumferential luminance difference.

According to a second aspect of the present invention, the feature amount includes a degree of symmetry representing the symmetry of the foreign substance candidate region, and the feature amount extraction means (55) determines the positions of the vertical center line and the horizontal center line of the foreign matter candidate region. When the foreign matter candidate areas are folded and overlapped with the vertical center line as the axis, the number of pixels where the foreign substance candidate areas do not overlap with each other, and when the foreign matter candidate area is folded and overlapped with the horizontal center line as the axis seeking total number of pixels the candidate region between do not overlap, it is not preferable to symmetry.
By using as a feature pairs Shodo or circumferential luminance difference, even if foreign matter or bubbles image of have a substantially perfect circle shape, because there are only several tens of pixels in size on the test image, Even when the corresponding foreign substance candidate region cannot accurately reproduce the shape of the image, information capable of discriminating between a circle and a shape other than a circle can be given. Therefore, an increase in the number of pixels and an increase in the data amount of the two-dimensional detector used for acquiring the inspection image can be suppressed.

Furthermore, as described in claim 3 , the foreign substance candidate region detection means (54) sets a small region including the target pixel for the target pixel of the inspection image, obtains a representative signal value of the small region, and sets it as the target pixel. By setting the value of the corresponding pixel, a background image is created, a difference image between the background image and the inspection image is created, an average pixel value of the difference image is calculated, and a predetermined binarization threshold is calculated based on the average pixel value Is preferably calculated. With such a configuration, even when the brightness of the inspection image locally changes due to illumination unevenness, it is possible to prevent a part other than a foreign substance or a bubble image from being erroneously detected as a foreign substance candidate area.

Furthermore, as described in claim 4 , the feature quantity extraction means (55) extracts at least two types of feature quantities, and the determination means (56) determines which one of the extracted feature quantities is each feature quantity. It is preferable to determine as a foreign object when a predetermined condition is satisfied.

Furthermore, as described in claim 5 , the amount of movement of the liquid during the light emission time per light emission of the light source unit (3) is the size of the pixel of the inspection image in the inspection image. It is preferable that the time is ½ or less. With such a configuration, since an image of a foreign object or a bubble can be taken in a stationary state, a feature amount that accurately reflects the shape of the foreign object candidate area can be extracted.

According to another aspect, a foreign object detection method according to the present invention is provided. The foreign matter detection method includes a step of obtaining an inspection image by photographing a liquid as an inspection object in synchronization with light emission of an illumination light source, and creating a binary image obtained by binarizing the inspection image with a predetermined threshold value. Detecting a low-intensity area of the binarized image as a foreign substance candidate area, extracting at least one feature quantity representing the circularity of the foreign substance candidate area, and at least one extracted by the feature quantity extraction means A step of determining that the foreign object candidate region is an image of a foreign object when the feature amount satisfies a predetermined condition.

  In addition, the code | symbol in the parenthesis attached | subjected to each said means is an example which shows a corresponding relationship with the specific means as described in embodiment mentioned later.

Hereinafter, a foreign object detection apparatus to which the present invention is applied will be described in detail with reference to the drawings.
As an example, a foreign object detection apparatus to which the present invention is applied includes a metal (for example, a diesel pump or the like) that is included in a cleaning liquid such as light oil used for cleaning a part that supplies fuel to an engine, or that is attached to or peeled off from a part (for example, , Iron, aluminum) and other foreign matters (length of about 50 μm to 150 μm). The foreign object detection apparatus to which the present invention is applied flows the cleaning liquid through an inspection tube, analyzes an inspection image obtained by photographing the cleaning liquid flowing through the tube, and determines a foreign object candidate region that may be an image of a foreign object or a bubble. To detect. Then, one or more feature amounts that tend to have different properties between the bubble and the foreign matter are obtained from the detected foreign matter candidate region, and when the feature amount satisfies a predetermined condition, an image shown in the foreign matter candidate region is It is determined that the object is a foreign object.

FIG. 1 shows a schematic configuration diagram of a foreign object detection apparatus 1 to which the present invention is applied.
The foreign substance inspection apparatus 1 to which the present invention is applied includes a measurement tube 2, a light source unit 3, an imaging unit 4, and a processing unit 5 in order to perform the above determination.
Hereinafter, each part of the foreign material detection apparatus 1 will be described in detail.

The measuring tube 2 is attached to a storage part 21 for temporarily storing the cleaning liquid, an on-off valve 22 that opens and closes the out-flow outlet, and is connected to the on-off valve 22. And a pipe 23 for flowing down the cleaning liquid stored in the storage unit 21.
The storage unit 21 is formed in a funnel shape, and collects the cleaning liquid used for cleaning the workpiece (diesel pump) 6 in the upper opening. Further, the outlet provided at the lowermost portion of the storage unit 21 can be opened and closed by an on-off valve 22. Therefore, the storage unit 21 can temporarily store the cleaning liquid before the measurement is started by closing the on-off valve 22 to reduce bubbles in the cleaning liquid generated at the time of recovery.

  The on-off valve 22 is controlled to open and close by the processing unit 5. When the cleaning liquid is collected, the control signal for closing the valve is received from the processing unit 5, the valve is closed, and the cleaning liquid is stored in the storage unit 21. On the other hand, at the start of measurement, a control signal for opening the valve is received from the processing unit 5, the valve is opened, and the cleaning liquid stored in the storage unit 21 flows down to the pipe 23. In addition, since the on-off valve 22 can be comprised using a known on-off valve, detailed description of a structure is abbreviate | omitted.

  The tube 23 is formed of a glass tube, optical plastic, or the like. Further, in the middle of the tube 23, a plane transmission unit 24 is provided for irradiating the illumination light illuminated from the light source 3 to the cleaning liquid flowing down the tube 23 and receiving the illumination light transmitted through the cleaning liquid by the imaging unit 4.

  2A is a schematic front view of the flat transmission portion 24, and FIG. 2B is a schematic cross-sectional view of the flat transmission portion 24 taken along the arrow AA 'in FIG. 2A. As shown in FIGS. 2A and 2B, in the plane transmission part 24, the tube 23 is wide in a direction substantially orthogonal to a straight line connecting the light source part 3 and the imaging part 4, and in a direction substantially parallel to the straight line. Is formed thin. Providing such a plane transmission unit 24 can suppress a decrease in the amount of illumination light that passes through the cleaning liquid, and can also reduce the possibility that foreign objects and bubbles in the cleaning liquid will be photographed. In the present embodiment, in the plane transmission part 24, the inner diameter of the tube 23 is 14 mm with respect to a direction substantially orthogonal to the straight line connecting the light source part 3 and the imaging part 4, and 2 mm with respect to a direction substantially parallel to the straight line. did. In addition, an inspection target region 25 imaged by the imaging unit 4 is set at a substantially central portion of the plane transmission unit 24. In the present embodiment, the inspection target area 25 is an area having a width of 8 mm and a height of 6 mm.

  The light source unit 3 irradiates the plane transmission unit 24 provided on the tube 23 with illumination light at regular time intervals so that the cleaning liquid flowing down the tube 23 can be continuously photographed. In addition, the light source unit 3 emits intense light in a very short time so that the image of foreign matter or bubbles does not blur on the inspection image due to the flow of the cleaning liquid as the subject. It is preferable to do. For example, it is preferable that the light emission time per time is set to a period in which the moving amount of the cleaning liquid during the light emission time is ½ pixel or less in the inspection image photographed by the imaging unit 4. Therefore, in the present embodiment, the light source unit 3 is composed of a strobe light, and the light emission time per time is set to 3.5 μsec.

  Based on the control signal received from the processing unit 5, the imaging unit 4 images the inspection target region 25 set on the flat transmission unit 24 provided on the tube 23 in synchronization with the light emission of the light source unit 3. Get an image. For this purpose, the imaging unit 4 includes a two-dimensional detector composed of a photoelectric converter such as a CCD or C-MOS sensor, and an imaging optical system that forms an image of the inspection target region 25 on the two-dimensional detector. Have In this embodiment, a CCD sensor having 1024 × 768 pixels is used as the two-dimensional detector. The imaging optical system was configured such that the inspection target area 25 was imaged on the CCD, and one pixel of the CCD represented about 8 μm × about 8 μm on the inspection target area 25. In addition, the imaging unit 4 performs imaging at a rate of about once every 80 milliseconds. Note that the number of pixels of the detector and the configuration of the imaging optical system are not limited to the above, and the size of the foreign object image to be detected is such that the feature amount can be appropriately extracted on the inspection image (for example, the inspection image The average size of the foreign object image is about several tens of pixels), and is optimized according to the embodiment to be implemented. Similarly, the shooting interval, the size of the shooting target area, and the like are also optimized according to the embodiment to be implemented.

  The acquired inspection image has 256 gradations (pixel value 0 to 255) per pixel, and has a larger pixel value as the amount of received light increases. The acquired inspection image is transmitted to the sequential processing unit 5.

  The processing unit 5 detects a foreign substance candidate region that may be an image of a foreign substance based on each inspection image acquired from the imaging unit 4, extracts a plurality of feature amounts from the foreign matter candidate region, and based on the feature amounts. The presence / absence of a foreign object is determined by performing a predetermined determination process.

FIG. 3 shows a functional block diagram of the processing unit 5. The processing unit 5 includes a storage unit 51, a control unit 52, a communication unit 53, a foreign substance candidate region detection unit 54, a feature amount extraction unit 55, and a determination unit 56, and includes, for example, a personal computer (PC) and its peripheral devices. Is done.
Hereinafter, each part of the processing unit 5 will be described in detail.

  The storage unit 51 includes a random access memory (RAM) or a nonvolatile memory such as a magnetic disk, an optical disk, or a flash memory, and temporarily stores the inspection image received from the imaging unit 4. The storage unit 51 stores a program for controlling the processing unit 5 and the like.

  The control means 52 is composed of a central processing unit (CPU) of a PC, a semiconductor memory such as a read only memory (ROM), a random access memory (RAM), etc., and operates according to a program read into the CPU. 2, each unit of the light source unit 3, the imaging unit 4, and the processing unit 5 is controlled. Further, the communication unit 53 is an input / output interface that transmits and receives control signals and image data between the processing unit 5, the measurement tube 2, the light source unit 3, and the imaging unit 4. USB, SCSI, RS232C, Ethernet (registration) (Trademark) and other various I / O ports and their drivers. Then, the processing unit 5 receives each inspection image from the imaging unit 4 through the communication unit 53. On the other hand, the control signal generated by the control unit 52 is transmitted to the measurement tube 2, the light source unit 3, and the imaging unit 4 through the communication unit 53. Further, the processing unit 5 outputs the detection result of the foreign matter determined by the determination unit 56 to an operation display unit (not shown) for displaying the determination result or the like or an external device through the communication unit 53. Or you may make it produce the image which changed the color of the foreign material candidate area | region determined with the image of the foreign material on the test | inspection image, and displays it on the operation display part.

  The foreign object candidate area detection unit 54, the feature amount extraction unit 55, and the determination unit 56 are implemented as program modules executed on the CPU, for example. Alternatively, the foreign object candidate area detection unit 54, the feature amount extraction unit 55, and the determination unit 56 may be mounted as a dedicated processing board including an image processing processor provided separately from the CPU.

The foreign object candidate area detecting means 54 distinguishes the image of the foreign object or the bubble from the area where only the cleaning liquid is photographed, and detects it as a foreign object candidate area.
FIG. 4 shows a schematic diagram of the inspection image. In FIG. 4, minute dark regions 401 and 402 included in the inspection image 400 are a foreign object image and a bubble image, respectively. If foreign matter or bubbles are present in the cleaning liquid, they scatter or absorb the illumination light. Therefore, as shown in FIG. 4, on the inspection image 400, the pixel value of the foreign object image 401 or the bubble image 402 is lower than that of the surrounding area. Therefore, the foreign object candidate area detection unit 54 detects the foreign object candidate area by binarizing the inspection image and labeling a collection area of pixels having pixel values different from the majority of pixels by binarization. .

  Specifically, the foreign substance candidate region detection unit 54 performs a local maximum value filter process on each inspection image acquired from the imaging unit 4 to obtain a background image. The local maximum value filter process is to obtain the maximum value in the mask as the pixel value of the pixel located at the center of the mask while sequentially moving the position of the mask of a predetermined size with respect to the image to be processed. In the present embodiment, the maximum value is obtained in a region larger than the foreign object image 401 and the bubble image 402, and the mask size is set to 31 in order to reliably distinguish the background (only the cleaning liquid portion) from the images 401 and 402. X 31 pixels. However, a larger or smaller mask may be used depending on the size of a foreign object to be detected.

  When creating the background image, the foreign substance candidate area detecting unit 54 obtains a difference image by subtracting the value of each corresponding pixel of the inspection image from each pixel of the background image. In this difference image, the overall luminance difference of the image due to the luminance unevenness of the illumination light is canceled out, and only an area having a pixel value that is locally low on the inspection image, such as the foreign object image 401 or the bubble image 402. Will have a large pixel value on the difference image. Therefore, the foreign substance candidate area detection unit 54 calculates the average pixel value of the difference image. Then, the foreign object candidate area detecting unit 54 binarizes the difference image using the calculated average pixel value as a binarization threshold value, and creates a binarized image. In the binarized image, only the region corresponding to the foreign object image 401 or the bubble image 402 has the pixel value “1”, and the other pixels have the pixel value “0”.

  When obtaining the binarized image, the foreign substance candidate area detecting unit 54 performs a labeling process on the pixel having the pixel value “1”. Then, an area where a predetermined number (for example, 10 pixels) or more of pixels having the pixel value “1” are connected is detected as a foreign substance candidate area. The predetermined number is optimized according to the size of the foreign object to be detected. The foreign substance candidate area detecting means 54 attaches a different label number Ln (= 1, 2,..., N) to each foreign substance candidate area in order to identify each foreign substance candidate area. Then, a label image having the same size as the binarized image is created. For example, in a label image, a pixel corresponding to a foreign substance candidate region with a label number Ln has a pixel value Ln. Since the labeling process is a well-known technique, detailed description is omitted here.

Note that the foreign substance candidate region detection unit 54 may use a value other than the average value of the difference images as the binarization threshold value. For example, a cumulative histogram of pixel values of the difference image may be calculated, and a pixel value that has a predetermined cumulative frequency (for example, 50%) may be used as the binarization threshold. Further, instead of performing the local maximum value filter processing, a background image may be obtained by performing local average value or local median filter processing. Alternatively, if the luminance unevenness of the illumination light does not cause a problem, a cumulative histogram of pixel values is directly obtained from the inspection image, and a pixel value having a predetermined cumulative frequency (for example, 5%) is used as the binarization threshold value. May be.
Furthermore, before labeling the binarized image, discontinuity connection or hole filling processing may be performed by performing a morphological closing operation on a pixel having a pixel value “1”. Conversely, an open operation may be performed to remove isolated points.

  The control unit 52 is configured so that the generated binarized image, the label image, and the number of foreign substance candidate regions (or the maximum value n of the label number Ln) can be used by other units of the processing unit 5. Is temporarily stored in a memory such as a RAM or storage means 51.

The feature quantity extraction unit 55 calculates a feature quantity that tends to have different values for the foreign substance and the bubble based on the foreign substance candidate area detected by the foreign substance candidate area detection unit 54.
FIGS. 5A to 5F show examples of a bubble image and a foreign object image in the inspection image. In FIGS. 5A and 5B, dark areas 501 and 502 are images of bubbles, respectively. In addition, the dark regions 503 to 506 in FIGS. 5C to 5F are images of foreign matters, respectively. As shown in FIGS. 5 (a) and 5 (b), since the bubbles have a relatively regular spherical shape, the image of the bubbles on the inspection image has a shape that is relatively close to a circle. . On the other hand, it is considered that the foreign matter is mainly a metal piece peeled off from the inner wall of the diesel pump. Therefore, as shown in FIGS. 5C to 5F, the image of the foreign material has a relatively irregular shape.
Therefore, in the present embodiment, the feature amount extraction unit 55 assumes that the foreign matter and the bubble tend to have different values, and represents four types of indicators (fillet diameter difference, occupancy rate, symmetry) that indicate the circularity of the foreign matter candidate region. Degree, circumferential luminance difference) is calculated as a feature amount. These feature amounts are calculated for each foreign substance candidate region. Since the feature amount extraction unit 55 performs the same process for each foreign matter candidate region, in the following description, the feature from one foreign matter candidate region is used. The amount extraction will be described.

Each feature amount extracted by the feature amount extraction unit 55 will be described using the table 600 shown in FIG.
In FIG. 6, one feature is described for each column of the table 600. For each row of the table 600, the name of each feature quantity to be extracted is shown in the row 601 from the top. A description of each feature amount is described in a row 602 below the row 602. The lower row 603 shows the tendency of the values assumed as the values of each feature amount for the bubble image, and the lowermost row 604 shows the values assumed as the values of the feature amounts for the foreign object image. Show the trend. Here, out of the feature amounts, the circumferential luminance difference is excluded from the binarized image or the label image obtained by the foreign substance candidate region detecting means 54. The circumferential luminance difference is extracted from the inspection image.

  First, the fillet diameter difference Fd is obtained as a difference between the width and height of a rectangle circumscribing the foreign substance candidate region. Since the bubble image is substantially circular, the fillet diameter difference Fd is substantially zero. On the other hand, since the image of the foreign material is likely to be an irregular shape, the fillet diameter difference Fd is likely to take a value other than zero. In order to obtain the fillet diameter difference Fd, the feature amount extraction unit 55 obtains the coordinates of pixels located at the left, right, top, and bottom ends of the target foreign substance candidate region in the binarized image or the label image. Then, the width of the circumscribed rectangle of the foreign substance candidate region is calculated based on the difference between the horizontal coordinate value of the right end pixel and the left end pixel of the region. Similarly, the height of the circumscribed rectangle of the foreign substance candidate region is calculated based on the difference between the vertical coordinate value of the lower end pixel and the vertical coordinate value of the upper end pixel of the region. And the difference of the calculated width and height is calculated | required and it is set as fillet diameter difference Fd.

  Next, the occupation rate Pr is expressed as the area of the foreign substance candidate region in the area of the circumscribed rectangle of the foreign substance candidate region. Since the bubble image is substantially circular, the occupation rate Pr is about 0.785. On the other hand, since an image of a foreign object is likely to have an irregular shape, the occupation rate Pr is likely to take a value smaller than 0.785. The feature quantity extraction means 55 calculates the area of the circumscribed rectangle by multiplying the width and height of the circumscribed rectangle used for calculating the fillet diameter difference Fd in order to obtain the occupation rate Pr. Also, the feature amount extraction unit 55 refers to the pixel value of the binarized image or the label image, counts the number of pixels corresponding to the foreign substance candidate area in the circumscribed rectangle, and sets the total as the area of the foreign substance candidate area. . Then, the feature amount extraction unit 55 calculates the occupation rate Pr by dividing the area of the foreign substance candidate region by the area of the circumscribed rectangle.

  The degree of symmetry Sd is expressed as the total number of pixels that do not match when the foreign object candidate region is folded around its vertical center line or horizontal center line. Since the image of the bubble is substantially circular, the center line is symmetric in both the vertical and horizontal directions, and the target degree Sd is almost zero. On the other hand, since the image of a foreign object is likely to have an irregular shape, it does not have center line symmetry in either the vertical direction or the horizontal direction or both, and the degree of symmetry Sd is likely to take a large value.

The feature quantity extraction unit 55 obtains the positions of the vertical center line and the horizontal center line in order to obtain the degree of symmetry Sd. Therefore, the feature amount extraction unit 55 sets the average value of the horizontal coordinate values of the left and right end pixels of the foreign substance candidate region obtained when calculating the fillet diameter difference Fd as the position of the vertical center line. Similarly, the average value of the vertical coordinate values of the upper and lower end pixels of the foreign substance candidate area is set as the position of the horizontal center line. Alternatively, the feature amount extraction unit 55 counts the number of pixels included in the foreign substance candidate area in order from the left end or the right end of the foreign substance candidate area, and the horizontal coordinates when the total reaches 50% of the area of the foreign substance candidate area. The value may be the position of the vertical center line. Similarly, in order from the top or bottom of the foreign substance candidate area, the number of pixels included in the foreign substance candidate area is counted, and the vertical coordinate value when the total reaches 50% of the area of the foreign substance candidate area is expressed as the horizontal center line. It is good also as a position.
When the vertical center line is obtained, the feature amount extraction unit 55 folds the right half of the foreign substance candidate area around the vertical center line and overlaps the left half of the foreign substance candidate area. Then, only one of the superimposed pixels counts the number of pixels included in the foreign substance candidate area. Similarly, the feature quantity extraction unit 55 folds the lower half of the foreign substance candidate area around the horizontal center line and overlaps the upper half of the foreign substance candidate area. Then, the number of pixels in which only one of the superimposed pixels is included in the foreign substance candidate region is further counted, and the total of the pixels is calculated. The total is set as the target degree Sd.

  Finally, the circumferential luminance difference Cb is expressed as the difference between the maximum value and the minimum value of the corresponding inspection image pixel values at eight points that are equidistant and equally spaced from the center of the foreign substance candidate region. Since the bubbles are almost spherical, the pixel values at the same distance from the center in the image are considered to be almost equal, so the circumferential luminance difference Cb is almost zero. On the other hand, since the image of a foreign object is likely to have an irregular shape, the circumferential luminance difference Cb is likely to take a relatively large value. In order to obtain the circumferential luminance difference Cb, the feature amount extraction unit 55 first obtains the gravity center position of the foreign substance candidate region based on the binarized image or the label image. Then, the obtained barycentric position is set as the center of the foreign substance candidate region. The feature amount extraction unit 55 then fills the fillet in four directions of up and down, left and right, and four directions of 45 ° obliquely from the upper left to the lower right and from the upper right to the lower left from the pixel of the inspection image corresponding to the center of the foreign substance candidate region. The positions of a total of 8 pixels that are separated by half the width of the circumscribed rectangle obtained when calculating the diameter difference Fd are specified. Then, the feature amount extraction unit 55 obtains the maximum value and the minimum value among the specified pixel values of the eight pixels, and sets the difference as the circumferential luminance difference Cb.

  In the above example, the pixel values of eight points are examined in order to obtain the circumferential luminance difference Cb, but the number of pixels to be examined is not limited to eight. For example, when the area of the foreign substance candidate area is relatively small (for example, 30 pixels or less), the circumferential luminance difference is based on the pixel values of four points that are equidistant from the center of the foreign substance candidate area in the vertical and horizontal directions. Cb may be obtained. Conversely, when the area of the foreign substance candidate region is relatively large (for example, 100 pixels or more), the number of pixels to be examined may be increased from 8 points. In addition, the distance from the center of the foreign object candidate region is not limited to half the width of the circumscribed rectangle, and may be half the height of the circumscribed rectangle, or may be either the width or the height of the circumscribed rectangle. The smaller half value may be used. Further, the value may be smaller than half of the width or height of the circumscribed rectangle.

  Even if the image of a foreign object or bubble has a substantially circular shape, these feature quantities are only about several tens of pixels on the inspection image, so the corresponding foreign object candidate region has the shape of the image. Even when it is not possible to accurately reproduce, it is possible to give information that can distinguish between a circle and a shape other than a circle. Therefore, an increase in the number of pixels and an increase in the data amount of the two-dimensional detector used for acquiring the inspection image can be suppressed.

  The feature quantity extraction unit 55 temporarily stores the four types of feature quantities obtained for each foreign substance candidate area in association with the corresponding foreign substance candidate area in a memory such as a RAM constituting the control unit 52 or the storage unit 51. . This association is, for example, a five-dimensional feature amount having each feature amount Fd, Pr, Sd, Cb and a label number Ln (= 1, 2,..., N) attached to each foreign substance candidate region as an element. A vector vf = (Ln, Fd, Pr, Sd, Cb) is generated and stored in units of feature vectors.

  Based on the feature amounts obtained by the feature amount extraction unit 55, the determination unit 56 determines whether the image is a foreign object image or a bubble image for each foreign object candidate region. Hereinafter, a specific procedure will be described with reference to a flowchart.

  FIG. 7 is a flowchart showing a determination procedure for identifying whether each foreign substance candidate region is an image of a foreign object or an image of a bubble. First, the determination unit 56 searches for and reads out the feature vector vf having the label number Ln (= 1, 2,..., N) of the foreign substance candidate area to be identified (step S101). ).

  Next, the determination unit 56 determines whether or not the occupation rate Pr is included in a predetermined range (step S102). In the present embodiment, the predetermined range is set to 0.67 <Pr <0.91. And when it remove | deviates from the predetermined range, the determination means 56 determines with the foreign material candidate area | region being an image of a foreign material. Then, the determination unit 56 associates the determination result R (= 1) representing the image of the foreign object with the label number Ln representing the foreign object candidate area, or the memory or storage unit 51 constituting the control unit 52 such as a RAM. Is temporarily stored (step S106). On the other hand, if it is within the predetermined range, the determination means 56 compares the fillet diameter difference Fd with the first threshold Th1 (step S103). In the present embodiment, the first threshold Th1 is set to 1. When the fillet diameter difference Fd is larger than the first threshold Th1, the determination unit 56 determines that the foreign substance candidate region is a foreign object image (step S106).

  When the fillet diameter difference Fd is equal to or smaller than the first threshold Th1 in step S103, the determination unit 56 compares (symmetry Sd−5% of the foreign substance candidate region area S) with the second threshold Th2 (step S103). S104). In the present embodiment, the second threshold Th2 is 10.6. If (symmetry Sd−5% of the area S of the foreign object candidate area) is larger than the second threshold Th2, the determination unit 56 determines that the foreign object candidate area is an image of a foreign object (step S106). . On the other hand, when (Symmetry Sd−5% of the foreign substance candidate region area S) is equal to or smaller than the second threshold Th2 in Step S104, the circumferential luminance difference Cb is compared with the third threshold Th3 (Step S105). In the present embodiment, the third threshold Th3 is set to 20. When the circumferential luminance difference Cb is larger than the third threshold Th3, the determination unit 56 determines that the foreign object candidate area is a foreign object image (step S106). On the other hand, when the circumferential luminance difference Cb is equal to or smaller than the third threshold Th3 in step S105, the determination unit 56 determines that the foreign substance candidate region is an image of a bubble. Then, the determination unit 56 temporarily stores the determination result R (= 0) indicating that it is an image of a bubble in association with the foreign object candidate region in the memory or the storage unit 51 constituting the control unit 52 such as a RAM. (Step S107).

  Note that the order of determination for each feature amount is not limited to the above order, and may be arbitrarily changed. Each threshold value and the predetermined range are not limited to the above values, and are optimized according to the nature of the foreign substance to be detected and the assumed size.

  When the determination unit 56 determines whether the image is a foreign object image or a bubble image for all the foreign object candidate areas, the determination unit 56 calculates the total number of foreign objects included in the inspection image and transmits the calculated total number of foreign objects to the operation display unit (not shown) through the communication unit 53. The user is notified of the detection result.

  The operation of the foreign object detection device 1 to which the present invention is applied will be described with reference to FIG. The operation of the foreign object detection device 1 is controlled by the control means 52 of the processing unit 5.

  As shown in FIG. 8, when the inspection is started, the control means 52 of the processing unit 5 transmits a control signal to open the valve to the on-off valve 22 of the measuring tube 2. Then, the flow of the cleaning liquid stored in the storage unit 21 is started (step S201). In addition, the elapsed time from the start of measurement is started. When the flow of the cleaning liquid is started, the imaging unit 4 performs imaging in synchronization with the light emission of the light source unit 3 and acquires an inspection image (step S202). Then, the inspection image is transmitted to the processing unit 5.

  When the processing unit 5 receives the inspection image through the communication unit 53, the foreign substance candidate region detection unit 54 detects the foreign substance candidate region (step S203). Then, a binarized image and a label image representing the foreign object candidate area are created. The detection of the foreign object candidate area is as described above.

  Next, the feature quantity extraction means 55 of the processing unit 5 is assigned a label number Ln (= 1, 2,..., N) as a foreign object candidate area to be determined as a foreign object image or a bubble image. A foreign object candidate area is selected (step S204). In the first selection, Ln = 1. Then, four types of feature quantities, ie, the fillet diameter difference Fd, the occupation ratio Pr, the symmetry degree Sd, and the circumferential luminance difference Cb are extracted from the target foreign substance candidate region (step S205). Note that the feature amount extraction method is as described above. When each feature amount is extracted, the determination unit 56 of the processing unit 5 determines whether the foreign object candidate area is a foreign object image or a bubble image according to the flow shown in FIG. It is stored in association with the foreign object candidate area (step S206).

  Thereafter, the feature quantity extraction means 55 of the processing unit 5 finishes the determination process as to whether or not Ln = n (where n is the number of foreign substance candidate areas), that is, whether all foreign substance candidate areas are foreign matters or bubbles. It is determined whether or not (step S207). If there is a foreign substance candidate area for which feature quantity extraction has not been completed, Ln is incremented by 1 (step S208), and control is returned to step S204. Then, the processes in steps S204 to S208 are repeated.

  On the other hand, when the feature amount extraction has been completed for all the foreign substance candidate regions in step S207, the control unit 52 of the processing unit 5 determines whether or not a predetermined time has elapsed from the start of measurement (step S209). ). If the predetermined time has not elapsed, the control is returned to before step S202, and the processes of steps S202 to S209 are repeated. On the other hand, if it is determined in step S209 that the predetermined time has elapsed, the control unit 52 outputs a detection result such as the total number of detected foreign objects to an operation display unit (not shown) or the like (step S210). Then, the process ends.

  As described above, the foreign matter detection apparatus 1 to which the present invention is applied detects foreign matter with good accuracy by distinguishing foreign matter from bubbles even when the measurement target cleaning liquid contains bubbles. can do. Further, unlike the particle counter, the measurement is performed while flowing the cleaning liquid through a tube having a certain diameter, so that the measurement can be performed in a short time.

The embodiments described above are for explaining the present invention, and the present invention is not limited to these embodiments.
For example, in the above-described embodiment, four types of fillet diameter difference, occupancy ratio, symmetry, and circumferential luminance difference are used as the feature amount. Depending on the required detection accuracy, one to one of these may be used. Only three types of feature values may be used. Alternatively, another feature amount may be used in combination with all or any of the above feature amounts. However, in order to obtain good detection accuracy, it is preferable to use at least two of the above-described feature amounts so that the shape of the foreign substance candidate region can be evaluated from different viewpoints.

  In the above-described embodiment, the determination unit 56 checks whether or not each feature amount satisfies the condition corresponding to the foreign matter, and determines that the feature is a foreign matter if any one of them corresponds to the foreign matter. In other cases, it was determined as a bubble. However, the determination unit 56 can also perform determination using another method, for example, a classification method such as discriminant analysis, neural network, support vector machine, or the like. For example, when discriminant analysis is used, a plurality of foreign substance candidate areas that are known in advance as images of foreign substances or bubbles are prepared in advance, and predetermined statistics (each feature is determined based on the feature value obtained for each foreign substance candidate area. A discriminant function is created by calculating the average value and variance of the quantity. The obtained discriminant function is stored in the storage unit 51 in advance. Then, when the foreign object candidate area to be discriminated is given, the judging means 56 reads the discriminant function from the storage means 51 and inputs the feature quantity extracted from the foreign object candidate area to be discriminated to the discriminant function. The result of discrimination from bubbles or foreign matters is obtained. When using a learning system such as a neural network or a support vector machine, similarly prepare a plurality of foreign substance candidate areas that are known in advance as images of foreign substances or bubbles, and based on the feature value obtained for each foreign substance candidate area. Then, the system is learned using a known learning method such as backpropagation. The learned system is stored in the storage means 51 in advance. Then, when a foreign object candidate area to be discriminated is given, the judging means 56 reads the learned system from the storage means 51 and learns the feature amount extracted from the foreign object candidate area to be discriminated. To obtain the result of discrimination from bubbles or foreign matter.

  Furthermore, the present invention can be applied in addition to the detection of foreign substances contained in the cleaning liquid. The present invention can be suitably applied to cases where foreign objects to be detected are mostly non-spherical unlike bubbles.

  As described above, various modifications can be made within the scope of the present invention according to the embodiment to be implemented.

It is a schematic block diagram of the foreign material detection apparatus to which this invention is applied. (A) is a schematic front view of a plane transmission part, (b) is a schematic sectional drawing of the plane transmission part by arrow AA 'of (a). It is a functional block diagram of a processing part. It is the schematic of a test | inspection image. (A) And (b) is an illustration figure of the image of the bubble on a test | inspection image, (c)-(f) is an illustration figure of the image of a foreign material. It is a schematic explanatory drawing of each feature-value extracted. It is a flowchart which shows the determination procedure of the foreign material candidate area | region in a determination means. It is a flowchart which shows operation | movement of the foreign material detection apparatus to which this invention is applied.

Explanation of symbols

DESCRIPTION OF SYMBOLS 1 Foreign object detection apparatus 2 Measuring tube 21 Storage part 22 On-off valve 23 Pipe 24 Plane transmission part 3 Light source part 4 Imaging part 5 Processing part 51 Memory | storage means 52 Control means 53 Communication means 54 Foreign substance candidate area | region detection means 55 Feature quantity extraction means 56 Determination means

Claims (5)

A measuring tube (2) that forms a flow path of a liquid that is an object to be inspected, and at least a part of the flow path has a transmission part (24) that is visible from the outside;
A light source part (3) for illuminating the transmission part (24);
An imaging unit (4) for acquiring an inspection image obtained by photographing the liquid flowing through the transmission unit (24) illuminated by the light source unit (3) in synchronization with light emission of the light source unit (3);
A foreign substance candidate area detection means (54) for binarizing the inspection image into a high luminance area and a low luminance area by binarizing with a predetermined threshold, and detecting the low luminance area as a foreign substance candidate area;
Feature quantity extraction means (55) for extracting at least one feature quantity representing the circularity of the foreign object candidate area;
A determination means (56) for determining that the foreign object candidate area is an image of a foreign object when at least one of the feature values extracted by the feature value extraction means (55) satisfies a predetermined condition;
I have a,
The feature amount is a circumferential luminance difference representing a luminance difference in the vicinity of the edge of the foreign substance candidate region,
The feature amount extraction means (55) obtains the barycentric position of the foreign substance candidate area, and for a plurality of pixels existing at a predetermined distance of 1/2 or less of the width or height of the foreign substance candidate area from the barycentric position, A foreign object detection device characterized in that a difference between a maximum value and a minimum value is calculated to obtain the circumferential luminance difference . The feature quantity extraction means (55) extracts a degree of symmetry representing the symmetry of the foreign substance candidate area together with the circumferential luminance difference ,
The feature quantity extraction means (55) determines the positions of the vertical center line and the horizontal center line of the foreign substance candidate area, and the foreign substance candidate when the foreign substance candidate area is folded and overlapped with the vertical center line as an axis. The total number of pixels that do not overlap each other when the foreign substance candidate areas are folded and overlapped around the horizontal center line as an axis, and the degree of symmetry is obtained. The foreign object detection device according to claim 1.   The foreign substance candidate region detecting means (54) sets a small region including the target pixel for the target pixel of the inspection image, obtains a representative signal value of the small region, and sets the pixel value corresponding to the target pixel. To create a background image, create a difference image between the background image and the inspection image, calculate an average pixel value of the difference image, and set the predetermined binarization threshold based on the average pixel value. The foreign object detection device according to claim 1, wherein the foreign object detection device calculates. The feature quantity extraction means (55) extracts at least two types of feature quantities,
The determination unit (56) according to any one of claims 1 to 3 , wherein the determination unit (56) determines that the extracted feature quantity is a foreign object when a predetermined condition is satisfied for each feature quantity. Foreign object detection device. The light emission time per time of the light source unit (3) is such that the amount of movement of the liquid during the light emission per time is less than or equal to ½ of the pixel size of the inspection image in the inspection image. The foreign object detection device according to any one of claims 1 to 4 , wherein the foreign object detection device is a period of time.