Image processing-based thermal protection function gradient material component identification method and system

Abstract

The application discloses a kind of based on image processing's thermal protection function gradient material component identification method and system, the method includes: obtaining thermal protection function gradient material original micrograph I;Thermal protection function gradient material original micrograph I is preprocessed, and preprocessed image is obtained in preprocessed image, on the basis of preprocessed image, through interactive image segmentation method, realize the component identification and segmentation of thermal protection function gradient material.The application is suitable for the identification of most thermal protection function gradient material microstructure component, can guarantee the precision and efficiency of identification simultaneously.

Description

TECHNICAL FIELD

[0001] The present application belongs to the technical field of microstructure images of functionally graded materials, and particularly relates to a method and system for identifying components of thermal protection functionally graded materials based on image processing. BACKGROUND

[0002] The performance of functionally graded materials is strongly dependent on the microstructure of the material, and the microstructure is one of the basic research contents in the science and engineering of functionally graded materials. Microstructure images are usually used to analyze the microstructure related to certain characteristics of the material, such as the microstructure of the material observed by scanning electron microscopy, including phase composition, distribution, and grain boundary feature distribution. Therefore, one of the key steps in the material structure design process is to effectively extract useful information and data from the images using algorithms and tools. The goal of image processing algorithms is to detect and separate each grain from the original microstructure image to obtain an accurate description of the microstructure, such as geometric and topological features.

[0003] Material microstructure data is an important type of material data, and the conversion of microstructure images into structural data through image processing technology can establish a causal relationship between structure and performance. Image segmentation is an important branch of image processing and the first step of image analysis, and it is also one of the most difficult problems in image processing. Traditional microstructure image analysis is mainly performed manually, which has the disadvantages of low efficiency, inaccurate analysis, and only obtaining original image features. In recent years, with the continuous development of materials science, scholars are increasingly eager to obtain ways to improve material performance from a micro perspective, find the internal mechanism of material properties, and guide the manufacturing process of materials. Therefore, applying computer image segmentation technology to analyze microstructure has become a research hotspot in the field of materials science.

[0004] Many scholars have studied the image segmentation method of material microstructure, which can be divided into low-level semantic segmentation, medium-level semantic segmentation, and high-level semantic segmentation according to the complexity of image features. The low-level semantic segmentation based on single-pixel features has the disadvantage of not fully utilizing the area information of each component in the image, and only processing the features of a single pixel. Therefore, the segmentation result is sensitive to noise, the segmentation accuracy is limited, and it cannot effectively segment complex images. The latest progress in high-level semantic segmentation of material microstructure images is to train a large number of data sets through deep learning to achieve automatic image segmentation. However, the automatic segmentation method requires a large number of image data for training when constructing a neural network, which has a high computational cost. At the same time, computer artificial intelligence has not yet reached the level of understanding high-level semantics, and it is difficult to obtain satisfactory results through automatic segmentation of images, which makes them not very convenient for the task of segmenting a small number of images. Therefore, it is necessary to establish a general segmentation method that can guarantee the segmentation effect and efficiency and is suitable for most thermal protection functionally graded material microstructure images. SUMMARY

[0005] The technical problem solved by the present application is to overcome the deficiencies of the prior art and provide a thermal protection function gradient material component identification method and system based on image processing, which is suitable for identifying most of the microstructure components of thermal protection function gradient materials and can ensure the accuracy and efficiency of identification.

[0006] To solve the above technical problems, the present application discloses a thermal protection function gradient material component identification method based on image processing, comprising:

[0007] Obtaining a thermal protection function gradient material original microstructure image I;

[0008] Pretreating the thermal protection function gradient material original microstructure image I to obtain a pretreated image

[0009] On the basis of the pretreated image , the component identification and segmentation of the thermal protection function gradient material are realized by an interactive image segmentation method.

[0010] In the above thermal protection function gradient material component identification method based on image processing, the thermal protection function gradient material original microstructure image I is pretreated by a k-means clustering algorithm to obtain a pretreated image

[0011] In the above thermal protection function gradient material component identification method based on image processing, the thermal protection function gradient material original microstructure image I is pretreated by a k-means clustering algorithm to obtain a pretreated image comprising:

[0012] Step 11, dividing the thermal protection function gradient material original microstructure image I into k clusters;

[0013] Step 12, initializing the cluster k;

[0014] Step 13, calculating the Euclidean distance between each centroid and each pixel to be clustered;

[0015] Step 14, according to the calculated Euclidean distance between each centroid and each input pixel to be clustered, all pixels to be clustered are assigned to the nearest centroid to obtain updated pixels to be clustered;

[0016] Step 15, updating the centroid according to the updated pixels to be clustered to obtain an updated centroid;

[0017] Step 16: Repeat steps 13 to 15 above, iterating continuously until the centroid obtained at the end does not change or meets the set error requirements. This centroid is recorded as the highest centroid, and the pixel to be clustered corresponding to the highest centroid is recorded as the clustering pixel.

[0018] Step 17: Reconstruct the image based on the clustered pixels to obtain the preprocessed image.

[0019] In the above image processing-based method for identifying the components of thermal protection functionally graded materials, the Euclidean distance is calculated as follows:

[0020] d ik =||g i (x,y)-c k ||

[0021] Where, d ki c represents the Euclidean distance between the k-th centroid and the i-th pixel to be clustered. k G represents the position of the k-th centroid. i (x,y) represents the i-th pixel to be clustered.

[0022] In the above image processing-based method for identifying the components of thermally protected functionally graded materials, the updated centroid calculation formula is as follows:

[0023]

[0024] Among them, g′ i (x,y) represents the updated i-th pixel to be clustered, and c′ k This indicates the position of the k-th centroid after the update.

[0025] In the above-mentioned image processing-based method for identifying the components of thermal protection functionally graded materials, the preprocessed image... Based on this, an interactive image segmentation method is used to identify and segment the components of thermally protective functionally graded materials, including:

[0026] Step 21, from the preprocessed image Extract several image patches of size M*M, each image patch representing a vector containing pixel intensity; divide the image patches into K clusters, the set of cluster centers representing the image content; where M represents pixel value;

[0027] Step 22: Use the allocation image A to uniquely allocate each image pixel (m,n) at the center of each image block to a corresponding dictionary block in the dictionary, and construct a double adjacency matrix B according to the allocation relationship;

[0028] Step 23: Encode the input information using label L; where the input information is a manually provided tag.

[0029] Step 24, according to the transformation matrix T derived from the double adjacent matrix B, the coded input information is propagated from the image space to the dictionary space, and the probability label of the whole image is obtained through the diffusion matrix P;

[0030] Step 25, according to the given additional label, the image segmentation result is improved, that is, the above steps 23-24 are repeated, and finally the segmentation image meeting the requirements is output, realizing the component identification and segmentation of the thermal protection function gradient material.

[0031] In the above-mentioned thermal protection function gradient material component identification method based on image processing, the assignment image A defines the binary relationship between the pixels assigned to the same dictionary and the corresponding image.

[0032] In the above-mentioned thermal protection function gradient material component identification method based on image processing, the assignment image A is obtained in an unsupervised manner; wherein the assignment image A contains information about the structural clustering of the preprocessed image .

[0033] In the above-mentioned thermal protection function gradient material component identification method based on image processing, the double adjacent matrix B is used to calculate the transformation of the dictionary probability and the image block probability, and to establish a linear index for the image pixels in the image block and the dictionary pixels in the dictionary block, that is, to define a linear mapping process of propagating information from the image to the dictionary.

[0034] Correspondingly, the application discloses a thermal protection function gradient material component identification system based on image processing, comprising:

[0035] The acquisition module is used for acquiring the original micro image I of the thermal protection function gradient material.

[0036] The preprocessing module is used for preprocessing the original micro image I of the thermal protection function gradient material to obtain the preprocessed image

[0037] The identification and segmentation module is used for realizing the component identification and segmentation of the thermal protection function gradient material through the interactive image segmentation method on the basis of the preprocessed image .

[0038] The application has the following advantages:

[0039] (1) The application discloses a thermal protection function gradient material component identification scheme based on image processing, and the universality of an interactive image segmentation method is utilized; for different research problems, input features can be controlled by manual marking, an identification algorithm is trained, different features of different components in an image can be flexibly and conveniently obtained, and different components are identified and segmented. While the segmentation effect is ensured, the identification efficiency of the method is high, and a small amount of effective marking can obtain a good segmentation result.

[0040] (2) The application discloses a thermal protection function gradient material component identification scheme based on image processing, and the universality and convenience of the interactive method segmentation method based on a pixel level are good for a function gradient material micro image with a real boundary or high contrast. BRIEF DESCRIPTION OF DRAWINGS

[0041] Fig. 1 is a step flow chart of a thermal protection function gradient material component identification method based on image processing in the embodiment of the application;

[0042] Fig. 2 a thermal protection function gradient material micro image in the embodiment of the application;

[0043] Fig. 3 an interactive image segmentation process chart in the embodiment of the application;

[0044] Fig. 4 a detail comparison chart of a segmentation result in the embodiment of the application. DETAILED DESCRIPTION

[0045] In order to make the purpose, technical scheme and advantages of the application more clear, the following will further describe the disclosed embodiments of the application in combination with the drawings.

[0046] One of the core ideas of the application is that: based on the principle of image segmentation, starting from a micro image of a function gradient material, micro features of each component are obtained through an interactive image segmentation method, and then different components are identified to realize segmentation of each component. The purpose is to identify each phase component in the micro image of the function gradient material through image segmentation, and then represent the content and spatial distribution of the phase component. In addition, the application utilizes the convenience of the interactive method, and presents the image features of interest marked by a person in real time in a program, and realizes re-segmentation of the image quickly through modification of the marking.

[0047] As Figs. 1-4 In the embodiment, the thermal protection function gradient material component identification method based on image processing comprises:

[0048] Step 101: obtaining a raw micro image I of a thermal protection function gradient material.

[0049] Step 102, pre-process the original micro image I of the thermal protection functional gradient material to obtain a pre-processed image

[0050] In the embodiment, the k-means clustering algorithm can be used to pre-process the original micro image I of the thermal protection functional gradient material to obtain a pre-processed image

[0051] Step 1021, divide the original micro image I of the thermal protection functional gradient material into k clusters.

[0052] Step 1022, initialize the cluster k.

[0053] Step 1023, calculate the Euclidean distance between each centroid and each pixel to be clustered.

[0054] Step 1024, according to the Euclidean distance between each centroid and each pixel to be clustered, assign all pixels to be clustered to the nearest centroid to obtain updated pixels to be clustered.

[0055] Step 1025, update the centroid according to the updated pixels to be clustered to obtain an updated centroid.

[0056] Step 1026, repeat the above steps 1023-1025, and iterate until the final centroid has no change or meets the set error requirement, denoted as the optimal centroid, and the pixels to be clustered corresponding to the optimal centroid are denoted as clustered pixels.

[0057] Step 1027, according to the clustered pixels, perform image reconstruction to obtain a pre-processed image

[0058] Preferably, the calculation formula of the Euclidean distance is as follows:

[0059] d ik =||g i (x,y)-c k ||

[0060] Where d ki represents the Euclidean distance between the kth centroid and the ith pixel to be clustered, c k represents the position of the kth centroid, and g i (x,y) represents the ith pixel to be clustered.

[0061] Preferably, the calculation formula of the updated centroid is as follows:

[0062]

[0063] Where g′ i(x,y) represents the updated i-th pixel to be clustered, c' k represents the position of the updated k-th centroid.

[0064] Preferably, the objective function J that minimizes the sum of distances from each pixel to be clustered to the centroid is:

[0065]

[0066] where R represents the total number of pixels to be clustered, w ik represents the weight. If the pixel to be clustered g i (x,y) belongs to the k-th cluster, w ik =1; otherwise, w ik =0. It can be seen that this is a minimization problem consisting of two parts. First, fix c k and calculate the minimum of J with respect to w ik . Then, fix w ik and calculate the minimum of J with respect to c k . Therefore:

[0067] The minimum of the first part can be expressed as:

[0068]

[0069] That is, the minimum of the first part is to assign g i (x,y) to the nearest cluster, according to the sum of the square of the distance between g i (x,y) and the centroid c k of the cluster, which corresponds to steps 1023 and 1024 described above.

[0070] The minimum of the second part can be expressed as:

[0071]

[0072] This is the mathematical expression for recalculating the centroid of each cluster to reflect the new assignment, which corresponds to step 1025 described above.

[0073] Step 103, based on the preprocessed image , the composition recognition and segmentation of the thermal protection function gradient material are realized by the interactive image segmentation method.

[0074] In the embodiment, the interactive image segmentation method can also be called a semi-automatic segmentation method, which is a medium-level semantic segmentation method. Through limited input of a user, more segmentation information is obtained as much as possible, so that the research object is accurately and conveniently segmented. The method has good universality and convenience, and can give different artificial labels for different functional gradient material micro images or different research problems to achieve the segmentation effect of "semi-automatic learning". Since the content and distribution of the phase composition of the functional gradient material are affected by the complex processing technology, it is difficult to obtain quantitative data. At the same time, the thermal mechanical properties of the functional gradient material are seriously dependent on the microstructure of the material, such as porosity, content and distribution, and even the performance of the functional gradient material with the same composition can be significantly different. Therefore, the purpose of the present application is to identify each phase composition in the micro image of the functional gradient material through image segmentation, and then represent the content and spatial distribution of the phase composition.

[0075] In the embodiment, the purpose of the pre-processing (step 102) is to find the structure in the image without providing labels. In order to obtain more accurate segmentation, the image needs to be further processed on the basis of the result of step 102. Since the micro image features of the composite material tend to form discrete clusters, and the image features in the same cluster may belong to the same class, but it cannot be assumed that each class has only one cluster representation, so more clusters than classes need to be created to capture various features existing in the image. Since the output of this step is a hierarchical image, each pixel position contains the probability of belonging to each defined class. According to the pixel probability, the segmentation result can be obtained by selecting the optimal possible class for each pixel, and the implementation steps are as follows:

[0076] Step 1031, a plurality of image blocks with a size of M*M are extracted from the pre-processed image , each image block represents a vector containing pixel intensity; the plurality of image blocks are clustered into K clusters, and the cluster center set represents the image content; wherein M represents the pixel value.

[0077] Step 1032, each image pixel (m, n) of the cluster center is uniquely assigned to a corresponding dictionary block in the dictionary by using the assignment image A, and a double-neighbor matrix B is constructed according to the assignment relationship. Wherein, the assignment image A can be obtained in an unsupervised manner, and the assignment image A contains information about the structure cluster in the pre-processed image ; that is, the assignment image A defines the binary relationship between the pixels assigned to the same dictionary and the corresponding image pixels.

[0078] Step 1033, the input information is encoded by using the label L. Wherein, the input information is the artificial label.

[0079] Step 1034, according to the transformation matrix T derived from the double-adjacent matrix B, the coded input information is propagated from the image space to the dictionary space, and the probability label of the whole image is obtained through the diffusion matrix P. Wherein, the double-adjacent matrix B is used to calculate the transformation of the dictionary probability and the image block probability, and a linear index is established for the image pixels in the image block and the dictionary pixels in the dictionary block, that is, a linear mapping process of propagating information from the image to the dictionary is defined.

[0080] Step 1035, according to the given additional label to improve the image segmentation result, that is, repeat the above steps 1033-1034, finally output the segmented image meeting the requirements, realize the component identification and segmentation of the thermal protection function gradient material.

[0081] It can be seen that the core part of the process is to construct a transformation matrix that converts the artificial label into pixel probability. The construction process of the transformation matrix will be described in principle as follows.

[0082] As mentioned earlier, the assignment image A obtained in an unsupervised manner contains information about the structure clustering in the image I. At the same time, the preprocessed image Also contains part of the label L provided by human. In order to combine the two information sources, a label dictionary is created to match the intensity dictionary. For each dictionary block, the assignment image A is used to mark the positions of all image blocks assigned to it. From these positions in the image grid, the corresponding blocks are extracted from the labeled image L. For the relevant set of labeled blocks, the pixel average value of each layer is calculated. Therefore, each M*M dictionary block has a corresponding M*M label representation composed of C layers. When the image is completely labeled, the sum of the label image L in each pixel is equal to 1, because only one C class of label is 1. Therefore, the representative label of each dictionary block also has a sum equal to 1 in each pixel. However, due to the averaging of pixels, the value of this representation is not binary, but decodes the normalized frequency of dictionary pixels labeled as belonging to the C class in the current label image. It can be considered that this representative label is the image probability of belonging to the C class, which is called dictionary probability. The dictionary probability can be pasted back into the X×Y image grid, also using the position information from the assignment image A, and averaging again for each pixel. In this way, an X×Y probability image P composed of C layers is obtained, where P is the diffusion form of L. That is, the self-similarity information encoded by the assignment image A propagates the human-provided label from L to other parts of the image.

[0083] Further, the construction of the transformation matrix for efficient calculation of the dictionary probability and the image probability is as follows:

[0084] The basis of this conversion is the relationship between the X×Y image grid and the M×M dictionary grid. This relationship is encoded using a double-adjacent matrix B of a×b. Wherein, a=XY, b=M2 K. To this end, a linear (single) indexing of the pixels in the image and the pixels in the dictionary grid is needed.

[0085] The linear index of the image pixel (m, n) is:

[0086]

[0087] For the dictionary grid, take (0, 0, K) as the center pixel of the Kth dictionary element, the coordinates of the other pixels in the block are defined according to the intra-block displacements Am and An, both from {-s,..., 0,..., s} and s = (M - 1) / 2. A dictionary pixel at coordinates (Am, An) has a linear index:

[0088] ξ = (Am + s) + (An + s)M + (K - 1)M 2

[0089] Each assignment of an image block centered at (m, n) to the Kth dictionary block centered at (0, 0, K) derives M 2 2 relations between image and dictionary pixels. Denote the relation between image and dictionary pixels with " ~ ", for all Am and An we have:

[0090]

[0091] Since the image blocks are overlapping, each non-boundary image pixel is related to M 2 dictionary pixels. The image pixels at the boundaries are related to less than M 2 dictionary pixels, and the pixels at the four corners are related to only one dictionary pixel. Therefore, there are in total (X - 2s)(Y - 2s)M 2 relations between image and dictionary pixels. Denote the relation between p image pixels and q dictionary pixels with a p x q bi-adjacency matrix B, the elements in B are:

[0092]

[0093] where m and n are the linear indices of the image pixels, and ξ is the linear index of the dictionary pixels.

[0094] The bi-adjacency matrix B defines a linear mapping process of propagating information from the image to the dictionary, and vice versa.

[0095] ​​​Consider the quantities defined on the image grid (e.g. the labels provided representing pixels belonging to class 1) and arrange them into a vector v of length l such that the i-th element contains the value of the i-th image pixel. Propagate these values into the dictionary by computing a vector D of length h:

[0096]

[0097] where 1 is a column vector of ones and diag(•) denotes a diagonal matrix with the argument defining the diagonal. Thus, the mapping from the image to the dictionary is defined as the transformation matrix T1:

[0098]

[0099] Likewise, the mapping from the dictionary to the image is given by the transformation matrix T2:

[0100] T2= diag(B1 m×1 ) -1 B

[0101] The two transformation matrices above are the basis of the method. The artificially provided labels propagate as follows:

[0102] P = T2T1L

[0103] where L is the vector L arranged in an n x C matrix and the resulting n x C matrix P needs to be arranged back into the layered image P.

[0104] In the present embodiment, there are two post-processing ways in this interactive method. The first way is to process the probability image, i.e. to obtain the features of interest in the segmentation or detection from the probability image. The second way is to extract the information stored in the dictionary and the probabilities associated with the dictionary.

[0105] Linear transformations are at the core of the post-processing, first transforming the user provided labels from L through the transformation matrix T1 into the dictionary space and then back into the image space through the transformation matrix T2. Consider the transformation from the labels to the dictionary space:

[0106] O = T1L

[0107] where the matrix O contains the pixel-level probabilities of the dictionary pixels (i.e. the dictionary probabilities) which are useful for processing images similar to I but not previously processed. That is, using a dictionary created with the classes of blocks in I, process a new pre-processed image All M x M classes of blocks need to be extracted for each pixel of I and assigned to the existing dictionary. Similarly to before, this process defines an image to dictionary procedure for which two associated transformation matrices can be computed. The dictionary to image transformation matrix is needed for the next step. Therefore, in order to calculate the probability image corresponding probability image the probability image matrix needs to be calculated

[0108]

[0109] The subsequent use of the method is in line with the framework of supervised learning. In this way, the original micro image I of the thermal protection functionally graded material and the calculated label L are for the training set. Then, the method can be used for new, unmarked preprocessed images generate a probability image Therefore, as long as the initial clustering can capture the features present in the image, the method can handle similar images.

[0110] To sum up, the present application starts from the micro image of the functionally graded material, obtains the micro features of each component based on the microstructure of the material through the interactive image segmentation method, and then identifies different components to realize the segmentation of each component. The interactive method segmentation method based on the pixel level has good universality and convenience for the micro image of the functionally graded material with a real boundary or high contrast.

[0111] On the basis of the above-mentioned embodiments, the present application further discloses a thermal protection functionally graded material component identification system based on image processing, comprising: an acquisition module configured to acquire a thermal protection functionally graded material original micro image I; a preprocessing module configured to preprocess the thermal protection functionally graded material original micro image I to obtain a preprocessed image An identification and segmentation module is configured to realize the component identification and segmentation of the thermal protection functionally graded material through an interactive image segmentation method based on the preprocessed image .

[0112] For the system embodiment, since it corresponds to the method embodiment, the description is relatively simple, and the related parts are described in the method embodiment part.

[0113] Although the present application has been disclosed with the above-mentioned preferred embodiments, it is not intended to limit the present application, and any person skilled in the art can make possible changes and modifications to the technical solutions of the present application by using the above-mentioned disclosed methods and technical contents without departing from the spirit and scope of the present application. Therefore, any simple modification, equivalent change and modification made to the above-mentioned embodiments according to the technical essence of the present application, which does not depart from the technical solutions of the present application, belongs to the protection scope of the technical solutions of the present application.

[0114] The contents not described in detail in the specification of the present application are the known technology of the person skilled in the art.

Claims

1. A method for identifying the composition of a thermal protection function gradient material based on image processing, characterized in that, The method comprises the following steps: An original micro image I of the thermal protection function gradient material is acquired; The k-means clustering algorithm is used for pre-processing of the thermal protection function gradient material original micro image I to obtain a pre-processed image The method comprises the following steps: step 11, dividing the thermal protection function gradient material original micro image I into k clusters; step 12, initializing the cluster k; step 13, calculating the Euclidean distance between each centroid and each pixel to be clustered; step 14, according to the Euclidean distance between each centroid and each input pixel to be clustered, all pixels to be clustered are assigned to the nearest centroid to obtain updated pixels to be clustered; step 15, according to the updated pixels to be clustered, the centroid is updated to obtain an updated centroid; step 16, repeating the above steps 13-15, and constantly iterating until the final centroid has no change or meets the set error requirement, and the optimal centroid is recorded as the optimal centroid, and the pixels to be clustered corresponding to the optimal centroid are recorded as the clustered pixels; step 17, according to the clustered pixels, the image is remodeled to obtain a pre-processed image On the basis of pre-processing image , component identification and segmentation of thermal protection function gradient material are realized by interactive image segmentation method; including: step 21, extracting a plurality of image blocks with size M*M from pre-processing image , each image block represents a vector containing pixel intensity, and M represents pixel value; a plurality of image blocks are clustered into K clusters, and the set of cluster centers represents image content; step 22, each image pixel (m, n) of the image block center is uniquely assigned to a corresponding dictionary block in the dictionary by assigning image A, and a double neighbor matrix B is constructed according to the assignment relationship; the assignment image A defines the binary relationship between the pixels assigned to the same dictionary and the corresponding image pixels; the double neighbor matrix B is used to calculate the transformation of dictionary probability and image block probability, and to establish a linear index between image pixels in the image block and dictionary pixels in the dictionary block, that is, to define a linear mapping process of information from image to dictionary; step 23, encoding the input information with label L, the input information is a manually provided label; step 24, according to the transformation matrix T derived from the double neighbor matrix B, the encoded input information is propagated from the image space to the dictionary space, and the probability label of the whole image is obtained through the diffusion matrix P; P=T2T1L; L is the vector of L; the mapping from image to dictionary is defined as transformation matrix T1: T1=diag(B T 1 n×1 ) -1 B T ; the mapping from dictionary to image is defined as transformation matrix T2: T2=diag(B1 m×1 ) -1 B; 1 represents a column vector full of 1, and diag(·) represents a diagonal matrix defined by the diagonal line parameter; step 25, according to the given additional label, the image segmentation result is improved, that is, the above steps 23-24 are repeated, and finally the segmentation image meeting the requirements is output, realizing the component identification and segmentation of thermal protection function gradient material.

2. The image processing based thermal protection function gradient material component identification method of claim 1, wherein, The solution formula of the Euclidean distance is as follows: d ik =||g i (x,y)-c k || wherein d ki represents the Euclidean distance between the kth centroid and the ith pixel to be clustered, c k represents the position of the kth centroid, g i (x, y) represents the ith pixel to be clustered.

3. The image processing based thermal protection function gradient material component identification method of claim 1, wherein, The solution formula of the updated centroid is as follows: where g i (x, y) denotes the updated i-th pixel to be clustered, c k denotes the position of the updated k-th centroid.

4. The image processing based thermal protection function gradient material component identification method of claim 1, wherein, obtaining an assignment image A in an unsupervised manner; wherein the assignment image A contains information about the structural clusters in the preprocessed image .

5. An image processing based thermal protection function gradient material component identification system for implementing the method of claim 1, wherein, The method comprises the following steps: An acquisition module is configured to acquire an original micro image I of the thermal protection function gradient material; A pre-processing module is configured to pre-process the original micrograph image I of the thermal protection function gradient material to obtain a pre-processed image The recognition and segmentation module is used for realizing component recognition and segmentation of the thermal protection function gradient material through an interactive image segmentation method on the basis of pre-processing images ​

Images