Concrete rebound strength detection method and device based on pattern recognition processing, storage medium and computer equipment
By acquiring rebound-related data from large concrete components and using pattern recognition processing to generate prior features, the accuracy problem of compressive strength testing for large concrete components is solved, and more accurate strength prediction is achieved.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-12-29
- Publication Date
- 2026-04-03
AI Technical Summary
In large concrete components, due to the large distance between different measurement points, the accuracy of the compressive strength obtained by directly calculating the average value is low, making it difficult to accurately reflect the compressive strength of large concrete components.
By acquiring rebound-related data detected by sensors, rebound data of multiple sub-regions are generated. The corresponding prior models are matched from the pre-built set of rebound-strength prior models, and prior features are generated using pattern recognition processing. Finally, the concrete rebound strength test results are obtained.
It improves the accuracy of predicting the compressive strength of concrete components by dynamically dividing semantically reasonable sub-region structures, reducing systematic biases and improving the precision of test results.
Smart Images
Figure CN121783745A_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of detection technology, and in particular to a method, apparatus, storage medium and computer equipment for detecting the rebound strength of concrete based on pattern recognition processing. Background Technology
[0002] For concrete components, compressive strength is one of the important performance indicators, affecting the overall quality and safety of concrete components and related engineering construction.
[0003] In related technologies, the rebound method can be used to detect the relationship between the surface hardness and strength of concrete components, thereby determining the compressive strength of the concrete components. Specifically, multiple measurement points are typically selected on the concrete component for measurement, and the compressive strength of the concrete component is calculated based on the average value of the measurement data from each point.
[0004] However, for some large concrete components, due to the large distance between different measurement points, the accuracy of the compressive strength obtained by directly calculating the average value is low, making it difficult to accurately reflect the compressive strength of large concrete components. Summary of the Invention
[0005] To address the aforementioned technical problems, this application proposes a method, apparatus, storage medium, and computer device for detecting the rebound strength of concrete based on pattern recognition processing, which can improve the accuracy of predicting the compressive strength of concrete components.
[0006] In a first aspect, embodiments of this application provide a method for detecting the rebound strength of concrete based on pattern recognition processing, including: Acquire rebound-related data detected by sensors in the concrete area to be tested; Based on the rebound-related data, multiple sub-region rebound data are generated; From a pre-built set of rebound-intensity prior models, at least one prior model corresponding to the rebound data of the plurality of sub-regions is matched, wherein the set of rebound-intensity prior models includes a plurality of prior models corresponding to a plurality of sets of rebound-intensity sample data, and each of the plurality of prior models is trained based on its corresponding rebound-intensity sample data; The rebound data of each sub-region is input into its corresponding prior model to generate prior features of the rebound data of each sub-region; Based on the prior features and the rebound-related data, the rebound strength test result of the concrete corresponding to the concrete area to be tested is obtained through pattern recognition processing.
[0007] Optionally, generating multiple sub-region rebound data based at least on the rebound-related data includes: Based on the spatial structure prompt information and the rebound-related data, a semantic sequence is generated, wherein the semantic sequence includes multiple semantic sub-region labels and the association relationship between the multiple semantic sub-region labels, and the multiple sub-region rebound data corresponds one-to-one with the multiple semantic sub-region labels; Based on the semantic sequence, the bounce-related data are aggregated to obtain the semantic sub-region bounce data corresponding to each of the multiple semantic sub-region labels; During the generation process of each semantic sub-region label, relevant feature information of each semantic sub-region label during the generation process is obtained, and the rebound intensity of the sub-region corresponding to each semantic sub-region label is determined based on the relevant feature information corresponding to each semantic sub-region label. Based on the sub-region bounce intensity and the semantic sub-region bounce data corresponding to each semantic sub-region label, sub-region bounce data corresponding to each semantic sub-region label is generated.
[0008] Optionally, the association relationship includes spatiotemporal association relationships; The generation of a semantic sequence based on spatial structure prompting information and rebound-related data includes: Determine the prompting features of the spatial structure prompting information and the rebound embedding features of the rebound-related data; The prompt features and the bounce embedding features are input into the semantic labeling neural network to obtain the multiple semantic sub-region labels output by the semantic labeling neural network; Spatiotemporal correlation analysis is performed based on the multiple semantic sub-region labels to obtain the spatiotemporal correlation relationship.
[0009] Optionally, the step of performing spatiotemporal correlation analysis based on the multiple semantic sub-region labels to obtain the spatiotemporal correlation relationship includes: Based on the multiple semantic sub-region labels, a third spatial clustering feature is obtained through spatial-level clustering processing. Based on the third spatial clustering features, the spatiotemporal correlation is generated through temporal correlation analysis.
[0010] Optionally, the step of obtaining third-space clustering features based on the multiple semantic sub-region labels through spatial-level clustering processing includes: Determine the label features of each of the multiple semantic sub-region labels; Based on the label features, spatial clustering is performed to obtain the first spatial clustering feature; The first spatial clustering feature is dimensionality reduced to obtain the second spatial clustering feature, and the label features and the second spatial clustering feature are fused to obtain the third spatial clustering feature.
[0011] Optionally, generating the spatiotemporal correlation based on the third spatial clustering features through temporal correlation analysis includes: Temporal correlation analysis is performed on the third spatial clustering features to obtain the first temporal correlation features; The first temporal correlation feature is dimensionality reduced to obtain the second temporal correlation feature, and the second temporal correlation feature is fused with the third spatial clustering feature to obtain the spatiotemporal correlation feature; The spatiotemporal correlation relationship is generated based on the spatiotemporal correlation features.
[0012] Optionally, a deep learning model is used to implement the pattern recognition processing, and the deep learning model includes fully connected layers; The step of obtaining the concrete rebound strength detection result corresponding to the concrete area to be detected through pattern recognition processing based on the prior features and the rebound-related data includes: For each sub-region rebound data, determine the sensor data in the rebound-related data that matches the rebound data of that sub-region, extract the sensor data features of the sensor data, and construct the feature pair corresponding to the rebound data of that sub-region based on the sensor data features and the prior features corresponding to the rebound data of that sub-region. Determine the importance weight of each sub-region corresponding to the rebound data of the multiple sub-regions; Based on the importance weights of the sub-regions, each feature pair is weighted and fused to obtain a weighted feature pair; The weighted feature pairs are input into the fully connected layer to obtain the concrete rebound strength test results output by the fully connected layer.
[0013] Secondly, embodiments of this application provide a concrete rebound strength testing device based on pattern recognition processing, comprising: The rebound data acquisition module is used to acquire rebound-related data detected by sensors in the concrete area to be tested. The sub-region rebound data generation module is used to generate multiple sub-region rebound data based at least on the rebound-related data; The prior model matching module is used to match at least one prior model corresponding to the rebound data of the plurality of sub-regions from a pre-built set of rebound-intensity prior models. The set of rebound-intensity prior models includes a plurality of prior models corresponding to a plurality of sets of rebound-intensity sample data. Each of the plurality of prior models is trained based on its corresponding rebound-intensity sample data. The prior feature generation module is used to input the rebound data of each sub-region into its corresponding prior model to generate the prior features of the rebound data of each sub-region. The pattern recognition module is used to obtain the concrete rebound strength detection result corresponding to the concrete area to be detected through pattern recognition processing based on the prior features and the rebound-related data.
[0014] Thirdly, embodiments of this application provide a non-transitory computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the steps of the method described in any of the above-mentioned embodiments.
[0015] Fourthly, embodiments of this application provide a computer device including a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein the processor executes the computer program to implement the steps of the method described in any of the preceding claims.
[0016] In summary, the embodiments of this application have at least the following beneficial effects: In this embodiment, rebound-related data detected by sensors in the concrete area to be tested is acquired; multiple sub-region rebound data are generated based on the rebound-related data; at least one prior model corresponding to the multiple sub-region rebound data is matched from a pre-constructed set of rebound-strength prior models, wherein the set of rebound-strength prior models includes multiple prior models corresponding to multiple sets of rebound-strength sample data, and each of the multiple prior models is trained based on its corresponding rebound-strength sample data; each sub-region rebound data is input into its corresponding prior model to generate prior features for each sub-region rebound data; based on the prior features and the rebound-related data, a concrete rebound strength detection result corresponding to the concrete area to be tested is obtained through pattern recognition processing. In this way, a corresponding prior model can be matched for each sub-region rebound data to obtain corresponding prior features, which can be combined with the originally measured rebound-related data to improve the recognition accuracy of pattern recognition processing, thereby obtaining a more accurate concrete rebound strength detection result, and thus improving the accuracy of compressive strength prediction for concrete components. Attached Figure Description
[0017] Figure 1 This is a schematic flowchart of the concrete rebound strength detection method based on pattern recognition processing provided in the embodiments of this application; Figure 2 This is a schematic diagram of the concrete rebound strength testing device based on pattern recognition processing provided in the embodiments of this application; Figure 3 This is a schematic diagram of the structure of the computer device provided in the embodiments of this application. Detailed Implementation
[0018] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments / examples are only a part of the embodiments / examples of this application, and not all of the embodiments / examples. Based on the embodiments / examples in this application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of this application.
[0019] In the description of this application, the terms "first," "second," "third," etc., are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. Therefore, a feature defined with "first," "second," "third," etc., may explicitly or implicitly include one or more of that feature. In the description of this application, unless otherwise stated, "multiple" means two or more. In the description of this application, the term "comprising" and its variations are open-ended, meaning "including but not limited to." The term "based on" means "at least partially based on." The term "according to" means "at least partially according to." The term "one embodiment / example" means "at least one embodiment / example"; the term "another embodiment / example" means "at least one additional embodiment / example"; the term "some embodiments / examples" means "at least some embodiments / examples."
[0020] In the description of this application, it should be noted that, unless otherwise expressly specified and limited, the terms "installation," "connection," and "linking" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal connection between two components. Those skilled in the art can understand the specific meaning of the above terms in this application based on the specific circumstances.
[0021] In the description of this application, it should be noted that, unless otherwise defined, all technical and scientific terms used in this application have the same meaning as commonly understood by one of ordinary skill in the art. The terminology used in this application is for the purpose of describing specific embodiments only and is not intended to limit the application. Those skilled in the art can understand the specific meaning of the above terms in this application according to the specific circumstances.
[0022] Firstly, see [the following] Figure 1 The diagram shows a schematic flowchart of a concrete rebound strength detection method based on pattern recognition processing provided in an embodiment of this application. This concrete rebound strength detection method based on pattern recognition processing can be applied to a computer device with data processing capabilities. The method includes S101-S105, as detailed below.
[0023] S101, acquire rebound-related data detected by the sensor in the concrete area to be tested.
[0024] In some examples, there can be multiple sensors, which can be deployed at multiple measurement points on the concrete area to be inspected. Each sensor can be used to collect measurement data at its corresponding measurement point. The rebound-related data can include the measurement data corresponding to each of the multiple sensors. For example, the multiple sensors can correspond to multiple rebound hammers, each containing a corresponding sensor. The sensor in each rebound hammer can include at least one of the following: at least one displacement sensor and at least one acceleration sensor. The measurement data can be used to indicate rebound parameters, such as at least one of the following rebound parameters: rebound value, angle, and carbonation depth.
[0025] In some examples, the concrete area to be inspected may include at least a portion of the surface area of a large concrete member, and / or multiple large concrete members (where each large concrete member is at least partially mechanically connected to other large concrete members, and each large concrete member may be mechanically connected to any other large concrete member via at least one large concrete member or directly to it). The multiple large concrete members may include at least one of the following: large beam members, large column members, large slab members, etc.
[0026] S102, at least based on the rebound-related data, generate multiple sub-region rebound data.
[0027] In some examples, the rebound-related data can be directly divided into regions according to the location of the measurement point to obtain multiple sub-region rebound data.
[0028] S103, from the pre-constructed set of rebound-intensity prior models, at least one prior model corresponding to the rebound data of the plurality of sub-regions is matched, wherein the set of rebound-intensity prior models includes a plurality of prior models corresponding to a plurality of sets of rebound-intensity sample data, and each of the plurality of prior models is trained based on its corresponding rebound-intensity sample data.
[0029] In some examples, the prior model can be a pre-trained machine learning model (such as a neural network, regression tree, etc.), and each prior model can correspond to the mapping relationship between rebound and strength under a specific set of working conditions (such as concrete type, age). The rebound-strength sample data corresponding to each prior model can be real test data collected under the corresponding specific set of working conditions.
[0030] Furthermore, the prior model can output the corresponding concrete rebound strength estimation result based on the sub-region rebound data (the prior features can be intermediate representations generated during the model's operation to output the concrete rebound strength estimation result, such as the output of the feature fusion and representation layer described below; the prior features can be used to characterize the strength trend information implied under the specific working conditions indicated by the sub-region rebound data). The prior model can be a model that has been trained to have the predictive ability to use sub-region rebound data as model input and concrete rebound strength estimation result as model output. In specific training, rebound-strength sample data (including sample rebound data and its corresponding sample expected strength) can be used as sample data to obtain the predicted strength generated by the model based on the sample rebound data. Based on the difference between the predicted strength and the sample expected strength, a general loss function is used to calculate the loss value, and a general training algorithm (such as gradient descent) is used to train the model based on the loss value, so that the trained model can have the above-mentioned capabilities. For example, the model may include an input representation layer, a feature fusion and representation layer, and a prediction output layer. The input representation layer can receive data from the input model and convert the data into the desired feature vector form. For example, the input representation layer can use word embeddings or pre-trained language models (such as bidirectional language representation models based on the Transformer architecture) to generate semantic vectors, and / or use embedding layers to generate dense vectors. The feature fusion and representation layer can be used to fuse the feature vectors converted by the input representation layer. For example, the feature fusion and representation layer can implement the fusion through fully connected layers or attention mechanism layers. The prediction output layer can be used to generate prediction results based on the fused features. For example, the prediction output layer can use a softmax layer to output the probability distribution for different categories, and then output the prediction result based on the probability (for example, it can output the top one or more classification results with the highest probability as the prediction result).
[0031] In some examples, a prior model corresponding to each sub-region's rebound data can be matched by analyzing the similarity between the operating conditions indicated by the rebound data of each sub-region and the specific operating conditions corresponding to each prior model. It is easy to understand that each of the at least one prior model mentioned above can correspond to at least one sub-region's rebound data.
[0032] S104, input the rebound data of each sub-region into its corresponding prior model to generate the prior features of the rebound data of each sub-region.
[0033] In some examples, the rebound data of each sub-region can be input into its corresponding prior model to obtain the prior features of the rebound data of each sub-region generated during model operation.
[0034] S105, based on the prior features and the rebound-related data, the concrete rebound strength detection result corresponding to the concrete area to be detected is obtained through pattern recognition processing.
[0035] In some examples, this pattern recognition process can use statistical, machine learning, or deep learning methods, with prior features as an aid, to identify regular patterns in rebound-related data, and then map the identified regular patterns to compressive strength, thereby forming a concrete rebound strength test result corresponding to the concrete area to be tested. It is understood that this concrete rebound strength test result is used to characterize the estimated compressive strength of the concrete in the concrete area to be tested, and can be expressed in MPa.
[0036] In this embodiment, the prior models are derived from real-collected rebound-strength data. Each prior model can be used to characterize a specific real working condition (which can be understood as a material-environment combination) without relying on a single empirical formula, thus effectively avoiding the black-box risks associated with empirical formulas. For example, for high-volume fly ash concrete, the error of empirical formulas is generally large, while this embodiment can establish a separate prior model for it, thereby significantly reducing the deviation.
[0037] In the embodiments of this application, the most suitable prior model can be matched for the rebound data of each sub-region to achieve on-demand correction. For example, for non-uniform pouring areas (such as honeycomb pits) in the concrete area to be tested, the corresponding prior model can be configured with lower weights to reduce systematic bias.
[0038] In one optional implementation, generating multiple sub-region rebound data based at least on the rebound-related data includes: Based on the spatial structure prompt information and the rebound-related data, a semantic sequence is generated, wherein the semantic sequence includes multiple semantic sub-region labels and the association relationship between the multiple semantic sub-region labels, and the multiple sub-region rebound data corresponds one-to-one with the multiple semantic sub-region labels; Based on the semantic sequence, the bounce-related data are aggregated to obtain the semantic sub-region bounce data corresponding to each of the multiple semantic sub-region labels; During the generation process of each semantic sub-region label, relevant feature information of each semantic sub-region label during the generation process is obtained, and the rebound intensity of the sub-region corresponding to each semantic sub-region label is determined based on the relevant feature information corresponding to each semantic sub-region label. Based on the sub-region bounce intensity and the semantic sub-region bounce data corresponding to each semantic sub-region label, sub-region bounce data corresponding to each semantic sub-region label is generated.
[0039] In this embodiment, instead of simply dividing the rebound data into sub-regions using a fixed grid, a semantically reasonable sub-region structure is dynamically created. This ensures that the semantics of each location within each sub-region are similar, resulting in high similarity in both spatial location and strength characteristics among locations within the same semantic sub-region. This allows for more accurate analysis of the concrete compressive strength of each semantic sub-region and / or the overall concrete compressive strength of the concrete region under test. For example, the concrete region under test can be divided into a central dense area, a weak edge area, and a corner carbonized area. These areas are not necessarily divided entirely according to physical location. Simply using a fixed grid for division might misclassify measurement data from points that actually belong to one area into another, leading to unreasonable region division. This embodiment, by combining the analyzed semantics, effectively reduces the probability of unreasonable region division.
[0040] In some examples, this spatial structure information can be used to indicate the aforementioned dense central area, weak edge area, and carbonized corner area.
[0041] In some examples, the semantic sequence can be an ordered or unordered sequence of labels, which may include multiple semantic sub-region labels. These semantic sub-region labels can be used to describe the semantic category to which each measurement point belongs. For example, spatial structure cue information and rebound-related data can be input into a pre-trained Transformer decoder or graph neural network to obtain the output semantic sequence. Here, each measurement point can be considered as a token. Alternatively, spatial structure cue information and rebound-related data can be input into a large model to obtain the semantic sequence output by that large model. The spatial structure cue information can be used to prompt the large model to analyze the rebound-related data to obtain the semantic sequence.
[0042] In some examples, data belonging to the same semantic sub-region label in the bounce-related data can be aggregated to form semantic sub-region bounce data corresponding to that semantic sub-region label.
[0043] In some examples, during the process of generating semantic sub-region labels (i.e., during the partial process of generating semantic sequences), intermediate representation information generated by the corresponding model can be obtained synchronously. For example, the intermediate representation information may include the decoder hidden state corresponding to the output label of the Transformer decoder, and the node embedding of the node corresponding to each semantic sub-region label of the graph neural network after message passing.
[0044] In some examples, the relevant feature information corresponding to each semantic sub-region label can be input into a regression head (MLP) to obtain the sub-region rebound intensity corresponding to each semantic sub-region label output by the MLP. For example, there can be multiple MLPs, and the relevant feature information corresponding to multiple semantic sub-region labels can be input into multiple MLPs in parallel, so that multiple MLPs output their respective sub-region rebound intensities in parallel.
[0045] In some examples, the rebound data of the sub-region may include the rebound strength and semantic rebound data of the corresponding sub-region, so that the rebound data of the sub-region is no longer the original measurement data, but may also include the rebound strength of the sub-region, in order to improve the accuracy of the concrete rebound strength detection results generated as input for subsequent pattern recognition processing.
[0046] In one optional implementation, the association relationship includes a spatiotemporal association relationship; The generation of a semantic sequence based on spatial structure prompting information and rebound-related data includes: Determine the prompting features of the spatial structure prompting information and the rebound embedding features of the rebound-related data; The prompt features and the bounce embedding features are input into the semantic labeling neural network to obtain the multiple semantic sub-region labels output by the semantic labeling neural network; Spatiotemporal correlation analysis is performed based on the multiple semantic sub-region labels to obtain the spatiotemporal correlation relationship.
[0047] In some examples, the spatial structure cue information can be textual modal information, in which case the cue features can be obtained by extracting textual features from the spatial structure cue information using a first encoder.
[0048] In some examples, bounce-related data can be encoded using a second encoder to obtain bounce-embedded features.
[0049] In some examples, the semantic labeling neural network, the first encoder, and the second encoder described above may be included in the Transformer decoder or the graph neural network described above to achieve the relevant functions involved in the above embodiments.
[0050] In one optional implementation, the step of performing spatiotemporal correlation analysis based on the plurality of semantic sub-region labels to obtain the spatiotemporal correlation relationship includes: Based on the multiple semantic sub-region labels, a third spatial clustering feature is obtained through spatial-level clustering processing. Based on the third spatial clustering features, the spatiotemporal correlation is generated through temporal correlation analysis.
[0051] In one optional implementation, the step of obtaining third spatial clustering features based on the plurality of semantic sub-region labels through spatial-level clustering processing includes: Determine the label features of each of the multiple semantic sub-region labels; Based on the label features, spatial clustering is performed to obtain the first spatial clustering feature; The first spatial clustering feature is dimensionality reduced to obtain the second spatial clustering feature, and the label features and the second spatial clustering feature are fused to obtain the third spatial clustering feature.
[0052] In one optional implementation, generating the spatiotemporal correlation based on the third spatial clustering features through temporal correlation analysis includes: Temporal correlation analysis is performed on the third spatial clustering features to obtain the first temporal correlation features; The first temporal correlation feature is dimensionality reduced to obtain the second temporal correlation feature, and the second temporal correlation feature is fused with the third spatial clustering feature to obtain the spatiotemporal correlation feature; The spatiotemporal correlation relationship is generated based on the spatiotemporal correlation features.
[0053] As can be seen from the above embodiments, the embodiments of this application provide an exemplary embodiment of "spatiotemporal correlation analysis for multiple semantic sub-region labels".
[0054] Determine the label features of each of the multiple semantic sub-region labels; Based on the label features, spatial clustering is performed to obtain the first spatial clustering feature; The first spatial clustering feature is dimensionality reduced to obtain the second spatial clustering feature, and the label features and the second spatial clustering feature are fused to obtain the third spatial clustering feature; Temporal correlation analysis is performed on the third spatial clustering features to obtain the first temporal correlation features; The first temporal correlation feature is dimensionality reduced to obtain the second temporal correlation feature, and the second temporal correlation feature is fused with the third spatial clustering feature to obtain the spatiotemporal correlation feature; The spatiotemporal correlation relationship is generated based on the spatiotemporal correlation features.
[0055] In this embodiment, a multi-level feature extraction, clustering, dimensionality reduction, and fusion mechanism is introduced. This mechanism can extract more robust and discriminative spatial structure patterns from semantic sub-region labels, and on this basis, construct dynamic correlations in the time dimension (quasi-temporal evolution). It should be noted that in some cases, concrete rebound detection can be a single static measurement. In this case, the time correlation described in this embodiment can be understood as implicit evolutionary modeling (such as multi-age trends, the order of data collection at each measurement point, or the logic of intensity changes). Therefore, this time correlation can be used to indicate the direction of intensity evolution in space, which can be abstractly understood as a pseudo-time flow. This embodiment can transform the physical intuition related to intensity distribution and change into computable time correlation features, enabling the model used for pattern recognition to not only predict intensity values but also understand the rationality of the intensity field, thereby significantly improving the accuracy, robustness, and interpretability of the detection.
[0056] For example, the above spatial clustering process can be implemented by pooling, such as attention-weighted pooling, global average pooling of CNN (Convolutional Neural Network), etc., so as to obtain a first spatial clustering feature that can characterize the whole of the concrete area to be detected.
[0057] For example, the first spatial clustering feature can be transformed into a low-dimensional space to achieve dimensionality reduction and obtain the second spatial clustering feature. The second spatial clustering feature obtained by dimensionality reduction is then fused with the initial label features to obtain the third spatial clustering feature, so that the third spatial clustering feature can retain the effective information corresponding to the initial label features while carrying the spatial relationship.
[0058] For example, a first temporal correlation feature can be obtained by performing temporal correlation analysis on the third spatial clustering feature, and then transformed into a low-dimensional space to achieve dimensionality reduction and obtain a second temporal correlation feature. This second temporal correlation feature is then further fused with the aforementioned third spatial clustering feature to obtain the final spatiotemporal correlation feature generated through spatiotemporal correlation analysis. Furthermore, this embodiment can employ models capable of processing time-series data, such as Transformer Encoder or RNN (Recurrent Neural Network), to perform this temporal correlation analysis.
[0059] In one alternative implementation, a deep learning model is used to implement the pattern recognition processing, the deep learning model including fully connected layers; The step of obtaining the concrete rebound strength detection result corresponding to the concrete area to be detected through pattern recognition processing based on the prior features and the rebound-related data includes: For each sub-region rebound data, determine the sensor data in the rebound-related data that matches the rebound data of that sub-region, extract the sensor data features of the sensor data, and construct the feature pair corresponding to the rebound data of that sub-region based on the sensor data features and the prior features corresponding to the rebound data of that sub-region. Determine the importance weight of each sub-region corresponding to the rebound data of the multiple sub-regions; Based on the importance weights of the sub-regions, each feature pair is weighted and fused to obtain a weighted feature pair; The weighted feature pairs are input into the fully connected layer to obtain the concrete rebound strength test results output by the fully connected layer.
[0060] In some examples, the Transformer model can be used to analyze and obtain the importance weights of each sub-region corresponding to the rebound data of multiple sub-regions.
[0061] Secondly, correspondingly, the embodiments of this application also provide a concrete rebound strength detection device based on pattern recognition processing, which can realize all the processes of the concrete rebound strength detection method based on pattern recognition processing provided in the above embodiments.
[0062] See Figure 2 The diagram shows a schematic of the concrete rebound strength testing device based on pattern recognition processing provided in this application embodiment. The concrete rebound strength testing device 200 based on pattern recognition processing includes: The rebound-related data acquisition module 201 is used to acquire rebound-related data detected by sensors in the concrete area to be tested. The sub-region rebound data generation module 202 is used to generate multiple sub-region rebound data based at least on the rebound-related data; The prior model matching module 203 is used to match at least one prior model corresponding to the rebound data of the plurality of sub-regions from a pre-constructed set of rebound-intensity prior models, wherein the set of rebound-intensity prior models includes a plurality of prior models corresponding to a plurality of sets of rebound-intensity sample data, and each of the plurality of prior models is trained based on its corresponding rebound-intensity sample data; The prior feature generation module 204 is used to input the rebound data of each sub-region into its corresponding prior model to generate prior features of the rebound data of each sub-region. The pattern recognition module 205 is used to obtain the concrete rebound strength detection result corresponding to the concrete area to be detected by pattern recognition processing based on the prior features and the rebound-related data.
[0063] In one optional implementation, generating multiple sub-region rebound data based at least on the rebound-related data includes: Based on the spatial structure prompt information and the rebound-related data, a semantic sequence is generated, wherein the semantic sequence includes multiple semantic sub-region labels and the association relationship between the multiple semantic sub-region labels, and the multiple sub-region rebound data corresponds one-to-one with the multiple semantic sub-region labels; Based on the semantic sequence, the bounce-related data are aggregated to obtain the semantic sub-region bounce data corresponding to each of the multiple semantic sub-region labels; During the generation process of each semantic sub-region label, relevant feature information of each semantic sub-region label during the generation process is obtained, and the rebound intensity of the sub-region corresponding to each semantic sub-region label is determined based on the relevant feature information corresponding to each semantic sub-region label. Based on the sub-region bounce intensity and the semantic sub-region bounce data corresponding to each semantic sub-region label, sub-region bounce data corresponding to each semantic sub-region label is generated.
[0064] In one optional implementation, the association relationship includes a spatiotemporal association relationship; The generation of a semantic sequence based on spatial structure prompting information and rebound-related data includes: Determine the prompting features of the spatial structure prompting information and the rebound embedding features of the rebound-related data; The prompt features and the bounce embedding features are input into the semantic labeling neural network to obtain the multiple semantic sub-region labels output by the semantic labeling neural network; Spatiotemporal correlation analysis is performed based on the multiple semantic sub-region labels to obtain the spatiotemporal correlation relationship.
[0065] In one optional implementation, the step of performing spatiotemporal correlation analysis based on the plurality of semantic sub-region labels to obtain the spatiotemporal correlation relationship includes: Based on the multiple semantic sub-region labels, a third spatial clustering feature is obtained through spatial-level clustering processing. Based on the third spatial clustering features, the spatiotemporal correlation is generated through temporal correlation analysis.
[0066] In one optional implementation, the step of obtaining third spatial clustering features based on the plurality of semantic sub-region labels through spatial-level clustering processing includes: Determine the label features of each of the multiple semantic sub-region labels; Based on the label features, spatial clustering is performed to obtain the first spatial clustering feature; The first spatial clustering feature is dimensionality reduced to obtain the second spatial clustering feature, and the label features and the second spatial clustering feature are fused to obtain the third spatial clustering feature.
[0067] In one optional implementation, generating the spatiotemporal correlation based on the third spatial clustering features through temporal correlation analysis includes: Temporal correlation analysis is performed on the third spatial clustering features to obtain the first temporal correlation features; The first temporal correlation feature is dimensionality reduced to obtain the second temporal correlation feature, and the second temporal correlation feature is fused with the third spatial clustering feature to obtain the spatiotemporal correlation feature; The spatiotemporal correlation relationship is generated based on the spatiotemporal correlation features.
[0068] In one alternative implementation, a deep learning model is used to implement the pattern recognition processing, the deep learning model including fully connected layers; The step of obtaining the concrete rebound strength detection result corresponding to the concrete area to be detected through pattern recognition processing based on the prior features and the rebound-related data includes: For each sub-region rebound data, determine the sensor data in the rebound-related data that matches the rebound data of that sub-region, extract the sensor data features of the sensor data, and construct the feature pair corresponding to the rebound data of that sub-region based on the sensor data features and the prior features corresponding to the rebound data of that sub-region. Determine the importance weight of each sub-region corresponding to the rebound data of the multiple sub-regions; Based on the importance weights of the sub-regions, each feature pair is weighted and fused to obtain a weighted feature pair; The weighted feature pairs are input into the fully connected layer to obtain the concrete rebound strength test results output by the fully connected layer.
[0069] Thirdly, embodiments of this application provide a non-transitory computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the steps of the method described in any of the above-mentioned embodiments.
[0070] Fourthly, embodiments of this application provide a computer device including a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein the processor executes the computer program to implement the steps of the method described in any of the preceding claims.
[0071] See Figure 3The computer device in this embodiment includes a processor 301, a memory 302, and a computer program stored in the memory 302 and executable on the processor 301, such as a concrete rebound strength detection program based on pattern recognition processing. When the processor 301 executes the computer program, it implements the steps in the various embodiments of the concrete rebound strength detection method based on pattern recognition processing described above, for example... Figure 1 The steps S101-S105 are shown.
[0072] For example, the computer program may be divided into one or more modules / units, which are stored in the memory 302 and executed by the processor 301 to complete this application. The one or more modules / units may be a series of computer program instruction segments capable of performing a specific function, which describe the execution process of the computer program in the computer device.
[0073] The computer device may be a desktop computer, laptop, handheld computer, or cloud server, etc. The computer device may include, but is not limited to, a processor 301 and a memory 302. Those skilled in the art will understand that the schematic diagram is merely an example of a computer device and does not constitute a limitation on the computer device. It may include more or fewer components than shown, or combine certain components, or different components. For example, the computer device may also include input / output devices, network access devices, buses, etc.
[0074] The processor 301 can be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor can be a microprocessor, or the processor 301 can be any conventional processor. The processor 301 is the control center of the computer device, connecting various parts of the entire computer device through various interfaces and lines.
[0075] The memory 302 can be used to store the computer programs and / or modules. The processor 301 implements various functions of the computer device by running or executing the computer programs and / or modules stored in the memory 302 and calling the data stored in the memory 302. The memory 302 may mainly include a program storage area and a data storage area. The program storage area may store the operating system, at least one application program required for a function (such as sound playback function, image playback function, etc.), etc.; the data storage area may store data created according to the use of the mobile phone (such as audio data, phonebook, etc.). In addition, the memory 302 may include high-speed random access memory, and may also include non-volatile memory, such as hard disk, memory, plug-in hard disk, smart media card (SMC), secure digital (SD) card, flash card, at least one disk storage device, flash memory device, or other volatile solid-state storage device.
[0076] Wherein, if the modules / units integrated into the computer device are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, all or part of the processes in the methods of the above embodiments can also be implemented by a computer program instructing related hardware. The computer program can be stored in a non-transitory computer-readable storage medium. When the computer program is executed by the processor 301, it can implement the steps of the various method embodiments described above. Wherein, the computer program includes computer program code, which can be in the form of source code, object code, executable file, or some intermediate form, etc. The computer-readable medium can include: any entity or device capable of carrying the computer program code, recording medium, USB flash drive, portable hard drive, magnetic disk, optical disk, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signal, telecommunication signal, and software distribution medium, etc.
[0077] In summary, the embodiments of this application have at least the following beneficial effects: In this embodiment, rebound-related data detected by sensors in the concrete area to be tested is acquired; multiple sub-region rebound data are generated based on the rebound-related data; at least one prior model corresponding to the multiple sub-region rebound data is matched from a pre-constructed set of rebound-strength prior models, wherein the set of rebound-strength prior models includes multiple prior models corresponding to multiple sets of rebound-strength sample data, and each of the multiple prior models is trained based on its corresponding rebound-strength sample data; each sub-region rebound data is input into its corresponding prior model to generate prior features for each sub-region rebound data; based on the prior features and the rebound-related data, a concrete rebound strength detection result corresponding to the concrete area to be tested is obtained through pattern recognition processing. In this way, a corresponding prior model can be matched for each sub-region rebound data to obtain corresponding prior features, which can be combined with the originally measured rebound-related data to improve the recognition accuracy of pattern recognition processing, thereby obtaining a more accurate concrete rebound strength detection result, and thus improving the accuracy of compressive strength prediction for concrete components.
[0078] Through the above description of the embodiments, those skilled in the art can clearly understand that this application can be implemented by means of software plus necessary hardware platforms, or it can be implemented entirely by hardware. Based on this understanding, all or part of the technical solutions of this application that contribute to the background technology can be embodied in the form of a software product. This computer software product can be stored in a storage medium, such as ROM (Read-Only Memory) / RAM (Random Access Memory), magnetic disk, optical disk, etc., including several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in various embodiments or some parts of the embodiments of this application.
[0079] The above description is the preferred embodiment of this application. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principle of this application, and these improvements and modifications are also considered to be within the scope of protection of this application.
Claims
1. A method for detecting the rebound strength of concrete based on pattern recognition processing, characterized in that, include: Acquire rebound-related data detected by sensors in the concrete area to be tested; Based on the rebound-related data, multiple sub-region rebound data are generated; From a pre-built set of rebound-intensity prior models, at least one prior model corresponding to the rebound data of the plurality of sub-regions is matched, wherein the set of rebound-intensity prior models includes a plurality of prior models corresponding to a plurality of sets of rebound-intensity sample data, and each of the plurality of prior models is trained based on its corresponding rebound-intensity sample data; The rebound data of each sub-region is input into its corresponding prior model to generate prior features of the rebound data of each sub-region; Based on the prior features and the rebound-related data, the rebound strength test result of the concrete corresponding to the concrete area to be tested is obtained through pattern recognition processing.
2. The method according to claim 1, characterized in that, The generation of multiple sub-region rebound data based at least on the rebound-related data includes: Based on the spatial structure prompt information and the rebound-related data, a semantic sequence is generated, wherein the semantic sequence includes multiple semantic sub-region labels and the association relationship between the multiple semantic sub-region labels, and the multiple sub-region rebound data corresponds one-to-one with the multiple semantic sub-region labels; Based on the semantic sequence, the bounce-related data are aggregated to obtain the semantic sub-region bounce data corresponding to each of the multiple semantic sub-region labels; During the generation process of each semantic sub-region label, relevant feature information of each semantic sub-region label during the generation process is obtained, and the rebound intensity of the sub-region corresponding to each semantic sub-region label is determined based on the relevant feature information corresponding to each semantic sub-region label. Based on the sub-region bounce intensity and the semantic sub-region bounce data corresponding to each semantic sub-region label, sub-region bounce data corresponding to each semantic sub-region label is generated.
3. The method according to claim 2, characterized in that, The relationships include spatiotemporal relationships; The generation of a semantic sequence based on spatial structure prompting information and rebound-related data includes: Determine the prompting features of the spatial structure prompting information and the rebound embedding features of the rebound-related data; The prompt features and the bounce embedding features are input into the semantic labeling neural network to obtain the multiple semantic sub-region labels output by the semantic labeling neural network; Spatiotemporal correlation analysis is performed based on the multiple semantic sub-region labels to obtain the spatiotemporal correlation relationship.
4. The method according to claim 3, characterized in that, The spatiotemporal correlation analysis based on the multiple semantic sub-region labels to obtain the spatiotemporal correlation relationship includes: Based on the multiple semantic sub-region labels, a third spatial clustering feature is obtained through spatial-level clustering processing. Based on the third spatial clustering features, the spatiotemporal correlation is generated through temporal correlation analysis.
5. The method according to claim 4, characterized in that, The third spatial clustering features are obtained based on the multiple semantic sub-region labels through spatial-level clustering processing, including: Determine the label features of each of the multiple semantic sub-region labels; Based on the label features, spatial clustering is performed to obtain the first spatial clustering feature; The first spatial clustering feature is dimensionality reduced to obtain the second spatial clustering feature, and the label features and the second spatial clustering feature are fused to obtain the third spatial clustering feature.
6. The method according to claim 4, characterized in that, The generation of the spatiotemporal correlation based on the third spatial clustering features through temporal correlation analysis includes: Temporal correlation analysis is performed on the third spatial clustering features to obtain the first temporal correlation features; The first temporal correlation feature is dimensionality reduced to obtain the second temporal correlation feature, and the second temporal correlation feature is fused with the third spatial clustering feature to obtain the spatiotemporal correlation feature; The spatiotemporal correlation relationship is generated based on the spatiotemporal correlation features.
7. The method according to claim 1, characterized in that, A deep learning model is used to implement the pattern recognition processing, and the deep learning model includes fully connected layers; The step of obtaining the concrete rebound strength detection result corresponding to the concrete area to be detected through pattern recognition processing based on the prior features and the rebound-related data includes: For each sub-region rebound data, determine the sensor data in the rebound-related data that matches the rebound data of that sub-region, extract the sensor data features of the sensor data, and construct the feature pair corresponding to the rebound data of that sub-region based on the sensor data features and the prior features corresponding to the rebound data of that sub-region. Determine the importance weight of each sub-region corresponding to the rebound data of the multiple sub-regions; Based on the importance weights of the sub-regions, each feature pair is weighted and fused to obtain a weighted feature pair; The weighted feature pairs are input into the fully connected layer to obtain the concrete rebound strength test result output by the fully connected layer.
8. A concrete rebound strength testing device based on pattern recognition processing, characterized in that, include: The rebound data acquisition module is used to acquire rebound-related data detected by sensors in the concrete area to be tested. The sub-region rebound data generation module is used to generate multiple sub-region rebound data based at least on the rebound-related data; The prior model matching module is used to match at least one prior model corresponding to the rebound data of the plurality of sub-regions from a pre-built set of rebound-intensity prior models. The set of rebound-intensity prior models includes a plurality of prior models corresponding to a plurality of sets of rebound-intensity sample data. Each of the plurality of prior models is trained based on its corresponding rebound-intensity sample data. The prior feature generation module is used to input the rebound data of each sub-region into its corresponding prior model to generate the prior features of the rebound data of each sub-region. The pattern recognition module is used to obtain the concrete rebound strength detection result corresponding to the concrete area to be detected through pattern recognition processing based on the prior features and the rebound-related data.
9. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the method described in any one of claims 1-7.
10. A computer device, characterized in that, The method includes a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein the processor, when executing the computer program, implements the method of any one of claims 1-7.