JPH10253522A - Image processing method - Google Patents

Abstract

PROBLEM TO BE SOLVED: To achieve a highly reliable recognition of images of liquid drops existing as dispersed phase in a dispersion medium in real time by reading a two-dimensional image data of a dispersion system into an X-Y coordinate for every pixel. SOLUTION: Thresholds are previously set for the discrimination of brightness, the extraction of edges and the coincidence of circle. A two-dimensional image data of a dispersion system is read into an X-Y coordinate system as F (x, y) for every pixel. Then, the image data is separated into masses in terms of brightness and shadow areas by the shading discrimination threshold. The set of the brightness area image data is arithmetically processed by the edge extraction threshold to extract edges thereby recognizing figures of the brightness areas while coordinates of the edges of the recognized figures are determined. Specific brightness area figures are recognized using the circle coincidence threshold. When the recognized brightness areas overlap each other, the areas with smaller radius of circle are excluded and the remaining areas are recognized to be circle areas. Then, the outer rims of the shadow areas extending to the outer circumferences of the respective circular areas are clarified and the minimum dimension of the distance to the outer rim from the center of the circle is defined as outline radius. Thus, an assumed circle area determined with the outline radium as radium is recognized to be an image of a liquid drop.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an image processing method for recognizing an image of a sphere by performing image processing on a two-dimensional image of a three-dimensional system in which spheres are dispersed in a medium. The present invention relates to an image processing method for performing image processing on a two-dimensional image of a dispersion system in which droplets are dispersed as a disperse phase in a dispersion medium to recognize an image of the droplets.

[0002]

2. Description of the Related Art In research, testing, or actual production activities, a two-dimensional image of a three-dimensional system consisting of a medium and a spherical body irregularly dispersed in the medium is processed to exist in the three-dimensional system. It is often necessary to recognize spherical objects.
For example, analysis of suspension polymerization of vinyl chloride, styrene and the like can be mentioned. In the suspension polymerization of polyvinyl chloride (PVC), a vinyl chloride monomer is suspended in water by stirring to carry out polymerization. In the initial stage of polymerization when the polymerization conversion rate is up to several percent, the monomer droplets are subject to shear force due to agitation, interfacial tension at the water / monomer interface, viscosity of the monomer droplets, the action of the dispersant at the water / monomer interface, Division and union are repeated by balance. The frequency (or speed) of fragmentation and coalescence of the monomer droplets per unit time is determined by various factors such as the polymerization reaction temperature, the number of rotations of the stirrer, the type and concentration of the dispersant, and the volume fraction of the monomer phase (dispersed phase). Greatly depends on the conditions. Further, this process has a great influence on the uniform dispersibility of the polymerization initiator, the polymerization additive, and the like in the monomer, the particle size distribution of the final PVC particles, the particle size, and the like, and determines the quality of the final PVC. It is a process. Therefore, for PVC suspension polymerization analysis, it is very useful and indispensable to collect droplet data under various conditions, such as droplet diameter, particle size distribution, and time-dependent change of monomer droplets. Conventionally, ethane dichloride (EDC), butyl chloride, etc. were suspended in water under various conditions as pseudo-vinyl chloride monomers, and sampled suspensions were photographed. Special cameras installed inside and outside the reactor For example, a photograph as shown in FIG. 7 is taken by photographing the suspension in the reactor. Then, a figure recognized as a monomer droplet is selected by human eyes from a large number of image figures in the photograph and recognized as a droplet, and necessary droplet size, particle size distribution, etc. are determined from the droplet image figure. I was getting data.

[0003]

However, a conventional image processing method for taking a photograph and selecting and recognizing a droplet image with human eyes from a myriad of individual complicated images shown in the photograph includes the following. There have been such various problems. The first is difficulty in recognizing image graphics. When light, especially parallel rays, is transmitted and radiated to the droplet of the dispersed phase, which has a spherical shape due to surface tension in the dispersion medium, a refraction phenomenon occurs at the interface between the dispersion medium and the droplet, and the image of the droplet is It is composed of a circular bright core region and a dark region formed annularly around the bright region, and the outline of the dark region is unclear. As a result, it has been extremely difficult to recognize the outline of the image graphic. Second, photographs of areas with a high density of droplets have a large number of individual images in the photograph, and a large number of images are superimposed in a complicated manner, which makes it difficult for human eyes. It was increasingly difficult to sort and recognize. For this reason, it is necessary to take a photograph of an area where the density of the droplet is low and collect necessary droplet data from the photograph, and the obtained data is not always true data.
Data reliability was low. In addition, since data is sorted by human eyes, individual differences occur, and the reliability of data is low in that respect. Third, since it takes a long time to take a photograph, develop, print, and recognize a droplet image by human eyes, it is difficult to collect droplet data in real time. Fourth, the conventional image processing method requires a lot of manpower, and it has been difficult to improve the productivity of data collection work of droplets. As described above, various problems have been described with respect to the conventional image processing method using human eyes, taking data collection of droplets in the analysis of PVC suspension polymerization as an example. However, the present invention is not limited to this. The same problem as described above also occurs when collecting shape data of a spherical body in a three-dimensional system composed of a spherical body floating irregularly in a medium.

Accordingly, it is an object of the present invention to perform image processing on a two-dimensional image of a dispersion system in which droplets are dispersed as a disperse phase in a dispersion medium, and to recognize a droplet image in real time with high reliability. By providing an image processing method and an image processing method for performing image processing on a two-dimensional image of a three-dimensional system in which spheres are dispersed in a medium and recognizing the image of the sphere in real time with high reliability. is there.

[0005]

In developing the image processing method of the present invention, (1) the circle that forms the outer periphery of the dark region of the droplet image is as clear as the circle that forms the periphery of the bright region. Since there is no, first, an easy-to-detect bright area graphic is detected, an image corresponding to the droplet is extracted, and then a dark area annularly present around the bright area of the extracted droplet candidate image is determined.
(2) Since the two-dimensional droplet image is a circular figure, a circular figure among the edge-extracted image figures should be used as a droplet image candidate, and when determining whether the image figure is a circular figure, A method of calculating the degree of circularity, a method of drawing circles having various radii with each pixel as a center, and obtaining drawing coordinates in advance, and comparing the edge coordinates of the image figure with the drawing coordinates, and the like (3. In the three-dimensional dispersion system, since the droplets are also dispersed in the depth direction of the dispersion system, the droplet image forms a circle having a radius of various sizes on a two-dimensional image of the dispersion system. , Are in an overlapping state. In order to collect droplet shape data, it is necessary to make individual separated droplet images. Therefore, among overlapping images, overlapping images with a small radius that are erroneously detected frequently are excluded. The present invention has been completed based on these points.

In order to achieve the above object, an image processing method according to the present invention performs image processing on a two-dimensional image of a dispersion system in which droplets are dispersed as a disperse phase in a dispersion medium to obtain an image of the droplets. An image processing method for recognizing image data, wherein a light / dark discrimination threshold value which is a threshold value for discriminating image data into bright region image data and dark region image data, and a bright region image data is recognized by extracting edges from the bright region image data Edge extraction threshold value,
A first step of presetting a circle coincidence threshold value, which is a threshold value for determining the circular coincidence degree of the bright region graphic whose edge has been extracted, and X- A second step of reading in the Y coordinate system, a third step of distinguishing image data into a set of bright area image data and a set of dark area image data using a light / dark determination threshold as a threshold, and a bright area using an edge extraction threshold as a threshold. A fourth step of performing arithmetic processing on a set of image data to extract edges and recognizing a graphic in each bright area and obtaining edge coordinates of the recognized graphic; A fifth step of certifying, and, when one of the bright regions (hereinafter, referred to as circle candidate regions) recognized as being circular and overlapping with the adjacent circle candidate region, the smaller one of the circle radii of the overlapping circle candidate regions. Weather A sixth step of excluding the area and recognizing the remaining circle candidate area as a circular area; and scanning each circular area in each radial direction from the center of the circle to obtain a dark area extending to the outer periphery of the circular area. The outer edge is determined, and then, a minimum step of the distance from the center of the circle of the circular region to the outer edge is defined as the outer radius, and a seventh step of obtaining an assumed circular region having the outer radius as a radius is performed. And an image recognition step of recognizing the image as a drop image.

In a first step, a light / dark discrimination threshold which is a threshold when discriminating image data into bright area image data and dark area image data, and an edge when light area image data is subjected to edge extraction to recognize a light area graphic. An edge extraction threshold value serving as a threshold value and a circle matching degree threshold value serving as a threshold value for determining a circle matching degree of a bright area graphic whose edge has been extracted are set in advance. In addition, the size of the entire image to be captured by the dispersion system is set in pixel units. The numerical values of the contrast threshold, the edge extraction threshold, and the circle coincidence threshold differ depending on the combination of the dispersing medium and the dispersing phase, the respective substances, the shape, dimensions, properties, etc. of the dispersing medium and the dispersing phase. It is determined in advance by a method such as obtaining. The dispersion medium may be either a gas or a liquid,
The type and properties of the dispersed phase constituting the droplet are not limited. The contrast threshold is set in units of density (gradation), luminance, and the like corresponding to the image data. For example, when image data is read at an inversion gradation of 256 gradations of gray, 200 is set. Image data having a small gradation is defined as image data of a bright area. The edge extraction threshold is, for example, generally about 100 when a bright area image is subjected to edge extraction by a Sobel operator from gray area image data, in terms of an inverted gray level of 256 gray levels. The circle coincidence threshold differs depending on the method of calculating the circle coincidence.

In the second step, the image data of the pixel unit of the dispersion system is read into the XY coordinate system, the X and Y coordinates are assigned to each pixel, and the density of the image for each pixel is indicated by F (x, y). . For example, 0 (black) to 255 (white)
Of the gray 256 gradation original image of 0 (white) to 255
(Black) and read it into the array. The reason for the inversion is to set the degree of overlap of black to a value close to 1.0 to simplify the calculation of the degree of overlap (coincidence) later. At this time, it is preferable to read as converted image data obtained by converting black into fine and strict, and white into large and gradual black-enhanced 8-tone images of 0 (white) to 7 (black). For example, 0 to 21 of 256 gray levels are gray levels 7 and 22 to 25 are 6 and 2 gray levels.
6 to 29 are gradation levels 5, 30 to 38 are gradation levels 4, 39 to 5
5 is the gradation 3, 56 to 82 are the gradation 2, 83 to 128 are the gradation 1, and 128 or more are the gradation 0. By processing the converted image data converted to the black enhanced 8-tone image, the processing in the subsequent steps can be performed at higher speed and with higher precision than in the case of the gray 256-gradation original image.

In the third step, the converted image data is distinguished into a set of bright area image data and a set of dark area image data using the lightness / darkness determination threshold value as a threshold value. By extracting only the inner light area from the light area and dark area of the droplet image,
Overlap between the droplet images can be eliminated, and the detection of the droplet images can be facilitated.

In a fourth step, a set of bright area image data is differentiated by, for example, a Sobel operator, edges are extracted using an edge extraction threshold as a threshold, and an image figure of each bright area is recognized and recognized. Find the edge coordinates of the figure. At the time of edge extraction, when the edge extraction operation value is larger than the edge extraction threshold, the coordinates are set as edge coordinates. Here, the Sobel operator is used for other Robert operators and normal (North) operators.
mal) This is because the outline of the edge can be made thicker than that of the operator. In an image in which figures do not overlap with each other, the detection of the image is facilitated by taking a thick outline. Note that instead of emphasis by differential operation, edge extraction may be performed by emphasis by Laplacian operation or emphasis by difference of local average.

In a fifth step, a specific bright area figure is recognized as a circle using the circle coincidence threshold as a threshold. The method of certification is not particularly limited.

One example of determining a circle in the fifth step is that, in the first step, the degree of circle coincidence between the drawing coordinates of the circle and the edge coordinates of the bright-area graphic extracted as an edge is determined as the circle coincidence threshold. Is set, and in the fifth step,
The drawing coordinates of a circle having a center at each of the coordinates in the bright area figure and different radiuses are obtained, and the image data of the drawing coordinates of the circle and the image data of the edge coordinates of the bright area figure are compared,
A specific bright area figure is recognized as a circle using the circle matching degree threshold as a threshold. In the first step, instead of reading the drawing coordinates of a circle having a center at each pixel (coordinate) and different radiuses from each other and obtaining the drawing coordinates in the fifth step, the drawing read in the first step is used. By using the coordinates, the processing in the fifth step can be speeded up.
In calculating the drawing coordinates, preferably, first, the minimum radius and the maximum radius are set in advance. Next, a circle having the minimum radius is drawn, and the coordinates of the circumference are stored in a file. Next, draw a circle with a radius longer by the unit length, store the coordinates of the circumference in a file, draw a circle with a radius longer by a unit length than that, and save each coordinate of the circumference in a file. This operation is repeated until the circle with the maximum radius is drawn, and the drawing coordinates are obtained. Next, in a fifth step, the image data of the drawing coordinates of the circle and the image data of the edge coordinates of the bright area graphic are compared, and the specific bright area graphic is recognized as a circle with the circle coincidence threshold as a threshold. When the specific bright area graphic is determined to be circular, for example, the image data of each specific bright area graphic is calculated in accordance with the following equation (1) to calculate the degree of circular coincidence. Circle coincidence = {(the edge of a bright area graphic, that is, the gradient of each coordinate on the circumference of the graphic)} / {(the number of pixels on the circumference of the drawing coordinate circle) × (255)} (1) In the calculation of the degree of circle coincidence according to (1), the gradient of each coordinate is represented by binary display, that is, 0 (non-edge portion) and 255 (edge portion). Between the values. The circle matching degree threshold value differs depending on the radius of the circle in the bright area, and is usually set for each of the large radius, the medium radius, and the small radius. For example, if the radius is 11 pixels to 3
For the 0 pixel large radius section, the threshold is 0.8 and the radius is 6
It is assumed that the threshold is 0.9 for a medium radius section of 10 pixels to 10 pixels, and 0.99 for a small radius section of 4 to 5 pixels in radius. If the circle matching degree is larger than the circle matching threshold, the bright area graphic is determined to be circular. If the circle matching degree is smaller than the circular matching threshold, the bright area graphic is not determined to be circular.

Another example of qualifying a circle in the fifth step is:
In the first step, a threshold value for comparing the area of a circle whose circumference is the circumference of the bright region graphic whose edges have been extracted with the area of the bright region graphic is set as a circle matching degree threshold value. The area of a circle whose circumference is the circumference of the bright region graphic whose edges have been extracted is compared with the area of the bright region graphic, and a specific bright region graphic is recognized as a circle using the circle coincidence threshold as a threshold. For example, assuming that the circumference of a bright area graphic is L and its area is S, the area S _{0 of} a circle having the circumference L is calculated by the following equation. S ₀ = L ² / (4π) Next, S / S ₀ is calculated, and when | 1−S / S ₀ | <circle coincidence threshold, the bright area figure is recognized as a circle.

In the sixth step, when one of the bright regions recognized as circular (hereinafter referred to as a circle candidate region) overlaps with the adjacent circle candidate region, the radius of the circle among the overlapping circle candidate regions is determined. The smaller circle candidate area is excluded, and the remaining circle candidate area is recognized as a circle area. For example, between one circle candidate area and another circle candidate area, L> R ₁ + R ₂ (2) where L: between the center of one circle candidate area and the center of another circle candidate area R ₁ : radius of one circle candidate area R ₂ : radius of another circle candidate area It is calculated whether or not the above equation (2) is satisfied. If not, the circle candidate area of smaller radius is calculated. To eliminate.
This calculation is sequentially performed for all combinations of circle candidate areas in the same manner. Here, the reason why the circle candidate region with the smaller radius is excluded is that there are many erroneous detections in the detection of the circle candidate region with the smaller radius.

In the seventh step, for each circular area, scanning is performed in the radial direction from the center of the circle to determine the outer edge of the dark area extending to the outer periphery of the circular area. The minimum dimension of the distances to is defined as the outer radius, and an assumed circular area having the outer radius as the radius is determined. For example, the second
Using the black-enhanced 8-tone image data converted in the step,
As shown in FIG. 8, the center position (C _X , C _Y ) of each circular area
From 0 (white) to 1 (black), then from 1 (black) to 0 (white), from 1 (black) to 0 (white). The distance between the changed point and the center position is determined, and the minimum value of the distance determined in eight directions is defined as the outer radius. In addition, as shown in the upper right direction of FIG. 8, in one of the eight directions, the value changes from 0 (white) to 1 (black) due to the degree of reflection of light when the image of the dispersion system is captured. Next, there may be no point where 1 (black) changes to 0 (white). In this case, there is no need to determine the distance in this one direction. In the seventh step, the ratio between the outer radius of the assumed circular region and the radius of the bright region is calculated for all the assumed circular regions, the average value of the calculated ratios is calculated, and the ratio average value is stored as the radius magnification. Good to do.

Instead of the above-described seventh step, the seventh step can be executed by another method described below. That is, the ratio between the radius of the bright region and the outer radius of the circular area, that is, the outer magnification is previously calculated theoretically or statistically. In the seventh step, the outer magnification is added to the radius of the circular area recognized in the sixth step. Is multiplied to obtain the outer radius.

In the last image recognition step, a circular area where the outer edge of the dark area is determined is recognized as an image of a droplet. Desired data such as the number and size of droplets can be obtained from the recognized image. For example, drawing the obtained image as a droplet,
The number of droplets, the center position of the droplet, and the radius of the droplet are obtained. Furthermore, a histogram showing the distribution of the number of droplets with respect to the droplet radius is drawn, and the shape, average radius, variance, and the like of the distribution are calculated.

In a preferred embodiment of the method according to the present invention, in the first step, an original image coincidence threshold is set as a threshold for judging the coincidence between the original image and the assumed circular area figure. In the step, when image data of the dispersion system is read in the XY coordinate system, the image data is converted into image data having a gradation number smaller than the gradation number of the original image data and read, and in the image recognition step, the image data obtained in the seventh step is obtained. Draws a circle having a radius obtained by multiplying each outer radius by a predetermined coefficient, inversely converts the gradation value of the coordinates on the circumference of the circle to the original image gradation value, and matches the original image matching threshold with the original image matching threshold. A highly circular area is recognized as a droplet image.

In the image recognition step, in order to obtain the original image coincidence, for example, a circle having a radius obtained by multiplying each of the outer radii obtained in the seventh step by the original image coincidence determination radius magnification as a predetermined coefficient is drawn. Then, the original image coincidence is calculated for the 256 gradation inverted image by the following equation (3). The original image coincidence determination radius magnification is a value smaller than 1.0, for example, about 0.9, and is input in advance. Original image coincidence = {(256 gradations of each coordinate on the drawn circle circumference)} / {(total number of pixels on the drawn circle circumference) × (gradation value (= 255)} ( 3) In the above calculation of the original image coincidence, the outer radius is multiplied by the original image coincidence determination radius magnification smaller than 1.0 because a circular shape having a dark area at the outer edge can be always formed. Next, when the original image coincidence threshold is set as a threshold and the original image coincidence is larger than the original image coincidence threshold, a circle region corresponding to the drawn circle is finally recognized as a droplet image. When the original image coincidence is smaller than the original image coincidence threshold, a circle region corresponding to the drawn circle is excluded from the image of the droplet. It can also be obtained from the relationship of the circle matching threshold x the original image matching threshold magnification. In this case, the original image coincidence threshold magnification is a value of about 0.8 smaller than 1.0, and the circle coincidence threshold varies depending on the radius division. And
The original image coincidence threshold magnification is set to the first instead of the original image coincidence threshold.
Set in steps.

In applying the method of the present invention, a special device is not required, and a two-dimensional image of a dispersed system can be captured using a known imaging device, for example, a CCD type imaging device, and an image can be obtained. Specifically, an imaging device provided with means capable of enlarging an actual image is preferable. In addition, a known small computer can be used for setting various thresholds, reading image data, converting, setting drawing coordinates, and performing arithmetic processing in each step.

An image processing method according to the present invention, which is a generalization of the above-described method, performs image processing on a two-dimensional image of a three-dimensional system in which spheres are dispersed in a medium to recognize an image of the sphere. A processing method, wherein a light / dark discrimination threshold that is a threshold when discriminating image data into light area image data and dark area image data, a threshold when recognizing a light area graphic by extracting edges of the light area image data, A first step of setting in advance an edge extraction threshold value and a circle coincidence threshold value that is a threshold value for determining the circular coincidence degree of the bright-area graphic whose edge has been extracted; and F (x, y) into the XY coordinate system, a third step of distinguishing image data into a set of bright area image data and a set of dark area image data using a light / dark discrimination threshold as a threshold, and an edge extraction threshold. As a threshold A fourth step of calculating a set of bright area image data to extract edges, recognizing a figure in each bright area, and finding an edge coordinate of the recognized figure, A fifth step of identifying a circle, and a step of overlapping a bright area identified as a circle (hereinafter referred to as a circle candidate area) with a circle candidate area adjacent thereto.
A sixth step of excluding the circle candidate area having the smaller radius of the circles from among the overlapped circle candidate areas and recognizing the remaining circle candidate area as a circle area; Scan in the radial direction to determine the outer edge of the dark area extending around the outer circumference of the circular area, then use the minimum dimension of the distance from the center of the circle of the circular area to the outer edge as the outer radius, and set the outer radius to the radius. A seventh step of finding an assumed circle area
An image recognition step of recognizing each assumed circular region as a spherical image. The medium of the three-dimensional system is a gas or a liquid, and the spherical body is a liquid or a solid and transmits light.

Further, the above image processing method is generalized,
An image processing method according to the present invention is an image processing method for recognizing a specific set of image data among input image data as a circular figure, and reads the input image data as pixel-by-pixel data into an XY coordinate system. A first step, a second step of comparing the data for each pixel with a first threshold to distinguish each pixel into a first region and a second region, and X-
A third step of obtaining the coordinates of an edge of the pixel set from the pixel set of the first area forming one set in the Y coordinate system; obtaining a circularity of the set of coordinates of the edge; comparing with a second threshold value To recognize if it is a circle
And a step.

[0023]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS An image processing of PVC droplets will be described below as an application example of the method of the present invention, and referring to the accompanying drawings.
Embodiments of the present invention will be described specifically and in detail. FIG.
3 is a flowchart of the method of the present invention. As a first step S _1, set as follows necessary settings such as thresholds, and inputs to the known image processing apparatus for carrying out the present invention method. Image size of dispersed system: 512 × 480 pixels Maximum radius of light area: 30 pixels Minimum radius of light area: 4 pixels Light / dark discrimination threshold: 200 luminance Edge extraction threshold: 100 luminance Circle coincidence threshold Large radius: 0.8 Medium radius: 0.9 Small radius: 0.99 However, the large radius is 11 pixels or more, the medium radius is 10 pixels or less, 6 pixels or more, and the small radius is 5 pixels or less. Original image matching judgment radius magnification: 0.875 Image coincidence threshold magnification: 0.8

In a second step S ₂ , image data of 256 gray levels of the original image shown in FIG. 2 is read and converted into 8-level black enhanced image data as shown in FIG. F (x, y) in which the density is represented by the inversion gradation
As shown. As a third step S _3, to determine the converted image data of the brightness determination threshold value greater than the gradient from the image shown in FIG. 3 as a light area image data to obtain inverted gradient light display region image as shown in FIG. In a fourth step S _4, by differentiating processing bright region image data in FIG. 4 by Sobel operator, the calculated value to recognize each bright area figure by extracting greater coordinate the edge extraction threshold value as an edge, FIG. 5
And the edge coordinates of each bright area figure are determined. In a fifth step S _5, has a center to each coordinate within the bright field graphic obtains the rendering coordinates of a circle for each mutually different radii, and comparing the edge coordinates of the drawing coordinates and bright areas of the circle, the circle Expression (1) using the coincidence threshold as a threshold
A specific light area is determined to be circular according to

In the sixth step S ₆ , when one of the bright regions determined to be circular (hereinafter referred to as a circle candidate region) overlaps with the adjacent circle candidate region, the formula ( According to 2), the circle candidate area with the smaller radius of the circle is excluded, and the remaining circle candidate area is recognized as a circle area.

[0026] In a seventh step S _7, for each circular region scans from the center of the circle in the radial direction, to confirm the outer edge of the dark region extending to the outer periphery of the circular region, then the circle of the circle area center The minimum dimension of the distance from the to the outer edge is defined as the outer radius, and an assumed circular area having the outer radius as the radius is obtained. In the eighth step S _8, the circle having a radius multiplied by a predetermined coefficient (original image matching judgment radius ratio) in each outer radius determined in the seventh step draws the gradation value of the circumference on the coordinate of the circle Inverting to the original image gradation value, and using the original image coincidence threshold as a threshold, the equation (3)
, A circle region having a high degree of coincidence with the original image data is recognized as a droplet image. In a ninth step S _9, as shown in FIG. 6, and draws the resulting image as droplets, the number of droplets, the center position of the drop, determine the radius of the droplet. In this example, the number of detected droplets was 179. Furthermore, a histogram showing the distribution of the number of droplets with respect to the droplet radius is drawn, and the shape, average radius, variance, and the like of the distribution are calculated.

[0027]

According to the structure of the method of the present invention, the dispersion system 2
By reading the image data of the two-dimensional image into XY coordinates for each pixel and performing digital image processing, it is possible to reliably and in real time recognize an image of a droplet existing as a disperse phase in a dispersion medium of a dispersion system. it can. Therefore, using the method of the present invention, the image of the droplet is recognized, and the number of droplets, the droplet radius, the droplet dispersion, and the like in the unit dispersion system are reliably and in real time determined from the image of the recognized droplet. Can be calculated. For example, PVC for producing PVC by a suspension polymerization method
By applying the method of the present invention to data collection of droplets,
The analysis of the suspension polymerization reaction of PVC is facilitated, and the growth of PVC droplets can be quantitatively monitored in real time in terms of actual production.
The productivity of VC manufacturing is improved.

[Brief description of the drawings]

FIG. 1 is a flowchart when the method of the present invention is performed.

FIG. 2 is a copy of a gray 256 tone original image.

FIG. 3 is a copy of an image obtained by converting the image of FIG. 2 into a black enhanced 8-tone image.

FIG. 4 is a copy of an image obtained by extracting a bright region in FIG. 2;

FIG. 5 is a copy of an image obtained by extracting an edge of the image of FIG. 3;

FIG. 6 is a copy of a finally obtained image of a droplet.

FIG. 7 is a transcript of a photograph of PVC in the form of droplets suspended and dispersed in a vinyl chloride monomer solution.

FIG. 8 is a diagram illustrating a dark region extending around the outer periphery of a bright region graphic.

Claims (8)

[Claims] 1. An image processing method for recognizing an image of a droplet by performing image processing on a two-dimensional image of a dispersion system in which droplets are dispersed as a disperse phase in a dispersion medium, comprising the steps of: Brightness / darkness determination threshold value that is a threshold value when discriminating between image data and dark region image data, edge extraction threshold value that is a threshold value when bright region image data is subjected to edge extraction to recognize a bright region graphic, and edge extracted bright region A first step of presetting a circle coincidence threshold value which is a threshold value for determining a circle coincidence degree of a figure; and a step of reading distributed image data into the XY coordinate system as F (x, y) for each pixel. 2 steps; a third step of distinguishing image data into a set of bright area image data and a set of dark area image data using a light / dark discrimination threshold as a threshold; calculating a set of bright area image data using an edge extraction threshold as a threshold Process A fourth step of extracting a figure, recognizing a figure in each bright area, and finding an edge coordinate of the recognized figure, a fifth step of using a circle coincidence threshold as a threshold to recognize a specific bright area figure as a circle, When one of the bright regions (hereinafter, referred to as a circle candidate region) that has been identified as being overlapped with the adjacent circle candidate region, the circle candidate region with the smaller radius of the circle is excluded from the overlapped circle candidate regions. A sixth step of recognizing the remaining circle candidate area as a circle area, and scanning each circle area in the radial direction from the center of the circle to determine the outer edge of the dark area extending around the circumference of the circle area; Next, a seventh step of determining an assumed circular area having a minimum radius as the outer radius and a minimum radius of the distance from the center of the circle to the outer edge of the circular area, and recognizing each assumed circular area as a droplet image Image recognition step to perform Wherein, the image processing method of an image of the droplet in the dispersion. 2. In a first step, a threshold value for determining a circle matching degree between a drawing coordinate of a circle and an edge coordinate of a bright area figure extracted as an edge is set as a circle matching degree threshold value. The drawing coordinates of a circle having a center at each coordinate and a radius different from each other are obtained, the image data of the drawing coordinates of the circle and the image data of the edge coordinates of the bright area figure are compared, and the circle matching degree threshold is set to the threshold. 2. The image processing method of an image of a droplet in a dispersion system according to claim 1, wherein the specific bright region graphic is recognized as a circle. 3. In a first step, a minimum radius and a maximum radius of a droplet are set in advance, and a drawing coordinate of a circle is calculated and stored for each mutually different radius between the minimum radius and the maximum radius. 3. The image processing method of an image of a droplet in a dispersion system according to claim 2, wherein 4. In the first step, a threshold value for comparing the area of a circle whose circumference is the circumference of the bright region graphic whose edges have been extracted with the area of the bright region graphic is set as a circle matching degree threshold value. In the fifth step, the area of the circle whose circumference is the circumference of the bright region graphic whose edges have been extracted is compared with the area of the bright region graphic, and the specific bright region graphic is recognized as a circle with the circle matching threshold value as the threshold value. 2. The image processing method for an image of a droplet in a dispersion system according to claim 1, wherein 5. The ratio between the radius of the bright region and the outer radius of the circular region, that is, the outer magnification, is calculated theoretically or statistically in advance. In the seventh step, the radius of the circular region recognized in the sixth step is determined. The image processing method of an image of a droplet in a dispersion system according to any one of claims 1 to 4, wherein an outer radius is obtained by multiplying the outer radius by an outer magnification. 6. In a first step, an original image coincidence threshold is set as a threshold when judging the coincidence between the original image and the assumed circular area graphic. In the second step, the distributed image data is set. Is read in the XY coordinate system, the original image data is converted and read into image data having a smaller number of gradations than the original number of gradations in order to sharpen the data of interest. Draw a circle having a radius obtained by multiplying each of the outer radii obtained in step 7 by a predetermined coefficient, inversely convert the gradation value of the coordinates on the circumference of the circle to the original image gradation value, and set the original image coincidence threshold value. 6. The image processing method for an image of a droplet in a dispersion system according to claim 1, wherein a circular region having a high degree of coincidence of the original image as a threshold is recognized as an image of the droplet. . 7. An image processing method for recognizing an image of a spherical body by performing image processing on a two-dimensional image of a three-dimensional system in which a spherical body is dispersed in a medium, wherein the image data is defined as bright area image data. Brightness / darkness determination threshold value which is a threshold value for discrimination with dark area image data, edge extraction threshold value which is a threshold value when bright area image data is extracted with an edge to recognize a bright area graphic, and a circle of the bright area graphic whose edge is extracted A first step of presetting a circle coincidence threshold which is a threshold for determining a coincidence, and a second step of reading image data of a distributed system as F (x, y) for each pixel into an XY coordinate system. A third step of distinguishing the image data into a set of bright area image data and a set of dark area image data using a light / dark discrimination threshold as a threshold, and performing an arithmetic processing on the set of bright area image data using the edge extraction threshold as a threshold Edge extraction Then, a fourth step of recognizing the graphic in each bright area and finding the edge coordinates of the recognized graphic, a fifth step of recognizing a specific bright area graphic as a circle using a circle coincidence threshold as a threshold, and recognizing a circular shape When one of the bright regions (hereinafter, referred to as a circle candidate region) overlaps with the adjacent circle candidate region, a circle candidate region having a smaller radius of the circle is excluded from the overlapped circle candidate regions and remains. A sixth step of recognizing the circle candidate area as a circle area; scanning each circle area in the radial direction from the center of the circle to determine the outer edge of the dark area extending around the outer circumference of the circle area; A seventh step of obtaining an assumed circular region having a minimum radius as the outer radius and a minimum radius of the distance from the center of the circle of the region to the outer edge, and image recognition for recognizing each assumed circular region as a spherical image; And a step To the image processing method of the image of spheroids dispersed within the 3-dimensional system. 8. An image processing method for recognizing a specific set of image data among input image data as a circular figure, comprising: a first step of reading input image data as data for each pixel into an XY coordinate system; A second step of comparing each pixel data with a first threshold to distinguish each pixel into a first area and a second area; and a first area forming one set in an XY coordinate system. A third step of obtaining the coordinates of the edge of the pixel set from the pixel set of (i), and a fourth step of determining the circularity of the set of coordinates of the edge and comparing with a second threshold value to determine whether the pixel is a circular figure. An image processing method comprising: