Image analyzing apparatus and program

Patent metadata

Field	Value
Publication number	US-2016247271-A1
Application number	US-201415025983-A
Country	US
Kind code	A1
Filing date	Sep 29, 2014
Priority date	Sep 30, 2013
Publication date	Aug 25, 2016
Grant date	—

Abstract

Proposed are an image analyzing apparatus and program in which the orientation of fiber bundles can be easily analyzed from a three-dimensional image of CMC. Provided is an image analyzing apparatus for analyzing an orientation of a fiber bundle from a three-dimensional image of a fiber-reinforced composite material, comprising an input unit which inputs the three-dimensional image, a binarization processing unit which binarizes the input three-dimensional image and acquires a binary image, an orientation estimation processing unit which estimates each orientation of foreground pixels in the binary image based on an orientation detection filter having a parameter for causing a shape of a detected cross section to have anisotropy, a center extraction processing unit which extracts center pixels showing a center of the fiber bundle from a foreground pixel group, in which the orientation thereof was estimated, based on the orientation detection filter, a fiber bundle connection processing unit which deems center pixels having a same or similar orientation to be a same fiber bundle with regard to the extracted center pixel group, and connects the center pixels indicating the same fiber bundle, and a meander determination processing unit which calculates a meandering amount of the connected center pixel group indicating the same fiber bundle.

First claim

1 . An image analyzing apparatus for analyzing an orientation of a fiber bundle from a three-dimensional image of a fiber-reinforced composite material, comprising:
an input unit which inputs the three-dimensional image; a binarization processing unit which binarizes the input three-dimensional image and acquires a binary image; an orientation estimation processing unit which estimates each orientation of foreground pixels in the binary image based on an orientation detection filter having a parameter for causing a shape of a detected cross section to have anisotropy; a center extraction processing unit which extracts center pixels indicating a center of the fiber bundle from a foreground pixel group, in which the orientation thereof was estimated, based on the orientation detection filter; a fiber bundle connection processing unit which deems center pixels having a same or similar orientation to be a same fiber bundle with regard to the extracted center pixel group, and connects the center pixels indicating the same fiber bundle; and a meander determination processing unit which calculates a meandering amount of the connected center pixel group indicating the same fiber bundle.

2 . The image analyzing apparatus according to claim 1 ,
wherein the orientation estimation processing unit changes attitude control angles of the orientation detection filter on the foreground pixels, searches for the attitude control angles in which a value of a convolution integral of a function indicated by the orientation detection filter and a function indicating the foreground pixels becomes maximum, and estimates a filter axial direction of the orientation detection filter in the attitude control angles in which the value of the convolution integral becomes maximum as the orientation of the foreground pixels.

3 . The image analyzing apparatus according to claim 1 ,
wherein the center extraction processing unit moves the orientation detection filter on the foreground pixel group in a direction that is vertical to the estimated orientation direction, searches for a position of the orientation detection filter in which a value of a convolution integral of a function indicated by the orientation detection filter and a function indicating the foreground pixel group becomes maximum, and extracts, as the center pixels, pixels on the foregoing pixel group corresponding to the position of the orientation detection filter in which the value of the convolution integral becomes maximum.

4 . The image analyzing apparatus according to claim 1 ,
wherein the fiber bundle connection processing unit groups successive center pixel groups that show a same or similar orientation direction with regard to the center pixel groups, extends a searching area to a direction that is the same as the orientation of the pixels of a starting point or an ending point of the grouped center pixel groups, and, when the extended searching areas overlap, deems that the pixels of the starting point or ending point belonging to different groups as pixels indicating the same fiber bundle, and connects the center pixels indicating the same fiber bundle by connecting the pixels of the starting point or ending point.

5 . The image analyzing apparatus according to claim 1 ,
wherein the meander determination processing unit calculates an appropriate straight line of the center pixel group indicating the same fiber bundle based on a least-squares method, applies the calculated approximate straight line to the center pixel group, and calculates, as the meandering amount, a maximum value among distances of a normal line lowered to the appropriate straight line from each of the center pixel groups.

6 . The image analyzing apparatus according to claim 1 , further comprising:
a higher precision processing unit which corrects each orientation of the center pixels based on the orientation detection filter and a gradient method after the center extraction processing unit extracts the center pixels from the foreground pixel group.

7 . The image analyzing apparatus according to claim 1 , further comprising:
a fiber bundle extraction image creation processing unit which creates a fiber bundle extraction image by adding the foregoing pixels indicating the fiber bundle after the fiber bundle connection processing unit connects the center pixels.

8 . The image analyzing apparatus according to claim 1 ,
wherein the three-dimensional image of the fiber-reinforced composite material is an image obtained by imaging a fabric produced via three-dimensional weaving or plain weaving with an X-ray CT device.

9 . A non-transitory computer-readable medium storing computer readable instructions thereon for analyzing an orientation of a fiber bundle from a three-dimensional image of a fiber-reinforced composite material, that when executed by a computer cause the computer to perform a method comprising:
inputting the three-dimensional image; binarizing the input three-dimensional image and acquiring a binary image; estimating each orientation of foreground pixels in the binary image based on an orientation detection filter having a parameter for causing a shape of a detected cross section to have anisotropy; extracting center pixels indicating a center of the fiber bundle from a foreground pixel group, in which the orientation thereof was estimated, based on the orientation detection filter; deeming center pixels having a same or similar orientation to be a same fiber bundle with regard to the extracted center pixel group, and connecting the center pixels indicating the same fiber bundle; and calculating a meandering amount of the connected center pixel group indicating the same fiber bundle.

10 . The non-transitory computer-readable medium according to claim 9 , further comprising:
correcting each orientation of the center pixels based on the orientation detection filter and a gradient method after the fourth step; and creating a fiber bundle extraction image by adding the foregoing pixels indicating the fiber bundle after the fifth step.

Assignees

• Ihi Corp
• Univ Tokyo

Inventors

• Hishida Hiroyuki
• Inagaki Koichi
• Nakamura Takeshi
• Yamauchi Yuta
• SUZUKI Hiroshima
• Michikawa Takashi
• Ohtake Yutaka

Classifications

• G01N23/046

  using tomography, e.g. computed tomography [CT] · CPC title

• G01N2223/419

  computed tomograph · CPC title

• G06T2207/30124

  Fabrics; Textile; Paper · CPC title

• G06T2207/10081

  Computed x-ray tomography [CT] · CPC title

• G06K9/4604

  Physics · mapped topic