JP6187036B2 - Image analysis apparatus and program - Google Patents

Description

  The present invention relates to an image analysis apparatus and program, and is particularly suitable for application to an image analysis apparatus and program for analyzing a polished cross-sectional image of a fiber-reinforced composite material.

  In recent years, development of ceramic matrix composites (CMC), which is a kind of fiber reinforced composite material, has been promoted. CMC is a composite material in which ceramic fibers as thin as hair are reinforced with a base material (matrix), and is characterized by being lightweight and excellent in heat resistance. Utilizing this feature, for example, the use of CMC as an aircraft engine member has been studied and is currently being put to practical use. By using CMC as an aircraft engine member, a significant improvement in fuel efficiency is expected.

  In general, CMC first prepares a woven fabric from a plurality of ceramic fiber bundles and coats the fiber surface with carbon or the like. Next, it is formed through a CVI (Chemical Vapor Infiltration) process for forming a matrix in the fiber bundle and a PIP (Polymer Impregnation and Pyrolysis) process for forming a matrix in the void between the fiber bundles. The ratio of the fiber volume per unit volume (fiber volume ratio) in CMC greatly affects the toughness of CMC.

  That is, the greater the fiber volume ratio, the higher the toughness and the better the heat resistance. Therefore, it is important to evaluate the fiber volume ratio in order to confirm whether the toughness of the formed CMC is sufficient.

  As a method for evaluating the fiber volume ratio of CMC, for example, a CMC polished cross-sectional image is displayed on a display screen, a fiber region is visually determined from the displayed polished cross-sectional image, and only the fiber region is selected and extracted. There is a method of calculating the fiber volume ratio. However, in general, the polished cross-sectional image is an image expressed in gray scale, and the fibers, matrix, and voids, which are main members included in the polished cross-sectional image, are all expressed in gray between white and black.

  In particular, it is difficult to distinguish between the fiber and the matrix covering the fiber because the difference in luminance value is the noise level. Therefore, if it is attempted to visually distinguish and extract only the fiber region from the polished cross-sectional image, there is a problem that it takes an enormous amount of time and the extraction range varies depending on the observer. Therefore, it is desirable to automatically extract the fiber region from the polished cross-sectional image by image analysis using a computer.

  In Patent Document 1, a computer inputs a stained image in which cell nuclei are stained, detects a region of the cell nucleus from the input stained image, calculates a shortest path around the cell including the detected cell nucleus, and calculates the shortest path. A technique for extracting a path as a contour of a cell is disclosed. According to this technique, it is possible to automatically extract the contour of an object (here, a cell) included in an object image.

JP 2008-146278 A

  However, Patent Document 1 discloses a technique for extracting the outline of a cell by using the feature that a target image is a stained image and a stained cell nucleus exists in the cell. Therefore, it is not possible to extract the contour of the target object from the target image that has not been processed such as staining.

  The polished cross-sectional image of CMC is originally a gray scale image, and as described above, it is difficult to distinguish between fibers and a matrix. Therefore, when extracting a fiber area | region from the grinding | polishing cross-sectional image of CMC, the technique of patent document 1 cannot be utilized.

  The method for creating a polished cross-sectional image is to first create a section of the sample, and then embed the section in a mold and fill the periphery of the section with resin. Next, a method is adopted in which the resin is cured, and finally the section is removed from the mold and the surface is polished. Therefore, depending on the resin material and the curing method, the polished cross-sectional image is not necessarily a gray scale image. However, since at least the fibers contained in the polished cross-sectional image cannot be dyed, the technique described in Patent Document 1 cannot be used when extracting the fiber region from the polished cross-sectional image.

  The present invention has been made in view of the above points, and proposes an image analysis apparatus and a program that can easily calculate a fiber volume ratio by extracting a fiber region from a polished cross-sectional image of a CMC.

  In order to solve this problem, in the present invention, in an image analysis apparatus that analyzes a polished cross-sectional image of a fiber-reinforced composite material, an input unit that inputs a polished cross-sectional image, and a calculation region is extracted from the input polished cross-sectional image. Based on the energy function using the initialization processing unit, the edge search processing unit that searches for the contour of the fiber included in the extracted calculation area, and the searched fiber contour is circular, the center and radius of the fiber are determined. And a fiber region determination processing unit that determines a region inside a circle formed by the estimated fiber center and radius as a fiber region.

  In order to solve such a problem, in the present invention, in a program for analyzing a polished cross-sectional image of a fiber reinforced composite material, a first step of inputting a polished cross-sectional image to a computer and calculation from the input polished cross-sectional image are performed. Based on the second step of extracting the region, the third step of searching for the contour of the fiber included in the extracted calculation region, and the energy function using the fact that the searched contour of the fiber is circular, And performing a fourth step of estimating a center and a radius and determining a region inside a circle formed by the estimated center and radius of the fiber as a fiber region.

  According to the present invention, the fiber volume ratio can be easily calculated by extracting the fiber region from the CMC polished cross-sectional image.

1 is an overall configuration diagram of an image analysis apparatus in the present embodiment. It is a flowchart which shows an image analysis process. It is a flowchart which shows the initialization process. It is a flowchart which shows an edge search process. It is a flowchart which shows a fiber area | region determination process. It is a flowchart which shows an end determination process. It is a flowchart which shows an optimization process. It is an image obtained by initialization processing. It is an image obtained by edge search processing. It is an image obtained by fiber area determination processing. It is an image obtained by the end determination process. It is an image obtained by an optimization process.

  Hereinafter, an embodiment of the present invention will be described in detail with reference to the drawings.

(1) Overall Configuration FIG. 1 shows the overall configuration of an image analysis apparatus 1 in the present embodiment. The image analysis apparatus 1 includes a CPU (Central Processing Unit) 11, an input unit 12, an image storage unit 13, a display unit 14, and a memory 15.

  The CPU 11 is a processor that comprehensively controls the operation of the image analysis apparatus 1 in cooperation with various programs stored in the memory 15. The input unit 12 is an interface that receives input from the user, and is, for example, a keyboard or a mouse. The input unit 12 in the present embodiment is also an interface for inputting a polished cross-sectional image of CMC (Ceramic Matrix Composites).

  Here, CMC refers to a CVI (Chemical Vapor Infiltration) process in which a woven fabric is produced with a plurality of ceramic fiber bundles, the fiber surface is coated with carbon, etc., and a matrix is formed in the fiber bundle, and voids between the fiber bundles ) Is a fiber reinforced composite material formed through a PIP (Polymer Impregnation and Pyrolysis) process for forming a matrix. The CMC polished cross-sectional image is an image obtained by polishing this CMC cross-section.

  In the present embodiment, a fiber region is automatically extracted from the CMC polished cross-sectional image. By automatically extracting the fiber region from the polished cross-sectional image, the ratio of the fiber area per unit area (fiber area ratio) in the polished cross-sectional image can be calculated. Moreover, the ratio (fiber volume ratio) of the fiber volume per unit volume of CMC can be calculated from this fiber area ratio. The CPU 11 may calculate the fiber volume ratio. Since the fiber volume ratio greatly affects the toughness of CMC, the toughness of the formed CMC can be evaluated.

  Returning to FIG. 1, the image storage unit 13 is a storage medium that stores a CMC polished cross-sectional image input from the input unit 12 and an image created by processing and correcting the polished cross-sectional image by various image processes. The display unit 14 is a display device such as an LCD (Liquid Crystal Display) that displays a polished cross-sectional image and an image created by processing and correction by various image processes.

  The memory 15 is a storage medium that stores various programs for executing image analysis processing in cooperation with the CPU 11. The various programs include an initialization processing unit 151, an edge search processing unit 152, a fiber region determination processing unit 153, an end determination processing unit 154, and an optimization processing unit 155. Image analysis processing (FIG. 2) executed by these various programs and details of each processing in the image analysis processing (FIGS. 3 to 7) will be described later.

(2) Overall Processing FIG. 2 shows a processing procedure of image analysis processing in the present embodiment. This image analysis process is executed in cooperation with various programs stored in the CPU 11 and the memory 15 when the input unit 12 receives an execution instruction from the user. For convenience of the following description, the processing entity will be described as various programs.

  First, when the initialization processing unit 151 inputs the CMC polishing cross-sectional image, the calculation region extraction image (see FIG. 5) that extracts the calculation region from the polishing cross-sectional image based on the user-specified parameter input from the user via the input unit 12. 8) and an initialization process for creating a fiber candidate region extraction image (FIG. 8) obtained by extracting fiber candidates from the calculation region extraction image (SP1).

  The user-specified parameters input in step SP1 include, for example, a mask range, a matrix luminance value, a void luminance value, and a minimum radius and a maximum radius of the fiber region. The range of the mask is used to create a calculation region extraction image from the polished cross-sectional image.

  The matrix luminance value and the void luminance value are used to roughly divide the matrix region and the void region from the calculation region extraction image to create a fiber candidate region extraction image. Information on the minimum radius and the maximum radius of the fiber region is used when the fiber region is determined by estimating the center and radius of the fiber region in the fiber region determination process (FIG. 5) described later.

  Next, the edge search processing unit 152 performs various image processes such as a contour extraction process using a Sobel filter, an automatic threshold process, and a thinning process on the calculation area extraction image created in step SP1, thereby generating voids. A void contour extraction image (FIG. 9) from which the contour is extracted, a fiber contour extraction image (FIG. 9) from which the fiber contour (edge) is extracted, and a fiber contour thinned image (FIG. 9) in which the fiber contour is thinned are created. Edge search processing is executed (SP2).

  It should be noted that the outline of the fiber cannot be appropriately extracted even if the binarization process is simply executed on the calculation region extraction image. Strictly speaking, when there are about a few fibers included in the calculation region extraction image, it is possible to extract the outline of the fibers simply by executing a binarization process. However, when the size of the calculation region extraction image becomes large, for example, 500 × 500 pixels (250,000 pixels), it is not necessary to execute the global threshold processing simply due to the difference in lighting or polishing state when inputting the image. The contour cannot be extracted properly.

  As described above, the CMC polished cross-sectional image is composed of fibers, matrices, and voids. In particular, the fibers and the matrix are made of the same material (for example, SiC), and the surface layer of the fibers has a matrix whose luminance value is very close to the fibers ( For example, CVI, SPI, PIP, MI matrix, etc.) exist. Therefore, even if the binarization process is simply executed, a part of the matrix is recognized as a fiber and the boundary with the matrix cannot be distinguished.

  Next, the fiber region determination processing unit 153 is based on the information on the minimum radius and the maximum radius of the fiber region in the user specified parameters and the energy function using the circular shape of the fiber region (for the energy function). As will be described later, the fiber region is determined by estimating the center and radius of the fiber region, and a fiber region determination process for creating a fiber region extraction image (FIG. 10) is executed (SP3).

  Next, the end determination processing unit 154 creates a divided fiber candidate extracted image (FIG. 11) by dividing the fiber region determined in step SP3 from the fiber candidate region extracted image created in step SP1, and creates this divided fiber candidate. A distance conversion process is executed on the extracted image to create a distance conversion image (FIG. 11). Then, the end determination processing unit 154 executes end determination processing for confirming that the distance in the distance conversion image is not equal to or greater than a predetermined threshold (SP4).

  Next, the optimization processing unit 155 calculates the union of the fiber regions for the fiber region extraction image after confirming that there is no omission in the extraction of the fiber region by the end determination process, and the adjacent fibers are at least adjacent to each other in the union. Error determination is executed based on the geometrical constraint that the distance exists beyond the radius, and error determination processing for creating an error location extraction image (FIG. 12) is executed (SP5).

  After each process of the above steps SP1-SP5 is performed, CPU11 complete | finishes this image analysis process.

(3) Details of Each Process FIGS. 3 to 7 show details of the initialization process, edge search process, fiber region determination process, end determination process, and optimization process in the image analysis process of FIG. 2 described above. Details of each process will be described below with reference to FIGS.

  FIG. 3 shows a detailed processing procedure of the initialization process. This initialization process is executed when the image analysis process of FIG. 2 shifts to the initialization process of step SP1.

  First, the initialization processing unit 151 inputs a polished cross-sectional image of CMC via the input unit 12 (SP11). Next, the initialization processing unit 151 applies a mask in a range specified by the user to the input CMC polished cross-sectional image, executes mask processing, and calculates a product set of the polished cross-sectional image and the mask range. Extract as a region (SP12).

  At this time, the initialization processing unit 151 creates an image showing the calculation area extracted in step SP12 as a calculation area extraction image. Next, the initialization processing unit 151 inputs the matrix luminance value, void luminance value, and information on the minimum and maximum radius of the fiber as user-specified parameters via the input unit 12 (SP13).

  The initialization processing unit 151 can clearly determine that the calculation region extraction image is a matrix and a void based on the matrix luminance value and the void luminance value among the user-specified parameters acquired in step SP13. The candidate fiber region is extracted by excluding the region (SP14), and the initialization process is terminated.

  At this time, the initialization processing unit 151 creates an image showing the fiber candidate region extracted in step SP14 as a fiber candidate region extracted image.

  FIG. 4 shows a detailed processing procedure of the edge search processing. This edge search process is executed when the image analysis process of FIG. 2 shifts to the edge search process of step SP2.

  First, the edge search processing unit 152 performs a contour extraction process using a Sobel filter on the calculation region extraction image created in the initialization process, and extracts a void contour in the calculation region extraction image (SP21).

  At this time, the edge search processing unit 152 creates an image showing the void outline extracted in step SP21 as a void outline extracted image. Next, the edge search processing unit 152 executes automatic threshold processing using the ISODATA method, which is a kind of K-means method, on the void contour extraction image (SP22), and again executes contour extraction processing using a Sobel filter. Then, the outline of the fiber is extracted (SP23).

  At this time, the edge search processing unit 152 creates an image showing the fiber contour extracted in step SP23 as a fiber contour extracted image. Finally, the edge search processing unit 152 executes thinning processing on the extracted fiber contour (SP24), and ends the edge search processing.

  At this time, the edge search processing unit 152 creates a fiber contour thinned image by superimposing the fiber contour thinned in step SP24 on the calculation region extraction image.

  FIG. 5 shows a detailed processing procedure of the fiber region determination processing. This fiber region determination process is executed when the image analysis process in FIG. 2 shifts to the fiber region determination process in step SP3.

  First, the fiber region determination processing unit 153 searches for a pixel near the contour of the thinned fiber as a center candidate of the fiber region in the fiber contour thinned image created by the edge search process (SP31). Next, the fiber region determination processing unit 153 forms a virtual circle centered on the center candidate obtained as a result of the search, and sets the radius of the formed circle as information on the minimum and maximum radii of the fibers in the user-specified parameters. Operate based on.

  And the fiber area | region determination process part 153 estimates the center and radius of a fiber area | region based on the energy function prepared beforehand (SP32).

  Here, the energy function is a function having a product of a noise rate and a hit rate as an evaluation value. For any one fiber region, there are many noise ratios on the circumference of a circle in which the center candidate is actually the center of the fiber region and the thinned outline pixel is centered on the center candidate. If you do, the value will increase.

  In other words, for any one of the fiber regions, a position that is not the center of the fiber region is set as a center candidate (for example, a pixel in a matrix region or a void region is set as a center candidate), or the center of the fiber region is set as a center candidate. Even when the circle is formed by manipulating the radius, if the circle formed is smaller or larger than the fiber region, the value of the noise ratio becomes smaller.

  In addition, for any one of the fiber regions, the hit rate has a large number of thin outline pixels on the circumference of the circle formed around the center candidate, and the length of the circumference of the formed circle is The shorter the value, the larger the value. In other words, for any one of the fiber regions, when the number of thin outline pixels existing on the circumference of the circle formed when a plurality of circles are formed around the center candidate is the same number The circle with the shorter circumference has a higher hit rate value.

  The energy function is R, the noise rate is Rn, the hit rate is Rh, and in the fiber contour thinned image, the thinning is included on the circumference of the circle formed by manipulating the radius of the circle centered on the center candidate. The energy function R is expressed by the following equation (1), where Non is the number of pixels of the contour, Nin is the number of pixels of the thinned outline contained in the circle, and Lon is the circumference length.

  The fiber region determination processing unit 153 determines a fiber region based on the estimated center and radius of the fiber region (SP33), and ends the fiber region determination process.

  At this time, the fiber region determination processing unit 153 creates an image indicating the fiber region determined in step SP33 as a fiber region extraction image.

  FIG. 6 shows a detailed processing procedure of the end determination processing. This end determination process is executed when the image analysis process in FIG. 2 shifts to the end determination process in step SP4.

  First, the end determination processing unit 154 divides the fiber region determined in step SP3 from the fiber candidate region extraction image created in step SP1 (SP41). At this time, the end determination processing unit 154 creates an image after dividing the fiber region in step SP41 as a post-division fiber candidate extraction image.

  Next, the end determination processing unit 154 compares the post-division fiber candidate extraction image with the pre-division image (referred to as pre-division image) (SP42), and determines whether there is a difference (SP43). When determining whether or not there is a difference, the end determination processing unit 154 performs distance conversion processing using the Dijkstra method on the post-division fiber candidate extraction image, and creates a distance conversion image.

  Then, the end determination processing unit 154 determines whether or not there is an area where the distance value is equal to or greater than a predetermined threshold in the created distance conversion image. If not, it is determined that there is no difference. As the threshold value, for example, the minimum fiber radius input as a user-specified parameter in step SP1 can be set. In this case, it can be determined that there is a possibility that at least the fiber having the minimum radius is included in the region where the distance value is equal to or greater than the threshold value (more than the minimum radius).

  Next, when there is a difference (SP43: Y), the end determination processing unit 154 sets the distance conversion image as a new fiber candidate region extraction image (SP44), and performs edge search processing again on the new fiber candidate region extraction image. And it transfers to step SP2 in order to perform a fiber area | region determination process. The distance conversion image is an image including a region including a fiber region. In contrast, if there is no difference (SP43: N), the end determination processing unit 154 ends this end determination processing.

  FIG. 7 shows a detailed processing procedure of the optimization processing. This optimization process is executed when the image analysis process of FIG. 2 shifts to the optimization process of step SP5.

  First, the optimization processing unit 155 calculates a union for the fiber region determined in step SP3 (SP51). Next, the optimization processing unit 155 performs error determination based on the geometric constraints between adjacent fibers (SP52).

  For example, the optimization processing unit 155 determines that the distance between the centers of adjacent fibers in the extracted union is based on the geometric constraint that the distance between the centers of adjacent fibers is at least larger than the sum of the radii of each other. A location where the distance is equal to or less than the sum of the radii of each other is detected and determined as an error. At this time, the optimization processing unit 155 creates an image indicating the error location as an error location extraction image.

  Next, the optimization processing unit 155 executes the edge search processing, the fiber region determination processing, and the end determination processing only once again for the fiber region including the error portion and the adjacent region (SP53), and this optimization processing is performed. Exit.

(4) Each image FIGS. 8-12 shows the specific example of each image produced by the initialization process in FIG. 3-7 mentioned above, an edge search process, a fiber area | region determination process, an end process, and an optimization process. . Hereinafter, each image will be described with reference to FIGS. 8 to 12 are displayed on the display screen by the display unit 14, for example.

FIG. 8 shows an image created by the initialization process.
The polished cross-sectional image G1 is an image obtained by polishing a cross section of the CMC, and is stored in the image storage unit 13 via the input unit 12. Each image described below is also stored in the image storage unit 13. The polished cross-sectional image G1 mainly includes a fiber region R1, a matrix region R2, and a void region R3.

  The mask image G2 is an image that has been subjected to mask processing by applying a mask in a range specified by the user. The calculation region extraction image G3 is an image of a product set of the polished cross-sectional image G1 and the mask image G2, and is an image to be calculated after the initialization process. The fiber candidate region extraction image G4 is an image after the matrix region R2 and the void region R3 are roughly excluded based on the matrix and void luminance values input via the input unit 12.

FIG. 9 shows an image created by the edge search process.
The void outline extraction image G5 is an image after the outline extraction process is performed on the calculation area extraction image G3, and is an image that has a high brightness value of the void outline E1 and is displayed in white. The fiber contour extraction image G6 is an image after the automatic threshold value processing and the contour extraction processing are performed again on the void contour extraction image G5, and is an image in which the brightness value of the fiber contour E2 is high and displayed in white. . The fiber contour thinned image G7 is an image obtained by executing the thinning process on the fiber contour E2 in the fiber contour extracted image G6 and superimposing the thinned contour E3 on the calculation region extracted image G3.

FIG. 10 shows an image created by the fiber region determination process.
The fiber region extraction image G8 estimates the center and radius of the fiber in the fiber contour thinned image G7 based on the fiber contour E3 in the fiber contour thinned image G7 and the energy function R prepared in advance. This is an image in which the center C1 and the circumference of the circle formed by the center C1 and the radius r1 are superimposed on the calculation region extraction image G3. The region surrounded by the circumference is the extracted fiber region.

FIG. 11 shows an image created by the end determination process.
The post-division fiber candidate extraction image G9 is an image after the fiber region extraction image G8 is divided from the fiber candidate region extraction image G4. The distance conversion image G10 is an image after the distance conversion processing is performed on the post-division fiber candidate extraction image G9. The fiber region re-extracted image G11 is an image after the edge search process and the fiber region determination process are performed again for a region whose distance is equal to or greater than the threshold in the distance conversion image G10.

FIG. 12 shows an image created by the optimization process.
The error location extraction image G12 is an image showing the error location ER1. The error location ER1 in the error location extraction image G12 is a location where some or all of the adjacent fiber regions overlap.

(5) Effect by this Embodiment As described above, according to the image analysis apparatus and program in this embodiment, the shape feature that the shape of the fiber included in the polished cross-sectional image of CMC is circular is utilized. Since the fiber region is automatically extracted from the polished cross-sectional image, the fiber area ratio can be calculated from the extracted fiber region, and the fiber volume ratio can be easily calculated from the calculated fiber area ratio. Therefore, it is possible to greatly reduce the time required for calculating the fiber volume ratio and to prevent the fiber volume ratio calculated for each observer from being varied.

(6) Other Embodiments In the present embodiment described above, when an error location where a part or all of adjacent fiber regions overlap is detected by the optimization process (FIG. 2: SP5), The edge search process, the fiber area determination process, and the end determination process (FIG. 2: SP2 to SP4) are executed once again for the error part and its vicinity area. It may be executed a plurality of times until it is not detected.

  Further, the optimization processing unit 155 determines whether or not the overlapping range is within a predetermined allowable range when part or all of the adjacent fiber regions are overlapped. Is divided from the calculation area extraction image G3 to create a new calculation area extraction image (for example, the calculation area extraction image G3 ′). The area determination process and the end determination process may be executed. Then, the final fiber region may be extracted by calculating the union of the fiber regions obtained from the calculation region extraction image G3 and the new calculation region extraction image G3 '.

  The allowable range is within the center pixel estimated as the center of the fiber region, in addition to considering the above geometric constraint that the distance between the centers of adjacent fiber regions is larger than the sum of the radii of each other. This area may also be determined in consideration of the fact that the center of the actual fiber is located in the center pixel and that the radius can actually take a free value other than the integer value of the pixel.

  As described above, according to another embodiment, even if the fibers do not overlap with each other, an allowable range is considered in consideration of a case where an error is determined as overlapping (and vice versa). Even if they are overlapped, if they are within the allowable range, they are excluded from recalculation targets, so that the accuracy of the finally extracted fiber region can be improved.

1 Image analysis device 11 CPU
12 Input unit 13 Image storage unit 14 Display unit 15 Memory 151 Initialization processing unit 152 Edge search processing unit 153 Fiber region determination processing unit 154 End determination processing unit 155 Optimization processing unit

Claims (10)

In an image analysis device that analyzes a polished cross-sectional image of a fiber-reinforced composite material,
An input unit for inputting the polished cross-sectional image;
An initialization processing unit for extracting a calculation region from the input polished cross-sectional image;
An edge search processing unit for searching for an outline of a fiber included in the extracted calculation region;
Based on the energy function that utilizes the circular shape of the searched fiber, the center and radius of the fiber are estimated, and the area inside the circle formed by the estimated center and radius of the fiber is a fiber region. An image analysis apparatus comprising: a fiber region determination processing unit determined as The energy function is
When a virtual circle including the fiber is formed, the evaluation value increases as the number of pixels indicating the outline of the fiber on the circumference of the virtual circle increases. The evaluation value decreases as there are many pixels indicating the outline of the fiber, and the evaluation value decreases as the circumference of the virtual circle increases.
The fiber region determination processing unit,
A plurality of virtual circles are formed by manipulating the center and radius of the virtual circle for the fiber, and the evaluation value obtained by calculating the energy function is the maximum among the plurality of virtual circles. The image analysis apparatus according to claim 1, wherein the center and radius of the circle are estimated as the center and radius of the fiber. The input unit is
Enter user specified parameters including information on minimum and maximum radii of the fibers,
The fiber region determination processing unit,
The image analysis apparatus according to claim 2, wherein a radius of the virtual circle is manipulated within a range of a minimum radius and a maximum radius of the fiber based on the user-specified parameter. Among the fibers included in the calculation region, when there is a fiber for which the fiber region is not determined by the fiber region determination processing unit, the edge for a region including a fiber for which the fiber region is not determined The image analysis apparatus according to claim 1, further comprising: an end determination processing unit that determines to execute the processing by the search processing unit and the fiber region determination processing unit again. The input unit is
Enter user-specified parameters including information on matrix luminance values and void luminance values,
The initialization processing unit
Based on the user-specified parameter, a fiber region is extracted by dividing a matrix region indicating a matrix and a void region indicating a void from the calculation region,
The termination determination processing unit
Dividing the fiber region determined by the fiber region determination processing unit from the fiber candidate region to extract a post-division fiber candidate region, performing a distance conversion process on the post-division fiber candidate region, and a distance conversion process When there is a pixel whose distance value is equal to or greater than a predetermined threshold among the pixels included in the image after performing the above, the fiber region determination processing unit out of the fibers included in the calculation region The image analysis apparatus according to claim 4, wherein it is determined that there is a fiber for which no determination has been made. Based on the geometric constraints between adjacent fibers, if there is a place where a part or all of the fiber region determined by the fiber region determination processing unit is overlapped, the overlapping part is indicated as an error part. The image analysis apparatus according to claim 1, further comprising: an optimization processing unit that detects as the image processing unit. The optimization processing unit includes:
The image analysis apparatus according to claim 6, wherein it is determined to execute again the processing by the edge search processing unit and the fiber region determination processing unit for a region including the error part. Based on the fiber region determined by the fiber region determination processing unit, the ratio of the fiber area per unit area included in the polished cross-sectional image is calculated, and based on the calculated ratio of the fiber area, the fiber reinforced composite The image analysis apparatus according to claim 1, further comprising: a processor that calculates a fiber volume ratio as a fiber volume ratio per unit volume contained in the material. In the program of the image analysis device that analyzes the polished cross-sectional image of the fiber reinforced composite material,
On the computer,
A first step of inputting the polished cross-sectional image;
A second step of extracting a calculation region from the inputted polished cross-sectional image;
A third step of searching for an outline of a fiber included in the extracted calculation region;
Based on the energy function that utilizes the circular shape of the searched fiber, the center and radius of the fiber are estimated, and the area inside the circle formed by the estimated center and radius of the fiber is a fiber region. A program for executing the fourth step determined as follows. In the computer,
Among the fibers included in the calculation region, when there is a fiber for which the fiber region is not determined by the fourth step, the third region is compared with a region including a fiber for which the fiber region is not determined. And a fifth step for deciding to re-execute the process of step 4 and the fourth step;
Based on geometric constraints between adjacent fibers, if there is a place where a part or all of the fiber region determined by the fourth step is duplicated, the duplicated part is designated as an error part. A sixth step of detecting;
Based on the fiber region determined in the fourth step, the ratio of the fiber area per unit area included in the polished cross-sectional image is calculated, and based on the calculated ratio of the unit area, the fiber-reinforced composite material The program of Claim 9 for performing the 7th step of calculating the ratio of the fiber volume per unit volume contained in as a fiber volume ratio.