A method for quantitatively evaluating multiple types of complex defects of a spacecraft

Abstract

The application discloses a kind of quantitative evaluation methods of multiple types complex defects of spacecraft, comprising: obtaining infrared reconstruction image characterized by defect feature, defect feature extraction is extracted to obtain defect feature extraction image;Obtain defect feature detection image;Obtain defect feature recognition image;Calibration is carried out to defect area, realize defect pixel rough positioning;Based on defect pixel rough positioning result, according to the defect boundary pixel point extracted, realize subpixel defect feature positioning;According to subpixel defect positioning result, to each calibrated defect area, contour feature size is in horizontal direction and vertical direction, according to the size relationship of specimen and image, it is respectively mapped to obtain the actual size of the different characteristic parameters of defect characterization;Defect feature quantization calculation is carried out.The application effectively improves the accuracy of defect positioning and quantization, improves the precision of defect positioning, solves the problem that horizontal direction and vertical direction pixel point pixel scale are different.

Description

Technical Field

[0001] This invention belongs to the field of damage detection and maintenance technology for aerospace vehicles. More specifically, this invention relates to a quantitative assessment method for multiple types of complex defects in spacecraft. Background Technology

[0002] To address the challenge of detecting various types of defects resulting from hypersonic impacts between space debris and spacecraft, numerous studies have been conducted on the quantitative assessment of defects using infrared thermal imaging (IRIM) technology. IRIM detection allows for the processing of infrared thermal image sequences to obtain reconstructed infrared images characterizing the defects, enabling preliminary determination of their morphological features through qualitative analysis. However, to ensure spacecraft operational safety, further quantitative assessment of these defect characteristics is necessary. Quantitative assessment results of the complex and diverse damage and defect characteristics of core spacecraft components can guide research on hypersonic impact damage, design of spacecraft protective structures, and spacecraft damage repair, thus ensuring the safe on-orbit operation of spacecraft.

[0003] Hypervelocity impact damage is characterized by its diversity, complexity, and wide distribution, posing challenges to defect detection and quantitative assessment. Furthermore, during the detection process, particulate matter splashed from the spacecraft surface and noise interference can affect the quantitative detection results. To address the difficulties in detecting hypervelocity impact damage and the objective influence of the testing environment, quantitative detection is divided into two parts: localization and quantification, to better identify the complex and multi-type hypervelocity impact damage. Localizing defects of different types and distributed in different locations is crucial for the identification and repair of complex and multi-type defect characteristics. Based on the localization results, the defect characteristic information of different types and regions is quantified, providing specific numerical references for the damage status of the specimen within six months. The morphological and physical characteristics of the complex and multi-type defects formed by hypervelocity impacts are described from both localization and quantification perspectives, and repair plans are formulated based on the characteristics of different damage types.

[0004] In the Chinese invention patent application published on June 1, 2021, with publication number CN112881467A and titled "A Method for Imaging and Quantitative Identification of Damage in Large-Size Composite Materials", the defect features in the infrared reconstructed image are extracted by segmentation during the quantitative detection of damage in large-size composite materials. The defect area is marked by the bounding rectangle, and the defect morphology features are then detected and quantified according to the marking results of the rectangle.

[0005] The method of using an circumscribed rectangle to delineate the defect area is effective for identifying and detecting defects in large-sized composite materials. However, when applied to the quantitative detection of ultra-high-speed impact damage, the types of ultra-high-speed impact damage are complex, and the defect area delineated by the circumscribed rectangle may deviate significantly from the actual defect area. This method can only provide a rough location of the defect, which affects the defect feature identification results and the efficiency of defect repair.

[0006] To address the limitation of only coarse localization using bounding boxes, the defect boundary contour can be extracted using boundary tracing. The defect boundary contour contains more detailed information than the bounding box, thus improving defect localization accuracy. However, since pixels have indivisible gaps, integer-pixel localization methods will exhibit significant errors when the defect boundary falls within these gaps. Furthermore, the discrete defect pixel contours become distorted when magnified, affecting defect region identification.

[0007] Furthermore, when quantifying the physical characteristics of defects, the results calculated based on the defect image are actually numerical features at the pixel scale. These need to be converted according to the ratio between the image size and the actual size of the test piece to obtain the actual dimensions. However, in actual calculations, because the image size and the actual size of the test piece differ, the pixel scale of pixels in the horizontal and vertical directions also differs. Therefore, there is no uniform ratio for calculating the actual physical information of the defects, resulting in some parameters that measure the physical characteristics of defects being unable to be calculated.

[0008] To address the aforementioned issues, this invention proposes a quantitative assessment method for complex defects of various types in spacecraft. Since pixels in infrared reconstructed images reflect continuous temperature changes and lack clear category classification, this invention employs a fuzzy C-means algorithm for segmentation. Considering the complexity of defect types and the impact of objectively existing thermal diffusion phenomena on the initially extracted defect features, post-processing is performed on the initially extracted defect feature images before quantitative detection, and region corrections are applied to defects of different types and locations. For the processed defect feature images, boundary features of different types and locations of damage are extracted to calibrate the defect regions, achieving coarse localization of defect pixels. Then, sub-pixel defect feature localization is achieved for each calibrated defect region using curve fitting. Before quantifying the feature parameters of each defect region, the fitted defect contours are size-mapped, and further quantified calculations are performed on features such as area, perimeter, major axis, and minor axis. Combining spatial location information and physical feature quantification information describes the defect features to meet the needs of quantitative identification of complex defects of various types. Summary of the Invention

[0009] One object of the present invention is to solve at least the above-mentioned problems and / or defects, and to provide at least the advantages described below.

[0010] To achieve these objectives and other advantages according to the present invention, a method for quantitatively evaluating multiple types of complex defects in spacecraft is provided, comprising the following steps: Step 1: Acquire infrared data, process the infrared thermal image sequence data, and obtain infrared reconstructed images characterizing defect features. Igs Furthermore, fuzzy C-clustering segmentation was used to reconstruct infrared images characterizing defect features. Igs Defect feature extraction is performed to obtain the defect feature extraction image. Ige ; Step 2: Extract image features based on membership degree. Ige Post-processing is performed to optimize the segmentation results and obtain defect feature detection images. Igt ; Step 3: Detect defect features in the image using the time-series alignment method. Igt Different types and locations of defects in the image are corrected to obtain defect feature recognition images. Igc ;; Step 4: Use 8-adjacent Freeman chain codes to detect the boundary features of different types and locations of defects in the defect feature recognition image, and calibrate the defect area to achieve coarse localization of defect pixels; Step 5: Based on the coarse localization results of defect pixels, sub-pixel defect feature localization is achieved for each labeled defect region using Bézier curve fitting based on the extracted defect boundary pixels. Step 6: Based on the subpixel defect localization results, for each calibrated defect area, the contour feature dimensions in the horizontal and vertical directions are mapped according to the size relationship between the specimen and the image to obtain the actual size of different feature parameters of the defect. Step 7: Perform defect feature quantification calculation.

[0011] Preferably, in step one, the specific steps for extracting defect features to obtain the defect feature extraction image include: Step 1.1: Set the number of clusters to [number]. C Initialize cluster centers v k , k =1,…, C Initialize the defect filter image Ige medium pixel x a Membership degree with cluster center v k membership degree u ak pixel x a membership degreeu ak satisfy ; Step 1.2: Calculate and update cluster centers v k : in, m For smoothing parameters, N For defect filtering images Igs The number of pixels in the image; Step 1.3: Update pixels x a Membership function u ak : Step 1.4: Calculate the current number. t The objective function at the next iteration J t : Step 1.5, Iteration Termination Condition Judgment: If the... t The objective function at the next iteration J t With the t The objective function at -1 iterations J t-1 The difference is less than the threshold for the iteration condition judgment. e ,Right now‖ J t - J t-1 ||< e Or reach the maximum number of iterations. T max The iteration terminates, yielding the final set of cluster centers. v k , k =1,…, C and defect-filtered images Igs Membership matrix of each pixel u ak ; Step 1.6, Defect Filtering Image Igs pixel x a Divided into membership matrix u ak The class with the largest value, defect-filtered image Igs All pixels x a The final defect feature extraction image is obtained after the segmentation is completed. Ige .

[0012] Preferably, step two includes the following specific steps: Step 2.1: Detect all possible misclassified pixels; detect all pixels. x a of r × r In the neighborhood region, the classification labels of each neighboring pixel are as follows: if the classification labels are the same, the original classification labels are kept unchanged; otherwise, pixels with inconsistent labels are considered to be misclassified and are extracted and saved. Step 2.2: Relabel the extracted potentially misclassified pixels; x e ,statistics x e of n × n Pixels in the neighborhood region x p Membership degree, recalculate pixel points x e Belongs to the k Membership degree of a class u ek ′: in, Oh e Represents pixels x e of r × r The set of neighboring pixels; express x e The neighboring pixels belong to the first k Membership degree of a class; C e express Oh e The number of pixels, according to u ek The calculation of ′ will x e The pixel is classified into the category with the highest membership value, and the pixel label is reset. It is important to note that when resetting pixel labels, the label being reset must be the same as the one used in step 2.2. r × r Classification labels appearing in the neighborhood region; image for extracting defect features. Ige The defect feature detection image is obtained by classifying and relabeling all possible defective pixels. Igt .

[0013] Preferably, in step three, the defect feature detection image obtained in step two is... Igt Divided into K The region, for the first k One defective area, k =1,…, K The size of the heat diffusion region is obtained by measuring the similarity of pixels in the defective region. in, In the i-th defect region, pixel k x i The distance between the reference point and the actual area of ​​the defect feature; For the first k The number of pixels in each defective region; For the first k The maximum distance between a pixel in a defect region and a reference point in the actual defect feature region; k e ttr_i For defective pixels k x i The thermal diffusion determination threshold is calculated by separately measuring the temperature thermal response correlation between the defect region pixels and the actual defect feature reference point, and the temperature thermal response correlation between the defect region pixels and the background region reference point. k Each defective region pixel k x i thermal diffusion judgment threshold k e ttr_i Taking into account the objectively existing thermal diffusion phenomenon, the defect area is corrected to obtain a defect feature recognition image. Igc The specific steps are as follows: Step 3.1: Calculate the first step using the timing alignment method. k Each defective region pixel k x i thermal diffusion judgment threshold k e ttr_i ; Step 3.1.1: Calculate the pixels in the defect area. k x i The correlation coefficient between the transient thermal response sequence and the background reference point; Step 3.1.1.1: Determine the transient thermal response sequence that needs to be time-shifted; select the pixel corresponding to the characteristic transient thermal response of the background region as the background region reference point. The corresponding temperature thermal response sequence is , T For thermal sequence acquisition time; for defective region pixels. k x i Transient thermal response sequence Find the reference points for the background areas respectively. The frame number where the temperature peak is located and defective area pixels k x i The frame number where the temperature peak is located k t 1; if ,but It is a transient thermal response sequence that needs to be moved, denoted as ttr 2. Remember for ttr 1; otherwise, It is a transient thermal response sequence that needs to be moved, denoted as ttr 1. Remember for ttr 1; Step 3.1.1.2, record ttr The frame number where the temperature peak is 1 is t max Set each move tt Time units, loop parameters t m Set to 1, there is ;right ttr 2. Perform time-series shifts, each shift... tt Each time unit will ttr subsequence of 2 and ttr subsequence of 1 Align and calculate their correlation coefficients. : in, E (·) represents the expectation of the transient thermal response sequence; Step 3.1.1.3 t m = t m +1, repeat step 3.1.1.2, calculating the corresponding Pearson correlation coefficient for each time interval, until... Until then; the set of correlation coefficients is obtained. ; Step 3.1.1.4, from Find the maximum value as the first k Each defective region pixel k x i Reference point with background area Temperature thermal response sequence correlation coefficient k r i&bg : Step 3.1.2: Following the same calculation method as in Step 3.1.1, calculate the... k The correlation coefficient between the temperature thermal response of each pixel in the defect region and the actual reference point of the defect feature region. k r i&bg According to the correlation coefficient k r i&bg and k r i&def Get the first k Each defective region pixel x i Threshold for determining the thermal diffusion region k e ttr_i : Step 3.2: Determine the threshold based on the calculated thermal diffusion region. k e ttr_i When the distance between the defect pixel and the reference point of the actual defect feature area satisfy: Consider pixels x i If the pixel is located in the heat diffusion area, remove it. x i Repeat step 3.2 until the defect feature recognition image is obtained. Igc The Middle k Each defective region pixel x i All pixels were accessed; remove all pixels located in the heat diffusion area. x i Proceed to step 3.3; Step 3.3, repeat steps 3.1 and 3.2 until... K Each pixel in the defect region was accessed, resulting in a defect feature recognition image. Igc .

[0014] Preferably, step four involves detecting defect feature recognition images using 8-adjacent Freeman chain codes. Igc By analyzing the boundary features of defects of different types and locations, the defect region is calibrated to achieve coarse localization of defect pixels. Specifically, for the first... k , k =1,…, K The specific process for pixel calibration and localization of each defect area is as follows: Select the k The starting pixel for boundary tracking of each defect region is denoted as... k E 0, with chaincode value dir The direction of =0 is the starting direction of the boundary search. Rotate 45° counterclockwise, following the direction... dir =1 to search for the next pixel; since the extracted defect feature image is a binary image, the pixel value representing the defect contour pixel is represented by "1" in this binary image; based on this, during the search for defect boundary pixels, if the pixel value is 1... k x t If the pixel value is "1", then the pixel is considered to be "1". k x t It is a defect boundary point, denoted as k E t And save; and k x t The chain code value is assigned to the pixel preceding it. k x t-1 ; to pixels k x t As the new starting point for the boundary search, rotate 90° counterclockwise, following the direction... dir =2 Search for the next defect boundary point; repeat the above steps until the starting point is found. k b At time 0, the defect area search was completed; during the entire search process, the detected boundary points... k E 0,…, k E M This refers to the boundary profile of the defect; the set of boundary points. k Oh L , k Oh L =( k E0,…, k E M The defects were coarsely located pixel by pixel; the above process was repeated until all defect areas were visited, resulting in... K Pixel calibration and localization results for each defect area.

[0015] Preferably, step five specifically includes: after coarse pixel positioning, calibration is obtained. K The first defect area was extracted, and the calibrated first defect area was extracted. k Defective areas M Each defect boundary pixel k =1,…, K , M The defect boundary pixels constitute a pixel set. k E i , i =0,1,…, M For adjacent defect boundary pixels respectively k E i and pixels k E i+1 Perform cubic Bézier curve fitting pairwise; because i The maximum value is M Therefore, pixel k E M+1 It does not exist; considering that the defect contour is a closed figure, we use... k E 0 as k E M+1 and k E M Perform cubic Bézier curve fitting; use the cubic Bézier curve to analyze the defect boundary pixels. k E i and k E i+1 The specific process of obtaining the sub-pixel defect closed contour by pairwise fitting is as follows: Step 5.1: For the edge contour pixels to be fitted... k E i and k E i+1 Construct control points between them. k K i and kC i : in, k a * and k b * It is any given positive number; k x i , k y i () represents the pixel at the defect boundary. k E i The positional parameters; due to the boundary pixel points of the defect detected by the boundary tracking algorithm. k E i ( k x i , k y i ), i =0,1,…, M Therefore, there are no pixels in the set of defect boundary pixels. k E -1 , k E M+1 , k E M+2 Its corresponding position parameter ( k x -1 , k y -1 ), ( k x M+1 , k y M+1 ), ( k x M+2 , k y M+2 Therefore, it does not exist; considering that the defect profile is closed, ( k x M , k y M The value of ) is used as ( kx -1 , k y -1 The value of ), ( k x 0, k y The value of 0) is used as ( k x M+1 , k y M+1 The value of ), ( k x 1, k y The value of 1) is used as ( k x M+2 , k y M+2 The value of ); Step 5.2: According to the definition of the Bézier curve, the defect boundary pixels... k E i and k E i+1 Expression of the cubic Bézier curve between for: The defect boundary pixels are obtained based on the expression for the cubic Bézier curve. k E i and pixels k E i+1 Fitting curves between; Step 5.3: Traverse the set of defect boundary pixels k E i All the pixels, i =0,1,…, M Pairwise fitting is performed to obtain the fitting curve between adjacent pixels; when i = M When, use k E 0 as k E M+1 and k E M Perform cubic Bézier curve fitting; Step 5.4: After traversing all defect boundary pixels, the result is obtained. MA cubic Bézier curve; M The first Bézier curve is smoothly connected to obtain the second Bézier curve. k Subpixel defect closed contour of each defect region k L Based on closed contours k L Achieve subpixel localization of defects; Step 5.5: Repeat the above process until... K All defect regions were accessed, and sub-pixel localization results of defects of different types and locations were obtained.

[0016] Preferably, the method for mapping the actual dimensions of different feature parameters of the inscribed defect in step six includes: [the method is described in the original text, but the provided text is incomplete and requires further context to be accurately translated.] k One defective area, k =1,…, K The coordinates of the sub-pixel boundary points are ( k x s , k y s The mapping is performed according to the following formula: In the formula, L × W The actual dimensions of the test specimen; M × N Given the size of the captured infrared image, obtain the mapped subpixel boundary point coordinates. k x s ′, k y s ′).

[0017] Preferably, the specific steps of step seven, which involves quantifying defect features, include: perimeter and area are conventional and directly reflect the size information of defects; based on the numerical values ​​of geometric structural feature parameters, the size information of each defect feature region can be directly quantified and evaluated, thereby reflecting the geometric size differences between each defect feature region; the area of ​​the defect can be used to quantify the size of the defect in the image; feature parameters of different defect regions are calculated separately; and the calibrated first... k One defective area, k =1,…, K For the mapped sub-pixel boundary points ( k x s ′, k ys The boundary contour of the defect in the actual size drawn by the set is denoted as ′) k L Since most defect feature areas are irregular, the defect boundary contour... k L The shape is irregular; the defect boundary profile is calculated directly. k L Determining the area of ​​the enclosed defect region is difficult; therefore, the defect boundary profile is used. k L Enclosed area k Oh pixels k p i The number of samples and the area feature parameters are used; since boundary size mapping was performed in step seven, area conversion between the specimen and the image is no longer required, and the defect boundary contour is statistically obtained at this point. k L Enclosed area k Oh The number of pixels is the area feature parameter of the defect: By statistically analyzing the area enclosed by the defect contour, the location information of the defect can be further determined, thus achieving the purpose of defect calibration. In the mapped defect contour image, the perimeter, major axis, and minor axis features of the defect are further calculated based on the defect boundary contour features. When calculating the perimeter of the defect, the fitted sub-pixel contour is obtained by sequentially fitting the defect boundary pixels one by one. k E i and k E i+1 The defect boundary feature size mapping does not change the positional relationship between sub-pixel boundary points, so the boundary pixel points can be obtained through curve integration. k E i and k E i+1 The actual length of the fitted curve between k P i : in, For defect boundary pixels k E i and k E i+1The expression for the cubic Bézier curve between; M The sum of the actual fitted curve lengths between each pair of boundary pixels is the perimeter of the defect. k P : Meanwhile, due to the irregularity of the defect feature region, the size of the minimum bounding rectangle obtained based on the sub-pixel defect boundary contour can also reflect the morphological characteristics of the defect; based on the fitted contour of the sub-pixel defect boundary contour... k L Find its minimum bounding rectangle, and use the minimum bounding rectangle to characterize the distribution of defects; the longer side of the minimum bounding rectangle is the major axis of the defect. k Dl The shorter side is the minor diameter of the defect. k Ds The minimum bounding rectangle of the subpixel defect boundary contour is determined using the area-based equally spaced rotation method: the subpixel defect boundary contour image is rotated equally within a 90° range, and the parameters of the bounding rectangle of the subpixel defect boundary contour are recorded for each rotation; the maximum horizontal coordinate of the subpixel defect boundary contour is then found. k x max and minimum value k x min The maximum value of the vertical coordinate. k y max and minimum value k y min Then the marked number k The longer side of the bounding rectangle of the defect region is: k Dl =max( k y max - k y min , k x max - k x min ) The shorter side is: k Ds =min( k y max - k y min ,k x max - k x min ); The area of ​​the circumscribed rectangle is k Dl × k Ds This invention defines the minimum bounding rectangle by area. The specific process of calculating the minimum bounding rectangle is as follows: First, based on the maximum and minimum values ​​of the sub-pixel defect boundary contour distribution coordinates, a bounding rectangle of the sub-pixel defect boundary contour region without rotation angle can be obtained. Then, the long side, short side, and area of ​​the minimum bounding rectangle in the current state are calculated, and the rotation angle is determined at this time. α =0°; Set equal-interval rotation angles i Rotate the subpixel defect boundary contour by an angle. i , α = α + i The bounding rectangle of the subpixel defect boundary contour is obtained in the same way, and the relevant parameters are recorded; the rotation process is repeated until the rotation angle is reached. α =90°; Finally, determine the minimum bounding rectangle based on the area size; Based on the above quantization process, calculate the physical characteristic parameters of the defect based on the sub-pixel defect boundary contour fitting results.

[0018] The present invention has at least the following beneficial effects: (1) This invention proposes a quantitative evaluation method for complex defects of various types in spacecraft. The quantitative detection is divided into two stages: localization and quantization. Defect characteristics are described by combining spatial location information and physical feature quantification information. During localization, considering the complexity of defect types and the interference of noise with the initial feature extraction results, a reclassification method based on membership degree fuzzy clustering is used to reclassify pixels that may be misclassified. When using infrared detection technology for damage detection, since thermal diffusion is an objective phenomenon, to eliminate the influence of thermal diffusion on defect identification, the correlation coefficients between the transient thermal response of pixels in the defect area and the reference transient thermal response of the defect area, and the correlation coefficients between the reference transient thermal response of the background area are calculated. The diffusion area judgment threshold is automatically determined based on the correlation coefficients, and the thermal diffusion area is removed using the judgment threshold. Based on the area correction results, defect boundary features are extracted using a boundary tracking method to locate defect pixels. Then, a method based on cubic Bézier curve subpixel contour fitting is used to accurately locate the defect. In the quantization section, the problem of different pixel scales is solved by mapping the feature size of the defect boundary. The area, perimeter, major axis and minor axis of the defect are calculated to describe the defect features from multiple aspects.

[0019] (2) The present invention corrects the defect region by post-processing and temporal alignment of defect feature extraction images through membership degree. It can optimize the defect feature extraction results based on segmentation method and remove the influence of the objectively existing thermal diffusion phenomenon in infrared detection technology, effectively improving the accuracy of defect localization and quantification.

[0020] (3) This invention locates the defect region by fitting the defect contour to subpixels. Subpixels can subdivide the microscopic gaps between pixels, improving the accuracy of defect location. Accurate defect location results can not only help to quickly find the defect shape and location, but also guide the secondary local detection of small defects, thereby improving the defect detection efficiency. Furthermore, the fitted defect contour is continuous and smooth. When the pixel blocks are magnified, they will be distorted, but the fitted curve still remains continuous and has stable defect characteristics.

[0021] (4) Before quantifying the defect features, this invention performs defect boundary feature size mapping, which solves the problem of different pixel scales. After mapping, the pixel scale in the horizontal and vertical directions is the same as the actual scale. The parameter size obtained by quantitative calculation is the actual size of the defect feature. Combined with different geometric feature information, the global features of the defect are characterized, providing a specific numerical reference for the damage of the specimen.

[0022] Other advantages, objectives and features of the present invention will become apparent in part from the following description, and in part from those skilled in the art through study and practice of the invention. Attached Figure Description

[0023] Figure 1 A flowchart illustrating the quantitative evaluation method for multiple types of complex defects in spacecraft provided by this invention; Figure 2 This is an infrared reconstructed image characterizing defect features in an embodiment of the present invention; Figure 3 This is an image for extracting defect features in an embodiment of the present invention; Figure 4 The image shown is a defect feature detection image in an embodiment of the present invention. Figure 5 This is a defect region marking image in an embodiment of the present invention; Figure 6 This is a defect feature recognition image in an embodiment of the present invention; Figure 7 This is an image of the defect region boundary extracted in an embodiment of the present invention; Figure 8 This is a subpixel defect region localization image in an embodiment of the present invention. Detailed Implementation

[0024] The present invention will now be described in further detail with reference to the accompanying drawings, so that those skilled in the art can implement it based on the description.

[0025] It should be understood that terms such as “having,” “comprising,” and “including” as used herein do not exclude the presence or addition of one or more other elements or combinations thereof.

[0026] The purpose of this invention is to overcome the shortcomings of existing technologies and provide a quantitative assessment method for multiple types of complex defects in spacecraft. Before quantitatively calculating defect features, segmentation is used to extract defect features. Considering that complex and multi-type defects can affect the accuracy of feature extraction and that the objectively existing thermal diffusion effect can interfere with the quantitative detection results, post-processing is performed on the initially extracted defect feature images, and then corrections are made for defect regions of different types and locations. Based on previous quantitative detection of damage to large-size composite materials, to address the quantitative detection requirements of multiple types of complex defects formed by hypervelocity impacts on spacecraft, defect features of different types and damage are first identified by extracting defect boundaries. Then, based on pixel localization results, sub-pixel defect feature localization is achieved for each identified defect region using curve fitting. Size mapping is performed on the fitted sub-pixel contours, and then the defect feature size parameters are quantified. Defect features are described by combining spatial location information and physical feature quantification information to meet the needs of quantitative identification of multiple types of complex defects.

[0027] like Figure 1 As shown, to achieve the above-mentioned objective, this invention proposes a quantitative evaluation method for multiple types of complex defects in spacecraft, characterized by the following steps: Step 1: Acquire infrared data, process the infrared thermal image sequence data, and obtain infrared reconstructed images characterizing defect features. Igs Furthermore, fuzzy C-clustering segmentation was used to reconstruct infrared images characterizing defect features. Igs Defect feature extraction is performed to obtain the defect feature extraction image. Ige This addresses the issue that infrared reconstructed images reflect temperature changes, and the division between regions is not clear, making them unusable directly by vision systems. The specific steps for defect feature extraction are as follows: Step 1.1: Set the number of clusters to [number]. C Initialize cluster centers v k , k =1,…, C Initialize the defect filter image Ige medium pixel x a Membership degree with cluster center v k membership degreeu ak pixel x a membership degree u ak satisfy ; Step 1.2: Calculate and update cluster centers v k : in, m For smoothing parameters, N For defect filtering images Igs The number of pixels in the image; Step 1.3: Update pixels x a Membership function u ak : Step 1.4: Calculate the current number. t The objective function at the next iteration J t : Step 1.5, Iteration Termination Condition Judgment: If the... t The objective function at the next iteration J t With the t The objective function at -1 iterations J t-1 The difference is less than the threshold for the iteration condition judgment. e ,Right now‖ J t - J t-1 ||< e Or reach the maximum number of iterations. T max The iteration terminates, yielding the final set of cluster centers. v k , k =1,…, C and defect-filtered images Igs Membership matrix of each pixel u ak ; Step 1.6, Defect Filtering Image Igs pixel x a Divided into membership matrix u ak The class with the largest value, defect-filtered image Igs All pixelsx a The final defect feature extraction image is obtained after the segmentation is completed. Ige .

[0028] Step 2: Extract image features based on membership degree. Ige Post-processing is performed to optimize the segmentation results and obtain defect feature detection images. Igt The specific steps are as follows: Step 2.1: Detect all possible misclassified pixels; detect all pixels. x a of r × r In the neighborhood region, the classification labels of each neighboring pixel are as follows: if the classification labels are the same, the original classification labels are kept unchanged; otherwise, pixels with inconsistent labels are considered to be misclassified and are extracted and saved. Step 2.2: Relabel the extracted potentially misclassified pixels; x e ,statistics x e of n × n Pixels in the neighborhood region x p Membership degree, recalculate pixel points x e Belongs to the k Membership degree of a class u ek ′: in, Oh e Represents pixels x e of r × r The set of neighboring pixels; express x e The neighboring pixels belong to the first k Membership degree of a class; C e express Oh e The number of pixels, according to u ek The calculation of ′ will x e The pixel is classified into the category with the highest membership value, and the pixel label is reset. It is important to note that when resetting pixel labels, the label being reset must be the same as the one used in step 2.2. r × rClassification labels appearing in the neighborhood region; image for extracting defect features. Ige The defect feature detection image is obtained by classifying and relabeling all possible defective pixels. Igt .

[0029] Step 3: Detect defect features in the image using the time-series alignment method. Igt Different types and locations of defects in the image are corrected to obtain defect feature recognition images. Ige After feature extraction, the defects were divided into... K The region, for the first k One defective area, k =1,…, K The size of the heat diffusion region is obtained by measuring the similarity of pixels in the defective region. in, Indicates the first k In the defective area, pixel k x i The distance between the reference point and the actual area of ​​the defect feature; For the first k The number of pixels in each defective region; For the first k The maximum distance between a pixel in a defect region and a reference point in the actual defect feature region; k e ttr_i For defective pixels k x i The thermal diffusion determination threshold is calculated by separately measuring the temperature thermal response correlation between the defect region pixels and the actual defect feature reference point, and the temperature thermal response correlation between the defect region pixels and the background region reference point. k Each defective region pixel k x i thermal diffusion judgment threshold k e ttr_i Considering the objectively existing thermal diffusion phenomenon, corrections are made to the defect area to obtain a defect feature recognition image. Igc The specific steps are as follows: Step 3.1: Calculate the first step using the timing alignment method. k Each defective region pixel k x i thermal diffusion judgment threshold k ettr_i ; Step 3.1.1: Calculate the pixels in the defect area. k x i The correlation coefficient between the transient thermal response sequence and the background reference point; Step 3.1.1.1: Determine the transient thermal response sequence that needs to be time-shifted; select the pixel corresponding to the characteristic transient thermal response of the background region as the background region reference point. The corresponding temperature thermal response sequence is , T For thermal sequence acquisition time; for defective region pixels. k x i Transient thermal response sequence Find the reference points for the background areas respectively. The frame number where the temperature peak is located and defective area pixels k x i The frame number where the temperature peak is located k t 1; if ,but It is a transient thermal response sequence that needs to be moved, denoted as ttr 2. Remember for ttr 1; otherwise, It is a transient thermal response sequence that needs to be moved, denoted as ttr 1. Remember for ttr 1; Step 3.1.1.2, record ttr The frame number where the temperature peak is 1 is t max Set each move tt Time units, loop parameters t m Set to 1, there is ;right ttr 2. Perform time-series shifts, each shift... tt Each time unit will ttr subsequence of 2 and ttr subsequence of 1 Align and calculate their correlation coefficients. : in, E (·) represents the expectation of the transient thermal response sequence; Step 3.1.1.3 tm = t m +1, repeat step 3.1.1.2, calculating the corresponding Pearson correlation coefficient for each time interval, until... Until then; the set of correlation coefficients is obtained. ; Step 3.1.1.4, from Find the maximum value as the first k Each defective region pixel k x i Reference point with background area Temperature thermal response sequence correlation coefficient k r i&bg : Step 3.1.2: Following the same calculation method as in Step 3.1.1, calculate the... k The correlation coefficient between the temperature thermal response of each pixel in the defect region and the actual reference point of the defect feature region. k r i&bg According to the correlation coefficient k r i&bg and k r i&def Get the first k Each defective region pixel x i Threshold for determining the thermal diffusion region k e ttr_i : Step 3.2: Determine the threshold based on the calculated thermal diffusion region. k e ttr_i When the distance between the defect pixel and the reference point of the actual defect feature area satisfy: Consider pixels x i If the pixel is located in the heat diffusion area, remove it. x i Repeat step 3.2 until the defect feature recognition image is obtained. Igc The Middle k Each defective region pixel x i All pixels were accessed; remove all pixels located in the heat diffusion area. xi Proceed to step 3.3; Step 3.3, repeat steps 3.1 and 3.2 until... K Each pixel in the defect region was accessed, resulting in a defect feature recognition image. Igc .

[0030] Step 4: Detect defect features in the image using 8-adjacency Freeman chain codes. Igc By analyzing the boundary features of defects of different types and locations, the defect region is calibrated to achieve coarse localization of defect pixels. For the first... k , k =1,…, K The specific process for pixel calibration and localization of each defect area is as follows: Select the k The starting pixel for boundary tracking of each defect region is denoted as... k E 0, with chaincode value dir The direction of =0 is the starting direction of the boundary search. Rotate 45° counterclockwise, following the direction... dir =1 to search for the next pixel; since the extracted defect feature image is a binary image, the pixel value representing the defect contour pixel is represented by "1" in this binary image; based on this, during the search for defect boundary pixels, if the pixel value is 1... k x t If the pixel value is "1", then the pixel is considered to be "1". k x t It is a defect boundary point, denoted as k E t And save; and k x t The chain code value is assigned to the pixel preceding it. k x t-1 ; to pixels k x t As the new starting point for the boundary search, rotate 90° counterclockwise, following the direction... dir =2 Search for the next defect boundary point; repeat the above steps until the starting point is found. k b At time 0, the defect area search was completed; during the entire search process, the detected boundary points... k E 0,…, k E M This refers to the boundary profile of the defect; the set of boundary points.k Oh L , k Oh L =( k E 0,…, k E M The defects were coarsely located pixel by pixel; the above process was repeated until all defect areas were visited, resulting in... K Pixel calibration and localization results for each defect area.

[0031] Step 5: Based on the coarse localization results of defect pixels, sub-pixel defect feature localization is achieved for each calibrated defect region using Bézier curve fitting based on the extracted defect boundary pixels.

[0032] After coarse pixel localization, calibration was obtained. K The first defect area was extracted, and the calibrated first defect area was extracted. k Defective areas M Each defect boundary pixel k =1,…, K , M The defect boundary pixels constitute a pixel set. k E i , i =0,1,…, M For adjacent defect boundary pixels respectively k E i and pixels k E i+1 Perform cubic Bézier curve fitting pairwise; because i The maximum value is M Therefore, pixel k E M+1 It does not exist; considering that the defect contour is a closed figure, we use... k E 0 as k E M+1 and k E M Perform cubic Bézier curve fitting; use the cubic Bézier curve to analyze the defect boundary pixels. k E i and k E i+1 The specific process of obtaining the sub-pixel defect closed contour by pairwise fitting is as follows: Step 5.1: For the edge contour pixels to be fitted... k E i and k E i+1 Construct control points between them. k K i and k C i : in, k a * and k b * It is any given positive number; k x i , k y i () represents the pixel at the defect boundary. k E i The positional parameters; due to the boundary pixel points of the defect detected by the boundary tracking algorithm. k E i ( k x i , k y i ), i =0,1,…, M Therefore, there are no pixels in the set of defect boundary pixels. k E -1 , k E M+1 , k E M+2 Its corresponding position parameter ( k x -1 , k y -1 ), ( k x M+1 , k y M+1 ), ( k xM+2 , k y M+2 Therefore, it does not exist; considering that the defect profile is closed, ( k x M , k y M The value of ) is used as ( k x -1 , k y -1 The value of ), ( k x 0, k y The value of 0) is used as ( k x M+1 , k y M+1 The value of ), ( k x 1, k y The value of 1) is used as ( k x M+2 , k y M+2 The value of ); Step 5.2: According to the definition of the Bézier curve, the defect boundary pixels... k E i and k E i+1 Expression of the cubic Bézier curve between for: The defect boundary pixels are obtained based on the expression for the cubic Bézier curve. k E i and pixels k E i+1 Fitting curves between; Step 5.3: Traverse the set of defect boundary pixels k E i All the pixels, i =0,1,…, M Pairwise fitting is performed to obtain the fitting curve between adjacent pixels; when i = M When, usek E 0 as k E M+1 and k E M Perform cubic Bézier curve fitting; Step 5.4: After traversing all defect boundary pixels, the result is obtained. M A cubic Bézier curve; M The first Bézier curve is smoothly connected to obtain the second Bézier curve. k Subpixel defect closed contour of each defect region k L Based on closed contours k L Achieve subpixel localization of defects; Step 5.5: Repeat the above process until... K All defect regions were accessed, and sub-pixel localization results of defects of different types and locations were obtained.

[0033] Step Six: Based on the sub-pixel defect localization results, for each calibrated defect region, map the contour feature dimensions in both the horizontal and vertical directions according to the size relationship between the specimen and the image to obtain the actual dimensions of different feature parameters characterizing the defect. For the calibrated... k One defective area, k =1,…, K The coordinates of the sub-pixel boundary points are ( k x s , k y s The mapping is performed according to the following formula: In the formula, L × W The actual dimensions of the test specimen; M × N Given the size of the captured infrared image, obtain the mapped subpixel boundary point coordinates. k x s ′, k y s ′).

[0034] Step 7: Defect Feature Quantification Calculation. Perimeter and area are conventional and direct indicators of defect size. Based on the numerical values ​​of geometric structural feature parameters, the size information of each defect feature region can be directly quantified and evaluated, thus reflecting the geometric size differences between each defect feature region. The area of ​​the defect can be used to quantify the size of the defect in the image. The feature parameters of different defect regions are calculated separately. The calibrated first... k One defective area, k =1,…, K For the mapped sub-pixel boundary points ( k x s ′, k y s The boundary contour of the defect in the actual size drawn by the set is denoted as ′) k L Since most defect feature areas are irregular, the defect boundary contour... k L The shape is irregular; the defect boundary profile is calculated directly. k L Determining the area of ​​the enclosed defect region is difficult; therefore, the defect boundary profile is used. k L Enclosed area k Oh pixels k p i The number of samples and the area feature parameters are used; since boundary size mapping was performed in step seven, area conversion between the specimen and the image is no longer required, and the defect boundary contour is statistically obtained at this point. k L Enclosed area k Oh The number of pixels is the area feature parameter of the defect: By statistically analyzing the area enclosed by the defect contour, the location information of the defect can be further determined, thus achieving the purpose of defect calibration. In the mapped defect contour image, the perimeter, major axis, and minor axis features of the defect are further calculated based on the defect boundary contour features. When calculating the perimeter of the defect, the fitted sub-pixel contour is obtained by sequentially fitting the defect boundary pixels one by one. k E i and k E i+1 The defect boundary feature size mapping does not change the positional relationship between sub-pixel boundary points, so the boundary pixel points can be obtained through curve integration. kE i and k E i+1 The actual length of the fitted curve between k P i : in, For defect boundary pixels k E i and k E i+1 The expression for the cubic Bézier curve between; M The sum of the actual fitted curve lengths between each pair of boundary pixels is the perimeter of the defect. k P : Meanwhile, due to the irregularity of the defect feature region, the size of the minimum bounding rectangle obtained based on the sub-pixel defect boundary contour can also reflect the morphological characteristics of the defect; based on the fitted contour of the sub-pixel defect boundary contour... k L Find its minimum bounding rectangle, and use the minimum bounding rectangle to characterize the distribution of defects; the longer side of the minimum bounding rectangle is the major axis of the defect. k Dl The shorter side is the minor diameter of the defect. k Ds The minimum bounding rectangle of the subpixel defect boundary contour is determined using the area-based equally spaced rotation method: the subpixel defect boundary contour image is rotated equally within a 90° range, and the parameters of the bounding rectangle of the subpixel defect boundary contour are recorded for each rotation; the maximum horizontal coordinate of the subpixel defect boundary contour is then found. k x max and minimum value k x min The maximum value of the vertical coordinate. k y max and minimum value k y min Then the marked number k The longer side of the bounding rectangle of the defect region is: k Dl =max( k y max - ky min , k x max - k x min ) The shorter side is: k Ds =min( k y max - k y min , k x max - k x min ); The area of ​​the circumscribed rectangle is k Dl × k Ds This invention defines the minimum bounding rectangle by area. The specific process of calculating the minimum bounding rectangle is as follows: First, based on the maximum and minimum values ​​of the sub-pixel defect boundary contour distribution coordinates, a bounding rectangle of the sub-pixel defect boundary contour region without rotation angle can be obtained. Then, the long side, short side, and area of ​​the minimum bounding rectangle in the current state are calculated, and the rotation angle is determined at this time. α =0°; Set equal-interval rotation angles i Rotate the subpixel defect boundary contour by an angle. i , α = α + i The bounding rectangle of the subpixel defect boundary contour is obtained in the same way, and the relevant parameters are recorded; the rotation process is repeated until the rotation angle is reached. α =90°; Finally, determine the minimum bounding rectangle based on the area size; Based on the above quantization process, calculate the physical characteristic parameters of the defect based on the sub-pixel defect boundary contour fitting results.

[0035] Simulation verification: To demonstrate the feasibility and correctness of this invention, the technical solution of this invention will be analyzed and explained below through experimental verification.

[0036] In this embodiment, a high-temperature ceramic experimental specimen with dimensions of 200mm × 200mm × 5mm was selected. Two different types of cylindrical damage were embedded in the subsurface of the material. The lower left corner shows the aperture. f The first defect is a 3mm layered defect; the remaining cylinders are inclusions filled with materials whose thermal conductivity differs from that of the high-temperature ceramic material. The upper right corner shows the aperture. f The diameter is 5mm, and the other two are internal inclusion defects. f All were 5mm. The infrared camera used for data acquisition had a resolution of 512×640, and 99 frames of infrared data were acquired throughout the experiment. The infrared data was stored in data blocks of size 512×640×99. By processing the infrared thermal image sequence, the reconstructed infrared images characterizing the defects were obtained, as shown below. Figure 2 As shown.

[0037] Defect features in the reconstructed infrared image are extracted using fuzzy C-cluster segmentation. The defect feature extraction image is shown below. Figure 3 As shown.

[0038] In the process of further quantitatively identifying defect features, considering the complexity of defect types and the presence of noise, pixels in the defect edge region may be incorrectly segmented. More accurate defect region segmentation can improve recognition accuracy. Therefore, preprocessing was performed on the defect feature-extracted image, and the membership degrees of edge pixels were recalculated. Based on these membership degrees, the edge pixels of the defect region were re-segmented. In the post-processing process, parameters were set... r =3, n =5, after post-processing the defect feature extraction image, the defect feature detection image is obtained as follows: Figure 4 As shown.

[0039] Based on the pixel division before and after post-processing, post-processing can re-divide the defect edge pixels, and the re-division result is consistent with the actual defect feature representation, demonstrating the effectiveness of the post-processing scheme. Due to the objective existence of thermal diffusion effects, the defect feature image obtained through feature extraction actually includes both the defect region and the thermal diffusion region, and the post-processing process cannot remove the influence of the thermal diffusion region. Considering the unknown degree of thermal diffusion, a comparative analysis was conducted on the transient thermal response curves of pixels located in the defect region. The defect region was marked, and the marking results are as follows: Figure 5 As shown, the heat diffusion area of ​​each defect region is then removed. For each defect region, the center pixel of the defect region is used as the reference point of the actual defect feature region. The distances between the remaining pixels and the reference point are calculated, and the maximum distance is found. The maximum similarity between the transient thermal response curves of each pixel in the defect region and the reference curve in the background region was calculated. r i&bg and the maximum similarity with the reference curve of the defect area r i&def Determine its threshold e ttr_i Partial thresholds for each region e ttr_i Calculation results and maximum distance The calculation results are shown in Table 1.

[0040] Table 1. Calculation results of some parameters for determining the thermal diffusion zone of each defect area. Based on the thermal diffusion region determination parameters, the thermal diffusion region in the post-processed defect feature extraction image is removed, resulting in... Figure 6 The defect feature recognition image is shown. Table 2 shows a comparison of the number of pixels in each defect region before and after heat diffusion region removal.

[0041] Table 2 Comparison of pixel count in each defect region before and after thermal diffusion removal As can be seen from the comparison results in Table 2, the method designed in this invention for calculating the maximum correlation coefficient by arranging time sequences can indeed achieve the purpose of correcting defective areas.

[0042] Edge contour information of the defect feature recognition image is extracted, and then boundary tracking is performed. The set of boundary pixels is used as the coarse localization result of the defect pixels. Pixels located at (123, 199) in defect region A, (427, 218) in defect region B, (428, 421) in defect region C, and (121, 411) in defect region D are selected as the starting points for boundary tracking of each defect region, resulting in the defect region boundary extraction image as shown below. Figure 7 As shown, boundary tracing extracted 44 boundary pixels for defect region A, 23 boundary pixels for defect region B, 39 boundary pixels for defect region C, and 24 boundary pixels for defect region D. The defect edge contour was fitted based on the set of defect edge contour pixels, and fitting parameters were set. a =0.25, b =0.25, and the partial control point set constructed to fit the contour of each defect region is shown in Table 3.

[0043] Table 3. Set of control points for contour fitting of each defect region The subpixel defect region localization results are as follows: Figure 8 As shown. From Figure 8 As shown in the subpixel defect region localization image, the fitted defect region shape is affected by the coarse localization of the defect boundary pixels, and the fitted defect contour is continuous and smooth. When magnified, pixel blocks show distortion, but the fitted curve remains continuous, exhibiting stable defect characteristics.

[0044] Further quantitative calculations of the physical characteristics of the defects were performed. First, size mapping was conducted on a 200mm × 200mm specimen, yielding an infrared reconstructed image of 640 × 512 pixels. After size mapping, the pixel dimensions of each parameter were calculated, representing the true size of the defect. The quantitative calculation results of defects in the high-temperature ceramic experimental specimen are shown in Table 4.

[0045] Table 4 Quantitative Calculation Results of Defects in High-Temperature Ceramic Experimental Specimens The quantitative results show that the defect extraction and identification method proposed in this invention has a certain degree of accuracy in identifying defect features. The error range for all defect features is between 0.72% and 6.75%, with the area error being the smallest. This indicates that the automatically determined threshold heat diffusion region removal method designed in this invention can automatically adjust the threshold according to different defect features, heating conditions, and degrees of heat diffusion, resulting in defect regions that are relatively close to the actual defect regions. The improved automation of the algorithm also enhances the accuracy of quantitative identification.

[0046] Applications, modifications and variations of the present invention will be readily apparent to those skilled in the art.

[0047] Although embodiments of the present invention have been disclosed above, they are not limited to the applications listed in the specification and embodiments. They can be applied to various fields suitable for the present invention. For those skilled in the art, other modifications can be easily made. Therefore, without departing from the general concept defined by the claims and their equivalents, the present invention is not limited to the specific details and illustrations shown and described herein.

Claims

1. A quantitative evaluation method for multiple types of complex defects in spacecraft, characterized in that, Includes the following steps: Step 1: Acquire infrared data, process the infrared thermal image sequence data, and obtain infrared reconstructed images characterizing defect features. Igs And fuzzy C-clustering segmentation is used to reconstruct infrared images characterizing defect features. Igs Defect feature extraction is performed to obtain the defect feature extraction image. Ige ; Step 2: Extract image features based on membership degree. Ige Post-processing is performed to optimize the segmentation results and obtain defect feature detection images. Igt ; Step 3: Detect defect features in the image using the time-series alignment method. Igt Different types and locations of defects in the image are corrected to obtain defect feature recognition images. Igc Taking into account the objectively existing thermal diffusion phenomenon, the defect area is corrected to obtain a defect feature recognition image. Igc The specific steps are as follows: Step 3.1: Calculate the first step using the timing alignment method. k Each defective region pixel k x i thermal diffusion judgment threshold k ε ttr_i ; Step 3.1.1: Calculate the pixels in the defect area. k x i The correlation coefficient between the transient thermal response sequence and the background reference point; Step 3.1.2: Following the same calculation method as in Step 3.1.1, calculate the... k The correlation coefficient between the temperature thermal response of each pixel in the defect region and the actual reference point of the defect feature region. k ρ i&bg According to the correlation coefficient k ρ i&bg and k ρ i&def Get the first k Each defective region pixel x i Threshold for determining the thermal diffusion region k ε ttr_i : Step 3.2: Determine the threshold based on the calculated thermal diffusion region. k ε ttr_i When the distance between the defect pixel and the reference point of the actual defect feature area satisfy: Consider pixels x i If the pixel is located in the heat diffusion area, remove it. x i Repeat step 3.2 until the defect feature recognition image is obtained. Igc The Middle k Each defective region pixel x i All pixels were accessed; remove all pixels located in the heat diffusion area. x i Proceed to step 3.3; Step 3.3, repeat steps 3.1 and 3.2 until... K Each pixel in the defect region was accessed, resulting in a defect feature recognition image. Igc ; Step 4: Detect defect features in the image using 8-adjacency Freeman chain codes. Igc By identifying the boundary features of defects of different types and locations, the defect area is calibrated to achieve coarse localization of defect pixels. Step 5: Based on the coarse localization results of defect pixels, sub-pixel defect feature localization is achieved for each labeled defect region using Bézier curve fitting based on the extracted defect boundary pixels. Step 6: Based on the subpixel defect localization results, for each calibrated defect area, the contour feature dimensions in the horizontal and vertical directions are mapped according to the size relationship between the specimen and the image to obtain the actual size of different feature parameters of the defect. Step 7: Perform defect feature quantification calculation.

2. The quantitative evaluation method for multiple types of complex defects in spacecraft as described in claim 1, characterized in that, In step one, defect feature extraction is performed to obtain a defect feature extraction image. Ige The specific steps include: Step 1.1: Set the number of clusters to [number]. C Initialize cluster centers v k , k =1,…, C Initialize the defect filter image Ige medium pixel x a Membership degree with cluster center v k membership degree u ak pixel x a membership degree u ak satisfy ; Step 1.2: Calculate and update cluster centers v k : in, m For smoothing parameters, N For defect filtering images Igs The number of pixels in the image; Step 1.3: Update pixels x a Membership function u ak : Step 1.4: Calculate the current number. t The objective function at the next iteration J t : Step 1.5, Iteration Termination Condition Judgment: If the... t The objective function at the next iteration J t With the t The objective function at -1 iterations J t-1 The difference is less than the threshold for the iteration condition judgment. ε ,Right now‖ J t - J t-1 ||< ε Or reach the maximum number of iterations. T max The iteration terminates, yielding the final set of cluster centers. v k , k =1,…, C and defect-filtered images Igs Membership matrix of each pixel u ak ; Step 1.6, Defect Filtering Image Igs pixel x a Divided into membership matrix u ak The class with the largest value, defect-filtered image Igs All pixels x a The final defect feature extraction image is obtained after the segmentation is completed. Ige .

3. The quantitative evaluation method for multiple types of complex defects in spacecraft as described in claim 1, characterized in that, The specific steps of step two include: Step 2.1: Detect all possible misclassified pixels; detect all pixels. x a of r × r In the neighborhood region, the classification labels of each neighboring pixel are as follows: if the classification labels are the same, the original classification labels are kept unchanged; otherwise, pixels with inconsistent labels are considered to be misclassified and are extracted and saved. Step 2.2: Relabel the extracted potentially misclassified pixels; x e ,statistics x e of n × n Pixels in the neighborhood region x p Membership degree, recalculate pixel points x e Belongs to the k Membership degree of a class u ek ′: in, Ω e Represents pixels x e of r × r The set of neighboring pixels; express x e The neighboring pixels belong to the first k Membership degree of a class; Γ e express Ω e The number of pixels, according to u ek The calculation of ′ will x e The pixel is classified into the category with the highest membership value, and the pixel label is reset. It is important to note that when resetting pixel labels, the label being reset must be the same as the one used in step 2.

2. r × r Classification labels appearing in the neighborhood region; image for extracting defect features. Ige The defect feature detection image is obtained by classifying and relabeling all possible defective pixels. Igt .

4. The quantitative evaluation method for multiple types of complex defects in spacecraft as described in claim 1, characterized in that, In step three, the defect feature detection image obtained in step two is... Igt Divided into K The region, for the first k One defective area, k =1,…, K The size of the heat diffusion region is obtained by measuring the similarity of pixels in the defective region. in, In the i-th defect region, pixel k x i The distance between the reference point and the actual area of ​​the defect feature; For the first k The number of pixels in each defective region; For the first k The maximum distance between a pixel in a defect region and a reference point in the actual defect feature region; k ε ttr_i For defective pixels k x i The thermal diffusion determination threshold is calculated by separately measuring the temperature thermal response correlation between the defect region pixels and the actual defect feature reference point, and the temperature thermal response correlation between the defect region pixels and the background region reference point. k Each defective region pixel k x i thermal diffusion judgment threshold k ε ttr_i ; The specific steps in step 3.1.1 include: Step 3.1.1.1: Determine the transient thermal response sequence that needs to be time-shifted; select the pixel corresponding to the characteristic transient thermal response of the background region as the background region reference point. The corresponding temperature thermal response sequence is , T For thermal sequence acquisition time; for defective region pixels. k x i Transient thermal response sequence Find the reference points for the background areas respectively. The frame number where the temperature peak is located and defective area pixels k x i The frame number where the temperature peak is located k t 1; if ,but It is a transient thermal response sequence that needs to be moved, denoted as ttr 2. Remember for ttr 1; otherwise, It is a transient thermal response sequence that needs to be moved, denoted as ttr 1, denoted as ttr 1 ; Step 3.1.1.2, record ttr The frame number where the temperature peak is 1 is t max Set each move tt Time units, loop parameters t m Set to 1, there is ;right ttr 2. Perform time-series shifts, each shift... tt Each time unit will ttr subsequence of 2 and ttr subsequence of 1 Align and calculate their correlation coefficients. : in, E (·) represents the expectation of the transient thermal response sequence; Step 3.1.1.3 t m = t m +1, repeat step 3.1.1.2, calculating the corresponding Pearson correlation coefficient for each time interval, until... Until then; the set of correlation coefficients is obtained. ; Step 3.1.1.4, from Find the maximum value as the first k Each defective region pixel k x i Reference point with background area Temperature thermal response sequence correlation coefficient k ρ i&bg : 。 5. The quantitative evaluation method for multiple types of complex defects in spacecraft as described in claim 1, characterized in that, Step four involves detecting defect feature recognition images using 8-adjacency Freeman chain codes. Igc By analyzing the boundary features of defects of different types and locations, the defect region is calibrated to achieve coarse localization of defect pixels. Specifically, for the first... k , k =1,…, K The specific process for pixel calibration and localization of each defect area is as follows: Select the k The starting pixel for boundary tracking of each defect region is denoted as... k E 0, with chaincode value dir The direction of =0 is the starting direction of the boundary search. Rotate 45° counterclockwise, following the direction... dir =1 to search for the next pixel; since the extracted defect feature image is a binary image, the pixel value representing the defect contour pixel is represented by "1"; based on this, during the search for defect boundary pixels, if the pixel value is 1... k x t If the pixel value is "1", then the pixel is considered to be "1". k x t It is a defect boundary point, denoted as k E t And save; and k x t The chain code value is assigned to the pixel preceding it. k x t-1 ; to pixels k x t As the new starting point for the boundary search, rotate 90° counterclockwise, following the direction... dir =2 Search for the next defect boundary point; repeat the above steps until the starting point is found. k b At time 0, the defect area search was completed; during the entire search process, the detected boundary points... k E 0,…, k E M This refers to the boundary profile of the defect; the set of boundary points. k Ω L , k Ω L =( k E 0,…, k E M The defects were coarsely located pixel by pixel; the above process was repeated until all defect areas were visited, resulting in... K Pixel calibration and localization results for each defect area.

6. The quantitative evaluation method for multiple types of complex defects in spacecraft as described in claim 1, characterized in that, The specific steps of step five include: after coarse pixel positioning, calibration is obtained. K The first defect area was extracted, and the calibrated first defect area was extracted. k Defective areas M Each defect boundary pixel k =1,…, K , M The defect boundary pixels constitute a pixel set. k E i , i =0,1,…, M For adjacent defect boundary pixels respectively k E i and pixels k E i+1 Perform cubic Bézier curve fitting pairwise; because i The maximum value is M Therefore, pixel k E M+1 It does not exist; considering that the defect contour is a closed figure, we use... k E 0 as k E M+1 and k E M Perform cubic Bézier curve fitting; use the cubic Bézier curve to analyze the defect boundary pixels. k E i and k E i+1 The specific process of obtaining the sub-pixel defect closed contour by pairwise fitting is as follows: Step 5.1: For the edge contour pixels to be fitted... k E i and k E i+1 Construct control points between them. k K i and k C i : in, k a * and k b * It is any given positive number; k x i , k y i () represents the pixel at the defect boundary. k E i The positional parameters; due to the boundary pixel points of the defect detected by the boundary tracking algorithm. k E i ( k x i , k y i ), i =0,1,…, M Therefore, there are no pixels in the set of defect boundary pixels. k E -1 , k E M+1 , k E M+2 Its corresponding position parameter ( k x -1 , k y -1 ), ( k x M+1 , k y M+1 ), ( k x M+2 , k y M+2 Therefore, it does not exist; considering that the defect profile is closed, ( k x M , k y M The value of ) is used as ( k x -1 , k y -1 The value of ), ( k x 0, k y The value of 0) is used as ( k x M+1 , k y M+1 The value of ), ( k x 1, k y The value of 1) is used as ( k x M+2 , k y M+2 The value of ); Step 5.2: According to the definition of the Bézier curve, the defect boundary pixels... k E i and k E i+1 Expression of the cubic Bézier curve between for: The defect boundary pixels are obtained based on the expression for the cubic Bézier curve. k E i and pixels k E i+1 Fitting curves between; Step 5.3: Traverse the set of defect boundary pixels k E i All the pixels, i =0,1,…, M Pairwise fitting is performed to obtain the fitting curve between adjacent pixels; when i = M When, use k E 0 as k E M+1 and k E M Perform cubic Bézier curve fitting; Step 5.4: After traversing all defect boundary pixels, the result is obtained. M A cubic Bézier curve; M The first Bézier curve is smoothly connected to obtain the second Bézier curve. k Subpixel defect closed contour of each defect region k L Based on closed contours k L Achieve subpixel localization of defects; Step 5.5: Repeat the above process until... K All defect regions were accessed, and sub-pixel localization results of defects of different types and locations were obtained.

7. The quantitative evaluation method for multiple types of complex defects in spacecraft as described in claim 1, characterized in that, The method for mapping the actual dimensions of different characteristic parameters of the inscribed defect in step six includes: mapping the calibrated first... k One defective area, k =1,…, K The coordinates of the sub-pixel boundary points are ( k x s , k y s The mapping is performed according to the following formula: In the formula, L × W The actual dimensions of the test specimen; M × N Given the size of the captured infrared image, obtain the mapped subpixel boundary point coordinates. k x s ′, k y s ′).

8. The quantitative evaluation method for multiple types of complex defects in spacecraft as described in claim 1, characterized in that, The specific steps of step seven, which involves quantifying defect features, include: perimeter and area are conventional and directly reflect the size information of defects. Based on the numerical values ​​of geometric structural feature parameters, the size information of each defect feature region can be directly quantified and evaluated, thereby reflecting the geometric size differences between each defect feature region; the area of ​​the defect can be used to quantify the size of the defect in the image; the feature parameters of different defect regions are calculated separately; and the calibrated first... k One defective area, k =1,…, K For the mapped sub-pixel boundary points ( k x s ′, k y s The boundary contour of the defect in the actual size drawn by the set is denoted as ′) k L Since most defect feature areas are irregular, the defect boundary contour... k L The shape is irregular; the defect boundary profile is calculated directly. k L Determining the area of ​​the enclosed defect region is difficult; therefore, the defect boundary profile is used. k L Enclosed area k Ω pixels k p i The number of samples and the area feature parameters are used; since boundary size mapping was performed in step seven, area conversion between the specimen and the image is no longer required, and the defect boundary contour is statistically obtained at this point. k L Enclosed area k Ω The number of pixels is the area feature parameter of the defect: By statistically analyzing the area enclosed by the defect contour, the location information of the defect can be further determined, achieving the purpose of defect calibration. In the mapped defect contour image, the perimeter, major axis, and minor axis features of the defect are further calculated based on the defect boundary contour features. When calculating the perimeter of the defect, the fitted sub-pixel contour is obtained by sequentially fitting the defect boundary pixels one by one. k E i and k E i+1 The defect boundary feature size mapping does not change the positional relationship between sub-pixel boundary points, so the boundary pixel points can be obtained through curve integration. k E i and k E i+1 The actual length of the fitted curve between k P i : Wherein, are the defect boundary pixels. k E i and k E i+1 The expression for the cubic Bézier curve between; M The sum of the actual fitted curve lengths between each pair of boundary pixels is the perimeter of the defect. k P : Meanwhile, due to the irregularity of the defect feature region, the size of the minimum bounding rectangle obtained based on the sub-pixel defect boundary contour can also reflect the morphological characteristics of the defect; based on the fitted contour of the sub-pixel defect boundary contour... k L Find its minimum bounding rectangle, and use the minimum bounding rectangle to characterize the distribution of defects; the longer side of the minimum bounding rectangle is the major axis of the defect. k Dl The shorter side is the minor diameter of the defect. k Ds The minimum bounding rectangle of the subpixel defect boundary contour is determined using the area-based equally spaced rotation method: the subpixel defect boundary contour image is rotated equally within a 90° range, and the parameters of the bounding rectangle of the subpixel defect boundary contour are recorded for each rotation; the maximum horizontal coordinate of the subpixel defect boundary contour is then found. k x max and minimum value k x min The maximum value of the vertical coordinate. k y max and minimum value k y min Then the marked number k The longer side of the bounding rectangle of the defect region is: k Dl =max( k y max - k y min , k x max - k x min ) The shorter side is: k Ds =min( k y max - k y min , k x max - k x min ); The area of ​​the circumscribed rectangle is k Dl × k Ds This invention defines the minimum bounding rectangle by area. The specific process of calculating the minimum bounding rectangle is as follows: First, based on the maximum and minimum values ​​of the sub-pixel defect boundary contour distribution coordinates, a bounding rectangle of the sub-pixel defect boundary contour region without rotation angle can be obtained. Then, the long side, short side, and area of ​​the minimum bounding rectangle in the current state are calculated, and the rotation angle is determined at this time. α =0°; Set equal-interval rotation angles θ Rotate the subpixel defect boundary contour by an angle. θ , α = α + θ The bounding rectangle of the subpixel defect boundary contour is obtained in the same way, and the relevant parameters are recorded; the rotation process is repeated until the rotation angle is reached. α =90°; Finally, determine the minimum bounding rectangle based on the area size; Based on the above quantization process, calculate the physical characteristic parameters of the defect based on the sub-pixel defect boundary contour fitting results.

Images