Catalyst image expert knowledge base construction method and device
By building a catalyst image expert knowledge base and utilizing electron microscope images and activity prediction models, the problem of material classification in transmission electron microscope images has been solved, efficient image recognition and accurate material classification have been achieved, supporting catalyst research and development.
Patent Information
- Application Number
- CN202410312502.1
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-03-19
- Publication Date
- 2025-09-19
AI Technical Summary
In the existing technology, it is difficult to accurately classify materials in transmission electron microscope images. Traditional text-based image retrieval methods are inefficient, require a lot of human resources, and have difficulty in accurately classifying unknown materials.
Construct a catalyst image expert knowledge base. By collecting electron microscope images, establish a catalyst activity prediction model, perform feature identification, generate an expert knowledge base, and use a similarity threshold to compare the catalyst images to be tested and identify similar images.
It improves the accuracy and efficiency of catalyst image recognition, reduces human resource waste, enriches material research directions, and supports subsequent catalyst research and development.
Smart Images
Figure CN120671785A_ABST
Abstract
Description
Technical Field
[0001] The present invention belongs to the technical field of image detection, and relates to a method and device for constructing a catalyst image expert knowledge base under catalyst electron microscope images. Background Art
[0002] The establishment of an expert database is fundamental and serves as a vehicle for storing expert information. It can provide strong talent and intellectual support for subsequent R&D work on the project. By establishing an expert knowledge base based on electron microscope catalyst images, sample information can be more comprehensively and accurately covered. Subsequently, through image retrieval, new catalyst images can be compared with images in the database to form prior knowledge, laying the foundation for subsequent catalyst R&D work, facilitating R&D workers' image analysis work, and improving the accuracy of subsequent catalyst R&D images.
[0003] With the rapid development and application of computers, networks, and multimedia technologies, the number of digital images is growing at an alarming rate. Content-based image retrieval technology can effectively solve the problem of retrieving relevant images from image databases. Traditional retrieval methods are text-based, which involves semantically describing all images in a digital image database, assigning a label and establishing a direct mapping between text and image. This transforms image searches into text searches, ultimately using matching algorithms to implement the text search process.
[0004] Transmission electron microscopy (TEM) was originally used to study cell biology. However, in recent years, TEM has also been widely used in the examination of cultural relics, crystal analysis, and the microelectronics industry. The wide variety of materials and their diverse microstructures greatly increases the difficulty of accurately classifying them. In real life, classifying unknown materials based on TEM images requires a considerable amount of time and effort from experts in various fields. Therefore, building an automated material classification system based on TEM images can not only reduce the waste of human resources but also enrich research in this area. Summary of the Invention
[0005] In response to the shortcomings of the existing technology, the present invention proposes a method and device for constructing an expert knowledge base of catalyst images. The present invention constructs an expert knowledge base of catalysts, compares new catalyst images with images in the library, forms prior knowledge, and improves image recognition accuracy.
[0006] In order to achieve the above objectives, the present invention provides a method for constructing a catalyst image expert knowledge base, comprising:
[0007] Collect electron microscope images of catalysts with different activities and obtain images of control samples;
[0008] Constructing a catalyst activity prediction model, and using the catalyst activity prediction model to perform feature identification on the control sample image, establishing a catalyst data set, and generating an expert knowledge base;
[0009] The catalyst image to be tested is compared with the images in the expert knowledge base, and when the similarity exceeds a preset similarity threshold, the catalyst image to be tested and / or the corresponding control sample image are identified.
[0010] In some embodiments, feature extraction is performed on the catalyst image to obtain a second feature;
[0011] The second feature is compared with the first feature, and when the similarity exceeds a first similarity threshold, the catalyst image to be tested is identified, wherein the first feature is a feature obtained by feature identification of the control sample image, and the preset similarity threshold includes the first similarity threshold.
[0012] In some embodiments, feature extraction is performed on the catalyst image to obtain a second feature;
[0013] The second feature is compared with the first feature. When the similarity exceeds a second similarity threshold, a corresponding control sample image is identified from the expert knowledge base, where the first feature is a feature obtained by feature identification of the control sample image, and the preset similarity threshold includes the second similarity threshold.
[0014] In some embodiments, the second feature is compared with the first feature using a Euclidean distance method.
[0015] In some embodiments, the 12-dimensional feature vector of each channel of the catalyst image to be tested is sequentially subtracted from the 12-dimensional feature vector of each channel of the image in the expert knowledge base to obtain a difference of 12 feature vectors;
[0016] The difference of each eigenvector is squared and then multiplied by the first weight of each eigenvector to obtain the distance value of each channel.
[0017] The distance value of each channel is multiplied by the second weight value occupied by the channel and then weighted to obtain a target distance, which represents the similarity between the catalyst image to be tested and the image in the expert knowledge base.
[0018] In some embodiments, before performing feature identification on the control sample image, the method further includes performing enhancement processing on the control sample image.
[0019] In some embodiments, a bilinear interpolation method is used to enhance the local information of the control sample image.
[0020] In some embodiments, the method further comprises adding basic information of catalysts to the expert knowledge base, wherein the basic information of catalysts comprises at least type information of each catalyst;
[0021] Reading first catalyst type information of the catalyst image to be detected,
[0022] If it is determined that the first catalyst type information does not exist in the catalyst basic information in the expert knowledge base, the catalyst image to be detected is added to the catalyst dataset, and the first catalyst type information is added to the catalyst basic information.
[0023] In some embodiments, the method further comprises:
[0024] The performance of the catalyst activity prediction model is tested, and if the accuracy does not meet the preset accuracy requirement, the parameters of the catalyst activity prediction model are adjusted.
[0025] Another aspect of the present invention provides a device for constructing a catalyst image expert knowledge base, adopting the above-mentioned catalyst image expert knowledge base construction method, and the device at least comprises:
[0026] An acquisition module is used to acquire electron microscope images of catalysts with different activities and obtain images of control samples;
[0027] An expert knowledge base generation module is used to construct a catalyst activity prediction model, and use the catalyst activity prediction model to perform feature identification on the control sample image, establish a catalyst data set, and generate an expert knowledge base;
[0028] The retrieval module is used to compare the catalyst image to be tested with the images in the expert knowledge base, and when the similarity exceeds a preset similarity threshold, identify the catalyst image to be tested and / or the corresponding control sample image.
[0029] It can be seen from the above scheme that the advantages of the present invention are:
[0030] The present invention provides a method for constructing an expert knowledge base for catalyst images. The method collects electron microscope images of catalysts with different activity levels to obtain reference sample images. A catalyst activity prediction model is then constructed, and the reference sample images are characterized using the model to establish a catalyst dataset and generate an expert knowledge base. The image of the catalyst to be tested is then compared with images in the expert knowledge base. When the similarity exceeds a preset similarity threshold, the catalyst image to be tested and / or the corresponding reference sample image are identified. This method establishes an expert knowledge base to store catalyst image information and assign values to the images. After the expert knowledge base is established, the catalyst image to be tested is searched and compared within the library to form prior knowledge, thereby improving image recognition accuracy and efficiency. BRIEF DESCRIPTION OF THE DRAWINGS
[0031] Figure 1 A schematic diagram of a process for constructing a catalyst image expert knowledge base according to an embodiment of the present invention;
[0032] Figure 2 is the control sample image;
[0033] Figure 3 This is the result of image retrieval and comparison through expert knowledge base;
[0034] Figure 4 A framework diagram of a device for constructing a catalyst image expert knowledge base according to the present invention;
[0035] in:
[0036] 300: catalyst image expert knowledge base construction device;
[0037] 301: acquisition module;
[0038] 302: expert knowledge base generation module;
[0039] 303: retrieval module;
[0040] S1-S3: steps. DETAILED DESCRIPTION
[0041] In order to make the above features and effects of the present invention more clearly understood, embodiments are given below and described in detail with reference to the accompanying drawings.
[0042] refer to Figure 1 As shown, Figure 1 Graph 1 shows the overall flow chart of the method for constructing a catalyst image expert knowledge base provided by an embodiment of the present invention.
[0043] A method for constructing a catalyst image expert knowledge base includes at least the following steps:
[0044] S1. Collect electron microscope images of catalysts with different activities and obtain images of control samples.
[0045] In this embodiment, electron microscope images of catalysts with different activities, such as TEM electron microscope images of hydrogenation catalysts and alumina with different activities, are collected and annotated according to manual identification and judgment by industry experts as reference sample images. Such sample images are established into an expert knowledge base as a reference standard for new catalyst images and a basis for judgment.
[0046] In addition, in this embodiment, the control sample image is further enhanced. Specifically, a bilinear interpolation method may be used to enhance the local information of the control sample image.
[0047] S2. Constructing a catalyst activity prediction model, and using the catalyst activity prediction model to perform feature identification on the control sample image, establishing a catalyst data set, and generating an expert knowledge base.
[0048] In this embodiment, a catalyst activity prediction model is further constructed based on the deep learning method, and the model is trained, published, modified, etc. The trained catalyst activity prediction model is used to perform feature identification on the control sample image, establish a catalyst data set, and generate an expert knowledge base.
[0049] In addition, in this embodiment, basic information of catalysts is further added to the expert knowledge base. The basic information of catalysts includes information such as type information and catalyst number of each catalyst, so as to facilitate search and query through the expert knowledge base.
[0050] In addition, in this embodiment, after the catalyst activity prediction model is constructed, the performance of the catalyst activity prediction model needs to be further tested. If the model accuracy does not meet the preset accuracy requirement, the parameters of the catalyst activity prediction model need to be adjusted.
[0051] S3. Compare the image of the catalyst to be tested with the images in the expert knowledge base. When the similarity exceeds a preset similarity threshold, identify the image of the catalyst to be tested and / or the corresponding control sample image.
[0052] In this embodiment, after constructing an expert knowledge base on catalysts, the image of the catalyst to be tested is compared with the images in the expert knowledge base, including two situations: identifying the image of the catalyst to be tested and identifying multiple reference sample images with high correlation from the expert knowledge base. Specifically:
[0053] One of the situations is: before comparing the catalyst image to be tested with the image in the expert knowledge base, first perform feature extraction on the catalyst image to be tested to obtain a second feature; then compare the second feature with the first feature obtained by feature identification of the control sample image. When the similarity exceeds a first similarity threshold, the catalyst image to be tested is identified. In this case, the preset similarity threshold includes the first similarity threshold.
[0054] In addition, another situation is: feature extraction is performed on the catalyst image to be tested to obtain a second feature; then the second feature is compared with the first feature obtained by feature identification of the control sample image. When the similarity exceeds the second similarity threshold, the corresponding control sample image is identified from the expert knowledge base. In this case, the preset similarity threshold includes the second similarity threshold. In this embodiment, the first similarity threshold and the second similarity threshold can be set to different values, and this embodiment does not specifically limit the size relationship between the two. Figure 3 As shown in the figure, JPE001 and JPE002 are related control sample images screened from the expert knowledge base based on the original image on the left.
[0055] In a specific implementation, the second feature is compared with the first feature using the Euclidean distance method, that is, the 12-dimensional feature vector of each channel of the catalyst image to be tested (specifically including the three channels H, S, and V) is subtracted from the 12-dimensional feature vector of each channel of the image in the expert knowledge base in sequence to obtain the difference of the 12 feature vectors; then, the difference of each feature vector is squared, multiplied by the first weight of each feature vector, and weighted to obtain the distance value of each channel. The distance value is the value obtained by multiplying the square of the difference of each feature vector by the first weight of each feature vector, and then taking the square root of the weighted sum. Finally, the distance value of each channel is multiplied by the second weight occupied by the channel and weighted to obtain the target distance, which can represent the similarity between the catalyst image to be tested and the image in the expert knowledge base.
[0056] Furthermore, in this embodiment, basic catalyst information is added to the expert knowledge base, which includes at least the type information of each catalyst. In practice, the type information of the catalyst image to be tested needs to be analyzed. If a catalyst image of that type does not yet exist in the expert knowledge base, it is added to the expert knowledge base to complete the image type and its corresponding label. Specifically, the first catalyst type information of the catalyst image to be tested is read. If it is determined that the first catalyst type information does not exist in the basic catalyst information in the expert knowledge base, the catalyst image to be tested is added to the catalyst dataset to enrich the expert knowledge base, and the first catalyst type information is added to the basic catalyst information.
[0057] In summary, the present invention provides a method for constructing an expert knowledge base for catalyst images. The method collects electron microscope images of catalysts with different activity levels to obtain control sample images. A catalyst activity prediction model is constructed, and the control sample images are characterized using the catalyst activity prediction model to establish a catalyst data set and generate an expert knowledge base. The image of the catalyst to be tested is then compared with the images in the expert knowledge base. When the similarity exceeds a preset similarity threshold, the image of the catalyst to be tested and / or the corresponding control sample image are identified. This method establishes an expert knowledge base to store catalyst image information and assign values to images. After the expert knowledge base is established, the image of the catalyst to be tested is searched and compared in the library to form prior knowledge, thereby improving the accuracy and efficiency of image recognition. In addition, the expert knowledge base can add new types of catalyst images, which are placed in the expert knowledge comparison library through expert analysis results. Users can manually add / delete database images, which is flexible and convenient.
[0058] Furthermore, according to the above catalyst image expert knowledge base construction method, the present invention also provides a catalyst image expert knowledge base construction device 300, referring to Figure 4 , Figure 4 The structure diagram of the catalyst image expert knowledge base construction device is shown. The device can be applied to personal terminals and host terminal devices, which can be realized by Figure 1 The catalyst image expert knowledge base construction method shown in the figure can realize each process implemented by the above method.
[0059] A catalyst image expert knowledge base construction device 300 includes at least:
[0060] An acquisition module 301 is used to acquire electron microscope images of catalysts with different activities to obtain a control sample image;
[0061] An expert knowledge base generation module 302 is used to construct a catalyst activity prediction model, and use the catalyst activity prediction model to perform feature identification on the control sample image, establish a catalyst data set, and generate an expert knowledge base;
[0062] The retrieval module 303 is configured to compare the catalyst image to be tested with images in the expert knowledge base, and identify the catalyst image to be tested and / or the corresponding control sample image when the similarity exceeds a preset similarity threshold.
[0063] In addition, it should be understood that the descriptions of the methods are also applicable to the devices according to the embodiments of the present application, and to avoid repetition, they will not be described in detail.
[0064] In addition, it should be understood that in the catalyst image expert knowledge base construction device 300 according to the embodiment of the present application, the division of the above-mentioned functional modules is only used as an example. In actual applications, the above-mentioned functions can be assigned to different functional modules as needed, that is, the device can be divided into functional modules different from the modules illustrated above to complete all or part of the functions described above.
[0065] It should be noted that, in this article, the terms "comprise", "include" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device comprising a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, an element defined by the statement "comprises a ..." does not exclude the presence of other identical elements in the process, method, article or device comprising the element. In addition, it should be noted that the scope of the methods and devices in the embodiments of the present application is not limited to performing functions in the order shown or discussed, and may also include performing functions in a substantially simultaneous manner or in the opposite order according to the functions involved. For example, the described method may be performed in an order different from that described, and various steps may also be applied, omitted, or combined. In addition, the features described with reference to certain examples may be combined in other examples.
[0066] The embodiments of the present application are described above in conjunction with the accompanying drawings, but the present application is not limited to the above-mentioned specific implementation methods. The above-mentioned specific implementation methods are merely illustrative and not restrictive. Under the guidance of this application, ordinary technicians in this field can also make many forms without departing from the purpose of this application and the scope of protection of the claims, all of which are within the protection of this application.
Claims
1. A method for constructing a catalyst image expert knowledge base, characterized in that: include: Collect electron microscope images of catalysts with different activities and obtain images of control samples; Constructing a catalyst activity prediction model, and using the catalyst activity prediction model to perform feature identification on the control sample image, establishing a catalyst data set, and generating an expert knowledge base; The catalyst image to be tested is compared with the images in the expert knowledge base, and when the similarity exceeds a preset similarity threshold, the catalyst image to be tested and / or the corresponding control sample image are identified.
2. The method according to claim 1, characterized in that performing feature extraction on the catalyst image to be tested to obtain a second feature; The second feature is compared with the first feature, and when the similarity exceeds a first similarity threshold, the catalyst image to be tested is identified, wherein the first feature is a feature obtained by feature identification of the control sample image, and the preset similarity threshold includes the first similarity threshold.
3. The method according to claim 1, characterized in that performing feature extraction on the catalyst image to be tested to obtain a second feature; The second feature is compared with the first feature. When the similarity exceeds a second similarity threshold, a corresponding control sample image is identified from the expert knowledge base, where the first feature is a feature obtained by feature identification of the control sample image, and the preset similarity threshold includes the second similarity threshold.
4. The method according to claim 2 or 3, characterized in that The second feature is compared with the first feature using a Euclidean distance method.
5. The method according to claim 4, characterized in that Subtracting the 12-dimensional feature vector of each channel of the catalyst image to be tested from the 12-dimensional feature vector of each channel of the image in the expert knowledge base in sequence to obtain a difference of 12 feature vectors; The difference of each eigenvector is squared and then multiplied by the first weight of each eigenvector to obtain the distance value of each channel. The distance value of each channel is multiplied by the second weight value occupied by the channel and then weighted to obtain a target distance, which represents the similarity between the catalyst image to be tested and the image in the expert knowledge base.
6. The method according to claim 1, characterized in that Before performing feature identification on the control sample image, the method further includes performing enhancement processing on the control sample image.
7. The method according to claim 6, characterized in that A bilinear interpolation method is used to enhance the local information of the control sample image.
8. The method according to claim 1, characterized in that Also includes: Adding basic information of catalysts to the expert knowledge base, wherein the basic information of catalysts at least includes type information of each catalyst; Reading first catalyst type information of the catalyst image to be detected, If it is determined that the first catalyst type information does not exist in the catalyst basic information in the expert knowledge base, the catalyst image to be detected is added to the catalyst dataset, and the first catalyst type information is added to the catalyst basic information.
9. The method according to claim 1, characterized in that Also includes: The performance of the catalyst activity prediction model is tested, and if the accuracy does not meet the preset accuracy requirement, the parameters of the catalyst activity prediction model are adjusted.
10. A catalyst image expert knowledge base construction device, characterized in that: The method for constructing a catalyst image expert knowledge base according to any one of claims 1 to 9 is characterized in that the device at least comprises: An acquisition module is used to acquire electron microscope images of catalysts with different activities and obtain images of control samples; An expert knowledge base generation module is used to construct a catalyst activity prediction model, and use the catalyst activity prediction model to perform feature identification on the control sample image, establish a catalyst data set, and generate an expert knowledge base; The retrieval module is used to compare the catalyst image to be tested with the images in the expert knowledge base, and when the similarity exceeds a preset similarity threshold, identify the catalyst image to be tested and / or the corresponding control sample image.