JP4529728B2 - Method and apparatus for inspecting foreign matter in container - Google Patents

Description

  The present invention uses image processing technology to remove bubbles and foreign substances in liquids such as bottles, ampoules, and packs filled with liquid foods and chemicals using image processing technology. The present invention relates to a method for inspecting foreign matter in a container to be identified and an apparatus therefor.

  In general, in containers made of translucent materials that contain liquids, such as food and chemical bottles, ampoules, and packs, it is necessary to check that foreign substances such as dust and insects are not mixed after the container is sealed. is there. Further, since the bubble is not a foreign substance, it is required to distinguish the bubble from the foreign substance. In order to perform this type of inspection, various techniques have been proposed in which a container is imaged by an imaging means such as a TV camera, and an image processing technique using a computer is applied to an image obtained by the imaging means.

As this type of technology, the external force is applied to the container to induce flow in the liquid, and the feature value of the differential image obtained by differentiating the difference image having the pixel value of the brightness difference of each pixel of the image captured at different times. It is considered that a bubble candidate region that is a candidate for a bubble existing region is used to identify a bubble and a foreign object by confirming displacement of the bubble candidate region with respect to difference images obtained at different times. (For example, refer to Patent Document 1). Further, Patent Document 1 uses a feature that the brightness of the peripheral portion of the bubble is lower than the brightness of the central portion in the image obtained by the transmitted illumination, and corresponds to a floating object in the differential image obtained by differentiating the difference image. Set the circumscribed rectangle circumscribing the area to be set, and set the inner area and the outer area in the ellipse inscribed in the circumscribed rectangle, and the average value of the inner area is set to the average value of the outer area. Also described is a technique for making a bubble candidate when it is higher (brighter) than a predetermined set value.
JP 2004-354100 A

  In the technique described in Patent Document 1 described above, it is determined whether a bubble is a bubble after setting a bubble candidate region. Therefore, it can be said that the accuracy of separation between a bubble and a foreign object is increased by a two-step determination. However, since the property that bubbles rise is used to determine whether or not it is a bubble, a plurality of images taken at different times are necessary, and there is a problem that inspection takes time.

  SUMMARY OF THE INVENTION The present invention has been made in view of the above-mentioned reasons, and an object of the present invention is to detect a foreign substance in a container and an apparatus thereof that can accurately identify bubbles and foreign substances without using a plurality of images. Is to provide.

The invention of claim 1 illuminates a container made of a translucent material containing liquid with a light source, images the light transmitted through the container with an imaging means, and performs image processing on the grayscale image obtained by the imaging means. A foreign matter inspection method for identifying bubbles and foreign matters in a liquid in a container by evaluating, and extracting a region having a change in brightness in a grayscale image as a candidate bubble region that is a candidate for a region containing bubbles A candidate area extraction process, an evaluation area setting process in which two areas, an inner area and an outer area, are set in the bubble candidate area using the distance from the outer peripheral edge of the bubble candidate area as a condition, and the average brightness of the inner area The value is higher than the average value of the brightness of the outer area , and the value obtained by normalizing the inter-class variance between the brightness of the pixels in the inner area and the brightness of the pixels in the outer area is used for the separation. If it is above the specified threshold The bubble candidate region and having a determining process and bubbles.

  According to this method, after extracting a bubble candidate region that is a candidate for a bubble existing region, it is determined whether or not the bubble candidate region is a bubble. A misjudgment with a foreign object can be prevented. In addition, for the bubble candidate area, not only the inner area and the outer area are set and the difference between the average brightness values is used to determine whether the bubble is a bubble, but the lightness separation degree between the inner area and the outer area is evaluated. Therefore, it is possible to accurately identify bubbles and foreign matters. In other words, it is not necessary to use a plurality of images because the bubble and the foreign object are identified by capturing the characteristic of the change in brightness related to the bubble instead of identifying the bubble and the foreign object by utilizing the rising of the bubble. Bubbles can be distinguished accurately in a short time.

  According to a second aspect of the present invention, in the first aspect of the invention, when the evaluation area setting process is such that the inner area and the outer area are in contact with each other, and when the determination process cannot determine a bubble, the evaluation area Change the condition distance in the setting process to widen the outer area and narrow the inner area and execute the determination process again, so that the next outer area can not be set in the evaluation area setting process so that it can be determined as a bubble in the determination process The evaluation region setting process and the determination process are repeated until it becomes.

  According to this method, it is possible to suppress the variation in the determination result due to the difference in conditions when setting the bubble candidate region, and to reduce the possibility of determining the bubble as a foreign object.

  The invention of claim 3 is characterized in that, in the invention of claim 1, the inner region and the outer region are set so as to have a gap between each other in the evaluation region setting process.

  According to this method, since a gap is formed between the inner region and the outer region, the number of pixels used for identifying the bubbles is reduced, and processing in a short time becomes possible. Moreover, the evaluation of the degree of separation is facilitated by not including pixels with lightness intermediate between the inner region and the outer region. That is, if there is a difference in lightness between the inner region and the outer region, the degree of separation becomes larger than when a pixel with intermediate lightness is included, and air bubbles can be easily identified.

  According to a fourth aspect of the present invention, in the third aspect of the invention, when it is not possible to determine a bubble in the determination process, the determination process is performed again by expanding the outer region by changing the distance of the condition in the evaluation region setting process. The evaluation area setting process and the determination process are repeated until it becomes impossible to set the next inner area in the evaluation area setting process or until it can be determined as a bubble in the determination process.

  According to this method, it is possible to suppress the variation in the determination result due to the difference in conditions when setting the bubble candidate region, and to reduce the possibility of determining the bubble as a foreign object.

  According to a fifth aspect of the present invention, in the first to fourth aspects of the invention, in the determination step, after obtaining an average value of the lightness of the inner region, it is determined whether it is large or small with respect to the average value of the lightness of the outer region. Instead, in the outer region, the presence or absence of a connected region that is equal to or more than a specified number of pixels having a lightness lower than the average value of the lightness of the inner region is determined, and when the connected region exists, the separation degree is compared with a threshold value. It is characterized by that.

  According to this method, since it is not necessary to obtain the average brightness value for the outer area after obtaining the average brightness value for the inner area, foreign matter can be removed without using information on all pixels in the outer area. As a result, the processing speed is improved.

  In the invention of claim 6, in the invention of claim 1 to claim 5, in the candidate region extraction process, the outer periphery of the previous candidate region obtained by using the difference in brightness value from the background is approximated by an ellipse, The inside of an ellipse is the bubble candidate region.

  According to this method, even if there is variation in the shape of the previous candidate area extracted from the grayscale image due to environmental conditions at the time of imaging, the inner area and the outer area are separated by using an elliptical area as the bubble candidate area. It can be set under certain conditions, and variation in determination results can be suppressed.

The invention of claim 7 is a foreign matter inspection device in a container, and images a light source that irradiates light to a container made of a light-transmitting material containing liquid and light that is irradiated from the light source to the container and passes through the container. An image processing unit includes an image processing unit, and an image processing device for identifying bubbles and foreign matters in the liquid in the container by performing image processing on the grayscale image obtained by the image capturing unit and evaluating the image. Use the distance between the candidate region extraction means for extracting the region where the brightness is changed as the bubble candidate region that is a candidate for the region where the bubble is present, and the outer periphery of the bubble candidate region as a condition, and place the inside in the bubble candidate region. an evaluation area setting means for setting a second region between the region and the outer region, the average value of the brightness of the inner region is higher than the average value of the brightness of the outer region, and the brightness of the pixels of the brightness and the outer region of the pixel in the inner region of whole the between-class variance of the Using the normalized value is dispersed in isolation, the bubble candidate region and having a determining means and bubbles when the separation degree is equal to or greater than a prescribed threshold value.

  According to this configuration, since the bubble candidate area that is a candidate for the bubble existence area is extracted, the candidate area extracting unit and the determination unit that determines whether the bubble candidate area is a bubble or not are provided. Therefore, it is possible to prevent erroneous recognition of bubbles and foreign matters. In addition, for the bubble candidate area, not only the inner area and the outer area are set and the difference between the average values of the lightness is used to determine whether the bubble is a bubble, but also the lightness separation degree between the inner area and the outer area. Therefore, it is possible to distinguish between bubbles and foreign objects with high accuracy. Therefore, it is not necessary to use a plurality of images, the process of identifying bubbles and foreign matters is simplified, and accurate discrimination can be performed in a short time.

  According to the present invention, since there is a two-stage process of extracting a bubble candidate area that is a candidate for a bubble existing area and a process of determining whether or not the bubble candidate area is a bubble, There is an advantage that misjudgment between a foreign object and a foreign object can be prevented. In addition to setting the inner region and the outer region as the bubble candidate region and not only determining whether the bubble is a bubble by using the difference between the average values of the two values, the degree of lightness separation between the inner region and the outer region can be determined. Since the evaluation is performed, there is an advantage that it is not necessary to use a plurality of images and it is possible to accurately identify bubbles and foreign matters in a short time.

(Embodiment 1)
In the present embodiment, as shown in FIG. 2, the side surface of the container X to be inspected is imaged by the TV camera 2 as the imaging means, and the container X is used by using the video signal including the grayscale information output from the TV camera 2. The presence / absence of foreign matter in the liquid filled in is determined. A light source 3 is arranged on the opposite side of the TV camera 2 across the container X. The light source 3 is used for transmitted illumination, and light transmitted through the container X emitted from the light source 3 is imaged by the TV camera 2.

  The output of the TV camera 2 is input to the image processing apparatus 1 configured by a computer that executes an appropriate program. The image processing apparatus 1 discriminates bubbles and foreign matters in the liquid filled in the container 1 by performing image processing to be described later using a video signal obtained from the TV camera 2. The image processing apparatus 1 can be connected to an external device 5 such as a monitor 4 or a programmable controller for device control.

  As shown in FIG. 3, the image processing apparatus 1 includes an A / D converter 11 that performs analog-digital conversion on a video signal output from the TV camera 2. The output of the A / D converter 11 is a grayscale image and is temporarily stored in a video RAM (VRAM) 12. By storing the grayscale image in the VRAM 12, the grayscale image stored in the VRAM 12 can be output to the monitor 4 through the D / A converter 14. The image processing apparatus 1 includes a memory 14 provided with an image memory area 14a serving as a work area and a program memory area 14b for storing a program used for image processing. Gray images are transferred from the VRAM 12 to the image memory area 14a. Is also passed. In the program memory area 14b, necessary image processing program modules prepared in advance in the FROM 15 as firmware are written. A program module to be written from the FROM 15 to the program memory area 14 b is specified by an operation unit (not shown) provided in the image processing apparatus 1 or a program support apparatus connected to the image processing apparatus 1.

  The program written in the program memory area 14b is executed by the arithmetic processing unit 10, and the processing result is given to the external device 5 through the interface 16 as necessary. The arithmetic processing unit 10 executes a program to perform image processing on the grayscale image stored in the image memory area 14a to identify bubbles and foreign matters in the liquid in the container X. That is, the arithmetic processing unit 10 realizes the following units by executing the program.

  Usually, the image captured by the TV camera 2 includes an image of the background of the container X, as shown in FIG. 4, so that only the inside of the container X is first inspected by the inspection region setting means 10a. An inspection area D1 is set. In the present embodiment, the inspection area setting means 10a extracts the boundary line between the outer peripheral surface of the container X and the background using the Hough transform, and automatically sets only the range inside the container X as the inspection area D1. . The technique for setting the inspection area D1 is not limited to the Hough transform, and a technique for determining the edge of the container 1 from the differential image and setting the rectangular inspection area D1 can also be used.

  When the inspection area D1 is set by the inspection area setting means 10a, the candidate area extraction means 10b evaluates the presence or absence of a bubble candidate area for the pixels in the inspection area D1, and if there is a bubble existence area candidate, the bubble candidate A region D2 (see FIG. 5) is extracted. In order to extract the bubble candidate region D2, first, pixels considered to be included in the bubble are extracted from the grayscale image.

  In this embodiment, as a technique for extracting pixels that are considered to be included in bubbles, in this embodiment, an edge extraction process that generates a differential image and extracts an edge having a maximum differential value, and an image that does not include bubbles as a template in advance is used. There is a background difference process for obtaining a difference between a captured image and a template.

  Each pixel extracted by the technique described above is labeled, and the distances between the labeled areas are evaluated to connect the respective areas in a range that can be regarded as one area. For example, expansion processing for pixels defined for one region is performed, and when connected to other regions by expansion processing, the two connected regions are integrated as one region. The region connected and integrated in this way is defined as a bubble candidate region D2. However, only a region including a predetermined number of pixels or more is used as a bubble candidate region D2 that is a candidate for a region where bubbles exist. This is because a minute area cannot be identified as a bubble or a foreign object from an image obtained by the TV camera 2, and such a minute area is treated as a measurement limit or less.

  The bubble candidate area D2 obtained in this way is superimposed on the gray image and used for masking the gray image. Here, masking means that only pixels in a range that overlaps the bubble candidate region D2 are targets for discrimination between bubbles and foreign matters. When a plurality of bubble candidate regions D2 are obtained, it is determined whether the bubbles are in order from the largest number of pixels.

  In order to determine whether or not the bubble candidate region D2 is a bubble, first, the pixels in the bubble candidate region D2 are labeled according to the distance (number of pixels) from the pixels outside the bubble candidate region D2 ( FIG. 5). In the illustrated example, the pixels outside the bubble candidate region D2 are labeled 0, the minimum distance pixel is labeled 1, and the other pixels are labeled with natural numbers corresponding to the distance. However, the distance is obtained for each pixel in eight vertical and horizontal diagonal directions, and the minimum distance is used. Therefore, the labels of the adjacent pixels in the direction in which the distance is minimum are arranged in numerical order. As described above, the process of labeling the pixels included in the bubble candidate area D2 according to the distance from the outer peripheral edge is referred to as a distance conversion process.

  The masking and distance conversion processing is executed by the evaluation area setting means 10c. Further, the evaluation area setting unit 10c uses the distance obtained by the distance conversion process as a condition, and as shown in FIG. 6, the inner area Da near the center and the outer area Db near the periphery in the bubble candidate area D2. 2 areas are set. Specifically, the inner area Da and the outer area Db are set in units of areas with the same label. In the example of FIG. 6, an area where the label is 1 is an outer area Db, and an area where the labels are 2 to 6 is an inner area Da. However, it is possible to appropriately set where the boundary between the inner area Da and the outer area Db is provided.

  After the inner region Da and the outer region Db are set by the evaluation region setting unit 10c, the determination unit 10d determines whether the bubble is a foreign object or not by using the brightness of each pixel in the inner region Da and the outer region Db. Since transmitted illumination is performed in the present embodiment, the amount of light incident on the TV camera in the case of bubbles is less in the peripheral portion than in the central portion. As an example, FIG. 7A shows an example of a grayscale image corresponding to bubbles, and FIGS. 8A and 9A show examples of grayscale images corresponding to different types of foreign matters. 7B, FIG. 8B, and FIG. 9B show brightness histograms corresponding to the grayscale images in FIG. 7A, FIG. 8A, and FIG. 9A, respectively. Show. As can be seen from the histogram, in the case of bubbles, it can be seen that multiple peaks occur and the contrast between the peaks is relatively large, and in the case of foreign matter, only one peak occurs (or It can be seen that even if multiple peaks occur, the difference in brightness is small. In the example of FIG. 9 (b), a plurality of peaks occur and the difference in brightness is large, but as is clear from FIG. 9 (a), the brightness is higher in the peripheral part than in the central part. The relationship between light and dark is reversed.

  That is, it can be said that in order to distinguish between a bubble and a foreign object, it is only necessary to evaluate the light / dark relationship between the central portion and the peripheral portion and the light / dark difference of the peak in the histogram. Therefore, as described above, the inner region Da and the outer region Db are set in the bubble candidate region D2. In the determination unit 10d, in order to compare the brightness relationship between the central portion and the peripheral portion, the average value of the brightness of the pixels included in the inner area Da and the outer area Db is obtained, and the magnitude of these average values is determined. Compare. The condition that the bubbles should satisfy is that the average value obtained from the inner area Da is larger (brighter) than the average value obtained from the outer area Db.

In addition, the determination unit 10d uses a degree of separation η described below as an evaluation value in order to evaluate the difference in brightness of the peak in the histogram. The degree of separation η is a value obtained by normalizing the interclass variance σ _b ² with the overall variance σ _T ² , and is defined by Equation 1.

Where N is the total number of pixels in the bubble candidate region D2, n _{1 is the} number of pixels in the inner region Da, n _{2 is the} number of pixels in the outer region Db, P _i is the brightness of the pixel i, and P ₁ with an overline , P ₂ , and P _m are the average brightness values of all the pixels in the inner area Da, the average brightness values of all the pixels in the outer area Db, and the average brightness values of all the pixels in the bubble candidate area D2. The degree of separation η takes a value between 0 and 1, and is 1 when the peaks are completely independent in the histogram. Therefore, a threshold value for the degree of separation η is defined, and when the degree of separation η is greater than or equal to this threshold value, the difference in brightness between peaks is sufficiently large in the histogram. That is, the condition that the bubbles should satisfy is that the degree of separation η is equal to or greater than the threshold value.

  In summary, the determination unit 10d satisfies the condition that the average value obtained from the inner area Da is larger than the average value obtained from the outer area Db and the degree of separation η is equal to or greater than a threshold value for the bubble candidate area D2. The bubble candidate area D2 is determined as a bubble.

  The processing of the present embodiment described above is summarized as a flowchart shown in FIG. That is, when a grayscale image is input to the image memory area 14a (S1), the inspection area D1 is set by the inspection area setting means 10a (S2). Next, after extracting the pixel considered to be included in the bubble (S3) in the candidate leakage area extraction unit 10b, the extracted pixel is labeled (S4), and the bubble is connected to extract the bubble candidate area D2. (S5). Further, the number of pixels of the bubble candidate region D2 is evaluated, and if the number of pixels is equal to or less than a predetermined number (threshold), it is excluded from the determination target. S6). Steps S3 to S6 are candidate area extraction processes.

  When the bubble candidate region D2 is extracted, the evaluation region setting means 10c performs masking (S7) and distance conversion processing on the bubble candidate region D2 (S8), and further includes the inner region Da and the outer region Db in the bubble candidate region D2. Are set (S9). Steps S7 to S9 are an evaluation area setting process.

  After the setting of the inner area Da and the outer area Db, the determination process in the determining means 10d is executed, the average value of the brightness of the inner area Da is higher than the average value of the brightness of the outer area Db, and the brightness of the inner area Da. When the separation degree η between the brightness of the outer area Db and the brightness of the outer area Db is equal to or greater than the threshold value, the bubble candidate area D2 is determined as a bubble (S10). Further, when one of the two conditions is not satisfied, it is determined as a foreign object.

  As described above, not only the average brightness of the inner area Da and the outer area Db but also whether the bubble is a foreign object or not is determined using the separation degree η. In both cases, it is possible to discriminate between bubbles and foreign matter, and the calculation of the degree of separation η is simple, so that the processing load for discriminating between bubbles and foreign matter is small, and high-speed processing is possible. Further, in the present embodiment, the inner region Da and the outer region Db are set in a shape reflecting the shape of the bubbles by the distance conversion process, so that the accuracy of the determination result is higher than when the shape of the bubbles is approximated by an ellipse or the like. Becomes higher.

(Embodiment 2)
In the first embodiment, when the condition is not satisfied in step S10, it is immediately determined as a foreign substance. However, in the present embodiment, when the condition is not satisfied in step S10, the range between the inner area Da and the outer area Db is determined. After the change, the condition determination in step S10 is executed again, and the range between the inner area Da and the outer area Db is changed so that when the condition is satisfied in step S10, it is determined as a bubble. That is, when the outer region Db is set as only the region having the label 1 as shown in FIG. 6, if the condition for determining the bubble in step S10 is not satisfied, the outer region Db is labeled 1 as shown in FIG. The area 2 is defined as an outer area Db, and the inner area Da is reduced to an area having labels 3-6. After changing the inner area Da and the outer area Db in this way, the condition determination in step S10 is performed.

  As described above, if the bubble condition is not satisfied in step S10, the outer area Db is enlarged inward and the inner area Da is reduced, and the determination of whether the bubble condition in step S10 is satisfied is repeated. Even in a state where the outer area Db is maximized (that is, a state where only the area labeled 6 is the inner area Da), if the bubble condition is not satisfied in step S10, it is determined as a foreign object. Since the outer area Db is gradually enlarged, it is desirable to set the minimum range that can be set when the outer area Db is initially set. That is, in the illustrated example, only the area with the label 1 is set as the first outer area Db. Other configurations and operations are the same as those of the first embodiment.

  In the present embodiment, the possibility that the bubble candidate region D2 can be determined as a bubble is increased, and therefore it is possible to reduce the waste of determining a bubble as a foreign object and classifying it as a defective product. In addition, the range of the bubble candidate region D2 may vary depending on the conditions for setting the bubble candidate region D2, but because the range between the inner region Da and the outer region Db is changed to identify whether it is a bubble or a foreign object, Bubbles can be accurately identified regardless of variations in the setting of the bubble candidate region D2. In addition, since the bubble candidate area D2 is set for the bubble and then determined under double conditions, the foreign object is not mistaken for a bubble.

(Embodiment 3)
In the embodiment described above, an example in which the boundary between the inner region Da and the outer region Db coincides with each other. However, in the present embodiment, as illustrated in FIG. 11, a gap is formed between the inner region Da and the outer region Db. Provided. In the example shown in FIG. 11, the inner area Da is an area having labels 5 and 6, and the outer area Db is an area having a label 1.

  As described above, when the gap is provided between the inner region Da and the outer region Db, when the bubble candidate region D2 is a bubble, regions having different brightness are clearly separated, and the degree of separation η is increased. Can do. Moreover, since the number of pixels to be handled is reduced, it is possible to expect high-speed processing with a small calculation load. Other configurations and operations are the same as those of the first embodiment.

(Embodiment 4)
In the present embodiment, as in the second embodiment, the technique of the third embodiment changes the size of the outer region Db instead of immediately determining the foreign object when the condition is not satisfied in the determination process of step S10 in FIG. Then, the determination in step S10 is performed again, thereby reducing the possibility of misidentifying bubbles as foreign matters. For example, in the state where the inner area Da and the outer area Db are set as shown in FIG. 11, when the condition of step S10 is not satisfied, the outer area Db is expanded inward as shown in FIG. The region where the labels are 1 and 2 are set as the outer region Db), and the determination in step S10 is performed again. In this way, if the condition of step S10 is satisfied, or the outer region Db is in contact with the inner region Da and the condition of step S10 is not satisfied, it is determined as a foreign object. Other configurations and operations are the same as those in the third embodiment.

(Embodiment 5)
In each of the above-described embodiments, it is assumed that the outer region Db is annularly continuous in the bubble candidate region D2, and the brightness of the outer region Db is substantially equal in the circumferential direction of the outer peripheral edge of the bubble candidate region D2. Depending on the positional relationship among the camera 2, the light source 3, and the bubble, as shown in FIG. 13, the outer region Db does not continue in a ring shape (see the region surrounded by a solid line), and the periphery of the outer peripheral edge of the bubble candidate region D 2 Often the brightness is uneven in the direction. However, even in such a case, if it is a bubble, the average value of the brightness of the inner area Da will be higher (brighter) than the average value of the brightness of the outer area Db.

  Therefore, in the present embodiment, the determination means 10d (step S10 in FIG. 1) first calculates the average value of the brightness of the inner area Da. Thereafter, instead of obtaining the average value of the brightness of the outer region Db, pixels having a lightness lower (darker) than the average value of the brightness of the inner region Da are extracted in the outer region Db, and these pixels are connected regions (target pixels). Are determined. It is determined whether or not the number of pixels in the connected region is equal to or greater than a specified number. If there is a connected area and the number of pixels in the connected area is equal to or greater than the specified number, the degree of separation η is obtained to determine whether or not it is a bubble. If only a pixel having a lightness lower than that of the inner area Da in the outer area Db is used as the degree of separation η, the degree of separation η is likely to increase, and the misperception that bubbles are regarded as foreign matters is reduced.

  Here, if the connected region does not exist, it is unnecessary to calculate the degree of separation η by considering it as a foreign object, so that unnecessary calculation can be omitted, and calculation for obtaining the average value of the outer region Db is unnecessary. Thus, when the presence of a connected region including a predetermined number or more of pixels is confirmed, the calculation shifts to the calculation of the degree of separation η. Therefore, the amount of calculation is reduced as compared with the case where information on all pixels in the outer region Db is used. Other configurations and operations are the same as those of the above-described embodiments.

(Embodiment 6)
In the present embodiment, an edge is extracted in the candidate area extraction process, and a bubble outline is extracted by performing Hough transform for pixels on the edge, and the bubble candidate area D2 is approximated by an ellipse.

  For example, in a grayscale image with uneven brightness in the outer region Db as shown in FIG. 13, even if the pixels are labeled and connected in the candidate region extraction process, the outer peripheral edge of the obtained bubble candidate region D2 The shape does not become smooth. Therefore, when the shape of the outer peripheral edge as it is is reflected as the shape of the bubble candidate region D2, pixels that should originally be light and have the inner region Da may be included in the outer region Db. The case where it is not performed normally occurs.

  Therefore, in the present embodiment, the bubble candidate region D2 extracted by the technique described in the first embodiment is used as a previous candidate region, the previous candidate region is approximated by an ellipse, and the approximated inside of the ellipse is used as the bubble candidate region D2. That is, in the candidate area extraction process, first, the previous candidate area is extracted from the grayscale image by the technique of the first embodiment. Next, an area where the change in density is maximum is extracted for the pixels in the previous candidate area. Although this type of processing method is well known and will not be described in detail, for example, a Sobel filter is used. An edge is extracted by thinning a region where the density change is maximized to one pixel width. The edges thus obtained are considered to be partially continuous and partially divided. For example, if an edge is extracted from the grayscale image shown in FIG. 13, it is considered that the edge is divided at the upper part and the lower right part. If it is an edge obtained from a bubble image, it is usually considered to be an ellipse (including a circle). Therefore, consider the divided edge as a part of the ellipse, perform Hough transform on the pixels on the edge, and approximate Extract possible ellipses. Techniques for approximating an ellipse by Hough transform are described in, for example, Takashi Watanabe and Toshihiro Shibata, “Detection of a missing ellipse using Hough transform and layered image,” Science theory Vol. J73-D-II No. 2 pp. 159-166, Feb. The technique described in 1990 is used.

  As described above, by extracting the edge and using the Hough transform, an ellipse that approximates the outer peripheral edge of the previous candidate region can be set, so the inside of this ellipse is used as the bubble candidate region D2. If the inner region Da and the outer region Db are set for the bubble candidate region D2 thus obtained, a region with high brightness at the center of the bubble candidate region D2 can be set as the inner region Da. The outer region Db can be set as intended, and the separation accuracy between bubbles and foreign matters in the determination process is increased. Note that the same result can be obtained by performing Hough transform on pixels on the outer peripheral edge of the previous candidate area instead of performing Hough transform on the pixels included in the edge.

  By the way, when the outer periphery of the previous candidate area is approximated by an ellipse, it is necessary to determine the major axis, the minor axis, and the inclination angle as the ellipse parameters from the previous candidate area. Normally, if no external force is applied, the bubble moves upward, and therefore the direction of the major axis of the ellipse coincides with the vertical direction in the image. Therefore, the inclination angle in the direction of the major axis of the ellipse with respect to the vertical direction of the image is set as the inclination angle, and the initial value when voting by Hough transform is set to 0 degree. The major axis and minor axis are set as a rectangular area circumscribing the previous candidate area (a rectangle having sides in the horizontal and vertical directions of the image), and the longer one of the lengths of each side of the rectangular area is the major axis and the shorter one. This is the short diameter. In this way, an operation for approximating an ellipse is performed using the parameters obtained from the previous candidate area as initial values.

  Instead of setting a rectangular area, the inertial moment of the previous candidate area is obtained, the major axis direction is taken as the major axis direction, the major axis is obtained on the major axis, and the short axis is obtained on the straight line perpendicular to the major axis and passing through the midpoint of the major axis. While obtaining the diameter, the angle of the main axis with respect to the vertical direction of the image may be used as the tilt angle. Other configurations and operations are the same as those of the above-described embodiments.

(Embodiment 7)
Using the method described in the literature shown in Embodiment 6, the procedure for obtaining an ellipse approximated by Hough transform from an edge that is partially continuous and partially divided is as follows.

  First, each pixel on the edge is set as a reference point, and the center of an ellipse serving as a template is positioned at a plurality of reference points, and each point on the template is voted on a voting space. Register the pixels having the maximum number of votes in the voting space, the number of votes, and the template parameters (major axis, minor axis, inclination angle). The position of the pixel from which the maximum number of votes is obtained after repeating the above-described processing while changing the parameter is considered to correspond to the center of the approximate ellipse, and the template parameter is the parameter of the approximate ellipse.

  In other words, using the property that the number of votes is maximized at the center of the approximate ellipse when the parameters of the approximate ellipse and the template ellipse match, the approximate ellipse using the template is obtained. Can do. Here, it is considered that a template is set for each of two reference points on the edge, voting is performed for each of the two templates, and the result of totaling the results is used as the number of votes. If this procedure is used, the difference in the number of votes tends to increase, and the difference in the number of votes can be easily found. In particular, if the distance between the centers of two templates is made large, intersection points are hardly obtained in a template smaller in size than an ellipse to be approximated, and many templates that do not match can be excluded from judgment. Therefore, the processing load is reduced.

  Therefore, in the present embodiment, as in the sixth embodiment, a rectangular area circumscribing the edge to be approximated by an ellipse is set, and the center of the rectangular area is set as the center of the ellipse to be approximated. The edge is divided into four regions by the extended straight line and the straight line extended in the minor axis direction. Then, the template is positioned at the reference points of the two regions that are diagonal positions across the center among the four divided regions, and the number of votes is obtained. Of the four divided areas, the upper right area is the first area, and the second area, the third area, and the fourth area are counterclockwise around the center, and the reference point between the first area and the third area. Each of the two combinations of templates set in the above and the two combinations of templates set as the reference points of the second area and the fourth area respectively vote to evaluate the number of votes.

Assuming that the number of reference points is the same in each of the first region to the fourth region, and the total number of reference points is n, the number of template combinations in this embodiment is (n / 4) × ( n / 4) a × ^2, becomes ⁿ 2/8. Actually, since the number of reference points in each region is not equally divided, it is not (n / 4), but this value is a measure of the number of combinations. Respect, the number of combinations for setting the template by combining all the reference points included in the edge two by two in this, a _n C _2, because it is (n 2 ^-n) / 2, this embodiment When the edge is divided into four regions as in the form and voting is performed by combining reference points included in the diagonal region, the processing load associated with the combination of templates can be reduced. Other configurations and operations are the same as those of the above-described embodiments.

It is a flowchart which shows the process sequence of embodiment of this invention. It is a schematic block diagram which shows an apparatus same as the above. It is a block diagram which shows the image processing apparatus used for the same as the above. It is a figure which shows the concept of the test | inspection area | region in the same as the above. It is a figure which shows the example of the distance conversion process in the same as the above. It is a figure which shows the example of a setting of the inner area | region and outer area | region in the same as the above. (A) is a figure which shows the example of an image corresponding to a bubble in the same as the above, (b) is a figure which shows the histogram corresponding to a bubble in the same as the above. (A) is a figure which shows the example of an image corresponding to a foreign material in the same as the above, (b) is a figure which shows the histogram corresponding to a foreign material in the same as the above. (A) is a figure which shows the example of an image corresponding to a foreign material in the same as the above, (b) is a figure which shows the histogram corresponding to a foreign material in the same as the above. It is a figure which shows the example of a setting of an inner area | region and an outer area | region in the same as the above. It is a figure which shows the example of a setting of an inner area | region and an outer area | region in the same as the above. It is a figure which shows the example of a setting of an inner area | region and an outer area | region in the same as the above. It is a figure which shows the example of an image of a bubble candidate area | region in the same as the above.

Explanation of symbols

DESCRIPTION OF SYMBOLS 1 Image processing apparatus 2 TV camera (imaging means)
3 Light source 10b Candidate area extraction means 10c Evaluation area setting means 10d Determination means X Container

Claims (7)

A container made of a translucent material containing a liquid is illuminated by a light source, and the light transmitted through the container is imaged by an imaging unit, and the grayscale image obtained by the imaging unit is subjected to image processing and evaluated to evaluate the inside of the container. A foreign substance inspection method for identifying bubbles and foreign substances in a liquid, wherein a candidate area extraction process of extracting an area in which lightness is changed in a grayscale image as a bubble candidate area that is a candidate for an area where bubbles exist; An evaluation region setting process for setting two regions, an inner region and an outer region, in the bubble candidate region using the distance from the outer peripheral edge of the bubble candidate region as a condition, and the average value of the lightness of the inner region is the lightness of the outer region When the value that is higher than the average value and the inter-class variance between the brightness of the pixels in the inner area and the brightness of the pixels in the outer area is normalized by the overall variance is used as the degree of separation, and the degree of separation is equal to or greater than the specified threshold The bubble candidate area Particle inspection method of a container and having a determining process and bubbles.   In the evaluation area setting process, the inner area and the outer area are set so that the boundary is in contact, and when it is not possible to determine the bubble in the determination process, the outer area is changed by changing the distance of the condition in the evaluation area setting process. Repeat the determination process again by narrowing the expanded inner area, and repeating the evaluation area setting process and the determination process until the next outer area cannot be set in the evaluation area setting process or until it can be determined as a bubble in the determination process. The method for inspecting foreign matter in a container according to claim 1, wherein:   2. The foreign substance inspection method in a container according to claim 1, wherein the inner region and the outer region are set to have a gap between each other in the evaluation region setting process.   If it is not possible to determine a bubble in the determination process, the determination process is performed again by expanding the outer area by changing the distance of the condition in the evaluation area setting process, and the next inner area cannot be set in the evaluation area setting process. 4. The foreign matter inspection method in a container according to claim 3, wherein the evaluation region setting process and the determination process are repeated until it can be determined that the bubble is determined in the determination process.   In the determination process, after obtaining the average value of the lightness of the inner region, instead of determining the lightness average value of the outer region, the lightness of the outer region is lower than the average value of the lightness of the inner region. 5. The method according to claim 1, further comprising: determining whether or not there is a connected region having a predetermined number of pixels or more, and comparing the separation degree with a threshold value when the connected region is present. The foreign substance inspection method in the container described.   2. The candidate area extraction process, wherein an outer periphery of a previous candidate area obtained by using a difference in brightness value from a background is approximated by an ellipse, and the inside of the ellipse is set as the bubble candidate area. The foreign matter inspection method in a container given in any 1 paragraph of Claim 5. A light source for irradiating light to a container made of a light-transmitting material containing liquid, an imaging means for imaging light that is irradiated from the light source to the container and transmitted through the container, and a grayscale image obtained by the imaging means is subjected to image processing. An image processing device that distinguishes bubbles and foreign substances in the liquid in the container by evaluating the image, and the image processing device is a candidate for a region where bubbles exist in a region where the brightness changes in a grayscale image. A candidate area extracting means for extracting as a certain bubble candidate area, an evaluation area setting means for setting two areas, an inner area and an outer area, in the bubble candidate area using the distance from the outer periphery of the bubble candidate area as a condition; The average value of the lightness of the inner region is higher than the average value of the lightness of the outer region , and the value obtained by normalizing the inter-class variance between the lightness of the pixels in the inner region and the lightness of the pixels in the outer region by the overall variance It used to, the degree of separation is defined Particle inspection apparatus in the container, characterized by having a bubble and determining means for determining the bubble candidate region when it is a value or more.