Concrete test block anti-counterfeiting detection method and system based on CoAtNet network

Abstract

The invention relates to the technical field of test block anti-counterfeiting detection, and discloses a concrete test block anti-counterfeiting detection method and system based on a CoAtNet network, and the method comprises the steps: obtaining a concrete test block, carrying out the image collection of the concrete test block before the test and the concrete test block after the test, obtaining a test block image before the test and a test block image after the test, inputting the test block image before the experiment and the test block image after the experiment into an initial convolution layer module, a multi-stage repeated convolution module, a feature integration module and feature fusion to obtain initial shallow layer features before the experiment and initial shallow layer features after the experiment, reducing feature dimensions, obtaining feature vectors before the experiment and feature vectors after the experiment, and obtaining a target image; and performing similarity calculation on the basis of the feature vectors before the experiment and the feature vectors after the experiment, and if the feature vectors before the experiment and the feature vectors after the experiment are similar, determining that the concrete test block is the same concrete test block before the experiment and after the experiment. According to the invention, labor and time are saved, and the engineering quality problem caused by replacement of the test block is avoided.

Description

Technical Field

[0001] The present invention relates to the technical field of test block anti-counterfeiting detection, and in particular to a concrete test block anti-counterfeiting detection method and system based on a CoAtNet network. Background Art

[0002] The quality inspection of concrete test blocks is the core link of the quality control of construction projects, and the reliability of its anti-counterfeiting technology is directly related to the safety assessment of engineering structures. Traditional anti-counterfeiting technology mainly relies on physical identification methods such as barcodes, QR codes or RFID tags. For example, the Chinese invention patent "Anti-counterfeiting traceability method for concrete test blocks based on chip fingerprints" (CN202410603044.7) combines RFID chips with PUF technology to give each test block a unique identity to improve anti-counterfeiting security. For another example, the Chinese invention patent "An anti-counterfeiting method for concrete test pieces based on image recognition based on AI" (CN202410742697.3) proposes a multi-scale detection model combining YOLOv8 and BiFPN and a dual-branch ViT feature extraction network, which comprehensively determines the authenticity of the test piece through dry and wet image alignment and deep learning similarity scores. The solution is to introduce an advanced multi-scale feature fusion mechanism, which significantly improves the target recognition accuracy under complex backgrounds, and strengthens the local feature correlation through a bidirectional attention module. However, its limitations cannot be ignored, and it relies on prefabricated labels with QR codes as reference features.

[0003] Contactless anti-counterfeiting technology based on image recognition is gradually emerging. For example, the Chinese invention patent "A method for preventing counterfeiting of concrete test blocks based on image recognition" (CN202110459242.7) discloses a method for collecting images of on-site and laboratory test blocks through a mobile terminal APP and performing similarity comparison. This method uses a template matching algorithm to improve the robustness of image matching through rotation and cropping operations. Its advantages are convenient operation and low cost, but there are some limitations. This method does not consider multi-angle feature fusion, which may lead to the omission of key texture information and reduce the comprehensive judgment ability of the anti-counterfeiting system.

[0004] In summary, the existing technology is either limited by the inherent defects of physical labels or fails to fully explore the potential of image recognition in anti-counterfeiting comparison. How to perform fine anti-counterfeiting under non-contact concrete blocks is still an extremely difficult problem. Through domestic and foreign patent and literature retrieval, there are many studies on the identification of physical identification methods for concrete anti-counterfeiting. However, the research on contactless anti-counterfeiting technology generally uses single-angle image detection, which is difficult to fully capture the surface texture details of the test block, and the classification accuracy may decrease when the size of the test block or the lighting conditions change. There is no multi-modal comparison anti-counterfeiting method that uses the CoAtNet model to optimize the surface features of concrete test blocks. Summary of the invention

[0005] The object of the present invention is to overcome one or more of the above-mentioned existing technical problems and to provide a concrete test block anti-counterfeiting detection method and system based on the CoAtNet network.

[0006] To achieve the above object, the present invention provides a concrete test block anti-counterfeiting detection method based on CoAtNet network, comprising: Acquire a concrete test block, collect images of the concrete test block before and after the test, and obtain images of the test block before and after the test; Input the test block image before the experiment and the test block image after the experiment into the initial convolution layer module to obtain the initial convolution features before the experiment and the initial convolution features after the experiment; Input the initial convolution features before the experiment and the initial convolution features after the experiment into a multi-stage repeated convolution module to obtain the deepening convolution features before the experiment and the deepening convolution features after the experiment; Input the pre-experimental deepening convolutional features and the post-experimental deepening convolutional features into the feature integration module to obtain the pre-experimental high-level convolutional features and the post-experimental high-level convolutional features; Based on the features of the image before the experiment and the features of the image after the experiment, feature fusion is performed on them respectively, and deeper features are gradually integrated into shallow features to obtain initial shallow features before the experiment and initial shallow features after the experiment; The feature dimensions of the initial shallow features before the experiment and the initial shallow features after the experiment are reduced to obtain the feature vectors before the experiment and the feature vectors after the experiment; The similarity is calculated based on the eigenvector before and after the experiment. If the two are similar, the concrete specimen is the same before and after the experiment.

[0007] According to one aspect of the present invention, the method for obtaining the test block image before the experiment and the test block image after the experiment is: Acquire a concrete test block, perform a first image acquisition on the current concrete test block, and obtain a first image of the concrete test block; Conducting an experiment on the current concrete test block, and performing a second image acquisition on the concrete test block after the experiment to obtain a second image of the concrete test block; Performing center cropping on the first image of the concrete test block and the second image of the concrete test block, and scaling the first image of the concrete test block and the second image of the concrete test block to a resolution size of 448×448 pixels; Based on the application of data enhancement technology, the first image of the concrete test block and the second image of the concrete test block are enhanced to increase the diversity of the first image of the concrete test block and the second image of the concrete test block and simulate the situations encountered in actual application scenarios; The first image of the concrete test block and the second image of the concrete test block are standardized to obtain the image of the test block before the test and the image of the test block after the test.

[0008] According to one aspect of the present invention, the method for obtaining the initial convolution features before the experiment and the initial convolution features after the experiment is: The initial convolution layer module includes a first convolution module and a second convolution module, and the first convolution module and the second convolution module are two 3×3 convolution kernels; The image of the test block before the experiment and the image of the test block after the experiment are input into the first convolution module to obtain the features of the first convolution module before the experiment and the features of the first convolution module after the experiment, where the formula is, ; ; in, represents the image of the test block before the experiment; Represents the image of the test block after the test; Represents a 3×3 convolution kernel; represents the first convolutional module; Represents the features of the first convolutional module before the experiment; Represents the features of the first convolutional module after the experiment; The features of the first convolution module before the experiment and the features of the second convolution module after the experiment are input into the second convolution module to obtain the initial convolution features before the experiment and the initial convolution features after the experiment, where the formula is, ; ; in, represents the second convolutional module; Represents the initial convolution features before the experiment; Represents the initial convolution features after the experiment.

[0009] According to one aspect of the present invention, the method for obtaining the pre-experimental deepening convolution feature and the post-experimental deepening convolution feature is: The multi-stage repeated convolution module includes a first repeated convolution module and a second repeated convolution module, wherein the first repeated convolution module and the second repeated convolution module are two 1×1 convolution kernels and one 4×4 depth separation convolution; The initial convolution features before the experiment and the initial convolution features after the experiment are input into the first repeated convolution module to obtain the first repeated convolution module features before the experiment and the first repeated convolution module features after the experiment, where the formula is, ; ; in, represents the first repeated convolution module; Represents a 1×1 convolution kernel; Indicates that the convolution kernel is 4×4 depth separation convolution; Represents the features of the first repeated convolution module before the experiment; Represents the features of the first repeated convolution module after the experiment; The features of the first repeated convolution module before the experiment and the features of the first repeated convolution module after the experiment are input into the second repeated convolution module to obtain the deepening convolution features before the experiment and the deepening convolution features after the experiment, where the formula is, ; ; in, represents the second repeated convolution module; Indicates deepening of convolution features before the experiment; Indicates deepening of convolution features after the experiment.

[0010] According to one aspect of the present invention, the method for obtaining the high-level convolution features before the experiment and the high-level convolution features after the experiment is: The feature integration module includes a first integration module and a second integration module, wherein the first integration module and the second integration module are repeated associative attention mechanisms and feedforward neural networks; The deepening convolution features before the experiment and the deepening convolution features after the experiment are input into the first integration module to obtain the first integration module features before the experiment and the first integration module features after the experiment, where the formula is, ; ; in, represents the first integration module; FFN stands for Feed-Forward Neural Network; Represents repeated associative attention mechanisms; represents the first integrated module characteristics before the experiment; It represents the first integrated module characteristics after the experiment; The features of the first integration module before the experiment and the features of the first integration module after the experiment are input into the second integration module to obtain the high-level convolution features before the experiment and the high-level convolution features after the experiment, where the formula is, ; ; in, represents the second integration module; Represents the high-level convolution features before the experiment; Represents high-level convolutional features after experiments.

[0011] According to one aspect of the present invention, the method for obtaining the initial shallow features before the experiment and the initial shallow features after the experiment is: The features of the pre-experiment image and the features of the post-experiment image extracted at each layer are stored in a pre-experiment feature list and a post-experiment feature list; Based on the FPN strategy, the deeper features are gradually integrated into the shallow features to obtain the initial shallow features before the experiment and the initial shallow features after the experiment, where the formula is: ; ; in, Indicates feature fusion; Represents the list of features before the experiment; Represents the feature list after the experiment; represents the initial shallow characteristics before the experiment; Represents the initial shallow features after the experiment.

[0012] According to one aspect of the present invention, the method for obtaining the feature vector before and after the experiment is: The initial shallow features before the experiment and the initial shallow features after the experiment are applied with adaptive average pooling to obtain the compressed features before the experiment and the compressed features after the experiment, where the formula is, ; ; in, Indicates feature fusion; represents the list of features before the experiment; Represents the feature list after the experiment; Apply the fully connected layer to the pre-experimental compressed features and the post-experimental compressed features to obtain the pre-experimental feature vector and the post-experimental feature vector, where the formula is, ; ; in, represents a fully connected layer; represents the feature vector before the experiment; represents the post-experimental feature vector.

[0013] To achieve the above object, the present invention provides a concrete test block anti-counterfeiting detection system based on CoAtNet network, comprising: Test block image acquisition module: acquire concrete test blocks, collect images of concrete test blocks before and after the test, and obtain images of test blocks before and after the test; Initial convolution feature acquisition module: input the test block image before the experiment and the test block image after the experiment into the initial convolution layer module to obtain the initial convolution features before the experiment and the initial convolution features after the experiment; Deepening convolution feature acquisition module: input the initial convolution features before the experiment and the initial convolution features after the experiment into the multi-stage repeated convolution module to obtain the deepening convolution features before the experiment and the deepening convolution features after the experiment; Advanced convolution feature acquisition module: input the pre-experimental deepening convolution features and the post-experimental deepening convolution features into the feature integration module to obtain the pre-experimental advanced convolution features and the post-experimental advanced convolution features; Initial shallow feature acquisition module: based on the features of the image before the experiment and the features of the image after the experiment, feature fusion is performed on them respectively, and deeper features are gradually integrated into shallow features to obtain initial shallow features before the experiment and initial shallow features after the experiment; Feature vector acquisition module: reducing the feature dimensions of the initial shallow features before the experiment and the initial shallow features after the experiment to obtain the feature vectors before the experiment and the feature vectors after the experiment; Similarity calculation module: perform similarity calculation based on the feature vector before the experiment and the feature vector after the experiment. If the two are similar, the concrete specimen is the same before and after the experiment.

[0014] To achieve the above-mentioned purpose, the present invention provides an electronic device, including a processor, a memory, and a computer program stored in the memory and executable on the processor. When the computer program is executed by the processor, the above-mentioned method for anti-counterfeiting detection of concrete test blocks based on the CoAtNet network is implemented.

[0015] To achieve the above-mentioned purpose, the present invention provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the above-mentioned anti-counterfeiting detection method for concrete test blocks based on the CoAtNet network is implemented.

[0016] Based on this, the beneficial effects of the present invention are as follows: by designing a concrete test block registration system, collecting high-resolution images, and constructing a full surface feature database of the test block, an improved CoAtNet model is proposed to address the problem of detail loss caused by traditional model size scaling, and a bottom-up semantic information fusion mechanism is innovatively introduced to dynamically combine deep global structures with shallow texture features; It can fully capture the surface texture, edge deformation and microscopic details of the test block, effectively avoid missing key features, and provide a high-integrity data foundation for subsequent comparative analysis. At the same time, it realizes fully automated operation, reduces labor costs and improves supervision efficiency. The model does not need to preset labels manually, and directly generates comparison benchmarks through multi-angle image features. It has high accuracy and less calculation time in high-resolution image classification tasks, and significantly improves the ability to capture features such as fine cracks and particle distribution on the surface of the test block, solving the problem of misjudgment caused by preprocessing distortion in existing technologies. In addition, the system is compatible with test blocks of different specifications and complex environmental lighting conditions, and has strong generalization capabilities, providing technical support for concrete quality supervision that is both reliable, efficient and universal. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 is a flow chart of a concrete test block anti-counterfeiting detection method based on a CoAtNet network according to an exemplary embodiment; Figure 2 is a module structure diagram of a concrete test block anti-counterfeiting detection method based on a CoAtNet network according to an exemplary embodiment; Figure 3 It is a flow chart of a concrete test block anti-counterfeiting detection system based on the CoAtNet network according to an exemplary embodiment. DETAILED DESCRIPTION

[0018] The present invention will now be discussed with reference to exemplary embodiments. It should be understood that the embodiments discussed are only for enabling those skilled in the art to better understand and thereby implement the present invention, rather than implying any limitation on the scope of the present invention.

[0019] As used herein, the term “including” and variations thereof are to be interpreted as open-ended terms meaning “including but not limited to.” The term “based on” is to be interpreted as “based, at least in part, on,” and the terms “one embodiment” and “an embodiment” are to be interpreted as “at least one embodiment.”

[0020] According to one embodiment of the present invention, Figure 1 is a flow chart of a concrete test block anti-counterfeiting detection method based on a CoAtNet network according to an exemplary embodiment. Figure 1As shown, in order to achieve the above-mentioned purpose, the present invention provides a concrete test block anti-counterfeiting detection method based on CoAtNet network, comprising: Acquire a concrete test block, collect images of the concrete test block before and after the test, and obtain images of the test block before and after the test; Figure 2 is a module structure diagram of a concrete test block anti-counterfeiting detection method based on a CoAtNet network according to an exemplary embodiment, such as Figure 2 As shown, the test block image before the experiment and the test block image after the experiment are input into the initial convolution layer module to obtain the initial convolution features before the experiment and the initial convolution features after the experiment; Input the initial convolution features before the experiment and the initial convolution features after the experiment into a multi-stage repeated convolution module to obtain the deepening convolution features before the experiment and the deepening convolution features after the experiment; Input the pre-experimental deepening convolutional features and the post-experimental deepening convolutional features into the feature integration module to obtain the pre-experimental high-level convolutional features and the post-experimental high-level convolutional features; Based on the features of the image before the experiment and the features of the image after the experiment, feature fusion is performed on them respectively, and deeper features are gradually integrated into shallow features to obtain initial shallow features before the experiment and initial shallow features after the experiment; The feature dimensions of the initial shallow features before the experiment and the initial shallow features after the experiment are reduced to obtain the feature vectors before the experiment and the feature vectors after the experiment; The similarity is calculated based on the eigenvector before and after the experiment. If the two are similar, the concrete specimen is the same before and after the experiment.

[0021] According to one embodiment of the present invention, the method for obtaining the test block image before the experiment and the test block image after the experiment is: Acquire a concrete test block, perform a first image acquisition on the current concrete test block, and obtain a first image of the concrete test block; Conducting an experiment on the current concrete test block, and performing a second image acquisition on the concrete test block after the experiment to obtain a second image of the concrete test block; The pixel size of the first image of the concrete test block and the second image of the concrete test block is 4000×3000. In order to meet the input requirements of the model, the collected first image of the concrete test block and the second image of the concrete test block are preprocessed; The first image of the concrete test block and the second image of the concrete test block are centrally cropped to remove unnecessary background information, and the first image of the concrete test block and the second image of the concrete test block are scaled to a resolution of 448×448 pixels, which can greatly reduce the impact of the compression size on the model performance; Enhance the first image of the concrete test block and the second image of the concrete test block based on the application of data enhancement technology, such as random rotation, translation and color adjustment, to increase the diversity of the first image of the concrete test block and the second image of the concrete test block and simulate situations encountered in actual application scenarios; The first image of the concrete test block and the second image of the concrete test block are standardized to have the same mean and standard deviation to ensure the consistency of model training, and the image of the test block before the experiment and the image of the test block after the experiment are obtained.

[0022] According to one embodiment of the present invention, a method for obtaining initial convolution features before and after the experiment is as follows: The initial convolution layer module includes a first convolution module and a second convolution module, and the first convolution module and the second convolution module are two 3×3 convolution kernels; The image of the test block before the experiment and the image of the test block after the experiment are input into the first convolution module to obtain the features of the first convolution module before the experiment and the features of the first convolution module after the experiment, and the image size is first reduced to 224×224, where the formula is, ; ; in, represents the image of the test block before the experiment; Represents the image of the test block after the test; Represents a 3×3 convolution kernel; represents the first convolutional module; Represents the features of the first convolutional module before the experiment; Represents the features of the first convolutional module after the experiment; The features of the first convolution module before the experiment and the features of the second convolution module after the experiment are input into the second convolution module to obtain the initial convolution features before the experiment and the initial convolution features after the experiment, and the image size is reduced to 112×112, where the formula is, ; ; in, represents the second convolutional module; Represents the initial convolution features before the experiment; Represents the initial convolution features after the experiment.

[0023] According to one embodiment of the present invention, the method for obtaining the pre-experiment deepening convolution feature and the post-experiment deepening convolution feature is: The multi-stage repeated convolution module includes a first repeated convolution module and a second repeated convolution module. The first repeated convolution module and the second repeated convolution module are two 1×1 convolution kernels and one 4×4 depth separation convolution, which gradually reduce the image size and deepen the features. The initial convolution features before the experiment and the initial convolution features after the experiment are input into the first repeated convolution module to obtain the first repeated convolution module features before the experiment and the first repeated convolution module features after the experiment, and the image size is reduced to 56×56, where the formula is, ; ; in, represents the first repeated convolution module; Represents a 1×1 convolution kernel; Indicates that the convolution kernel is 4×4 depth separation convolution; Represents the features of the first repeated convolution module before the experiment; Represents the features of the first repeated convolution module after the experiment; The features of the first repeated convolution module before the experiment and the features of the first repeated convolution module after the experiment are input into the second repeated convolution module to obtain the deepening convolution features before the experiment and the deepening convolution features after the experiment, and reduce the image size to 28×28, where the formula is, ; ; in, represents the second repeated convolution module; Indicates deepening of convolution features before the experiment; Indicates deepening of convolution features after the experiment.

[0024] According to one embodiment of the present invention, the method for obtaining the high-level convolution features before the experiment and the high-level convolution features after the experiment is: After deep processing is completed, the model enters a more complex high-level feature integration stage. The key to this stage is to use repeated associative attention mechanisms and feedforward neural networks to further refine and fuse features and enhance the model's ability to learn the relationship between features of different scales. The feature integration module includes the first integration module and the second integration module. The first integration module and the second integration module are repeated associative attention mechanisms and feedforward neural networks. The deep convolution features before the experiment and the deep convolution features after the experiment are input into the first integration module to obtain the first integration module features before the experiment and the first integration module features after the experiment, and the image size is reduced to 14×14, where the formula is, ; ; in, represents the first integration module; FFN stands for Feed-Forward Neural Network; Represents repeated associative attention mechanisms; represents the first integrated module characteristics before the experiment; It represents the first integrated module characteristics after the experiment; The features of the first integration module before the experiment and the features of the first integration module after the experiment are input into the second integration module to obtain the high-level convolution features before the experiment and the high-level convolution features after the experiment, and the image size is reduced to 7×7, where the formula is, ; ; in, represents the second integration module; Represents the high-level convolution features before the experiment; Represents high-level convolutional features after experiments.

[0025] According to one embodiment of the present invention, the method for obtaining the initial shallow features before the experiment and the initial shallow features after the experiment is: The features of the pre-experimental image and the features of the post-experimental image extracted at each layer are stored in the pre-experimental feature list and the post-experimental feature list, and the feature maps of different layers of these features have a gradual change from high-resolution shallow information to low-resolution deep semantic information; Through the concept of feature pyramid network FPN, it is optimized to more effectively integrate features of different scales and improve the performance of concrete specimen images in similarity calculation. Based on the FPN strategy, deeper features are gradually integrated into shallow features to obtain the initial shallow features before the experiment and the initial shallow features after the experiment. The formula is: ; ; in, Indicates feature fusion; represents the list of features before the experiment; Represents the feature list after the experiment; represents the initial shallow characteristics before the experiment; Represents the initial shallow features after the experiment.

[0026] This process ensures that the deep semantic information is effectively integrated into the shallow features, thereby ensuring that the information at different scales is fully utilized. The final shallow features, the initial shallow features before the experiment and the initial shallow features after the experiment, will contain rich details and semantic information, which is an ideal feature representation for image comparison; According to one embodiment of the present invention, the method for obtaining the feature vector before and after the experiment is: The initial shallow features before the experiment and the initial shallow features after the experiment are applied with adaptive average pooling to obtain the compressed features before the experiment and the compressed features after the experiment, where the formula is, ; ; in, Indicates feature fusion; represents the list of features before the experiment; Represents the feature list after the experiment; Apply the fully connected layer to the pre-experimental compressed features and the post-experimental compressed features to obtain the pre-experimental feature vector and the post-experimental feature vector, where the formula is, ; ; in, represents a fully connected layer; represents the feature vector before the experiment; represents the post-experimental feature vector.

[0027] According to one embodiment of the present invention, the obtained pre-experimental feature vectors and post-experimental feature vectors are trained through a cosine similarity loss function to ensure that similar images are closer in the feature space, while dissimilar images are farther away. This loss function not only optimizes the expressive power of the features, but also ensures accuracy and robustness in practical applications.

[0028] According to one embodiment of the present invention, a decision is made whether to accept or reject the authenticity of the concrete test block based on the similarity score. If the similarity exceeds a preset threshold, (For example =0.85), the two concrete test blocks are considered to have the same source, otherwise they are considered to have different sources. The formal expression is: ; in, Indicates decision making; To express acceptance; To express refusal; Indicates similarity; Indicates the threshold value.

[0029] According to one embodiment of the present invention, for areas where the contrast features do not match, statistical analysis is used to further screen out abnormal areas, and detection areas that are too small or obviously do not match the characteristic morphology are filtered out based on the area or shape of the area. The area is reviewed again, and after confirmation, it is reported to the system platform to warn of the existence of replacement of the test block.

[0030] In addition, in order to achieve the above-mentioned purpose of the invention, the present invention also provides a concrete test block anti-counterfeiting detection system based on the CoAtNet network. Figure 3 is a flow chart of a concrete test block anti-counterfeiting detection system based on a CoAtNet network according to an exemplary embodiment. Figure 3 As shown, a concrete test block anti-counterfeiting detection system based on the CoAtNet network in the present invention includes: Test block image acquisition module: acquire concrete test blocks, collect images of concrete test blocks before and after the test, and obtain images of test blocks before and after the test; Initial convolution feature acquisition module: input the test block image before the experiment and the test block image after the experiment into the initial convolution layer module to obtain the initial convolution features before the experiment and the initial convolution features after the experiment; Deepening convolution feature acquisition module: input the initial convolution features before the experiment and the initial convolution features after the experiment into the multi-stage repeated convolution module to obtain the deepening convolution features before the experiment and the deepening convolution features after the experiment; Advanced convolution feature acquisition module: input the pre-experimental deepening convolution features and the post-experimental deepening convolution features into the feature integration module to obtain the pre-experimental advanced convolution features and the post-experimental advanced convolution features; Initial shallow feature acquisition module: based on the features of the image before the experiment and the features of the image after the experiment, feature fusion is performed on them respectively, and deeper features are gradually integrated into shallow features to obtain initial shallow features before the experiment and initial shallow features after the experiment; Feature vector acquisition module: reducing the feature dimensions of the initial shallow features before the experiment and the initial shallow features after the experiment to obtain the feature vectors before the experiment and the feature vectors after the experiment; Similarity calculation module: perform similarity calculation based on the feature vector before the experiment and the feature vector after the experiment. If the two are similar, the concrete specimen is the same before and after the experiment.

[0031] To achieve the above-mentioned purpose of the invention, the present invention also provides an electronic device, which includes: a processor, a memory, and a computer program stored in the memory and executable on the processor. When the computer program is executed by the processor, the above-mentioned concrete test block anti-counterfeiting detection method based on the CoAtNet network is implemented.

[0032] To achieve the above-mentioned purpose of the invention, the present invention also provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the above-mentioned concrete test block anti-counterfeiting detection method based on the CoAtNet network is implemented.

[0033] Those of ordinary skill in the art will appreciate that the modules and algorithm steps described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of the present invention.

[0034] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the above-described devices and equipment can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0035] In the embodiments provided in the present application, it should be understood that the disclosed devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of the modules is only a logical function division. There may be other division methods in actual implementation, such as multiple modules or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or modules, which can be electrical, mechanical or other forms.

[0036] The modules described as separate components may or may not be physically separated, and the components shown as modules may or may not be physical modules, that is, they may be located in one place or distributed on multiple network modules. Some or all of the modules may be selected according to actual needs to achieve the purpose of the embodiments of the present invention.

[0037] In addition, each functional module in the embodiment of the present invention may be integrated into one processing module, or each module may exist physically separately, or two or more modules may be integrated into one module.

[0038] If the functions are implemented in the form of software function modules and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention can be essentially or partly embodied in the form of a software product that contributes to the prior art. The computer software product is stored in a storage medium and includes several instructions for enabling a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the energy-saving signal sending / receiving method of each embodiment of the present invention. The aforementioned storage medium includes: various media that can store program codes, such as a USB flash drive, a mobile hard disk, a ROM, a RAM, a magnetic disk, or an optical disk.

[0039] The above description is only a preferred embodiment of the present application and an explanation of the technical principles used. Those skilled in the art should understand that the scope of the invention involved in the present application is not limited to the technical solution formed by a specific combination of the above technical features, but should also cover other technical solutions formed by any combination of the above technical features or their equivalent features without departing from the inventive concept. For example, the above features are replaced with (but not limited to) technical features with similar functions disclosed in the present application.

[0040] It should be understood that the size of the serial numbers of each step in the content of the invention and the embodiments of the present invention does not absolutely mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention.

Claims

1. A concrete test block anti-counterfeiting detection method based on CoAtNet network, characterized in that: include: Acquire a concrete test block, collect images of the concrete test block before and after the test, and obtain images of the test block before and after the test; Input the test block image before the experiment and the test block image after the experiment into the initial convolution layer module to obtain the initial convolution features before the experiment and the initial convolution features after the experiment; Input the initial convolution features before the experiment and the initial convolution features after the experiment into a multi-stage repeated convolution module to obtain the deepening convolution features before the experiment and the deepening convolution features after the experiment; Input the pre-experimental deepening convolutional features and the post-experimental deepening convolutional features into the feature integration module to obtain the pre-experimental high-level convolutional features and the post-experimental high-level convolutional features; Based on the features of the image before the experiment and the features of the image after the experiment, feature fusion is performed on them respectively, and deeper features are gradually integrated into shallow features to obtain initial shallow features before the experiment and initial shallow features after the experiment; The feature dimensions of the initial shallow features before the experiment and the initial shallow features after the experiment are reduced to obtain the feature vectors before the experiment and the feature vectors after the experiment; The similarity is calculated based on the eigenvector before and after the experiment. If the two are similar, the concrete specimen is the same before and after the experiment.

2. A concrete test block anti-counterfeiting detection method based on CoAtNet network as claimed in claim 1, characterized in that: The method for obtaining the test block image before and after the test is as follows: Acquire a concrete test block, perform a first image acquisition on the current concrete test block, and obtain a first image of the concrete test block; Conducting an experiment on the current concrete test block, and performing a second image acquisition on the concrete test block after the experiment to obtain a second image of the concrete test block; Performing center cropping on the first image of the concrete test block and the second image of the concrete test block, and scaling the first image of the concrete test block and the second image of the concrete test block to a resolution size of 448×448 pixels; Based on the application of data enhancement technology, the first image of the concrete test block and the second image of the concrete test block are enhanced to increase the diversity of the first image of the concrete test block and the second image of the concrete test block and simulate the situations encountered in actual application scenarios; The first image of the concrete test block and the second image of the concrete test block are standardized to obtain the image of the test block before the test and the image of the test block after the test.

3. A concrete test block anti-counterfeiting detection method based on CoAtNet network as claimed in claim 2, characterized in that: The method for obtaining the initial convolution features before the experiment and the initial convolution features after the experiment: The initial convolution layer module includes a first convolution module and a second convolution module, and the first convolution module and the second convolution module are two 3×3 convolution kernels; The image of the test block before the experiment and the image of the test block after the experiment are input into the first convolution module to obtain the features of the first convolution module before the experiment and the features of the first convolution module after the experiment, where the formula is, ； ； in, represents the image of the test block before the experiment; Represents the image of the test block after the test; Represents a 3×3 convolution kernel; represents the first convolutional module; Represents the features of the first convolutional module before the experiment; Represents the features of the first convolutional module after the experiment; The features of the first convolution module before the experiment and the features of the second convolution module after the experiment are input into the second convolution module to obtain the initial convolution features before the experiment and the initial convolution features after the experiment, where the formula is, ； ； in, represents the second convolutional module; Represents the initial convolution features before the experiment; Represents the initial convolution features after the experiment.

4. A concrete test block anti-counterfeiting detection method based on CoAtNet network as claimed in claim 3, characterized in that: The method for obtaining the deepening convolution features before the experiment and the deepening convolution features after the experiment is: The multi-stage repeated convolution module includes a first repeated convolution module and a second repeated convolution module, wherein the first repeated convolution module and the second repeated convolution module are two 1×1 convolution kernels and one 4×4 depth separation convolution; The initial convolution features before the experiment and the initial convolution features after the experiment are input into the first repeated convolution module to obtain the first repeated convolution module features before the experiment and the first repeated convolution module features after the experiment, where the formula is, ； ； in, represents the first repeated convolution module; Represents a 1×1 convolution kernel; Indicates that the convolution kernel is 4×4 depth separation convolution; Represents the features of the first repeated convolution module before the experiment; Represents the features of the first repeated convolution module after the experiment; The features of the first repeated convolution module before the experiment and the features of the first repeated convolution module after the experiment are input into the second repeated convolution module to obtain the deepening convolution features before the experiment and the deepening convolution features after the experiment, where the formula is, ； ； in, represents the second repeated convolution module; Indicates deepening of convolution features before the experiment; Indicates deepening of convolution features after the experiment.

5. A concrete test block anti-counterfeiting detection method based on CoAtNet network as claimed in claim 4, characterized in that: The method for obtaining the high-level convolution features before and after the experiment is as follows: The feature integration module includes a first integration module and a second integration module, wherein the first integration module and the second integration module are repeated associative attention mechanisms and feedforward neural networks; The deepening convolution features before the experiment and the deepening convolution features after the experiment are input into the first integration module to obtain the first integration module features before the experiment and the first integration module features after the experiment, where the formula is, ； ； in, represents the first integration module; FFN stands for Feed-Forward Neural Network; Represents repeated associative attention mechanisms; represents the first integrated module characteristics before the experiment; It represents the first integrated module characteristics after the experiment; The features of the first integration module before the experiment and the features of the first integration module after the experiment are input into the second integration module to obtain the high-level convolution features before the experiment and the high-level convolution features after the experiment, where the formula is, ； ； in, represents the second integration module; Represents the high-level convolution features before the experiment; Represents high-level convolutional features after experiments.

6. A concrete test block anti-counterfeiting detection method based on CoAtNet network as claimed in claim 5, characterized in that: The method for obtaining the initial shallow features before the experiment and the initial shallow features after the experiment is: The features of the pre-experiment image and the features of the post-experiment image extracted at each layer are stored in a pre-experiment feature list and a post-experiment feature list; Based on the FPN strategy, the deeper features are gradually integrated into the shallow features to obtain the initial shallow features before the experiment and the initial shallow features after the experiment, where the formula is: ； ； in, Indicates feature fusion; Represents the list of features before the experiment; Represents the feature list after the experiment; represents the initial shallow characteristics before the experiment; Represents the initial shallow features after the experiment.

7. A concrete test block anti-counterfeiting detection method based on CoAtNet network as claimed in claim 6, characterized in that: The method for obtaining the feature vector before and after the experiment is: The initial shallow features before the experiment and the initial shallow features after the experiment are applied with adaptive average pooling to obtain the compressed features before the experiment and the compressed features after the experiment, where the formula is, ； ； in, Indicates feature fusion; Represents the list of features before the experiment; Represents the feature list after the experiment; Apply the fully connected layer to the pre-experimental compressed features and the post-experimental compressed features to obtain the pre-experimental feature vector and the post-experimental feature vector, where the formula is, ； ； in, represents a fully connected layer; represents the feature vector before the experiment; represents the post-experimental feature vector.

8. A concrete test block anti-counterfeiting detection system based on CoAtNet network, characterized in that: include: Test block image acquisition module: acquire concrete test blocks, collect images of concrete test blocks before and after the test, and obtain images of test blocks before and after the test; Initial convolution feature acquisition module: input the test block image before the experiment and the test block image after the experiment into the initial convolution layer module to obtain the initial convolution features before the experiment and the initial convolution features after the experiment; Deepening convolution feature acquisition module: input the initial convolution features before the experiment and the initial convolution features after the experiment into the multi-stage repeated convolution module to obtain the deepening convolution features before the experiment and the deepening convolution features after the experiment; Advanced convolution feature acquisition module: input the deep convolution features before the experiment and the deep convolution features after the experiment into the feature integration module to obtain the advanced convolution features before the experiment and the advanced convolution features after the experiment; Initial shallow feature acquisition module: based on the features of the image before the experiment and the features of the image after the experiment, feature fusion is performed on them respectively, and deeper features are gradually integrated into shallow features to obtain initial shallow features before the experiment and initial shallow features after the experiment; Feature vector acquisition module: reducing the feature dimensions of the initial shallow features before the experiment and the initial shallow features after the experiment to obtain the feature vectors before the experiment and the feature vectors after the experiment; Similarity calculation module: perform similarity calculation based on the feature vector before the experiment and the feature vector after the experiment. If the two are similar, the concrete specimen is the same before and after the experiment.

9. An electronic device, characterized in that: The invention comprises a processor, a memory and a computer program stored in the memory and executable on the processor, wherein when the computer program is executed by the processor, a concrete test block anti-counterfeiting detection method based on the CoAtNet network as claimed in any one of claims 1 to 7 is implemented.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and when the computer program is executed by the processor, the method for anti-counterfeiting detection of concrete test blocks based on the CoAtNet network as claimed in any one of claims 1 to 7 is implemented.

Images