CN110060253B - Composite sleeper pore defect identification method based on Gabor multi-feature extraction and optimization - Google Patents

Composite sleeper pore defect identification method based on Gabor multi-feature extraction and optimization Download PDF

Info

Publication number
CN110060253B
CN110060253B CN201910370740.7A CN201910370740A CN110060253B CN 110060253 B CN110060253 B CN 110060253B CN 201910370740 A CN201910370740 A CN 201910370740A CN 110060253 B CN110060253 B CN 110060253B
Authority
CN
China
Prior art keywords
image
defect
pore
gabor
characteristic
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Active
Application number
CN201910370740.7A
Other languages
Chinese (zh)
Other versions
CN110060253A (en
Inventor
张周锁
孟梨斌
赵海强
贾俊康
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Xian Jiaotong University
Original Assignee
Xian Jiaotong University
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Xian Jiaotong University filed Critical Xian Jiaotong University
Priority to CN201910370740.7A priority Critical patent/CN110060253B/en
Publication of CN110060253A publication Critical patent/CN110060253A/en
Application granted granted Critical
Publication of CN110060253B publication Critical patent/CN110060253B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
    • G01N23/00—Investigating or analysing materials by the use of wave or particle radiation, e.g. X-rays or neutrons, not covered by groups G01N3/00 – G01N17/00, G01N21/00 or G01N22/00
    • G01N23/02—Investigating or analysing materials by the use of wave or particle radiation, e.g. X-rays or neutrons, not covered by groups G01N3/00 – G01N17/00, G01N21/00 or G01N22/00 by transmitting the radiation through the material
    • G01N23/04—Investigating or analysing materials by the use of wave or particle radiation, e.g. X-rays or neutrons, not covered by groups G01N3/00 – G01N17/00, G01N21/00 or G01N22/00 by transmitting the radiation through the material and forming images of the material
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
    • G06F18/211—Selection of the most significant subset of features
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/23—Clustering techniques
    • G06F18/232—Non-hierarchical techniques
    • G06F18/2321—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions
    • G06F18/23213—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions with fixed number of clusters, e.g. K-means clustering
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T5/00—Image enhancement or restoration
    • G06T5/20—Image enhancement or restoration using local operators
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T5/00—Image enhancement or restoration
    • G06T5/70—Denoising; Smoothing
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T5/00—Image enhancement or restoration
    • G06T5/73—Deblurring; Sharpening
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/0002—Inspection of images, e.g. flaw detection
    • G06T7/0004—Industrial image inspection
    • G06T7/0008—Industrial image inspection checking presence/absence
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/10—Segmentation; Edge detection
    • G06T7/11—Region-based segmentation
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/40—Extraction of image or video features
    • G06V10/46—Descriptors for shape, contour or point-related descriptors, e.g. scale invariant feature transform [SIFT] or bags of words [BoW]; Salient regional features
    • G06V10/462—Salient features, e.g. scale invariant feature transforms [SIFT]
    • G06V10/464—Salient features, e.g. scale invariant feature transforms [SIFT] using a plurality of salient features, e.g. bag-of-words [BoW] representations
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/10—Image acquisition modality
    • G06T2207/10116—X-ray image
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/20—Special algorithmic details
    • G06T2207/20024—Filtering details
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/30—Subject of image; Context of image processing
    • G06T2207/30108—Industrial image inspection

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Data Mining & Analysis (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Evolutionary Computation (AREA)
  • General Engineering & Computer Science (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Evolutionary Biology (AREA)
  • Artificial Intelligence (AREA)
  • Health & Medical Sciences (AREA)
  • Probability & Statistics with Applications (AREA)
  • Multimedia (AREA)
  • Quality & Reliability (AREA)
  • Chemical & Material Sciences (AREA)
  • Analytical Chemistry (AREA)
  • Biochemistry (AREA)
  • General Health & Medical Sciences (AREA)
  • Immunology (AREA)
  • Pathology (AREA)
  • Analysing Materials By The Use Of Radiation (AREA)

Abstract

The invention discloses a composite sleeper pore defect identification method based on Gabor multi-feature extraction and optimization. Firstly, based on the attenuation characteristic of the composite sleeper to X-rays, carrying out X-ray transillumination detection on the composite sleeper by using an optimal detection system parameter to obtain an original projection image; then, preprocessing the original radiographic image, including row minimum noise reduction and fuzzy enhancement; then, performing multi-scale and multi-direction feature extraction on the radiographic image by using a designed Gabor filter group, and screening according to an energy criterion to obtain an image with prominent pore defect features; then, vectorizing the characteristic image, carrying out mean value removal and normalization, and finishing the clustering and segmentation of the pore defects by using a fuzzy C mean value method; and finally, removing the miscellaneous points and the false defects by using morphological open-close operation, calculating geometrical characteristic parameters of the defects, and realizing quantitative identification of the pore defects.

Description

Composite sleeper pore defect identification method based on Gabor multi-feature extraction and optimization
Technical Field
The invention relates to a composite material defect detection method, in particular to a composite sleeper pore defect identification method based on Gabor multi-feature extraction and optimization.
Background
The composite sleeper is a composite material structure made of rigid polyurethane resin foam reinforced by glass fiber yarn bundles and is manufactured by a pultrusion method. Compared with other sleepers, the material has various excellent characteristics of light weight, corrosion resistance, strong vibration reduction, good fatigue resistance and the like, and is applied to the field of rail traffic engineering such as subways, high-speed rails and the like. However, in the production process, the composite sleeper has the defect of pores with over-standard sizes, so that the service performance of the composite sleeper structure is seriously degraded, and economic loss and safety accidents are caused. Therefore, with the increase of the application and the demand of the composite sleeper, the nondestructive detection method which is high in research and detection efficiency and can be used for online real-time detection has important significance for detecting the internal pore defects of the composite sleeper, and is an urgent actual demand of the current engineering.
In the actual production of the existing composite sleeper, a destructive sampling inspection method of cutting a sample and observing a section is adopted for detection, so that the defects of high cost, low efficiency and the like exist. At present, the common nondestructive detection methods for structural defects such as vibration response analysis, ultrasonic C scanning and the like are difficult to meet the requirements of online and quantitative detection of the composite sleeper. The X-ray real-time imaging detection technology has the characteristics of visual results, real-time online performance and the like, and can meet the actual requirements of the engineering. However, due to the complexity of the composite sleeper material and the uncertainty of the void defects, how to accurately extract weak void defect characteristic information from the original sleeper X-ray image so as to realize the quantitative identification of the defects is a key problem in the detection of the composite void defects.
Disclosure of Invention
The invention aims to provide a composite sleeper pore defect identification method based on Gabor multi-feature extraction and optimization, which overcomes the defects of the prior art, realizes accurate identification of pore defects in X-ray images through preprocessing, Gabor multi-feature extraction, energy criterion optimization and clustering segmentation processes, has short time consumption, high precision and accurate and visual results, and can meet the detection requirements of the actual production process.
In order to achieve the purpose, the invention adopts the following technical scheme:
a composite sleeper pore defect identification method based on Gabor multi-feature extraction and optimization comprises the following steps:
(1) raw radiographic image acquisition. And the selection of the tube voltage and the tube current of the detection system is completed based on the attenuation characteristic of the composite sleeper to X-rays and according to a gray average gradient (GMG) index. And carrying out X-ray transillumination detection on the composite sleeper by using the optimal detection parameters to obtain an original projection image.
(2) And (5) image preprocessing. And respectively reducing noise and enhancing the original radiographic image of the composite sleeper pore defect, reducing Gaussian noise and stripe background noise of the original radiographic image, improving contrast and enhancing the appearance degree of the pore defect in the radiographic image.
(3) And extracting and optimizing defect features. And performing multi-scale and multi-direction filtering on the radiation image by using a designed Gabor filter group, and selecting a filtering image with prominent pore defect characteristics as a final characteristic image according to an energy principle.
(4) And clustering and dividing defect features. Firstly, converting a feature image with prominent pore defects into a feature vector, then carrying out zero mean and normalization on the feature vector to eliminate the difference of the sizes of all features, and finally clustering the feature vector by using a fuzzy C mean clustering algorithm to finish the segmentation of the pore defects.
(5) And (5) image post-processing. And (4) performing morphological opening and closing operation on the clustering result image to remove the pseudo pixel points to obtain a final binary image of the pore defect. And judging whether the size of the pore defect exceeds the standard or not by calculating the characteristic parameters of the pore defect, thereby realizing the identification of the pore defect.
Further, the step (2) is specifically as follows:
firstly, a sleeper original ray image is enhanced by using a classic S.K.Pal fuzzy enhancement algorithm, and the image gray level distribution range is improved. Then, aiming at the noise distribution characteristics of the composite sleeper structure, removing the stripe background noise by using a line minimum denoising algorithm, wherein the mathematical expression is as follows:
Figure BDA0002049848070000021
wherein the content of the first and second substances,
Figure BDA0002049848070000031
and f (x, y) represents the gray value of (x, y) points in the image after and before denoising, the image size is an M multiplied by N matrix, and t is each column of data in the corresponding row of the image.
Further, the step (3) is specifically:
firstly, constructing 5 scales (f) according to the characteristic that the pore defects in the composite sleeper radiographic image are weak1，f2，f3，f4，f5) And 8 directions (theta)1，θ2，θ3，θ4，θ5，θ6，θ7，θ8) The Gabor wavelet filter bank of (a) is used to extract features.
After the Gabor filter bank is obtained, each Gabor filter is convolved with the preprocessed image. Let the original image be I (x, y), and the Gabor filter of a particular scale p and direction q be defined as Gp,q(x, y), then the filtered image Ip,q(x, y) is:
Figure BDA0002049848070000032
wherein, p is a scale ordinal number, and p is 1,2, …, 5; q is an orientation number, q is 1,2, …, 8; gp,q(x,y)RAnd Gp,q(x,y)IThe real and imaginary parts of the Gabor filter at the corresponding scale and orientation, respectively.
Then, after obtaining a plurality of Gabor features of the radiographic image, optimizing the plurality of features by using an energy principle-based optimization method, which comprises the following specific processes:
1) calculating the total energy of the obtained filtering result image
Figure BDA0002049848070000033
And energy of each filtered image
Figure BDA0002049848070000034
2) Ordering the energy of each filtering image from large to small, defining
Figure BDA0002049848070000035
For the energy ratio coefficient, find the image sequence satisfying R2The first n filtered images of more than or equal to 0.95 are taken as the preferred characteristic image rk(x,y)，k＝1,2,…,n。
Finally, after obtaining the preferred characteristic image, smoothing and filtering the preferred characteristic image by using a Gaussian window to obtain a characteristic matrix T (x, y) formed by n characteristic images finally describing the composite sleeper pore defect:
T(x,y)＝[e1(x,y),e2(x,y),…,en(x,y)]
wherein e isk(x, y), k 1,2, …, n representing each preferred image rk(x, y) results of the processing.
Further, the step (4) is specifically as follows:
firstly, vectorizing a feature matrix T (X, y), and changing the feature matrix into a data set X containing M × N pixel point samples in the order from left to right and from top to bottom, wherein the data set X is equal to { X ═ X1,X2,…,Xi,…,XS(S) ═ mxn), where the ith sample Xi＝{xi1,xi2,…,xik,…,xinIs an n-dimensional vector, xikIs a corresponding pixel point sample XiThe kth Gabor feature of (1).
Then, for the feature vector Xi＝{xi1,xi1,…,xik,…,xinCarrying out zero mean value and normalization, eliminating difference of each characteristic quantity value, and obtaining a characteristic vector
Figure BDA0002049848070000041
And finally, clustering each pixel point in the data sample by utilizing fuzzy C-means clustering, and realizing final pore defect segmentation to obtain a binary image reflecting the pore defects.
Further, the step (5) is specifically as follows:
firstly, selecting a disc-shaped structural element to sequentially perform morphological opening operation and closing operation on a defect image obtained by segmentation by adopting a mathematical morphology method to obtain a final pore defect result.
Then, the longest vertical axis of the defect region in the image is taken as the diameter of the pore defect. Number of pixel points N according to longest axismaxAnd the obtained image resolution P is calculated to obtain the actual defect diameter D ═ NmaxAnd x P, realizing quantitative identification.
Compared with the prior art, the invention has the following beneficial technical effects:
1) the composite sleeper detection is carried out by adopting an X-ray real-time imaging technology, so that an original ray image containing defect information can be quickly obtained, and the requirement of online real-time detection is met; 2) by adopting a Gabor multi-feature extraction and optimization method, the superior local time-frequency analysis capability of Gabor wavelets is fully utilized, and weak, edge-fuzzy and noise-coupled defect features in the image can be accurately extracted; 3) the identification method combines the row minimum value noise reduction preprocessing, the clustering segmentation and the morphology post-processing, fully considers the defect distribution rule and the geometric characteristics, and further improves the identification precision and accuracy.
Drawings
FIG. 1 is a flow chart of the composite tie void defect identification of the present invention;
FIG. 2 is a composite sleeper test sample designed and manufactured; as shown in the figure, the size of the circular hole is 30.0 multiplied by 20.0mm, and circular hole defects with the diameters of 5mm, 4mm, 3mm, 2mm and 1mm are artificially processed along the forming direction;
FIG. 3 is a specific flow diagram of Gabor multi-feature extraction and optimization;
FIG. 4 is an original image and a pre-processed image obtained by X-ray detection; wherein (a) is an original image, and (b) is a preprocessed image;
FIG. 5 is a result image of the specimen radiographic image using a Gabor filter bank to extract features; the image in the frame is a defect feature salient image which is preferably selected according to an energy criterion;
FIG. 6 is a subsequent image fuzzy C-means clustering segmentation and image post-processing result diagram; wherein (a) is a binary image obtained by original clustering segmentation, and (b) is a defect identification result image after morphological post-processing.
Detailed Description
The invention is described in detail below with reference to the following figures and detailed description:
in order to make the aforementioned objects, features and advantages of the present invention comprehensible, the present invention accompanied with figures and detailed description thereof is described in detail as follows:
referring to fig. 1, which is a flowchart for identifying a composite sleeper pore defect, first, based on the attenuation characteristic of the composite sleeper to X-rays, the composite sleeper is subjected to X-ray transillumination detection by using the preferred detection system parameters, and an original radiographic image is obtained; then, preprocessing the original radiographic image, including row minimum noise reduction and fuzzy enhancement; then, performing multi-scale and multi-direction feature extraction on the radiographic image by using a designed Gabor filter group, and screening according to an energy criterion to obtain an image with prominent pore defect features; then, vectorizing the characteristic image, carrying out mean value removal and normalization, and finishing the clustering and segmentation of the pore defects by using a fuzzy C mean value method; and finally, removing the miscellaneous points and the false defects by using morphological open-close operation, calculating geometrical characteristic parameters of the defects, and realizing quantitative identification of the pore defects.
The method utilizes an X-ray real-time imaging technology to obtain sleeper image, and carries out pore defect identification by a Gabor multi-feature extraction and optimization method according to the following specific steps:
(1) raw radiographic image acquisition
And the selection of the tube voltage and the tube current of the detection system is completed based on the attenuation characteristic of the composite sleeper to X-rays and according to a gray average gradient (GMG) index. And carrying out X-ray transillumination detection on the composite sleeper by using the optimal detection parameters to obtain an original projection image.
(2) Image pre-processing
Firstly, a sleeper original ray image is enhanced by using a classic S.K.Pal fuzzy enhancement algorithm, and the image gray level distribution range is improved. Then, aiming at the noise distribution characteristics of the composite sleeper structure, removing the stripe background noise by using a line minimum denoising algorithm, wherein the mathematical expression is as follows:
Figure BDA0002049848070000061
wherein the content of the first and second substances,
Figure BDA0002049848070000062
and f (x, y) represents the gray value of (x, y) points in the image after and before denoising, the image size is an M multiplied by N matrix, and t is each column of data in the corresponding row of the image.
(3) Defect feature extraction and optimization
Firstly, constructing 5 scales (f) according to the characteristic that the pore defects in the composite sleeper radiographic image are weak1，f2，f3，f4，f5) And 8 directions (theta)1，θ2，θ3，θ4，θ5，θ6，θ7，θ8) The Gabor wavelet filter bank of (a) is used to extract features.
After the Gabor filter bank is obtained, each Gabor filter is convolved with the preprocessed image. Let the original image be I (x, y), and the Gabor filter of a particular scale p and direction q be defined as Gp,q(x, y), then the filtered image Ip,q(x, y) is:
Figure BDA0002049848070000063
wherein, p is a scale ordinal number, and p is 1,2, …, 5; q is an orientation number, q is 1,2, …, 8; gp,q(x,y)RAnd Gp,q(x,y)IThe real and imaginary parts of the Gabor filter at the corresponding scale and orientation, respectively.
Then, after obtaining a plurality of Gabor features of the radiographic image, optimizing the plurality of features by using an energy principle-based optimization method, which comprises the following specific processes:
1) calculating the total energy of the obtained filtering result image
Figure BDA0002049848070000071
And energy of each filtered image
Figure BDA0002049848070000072
2) Ordering the energy of each filtering image from large to small, defining
Figure BDA0002049848070000073
For the energy ratio coefficient, find the image sequence satisfying R2The first n filtered images of more than or equal to 0.95 are taken as the preferred characteristic image rk(x,y)，k＝1,2,…,n。
Finally, after obtaining the preferred characteristic image, smoothing and filtering the preferred characteristic image by using a Gaussian window to obtain a characteristic matrix T (x, y) formed by n characteristic images finally describing the composite sleeper pore defect:
T(x,y)＝[e1(x,y),e2(x,y),…,en(x,y)]
wherein e isk(x, y), k 1,2, …, n representing each preferred image rk(x, y) results of the processing.
(4) Defect feature clustering and segmentation
Firstly, vectorizing a feature matrix T (X, y), and changing the feature matrix into a data set X containing M × N pixel point samples in the order from left to right and from top to bottom, wherein the data set X is equal to { X ═ X1,X2,…,Xi,…,XS(S) ═ mxn), where the ith sample Xi＝{xi1,xi2,…,xik,…,xinIs an n-dimensional vector, xikIs a corresponding pixel point sample XiThe kth Gabor feature of (1).
Then, for the feature vector Xi＝{xi1,xi1,…,xik,…,xinCarrying out zero mean value and normalization, eliminating difference of each characteristic quantity value, and obtaining a characteristic vector
Figure BDA0002049848070000074
And finally, clustering each pixel point in the data sample by utilizing fuzzy C-means clustering, and realizing final pore defect segmentation to obtain a binary image reflecting the pore defects.
(5) Image post-processing
Firstly, selecting a disc-shaped structural element to sequentially perform morphological opening operation and closing operation on a defect image obtained by segmentation by adopting a mathematical morphology method to obtain a final pore defect result.
Then, the longest vertical axis of the defect region in the image is taken as the diameter of the pore defect. Number of pixel points N according to longest axismaxAnd the obtained image resolution P is calculated to obtain the actual defect diameter D ═ NmaxAnd x P, realizing quantitative identification.
A practical and effective way is provided for identifying the composite sleeper pore defects based on Gabor multi-feature extraction and an optimal composite sleeper pore defect identification method. The method can accurately and intuitively extract weak and strong coupling pore defect information characteristics from the original X-ray image, realizes quantitative identification, and has important engineering application value.
A specific application example process is given below, while verifying the effectiveness of the invention in engineering applications:
and designing and manufacturing a composite sleeper test sample to verify and explain the identification method. As shown in fig. 2, the test piece dimensions were 30.0 × 30.0 × 20.0(mm), and the cross-sectional dimensions were 30.0 × 20.0 (mm). Circular hole defects with diameters of 5mm, 4mm, 3mm, 2mm and 1mm are artificially processed on the cross section.
Firstly, the composite sleeper X-ray transillumination detection process is completed on a test sample piece according to the optimized detection system parameters, and an original projection image is obtained (as shown in (a) in fig. 4). It can be seen that the defect features in the original image are weak and coupled with the background fringe noise, larger defects are clearer, but smaller defects are difficult to directly distinguish.
With the method of the present invention, image preprocessing is first performed, including line minimum noise reduction and blur enhancement, to obtain a preprocessed image (as shown in fig. 4 (b)). It can be seen that the image contrast is significantly improved after the pretreatment, most of the background noise is suppressed, and the pore defects are highlighted.
Next, multi-feature extraction and optimization are performed on the image by using a designed Gabor filter bank (the specific flow is shown in fig. 3), and first, 40 different feature images in 8 directions of 5 scales are obtained by multi-scale extraction, and the result is shown in fig. 5. It can be seen that Gabor filters can effectively enhance the response of a void defect region when the filter dimensions and orientation parameters are consistent with the void defect region. Then, the energy principle is utilized to optimize a plurality of characteristics to obtain the condition that R is satisfied2A total of 8 Gabor feature images required of ≧ 0.95, as shown in the box of FIG. 5. Finally, local energy calculation is carried out on the optimized Gabor characteristic image by utilizing the weighted Gaussian window, abrupt noise in the image after Gabor conversion is reduced, and finally a Gabor characteristic matrix T reflecting pore defect information is obtainedA(x, y), the matrix size is 418 × 465 × 8.
Finally, the feature matrix is converted into a vector form
Figure BDA0002049848070000091
And the feature vectors are subjected to zero mean and normalization processing, and are subjected to clustering segmentation by using a fuzzy C mean method, and the result is shown in FIG. 6 (a). Pore defects of 5mm, 4mm and 3mm in diameter are clearly and visually apparent from the figure. Meanwhile, due to the inherent texture and clustering errors of the radiographic image, the defect edges in the segmentation result are not smooth, and some unexpected false targets (the part indicated in (a) in fig. 6) appear. Therefore, filtering the clustering result by using mathematical morphology, and obtaining the final recognition result is shown in fig. 6 (b). The sizes of the pore defects calculated according to the illustrated defect identification result are shown in the following table 1, and it can be seen that the minimum size of the identified pore defects reaches 3mm, and the identification precision is higher than 90%, so that the effectiveness of the composite sleeper pore defect identification is verified.
Table 1 test piece pore defect size calculation results
Figure BDA0002049848070000092

Claims (4)

1. A composite sleeper pore defect identification method based on Gabor multi-feature extraction and optimization is characterized by comprising the following steps:
(1) acquisition of original radiographic images: carrying out X-ray transillumination detection on the composite sleeper to obtain an original radiographic image;
(2) image preprocessing: respectively enhancing and denoising the original radiographic image of the composite sleeper pore defect;
(3) defect feature extraction and optimization: carrying out multi-scale and multi-direction filtering on the preprocessed radiation image by using a designed Gabor filter group, and selecting a filtering image with prominent pore defect characteristics as a final characteristic image according to an energy principle; the method specifically comprises the following steps:
firstly, constructing Gabor filter banks with 5 scales and 8 directions for extracting features according to the characteristic that pore defects in a composite sleeper radiographic image are weak, wherein the 5 scales are f1，f2，f3，f4And f5And 8 directions are respectively theta1，θ2，θ3，θ4，θ5，θ6，θ7And theta8；
After the Gabor filter bank is obtained, each Gabor filter is respectively convolved with the preprocessed image, the original image is set as I (x, y), and the Gabor filter with a specific scale p and a direction q is defined as Gp,q(x, y), then the filtered image Ip,q(x, y) is:
Figure FDA0002756961820000011
wherein, p is a scale ordinal number, and p is 1,2, …, 5; q is an orientation number, q is 1,2, …, 8; gp,q(x,y)RAnd Gp,q(x,y)IThe real part and the imaginary part of the Gabor filter under the corresponding scale and direction are respectively;
then, after obtaining a plurality of Gabor features of the radiographic image, optimizing the plurality of features by utilizing an energy principle-based optimization method;
finally, after obtaining the preferred characteristic image, smoothing and filtering the preferred characteristic image by using a Gaussian window to obtain a characteristic matrix T (x, y) formed by n characteristic images finally describing the composite sleeper pore defect:
T(x,y)＝[e1(x,y),e2(x,y),…,en(x,y)]
wherein e isk(x, y), k 1,2, …, n representing each preferred image rk(x, y) processing results;
the specific process of optimizing a plurality of characteristics by utilizing an optimal method based on an energy principle comprises the following steps:
1) calculating the total energy of the obtained filtering result image
Figure FDA0002756961820000021
And energy of each filtered image
Figure FDA0002756961820000022
2) Ordering the energy of each filtering image from large to small, defining
Figure FDA0002756961820000023
For the energy ratio coefficient, find the image sequence satisfying R2The first n filtered images of more than or equal to 0.95 are taken as the preferred characteristic image rk(x,y)，k＝1,2,…,n；
(4) Defect feature clustering and segmentation: firstly, converting a feature image with highlighted pore defects into a feature vector, then carrying out zero mean and normalization on the feature vector, and finally clustering the feature vector by using a fuzzy C mean clustering algorithm to finish the segmentation of the pore defects;
(5) image post-processing: removing pseudo pixel points by combining morphological open-close operation on the clustering result image to obtain a final binary image of the pore defects; and judging whether the size of the pore defect exceeds the standard or not by calculating the characteristic parameters of the pore defect, thereby realizing the identification of the pore defect.
2. The method for identifying the porosity defect of the composite sleeper based on Gabor multi-feature extraction and optimization as claimed in claim 1, wherein the step (2) is specifically as follows:
firstly, enhancing a sleeper original ray image by using an S.K.Pal fuzzy enhancement algorithm, and then removing stripe background noise by using a line minimum denoising algorithm, wherein the mathematical expression is as follows:
Figure FDA0002756961820000024
wherein the content of the first and second substances,
Figure FDA0002756961820000025
and f (x, y) represents the gray value of (x, y) points in the image after and before denoising, the image size is an M multiplied by N matrix, and t is each column of data in the corresponding row of the image.
3. The method for identifying the porosity defect of the composite sleeper based on Gabor multi-feature extraction and optimization as claimed in claim 1, wherein the step (4) is specifically as follows:
firstly, vectorizing a feature matrix T (X, y), and changing the feature matrix into a data set X containing M × N pixel point samples in the order from left to right and from top to bottom, wherein the data set X is equal to { X ═ X1,X2,…,Xi,…,XSWhere, M × N, the ith sample Xi＝{xi1,xi2,…,xik,…,xinIs an n-dimensional vector, xikIs a corresponding pixel point sample XiThe kth Gabor feature of (1);
then, for the feature vector Xi＝{xi1,xi1,…,xik,…,xinCarrying out zero mean value and normalization, eliminating difference of each characteristic quantity value, and obtaining a characteristic vector
Figure FDA0002756961820000031
And finally, clustering each pixel point in the data sample by utilizing fuzzy C-means clustering, and realizing final pore defect segmentation to obtain a binary image reflecting the pore defects.
4. The method for identifying the porosity defect of the composite sleeper based on Gabor multi-feature extraction and optimization as claimed in claim 1, wherein the step (5) is specifically as follows:
firstly, selecting a disc-shaped structural element to sequentially perform morphological opening operation and closing operation on a defect image obtained by segmentation by adopting a mathematical morphology method to obtain a final pore defect result;
then, the longest axis in the vertical direction of the defect area in the image is taken as the diameter of the pore defect, and the number N of pixel points of the longest axis is usedmaxAnd the obtained image resolution P is calculated to obtain the actual defect diameter D ═ NmaxAnd x P, realizing quantitative identification.
CN201910370740.7A 2019-05-06 2019-05-06 Composite sleeper pore defect identification method based on Gabor multi-feature extraction and optimization Active CN110060253B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201910370740.7A CN110060253B (en) 2019-05-06 2019-05-06 Composite sleeper pore defect identification method based on Gabor multi-feature extraction and optimization

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201910370740.7A CN110060253B (en) 2019-05-06 2019-05-06 Composite sleeper pore defect identification method based on Gabor multi-feature extraction and optimization

Publications (2)

Publication Number Publication Date
CN110060253A CN110060253A (en) 2019-07-26
CN110060253B true CN110060253B (en) 2021-04-20

Family

ID=67322204

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201910370740.7A Active CN110060253B (en) 2019-05-06 2019-05-06 Composite sleeper pore defect identification method based on Gabor multi-feature extraction and optimization

Country Status (1)

Country Link
CN (1) CN110060253B (en)

Families Citing this family (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110458850B (en) * 2019-08-01 2020-12-11 北京灵医灵科技有限公司 Segmentation method and segmentation system for large joint tissues

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US8809718B1 (en) * 2012-12-20 2014-08-19 Mss, Inc. Optical wire sorting
CN105719266A (en) * 2014-12-02 2016-06-29 马鞍山森格电子科技有限公司 Method for detecting surface defect of magnetic shoe based on texture feature clustering
CN108133473A (en) * 2017-12-21 2018-06-08 江南大学 Warp knitted jacquard fabric defect detection method based on Gabor filtering and deep neural network
CN108765402A (en) * 2018-05-30 2018-11-06 武汉理工大学 Non-woven fabrics defects detection and sorting technique
CN109671056A (en) * 2018-12-03 2019-04-23 西安交通大学 A kind of compound sleeper porosity defects detection method based on radioscopic image

Family Cites Families (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US6753965B2 (en) * 2001-01-09 2004-06-22 The University Of Hong Kong Defect detection system for quality assurance using automated visual inspection
CN103198322B (en) * 2013-01-18 2017-02-08 江南大学 Magnetic tile surface defect feature extraction and defect classification method based on machine vision
CN104680127A (en) * 2014-12-18 2015-06-03 闻泰通讯股份有限公司 Gesture identification method and gesture identification system
JP6537376B2 (en) * 2015-07-01 2019-07-03 キヤノン株式会社 Image processing apparatus and image processing method
CN108073932A (en) * 2016-11-16 2018-05-25 中国科学院沈阳计算技术研究所有限公司 A kind of workpiece image Feature extraction and recognition method based on Gabor filtering

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US8809718B1 (en) * 2012-12-20 2014-08-19 Mss, Inc. Optical wire sorting
CN105719266A (en) * 2014-12-02 2016-06-29 马鞍山森格电子科技有限公司 Method for detecting surface defect of magnetic shoe based on texture feature clustering
CN108133473A (en) * 2017-12-21 2018-06-08 江南大学 Warp knitted jacquard fabric defect detection method based on Gabor filtering and deep neural network
CN108765402A (en) * 2018-05-30 2018-11-06 武汉理工大学 Non-woven fabrics defects detection and sorting technique
CN109671056A (en) * 2018-12-03 2019-04-23 西安交通大学 A kind of compound sleeper porosity defects detection method based on radioscopic image

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
"Fast Surface Defect Detection Using Improved Gabor Filters";Jiaxu Ma 等;《2018 25th IEEE International Conference on Image Processing (ICIP)》;20180906;1508-1512 *
"基于Gabor变换的木材表面缺陷识别方法的研究";王林;《中国优秀硕士学位论文全文数据库-信息科技辑》;20110415;第2011年卷(第4期);摘要，第1.2、3.3、3.4、4.1、4.4节 *

Also Published As

Publication number Publication date
CN110060253A (en) 2019-07-26

Similar Documents

Publication Publication Date Title
CN109242888B (en) Infrared and visible light image fusion method combining image significance and non-subsampled contourlet transformation
CN110211101A (en) A kind of rail surface defect rapid detection system and method
CN111462076B (en) Full-slice digital pathological image fuzzy region detection method and system
CN1335496A (en) Image inspecting method and device for detecting faults
CN109671056B (en) Composite sleeper pore defect detection method based on X-ray image
CN109685766A (en) A kind of Fabric Defect detection method based on region fusion feature
CN108550145B (en) SAR image quality evaluation method and device
CN105445160A (en) Void characteristic of asphalt mixture and extraction method of void characteristic
CN114445386B (en) PVC pipe quality detection and evaluation method and system based on artificial intelligence
CN113536963B (en) SAR image airplane target detection method based on lightweight YOLO network
CN115311507B (en) Building board classification method based on data processing
Zuo et al. Fabric defect detection based on texture enhancement
CN110060253B (en) Composite sleeper pore defect identification method based on Gabor multi-feature extraction and optimization
CN113936132A (en) Method and system for detecting water pollution of chemical plant based on computer vision
CN115546127A (en) Engine lubricating oil abrasive particle image analysis method and device, medium and equipment
CN109934102B (en) Finger vein identification method based on image super-resolution
CN113610024B (en) Multi-strategy deep learning remote sensing image small target detection method
CN114821316A (en) Three-dimensional ground penetrating radar crack disease identification method and system
CN114842201A (en) Sandstone aggregate image segmentation method based on improved Mask _ Rcnn
CN103065296B (en) High-resolution remote sensing image residential area extraction method based on edge feature
CN114627283B (en) System and method for extracting interested brain region fiber bundles based on cluster denoising
CN116977266A (en) Canopy defect detection method and system based on small sample image feature extraction
Zhang et al. Research of glass defects detection based on DFT and optimal threshold method
CN112508862B (en) Method for enhancing magneto-optical image of crack by improving GAN
CN114565841A (en) Vehicle type recognition method based on image processing

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant