CN101339652B - Solid engines CT image division method - Google Patents

Abstract

The invention discloses a splitting method of a CD image of a solid engine. The method comprises the following steps: a CD faultage image data of the solid engine is read in, an artifact which is formed outside the solid engine owning to the air agitation formed in the CD detection is eliminated; the image, the artifact of which is eliminated, filters and eliminates background noise by median filter, meanwhile, the edges of each component of the solid engine are sharpened in the CD image and a casing, a propelling agent and an astropyle of the solid engine is cut in the CD image; a defect is cut from the propelling agent. By adopting the method, an outer layer air ring artifact, the casing, the propelling agent, the astropyle and various defects of the solid engine in the CD image are cut automatically and accurately, thus solving the problem of low efficiency and poor accuracy in the CD detection image of the solid engine by the manual interpretation as well as being not limited by the eigenvalue of image information.

Description

A kind of dividing method of solid engines CT image

Technical field

The present invention relates to the image Segmentation Technology field, more specifically, the present invention relates to a kind of dividing method of solid engines CT image.

Background technology

The security performance of solid engines is the hot issue that people study always, and the existence of inner any one excessive defect of solid engines all may influence the overall performance of engine, even cause catastrophic accident.Since the eighties in 20th century, industry CT is as a kind of advanced person's nondestructiving detecting means, obtains to develop rapidly and be widely used.Utilization industry CT technology for detection solid engines can detect the problem that many conventional lossless detection methods are difficult to check out, has improved the security performance of solid engines greatly.And during solid engines CT detected at present, faultage image relied on manual each ingredient of interpretation of testing staff and inherent vice fully, need expend a large amount of time, constantly was engaged in repetitive operation, and maloperation appears in inefficiency easily.

Industry CT detects solid engines and generates gray-scale image, the zone that can be divided into solid engines CT image each tool characteristic by image Segmentation Technology, extract interested target area, thereby carry out defect recognition, so cutting apart of solid engines CT image is the basis of automatic identification of engine interior defective and 3-D view reconstruct.Existing image partition method comprises thousands of kinds of dividing methods of all kinds such as parallel border class, serial border class, parallel regional class and serial zone class, but in these methods, still needleless to all solids engines CT image automatically, dividing method accurately.For example, in the rim detection split plot design based on the border, various detection operators are the variation that utilizes the picture amplitude value in essence, carry out local operation, though calculated amount is little, calculate very sensitively to noise, and affected by noise when big when image, segmentation effect is relatively poor; The unimodal intensity profile image request that thresholding method was fit to: the gray-scale value between the inner neighbor of target and background is a height correlation, but the pixel on target and background intersection both sides has very big difference on gray-scale value.The space clustering split plot design is a kind of method of the overall situation, though than the more antinoise of rim detection split plot design, the cluster regular meeting of feature space causes producing the disconnected cut zone of image space.

Above-mentioned these methods all are to utilize that some part special characteristic comes cut zone in the image information, segmentation effect is relatively poor, and the whole bag of tricks has limitation and specific aim, does not have general applicability, can only design suitable dividing method at the demand of various specific areas.

Summary of the invention

Poor for the segmentation effect that overcomes solid engines CT image dividing method in the prior art, do not have a defective of universality, the present invention proposes a kind of dividing method of solid engines CT image.

A kind of dividing method of solid engines CT image comprises:

Step 10), read in the solid engines CT image data;

Step 20), according to the gray-scale value of described CT image, fill described image shell rim inside, remove the air ring artifacts of described CT image;

Step 30), the ground unrest in the described CT image of filtering, the edge of each ingredient of solid engines in the CT of the sharpening simultaneously image;

Step 40), according to many threshold methods, calculate the maximum between-cluster variance of housing, the propellant that contains defectiveness, star hole and background, determine the threshold value in housing, the propellant that contains defectiveness and the star hole of the described image of differentiation in the described CT gradation of image scope, cut apart the housing in the described image, the propellant that contains defectiveness and star hole;

Step 50), according to many threshold methods, determine the propellant of the described image of differentiation in the described CT gradation of image scope and the threshold value of defective, cut apart the defective in the propellant.

Wherein, step 20) in,, obtains the binaryzation gradient image of described CT image, determine the edge of described housing according to the difference of the gray-scale value of described CT image housing outward flange and air ring artifacts.

Wherein, step 20) in,, fills shell rim, obtain removing the binaryzation mask image of the described CT image of air ring artifacts based on the two-value mathematical morphology.

Wherein, step 20) further comprise, the respective coordinates data in the matrix of described binaryzation mask images and described CT image are multiplied each other respectively, obtain removing the solid engines CT image of air ring artifacts.

Wherein, step 30) in, use median filtering method, guaranteeing on the sharp-edged basis of described each ingredient of CT image filter out background noise.

Wherein, step 40) in, further comprise:

Step 410), calculate overall average gray level and the housing of setting, the propellant that contains defectiveness, star hole and the background 4 class targets inside average gray level separately of described CT image;

Step 420), the inside average gray level of calculating housing, the propellant that contains defectiveness, star hole and target context accounts for the ratio of the overall average gray level of image separately;

Step 430), calculate housing, contain the maximum between-cluster variance of propellant, star hole and the target context of defectiveness, determine that making the parameter of inter-class variance maximum is described threshold value.

Wherein, step 40) in, according to the threshold value of determining, generation comprises housing, contains the dicing masks image of the propellant of defectiveness, star hole and background, after will cutting apart housing, the propellant that contains defectiveness, the star hole coordinate data corresponding in the mask images matrix and distinguishing normalization, multiply each other respectively with coordinate data in the described CT image array with background.

Wherein, step 50) in, further comprise:

Step 510), calculate overall average gray level and the defective of setting and the inner separately average gray level of propellant target of described CT image;

Step 520), the inner separately average gray level of calculating defective and propellant accounts for the ratio of the overall average gray level of image;

Step 530), calculate the maximum between-cluster variance of defective and propellant, determine that making the parameter of inter-class variance maximum is described threshold value.

Wherein, the dicing masks image that will contain defectiveness and propellant uses unlatching and the closed combinatorial operation filtering defective in the mathematical morphology, and wherein, the structural element in the mathematical morphology is chosen for:

$\left[\begin{array}{ccc}0& 1& 0\\ 1& 1& 1\\ 0& 1& 0\end{array}\right].$

Wherein, use the interior respective coordinates data of the mask image cut apart defective and the matrix of described CT image to multiply each other respectively, eliminate the defective in the described image.

The present invention contains on the basis of the solid engines CT image architectural feature of defectiveness and intensity profile in analysis, integrated application rim detection, mathematical morphology and many thresholding methods, overcome the influence of the pseudo-shadow of air in the CT detection, realized the automatic accurate of solid engines CT image cut apart, provide more comprehensive and accurate analysis result for can't harm detection automatically and image reconstruction.

By using the present invention, can be partitioned into outer air ring artifacts, housing, propellant, star hole and various defective in the solid engines CT image automatically, exactly, solved the problem of the low and poor accuracy of manual interpretation solid engines CT detected image efficient, and the spy who is not subjected to image information

Description of drawings

Fig. 1 is a solid engines CT image dividing method process flow diagram according to an embodiment of the invention;

Fig. 2 is the solid engines CT image that contains defectiveness according to an embodiment of the invention;

Fig. 3 is a solid engines CT image of removing the air ring artifacts according to an embodiment of the invention;

Fig. 4 is cut apart housing according to an embodiment of the invention, is contained the propellant of defectiveness and the solid engines CT image in star hole;

Fig. 5 is a solid engines CT image of cutting apart defective according to an embodiment of the invention from propellant.

Embodiment

Below in conjunction with the drawings and specific embodiments, the dividing method of a kind of solid engines CT image provided by the invention is described further.

Fig. 1 illustrates solid engines CT image dividing method overall procedure according to an embodiment of the invention, at first reads in solid engines CT faultage image (below abbreviate the CT image as) data; According to rim detection and morphological method, the air turbulence when removing owing to the CT detection is at the air ring artifacts of solid engines outside formation then; Pseudo-movie queen's image process medium filtering filter out background noise, the edge of each ingredient of solid engines in the CT of the sharpening simultaneously image will be gone to; Based on many threshold methods, cut apart solid engines housing in the CT image, propellant and star hole; At last defective is cut apart from propellant.

In described embodiment, at first read in the solid engines CT faultage image data.

Fig. 2 is the solid engines CT image synoptic diagram that contains defectiveness, wherein, Fig. 2 A is the former figure of CT image that reads in, Fig. 2 B is the image of solid engines CT image behind histogram equalization, comprises air ring artifacts 1, housing 2, star hole 3, propellant 4 and defective 5 in the solid engines CT image of Fig. 2.

Shown in Fig. 2 B, the gray-scale value of the outward flange of housing 2 and air ring artifacts 1 differs greatly.Adopt traditional 3 * 3 Suo Beier (Sobel) WITH CROSS GRADIENTS operator, calculate the binaryzation gradient image of CT image shown in Fig. 2 A, can draw the complete edge of housing 2, be used to remove the air ring artifacts.Also can adopt other methods, determine the complete edge of housing 2 such as Variance operator, Roberts operator.

Fig. 3 is a synoptic diagram of removing the air ring artifacts, wherein, Fig. 3 A adopts traditional 3 * 3 Sobel WITH CROSS GRADIENTS operator that Fig. 2 A is carried out bianry image after the rim detection, and comprising frontier point w, and two ring-type curves shown in the figure are respectively the inner and outer boundary at edge; Fig. 3 B is the binaryzation mask images of the removal air ring artifacts after mathematical morphology is filled; Fig. 3 C is a solid engines CT image of removing the air ring artifacts.

In the two-value mathematical morphology, obtain the zone by filling according to known boundaries.To the w of bianry image frontier point shown in Fig. 3 A, by the image border inner and outer boundary around the set A of determining, by with structural element B pair set A expansion, supplement with ask friendship to come the fill area, as described below:

At first, will be that any point assignment in the A is 1 in Fig. 3 A frontier point w, establishing taken point is X ₀=1, fill according to following iterative formula then:

X _k＝(X _k-1_B)∩A ^c k＝1，2，3，…；

Wherein, structural element B is $\left[\begin{array}{ccc}0& 1& 0\\ 1& 1& 1\\ 0& 1& 0\end{array}\right],$ Structure centre is located A in " 1 " at matrix center ^cSupplementary set for A.

Work as X _k=X _K-1The time, stop iteration, at this moment X _kThe intra-zone that comprises frontier point w with the common factor of A and filled is just removed the binaryzation mask images of air ring artifacts, shown in Fig. 3 B.

Fig. 3 B mask images and the former figure of Fig. 2 A solid engines CT are multiplied each other, and the interior respective coordinates data of matrix that are about to Fig. 3 B mask images and the former figure of Fig. 2 A solid engines CT multiply each other respectively, obtain removing the solid engines CT image of air ring artifacts, shown in Fig. 3 C.

There are many noises in the solid engines CT image behind the removal air ring artifacts, in the present embodiment, adopt median filtering method, on the basis that guarantees each ingredient marginal sharpness of solid engines CT image, filtering noise, the edge of each ingredient of solid engines in the CT of the sharpening simultaneously image.

After the filtering noise, in the solid engines CT image of removing the air ring artifacts, housing, the propellant that contains defectiveness and star hole three's grey value profile be not or not an interval.In the present embodiment, adopt the Otsu multi-threshold image segmentation method, the number of targets that setting is cut apart is housing, contains propellant, star hole and background 4 classes of defectiveness that then required definite number of threshold values is 3, is respectively k ₁, k ₂, k ₃

Being calculated as follows of the overall average gray level of image:

$\mathrm{\μ}=\underset{i=0}{\overset{255}{\mathrm{\Σ}}}{\mathrm{iP}}_i;$

Wherein, μ is the overall average gray level of image, P _iBe the probability that i level pixel occurs, the inner average gray level of j+1 classification target is

${\mathrm{\μ}}_j\left(k\right)=\underset{i=k_j}{\overset{k_{j+1}-1}{\mathrm{\Σ}}}{\mathrm{iP}}_i$

Proportion is

${\mathrm{\ω}}_j=\underset{i=k_j}{\overset{k_{j+1}-1}{\mathrm{\Σ}}}{P}_i$

Wherein, ω _jBe the ratio that the inner average gray level of j+1 classification target accounts for the overall average gray level of image, j from 0 to 2, in addition threshold boundaries k ₀=0, k ₄=255.Order

μ _j＝μ _j(k)/ω _j

According to the principle of maximum variance between clusters, inter-class variance is:

${\mathrm{\σ}}^2(k_1,k_2,k_3)=\underset{i=0}{\overset{3}{\mathrm{\Σ}}}{\mathrm{\ω}}_i{(\mathrm{\μ}-{\mathrm{\μ}}_i)}^2$

Make k ₁, k ₂, k ₃Satisfy k ₁＜k ₂＜k ₃, and k ₁, k ₂, k ₃Change from 0 to 255, that is tried to achieve makes σ ²(k ₁, k ₂, k ₃) maximum k ₁, k ₂, k ₃Be the threshold value of asking.According to the threshold value k that tries to achieve ₁, k ₂, k ₃, determine that gray-scale value is less than k ₁Be the star hole, gray-scale value is between k ₁With k ₂Between for containing the propellant of defectiveness, gray-scale value is between k ₂And k ₃Between be housing, gray-scale value is greater than k ₃Be background, generate and cut apart mask images.

Fig. 4 is propellant and the star hole synoptic diagram of cutting apart housing, containing defectiveness; Wherein, A is that mask images is cut apart in housing, the propellant that contains defectiveness and the star hole of adopting the many thresholding methods of improved Otsu to draw, and B is the housing image that is partitioned into, and C is the image of the propellant that contains defectiveness that is partitioned into, and D is the star hole pattern picture that is partitioned into.

What generate as calculated cuts apart mask images shown in Fig. 4 A, with mask the former figure of solid engines CT shown in Fig. 2 A is made dividing processing, Fig. 4 A mask images and the former figure of Fig. 2 A solid engines CT are multiplied each other, after soon the coordinate data corresponding with background of housing, the propellant that contains defectiveness, the star hole in the matrix of Fig. 4 A mask images distinguished normalization, multiply each other respectively with the coordinate data in the matrix of the former figure of Fig. 2 A solid engines CT, the housing that obtains, the propellant that comprises defective and star hole pattern picture are respectively shown in Fig. 4 B, 4C, 4D.

Shown in Fig. 4 C, though defective and propellant belong between same gray area, it still differs bigger with the gray scale of neighborhood on every side.In like manner adopt aforesaid cutting procedure, promptly adopt the Otsu multi-threshold image segmentation method, set the target of cutting apart among Fig. 4 C and be made up of defective, propellant and background three, then required definite number of threshold values is 2, is respectively k ₅, k ₆, generate and cut apart mask images.

Fig. 5 is a synoptic diagram of cutting apart defective from propellant, wherein, A is that the defective, the propellant that adopt the many thresholding methods of improved Otsu to draw are cut apart mask images, and B is defective and propellant separate picture, C is the image after the B filtering defective, and D is the image after the defective that is partitioned into is amplified 4 * 4 times.

The defective of separating with propellant among Fig. 5 B can be with the unlatching in the mathematical morphology and closed combinatorial operation with its filtering:

Choose structural element $B=\left[\begin{array}{ccc}0& 1& 0\\ 1& 1& 1\\ 0& 1& 0\end{array}\right],$ Structure centre is located in " 1 " at matrix center.

Earlier with B Fig. 5 B is carried out open operation, carry out closed procedure with B to opening operating result again, whole process formula is

{[(A_B)_B]_B}_B＝(AоB)·B

Result of calculation is shown in Fig. 5 C, and defective is filtered out.

Again Fig. 5 B and Fig. 5 C are made XOR, the result who obtains is a mask images of cutting apart defective.Cut apart mask with defective the former figure of solid engines CT shown in Fig. 2 A is made dividing processing, the respective coordinates data that are about in the matrix of described mask images and the former figure of Fig. 2 A solid engines CT multiply each other respectively, after the defect image that is partitioned into that obtains amplifies 4 * 4 times shown in Fig. 5 D.

It should be noted that at last, above embodiment is only in order to illustrate that technical scheme of the present invention is not intended to limit, and on using, can extend to other modification, variation, application and embodiment, think that simultaneously all such modifications, variation, application, embodiment are within the spirit and scope of the present invention.

Claims (7)

1. the dividing method of a solid engines CT image comprises:

Step 10), read in the solid engines CT image data;

Step 20), according to the gray-scale value of described CT image, fill described image shell rim inside, remove the air ring artifacts of described CT image;

Step 30), the ground unrest in the described CT image of filtering, the edge of each ingredient of solid engines in the CT of the sharpening simultaneously image;

Step 40), according to many threshold methods, calculate the maximum between-cluster variance of housing, the propellant that contains defectiveness, star hole and background, determine the threshold value in housing, the propellant that contains defectiveness and the star hole of the described image of differentiation in the described CT gradation of image scope, cut apart the housing in the described image, the propellant that contains defectiveness and star hole;

Step 50), according to many threshold methods, determine the propellant of the described image of differentiation in the described CT gradation of image scope and the threshold value of defective, cut apart the defective in the propellant;

Wherein, step 40) in, further comprise:

Step 410), calculate overall average gray level and the housing of setting, the propellant that contains defectiveness, star hole and the background 4 class targets inside average gray level separately of described CT image;

Step 420), the inside average gray level of calculating housing, the propellant that contains defectiveness, star hole and target context accounts for the ratio of the overall average gray level of image separately;

Step 430), calculate housing, contain the maximum between-cluster variance of propellant, star hole and the target context of defectiveness, determine that making the parameter of inter-class variance maximum is described threshold value; According to the threshold value of determining, generation comprises housing, contains the dicing masks image of the propellant of defectiveness, star hole and background, after will cutting apart housing, the propellant that contains defectiveness, the star hole coordinate data corresponding in the mask images matrix and distinguishing normalization, multiply each other respectively with coordinate data in the described CT image array with background;

Wherein, step 50) in, further comprise:

Step 510), calculate overall average gray level and the defective of setting and the inner separately average gray level of propellant target of described CT image;

Step 520), the inner separately average gray level of calculating defective and propellant accounts for the ratio of the overall average gray level of image;

Step 530), calculate the maximum between-cluster variance of defective and propellant, determine that making the parameter of inter-class variance maximum is described threshold value.

2. the process of claim 1 wherein step 20) in, according to the difference of the gray-scale value of described CT image housing outward flange and air ring artifacts, obtain the binaryzation gradient image of described CT image, determine the edge of described housing.

3. the process of claim 1 wherein step 20) in, based on the two-value mathematical morphology, fill shell rim, obtain removing the binaryzation mask image of the described CT image of air ring artifacts.

4. the method for claim 3, wherein, step 20) further comprise, the respective coordinates data in the matrix of described binaryzation mask images and described CT image are multiplied each other respectively, obtain removing the solid engines CT image of air ring artifacts.

5. the process of claim 1 wherein step 30) in, use median filtering method, guaranteeing on the sharp-edged basis of described each ingredient of CT image filter out background noise.

6. the process of claim 1 wherein that the dicing masks image that will contain defectiveness and propellant uses unlatching and the closed combinatorial operation filtering defective in the mathematical morphology, wherein, the structural element in the mathematical morphology is chosen for:

$\left[\begin{array}{ccc}0& 1& 0\\ 1& 1& 1\\ 0& 1& 0\end{array}\right].$

7. the process of claim 1 wherein, use the interior respective coordinates data of the mask image cut apart defective and the matrix of described CT image to multiply each other respectively, eliminate the defective in the described image.

Images