Casting image enhancement method for metallographic analysis

Abstract

A cast image enhancement method for metallographic analysis is disclosed. According to the method, a casting metallographic image is collected, analysis and feature recognition of the casting metallographic image are carried out at the rear end through an image processing and analysis algorithm based on artificial intelligence and deep learning, and therefore multi-dimensional representation features related to a casting metallographic phase are captured from the casting metallographic image, and enhancement processing of the casting metallographic image is carried out. In this way, analysis and control over the nodular cast iron inoculation process can be conducted more finely in the follow-up process, and reliable support is provided for production quality control over materials such as nodular cast iron.

Description

Technical Field

[0001] The present disclosure relates to the field of metallographic analysis, and particularly to a casting image enhancement method for metallographic analysis. Background Art

[0002] The inoculation treatment of ductile iron refers to adding an inoculant to the molten iron during the casting process of ductile iron, and the main functions are as follows: First, it helps to optimize the graphite morphology, promote the formation of spherical graphite and the uniform distribution of spherical graphite, enhancing the overall mechanical properties of the material; Second, it helps to reduce the microscopic defects in the casting, such as pores, inclusions, etc. If these defects are not controlled, they may cause stress concentration, thereby reducing the material strength. And the preparation of the inoculant helps to enhance the inoculation effect and reduce the production cost.

[0003] Metallographic analysis is one of the important analysis methods in materials science. By observing and analyzing the metallographic microstructure, the organizational structure, phase composition and performance characteristics of the material can be revealed. In metallographic analysis, the quality of the image, such as clarity and detail performance, etc., is crucial because they directly affect the accuracy of the analysis results. For example, as a high-performance casting material, ductile iron, its microscopic structural characteristics, such as graphite morphology and distribution, have a decisive impact on the material properties. And the casting metallographic images are usually affected by factors such as lighting conditions and noise, making the quality of the image clarity and contrast not high, restricting the accuracy and depth of further analysis.

[0004] Traditional metallographic image analysis relies on optical microscopes and simple image processing techniques, such as histogram equalization, filtering, etc. However, due to the complexity of the material itself and the limitations of the optical microscope, the obtained images often fail to meet the requirements of high-precision analysis. In addition, the microscopic defects often involved in the inoculation treatment process of ductile iron, such as pores and inclusions, require more refined and high-quality image materials for subsequent identification and analysis.

[0005] Therefore, an optimized casting image enhancement scheme for metallographic analysis is desired, which can achieve a higher quality image enhancement effect, so as to facilitate more refined analysis and control of the inoculation process of ductile iron, and provide reliable support for the production quality control of materials such as ductile iron. Summary of the Invention

[0006] The present disclosure is made in consideration of the above problems. An object of the present disclosure is to provide a casting image enhancement method for metallographic analysis.

[0007] Embodiments of the present disclosure provide a casting image enhancement method for metallographic analysis, which includes:

[0008] Obtain the casting metallographic image to be processed;

[0009] Perform a color space transformation on the to-be-processed cast metallographic image to obtain a YCbCr to-be-processed cast metallographic image;

[0010] Perform feature extraction on the to-be-processed cast metallographic image and the YCbCr to-be-processed cast metallographic image respectively to obtain a cast metallographic RGB feature map and a cast metallographic YCbCr feature map;

[0011] Perform feature fusion processing based on gated screening adaptive enhancement on the cast metallographic RGB feature map and the cast metallographic YCbCr feature map to obtain an adaptively enhanced multi-dimensional representation feature map of the cast metallographic;

[0012] Based on the adaptively enhanced multi-dimensional representation feature map of the cast metallographic, determine an enhanced cast metallographic image.

[0013] For example, in the cast image enhancement method for metallographic analysis according to an embodiment of the present disclosure, wherein performing feature extraction on the to-be-processed cast metallographic image and the YCbCr to-be-processed cast metallographic image respectively to obtain a cast metallographic RGB feature map and a cast metallographic YCbCr feature map includes: inputting the to-be-processed cast metallographic image and the YCbCr to-be-processed cast metallographic image into a metallographic feature extractor based on a dilated convolutional neural network model to obtain the cast metallographic RGB feature map and the cast metallographic YCbCr feature map.

[0014] For example, in the cast image enhancement method for metallographic analysis according to an embodiment of the present disclosure, wherein performing feature fusion processing based on gated screening adaptive enhancement on the cast metallographic RGB feature map and the cast metallographic YCbCr feature map to obtain an adaptively enhanced multi-dimensional representation feature map of the cast metallographic includes:

[0015] Fuse the cast metallographic RGB feature map and the cast metallographic YCbCr feature map to obtain a multi-dimensional representation feature map of the cast metallographic;

[0016] Input the multi-dimensional representation feature map of the cast metallographic into a gated screening adaptive attention module to obtain the adaptively enhanced multi-dimensional representation feature map of the cast metallographic.

[0017] For example, in the cast image enhancement method for metallographic analysis according to an embodiment of the present disclosure, wherein inputting the multi-dimensional representation feature map of the cast metallographic into a gated screening adaptive attention module to obtain the adaptively enhanced multi-dimensional representation feature map of the cast metallographic includes:

[0018] Perform global pooling processing based on the maximum value, global pooling processing based on random values, and global pooling processing based on the average value on the cast metallographic multi-dimensional representation feature map to obtain a cast metallographic multi-dimensional representation global maximum pooling feature vector, a cast metallographic multi-dimensional representation global mean pooling feature vector, and a cast metallographic multi-dimensional representation global random value pooling feature vector;

[0019] Calculate the position-wise weighted sum among the cast metallographic multi-dimensional representation global maximum pooling feature vector, the cast metallographic multi-dimensional representation global mean pooling feature vector, and the cast metallographic multi-dimensional representation global random value pooling feature vector to obtain a cast metallographic multi-dimensional representation multi-scale pooling representation vector;

[0020] Perform per-channel semantic feature interaction and feature activation based on a fully connected layer on the cast metallographic multi-dimensional representation multi-scale pooling representation vector to obtain a cast metallographic multi-dimensional representation channel semantic association feature vector;

[0021] Input the cast metallographic multi-dimensional representation channel semantic association feature vector into a gated unit to perform masking processing on each position feature value in the cast metallographic multi-dimensional representation channel semantic association feature vector and then perform a normalization operation to obtain a normalized cast metallographic multi-dimensional representation semantic weight vector;

[0022] Use the feature values at each position in the normalized cast metallographic multi-dimensional representation semantic weight vector as weighting coefficients to perform weighted optimization on each feature matrix along the channel dimension in the cast metallographic multi-dimensional representation feature map to obtain the self-adaptive enhanced cast metallographic multi-dimensional representation feature map.

[0023] For example, in the cast image enhancement method for metallographic analysis according to an embodiment of the present disclosure, wherein calculating the position-wise weighted sum among the cast metallographic multi-dimensional representation global maximum pooling feature vector, the cast metallographic multi-dimensional representation global mean pooling feature vector, and the cast metallographic multi-dimensional representation global random value pooling feature vector to obtain a cast metallographic multi-dimensional representation multi-scale pooling representation vector includes:

[0024] Calculate the position-wise product between the cast metallographic multi-dimensional representation global maximum pooling feature vector and a first preset trainable weight value to obtain a weight-modulated cast metallographic multi-dimensional representation global maximum pooling feature vector;

[0025] Calculate the position-wise product between the cast metallographic multi-dimensional representation global mean pooling feature vector and a second preset trainable weight value to obtain a weight-modulated cast metallographic multi-dimensional representation global mean pooling feature vector;

[0026] Calculate the element-wise product between the globally randomly pooled feature vector of the cast metallographic multi-dimensional representation and the third preset trainable weight value to obtain the weight-modulated globally randomly pooled feature vector of the cast metallographic multi-dimensional representation;

[0027] Perform element-wise addition on the weight-modulated globally maximum pooled feature vector of the cast metallographic multi-dimensional representation, the weight-modulated globally mean pooled feature vector of the cast metallographic multi-dimensional representation, and the weight-modulated globally randomly pooled feature vector of the cast metallographic multi-dimensional representation to obtain the multi-scale pooled representation vector of the cast metallographic multi-dimensional representation.

[0028] For example, in the method for enhancing cast images for metallographic analysis according to an embodiment of the present disclosure, wherein the globally channel-semantically associated feature vector of the cast metallographic multi-dimensional representation is input into a gated unit, and after masking each position feature value in the globally channel-semantically associated feature vector of the cast metallographic multi-dimensional representation and then performing a normalization operation to obtain a normalized globally channel-semantically weighted vector of the cast metallographic multi-dimensional representation, it includes:

[0029] Mask each position feature value in the globally channel-semantically associated feature vector of the cast metallographic multi-dimensional representation based on a gating mechanism to obtain a masked globally channel-semantically associated feature vector of the cast metallographic multi-dimensional representation;

[0030] Normalize each position feature value in the masked globally channel-semantically associated feature vector of the cast metallographic multi-dimensional representation to obtain the normalized globally channel-semantically weighted vector of the cast metallographic multi-dimensional representation.

[0031] For example, in the method for enhancing cast images for metallographic analysis according to an embodiment of the present disclosure, wherein masking each position feature value in the globally channel-semantically associated feature vector of the cast metallographic multi-dimensional representation based on a gating mechanism to obtain a masked globally channel-semantically associated feature vector of the cast metallographic multi-dimensional representation includes: in response to each position feature value in the globally channel-semantically associated feature vector of the cast metallographic multi-dimensional representation being greater than or equal to a predetermined threshold, taking the original value of each position feature value in the globally channel-semantically associated feature vector of the cast metallographic multi-dimensional representation, otherwise setting it to zero.

[0032] For example, in the method for enhancing cast images for metallographic analysis according to an embodiment of the present disclosure, wherein determining an enhanced cast metallographic image based on the adaptively enhanced globally multi-dimensional representation feature map of the cast metallographic includes: inputting the adaptively enhanced globally multi-dimensional representation feature map of the cast metallographic into an image enhancer based on a generative adversarial network to obtain an enhanced cast metallographic image.

[0033] The casting image enhancement method for metallographic analysis according to the present disclosure collects casting metallographic images and uses image processing and analysis algorithms based on artificial intelligence and deep learning at the backend to analyze and identify features of the casting metallographic images, thereby capturing multi-dimensional representation features regarding casting metallography and performing enhancement processing on the casting metallographic images. In this way, it can help to more precisely analyze and control the inoculation process of ductile iron, providing reliable support for the production quality control of materials such as ductile iron. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] To more clearly illustrate the technical solutions of the embodiments of the present disclosure, the accompanying drawings of the embodiments of the present disclosure will be briefly introduced below. Obviously, the drawings described below only relate to some embodiments of the present disclosure and do not limit the present disclosure.

[0035] Figure 1 FIG. shows a schematic application architecture diagram of the casting image enhancement method for metallographic analysis in an embodiment of the present disclosure;

[0036] Figure 2 FIG. shows a flowchart of the casting image enhancement method for metallographic analysis in an embodiment of the present disclosure;

[0037] Figure 3 FIG. shows a flowchart of sub-step S540 of the casting image enhancement method for metallographic analysis in an embodiment of the present disclosure;

[0038] Figure 4 FIG. shows a flowchart of sub-step S542 of the casting image enhancement method for metallographic analysis in an embodiment of the present disclosure;

[0039] Figure 5 FIG. shows a schematic structural diagram of the casting image enhancement system for metallographic analysis in an embodiment of the present disclosure;

[0040] Figure 6 FIG. shows an application scenario diagram of the casting image enhancement method for metallographic analysis in an embodiment of the present disclosure; and

[0041] Figure 7 FIG. shows a schematic diagram of a storage medium according to an embodiment of the present disclosure. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0042] The technical solutions in the embodiments of the present disclosure will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some of the embodiments of the present disclosure, rather than all of them. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present disclosure without creative efforts also belong to the scope of protection of the present disclosure.

[0043] The terms used in this specification are those general terms that are currently widely used in the art in consideration of the functions of the present disclosure. However, these terms may vary according to the intention of those of ordinary skill in the art, precedents, or new technologies in the art. In addition, specific terms may be selected by the applicant, and in such cases, their detailed meanings will be described in the detailed description of the present disclosure. Therefore, the terms used in the specification should not be construed as mere names, but rather based on the meanings of the terms and the overall description of the present disclosure.

[0044] Although the present disclosure makes various references to certain modules in the system according to the embodiments of the present disclosure, any number of different modules may be used and run on the user terminal and / or the server. The modules are merely illustrative, and different aspects of the system and method may use different modules.

[0045] Flowcharts are used in the present disclosure to illustrate the operations performed by the system according to the embodiments of the present disclosure. It should be understood that the preceding or following operations are not necessarily executed precisely in order. Instead, various steps may be processed in reverse order or simultaneously as needed. At the same time, other operations may also be added to these processes, or one or more steps may be removed from these processes.

[0046] Figure 1 The application architecture diagram of the casting image enhancement method for metallographic analysis in the embodiments of the present disclosure is shown, including a server 100 and a terminal device 200.

[0047] The terminal device 200 and the server 100 can be connected through the Internet to achieve communication with each other. Optionally, the above Internet uses standard communication technologies and / or protocols. The Internet is usually the Internet, but can also be any network, including but not limited to any combination of a local area network (LAN), a metropolitan area network (MAN), a wide area network (WAN), a mobile, wired or wireless network, a private network or a virtual private network. In some embodiments, technologies and / or formats including Hyper Text Markup Language (HTML), Extensible Markup Language (XML), etc. are used to represent the data exchanged through the network. In addition, conventional encryption technologies such as Secure Socket Layer (SSL), Transport Layer Security (TLS), Virtual Private Network (VPN), Internet Protocol Security (IPsec), etc. can be used to encrypt all or some of the links. In other embodiments, customized and / or dedicated data communication technologies can also be used to replace or supplement the above data communication technologies.

[0048] The server 100 can provide various network services for the terminal device 200. Among them, the server 100 can be a single server, a server cluster composed of several servers, or a cloud computing center. Specifically, the server 100 can include a processor 110 (Central Processing Unit, CPU), a memory 120, an input device 130, an output device 140, etc. The input device 130 can include a keyboard, a mouse, a touch screen, etc. The output device 140 can include a display device, such as a liquid crystal display (LCD), a cathode ray tube (CRT), etc.

[0049] The memory 120 can include a read-only memory (ROM) and a random access memory (RAM), and provide the program instructions and data stored in the memory 120 to the processor 110. In the embodiments of the present disclosure, the memory 120 can be used to store the program of the casting image enhancement method for metallographic analysis in the embodiments of the present disclosure.

[0050] The processor 110 calls the program instructions stored in the memory 120, and the processor 110 is used to execute the steps of any one of the casting image enhancement methods for metallographic analysis in the embodiments of the present disclosure according to the obtained program instructions.

[0051] In addition, the application architecture diagram in the embodiment of the present disclosure is intended to more clearly illustrate the technical solution in the embodiment of the present disclosure, and does not constitute a limitation on the technical solution provided in the embodiment of the present disclosure. Of course, for other application architectures and business applications, the technical solution provided in the embodiment of the present disclosure is also applicable to similar problems.

[0052] The following is a non-limiting description of the casting image enhancement method for metallographic analysis provided according to at least one embodiment of the present disclosure through several examples or embodiments. As described below, different features in these specific examples or embodiments can be combined with each other without conflicting with each other to obtain new examples or embodiments, and these new examples or embodiments also fall within the scope of protection of the present disclosure.

[0053] The commonly used ductile iron grade is QT400-18AL. How to increase it to QT500-18AL at a lower cost, that is, to increase the tensile strength from 400 to 500, is the technical goal that is being worked on. In this process, a large number of tests are required, and microscopic analysis is also performed during the production process as a means of controlling production quality. This has generated a lot of test data, including a large number of metallographic microscopic analysis pictures, which were previously analyzed and processed manually with relatively low efficiency. Image recognition technology can significantly improve the processing efficiency of metallographic microscopic analysis, and can quickly, batch-wise, and automatically process a large number of metallographic microscopic analysis pictures, which helps to improve the test and production efficiency of ductile iron.

[0054] In response to the above technical problems, the technical concept of the present disclosure is to collect casting metallographic images, and use image processing and analysis algorithms based on artificial intelligence and deep learning in the back end to analyze and identify the features of the casting metallographic images, so as to capture the multi-dimensional representation features of the casting metallographic images, so as to enhance the casting metallographic images. In this way, higher quality analysis materials can be provided for the subsequent ductile iron inoculation process, which will help to analyze and control the ductile iron inoculation process more finely in the future, and provide reliable support for the production quality control of ductile iron and other materials.

[0055] Figure 2 The flowchart of the casting image enhancement method for metallographic analysis in the embodiment of the present disclosure is shown. For example, the casting image enhancement method for metallographic analysis can be executed by a server, which can be Figure 1 The server 100 shown in FIG. Figure 2As shown, a casting image enhancement method for metallographic analysis according to an embodiment of the present disclosure includes the steps of: S510, obtaining a casting metallographic image to be processed; S520, performing color space transformation on the casting metallographic image to be processed to obtain a YCbCr casting metallographic image to be processed; S530, respectively performing feature extraction on the casting metallographic image to be processed and the YCbCr casting metallographic image to be processed to obtain a casting metallographic RGB feature map and a casting metallographic YCbCr feature map; S540, performing feature fusion processing based on gated screening adaptive enhancement on the casting metallographic RGB feature map and the casting metallographic YCbCr feature map to obtain an adaptively enhanced casting metallographic multi-dimensional representation feature map; S550, determining an enhanced casting metallographic image based on the adaptively enhanced casting metallographic multi-dimensional representation feature map.

[0056] Specifically, in the technical solution of the present disclosure, first, a casting metallographic image to be processed is obtained. Next, considering that in the process of processing the casting metallographic image to be processed, the characteristics of the casting metallographic structure are often related to the brightness and chromaticity of the image. Therefore, in order to better analyze the brightness information and chromaticity information in the casting metallographic image to be processed, it is necessary to separate the two, which is conducive to more effectively performing feature extraction and enhancement of the casting metallographic image. Based on this, in the technical solution of the present disclosure, color space transformation is performed on the casting metallographic image to be processed to obtain a YCbCr casting metallographic image to be processed. It should be understood that since the YCbCr color space is a color coding method that separates brightness (Y) and chromaticity (Cb, Cr) information. In the YCbCr color space, the Y channel contains the brightness information of the image, and the Cb and Cr channels contain the chromaticity information of the image. By converting the original casting metallographic image to be processed into the YCbCr color space, the brightness information (Y channel) and chromaticity information (Cb, Cr channels) can be processed separately. This can focus more on capturing the detailed information of the image, which helps to better retain or enhance the texture, edges, and color information of the image during the image enhancement process, providing a better basis for subsequent metallographic feature extraction and image enhancement, making the enhancement process of the casting metallographic image more flexible and accurate.

[0057] Then, considering that the basic color information and texture features of the casting metallographic image can be reflected, which is helpful for analyzing the color distribution and morphological structure of the metallographic structure, and the Y channel in the YCbCr to-be-processed casting metallographic image retains the luminance information of the image, while the Cb and Cr channels contain the chrominance information. Using the YCbCr color space can better separate the luminance and chrominance information, improving the discrimination and expression ability of features. Based on this, in order to more comprehensively capture the information of the casting metallographic structure, thereby improving the expression ability and richness of the casting metallographic features, in the technical solution of the present disclosure, the to-be-processed casting metallographic image and the YCbCr to-be-processed casting metallographic image are further input into a metallographic feature extractor based on a dilated convolutional neural network model to respectively extract the implicit multi-scale metallographic semantic features in the to-be-processed casting metallographic image and the YCbCr to-be-processed casting metallographic image, so as to obtain a casting metallographic RGB feature map and a casting metallographic YCbCr feature map. It should be understood that the dilated convolutional neural network can allow the network to extract features at different scales by changing the dilation rate, which means that it can simultaneously capture the local details and global structure information in the image. This is particularly important for metallographic image analysis because the microstructures and defects of the casting may be presented at different scales. Therefore, through the processing of the metallographic feature extractor based on the dilated convolutional neural network model, the complex feature representation of the metallographic structure can be learned and captured, so as to better distinguish different nodular cast iron metallographic structures and details, which is helpful for generating enhanced casting metallographic images for more refined analysis and control of the subsequent nodular cast iron inoculation process.

[0058] Correspondingly, in step S530, feature extraction is respectively performed on the to-be-processed casting metallographic image and the YCbCr to-be-processed casting metallographic image to obtain a casting metallographic RGB feature map and a casting metallographic YCbCr feature map, including: inputting the to-be-processed casting metallographic image and the YCbCr to-be-processed casting metallographic image into a metallographic feature extractor based on a dilated convolutional neural network model to obtain the casting metallographic RGB feature map and the casting metallographic YCbCr feature map.

[0059] Furthermore, since the cast metallographic RGB feature map contains the color and brightness information in the cast metallographic image to be processed, and the cast metallographic YCbCr feature map focuses more on color and texture details, therefore, the cast metallographic RGB feature map and the cast metallographic YCbCr feature map are further fused to obtain a cast metallographic multi-dimensional representation feature map. By combining the cast metallographic RGB and YCbCr feature maps, the advantages of different types of image spatial features can be integrated to achieve information complementarity, thereby obtaining a more comprehensive image feature representation. In this way, a more comprehensive and rich metallographic feature representation can be provided, which helps to accurately describe the morphology, color, and texture features of the metallographic structure, so as to facilitate multi-faceted and detailed enhancement of the cast metallographic image in the subsequent process, and improve the accuracy and reliability of metallographic analysis.

[0060] It should be understood that the cast metallographic multi-dimensional representation feature map contains the metallographic feature representation after the fusion of different types of image features, and can describe the morphology, color, and texture features of the metallographic structure more comprehensively and accurately. However, some of this information contributes more to the expression of metallographic features, while there are still some redundant and less correlated features. Therefore, in order to strengthen and screen the important information in the cast metallographic multi-dimensional representation feature map, thereby improving the quality and information expression ability of the metallographic image, in the technical solution of the present disclosure, the cast metallographic multi-dimensional representation feature map is further input into a gated screening adaptive attention module to obtain an adaptively enhanced cast metallographic multi-dimensional representation feature map. It should be understood that in metallographic analysis, different material microstructures may require different enhancement strategies, and the adaptive attention mechanism can provide this flexibility. Specifically, through the processing of the gated screening adaptive attention module, the correlation relationships and features between different contents in the metallographic image can be automatically learned and captured, so as to automatically adjust its focus according to the different contents of the metallographic image. This enables the image enhancement process to be optimized according to the specific regions and features of the image. That is to say, the adaptive attention mechanism can dynamically adjust the weights according to the importance in the feature map, making the network pay more attention to the features that are important for the current task. Therefore, by applying the gated screening adaptive attention module, on the basis of retaining important features, features that are irrelevant or have a large noise interference to the cast metallographic image characterization and enhancement task can be weakened or filtered out, thereby improving the perception ability and distinguishability of the key features in the metallographic image, highlighting the key details and structural features of the metallographic structure, and providing more powerful support for subsequent image enhancement and analysis.

[0061] Correspondingly, in step S540, as Figure 3As shown, perform feature fusion processing based on gated screening adaptive enhancement on the cast metallographic RGB feature map and the cast metallographic YCbCr feature map to obtain an adaptively enhanced multi-dimensional representation feature map of the cast metallography, including: S541, fuse the cast metallographic RGB feature map and the cast metallographic YCbCr feature map to obtain a multi-dimensional representation feature map of the cast metallography; S542, input the multi-dimensional representation feature map of the cast metallography into a gated screening adaptive attention module to obtain the adaptively enhanced multi-dimensional representation feature map of the cast metallography.

[0062] Specifically, in step S542, as Figure 4 shown, input the multi-dimensional representation feature map of the cast metallography into a gated screening adaptive attention module to obtain the adaptively enhanced multi-dimensional representation feature map of the cast metallography, including: S5421, perform global pooling processing based on the maximum value, global pooling processing based on random values, and global pooling processing based on the average value on the multi-dimensional representation feature map of the cast metallography to obtain a global maximum pooling feature vector of the multi-dimensional representation of the cast metallography, a global mean pooling feature vector of the multi-dimensional representation of the cast metallography, and a global random value pooling feature vector of the multi-dimensional representation of the cast metallography; S5422, calculate the position-wise weighted sum between the global maximum pooling feature vector of the multi-dimensional representation of the cast metallography, the global mean pooling feature vector of the multi-dimensional representation of the cast metallography, and the global random value pooling feature vector of the multi-dimensional representation of the cast metallography to obtain a multi-scale pooling representation vector of the multi-dimensional representation of the cast metallography; S5423, perform per-channel semantic feature interaction and feature activation based on a fully connected layer on the multi-scale pooling representation vector of the multi-dimensional representation of the cast metallography to obtain a channel semantic association feature vector of the multi-dimensional representation of the cast metallography; S5424, input the channel semantic association feature vector of the multi-dimensional representation of the cast metallography into a gated unit to perform masking processing on each position feature value in the channel semantic association feature vector of the multi-dimensional representation of the cast metallography and then perform a normalization operation to obtain a normalized semantic weight vector of the multi-dimensional representation of the cast metallography; S5425, use the feature values at each position in the normalized semantic weight vector of the multi-dimensional representation of the cast metallography as weighting coefficients to perform weighted optimization on each feature matrix along the channel dimension in the multi-dimensional representation feature map of the cast metallography to obtain the adaptively enhanced multi-dimensional representation feature map of the cast metallography.

[0063] Among them, in step S5422, calculating the position-wise weighted sum between the global maximum pooling feature vector of the cast microstructure multi-dimensional representation, the global average pooling feature vector of the cast microstructure multi-dimensional representation, and the global random value pooling feature vector of the cast microstructure multi-dimensional representation to obtain the multi-scale pooling representation vector of the cast microstructure multi-dimensional representation includes: calculating the position-wise product between the global maximum pooling feature vector of the cast microstructure multi-dimensional representation and a first preset trainable weight value to obtain the weight-modulated global maximum pooling feature vector of the cast microstructure multi-dimensional representation; calculating the position-wise product between the global average pooling feature vector of the cast microstructure multi-dimensional representation and a second preset trainable weight value to obtain the weight-modulated global average pooling feature vector of the cast microstructure multi-dimensional representation; calculating the position-wise product between the global random value pooling feature vector of the cast microstructure multi-dimensional representation and a third preset trainable weight value to obtain the weight-modulated global random value pooling feature vector of the cast microstructure multi-dimensional representation; and performing position-wise addition on the weight-modulated global maximum pooling feature vector of the cast microstructure multi-dimensional representation, the weight-modulated global average pooling feature vector of the cast microstructure multi-dimensional representation, and the weight-modulated global random value pooling feature vector of the cast microstructure multi-dimensional representation to obtain the multi-scale pooling representation vector of the cast microstructure multi-dimensional representation.

[0064] Among them, in step S5424, inputting the channel semantic association feature vector of the cast microstructure multi-dimensional representation into a gated unit to perform masking processing on each position feature value in the channel semantic association feature vector of the cast microstructure multi-dimensional representation and then performing a normalization operation to obtain a normalized semantic weight vector of the cast microstructure multi-dimensional representation includes: performing masking processing on each position feature value in the channel semantic association feature vector of the cast microstructure multi-dimensional representation based on a gating mechanism to obtain a masked channel semantic association feature vector of the cast microstructure multi-dimensional representation; and performing normalization processing on each position feature value in the masked channel semantic association feature vector of the cast microstructure multi-dimensional representation to obtain the normalized semantic weight vector of the cast microstructure multi-dimensional representation. Specifically, performing masking processing on each position feature value in the channel semantic association feature vector of the cast microstructure multi-dimensional representation based on a gating mechanism to obtain a masked channel semantic association feature vector of the cast microstructure multi-dimensional representation includes: in response to each position feature value in the channel semantic association feature vector of the cast microstructure multi-dimensional representation being greater than or equal to a predetermined threshold, taking the original value of each position feature value in the channel semantic association feature vector of the cast microstructure multi-dimensional representation, otherwise setting it to zero.

[0065] More specifically, in one example, inputting the cast metallographic multi-dimensional representation feature map into the gated screening adaptive attention module to obtain the adaptive enhanced cast metallographic multi-dimensional representation feature map includes: inputting the cast metallographic multi-dimensional representation feature map into the gated screening adaptive attention module and processing it according to the following gated screening adaptive enhancement formula to obtain the adaptive enhanced cast metallographic multi-dimensional representation feature map; where the gated screening adaptive enhancement formula is:

[0066] V c = α·AvgPool(F)+β·MaxPool(F)+γ·StoPool(F)

[0067] A = σ(W a -V c +B a )

[0068] A m = mask(A)

[0069]

[0070] F′ = A′⊙F

[0071] Where F is the cast metallographic multi-dimensional representation feature map, AvgPool(·), MaxPool(·), and StoPool(·) are respectively performing global mean pooling, maximum pooling, and random value pooling on each feature matrix along the channel dimension of the feature map, α, β, and γ are all preset trainable weight values, V c is the cast metallographic multi-dimensional representation multi-scale pooling representation vector, W a and B a are the weight matrix and the bias vector respectively, σ is the sigmoid activation function, A is the cast metallographic multi-dimensional representation channel semantic association feature vector, mask(·) represents masking processing, A m is the masked cast metallographic multi-dimensional representation channel semantic association feature vector, is the feature value at the i-th position in the masked cast metallographic multi-dimensional representation channel semantic association feature vector, n represents the number of feature values in the masked cast metallographic multi-dimensional representation channel semantic association feature vector, A i ′ is the feature value at the i-th position in the normalized cast metallographic multi-dimensional representation semantic weight vector, A ′ is the normalized cast metallographic multi-dimensional representation semantic weight vector, ⊙ represents weighted multiplication along the channel dimension, F ′ is the adaptive enhanced cast metallographic multi-dimensional representation feature map.

[0072] Furthermore, input the multi-dimensional representation feature map of the adaptively enhanced casting metallographic structure into an image enhancer based on a generative adversarial network to obtain an enhanced casting metallographic image. That is to say, use the multi-dimensional representation features of the adaptively enhanced casting metallographic structure to generate an enhanced casting metallographic image, and thus perform adaptive enhancement processing on the casting metallographic image according to the semantic content of the important parts in the casting metallographic image. In this way, it is possible to provide higher-quality analysis materials for the subsequent inoculation process of ductile iron, which helps to analyze and control the inoculation process of ductile iron more precisely and provides reliable support for the production quality control of materials such as ductile iron.

[0073] Correspondingly, in step S550, based on the multi-dimensional representation feature map of the adaptively enhanced casting metallographic structure, determining the enhanced casting metallographic image includes: inputting the multi-dimensional representation feature map of the adaptively enhanced casting metallographic structure into an image enhancer based on a generative adversarial network to obtain an enhanced casting metallographic image.

[0074] In the embodiments of the present disclosure, the RGB feature map of the casting metallographic structure and the YCbCr feature map of the casting metallographic structure respectively represent the image semantic features of the to-be-processed casting metallographic image and the YCbCr to-be-processed casting metallographic image based on a dilated convolutional neural network model. When performing feature fusion and gated screening for adaptive attention enhancement, local fusion overflow features may be caused by differences in attention weights, thereby affecting the accuracy of the enhanced casting metallographic image obtained by inputting the multi-dimensional representation feature map of the adaptively enhanced casting metallographic structure into an image enhancer based on a generative adversarial network.

[0075] Preferably, inputting the multi-dimensional representation feature map of the adaptively enhanced casting metallographic structure into an image enhancer based on a generative adversarial network to obtain an enhanced casting metallographic image includes the following steps:

[0076] Calculate the L1 norm aggregation value of each eigenvalue of the multi-dimensional representation feature map of the adaptively enhanced casting metallographic structure to obtain the first multi-dimensional reference structure quantity of the adaptively enhanced casting metallographic structure, expressed as:

[0077]

[0078] n = W × H × C

[0079] where f i represents each eigenvalue of the multi-dimensional representation feature map of the adaptively enhanced casting metallographic structure, n represents the scale of the multi-dimensional representation feature map of the adaptively enhanced casting metallographic structure, represents the first multi-dimensional reference structure quantity of the adaptively enhanced casting metallographic structure, and W, H, and C respectively represent the width, height, and number of channels of the multi-dimensional representation feature map of the adaptively enhanced casting metallographic structure;

[0080] Calculate the L2 norm aggregation value of each eigenvalue of the adaptive enhanced casting metallographic multi-dimensional representation feature map to obtain the second adaptive enhanced casting metallographic multi-dimensional reference structure quantity, expressed as:

[0081]

[0082] where ω represents the second adaptive enhanced casting metallographic multi-dimensional reference structure quantity;

[0083] Multiply each eigenvalue of the adaptive enhanced casting metallographic multi-dimensional representation feature map by the first adaptive enhanced casting metallographic multi-dimensional reference structure quantity and the second adaptive enhanced casting metallographic multi-dimensional reference structure quantity respectively to obtain the first adaptive enhanced casting metallographic multi-dimensional structure correlation factor and the second adaptive enhanced casting metallographic multi-dimensional structure correlation factor corresponding to each eigenvalue;

[0084] Multiply each eigenvalue of the adaptive enhanced casting metallographic multi-dimensional representation feature map by the scale of the adaptive enhanced casting metallographic multi-dimensional representation feature map and the square root of the scale respectively to obtain the first adaptive enhanced casting metallographic multi-dimensional scale evolution quantity and the second adaptive enhanced casting metallographic multi-dimensional scale evolution quantity corresponding to each eigenvalue;

[0085] Divide the first adaptive enhanced casting metallographic multi-dimensional structure correlation factor by the difference between the first adaptive enhanced casting metallographic multi-dimensional reference structure quantity and the first adaptive enhanced casting metallographic multi-dimensional scale evolution quantity to obtain the first adaptive enhanced casting metallographic multi-dimensional transformation modulation coefficient, expressed as:

[0086]

[0087] where f 1i represents the first adaptive enhanced casting metallographic multi-dimensional transformation modulation coefficient;

[0088] Divide the second adaptive enhanced casting metallographic multi-dimensional structure correlation factor by the difference between the second adaptive enhanced casting metallographic multi-dimensional reference structure quantity and the second adaptive enhanced casting metallographic multi-dimensional scale evolution quantity to obtain the second adaptive enhanced casting metallographic multi-dimensional transformation modulation coefficient, expressed as:

[0089]

[0090] where f 2i represents the second adaptive enhanced casting metallographic multi-dimensional transformation modulation coefficient;

[0091] Calculate the weighted sum of the first adaptive enhanced casting metallographic multi-dimensional transformation modulation coefficient and the second adaptive enhanced casting metallographic multi-dimensional transformation modulation coefficient to obtain each eigenvalue of the optimized adaptive enhanced casting metallographic multi-dimensional representation feature map, expressed as:

[0092]

[0093] where denotes element-wise addition, τ denotes the weighted hyperparameter, ⊙ denotes element-wise multiplication, F1 denotes the feature map composed of the first adaptive enhanced casting metallographic multi-dimensional transformation modulation coefficient, F2 denotes the feature map composed of the second adaptive enhanced casting metallographic multi-dimensional transformation modulation coefficient, and F′ denotes the optimized adaptive enhanced casting metallographic multi-dimensional representation feature map;

[0094] Pass the optimized adaptive enhanced casting metallographic multi-dimensional representation feature map through an image enhancer based on a generative adversarial network to obtain an enhanced casting metallographic image.

[0095] That is, by performing high-dimensional characterization analysis on the geometric topological features of the adaptive enhanced casting metallographic multi-dimensional representation feature map, constructing a benchmark observation domain using a standardized spatial form descriptor, and implementing a cross-resolution domain transformation operation to perform multi-granularity reconstruction on the eigenvalues. Combining a focus weight modulation mechanism to dynamically calibrate the spatial correlation of the feature vectors, ensuring geometric deformation robustness (covering translation, scaling, and rotation transformations) during the feature space interaction process, effectively improving the generation modeling regression quality of the adaptive enhanced casting metallographic multi-dimensional representation feature map under complex geometric topological conditions, and significantly enhancing the optimization efficiency and generalization performance. In this way, robust convergence of the adaptive enhanced casting metallographic multi-dimensional representation feature map to the target generation regression domain is achieved, and the image generation quality of the enhanced casting metallographic image obtained by passing the adaptive enhanced casting metallographic multi-dimensional representation feature map through an image enhancer based on a generative adversarial network is improved. In this way, the enhancement process of the casting metallographic image can be carried out more accurately, providing higher-quality analysis materials for the subsequent nodular cast iron inoculation process, thus contributing to more refined analysis and control of the nodular cast iron inoculation process and providing reliable support for the production quality control of materials such as nodular cast iron.

[0096] Based on the above embodiments, refer to Figure 5As shown in the figure, it is a schematic structural diagram of a casting image enhancement system 800 for metallographic analysis in an embodiment of the present disclosure. The casting image enhancement system 800 for metallographic analysis includes: an image acquisition module 810, configured to acquire a casting metallographic image to be processed; a color space transformation module 820, configured to perform color space transformation on the casting metallographic image to be processed to obtain a YCbCr casting metallographic image to be processed; a feature extraction module 830, configured to perform feature extraction on the casting metallographic image to be processed and the YCbCr casting metallographic image to be processed respectively to obtain a casting metallographic RGB feature map and a casting metallographic YCbCr feature map; a feature enhancement and fusion module 840, configured to perform feature fusion processing based on gated screening and adaptive enhancement on the casting metallographic RGB feature map and the casting metallographic YCbCr feature map to obtain an adaptively enhanced multi-dimensional representation feature map of the casting metallographic; an enhancement module 850, configured to determine an enhanced casting metallographic image based on the adaptively enhanced multi-dimensional representation feature map of the casting metallographic.

[0097] Here, those skilled in the art can understand that the specific functions and operations of each module in the above-mentioned casting image enhancement system 800 for metallographic analysis have been introduced in detail in the description of the casting image enhancement method for metallographic analysis above, and therefore, the repeated description thereof will be omitted. Figures 2 to 4 in the description of the casting image enhancement method for metallographic analysis, and therefore, the repeated description thereof will be omitted.

[0098] Figure 6 is an application scenario diagram of the casting image enhancement method for metallographic analysis according to an embodiment of the present disclosure. As Figure 6 shown, in this application scenario, first, a casting metallographic image to be processed (for example, Figure 6 D shown in the figure) is acquired, and then, the casting metallographic image to be processed is input into a server (for example, Figure 6 S shown in the figure) deployed with a casting image enhancement algorithm for metallographic analysis, where the server can use the casting image enhancement algorithm for metallographic analysis to process the casting metallographic image to be processed to obtain an enhanced casting metallographic image.

[0099] Based on the above embodiments, an electronic device in another exemplary implementation manner is further provided in an embodiment of the present disclosure. In some possible implementation manners, the electronic device in an embodiment of the present disclosure may include a memory, a processor, and a computer program stored on the memory and executable on the processor. When the processor executes the program, the steps of the casting image enhancement method for metallographic analysis in the above embodiments can be implemented.

[0100] For example, taking the electronic device as the present disclosure Figure 1Taking the server 100 in [it] as an example for illustration, the processor in this electronic device is the processor 110 in the server 100, and the memory in this electronic device is the memory 120 in the server 100.

[0101] Embodiments of the present disclosure also provide a computer-readable storage medium. Figure 7 The schematic diagram of a computer-readable storage medium 1000 according to an embodiment of the present disclosure is shown. As Figure 7 shown, computer-executable instructions 1001 are stored on the computer-readable storage medium 1000. When the computer-executable instructions 1001 are run by a processor, the casting image enhancement method for metallographic analysis according to the embodiments of the present disclosure described with reference to the above drawings can be executed. The computer-readable storage medium includes but is not limited to, for example, volatile memory and / or non-volatile memory. The volatile memory may include, for example, random access memory (RAM) and / or cache memory, etc. The non-volatile memory may include, for example, read-only memory (ROM), hard disk, flash memory, etc.

[0102] Embodiments of the present disclosure also provide a computer program product or a computer program. The computer program product or the computer program includes computer-executable instructions, and the computer-executable instructions are stored in a computer-readable storage medium. The processor of a computer device reads the computer-executable instructions from the computer-readable storage medium, and the processor executes the computer-executable instructions, so that the computer device executes the casting image enhancement method for metallographic analysis according to the embodiments of the present disclosure.

[0103] Those skilled in the art can understand that the content disclosed in the present disclosure can have various variations and improvements. For example, the various devices or components described above can be implemented by hardware, or can be implemented by software, firmware, or some or all of the combinations of the three.

[0104] In addition, although the present disclosure makes various references to certain units in the systems according to the embodiments of the present disclosure, however, any number of different units can be used and run on the client and / or the server. The units are only illustrative, and different aspects of the system and method can use different units.

[0105] Those of ordinary skill in the art will understand that all or part of the steps in the above methods can be completed by instructing relevant hardware through a program, and the program can be stored in a computer-readable storage medium, such as a read-only memory, a magnetic disk, or an optical disc, etc. Optionally, all or part of the steps of the above embodiments can also be implemented using one or more integrated circuits. Correspondingly, each module / unit in the above embodiments can be implemented in the form of hardware or in the form of a software function module. The present disclosure is not limited to any specific combination of hardware and software.

[0106] Unless otherwise defined, all terms (including technical and scientific terms) used herein have the same meaning as commonly understood by those of ordinary skill in the art to which this disclosure belongs. It should also be understood that terms such as those defined in a commonly used dictionary should be interpreted as having a meaning consistent with their meaning in the context of the relevant art, and should not be interpreted in an idealized or overly formal sense unless explicitly defined as such herein.

[0107] The above is an explanation of the present disclosure and should not be regarded as a limitation thereof. Although several exemplary embodiments of the present disclosure have been described, those skilled in the art will readily understand that many modifications can be made to the exemplary embodiments without departing from the novel teachings and advantages of the present disclosure. Therefore, all such modifications are intended to be included within the scope of the present disclosure as defined by the claims. It should be understood that the above is an explanation of the present disclosure and should not be considered limited to the specific embodiments disclosed, and modifications to the disclosed embodiments and other embodiments are intended to be included within the scope of the appended claims. The present disclosure is defined by the claims and their equivalents.

Claims

1. A casting image enhancement method for metallographic analysis, characterized in that: include: Acquire metallographic images of castings to be processed; Performing color space transformation on the to-be-processed casting metallographic image to obtain a YCbCr to-be-processed casting metallographic image; Performing feature extraction on the to-be-processed casting metallographic image and the to-be-processed YCbCr casting metallographic image to obtain a casting metallographic RGB feature map and a casting metallographic YCbCr feature map respectively; Performing feature fusion processing based on gated screening adaptive enhancement on the casting metallographic RGB feature map and the casting metallographic YCbCr feature map to obtain an adaptive enhanced casting metallographic multi-dimensional representation feature map; Based on the adaptive enhanced casting metallographic multi-dimensional representation feature map, an enhanced casting metallographic image is determined.

2. The casting image enhancement method for metallographic analysis according to claim 1, characterized in that: The to-be-processed casting metallographic image and the to-be-processed YCbCr casting metallographic image are respectively subjected to feature extraction to obtain a casting metallographic RGB feature map and a casting metallographic YCbCr feature map, including: inputting the to-be-processed casting metallographic image and the to-be-processed YCbCr casting metallographic image into a metallographic feature extractor based on a void convolutional neural network model to obtain the casting metallographic RGB feature map and the casting metallographic YCbCr feature map.

3. The casting image enhancement method for metallographic analysis according to claim 2, characterized in that: The casting metallographic RGB feature map and the casting metallographic YCbCr feature map are subjected to feature fusion processing based on gated screening adaptive enhancement to obtain an adaptive enhanced casting metallographic multi-dimensional representation feature map, including: The casting metallographic RGB feature map and the casting metallographic YCbCr feature map are merged to obtain a casting metallographic multi-dimensional representation feature map; The casting metallographic multi-dimensional representation feature map is input into a gated screening adaptive attention module to obtain the adaptive enhanced casting metallographic multi-dimensional representation feature map.

4. The casting image enhancement method for metallographic analysis according to claim 3, characterized in that: Inputting the casting metallographic multi-dimensional representation feature map into the gated screening adaptive attention module to obtain the adaptive enhanced casting metallographic multi-dimensional representation feature map, including: Performing global pooling processing based on maximum value, global pooling processing based on random value and global pooling processing based on average value on the casting metallographic multi-dimensional representation feature map to obtain a casting metallographic multi-dimensional representation global maximum value pooling feature vector, a casting metallographic multi-dimensional representation global mean value pooling feature vector and a casting metallographic multi-dimensional representation global random value pooling feature vector; Calculating the position-weighted sum of the global maximum pooling feature vector of the casting metallographic multi-dimensional representation, the global mean pooling feature vector of the casting metallographic multi-dimensional representation, and the global random value pooling feature vector of the casting metallographic multi-dimensional representation to obtain a casting metallographic multi-dimensional representation multi-scale pooling representation vector; Performing channel-by-channel semantic feature interaction and feature activation based on a fully connected layer on the multi-scale pooling representation vector of the casting metallographic multi-dimensional representation to obtain a channel semantic association feature vector of the casting metallographic multi-dimensional representation; Inputting the casting metallographic multi-dimensional representation channel semantic association feature vector into a gating unit, performing masking processing on each position feature value in the casting metallographic multi-dimensional representation channel semantic association feature vector and then performing a normalization operation to obtain a normalized casting metallographic multi-dimensional representation semantic weight vector; The eigenvalues ​​of each position in the normalized casting metallographic multi-dimensional representation semantic weight vector are used as weighting coefficients, and weighted optimization is performed on each feature matrix along the channel dimension in the casting metallographic multi-dimensional representation feature map to obtain the adaptive enhanced casting metallographic multi-dimensional representation feature map.

5. The casting image enhancement method for metallographic analysis according to claim 4, characterized in that: Calculating the position-weighted sum of the global maximum pooling feature vector of the casting metallographic multi-dimensional representation, the global mean pooling feature vector of the casting metallographic multi-dimensional representation, and the global random value pooling feature vector of the casting metallographic multi-dimensional representation to obtain the casting metallographic multi-dimensional representation multi-scale pooling representation vector, including: Calculating the positional product between the casting metallographic multi-dimensional representation global maximum pooling feature vector and the first preset trainable weight value to obtain the weight-modulated casting metallographic multi-dimensional representation global maximum pooling feature vector; Calculating the positional product between the casting metallographic multi-dimensional representation global mean pooling feature vector and a second preset trainable weight value to obtain the weight-modulated casting metallographic multi-dimensional representation global mean pooling feature vector; Calculating the positional product between the casting metallographic multi-dimensional representation global random value pooling feature vector and a third preset trainable weight value to obtain the weight-modulated casting metallographic multi-dimensional representation global random value pooling feature vector; The global maximum pooling feature vector of the multi-dimensional representation of the casting metallographic phase after weight modulation, the global mean pooling feature vector of the multi-dimensional representation of the casting metallographic phase after weight modulation, and the global random value pooling feature vector of the multi-dimensional representation of the casting metallographic phase after weight modulation are added by position to obtain the multi-scale pooling representation vector of the multi-dimensional representation of the casting metallographic phase.

6. The casting image enhancement method for metallographic analysis according to claim 5, characterized in that: The casting metallographic multi-dimensional representation channel semantic association feature vector is input into the gating unit, so as to perform masking processing on each position feature value in the casting metallographic multi-dimensional representation channel semantic association feature vector and then perform normalization operation to obtain a normalized casting metallographic multi-dimensional representation semantic weight vector, including: Based on the gating mechanism, masking is performed on each position feature value in the casting metallographic multi-dimensional representation channel semantic association feature vector to obtain a masked casting metallographic multi-dimensional representation channel semantic association feature vector; Each position feature value in the masked casting metallographic multi-dimensional representation channel semantic association feature vector is normalized to obtain the normalized casting metallographic multi-dimensional representation semantic weight vector.

7. The casting image enhancement method for metallographic analysis according to claim 6, characterized in that: Based on the gating mechanism, each position feature value in the semantic association feature vector of the casting metallographic multi-dimensional representation channel is masked to obtain a masked casting metallographic multi-dimensional representation channel semantic association feature vector, including: in response to each position feature value in the semantic association feature vector of the casting metallographic multi-dimensional representation channel being greater than or equal to a predetermined threshold, each position feature value in the semantic association feature vector of the casting metallographic multi-dimensional representation channel is taken to be the original value, otherwise it is set to zero.

8. The casting image enhancement method for metallographic analysis according to claim 7, characterized in that: Based on the adaptive enhanced casting metallographic multi-dimensional representation feature map, an enhanced casting metallographic image is determined, including: inputting the adaptive enhanced casting metallographic multi-dimensional representation feature map into an image enhancer based on a generative adversarial network to obtain an enhanced casting metallographic image.

Images