CN114549448A - Complex multi-type defect detection and evaluation method based on infrared thermal imaging data analysis - Google Patents

Complex multi-type defect detection and evaluation method based on infrared thermal imaging data analysis Download PDF

Info

Publication number
CN114549448A
CN114549448A CN202210148014.2A CN202210148014A CN114549448A CN 114549448 A CN114549448 A CN 114549448A CN 202210148014 A CN202210148014 A CN 202210148014A CN 114549448 A CN114549448 A CN 114549448A
Authority
CN
China
Prior art keywords
image
infrared
defect
damage
irri
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.)
Granted
Application number
CN202210148014.2A
Other languages
Chinese (zh)
Other versions
CN114549448B (en
Inventor
黄雪刚
雷光钰
殷春
罗健浩
于哲峰
罗庆
石安华
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.)
Ultra High Speed Aerodynamics Institute China Aerodynamics Research and Development Center
Original Assignee
Ultra High Speed Aerodynamics Institute China Aerodynamics Research and Development Center
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 Ultra High Speed Aerodynamics Institute China Aerodynamics Research and Development Center filed Critical Ultra High Speed Aerodynamics Institute China Aerodynamics Research and Development Center
Priority to CN202210148014.2A priority Critical patent/CN114549448B/en
Publication of CN114549448A publication Critical patent/CN114549448A/en
Application granted granted Critical
Publication of CN114549448B publication Critical patent/CN114549448B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • 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
    • 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
    • G06F18/217—Validation; Performance evaluation; Active pattern learning techniques
    • G06F18/2193—Validation; Performance evaluation; Active pattern learning techniques based on specific statistical tests
    • 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
    • G06T5/30—Erosion or dilatation, e.g. thinning
    • 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
    • G06T5/90—Dynamic range modification of images or parts thereof
    • G06T5/92—Dynamic range modification of images or parts thereof based on global image properties
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/10—Segmentation; Edge detection
    • G06T7/13—Edge detection
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/10—Segmentation; Edge detection
    • G06T7/155—Segmentation; Edge detection involving morphological operators
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/40—Analysis of texture
    • G06T7/41—Analysis of texture based on statistical description of texture
    • G06T7/44—Analysis of texture based on statistical description of texture using image operators, e.g. filters, edge density metrics or local histograms
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/60—Analysis of geometric attributes
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/60—Analysis of geometric attributes
    • G06T7/62—Analysis of geometric attributes of area, perimeter, diameter or volume
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/60—Analysis of geometric attributes
    • G06T7/64—Analysis of geometric attributes of convexity or concavity
    • 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
    • G06T2207/10048—Infrared 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
    • G06T2207/30—Subject of image; Context of image processing
    • G06T2207/30108—Industrial image inspection
    • 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
    • G06T2207/30168—Image quality inspection

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Geometry (AREA)
  • Data Mining & Analysis (AREA)
  • Probability & Statistics with Applications (AREA)
  • Quality & Reliability (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Artificial Intelligence (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Evolutionary Biology (AREA)
  • Evolutionary Computation (AREA)
  • General Engineering & Computer Science (AREA)
  • Image Processing (AREA)
  • Radiation Pyrometers (AREA)
  • Investigating Or Analyzing Materials Using Thermal Means (AREA)

Abstract

The invention discloses a complex multi-type defect detection and evaluation method based on infrared thermal imaging data analysis, which comprises the following steps: introducing thermal radiation contrast mass fraction and damage form mass fraction, firstly carrying out intensity scaling on infrared reconstruction, then extracting defect characteristics, and highlighting defect information; evaluating the shape and texture characteristics of the defect by using a matrix function; the distribution function of the matrix function is analogized with the power density of the image, and two defect characteristic mass fractions are respectively calculated; the extraction condition of defect features is objectively evaluated from two aspects of regional comparison and damage form; objectively evaluating the quality of the infrared reconstruction image from a quantization angle by utilizing the peak signal-to-noise ratio; detecting the infrared reconstruction images of each damage type respectively, and evaluating the reliability of the detection scheme on the detection condition of each type of damage; and detecting the infrared reconstruction image of the background area so as to judge the completeness of the extraction of different types of defects of the same test piece.

Description

Complex multi-type defect detection and evaluation method based on infrared thermal imaging data analysis
Technical Field
The invention belongs to the technical field of defect detection, and particularly relates to a complex multi-type defect detection and evaluation method based on infrared thermal imaging data analysis.
Background
The spacecraft may be hit by the tiny meteoroid bodies/space debris at ultra high speed when in service in the space, and various, complex types (such as coupling and the like) and wide distribution areas of damages are formed on the surface/subsurface of the spacecraft. Therefore, the research on the defect detection technology of the complex types of the spacecraft materials is very necessary for the operation safety of the spacecraft. At present, a plurality of defect detection schemes based on infrared thermal imaging technology are used for spacecraft defect detection. For example, a thermal signal reconstruction theory, an independent component analysis method and the like are all used for obtaining a plurality of infrared reconstruction images containing different damage types by extracting and mining the characteristic information of an infrared sequence and reconstructing the infrared reconstruction images back to a defect characteristic space, and each infrared reconstruction image visualizes one type of damage, so that the purpose of identifying complex and multi-defects is achieved. Due to the fact that the ultra-high-speed impact test piece has multiple defect types and complex defect conditions, the detection result of the detection scheme needs to be subjected to targeted judgment.
Therefore, objective evaluation criteria are required to be set to evaluate the targeted detection performance of the detection scheme. The overall integrity and the high quality of characteristic characterization of all damage types extracted from the infrared reconstructed image are very important. In addition, the infrared image is susceptible to noise, and the quality of the infrared reconstructed image needs to be guaranteed. Meanwhile, the overall judgment index of one infrared detection scheme is considered together with the whole detection process of the infrared detection scheme. And aiming at characteristic quality judgment, introducing a thermal radiation comparison mass fraction and a damage form mass fraction. And (4) firstly carrying out intensity scaling on the infrared reconstruction and then extracting defect characteristics to highlight defect information. And evaluating the characteristics of the defect such as shape, texture and the like by using a matrix function. And (5) carrying out analogy on the distribution function of the matrix function and the power density of the image, and respectively calculating two defect characteristic quality scores. And the extraction condition of the defect features is objectively evaluated from two aspects of regional comparison and damage form. In addition, the quality of the infrared reconstructed image is objectively evaluated from a quantitative point of view by using the peak signal-to-noise ratio.
The detection scheme is directed to extracting defect type integrity. And respectively detecting the infrared reconstruction images of each damage type, and evaluating the reliability of the detection scheme on the detection condition of each type of damage. And detecting the infrared reconstruction image of the background area so as to judge the completeness of the extraction of different types of defects of the same test piece.
Based on the method, a comprehensive objective evaluation mode is designed to evaluate the damage detection condition of the infrared defect detection scheme. And comprehensively considering various damage characteristics and infrared reconstructed images extracted by the infrared detection scheme, setting quantitative objective indexes, and evaluating the detection capability of the infrared detection scheme on various ultra-high-speed impact damages through the infrared reconstructed images.
Disclosure of Invention
An object of the present invention is to solve at least the above problems and/or disadvantages and to provide at least the advantages described hereinafter.
To achieve these objects and other advantages in accordance with the purpose of the invention, there is provided a complex multi-type defect detection evaluation method based on infrared thermal imaging data analysis, comprising: aiming at characteristic quality judgment, introducing thermal radiation contrast mass fraction and damage form mass fraction, firstly carrying out intensity scaling on infrared reconstruction, then extracting defect characteristics, and highlighting defect information; evaluating the shape and texture characteristics of the defect by using a matrix function; carrying out analogy on the distribution function of the matrix function and the power density of the image, and respectively calculating two defect characteristic mass fractions; the extraction condition of defect features is objectively evaluated from two aspects of regional comparison and damage form; in addition, the quality of the infrared reconstructed image is objectively evaluated from the quantization angle by utilizing the peak signal-to-noise ratio;
aiming at the detection scheme, the integrity of defect types is extracted, the infrared reconstruction images of all damage types are respectively detected, and the credibility of the detection scheme on the detection conditions of all types of damage is evaluated; and detecting the infrared reconstruction image of the background area so as to judge the extraction completeness of different types of defects of the same test piece.
Preferably, the method for detecting and evaluating the complex multi-type defects based on the infrared thermal imaging data analysis further comprises the following steps:
s1, acquiring an infrared sequence by using a detection scheme T to be evaluated, and performing feature extraction to obtain a background area infrared reconstruction image and Q infrared reconstruction images which respectively at least comprise perforations, bulges, impact pits and peeled Q ultra-high-speed impact damages;
step S2, respectively preprocessing the infrared reconstruction images for representing the k-th damage or background area obtained in the step S1 to obtain defect characteristic images;
step S3, objectively evaluating the defect feature extraction condition of the k-th auxiliary defect feature image obtained in the step S2 from the aspects of damage form and region comparison;
step S4, objectively evaluating the self image quality of the k-th sub-defect feature image obtained in the step S2 by utilizing the peak signal-to-noise ratio;
step S5, calculating total mass fractions of the infrared reconstructed images of the background area and each type of damage by integrating the damage characteristic evaluation result obtained in the step S3 and the image characteristic evaluation result obtained in the step S4;
s6, evaluating the credibility of the infrared reconstruction image based on the total mass fraction of the infrared reconstruction image obtained in the S5;
s7, calculating the objective quality score of the infrared detection scheme by using the credible infrared reconstruction images of the various types of damages obtained in the step S6;
s8, evaluating the identification performance of all ultra-high speed impact damage types based on the infrared detection scheme objective quality scores obtained in the S7;
step S9, based on the evaluation results of the step S8 on the identification performance of all the ultra-high speed impact types, under the condition that the identification performance of the current detection scheme on all the ultra-high speed impact damage is good, quantitative calculation is respectively carried out on the damage characteristics of all types; including quantitatively calculating the width, area, curvature characteristics of each type of lesion.
Preferably, the infrared reconstructed image of the background area obtained in the step S1 is regarded as IRRIo(0)Recording the infrared image of the ultra-high speed impact damage as IRRIo(q)，q＝1,…,Q；
In step S2, the infrared reconstructed image IRRI obtained in step S1 and used for characterizing the k-th lesion (i.e., k ═ 1,2, …, Q) or the background region (i.e., k ═ 0) is processedo(k)Respectively carrying out pretreatment, mainly comprising the following steps: firstly, respectively reconstructing the k infrared reconstruction image IRRI by utilizing an inverse tangent atan normalization methodo(k)Normalization, aiming at improving the stability of the defect characteristics of the infrared reconstructed image visualization, and recording the obtained normalized image as
Figure BDA0003509227520000031
k is 0,1, …, Q; then, expansion corrosion morphology is carried out on the normalized imageThe image characteristics are enhanced, and noise interference is reduced; performing convolution filtering on the image after morphological processing by using a Gabor convolution filter to extract the edge and texture characteristics of the defect; finally obtaining a defect characteristic image IRRIdf(k)And k is 0,1, … and Q, and the specific steps are as follows:
step S21, reconstructing the infrared image IRRI based on the arctangent atan normalization methodo(k)K is 0,1, …, Q normalization, and stability of defect characteristics of infrared reconstructed image visualization is improved; IRRI (infrared image reconstruction) image by arc tangent atan normalization methodo(k)K is 0,1, …, and the data point R in Q is located at ith row and jth columnijCorresponding radiation value xijZooming to obtain new radiation value xi(new)j(new)：
Figure BDA0003509227520000041
Wherein, atan (x)ij) Is xijThe arctan function of; new radiation value x obtained after radiation value normalization of all infrared data pointsi(new)j(new)Forming a normalized image
Figure BDA0003509227520000042
k＝0,1,…,Q；
Step S22, normalizing the image
Figure BDA0003509227520000043
Q is subjected to a morphological treatment of dilation followed by erosion, resulting in a morphologically processed image
Figure BDA0003509227520000044
k＝0,1,…,Q；
First pair normalized image
Figure BDA0003509227520000045
k is 0,1, …, Q is subjected to expansion processing to obtain an expansion image
Figure BDA0003509227520000046
Figure BDA0003509227520000047
Wherein C is a neighborhood window of r × r, r is an odd number, the center pixel of the neighborhood window is (r +1)/2,
Figure BDA0003509227520000048
as an image
Figure BDA0003509227520000049
The expansion treatment process comprises the following steps: in the process of moving the neighborhood window C, if the neighborhood window C is matched with the image
Figure BDA00035092275200000410
If there is an overlapping area, recording the position, all moving neighborhood windows C and the image
Figure BDA00035092275200000411
The set of positions where intersections exist is an image
Figure BDA00035092275200000412
Dilated image under neighborhood window C
Figure BDA00035092275200000413
Then to the expansion image
Figure BDA00035092275200000414
k is 0,1, …, and Q is subjected to etching treatment:
Figure BDA00035092275200000415
theta is an image
Figure BDA00035092275200000416
And (3) corrosion treatment process: moving the neighborhood window C, if the neighborhood window C and the image
Figure BDA00035092275200000417
The intersection of (A) completely belongs to the image
Figure BDA00035092275200000418
The location point is saved and all points satisfying the condition constitute an image
Figure BDA00035092275200000419
Corroded image obtained by corrosion of neighborhood window C
Figure BDA00035092275200000420
Etching image obtained by ending etching operation
Figure BDA00035092275200000421
I.e. morphologically processed images obtained by morphological processing of dilation followed by erosion
Figure BDA00035092275200000422
k＝0,1,…,Q；
Step S23, morphological processing image
Figure BDA00035092275200000423
k is 0,1, …, Q is processed by convolution filtering to extract the edge, texture and other characteristics of the defect, and a defect characteristic image IRRI is obtaineddf(k)，k＝0,1,…,Q：
Figure BDA00035092275200000424
Where F is a Gabor filter kernel of size n × n, representing the pair image
Figure BDA00035092275200000425
Performing Gabor convolution operation: image IRRIdf(k)Dividing the image into image blocks, selecting a scales and b directions to form ab Gabor filter banks; the Gabor filter bank is convolved with each image block in a space domain, each image block can obtain ab filter outputs, and the average value of the filter outputs of the image blocks is extracted to form a sumThe column vector of 1 is taken as the defect characteristic of the image block; the defect features of all the image blocks form a defect feature image IRRIdf(k)The definition of each position (x, y) of the Gabor filter kernel is as follows:
Figure BDA0003509227520000051
wherein ,
x'＝x·cosθ+y·sinθ
y'＝-x·sinθ+y·cosθ，
wherein λ is a sine function wavelength; θ is the direction of the Gabor kernel function; ψ is a phase offset; ε is the bandwidth; γ is the aspect ratio of the space, which determines the shape of the Gabor kernel.
Preferably, the step S3 is to compare the k-th defect feature image IRRI obtained in the step S2 from two aspects of damage morphology and region comparisondf(k)The defect feature extraction condition of the Q is objectively evaluated, wherein k is 0,1, …; because the radiation value of each data point in the infrared reconstructed image can reflect damage information, thermal radiation contrast mass fraction and damage form mass fraction are introduced on the basis, and characteristics such as shape, texture and the like of a defect are evaluated by utilizing a matrix function; the distribution function of the matrix function is compared with the power density of the image, and two defect characteristic mass fractions are respectively calculated by utilizing the central limit theorem, and the method specifically comprises the following steps:
step S31, calculating a normalized image
Figure BDA0003509227520000052
k 0,1, …, first order (mean) moment function of the radiation values of the Q data points
Figure BDA0003509227520000053
k 0,1, …, Q and a second order (variance) moment function
Figure BDA0003509227520000054
k＝0,1,…,Q：
Figure BDA0003509227520000055
Figure BDA0003509227520000056
Wherein Z is a normalized image
Figure BDA0003509227520000057
k is 0,1, …, number of mid-infrared data points in Q, xvTo normalize the image
Figure BDA0003509227520000058
k is 0,1, …, the radiation value of the infrared data point in Q;
step S32, calculating a defect characteristic image IRRIdf(k)K is 0,1, …, a first moment function μ of Qdf(k)K is 0,1, …, Q and a second moment function
Figure BDA0003509227520000059
k＝0,1,…,Q：
Figure BDA00035092275200000510
Figure BDA00035092275200000511
Wherein k is 0,1, …, Q, P is defect characteristic image IRRIdf(k)Number of mid-infrared data points, xnFor defective characteristic images IRRIdf(k)K is 0,1, …, Q infrared data point radiation value;
step S33, defect characteristic image IRRIdf(k)K is 0,1, …, Q demarcates the regions by radiance: dividing infrared data points with the same radiation value into one class, representing one characteristic of the defect, and recording the characteristic as IRTdf_l(k)，k＝0,1,…,Q；
Step S34, estimating characteristic IRTdf_l(k)K is 0,1, …, a function of the moment of the Q radiation value μdf_(l)kK is 0,1, …, distribution obeyed by Q: mean value μ according to the central limit theoremdf_l(k)Obey mean value of
Figure BDA0003509227520000061
Normal distribution of (a):
Figure BDA0003509227520000062
where k is 0,1, …, Q, p (μ)df_l(k)) Is mudf_l(k)Is determined by the probability density function of (a),
Figure BDA0003509227520000063
for the normalized image obtained in step S21
Figure BDA0003509227520000064
The mean value of the radiation values of the infrared data points,
Figure BDA0003509227520000065
is mudf_l(k)The variance of the obeyed normal distribution;
step S35, estimating characteristic IRTdf_l(k)K is 0,1, …, a function of the moment of the Q radiation value
Figure BDA0003509227520000066
k is 0,1, …, Q obeys a distribution: using the same estimate, variance, in step S34
Figure BDA0003509227520000067
Distribution of (2) is close to the mean
Figure BDA0003509227520000068
Normal distribution of (a):
Figure BDA0003509227520000069
wherein k is 0,1, …, Q,
Figure BDA00035092275200000610
is that
Figure BDA00035092275200000611
Is determined by the probability density function of (a),
Figure BDA00035092275200000612
for the normalized image obtained in step S21
Figure BDA00035092275200000613
The variance of the radiation values of the infrared data points,
Figure BDA00035092275200000614
is composed of
Figure BDA00035092275200000615
The variance of the obeyed normal distribution;
step S36, calculating the three sigma principle respectively
Figure BDA00035092275200000616
k is 0,1, …, Q and
Figure BDA00035092275200000617
k＝0,1,…,Q：
Figure BDA00035092275200000618
Figure BDA00035092275200000619
wherein k is 0,1, …, Q,
Figure BDA00035092275200000620
is mudf_l(k)Distributed in intervals
Figure BDA00035092275200000621
The probability of (a) of (b) being,
Figure BDA00035092275200000622
for the normalized image obtained in step S21
Figure BDA00035092275200000623
The mean value of the infrared data point radiation values;
Figure BDA00035092275200000624
is composed of
Figure BDA00035092275200000625
Distributed in intervals
Figure BDA00035092275200000626
The probability of (a) of (b) being,
Figure BDA00035092275200000627
for the normalized image obtained in step S21
Figure BDA00035092275200000628
The variance of the radiation values of the pixels;
Figure BDA0003509227520000071
a distribution function that is a standard normal distribution;
step S37, calculating the value of mudf_l(k)K is 0,1, …, normal distribution function of Q:
μdf_l(k)k is 0,1, …, and Q follows a normal distribution of
Figure BDA0003509227520000072
k is 0,1, …, Q, random variable RVμ(k)K is 0,1, …, Q for μdf_(l)kNormal distribution function of Q, k being 0,1, …
Figure BDA0003509227520000073
k is 0,1, …, Q is:
Figure BDA0003509227520000074
step S38, calculate about
Figure BDA0003509227520000075
k is 0,1, …, normal distribution function of Q:
Figure BDA0003509227520000076
k is 0,1, …, and Q follows a normal distribution of
Figure BDA0003509227520000077
k is 0,1, …, Q, random variable RVσ(k)K is 0,1, …, Q relates to
Figure BDA0003509227520000078
Normal distribution function of Q with k equal to 0,1, …
Figure BDA0003509227520000079
k is 0,1, …, Q is:
Figure BDA00035092275200000710
step S39, calculating thermal radiation contrast mass fraction TRC representing the background area and the infrared reconstruction image defect characteristics of various types of damage(k)，k＝0,1,…,Q：
Figure BDA00035092275200000711
Wherein k is 0,1, …, Q,
Figure BDA00035092275200000712
is a random variable RVσ(k)About
Figure BDA00035092275200000713
A normal distribution function of; random variable
Figure BDA00035092275200000714
Is by standardization
Figure BDA00035092275200000715
Mean value of obeys of
Figure BDA00035092275200000716
Variance of
Figure BDA00035092275200000719
Obtained by normal distribution;
step S310, calculating damage form quality scores DM of infrared reconstruction image defect characteristics of representing background areas and various types of damages(k)，k＝0,1,…,Q：
Figure BDA00035092275200000717
Wherein k is 0,1, …, Q,
Figure BDA00035092275200000718
is a random variable RVμ(k)About mudf_l(k)The normal distribution function of (2); random variable
Figure BDA0003509227520000081
Is obtained by normalizing mudf_l(k)Mean value of obeys of
Figure BDA0003509227520000082
Variance of
Figure BDA0003509227520000083
Obtained by normal distribution;
in step S311, if k is equal to 0, the background infrared reconstructed image IRRI is calculatedo(0)The mass fraction of (A): if any TRC is present(0)+DM(0)And if yes, determining that the background image has unseparated damage, returning to the step S1 to detect the damage features which are not identified, and extracting the defect features, otherwise, executing the step S4.
Preferably, wherein said stepS4 comparing the k-th defect feature image IRRI obtained in step S2 by using peak signal-to-noise ratiodf(k)And (5) objectively evaluating the self image quality of the image with k being 0,1, … and Q, and calculating the IRRI of each type of damage characteristic imagedf(k)K is 0,1, …, peak signal-to-noise ratio P of Q(k)：
Figure BDA0003509227520000084
Wherein k is 0,1, …, Q,
Figure BDA0003509227520000085
indicating defect characteristics IRTdf_l(k)And defect feature image IRRIdf(k)M × N denotes the defect feature image IRRIdf(k)Size of (IRT)df_l(k)(i, j) denotes the Defect feature IRTdf_l(k)Radiation value at position (i, j), IRRIdf(i, j) Defect feature image IRRIdf(k)The radiation value at position (i, j);
Figure BDA0003509227520000086
indicating defect characteristics IRTdf_l(k)Maximum value of the radiation value of (1).
Preferably, in the step S5, the total quality scores TQS of the background region and the infrared reconstructed image of each type of damage are calculated by integrating the damage characteristic evaluation result obtained in the step S3 and the image characteristic evaluation result obtained in the step S4 respectively(k)，k＝0,1,…,Q：
TQS(k)＝λ1·TRC(k)+λ2·DM(k)+λ3·P(k),k＝0,1,...,Q
wherein ,
Figure BDA0003509227520000087
are weight coefficients.
Preferably, in the step S6, the total quality score TQS of the infrared reconstructed image obtained in the step S5 is used as the basis(k)K 0, 1.. Q evaluates the reliability of the infrared reconstructed imageThe estimation comprises the following specific steps: if there is TQS(k)If η, k is 0,1, …, Q, where η is the damage assessment condition threshold, the current infrared reconstructed image is deemed to be unreliable for the k-th damage, and the process returns to step S1 to re-extract the kth infrared reconstructed image IRRI from the infrared sequenceo(k)Evaluating the extraction result of the kth damage; if the extracted result of the k-th damage is reevaluated, TQS still remains(k)If the measured value is less than eta, the method returns to the step S1 to detect the test piece to be detected again, acquires the infrared sequence again, and extracts the infrared reconstruction image IRRI of the visual k-th damageo(k)And then, evaluation was performed.
Preferably, in the step S7, for the trusted infrared reconstructed image of each type of damage obtained in the step S6, the trusted infrared reconstructed image is used to calculate the objective quality score of the infrared detection scheme, and the specific steps are as follows: q total mass fractions TQS of infrared reconstruction images respectively containing Q collision damages(k)After the calculation is finished, calculating the objective quality score Id of the infrared detection schemeoq：
Figure BDA0003509227520000091
By objective quality score IdoqTo judge the defect extraction condition, Id, of the infrared detection schemeoqThe larger the value of the defect identification result is, the better the detection identification result of the current infrared detection scheme on the complex multi-type defects formed by the ultrahigh-speed impact on the spacecraft is.
Preferably, in the step S8, the objective quality score Id is obtained based on the infrared detection scheme obtained in the step S7oqEvaluating the identification performance of all the ultra-high speed impact damage types; the method comprises the following specific steps: if any, IdoqIf the speed is higher than epsilon, wherein epsilon is a damage identification evaluation threshold value, the current detection scheme is considered to be capable of better identifying all types of ultra-high speed impact damage, and the accuracy of quantitative calculation of the damage is higher based on the infrared reconstructed image which is extracted from the current detection scheme and represents each type of damage; step S9 is transferred to carry out quantitative calculation on the ultra-high speed impact damage of each type; whether or notThen, the flow returns to step S1 to obtain the total quality score TQS(k)N infrared reconstructed images of < epsilon
Figure BDA0003509227520000092
Re-evaluating the infrared reconstructed image when
Figure BDA0003509227520000093
When the fraction is the fraction, rounding up; reextracting individual characterizations from infrared sequences
Figure BDA0003509227520000094
Damage by ultra-high speed impact
Figure BDA0003509227520000095
The infrared reconstructed image is displayed, the characterization effect of the infrared reconstructed image on the damage is improved, and the infrared reconstructed image is subjected to characterization again
Figure BDA0003509227520000096
Seed damage was assessed.
Through the steps, the objective quality score Id of the detection scheme T is obtainedoqThus, objective evaluation of the detection scheme T is realized. The detection scheme is objectively and effectively evaluated, so that complete complex multi-type ultrahigh-speed impact damage information can be extracted and mined in the early stage of defect quantitative identification, and a high-quality infrared reconstructed image with clear damage characteristics is obtained. The accuracy of further defect quantitative analysis is improved, and the precision of the whole ultrahigh-speed impact damage identification process is improved.
The invention at least comprises the following beneficial effects:
(1) the quality evaluation method provided by the invention can be used for evaluating the quality of the whole infrared detection scheme by combining the quality of the infrared image and the detection flow of the detection scheme, and judging the quality of the detection scheme according to the quality fraction so as to improve the applicability and detection precision of the detection scheme to complex multi-type damage formed by ultra-high speed impact and comprehensively evaluate the detection scheme.
(2) The quality evaluation method provided by the invention carries out normalization processing on the infrared reconstruction image. The image normalization can improve the contrast between the defect region and the background region and enhance the stability of the image on the premise of not changing the damage information stored in the infrared reconstruction image.
(3) The quality evaluation method provided by the invention emphasizes the defect characteristics by using morphological processing and filter filtering. For the problem that the infrared reconstruction image is not clearly divided, the defect features are highlighted, so that the complex defect features are more stable, and the defect extraction condition can be further objectively evaluated conveniently.
Additional advantages, objects, and features of the invention will be set forth in part in the description which follows and in part will become apparent to those having ordinary skill in the art upon examination of the following or may be learned from practice of the invention.
Drawings
FIG. 1 is a flow chart of a complex multi-type defect detection and evaluation method based on infrared thermal imaging data analysis according to the present invention;
FIG. 2 is a flow chart of objective assessment of defect feature extraction from the aspect of damage features according to the present invention;
FIG. 3 is an infrared reconstructed image of the interior of the impact pit containing various damage types evaluated in this example;
fig. 4 is an infrared reconstructed image in which a background region to be evaluated in this embodiment includes a plurality of damage types;
FIG. 5 is an infrared reconstructed image of impact pit edges containing various damage types evaluated in this example;
FIG. 6 is a normalized image obtained by preprocessing an infrared reconstructed image of an impact center;
FIG. 7 is a defect feature image obtained by impact center infrared reconstruction image preprocessing;
FIG. 8 is a normalized image obtained by preprocessing an impact edge infrared reconstructed image;
FIG. 9 is a defect feature image obtained by impact edge infrared reconstructed image preprocessing.
Detailed Description
The present invention is further described in detail below with reference to the attached drawings so that those skilled in the art can implement the invention by referring to the description text.
It will be understood that terms such as "having," "including," and "comprising," as used herein, do not preclude the presence or addition of one or more other elements or groups thereof.
As shown in fig. 1, a flowchart of a complex multi-type defect detection and evaluation method based on infrared thermal imaging data analysis includes the following steps: aiming at characteristic quality judgment, introducing thermal radiation contrast mass fraction and damage form mass fraction, firstly carrying out intensity scaling on infrared reconstruction, then extracting defect characteristics, and highlighting defect information; evaluating the shape and texture characteristics of the defect by using a matrix function; carrying out analogy on the distribution function of the matrix function and the power density of the image, and respectively calculating two defect characteristic mass fractions; objectively evaluating the extraction condition of the defect characteristics from two aspects of area comparison and damage forms; in addition, the quality of the infrared reconstructed image is objectively evaluated from the quantization angle by utilizing the peak signal-to-noise ratio;
aiming at the detection scheme, the integrity of defect types is extracted, the infrared reconstruction images of all damage types are respectively detected, and the credibility of the detection scheme on the detection conditions of all types of damage is evaluated; and detecting the infrared reconstruction image of the background area so as to judge the completeness of the extraction of different types of defects of the same test piece.
The complex multi-type defect detection and evaluation method based on infrared thermal imaging data analysis further comprises the following steps:
s1, acquiring an infrared sequence by using a detection scheme T to be evaluated, and performing feature extraction to obtain a background area infrared reconstruction image and Q infrared reconstruction images respectively containing at least perforations, bulges, impact pits and peeled Q ultra-high-speed impact damages;
step S2, respectively preprocessing the infrared reconstruction images for representing the k-th damage or background area obtained in the step S1 to obtain defect characteristic images;
step S3, objectively evaluating the defect feature extraction condition of the k-th auxiliary defect feature image obtained in the step S2 from the aspects of damage form and region comparison;
step S4, objectively evaluating the self image quality of the k-th sub-defect feature image obtained in the step S2 by utilizing the peak signal-to-noise ratio;
step S5, calculating total mass fractions of the infrared reconstructed images of the background area and each type of damage by integrating the damage characteristic evaluation result obtained in the step S3 and the image characteristic evaluation result obtained in the step S4;
s6, evaluating the credibility of the infrared reconstruction image based on the total mass fraction of the infrared reconstruction image obtained in the S5;
s7, calculating the objective quality score of the infrared detection scheme by using the credible infrared reconstruction images of the various types of damages obtained in the step S6;
s8, evaluating the identification performance of all ultra-high speed impact damage types based on the infrared detection scheme objective quality scores obtained in the S7;
step S9, based on the evaluation results of the step S8 on the identification performance of all the ultra-high speed impact types, under the condition that the identification performance of the current detection scheme on all the ultra-high speed impact damage is good, quantitative calculation is respectively carried out on the damage characteristics of all types; including quantitatively calculating the width, area, curvature characteristics of each type of lesion.
In the above technical solution, the infrared reconstructed image of the background area obtained in the step S1 is recorded as IRRIo(0)Recording the infrared image of the ultra-high speed impact damage as IRRIo(q)，q＝1,…,Q；
In step S2, the infrared reconstructed image IRRI obtained in step S1 for characterizing the kth lesion (i.e., k ═ 1,2, …, Q) or the background region (i.e., k ═ 0) is processedo(k)Respectively carrying out pretreatment, mainly comprising the following steps: firstly, respectively reconstructing the k infrared reconstruction image IRRI by utilizing an inverse tangent atan normalization methodo(k)Normalization, aiming at improving the stability of the defect characteristics of the infrared reconstructed image visualization, and recording the obtained normalized image as
Figure BDA0003509227520000121
k is 0,1, …, Q; then, performing expansion corrosion morphological operation on the normalized image, enhancing the image characteristics and reducing noise interference; performing convolution filtering on the image subjected to morphological processing by using a Gabor convolution filter to extract the characteristics of the edges, the textures and the like of the defects; finally obtaining a defect characteristic image IRRIdf(k)And k is 0,1, … and Q, and the specific steps are as follows:
step S21, reconstructing the IRRI of the infrared image based on the arctangent atan normalization methodo(k)K is 0,1, …, Q normalization, and stability of defect characteristics of infrared reconstructed image visualization is improved; IRRI (infrared image reconstruction) image by arc tangent atan normalization methodo(k)K is 0,1, …, and the data point R in Q is located at ith row and jth columnijCorresponding radiation value xijZooming to obtain new radiation value xi(new)j(new)：
Figure BDA0003509227520000122
Wherein, atan (x)ij) Is xijThe arctan function of; new radiation value x obtained after radiation value normalization of all infrared data pointsi(new)j(new)Forming a normalized image
Figure BDA0003509227520000123
k＝0,1,…,Q；
Step S22, normalizing the image
Figure BDA0003509227520000124
k is 0,1, …, Q is subjected to morphological processing of dilation followed by erosion to obtain a morphologically processed image
Figure BDA0003509227520000125
k＝0,1,…,Q；
First pair normalized image
Figure BDA0003509227520000126
k is 0,1, …, Q is subjected to expansion processing to obtain an expansion image
Figure BDA0003509227520000127
Figure BDA0003509227520000131
Wherein C is a neighborhood window of r × r, r is an odd number, the center pixel of the neighborhood window is (r +1)/2,
Figure BDA0003509227520000132
as an image
Figure BDA0003509227520000133
The expansion treatment process comprises the following steps: in the process of moving the neighborhood window C, if the neighborhood window C is matched with the image
Figure BDA0003509227520000134
If there is an overlapping area, recording the position, all moving neighborhood windows C and the image
Figure BDA0003509227520000135
The set of positions where intersections exist is an image
Figure BDA0003509227520000136
Dilated image under neighborhood window C
Figure BDA0003509227520000137
For the dilated image
Figure BDA0003509227520000138
k is 0,1, …, and Q is subjected to etching treatment:
Figure BDA0003509227520000139
theta is an image
Figure BDA00035092275200001310
And (3) corrosion treatment process: moving the neighborhood window C if it is adjacentDomain window C and image
Figure BDA00035092275200001311
The intersection of (A) completely belongs to the image
Figure BDA00035092275200001312
The location point is saved and all points satisfying the condition constitute an image
Figure BDA00035092275200001313
Corroded image obtained by corrosion of neighborhood window C
Figure BDA00035092275200001314
Etching image obtained by ending etching operation
Figure BDA00035092275200001315
I.e. morphologically processed images obtained by morphological processing of dilation followed by erosion
Figure BDA00035092275200001316
k＝0,1,…,Q；
Step S23, morphological processing image
Figure BDA00035092275200001317
k is 0,1, …, Q is processed by convolution filtering to extract the edge, texture and other characteristics of the defect, and a defect characteristic image IRRI is obtaineddf(k)，k＝0,1,…,Q：
Figure BDA00035092275200001318
Where F is a Gabor filter kernel of size n × n, representing the pair image
Figure BDA00035092275200001319
Performing Gabor convolution operation: image IRRIdf(k)Dividing the image into image blocks, selecting a scales and b directions to form ab Gabor filter groups; gabor filter bank and each imageThe block is convolved in a space domain, each image block can obtain ab filter outputs, and the average value of the filter outputs of the image blocks is extracted to form a column vector with the size of ab multiplied by 1 to be used as the defect characteristic of the image block; the defect characteristics of all image blocks form a defect characteristic image IRRIdf(k)The definition of each position (x, y) of the Gabor filter kernel is as follows:
Figure BDA00035092275200001320
wherein ,
x'＝x·cosθ+y·sinθ
y'＝-x·sinθ+y·cosθ，
wherein λ is a sine function wavelength; θ is the direction of the Gabor kernel function; ψ is a phase offset; ε is the bandwidth; γ is the aspect ratio of the space, which determines the shape of the Gabor kernel.
In the above technical solution, in the step S3, the k-th defect feature image IRRI obtained in the step S2 is compared from two aspects of damage morphology and region comparisondf(k)The defect feature extraction condition of the Q is objectively evaluated, wherein k is 0,1, …; because the radiation value of each data point in the infrared reconstructed image can reflect damage information, thermal radiation contrast mass fraction and damage form mass fraction are introduced on the basis, and characteristics such as shape, texture and the like of a defect are evaluated by utilizing a matrix function; the distribution function of the matrix function is compared with the power density of the image, and two defect characteristic mass fractions are respectively calculated by utilizing the central limit theorem, and the method specifically comprises the following steps:
step S31, calculating a normalized image
Figure BDA0003509227520000141
k 0,1, …, first order (mean) moment function of the radiation values of the Q data points
Figure BDA0003509227520000142
k is 0,1, …, Q and a second order (variance) moment function
Figure BDA0003509227520000143
k＝0,1,…,Q：
Figure BDA0003509227520000144
Figure BDA0003509227520000145
Wherein Z is a normalized image
Figure BDA0003509227520000146
k is 0,1, …, number of mid-infrared data points in Q, xvTo normalize the image
Figure BDA0003509227520000147
k is 0,1, …, the radiation value of the infrared data point in Q;
step S32, calculating a defect characteristic image IRRIdf(k)K is a first order moment function μ of 0,1, …, Qdf(k)K is 0,1, …, Q and a second moment function
Figure BDA0003509227520000148
k＝0,1,…,Q：
Figure BDA0003509227520000149
Figure BDA00035092275200001410
Wherein k is 0,1, …, Q, P is defect characteristic image IRRIdf(k)Number of mid-infrared data points, xnFor defective characteristic images IRRIdf(k)K is 0,1, …, Q infrared data point radiation value;
step S33, defect characteristic image IRRIdf(k)K is 0,1, …, Q demarcates the regions by radiance: dividing infrared data points with the same radiation value into one class, representing one characteristic of the defect, and recording the characteristic as IRTdf_l(k)，k＝0,1,…,Q；
Step S34, estimating characteristic IRTdf_l(k)K 0, 1.. times, the moment function μ of the Q radiation valuedf_(l)kK is 0,1, a distribution obeyed by Q: mean value μ according to the central limit theoremdf_l(k)Obey mean value of
Figure BDA00035092275200001411
Normal distribution of (a):
Figure BDA0003509227520000151
wherein k is 0,1,.., Q, p (μ ═ p)df_l(k)) Is mudf_l(k)Is determined by the probability density function of (a),
Figure BDA0003509227520000152
for the normalized image obtained in step S21
Figure BDA0003509227520000153
The mean value of the radiation values of the infrared data points,
Figure BDA0003509227520000154
is mudf_l(k)The variance of the obeyed normal distribution;
step S35, estimating characteristic IRTdf_l(k)K is 0,1, …, a function of the moment of the Q radiation value
Figure BDA0003509227520000155
k is 0,1, …, Q obeys a distribution: using the same estimate, variance, in step S34
Figure BDA0003509227520000156
Distribution of (2) is close to mean
Figure BDA0003509227520000157
Normal distribution of (a):
Figure BDA0003509227520000158
where k is 0,1, …, Q,
Figure BDA0003509227520000159
is that
Figure BDA00035092275200001510
Is determined by the probability density function of (a),
Figure BDA00035092275200001511
for the normalized image obtained in step S21
Figure BDA00035092275200001512
The variance of the radiation values of the infrared data points,
Figure BDA00035092275200001513
is composed of
Figure BDA00035092275200001514
The variance of the obeyed normal distribution;
step S36, calculating the three sigma principle respectively
Figure BDA00035092275200001515
k is 0,1, …, Q and
Figure BDA00035092275200001516
k＝0,1,…,Q：
Figure BDA00035092275200001517
Figure BDA00035092275200001518
where k is 0,1, …, Q,
Figure BDA00035092275200001519
is mudf_l(k)Distributed in intervals
Figure BDA00035092275200001520
The probability of (a) of (b) being,
Figure BDA00035092275200001521
for the normalized image obtained in step S21
Figure BDA00035092275200001522
The mean value of the infrared data point radiation values;
Figure BDA00035092275200001523
is composed of
Figure BDA00035092275200001524
Distributed in intervals
Figure BDA00035092275200001525
The probability of (a) of (b) being,
Figure BDA00035092275200001526
for the normalized image obtained in step S21
Figure BDA00035092275200001527
The variance of the radiation values of the pixels;
Figure BDA00035092275200001528
a distribution function that is a standard normal distribution;
step S37, calculating the value of mudf_l(k)K is 0,1, …, the normal distribution function of Q:
μdf_l(k)k is 0,1, …, and Q follows a normal distribution of
Figure BDA00035092275200001529
k is 0,1, …, Q, random variable RVμ(k)K is 0,1, …, Q for μdf_(l)kNormal distribution function of Q, k being 0,1, …
Figure BDA0003509227520000161
k is 0,1, …, Q is:
Figure BDA0003509227520000162
step S38, calculate about
Figure BDA0003509227520000163
k is 0,1, …, normal distribution function of Q:
Figure BDA0003509227520000164
k is 0,1, …, and Q follows a normal distribution of
Figure BDA0003509227520000165
k is 0,1, …, Q, random variable RVσ(k)K is 0,1, …, Q relates to
Figure BDA0003509227520000166
Normal distribution function of Q with k equal to 0,1, …
Figure BDA0003509227520000167
k is 0,1, …, Q is:
Figure BDA0003509227520000168
step S39, calculating thermal radiation contrast mass fraction TRC representing the background area and the infrared reconstruction image defect characteristics of various types of damage(k)，k＝0,1,…,Q：
Figure BDA0003509227520000169
Wherein k is 0,1, …, Q,
Figure BDA00035092275200001610
is a random variable RVσ(k)About
Figure BDA00035092275200001611
A normal distribution function of; random variable
Figure BDA00035092275200001612
Is by standardization
Figure BDA00035092275200001613
Obey an average of
Figure BDA00035092275200001614
Variance of
Figure BDA00035092275200001615
Obtained by normal distribution;
step S310, calculating damage form quality scores DM of infrared reconstruction image defect characteristics of representing background areas and various types of damages(k)，k＝0,1,…,Q：
Figure BDA00035092275200001616
Wherein k is 0,1, …, Q,
Figure BDA00035092275200001617
is a random variable RVμ(k)About mudf_l(k)A normal distribution function of; random variable
Figure BDA00035092275200001618
Is by normalizing μdf_l(k)Mean value of obeys of
Figure BDA00035092275200001619
Variance of
Figure BDA00035092275200001620
Obtained by normal distribution;
in step S311, if k is equal to 0, the background infrared reconstructed image IRRI is calculatedo(0)The mass fraction of (A): if TRC is present(0)+DM(0)Xi, wherein xi is defect feature threshold, the background image is considered to have the damage which is not separated, and the step S1 is returned to identify the damageAnd detecting the obtained damage features, extracting defect features, and otherwise, executing the step S4.
In the above technical solution, the step S4 uses the peak signal-to-noise ratio to compare the k-th defect feature image IRRI obtained in the step S2df(k)And k is 0,1, …, and the self image quality of Q is objectively evaluated to calculate the IRRI of each type of damage characteristic imagedf(k)K 0,1, …, Q peak signal-to-noise ratio P(k)：
Figure BDA0003509227520000171
Wherein k is 0,1, …, Q,
Figure BDA0003509227520000172
indicating defect characteristics IRTdf_l(k)And defect feature image IRRIdf(k)M × N denotes the defect feature image IRRIdf(k)Size of (IRT)df_l(k)(i, j) denotes the Defect feature IRTdf_l(k)Radiation value at position (i, j), IRRIdf(i, j) Defect feature image IRRIdf(k)The radiation value at position (i, j); MAXIRTdf_l(k)Indicating defect characteristics IRTdf_l(k)Maximum value of the radiation value of (1).
In the above technical solution, in the step S5, the damage characteristic evaluation result obtained in the step S3 and the image characteristic evaluation result obtained in the step S4 are integrated to calculate the total quality scores TQS of the infrared reconstructed image of the background region and each type of damage respectively(k)，k＝0,1,…,Q：
TQS(k)＝λ1·TRC(k)+λ2·DM(k)+λ3·P(k),k＝0,1,…,Q
wherein ,
Figure BDA0003509227520000173
are weight coefficients.
In the above technical solution, in the step S6, the total quality score TQS of the infrared reconstruction image obtained in the step S5(k)And k is 0,1, …, and Q is used for evaluating the reliability of the infrared reconstructed image, and the specific steps are as follows: if there is TQS(k)If η, k is 0,1, …, Q, where η is the damage assessment condition threshold, the current infrared reconstructed image is deemed to be unreliable for the k-th damage, and the process returns to step S1 to re-extract the kth infrared reconstructed image IRRI from the infrared sequenceo(k)Evaluating the extraction result of the kth damage; if the extracted result of the k-th damage is reevaluated, TQS still remains(k)If the measured value is less than eta, the method returns to the step S1 to detect the test piece to be detected again, acquires the infrared sequence again, and extracts the infrared reconstruction image IRRI of the visual k-th damageo(k)And then, evaluation was performed.
In the above technical solution, in the step S7, for the trusted infrared reconstructed images of the various types of damages obtained in the step S6, the objective quality score of the infrared detection scheme is calculated by using the trusted infrared reconstructed images, and the specific steps are as follows: q total mass fractions TQS of infrared reconstruction images respectively containing Q collision damages(k)After the calculation is finished, calculating the objective quality score Id of the infrared detection schemeoq：
Figure BDA0003509227520000181
By objective quality score IdoqTo judge the defect extraction condition, Id, of the infrared detection schemeoqThe larger the value of the defect identification result is, the better the detection identification result of the current infrared detection scheme on the complex multi-type defects formed by the ultrahigh-speed impact on the spacecraft is.
In the above technical solution, in the step S8, the objective quality score Id of the infrared detection scheme is obtained based on the step S7oqEvaluating the identification performance of all the ultra-high speed impact damage types; the method comprises the following specific steps: if any, IdoqIf the speed is higher than epsilon, wherein epsilon is a damage identification evaluation threshold value, the current detection scheme is considered to be capable of better identifying all types of ultra-high speed impact damage, and the accuracy of quantitative calculation of the damage is higher based on the infrared reconstructed image which is extracted from the current detection scheme and represents each type of damage; go to the stepS9, carrying out quantitative calculation on the ultra-high speed impact damage of each type; otherwise, return to step S1 from the Total quality score TQS(k)N infrared reconstructed images of < epsilon
Figure BDA0003509227520000182
Re-evaluating the infrared reconstructed image when
Figure BDA0003509227520000183
When the fraction is the fraction, rounding up; reextracting individual characterizations from infrared sequences
Figure BDA0003509227520000184
Damage by ultra-high speed impact
Figure BDA0003509227520000185
The infrared reconstructed image is displayed, the characterization effect of the infrared reconstructed image on the damage is improved, and the infrared reconstructed image is subjected to characterization again
Figure BDA0003509227520000186
Seed damage was assessed.
Through the steps, the objective quality score Id of the detection scheme T is obtainedoqThus, objective evaluation of the detection scheme T is realized. The detection scheme is objectively and effectively evaluated, so that complete complex multi-type ultrahigh-speed impact damage information can be extracted and mined in the early stage of defect quantitative identification, and a high-quality infrared reconstructed image with clear damage characteristics is obtained. The accuracy of further defect quantitative analysis is improved, and the precision of the whole ultrahigh-speed impact damage identification process is improved.
Examples
In the experiment, the thermal infrared imager collected 502 frames of images with pixel size of 512 × 640. The infrared sequence processing is performed to obtain 3 infrared reconstructed images with the size of 512 × 640, and the infrared defect reconstructed images are shown in fig. 3, 4 and 5. The type of the expression area in the reconstructed image can be judged according to the highlighted area of the infrared reconstructed image color, and the area types of the test piece are (a) the inside of the impact pit, (b) the background area and (c) the edge of the impact pit respectively correspond to fig. 3, fig. 4 and fig. 5.
Firstly, the IRRI image of the internal defect of the impact pit is reconstructedo(1)The quality was evaluated objectively. In the invention, in order to ensure the stability of data points of a defect region in subsequent processing, an infrared reconstructed image IRRI of the defect in the impact pit is obtainedo(1)Normalization processing is carried out, and images are normalized by the impact center
Figure BDA0003509227520000191
As shown in fig. 6 and 7. To illustrate the effectiveness of the method, we intercepted the radiation values of the infrared data points before and after normalization for comparison, as shown in table 1. Experimental results prove that the normalization method adopted by the invention is effective, the normalization does not change the relation between the radiation values of the data points, and the damage information stored in the infrared reconstruction image is not changed. It is reasonable to further evaluate the quality of the detection scheme based on the normalized image. Calculating to obtain an impact center normalized image
Figure BDA0003509227520000192
First order matrix function of
Figure BDA0003509227520000193
0.3974, second order matrix function
Figure BDA0003509227520000194
Is 0.2187.
TABLE 1 partial IR data point radiation value comparison before and after normalization of IR reconstructed image at impact center
Figure BDA0003509227520000195
Setting neighborhood window of 3X 3 size to normalize the image of the center of the collision
Figure BDA0003509227520000196
Performing morphological processing, and setting the bandwidth epsilon to be 2 pi; the aspect ratio gamma of the space is 0.5; sine functionThe wavelength lambda is 12; theta is 0, the phase shift psi is 0, and an impact center defect characteristic image IRRI is obtaineddf(1)As shown in fig. 7. As can be seen from FIG. 7, the extracted impact center defect feature image IRRIdf(1)Although some background information is obscured, the defect features are left intact. And the defect information is highlighted, the comparison among all the regions is more obvious, and the objective evaluation on the defect extraction condition is further facilitated. Calculating to obtain an IRRI image of the defect of the impact centerdf(1)First order matrix function mudf(1)0.1749, second order matrix function
Figure BDA0003509227520000197
Is 0.2975.
Defect feature image IRRI at impact centerdf(1)Middle, characteristic IRTdf_l(1)Mean value of the radiation values mudf_l(1)The normal distribution obeyed is:
Figure BDA0003509227520000198
characteristic IRTdf_l(1)Variance of radiation value
Figure BDA0003509227520000199
The normal distribution obeyed is:
Figure BDA00035092275200001910
respectively calculating different characteristic IRT by utilizing three-sigma principledf_l(1)The variance of the distribution of the normal distribution to which each matrix function of radiation values follows. Calculating to obtain impact center image characteristics IRTdf_l(1)Mean value of the radiation values mudf_l(1)Variance of obeyed normal distribution
Figure BDA00035092275200001911
Is 0.13252Then there is μdf_l(1)～N(0.3974,0.13252) (ii) a Variance (variance)
Figure BDA00035092275200001912
Variance of obeyed normal distribution
Figure BDA00035092275200001913
Is 0.07292Then there is
Figure BDA0003509227520000201
Specific parameters are shown in table 2.
TABLE 2 impact center image Defect feature quality score calculation
Figure BDA0003509227520000202
By standardisation
Figure BDA0003509227520000203
Obeying normal distribution to obtain random variable RVσ(1)Then based on the random variable RVσ(1)About
Figure BDA0003509227520000204
Normal distribution function of
Figure BDA0003509227520000205
Calculating thermal radiation contrast mass fraction TRC(1). When the feature number l of the defect tends to infinity, the central defect feature IRT is impacteddf_l(1)Values of the respective matrix functions of the radiation values and the impact center defect feature image IRRIdf(1)In which the values of the respective matrix functions are the same, i.e. mudf_l(1)＝μdf(1)，
Figure BDA0003509227520000206
So when the number of defective features/tends to infinity, the impact center image feature IRTdf_l(1)Mean value of the radiation values mudf_l(1)0.1749, variance
Figure BDA0003509227520000207
Is 0.2975. Normalized impact center image feature IRTdf_l(1)Variance of radiation value
Figure BDA0003509227520000208
Obeying normal distribution to obtain random variable RVσ(1)Is 1.0801. Calculating thermal radiation contrast mass fraction TRC(1)Is 0.8600. Thermal radiation versus mass fraction TRC(1)The relevant parameters are shown in table 2.
By normalizing μdf_l(1)Obeying normal distribution to obtain random variable RVμ(1)Then based on the random variable RVμ(1)About mudf_l(1)Normal distribution function of
Figure BDA0003509227520000209
Calculating the Damage morphology mass fraction DM(1). Normalized impact center image feature IRTdf_l(1)Mean value of the radiation values mudf_l(1)Obeying normal distribution to obtain random variable RVμ(1)To 0.2586, a damage morphology mass fraction DM was calculated(1): currently there is mudf_l(1)＝0.1749，
Figure BDA00035092275200002010
Is provided with
Figure BDA00035092275200002011
So the current damage morphology mass fraction DM(1)The specific expression is as follows:
Figure BDA00035092275200002012
calculating to obtain the quality fraction DM of the damage form(1)Is 0.0931. Mass fraction of damage morphology DM(1)The relevant parameters are shown in table 3.
TABLE 3 image quality-related parameter calculation of impact center images
Figure BDA00035092275200002014
Mean square error MSE of impact center infrared reconstructed image(1)0.2596, characteristic maximum radiation value
Figure BDA00035092275200002013
To 1.7221, the impact was calculatedPeak signal ratio P of central infrared reconstructed image(1)Is 10.5758. Peak signal to noise ratio P(1)The specific calculation results of the relevant parameters are shown in table 4. Heat radiation contrast mass fraction TRC of impact center infrared reconstructed image(1)0.8600, the weighting coefficient lambda thereof1Set to 0.65; mass fraction of damage form DM(1)0.0931, the weighting coefficient lambda thereof2Is 0.3; peak signal to noise ratio of P(1)10.5758, weighting coefficient lambda thereof3Set to 0.1, the total mass fraction TQS of the infrared reconstructed image(1)Is 1.1158.
TABLE 4 impact center image Total quality score TQS(1)Calculating out
Figure BDA0003509227520000211
Infrared reconstructed image IRRI of impact pit edge defect according to same processo(2)The quality was evaluated objectively. Firstly, the IRRI image of the impact pit edge defect is reconstructedo(2)Normalization processing is carried out, and pit edge normalization images are impacted
Figure BDA0003509227520000212
As shown in fig. 8. To illustrate the effectiveness of the method, we intercepted the radiation values of the partial infrared data points before and after normalization for comparison, as shown in table 5. Calculating to obtain an impact edge normalized image
Figure BDA0003509227520000213
First order matrix function of
Figure BDA0003509227520000214
Is-0.2683, second order matrix function
Figure BDA0003509227520000215
Is 0.0814. As shown in fig. 9, the impact edge feature image IRRI is extracteddf(2)Calculating to obtain an impact edge defect characteristic image IRRIdf(2)First order matrix function mudf(2)Is 0.0900, twoOrder matrix function
Figure BDA0003509227520000216
Is 0.0901.
TABLE 5 partial IR data point radiation value comparison before and after normalization of impact edge IR reconstructed image
Figure BDA0003509227520000217
Impact edge image feature IRTdf_l(2)Mean value of the radiation values mudf_l(2)Variance of obeyed normal distribution
Figure BDA0003509227520000218
Is (-0.0894)2Then there is μdf_l(2)～N[-0.2683,(-0.0894)2](ii) a Variance (variance)
Figure BDA0003509227520000219
Variance of obeyed normal distribution
Figure BDA00035092275200002110
Is 0.02712Then there is
Figure BDA00035092275200002111
Specific parameters are shown in table 3. Normalized impact edge image feature IRTdf_l(2)Variance of radiation value
Figure BDA00035092275200002112
Obeying normal distribution to obtain random variable RVσ(2)To 0.3225, a thermal radiation contrast mass fraction TRC is calculated(2)Is 0.6265. Thermal radiation versus mass fraction TRC(2)The relevant parameters are shown in table 6. Normalized impact edge image feature IRTdf_l(2)Mean value of radiation values mudf_l(2)Obeying normal distribution to obtain random variable RVμ(2)To 0.5720, a damage morphology mass fraction DM was calculated(2)Is 0.0001. Mass fraction of damage form DM(2)The relevant parameters are shown in table 6.
TABLE 6 impact edge image Defect feature quality score calculation
Figure BDA0003509227520000221
Mean square error MSE of impact edge infrared reconstructed image(2)0.1340, characteristic maximum radiation value
Figure BDA0003509227520000222
To 1.4282, a peak signal ratio P is calculated(2)Is 11.8248. Peak signal to noise ratio P(2)The specific calculation results of the relevant parameters are shown in table 7:
TABLE 7 image quality-related parameter calculation of impact edge images
Figure BDA0003509227520000223
Calculating to obtain total mass fraction TQS of impact edge infrared reconstruction image(2)Is 0.9985. Total mass fraction TQS(2)The results of the calculation of the correlation parameters are shown in table 8:
TABLE 8 impact edge image Total quality score TQS(2)Computing
Figure BDA0003509227520000224
Infrared reconstructed image IRRI of background area according to same processo(0)And (5) performing quality evaluation. The threshold ξ is set at 0.5. Calculating to obtain an infrared reconstructed image IRRI of a background areao(0)Thermal radiation versus mass fraction TR of(C0)0.3179; mass fraction of damage form DM(0)Was 0.0129. With TRC(0)+DM(0)0.3308 < ξ. The lesion features have been fully identified. The calculation results of the infrared reconstructed image objective quality scores of the background regions are shown in table 9:
TABLE 9 Observation quality score calculation for infrared reconstructed image of background area
Figure BDA0003509227520000225
The damage identification evaluation threshold epsilon was set to 0.9. Reconstruction of images IRRI from mid-infrared of head-on collisiono(1)Impact edge reconstructed image IRRIo(2)And background region infrared reconstruction image IRRIo(0)And finally calculating to obtain the objective quality score Id of the current infrared detection schemeoqComprises the following steps:
Figure BDA0003509227520000231
Idoqand epsilon indicates that the current detection scheme has a good recognition effect on all types of ultra-high-speed impact damage. The experimental result proves that the defect evaluation mode designed by the invention can objectively evaluate the extraction condition of the infrared detection scheme on the defects by combining the infrared image quality and the detection flow of the detection scheme. The detection scheme is objectively and effectively evaluated, so that complete complex multi-type ultrahigh-speed impact damage information can be extracted and mined in the early stage of defect quantitative identification, and a high-quality infrared reconstructed image with clear damage characteristics is obtained. The accuracy of further defect quantitative analysis is improved, and the precision of the whole ultrahigh-speed impact damage identification process is improved.
The number of apparatuses and the scale of the process described herein are intended to simplify the description of the present invention. Applications, modifications and variations of the present invention will be apparent to those skilled in the art.
While embodiments of the invention have been described above, it is not limited to the applications set forth in the description and the embodiments, which are fully applicable in various fields of endeavor to which the invention pertains, and further modifications may readily be made by those skilled in the art, it being understood that the invention is not limited to the details shown and described herein without departing from the general concept defined by the appended claims and their equivalents.

Claims (9)

1. A complex multi-type defect detection and evaluation method based on infrared thermal imaging data analysis is characterized by comprising the following steps: aiming at characteristic quality judgment, introducing thermal radiation contrast mass fraction and damage form mass fraction, firstly carrying out intensity scaling on infrared reconstruction, then extracting defect characteristics, and highlighting defect information; evaluating the shape and texture characteristics of the defect by using a matrix function; the distribution function of the matrix function is analogized with the power density of the image, and two defect characteristic mass fractions are respectively calculated; the extraction condition of defect features is objectively evaluated from two aspects of regional comparison and damage form; in addition, the quality of the infrared reconstructed image is objectively evaluated from the quantization angle by utilizing the peak signal-to-noise ratio;
aiming at the detection scheme, the integrity of defect types is extracted, the infrared reconstruction images of all damage types are respectively detected, and the credibility of the detection scheme on the detection conditions of all types of damage is evaluated; and detecting the infrared reconstruction image of the background area so as to judge the completeness of the extraction of different types of defects of the same test piece.
2. The method for complex multi-type defect detection and evaluation based on infrared thermal imaging data analysis of claim 1, further comprising the steps of:
s1, acquiring an infrared sequence by using a detection scheme T to be evaluated, and performing feature extraction to obtain a background area infrared reconstruction image and Q infrared reconstruction images respectively containing at least perforations, bulges, impact pits and peeled Q ultra-high-speed impact damages;
step S2, respectively preprocessing the infrared reconstruction images for representing the k-th damage or background area obtained in the step S1 to obtain defect characteristic images;
step S3, objectively evaluating the defect feature extraction condition of the k-th auxiliary defect feature image obtained in the step S2 from the aspects of damage form and region comparison;
step S4, objectively evaluating the self image quality of the k-th sub-defect feature image obtained in the step S2 by utilizing the peak signal-to-noise ratio;
step S5, calculating the total mass fraction of the infrared reconstructed image of the background area and each type of damage by integrating the damage characteristic evaluation result obtained in the step S3 and the image characteristic evaluation result obtained in the step S4;
s6, evaluating the credibility of the infrared reconstruction image based on the total mass fraction of the infrared reconstruction image obtained in the S5;
s7, calculating the objective quality score of the infrared detection scheme by using the credible infrared reconstruction images of the various types of damages obtained in the step S6;
s8, evaluating the identification performance of all ultra-high speed impact damage types based on the infrared detection scheme objective quality scores obtained in the S7;
step S9, based on the evaluation results of the step S8 on the identification performance of all the ultra-high speed impact types, under the condition that the identification performance of the current detection scheme on all the ultra-high speed impact damage is good, quantitative calculation is respectively carried out on the damage characteristics of all types; including quantitatively calculating the width, area, curvature characteristics of each type of lesion.
3. The method for detecting and evaluating the defects of multiple complex types based on the infrared thermal imaging data analysis of claim 2, wherein the infrared reconstructed image of the background area obtained in the step S1 is recorded as IRRIo(0)Recording the infrared image of the ultra-high speed impact damage as IRRIo(q)，q＝1，…，Q；
In the step S2, the infrared reconstructed image IRRI for characterizing the k-th injury or background region obtained in the step S1 is processedo(k)Respectively carrying out pretreatment, mainly comprising the following steps: firstly, respectively reconstructing the k infrared reconstruction image IRRI by utilizing an inverse tangent atan normalization methodo(k)Normalization, aiming at improving the stability of the defect characteristics of the infrared reconstructed image visualization, and recording the obtained normalized image as
Figure FDA0003509227510000021
Figure FDA0003509227510000022
Then, performing expansion corrosion morphological operation on the normalized image, enhancing image characteristics and reducing noise interference; performing convolution filtering on the image subjected to morphological processing by using a Gabor convolution filter to extract the characteristics of the edges, the textures and the like of the defects; finally obtaining a defect characteristic image IRRIdf(k)And k is 0,1, … and Q, and the specific steps are as follows:
step S21, reconstructing the IRRI of the infrared image based on the arctangent atan normalization methodo(k)K is 0,1, …, Q normalization, and stability of defect characteristics of infrared reconstructed image visualization is improved; IRRI (infrared image reconstruction) image by arc tangent atan normalization methodo(k)K is 0,1, …, and the data point R in Q is located at ith row and jth columnijCorresponding radiation value xijZooming to obtain new radiation value xi(new)j(new)：
Figure FDA0003509227510000023
Wherein, atan (x)ij) Is xijThe arctan function of; new radiation value x obtained after radiation value normalization of all infrared data pointsi(new)j(new)Forming a normalized image
Figure FDA0003509227510000024
Step S22, normalizing the image
Figure FDA0003509227510000025
Performing morphological processing of expansion and corrosion to obtain morphological processing image
Figure FDA0003509227510000026
First pair normalized image
Figure FDA0003509227510000027
Performing expansion processing to obtain an expansion image
Figure FDA0003509227510000028
Figure FDA0003509227510000029
Wherein C is a neighborhood window of r × r, r is an odd number, the center pixel of the neighborhood window is (r +1)/2,
Figure FDA0003509227510000031
as an image
Figure FDA0003509227510000032
The expansion treatment process comprises the following steps: in the process of moving the neighborhood window C, if the neighborhood window C is matched with the image
Figure FDA0003509227510000033
If there is an overlapping area, recording the position, all moving neighborhood windows C and the image
Figure FDA0003509227510000034
The set of positions where intersections exist is an image
Figure FDA0003509227510000035
Dilated image under neighborhood window C
Figure FDA0003509227510000036
For the dilated image
Figure FDA0003509227510000037
Carrying out corrosion treatment:
Figure FDA0003509227510000038
theta is an image
Figure FDA0003509227510000039
And (3) corrosion treatment process: moving the neighborhood window C, if the neighborhood window C and the image
Figure FDA00035092275100000310
The intersection of (A) completely belongs to the image
Figure FDA00035092275100000311
The location point is saved and all points satisfying the condition constitute an image
Figure FDA00035092275100000312
Corroded image obtained by corrosion of neighborhood window C
Figure FDA00035092275100000313
Etching image obtained by ending etching operation
Figure FDA00035092275100000314
I.e. morphologically processed images obtained by morphological processing of dilation followed by erosion
Figure FDA00035092275100000315
Step S23, morphological processing image
Figure FDA00035092275100000316
Convolution filtering is carried out to extract the characteristics of the edge, the texture and the like of the defect so as to obtain a defect characteristic image IRRIdf(k)，k＝0，1，…，Q：
Figure FDA00035092275100000317
Where F is a Gabor filter kernel of size n × n, representing the pair image
Figure FDA00035092275100000318
Performing Gabor convolution operation: image IRRIdf(k)Dividing the image into image blocks, selecting a scales and b directions to form ab Gabor filter groups; the Gabor filter bank is convolved with each image block in a space domain, each image block can obtain ab filter outputs, and the average value of the filter outputs of the image blocks is extracted to form a column vector with the size of ab x 1 to serve as the defect characteristic of the image block; the defect characteristics of all image blocks form a defect characteristic image IRRIdf(k)The definition of each position (x, y) of the Gabor filter kernel is as follows:
Figure FDA00035092275100000319
wherein ,
x′＝x·cosθ+y·sinθ
y′＝-x·sinθ+y·cosθ，
wherein λ is a sine function wavelength; θ is the direction of the Gabor kernel function; ψ is a phase offset; ε is the bandwidth; γ is the aspect ratio of the space, which determines the shape of the Gabor kernel.
4. The method for detecting and evaluating the defects of multiple types based on the IR thermographic data analysis of claim 3, wherein said step S3 is implemented for the k-th defect feature image IRRI obtained in step S2 according to both the shape and the area contrast of the damaged areadf(k)The defect feature extraction condition of the Q is objectively evaluated, wherein k is 0,1, …; because the radiation value of each data point in the infrared reconstructed image can reflect damage information, thermal radiation contrast mass fraction and damage form mass fraction are introduced on the basis, and characteristics such as shape, texture and the like of a defect are evaluated by utilizing a matrix function; the distribution function of the matrix function is analogized with the power density of the image, and two defect characteristic mass fractions are respectively calculated by utilizing the central limit theorem, and the method specifically comprises the following steps:
step S31, calculating a normalized image
Figure FDA0003509227510000041
First moment function of data point radiation value
Figure FDA0003509227510000042
And second moment function
Figure FDA0003509227510000043
Figure FDA0003509227510000044
Figure FDA0003509227510000045
Wherein Z is a normalized image
Figure FDA0003509227510000046
Number of mid-infrared data points, xvTo normalize the image
Figure FDA0003509227510000047
The radiation value of the mid-infrared data point;
step S32, calculating a defect characteristic image IRRIdf(k)K is 0,1, …, a first moment function μ of Qdf(k)K is 0,1, …, Q and a second moment function
Figure FDA0003509227510000048
Figure FDA0003509227510000049
Figure FDA00035092275100000410
Wherein k is 0,1, …, Q, P is a defect characteristic image IRRIdf(k)Number of mid-infrared data points, xnFor defect feature images IRRIdf(k)K is 0,1, …, Q infrared data point radiation value;
step S33, defect characteristic image IRRIdf(k)K is 0,1, …, Q demarcates the regions by radiance: dividing infrared data points with the same radiation value into one class, representing one characteristic of the defect, and recording the characteristic as IRTdf_l(k)，k＝0，1，…，Q；
Step S34, estimating characteristic IRTdf_l(k)K is 0,1, …, a function of the moment of the Q radiation value μdf_l(k)K is 0,1, …, the distribution obeyed by Q: mean value μ according to the central limit theoremdf_l(k)Obey mean value of
Figure FDA00035092275100000411
Normal distribution of (a):
Figure FDA0003509227510000051
where k is 0,1, …, Q, p (μ)df_l(k)) Is mudf_l(k)Is determined by the probability density function of (a),
Figure FDA0003509227510000052
for the normalized image obtained in step S21
Figure FDA0003509227510000053
The mean value of the radiation values of the infrared data points,
Figure FDA0003509227510000054
is mudf_l(k)The variance of the obeyed normal distribution;
step S35, estimating characteristic IRTdf_l(k)K is 0,1, …, a function of the moment of the Q radiation value
Figure FDA0003509227510000055
Figure FDA00035092275100000530
Distribution of compliance: using the same estimate, variance, in step S34
Figure FDA0003509227510000056
Distribution of (2) is close to mean
Figure FDA0003509227510000057
Normal distribution of (a):
Figure FDA0003509227510000058
where k is 0,1, …, Q,
Figure FDA0003509227510000059
is that
Figure FDA00035092275100000510
Is determined by the probability density function of (a),
Figure FDA00035092275100000511
for the normalized image obtained in step S21
Figure FDA00035092275100000512
The variance of the radiation values of the infrared data points,
Figure FDA00035092275100000513
is composed of
Figure FDA00035092275100000514
A variance of the obeyed normal distribution;
step S36, calculating the three sigma principle respectively
Figure FDA00035092275100000515
And
Figure FDA00035092275100000516
Figure FDA00035092275100000517
Figure FDA00035092275100000518
wherein k is 0,1, …, Q,
Figure FDA00035092275100000519
is mudf_l(k)Distributed in intervals
Figure FDA00035092275100000520
The probability of (a) of (b) being,
Figure FDA00035092275100000521
for the normalized image obtained in step S21
Figure FDA00035092275100000522
The mean value of the infrared data point radiation values;
Figure FDA00035092275100000523
is composed of
Figure FDA00035092275100000524
Is distributed in the interval
Figure FDA00035092275100000525
The probability of (a) of (b) being,
Figure FDA00035092275100000526
for the normalized image obtained in step S21
Figure FDA00035092275100000527
The variance of the radiation values of the pixels;
Figure FDA00035092275100000528
a distribution function that is a standard normal distribution;
step S37, calculating the value of mudf_l(k)K is 0,1, …, normal distribution function of Q:
μdf_l(k)k is 0,1, …, and Q follows a normal distribution of
Figure FDA00035092275100000529
Random variable RVμ(k)K is 0,1, …, Q for μdf_l(k)Normal distribution function of Q, k being 0,1, …
Figure FDA0003509227510000061
Comprises the following steps:
Figure FDA0003509227510000062
step S38, calculate about
Figure FDA0003509227510000063
Normal distribution function of (a):
Figure FDA0003509227510000064
Figure FDA0003509227510000065
obey a normal distribution of
Figure FDA0003509227510000066
Random variable RVσ(k)K is 0,1, …, Q relates to
Figure FDA0003509227510000067
Normal distribution function of
Figure FDA0003509227510000068
Figure FDA0003509227510000069
Comprises the following steps:
Figure FDA00035092275100000610
step S39, calculating thermal radiation contrast mass fraction TRC representing the background area and the infrared reconstruction image defect characteristics of various types of damage(k)，k＝0，1，…，Q：
Figure FDA00035092275100000611
Wherein k is 0,1, …, Q,
Figure FDA00035092275100000612
is a random variable RVσ(k)About
Figure FDA00035092275100000613
A normal distribution function of; random variable
Figure FDA00035092275100000614
Is by standardization
Figure FDA00035092275100000615
Obey an average of
Figure FDA00035092275100000616
Variance of
Figure FDA00035092275100000617
Normal scoreObtaining the cloth;
step S310, calculating damage form quality scores DM of infrared reconstruction image defect characteristics of representing background areas and various types of damages(k)，k＝0，1，…，Q：
Figure FDA00035092275100000618
Wherein k is 0,1, …, Q,
Figure FDA00035092275100000619
is a random variable RVμ(k)About mudf_l(k)A normal distribution function of; random variable
Figure FDA00035092275100000620
Is by normalizing μdf_l(k)Obey an average of
Figure FDA00035092275100000621
Variance of
Figure FDA00035092275100000622
Obtained by normal distribution;
in step S311, if k is equal to 0, the background infrared reconstructed image IRRI is calculatedo(0)The mass fraction of (A): if TRC is present(0)+DM(0)And if yes, determining that the background image has unseparated damage, returning to the step S1 to detect the unrecognized damage characteristic, and extracting the defect characteristic, otherwise, executing the step S4.
5. The method for detecting and evaluating the defects of multiple types based on the IR thermographic data analysis of claim 4, wherein said step S4 utilizes the peak SNR for the k-th IRRI image obtained in step S2df(k)And (5) objectively evaluating the self image quality of the image with k being 0,1, … and Q, and calculating the IRRI of each type of damage characteristic imagedf(k)K 0,1, …, Q peak signal-to-noise ratio P(k)：
Figure FDA0003509227510000071
Wherein k is 0,1, …, Q,
Figure FDA0003509227510000072
indicating defect characteristics IRTdf_l(k)And defect feature image IRRIdf(k)M × N denotes the defect feature image IRRIdf(k)Size of (IRT)df_l(k)(i, j) denotes the Defect feature IRTdf_l(k)Radiation value at position (i, j), IRRIdf(i, j) Defect feature image IRRIdf(k)The radiation value at position (i, j);
Figure FDA0003509227510000073
characterizing defects IRTdf_l(k)Maximum value of the radiation value of (1).
6. The method of claim 5, wherein in step S5, the evaluation result of the damage characteristic obtained in step S3 and the evaluation result of the image characteristic obtained in step S4 are combined to calculate the total quality score TQS of the infrared reconstructed image of the background region and each type of damage respectively(k)，k＝0，1，…，Q：
TQS(k)＝λ1·TRC(k)+λ2·DM(k)+λ3·P(k)，k＝0，1，…，Q
wherein ,
Figure FDA0003509227510000074
are weight coefficients.
7. The complex multi-type defect detection based on infrared thermographic data analysis of claim 6The evaluation method is characterized in that in the step S6, the total quality score TQS based on the infrared reconstruction image obtained in the step S5(k)Evaluating the reliability of the infrared reconstructed image, wherein k is 0,1, … and Q, and the specific steps are as follows: if there is TQS(k)If η, k is 0,1, …, Q, where η is the damage assessment condition threshold, the current infrared reconstructed image is deemed to be unreliable for the k-th damage, and the process returns to step S1 to re-extract the kth infrared reconstructed image IRRI from the infrared sequenceo(k)Evaluating the extraction result of the kth damage; if the extracted result of the k-th damage is reevaluated, TQS still remains(k)If the measured value is less than eta, the method returns to the step S1 to detect the test piece to be detected again, acquires the infrared sequence again, and extracts the infrared reconstruction image IRRI of the visual k-th damageo(k)And then, evaluation was performed.
8. The method for detecting and evaluating the defects of multiple types based on the infrared thermal imaging data analysis as claimed in claim 7, wherein in the step S7, the steps of obtaining the credible infrared reconstruction images of the damage types in the step S6 and calculating the objective quality score of the infrared detection scheme by using the credible infrared reconstruction images comprise: q total mass fractions TQS of infrared reconstruction images respectively containing Q collision damages(k)After the calculation is finished, calculating the objective quality score Id of the infrared detection schemeoq：
Figure FDA0003509227510000081
By objective quality score IdoqTo judge the defect extraction condition, Id, of the infrared detection schemeoqThe larger the value of the defect identification result is, the better the detection identification result of the current infrared detection scheme on the complex multi-type defects formed by the ultrahigh-speed impact on the spacecraft is.
9. The method for detecting and evaluating complex multi-type defects based on infrared thermal imaging data analysis according to claim 8, wherein in the step S8,objective quality score Id of infrared detection scheme obtained based on step S7oqEvaluating the identification performance of all the ultra-high speed impact damage types; the method comprises the following specific steps: if any, IdoqIf the speed is higher than epsilon, wherein epsilon is a damage identification evaluation threshold value, the current detection scheme is considered to be capable of better identifying all types of ultra-high speed impact damage, and the accuracy of quantitative calculation of the damage is higher based on the infrared reconstructed image which is extracted from the current detection scheme and represents each type of damage; step S9 is transferred to carry out quantitative calculation on the ultra-high speed impact damage of each type; otherwise, return to step S1 from the Total quality score TQS(k)N infrared reconstructed images of < epsilon
Figure FDA0003509227510000082
Re-evaluating the infrared reconstructed image when
Figure FDA0003509227510000083
When the fraction is the fraction, rounding up; reextracting individual signatures from infrared sequences
Figure FDA0003509227510000084
Damage by ultra-high speed impact
Figure FDA0003509227510000085
The infrared reconstructed image is displayed, the characterization effect of the infrared reconstructed image on the damage is improved, and the infrared reconstructed image is subjected to characterization again
Figure FDA0003509227510000086
Seed damage was assessed.
CN202210148014.2A 2022-02-17 2022-02-17 Complex multi-type defect detection evaluation method based on infrared thermal imaging data analysis Active CN114549448B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202210148014.2A CN114549448B (en) 2022-02-17 2022-02-17 Complex multi-type defect detection evaluation method based on infrared thermal imaging data analysis

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202210148014.2A CN114549448B (en) 2022-02-17 2022-02-17 Complex multi-type defect detection evaluation method based on infrared thermal imaging data analysis

Publications (2)

Publication Number Publication Date
CN114549448A true CN114549448A (en) 2022-05-27
CN114549448B CN114549448B (en) 2023-08-11

Family

ID=81676312

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202210148014.2A Active CN114549448B (en) 2022-02-17 2022-02-17 Complex multi-type defect detection evaluation method based on infrared thermal imaging data analysis

Country Status (1)

Country Link
CN (1) CN114549448B (en)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN115272301A (en) * 2022-09-20 2022-11-01 江苏新世嘉家纺高新科技股份有限公司 Automatic cheese defect detection method based on robot

Citations (21)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2006022568A1 (en) * 2004-08-13 2006-03-02 Zao 'koordinacionny Centr Po Nadeznosty, Bezopasnosty I Resursu Oborudovania I Truboprovodam Atomnyh Stancy' Method for determining a product initial and residual defectiveness, a defect detection probability and a product quality according to certainty and probability components of a residual defectiveness
CN107918216A (en) * 2017-12-13 2018-04-17 深圳Tcl新技术有限公司 Image Mura defects appraisal procedure, system and readable storage medium storing program for executing
US20180218502A1 (en) * 2017-01-27 2018-08-02 Arterys Inc. Automated segmentation utilizing fully convolutional networks
CN108896389A (en) * 2018-07-17 2018-11-27 北华大学 A kind of damages of concrete structures monitoring method based on piezoelectric intelligent aggregate
US20190197678A1 (en) * 2016-08-29 2019-06-27 Wuhan Jingce Electronic Group Co., Ltd. Gpu-based tft-lcd mura defect detection method
US10620133B1 (en) * 2016-08-17 2020-04-14 United States Of America As Represented By The Administrator Of The National Aeronautics And Space Administration Contrast-based imaging and analysis computer-implemented methods to analyze thermography data for nondestructive evaluation
CN111652252A (en) * 2020-06-11 2020-09-11 中国空气动力研究与发展中心超高速空气动力研究所 Ultrahigh-speed impact damage quantitative identification method based on ensemble learning
CN112016628A (en) * 2020-09-04 2020-12-01 中国空气动力研究与发展中心超高速空气动力研究所 Space debris impact damage interpretation method based on dynamic multi-target prediction
US20210001171A1 (en) * 2012-08-31 2021-01-07 Blue Goji Llc System and method for evaluation, detection, conditioning, and treatment of neurological functioning and conditions
CN113409314A (en) * 2021-08-18 2021-09-17 南京市特种设备安全监督检验研究院 Unmanned aerial vehicle visual detection and evaluation method and system for corrosion of high-altitude steel structure
CN113409290A (en) * 2021-06-29 2021-09-17 北京兆维电子(集团)有限责任公司 Method and device for detecting appearance defects of liquid crystal display and storage medium
CN113498528A (en) * 2020-01-21 2021-10-12 京东方科技集团股份有限公司 Image defect determining method and device, electronic equipment and storage medium
CN113538418A (en) * 2021-08-27 2021-10-22 浙江工业大学 Tire X-ray image defect extraction model construction method based on morphological analysis
CN113570605A (en) * 2021-09-28 2021-10-29 深圳市绘晶科技有限公司 Defect detection method and system based on liquid crystal display panel
CN113763367A (en) * 2021-09-13 2021-12-07 中国空气动力研究与发展中心超高速空气动力研究所 Comprehensive interpretation method for infrared detection characteristics of large-size test piece
US20210383563A1 (en) * 2021-06-21 2021-12-09 University Of Electronic Science And Technology Of China Method for quantitatively identifying the defects of large-size composite material based on infrared image sequence
CN113781445A (en) * 2021-09-13 2021-12-10 中国空气动力研究与发展中心超高速空气动力研究所 Multi-region complex damage defect feature extraction fusion method
CN113793318A (en) * 2021-09-13 2021-12-14 中国空气动力研究与发展中心超高速空气动力研究所 Multi-region complex damage defect characteristic comprehensive analysis method
CN113887667A (en) * 2021-10-28 2022-01-04 中国人民解放军空军工程大学 Composite material defect detection method and system based on infrared and ultrasonic signal fusion
CN113887666A (en) * 2021-10-28 2022-01-04 中国人民解放军空军工程大学 Defect detection method and system for composite material based on Transformer
CN113989224A (en) * 2021-10-27 2022-01-28 西安工程大学 Color texture fabric defect detection method based on generation countermeasure network

Patent Citations (21)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2006022568A1 (en) * 2004-08-13 2006-03-02 Zao 'koordinacionny Centr Po Nadeznosty, Bezopasnosty I Resursu Oborudovania I Truboprovodam Atomnyh Stancy' Method for determining a product initial and residual defectiveness, a defect detection probability and a product quality according to certainty and probability components of a residual defectiveness
US20210001171A1 (en) * 2012-08-31 2021-01-07 Blue Goji Llc System and method for evaluation, detection, conditioning, and treatment of neurological functioning and conditions
US10620133B1 (en) * 2016-08-17 2020-04-14 United States Of America As Represented By The Administrator Of The National Aeronautics And Space Administration Contrast-based imaging and analysis computer-implemented methods to analyze thermography data for nondestructive evaluation
US20190197678A1 (en) * 2016-08-29 2019-06-27 Wuhan Jingce Electronic Group Co., Ltd. Gpu-based tft-lcd mura defect detection method
US20180218502A1 (en) * 2017-01-27 2018-08-02 Arterys Inc. Automated segmentation utilizing fully convolutional networks
CN107918216A (en) * 2017-12-13 2018-04-17 深圳Tcl新技术有限公司 Image Mura defects appraisal procedure, system and readable storage medium storing program for executing
CN108896389A (en) * 2018-07-17 2018-11-27 北华大学 A kind of damages of concrete structures monitoring method based on piezoelectric intelligent aggregate
CN113498528A (en) * 2020-01-21 2021-10-12 京东方科技集团股份有限公司 Image defect determining method and device, electronic equipment and storage medium
CN111652252A (en) * 2020-06-11 2020-09-11 中国空气动力研究与发展中心超高速空气动力研究所 Ultrahigh-speed impact damage quantitative identification method based on ensemble learning
CN112016628A (en) * 2020-09-04 2020-12-01 中国空气动力研究与发展中心超高速空气动力研究所 Space debris impact damage interpretation method based on dynamic multi-target prediction
US20210383563A1 (en) * 2021-06-21 2021-12-09 University Of Electronic Science And Technology Of China Method for quantitatively identifying the defects of large-size composite material based on infrared image sequence
CN113409290A (en) * 2021-06-29 2021-09-17 北京兆维电子(集团)有限责任公司 Method and device for detecting appearance defects of liquid crystal display and storage medium
CN113409314A (en) * 2021-08-18 2021-09-17 南京市特种设备安全监督检验研究院 Unmanned aerial vehicle visual detection and evaluation method and system for corrosion of high-altitude steel structure
CN113538418A (en) * 2021-08-27 2021-10-22 浙江工业大学 Tire X-ray image defect extraction model construction method based on morphological analysis
CN113763367A (en) * 2021-09-13 2021-12-07 中国空气动力研究与发展中心超高速空气动力研究所 Comprehensive interpretation method for infrared detection characteristics of large-size test piece
CN113781445A (en) * 2021-09-13 2021-12-10 中国空气动力研究与发展中心超高速空气动力研究所 Multi-region complex damage defect feature extraction fusion method
CN113793318A (en) * 2021-09-13 2021-12-14 中国空气动力研究与发展中心超高速空气动力研究所 Multi-region complex damage defect characteristic comprehensive analysis method
CN113570605A (en) * 2021-09-28 2021-10-29 深圳市绘晶科技有限公司 Defect detection method and system based on liquid crystal display panel
CN113989224A (en) * 2021-10-27 2022-01-28 西安工程大学 Color texture fabric defect detection method based on generation countermeasure network
CN113887667A (en) * 2021-10-28 2022-01-04 中国人民解放军空军工程大学 Composite material defect detection method and system based on infrared and ultrasonic signal fusion
CN113887666A (en) * 2021-10-28 2022-01-04 中国人民解放军空军工程大学 Defect detection method and system for composite material based on Transformer

Non-Patent Citations (6)

* Cited by examiner, † Cited by third party
Title
G.LEI等: ""Using an Optimal Multi-Target Image Segmentation Based Feature Extraction Method to Detect Hypervelocity Impact Damage for Spacecraft"", 《IEEE SENSOR JOURNAL》, vol. 21, no. 18, pages 20258 - 20272, XP011877619, DOI: 10.1109/JSEN.2021.3092432 *
HUANG X等: ""Adaptive rapid defect identification in ECPT based on K-means and qutomatic segmentation algorithm"", 《JOURNAL OF AMBIENT INTELLIGENCE AND HUMANIZED COMPUTING》, pages 1 - 15 *
ZHANG H等: ""Design of Hypervelocity-Impact Damage Assessment Technique Based on Variational Bayesian"", 《IEEE ACCESS》, vol. 53, no. 2, pages 814 - 819 *
杨晓: ""基于红外热波技术的大尺寸复合材料缺陷检测研究"", 《中国优秀硕士学位论文全文数据库》, no. 1, pages 020 - 397 *
陈小伟等: ""超高速碰撞产生的碎片研究进展"", 《现代应用物理》, pages 2016 - 2047 *
韩航迪等: ""基于改进Tiny-YOLOv3网络的航天电子焊点缺陷主动红外检测研究"", 《仪器仪表学报》, vol. 41, no. 11, pages 42 - 49 *

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN115272301A (en) * 2022-09-20 2022-11-01 江苏新世嘉家纺高新科技股份有限公司 Automatic cheese defect detection method based on robot

Also Published As

Publication number Publication date
CN114549448B (en) 2023-08-11

Similar Documents

Publication Publication Date Title
CN111080629B (en) Method for detecting image splicing tampering
Yoon et al. LFIQ: Latent fingerprint image quality
EP1593094B1 (en) Image analysis for the purpose of assessing cancer
US20080013803A1 (en) Method and apparatus for determining print image quality
CN103400357B (en) A kind of method removing image salt-pepper noise
CN108932721A (en) A kind of infrared Image Segmentation and fusion method for crusing robot
CN114758219B (en) Trace identification method based on spectral data and infrared temperature data fusion
CN116188786B (en) Image segmentation system for hepatic duct and biliary tract calculus
CN114155202A (en) Thyroid nodule ultrasonic image classification method based on feature fusion and transfer learning
CN114549448A (en) Complex multi-type defect detection and evaluation method based on infrared thermal imaging data analysis
CN115937518A (en) Pavement disease identification method and system based on multi-source image fusion
Li et al. Automated image analysis of uterine cervical images
CN113537236B (en) Quantitative identification method for defect of thermal diffusion effect for infrared detection of spacecraft damage
CN109934102B (en) Finger vein identification method based on image super-resolution
CN106407975B (en) Multiple dimensioned layering object detection method based on space-optical spectrum structural constraint
CN113822361B (en) SAR image similarity measurement method and system based on Hamming distance
Luo et al. Rain-like layer removal from hot-rolled steel strip based on attentive dual residual generative adversarial network
Yang et al. Image Classification for Automobile Pipe Joints Surface Defect Detection Using Wavelet Decomposition and Convolutional Neural Network
Zuo et al. Global and local quality measures for NIR iris video
CN110211103A (en) Comentropy additivity based on infrared thermal imaging obscures defect characteristic and analyzes reconstructing method
Kang Vein pattern extraction based on vectorgrams of maximal intra-neighbor difference
CN110222740B (en) Infrared technology defect reconstruction and feature extraction method based on additive fuzzy
CN114529515A (en) Method for automatically identifying internal defects of solar cell
Nimkar et al. Fingerprint segmentation algorithms: A literature review
Elanangai et al. Automated system for defect identification and character recognition using IR images of SS-plates

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