Concrete crack detection method and system combined with time sequence neural network

Abstract

The invention discloses a concrete crack detection method and system combined with a time sequence neural network, and mainly relates to the technical field of concrete detection. Comprising the steps of performing concrete crack real-time detection on a target application scene by using a detection assembly to obtain a real-time concrete crack detection result; calling a target scene environment clue template based on the target application scene to obtain a scene environment clue sensing data sequence; performing environment interference trend characteristic interaction analysis on the scene environment clue sensing data sequence by using a long and short time sequence neural network to obtain environment interference trend characteristics; and performing causal perception fusion to obtain a fused concrete crack detection result. The method has the beneficial effects that the technical problem that the actual development condition of the concrete crack cannot be deeply detected and analyzed due to neglect of potential influence factors in the concrete crack detection in the prior art is solved, and the technical effects of obtaining the crack detection result containing the environmental interference trend and improving the concrete crack detection reliability are achieved.

Description

Technical Field

[0001] The present invention relates to the technical field of concrete detection, and in particular to a concrete crack detection method and system combined with a temporal neural network. Background Art

[0002] Concrete crack detection technology is widely used in building structural health monitoring to promptly detect cracks and assess their impact on structural safety. Traditional concrete crack detection methods typically rely on image processing, sensor detection, and manual inspection. However, these methods lack temporal continuity in the data collected and ignore environmental interference, resulting in a lag in concrete crack detection results.

[0003] The existing technology has a technical problem that concrete crack detection ignores potential influencing factors, resulting in the inability to conduct in-depth detection and analysis of the actual development of concrete cracks. Summary of the Invention

[0004] The present application provides a concrete crack detection method and system combined with a time series neural network, which is used to solve the technical problem that concrete crack detection in the prior art ignores potential influencing factors, resulting in the inability to conduct in-depth detection and analysis of the actual development of concrete cracks.

[0005] In view of the above problems, the present application provides a concrete crack detection method and system combined with a time series neural network.

[0006] In a first aspect of the present application, a method for detecting concrete cracks in combination with a temporal neural network is provided, the method comprising: A detection component is used to perform real-time detection of concrete cracks in a target application scenario to obtain real-time concrete crack detection results, wherein the concrete crack detection results include crack width-depth-length and crack extension direction; a target scene environment clue template is called based on the target application scenario, and continuous environmental perception data is extracted based on the target scene environment clue template to obtain a scene environment clue perception data sequence; a long-short time series neural network is used to interactively analyze the environmental interference trend characteristics of the scene environment clue perception data sequence to obtain environmental interference trend characteristics; a causal perception fusion of crack width-depth-length and crack extension direction is performed based on the environmental interference trend characteristics to obtain a fused concrete crack detection result.

[0007] Optionally, a scene environment clue template is called based on the target application scenario, and continuous environmental perception data is extracted based on the target scene environment clue template to obtain a scene environment clue perception data sequence, including: constructing a scene environment clue template library, wherein each scene environment clue template includes a scene feature vector; performing feature extraction on the target application scenario information of the target application scenario according to preset scene indicators to construct a target application scene feature vector; performing approximate matching based on the target application scene feature vector and the scene feature vector of each scene environment clue template in the scene environment clue template library, and taking the scene environment clue template corresponding to the maximum matching similarity as the target scene environment clue template.

[0008] Optionally, a scene environment clue template library is constructed, wherein each scene environment clue template includes a scene feature vector, including: obtaining a sample application scenario information set and a corresponding sample abnormal environment clue set; performing feature extraction on the sample application scenario information set according to preset scene indicators to obtain a sample application scenario feature vector set, and performing similar aggregation on the sample application scenario feature vector set to obtain K aggregated sample application scenario feature vector sets, wherein K is a positive integer; mapping and aggregating the sample abnormal environment clue set according to the K aggregated sample application scenario feature vector sets to obtain K aggregated sample abnormal environment clue sets, and performing union on the K aggregated sample abnormal environment clue sets to obtain K integrated sample abnormal environment clues; centrally integrating the K aggregated sample application scenario feature vector sets to obtain K scene feature vectors, and using the K scene feature vectors to identify the K integrated sample abnormal environment clues to construct a scene environment clue template library.

[0009] Optionally, K aggregated sample application scenario feature vector sets are centrally integrated to obtain K scenario feature vectors, including: extracting a first aggregated sample application scenario feature vector set from the K aggregated sample application scenario feature vector sets; combining the first aggregated sample application scenario feature vector sets in pairs, calculating the combined similarities respectively, and obtaining a combined similarity set; extracting the combined similarity set with any one of the first aggregated sample application scenario feature vectors in the first aggregated sample application scenario feature vector set, and performing mean calculation on the extracted results to obtain an integrated similarity set, wherein each integrated similarity corresponds to a first aggregated sample application scenario feature vector; taking the first aggregated sample application scenario feature vector corresponding to the maximum value in the integrated similarity set as the first scenario feature vector; performing combined similarity calculation on the K aggregated sample application scenario feature vector sets respectively, and integrating according to the calculation results to obtain K scenario feature vectors.

[0010] Optionally, a long-short time series neural network is used to interactively analyze the environmental interference trend characteristics of the scene environment clue perception data sequence to obtain the environmental interference trend characteristics, including: extracting the long time series trend feature extraction branch of the long-short time series neural network to perform feature extraction on the scene environment clue perception data sequence to obtain the long time series environmental interference trend characteristics; extracting the short time series trend feature extraction branch of the long-short time series neural network to perform feature extraction on the scene environment clue perception data sequence to obtain the short time series environmental interference trend characteristics; interactively analyzing the long time series environmental interference trend characteristics and the short time series environmental interference trend characteristics to obtain the environmental interference trend characteristics.

[0011] Optionally, the long-term environmental interference trend characteristics and the short-term environmental interference trend characteristics are interactively analyzed to obtain the environmental interference trend characteristics, including: using the cosine similarity formula to calculate the feature interaction similarity set of the long-term environmental interference trend characteristics and the short-term environmental interference trend characteristics; constructing an interactive analysis matrix based on the feature interaction similarity set, and using the interactive analysis matrix to enhance the short-term environmental interference trend characteristics to obtain the environmental interference trend characteristics.

[0012] Optionally, the method includes: constructing a causal perception fusion device; using the causal perception fusion device to analyze the environmental interference trend characteristics, crack width-depth-length and crack extension direction to obtain the fused concrete crack detection result.

[0013] A second aspect of the present application provides a concrete crack detection system combined with a temporal neural network, the system comprising: The crack detection result acquisition module is used to use the detection component to perform real-time detection of concrete cracks in the target application scenario to obtain real-time concrete crack detection results, wherein the concrete crack detection results include crack width-depth-length and crack extension direction; the perception data sequence acquisition module is used to call the target scene environment clue template based on the target application scenario, perform continuous environmental perception data extraction based on the target scene environment clue template, and obtain a scene environment clue perception data sequence; the environmental interference trend feature acquisition module is used to use the long and short time series neural network to perform environmental interference trend feature interactive analysis on the scene environment clue perception data sequence to obtain environmental interference trend features; the fused concrete crack detection result acquisition module is used to perform causal perception fusion of crack width-depth-length and crack extension direction based on the environmental interference trend features to obtain a fused concrete crack detection result.

[0014] One or more technical solutions provided in this application have at least the following technical effects or advantages: This application uses a detection component to perform real-time detection of concrete cracks in a target application scenario to obtain real-time concrete crack detection results, wherein the concrete crack detection results include crack width, depth, length and crack extension direction; based on the target application scenario, the target scene environment clue template is called, and the environment continuous perception data is extracted based on the target scene environment clue template to obtain a scene environment clue perception data sequence; a long-short time series neural network is used to interactively analyze the environmental interference trend characteristics of the scene environment clue perception data sequence to obtain environmental interference trend characteristics; based on the environmental interference trend characteristics, a causal perception fusion of the crack width, depth, length and crack extension direction is performed to obtain a fused concrete crack detection result. The technical effect of obtaining crack detection results that include environmental interference trends and improving the reliability of concrete crack detection is achieved. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Attachment Figure 1 This is a flow chart of a concrete crack detection method combined with a temporal neural network provided by an embodiment of the present invention.

[0016] Attachment Figure 2 This is a structural diagram of a concrete crack detection system combined with a time series neural network provided by an embodiment of the present invention.

[0017] Reference numerals shown in the accompanying drawings: Crack detection result acquisition module 11, perception data sequence acquisition module 12, environmental interference trend feature acquisition module 13, fusion concrete crack detection result acquisition module 14. DETAILED DESCRIPTION

[0018] The present invention will be further described below in conjunction with specific examples. It should be understood that these examples are only used to illustrate the present invention and are not intended to limit the scope of the present invention. In addition, it should be understood that after reading the content taught by the present invention, those skilled in the art can make various changes or modifications to the present invention, and these equivalent forms also fall within the scope defined by the claims attached to this application. It should be noted that the terms "including" and "having" are intended to cover non-exclusive inclusions. For example, a process, method, system, product or server that includes a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or modules that are not clearly listed or inherent to these processes, methods, products or devices.

[0019] Example 1, as shown in the attached Figure 1 As shown, the present application provides a concrete crack detection method combined with a temporal neural network, the method comprising: S1: Use the detection component to perform real-time detection of concrete cracks in the target application scenario to obtain real-time concrete crack detection results, where the concrete crack detection results include crack width, depth, length and crack extension direction; In one possible embodiment, the detection component refers to a device or sensor used for concrete crack detection, including a crack width meter, crack depth meter, laser scanner, etc. The detection component is used to obtain different characteristic data of concrete cracks. The target application scenario refers to the specific building or structural environment where crack detection is required, such as bridges, floor slabs, tunnels, etc. The concrete in these scenarios may have cracks, affecting the safety of the structure. The concrete crack detection results reflect the real-time status of the concrete cracks in the target application scenario, including the crack width, depth, length and crack extension direction.

[0020] Where width is the lateral size of the crack. Depth is the distance from the surface to its deepest point. Length is the distance the crack extends above the surface. Direction is the direction the crack extends along the surface of the structure, typically expressed as vertical, horizontal, or diagonal.

[0021] Crack width meters are preferably installed on both sides of a concrete crack or around its perimeter to measure the crack's width. Crack depth meters use probes, ultrasonic waves, or laser technology to measure crack depth. During the measurement process, the crack depth meter's probe or laser beam is inserted into the crack and accurately measures its depth. Laser scanners scan the concrete surface with a laser beam and use the reflected wave data to accurately construct a three-dimensional image of the crack. Laser scanners are able to capture the entire length and direction of the crack. This process is usually fully automated and can determine the crack's length and its propagation path across the surface.

[0022] By combining a variety of detection equipment (crack width meters, crack depth meters, laser scanners, etc.) to comprehensively measure the geometric characteristics of concrete cracks, the technical effect of providing basic information for the subsequent integration of crack detection results and environmental interference conditions is achieved.

[0023] S2: calling a target scene environment clue template based on the target application scenario, extracting continuous environment perception data based on the target scene environment clue template, and obtaining a scene environment clue perception data sequence; Furthermore, based on the target application scenario, a scene environment clue template is called, and based on the target scene environment clue template, continuous environment perception data is extracted to obtain a scene environment clue perception data sequence. In this embodiment, step S2 of the present application further includes: Constructing a scene environment clue template library, wherein each scene environment clue template includes a scene feature vector; Extracting features of target application scenario information of the target application scenario according to preset scenario indicators to construct a target application scenario feature vector; Based on the approximate matching between the target application scene feature vector and the scene feature vector of each scene environment clue template in the scene environment clue template library, the scene environment clue template corresponding to the maximum matching similarity is used as the target scene environment clue template.

[0024] In one possible embodiment, the target scenario environmental cue template is a template that reflects the environmental factors that interfere with concrete cracks in the target application scenario, including characteristics of environmental factors that may affect crack development in that scenario (such as temperature, humidity, and load). Each environmental cue template includes a scenario feature vector that represents the manifestation of these environmental factors in that scenario.

[0025] Using the target scene environmental clue template as the perception data extraction target, the system continuously extracts environmental perception data from environmental sensors deployed within the target application scene, obtaining a scene environmental clue perception data sequence. This scene environmental clue perception data sequence reflects the temporal changes in environmental data associated with concrete cracks within the target application scene. By obtaining this scene environmental clue perception data sequence, the system achieves the technical effect of providing data support for subsequent analysis of environmental interference trend characteristics within the target application scene.

[0026] Preferably, a scene environment clue template library is pre-built, wherein each scene environment clue template includes a scene feature vector. The scene feature vector is used to describe the representation of each scene environment clue template in the corresponding type of application scenario. The preset scene indicators are indicators used to describe the environmental conditions of the application scenario, including temperature, humidity, load, etc. The target application scenario information is used to describe the environmental conditions of the target application scenario.

[0027] In one embodiment, the target application scenario information is subjected to feature extraction using the preset scenario index as an index, and the extracted index is filled into an initially empty vector to obtain the target application scenario feature vector, wherein the target application scenario feature vector reflects the actual environment of the target application scenario.

[0028] Next, the cosine similarity formula is used to calculate the similarity between the target application scenario feature vector and the scene feature vector of each scene environment cue template in the scene environment cue template library, thereby obtaining a matching similarity set. The matching similarity set reflects the degree of similarity between the target application scenario feature vector and the scene feature vector corresponding to the template in the scene environment cue template library. The scene environment cue template corresponding to the maximum matching similarity is extracted and used as the target scene environment cue template.

[0029] Furthermore, a scene environment clue template library is constructed, wherein each scene environment clue template includes a scene feature vector. Step S2 of the embodiment of the present application further includes: Obtaining a sample application scenario information set and a corresponding sample abnormal environment clue set; Extract features from the sample application scenario information set according to preset scenario indicators to obtain a sample application scenario feature vector set, and aggregate the sample application scenario feature vector sets of the same type to obtain K aggregated sample application scenario feature vector sets, where K is a positive integer; Mapping and aggregating the sample abnormal environment clue set according to the K aggregated sample application scenario feature vector sets to obtain K aggregated sample abnormal environment clue sets, and performing a union on the K aggregated sample abnormal environment clue sets to obtain K integrated sample abnormal environment clues; The K aggregated sample application scene feature vector sets are centrally integrated to obtain K scene feature vectors, and the K scene feature vectors are used to identify abnormal environmental clues of the K integrated samples to construct a scene environmental clue template library.

[0030] In one possible embodiment, the sample application scenario information set refers to a collection of environmental information related to crack detection collected across multiple different application scenarios. Each sample scenario information typically includes parameters such as temperature, humidity, load, and vibration, which can affect crack evolution. The sample abnormal environmental clue set refers to abnormal environmental signals or features collected within the sample application scenarios, typically manifesting as environmental fluctuations or factors that affect crack propagation under specific circumstances, such as severe temperature fluctuations or load overload.

[0031] Using the preset scenario index as an index, feature extraction is performed on the sample application scenario information set to obtain a sample application scenario feature vector set. Furthermore, the sample application scenario feature vector sets are clustered in the same category, that is, relatively similar sample application scenario feature vectors are aggregated into one set to obtain the K aggregated sample application scenario feature vector sets.

[0032] Preferably, a first sample application scenario feature vector is randomly extracted from the sample application scenario feature vector set, the similarity between the sample application scenario feature vector set and the first sample application scenario feature vector is traversed and calculated, and the sample application scenario feature vectors that meet a preset similarity threshold in the calculation results are added to the first aggregated sample application scenario feature vector set. Then, the first aggregated sample application scenario feature vector set is removed from the sample application scenario feature vector set, and a second sample application scenario feature vector is randomly extracted again. Based on the same principle as for obtaining the first aggregated sample application scenario feature vector set, similar aggregation is performed to obtain a second aggregated sample application scenario feature vector set. This process is repeated in this way to obtain K aggregated sample application scenario feature vector sets.

[0033] Since the sample application scenario feature vectors correspond one-to-one to the sample abnormal environmental clues, the sample abnormal environmental clues set can be mapped and aggregated based on the K aggregated sample application scenario feature vectors to obtain the corresponding K aggregated sample abnormal environmental clues sets. For each of the K aggregated sample abnormal environmental clues sets, only one duplicate abnormal environmental clue factor is retained. In other words, the K aggregated sample abnormal environmental clues sets are unioned to obtain the K integrated sample abnormal environmental clues. The K integrated sample abnormal environmental clues are used to describe the abnormal environmental factors that are likely to interfere with concrete cracks in different application scenarios.

[0034] Then, the K aggregated sample application scene feature vector sets are centrally integrated to obtain a feature vector representing each aggregated set, namely the K scene feature vectors. Then, based on the K scene feature vectors, the K integrated sample abnormal environmental clues are identified, and the identified K integrated sample abnormal environmental clues are each used as a scene environmental clue template, which is then compiled into a scene environmental clue template library.

[0035] By building a scene environment clue template library, we can provide a customized environmental analysis template for each type of application scenario, provide data support for the subsequent determination of environmental interference factors in different application scenarios, and achieve the technical effect of improving the efficiency of environmental interference analysis.

[0036] Furthermore, the K aggregated sample application scenario feature vector sets are centrally integrated to obtain K scenario feature vectors. In this embodiment of the application, step S2 further includes: Extracting a first aggregated sample application scenario feature vector set from the K aggregated sample application scenario feature vector sets; Combining the first aggregated sample application scenario feature vector sets in pairs, calculating the combined similarities respectively, and obtaining a combined similarity set; Extracting the combined similarity set using any one of the first aggregated sample application scenario feature vectors in the first aggregated sample application scenario feature vector set, and performing mean calculation on the extracted results to obtain an integrated similarity set, wherein each integrated similarity corresponds to a first aggregated sample application scenario feature vector; Using the first aggregated sample application scene feature vector corresponding to the maximum value in the integrated similarity set as the first scene feature vector; The combined similarity calculation is performed on the K aggregated sample application scene feature vector sets respectively, and the calculation results are integrated to obtain K scene feature vectors.

[0037] In one embodiment of the present application, a first set of aggregated sample application scenario feature vectors is extracted from the K sets of aggregated sample application scenario feature vectors. Then, the first aggregated sample application scenario feature vectors in the first set of aggregated sample application scenario feature vectors are combined in pairs to obtain a combination set. The similarity of each combination is then calculated using the cosine similarity formula to obtain a combination similarity set. Each combination similarity reflects the degree of similarity between two first aggregated sample application scenario feature vectors within a combination.

[0038] Then, using any first aggregated sample application scenario feature vector in the first aggregated sample application scenario feature vector set as an index, the combined similarity set is extracted to obtain the combined similarity associated with the first aggregated sample application scenario feature vector as an extraction result. The extraction results are then averaged to obtain the integrated similarity corresponding to the first aggregated sample application scenario feature vector. Based on the same acquisition principle, the first aggregated sample application scenario feature vector set is extracted and analyzed to obtain the integrated similarity set.

[0039] Then, the first aggregated sample application scenario feature vector corresponding to the maximum value in the integrated similarity set is used as the first scenario feature vector. In other words, the first scenario feature vector is the first aggregated sample application scenario feature vector with the highest average correlation with the other first aggregated sample application scenario feature vectors. Based on the same principle as for obtaining the first scenario feature vector, combined similarity calculations are performed on each of the K aggregated sample application scenario feature vector sets, and then they are integrated to obtain K scenario feature vectors.

[0040] S3: using a long-short time series neural network to interactively analyze the environmental interference trend characteristics of the scene environmental clue perception data sequence to obtain the environmental interference trend characteristics; Furthermore, the long-short time series neural network is used to interactively analyze the environmental interference trend characteristics of the scene environment clue perception data sequence to obtain the environmental interference trend characteristics. In this embodiment of the application, step S3 further includes: Extracting the long time series trend feature extraction branch of the long-short time series neural network to perform feature extraction on the scene environment clue perception data sequence to obtain a long time series environmental interference trend feature; Extracting the short time series trend feature extraction branch of the long-short time series neural network to perform feature extraction on the scene environment clue perception data sequence to obtain a short time series environmental interference trend feature; The long-term environmental interference trend feature and the short-term environmental interference trend feature are interactively analyzed to obtain the environmental interference trend feature.

[0041] Furthermore, interactive analysis is performed on the long-term environmental interference trend characteristics and the short-term environmental interference trend characteristics to obtain the environmental interference trend characteristics. In this embodiment of the application, step S3 further includes: The cosine similarity formula is used to calculate the feature interaction similarity set of long-term environmental interference trend features and short-term environmental interference trend features; An interaction analysis matrix is constructed based on the feature interaction similarity set, and the short-time series environmental interference trend feature is enhanced using the interaction analysis matrix to obtain the environmental interference trend feature.

[0042] In one embodiment, a long-short time series neural network is a neural network architecture specifically designed for processing and predicting time series data, capable of capturing long-term dependencies. This architecture utilizes a long-short time series trend feature extraction branch to extract features from long-short time series data, while a short-short time series trend feature extraction branch extracts features from short-short time series data (e.g., environmental changes over a short period of time). The environmental interference trend features are used to integrate long-short time series analysis, reliably reflecting the trends of environmental factors that interfere with concrete cracking within the target application scenario.

[0043] Preferably, the long time series trend feature extraction branch and the short time series trend feature extraction branch of the long and short time series neural network are connected in parallel. A plurality of sample scene environment clue perception data sequences, a plurality of sample long time series environmental interference trend features and a plurality of sample short time series environmental interference trend features are obtained as training data sets. A plurality of sample scene environment clue perception data sequences and a plurality of sample long time series environmental interference trend features in the training data set are extracted to perform supervised training on the framework constructed based on the time series neural network until the training converges, and the trained long time series trend feature extraction branch is obtained. A plurality of sample scene environment clue perception data sequences and a plurality of sample short time series environmental interference trend features in the training data set are extracted to perform supervised training on the framework constructed based on the time series neural network, and the network parameters of the framework are updated according to the output results during training, until the training converges, and the trained short time series trend feature extraction branch is obtained.

[0044] The cosine similarity formula is used to calculate the feature interaction similarity set of the long-term environmental interference trend features and the short-term environmental interference trend features. The feature interaction similarity set reflects the degree of feature similarity between the long-term environmental interference trend features and the short-term environmental interference trend features. Furthermore, the feature interaction similarity set is normalized using the softmax formula, and the processed data is filled into the initially empty matrix to obtain an interaction analysis matrix, which is a diagonal matrix.

[0045] The interactive analysis matrix is then used to enhance the short-term environmental interference trend characteristics to obtain the environmental interference trend characteristics. Preferably, the interactive analysis matrix is used as a weight matrix to perform weighted calculations on the short-term environmental interference trend characteristics to obtain the environmental interference trend characteristics. This achieves the goal of comprehensively analyzing the characteristics of long and short time series, improving the reliability of environmental interference trend analysis, and thus improving the reliability of subsequent concrete crack detection results.

[0046] S4: Based on the environmental interference trend characteristics, causal perception fusion is performed on the crack width-depth-length and the crack extension direction to obtain a fused concrete crack detection result.

[0047] Furthermore, step S4 of the embodiment of the present application further includes: Build a causal-aware fuser; The causal perception fusion device is used to analyze the environmental interference trend characteristics, crack width-depth-length and crack extension direction to obtain the fused concrete crack detection result.

[0048] Preferably, the causal-aware fusion unit is a functional unit for performing causal trend fusion on real-time concrete crack detection results based on environmental interference. Training data for the fusion unit is obtained by acquiring multiple sample environmental interference trend features, multiple sample crack widths, depths, and lengths, and multiple sample crack extension directions, along with the corresponding multiple fused concrete crack detection results. A feedforward neural network architecture is obtained, comprising an input layer, a hidden layer, and an output layer. The environmental interference trend features, crack widths, depths, and lengths, and fused crack detection results for each sample are used as network inputs. Forward propagation is performed through the feedforward neural network to calculate predicted crack extension results. The error between the predicted results and the actual labels (e.g., true crack widths and depths) is calculated, using a commonly used loss function such as mean squared error (MSE). A backpropagation algorithm is used to calculate the error, and network parameters (weights and biases) are updated based on the gradient. An optimization algorithm (e.g., gradient descent or its variants, such as the Adam optimizer) is typically used to update model weights until training converges, resulting in the completed causal-aware fusion unit.

[0049] The trained causal perception fusion unit is used to analyze the environmental interference trend characteristics, crack width, depth, length, and crack extension direction to obtain the fused concrete crack detection result. The fused concrete crack detection result is the concrete crack detection result that implicitly incorporates the environmental interference trend. This achieves the technical effect of improving the reliability of the crack detection result.

[0050] In summary, the embodiments of the present application have at least the following technical effects: 1. This application introduces interference analysis of environmental factors through real-time data acquisition and the application of time series neural networks, which can effectively capture the impact of factors such as temperature, humidity, and load on crack development, achieving the technical effect of improving the accuracy of crack detection results.

[0051] 2. This application achieves the technical effect of further improving the accuracy of crack detection results by integrating the geometric characteristics of cracks with environmental interference trends to obtain crack detection results that include implicit environmental interference conditions.

[0052] Example 2, based on the same inventive concept as the method for detecting concrete cracks in combination with a temporal neural network in the above embodiment, as shown in the attached Figure 2 As shown, the present application provides a concrete crack detection system combined with a time series neural network. The system and method embodiments in the present application are based on the same inventive concept. The system includes: The crack detection result acquisition module 11 is used to perform real-time detection of concrete cracks in the target application scenario using the detection component to obtain real-time concrete crack detection results, wherein the concrete crack detection results include crack width-depth-length and crack extension direction; A perception data sequence acquisition module 12 is configured to call a target scene environment clue template based on the target application scenario, extract continuous environment perception data based on the target scene environment clue template, and obtain a scene environment clue perception data sequence; An environmental interference trend feature acquisition module 13 is configured to perform an interactive analysis of environmental interference trend features on the scene environment clue perception data sequence using a long-short time series neural network to acquire environmental interference trend features; The fused concrete crack detection result obtaining module 14 is used to perform causal perception fusion on the crack width-depth-length and the crack extension direction based on the environmental interference trend characteristics to obtain a fused concrete crack detection result.

[0053] Furthermore, the sensing data sequence obtaining module 12 is configured to perform the following steps: Constructing a scene environment clue template library, wherein each scene environment clue template includes a scene feature vector; Extracting features of target application scenario information of the target application scenario according to preset scenario indicators to construct a target application scenario feature vector; Based on the approximate matching between the target application scene feature vector and the scene feature vector of each scene environment clue template in the scene environment clue template library, the scene environment clue template corresponding to the maximum matching similarity is used as the target scene environment clue template.

[0054] Furthermore, the sensing data sequence obtaining module 12 is configured to perform the following steps: Obtaining a sample application scenario information set and a corresponding sample abnormal environment clue set; Extract features from the sample application scenario information set according to preset scenario indicators to obtain a sample application scenario feature vector set, and aggregate the sample application scenario feature vector sets of the same type to obtain K aggregated sample application scenario feature vector sets, where K is a positive integer; Mapping and aggregating the sample abnormal environment clue set according to the K aggregated sample application scenario feature vector sets to obtain K aggregated sample abnormal environment clue sets, and performing a union on the K aggregated sample abnormal environment clue sets to obtain K integrated sample abnormal environment clues; The K aggregated sample application scene feature vector sets are centrally integrated to obtain K scene feature vectors, and the K scene feature vectors are used to identify abnormal environmental clues of the K integrated samples to construct a scene environmental clue template library.

[0055] Furthermore, the sensing data sequence obtaining module 12 is configured to perform the following steps: Extracting a first aggregated sample application scenario feature vector set from the K aggregated sample application scenario feature vector sets; Combining the first aggregated sample application scenario feature vector sets in pairs, calculating the combined similarities respectively, and obtaining a combined similarity set; Extracting the combined similarity set using any one of the first aggregated sample application scenario feature vectors in the first aggregated sample application scenario feature vector set, and performing mean calculation on the extracted results to obtain an integrated similarity set, wherein each integrated similarity corresponds to a first aggregated sample application scenario feature vector; Using the first aggregated sample application scene feature vector corresponding to the maximum value in the integrated similarity set as the first scene feature vector; The combined similarity calculation is performed on the K aggregated sample application scene feature vector sets respectively, and the calculation results are integrated to obtain K scene feature vectors.

[0056] Furthermore, the environmental interference trend feature acquisition module 13 is configured to perform the following steps: Extracting the long time series trend feature extraction branch of the long-short time series neural network to perform feature extraction on the scene environment clue perception data sequence to obtain a long time series environmental interference trend feature; Extracting the short time series trend feature extraction branch of the long-short time series neural network to perform feature extraction on the scene environment clue perception data sequence to obtain a short time series environmental interference trend feature; The long-term environmental interference trend feature and the short-term environmental interference trend feature are interactively analyzed to obtain the environmental interference trend feature.

[0057] Furthermore, the environmental interference trend feature acquisition module 13 is configured to perform the following steps: The cosine similarity formula is used to calculate the feature interaction similarity set of long-term environmental interference trend features and short-term environmental interference trend features; An interaction analysis matrix is constructed based on the feature interaction similarity set, and the short-time series environmental interference trend feature is enhanced using the interaction analysis matrix to obtain the environmental interference trend feature.

[0058] Furthermore, the fusion concrete crack detection result obtaining module 14 is used to perform the following steps: Build a causal-aware fuser; The causal perception fusion device is used to analyze the environmental interference trend characteristics, crack width-depth-length and crack extension direction to obtain the fused concrete crack detection result.

[0059] It should be noted that the order in which the embodiments of the present application are presented is for illustrative purposes only and does not necessarily represent the superiority or inferiority of the embodiments. Furthermore, the foregoing descriptions of specific embodiments of this specification are provided. Furthermore, the processes depicted in the accompanying drawings do not necessarily require the specific order or sequential sequence shown to achieve the desired results. In certain embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0060] The above description is only a preferred embodiment of the present application and is not intended to limit the present application. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present application should be included in the scope of protection of the present application.

[0061] This specification and drawings are merely illustrative of the present application and are intended to cover any and all modifications, variations, combinations, or equivalents within the scope of this application. Obviously, those skilled in the art may make various modifications and variations to this application without departing from the scope of this application. Thus, this application is intended to include such modifications and variations as fall within the scope of this application and its equivalents.

Claims

1. A concrete crack detection method combined with a time series neural network, characterized in that: The method comprises: Use the detection component to perform real-time detection of concrete cracks in the target application scenario and obtain real-time concrete crack detection results, where the concrete crack detection results include crack width, depth, length and crack extension direction; Calling a target scene environment clue template based on the target application scenario, extracting continuous environment perception data based on the target scene environment clue template, and obtaining a scene environment clue perception data sequence; Using a long-short time series neural network to interactively analyze the environmental interference trend characteristics of the scene environmental clue perception data sequence to obtain the environmental interference trend characteristics; Based on the environmental interference trend characteristics, causal perception fusion of crack width-depth-length and crack extension direction is performed to obtain a fused concrete crack detection result.

2. The concrete crack detection method combined with a temporal neural network according to claim 1, characterized in that: Calling a scene environment clue template based on the target application scenario, extracting continuous environment perception data based on the target scene environment clue template, and obtaining a scene environment clue perception data sequence, including: Constructing a scene environment clue template library, wherein each scene environment clue template includes a scene feature vector; Extracting features of target application scenario information of the target application scenario according to preset scenario indicators to construct a target application scenario feature vector; Based on the approximate matching between the target application scene feature vector and the scene feature vector of each scene environment clue template in the scene environment clue template library, the scene environment clue template corresponding to the maximum matching similarity is used as the target scene environment clue template.

3. The concrete crack detection method combined with a temporal neural network according to claim 2, characterized in that: Construct a scene environment clue template library, wherein each scene environment clue template includes a scene feature vector, including: Obtaining a sample application scenario information set and a corresponding sample abnormal environment clue set; Extract features from the sample application scenario information set according to preset scenario indicators to obtain a sample application scenario feature vector set, and aggregate the sample application scenario feature vector sets of the same type to obtain K aggregated sample application scenario feature vector sets, where K is a positive integer; Mapping and aggregating the sample abnormal environment clue set according to the K aggregated sample application scenario feature vector sets to obtain K aggregated sample abnormal environment clue sets, and performing a union on the K aggregated sample abnormal environment clue sets to obtain K integrated sample abnormal environment clues; The K aggregated sample application scene feature vector sets are centrally integrated to obtain K scene feature vectors, and the K scene feature vectors are used to identify abnormal environmental clues of the K integrated samples to construct a scene environmental clue template library.

4. The method for detecting concrete cracks in combination with a temporal neural network according to claim 3, wherein: The K aggregated sample application scene feature vector sets are centrally integrated to obtain K scene feature vectors, including: Extracting a first aggregated sample application scenario feature vector set from the K aggregated sample application scenario feature vector sets; Combining the first aggregated sample application scenario feature vector sets in pairs, calculating the combined similarities respectively, and obtaining a combined similarity set; Extracting the combined similarity set using any one of the first aggregated sample application scenario feature vectors in the first aggregated sample application scenario feature vector set, and performing mean calculation on the extracted results to obtain an integrated similarity set, wherein each integrated similarity corresponds to a first aggregated sample application scenario feature vector; Using the first aggregated sample application scene feature vector corresponding to the maximum value in the integrated similarity set as the first scene feature vector; The combined similarity calculation is performed on the K aggregated sample application scene feature vector sets respectively, and the calculation results are integrated to obtain K scene feature vectors.

5. The method for detecting concrete cracks in combination with a temporal neural network according to claim 1, wherein: Using a long-short time series neural network to interactively analyze the environmental interference trend characteristics of the scene environmental clue perception data sequence, the environmental interference trend characteristics are obtained, including: Extracting the long time series trend feature extraction branch of the long-short time series neural network to perform feature extraction on the scene environment clue perception data sequence to obtain a long time series environmental interference trend feature; Extracting the short time series trend feature extraction branch of the long-short time series neural network to perform feature extraction on the scene environment clue perception data sequence to obtain a short time series environmental interference trend feature; The long-term environmental interference trend feature and the short-term environmental interference trend feature are interactively analyzed to obtain the environmental interference trend feature.

6. The method for detecting concrete cracks in combination with a temporal neural network according to claim 1, wherein: Interactively analyzing the long-term environmental interference trend characteristics and the short-term environmental interference trend characteristics to obtain the environmental interference trend characteristics includes: The cosine similarity formula is used to calculate the feature interaction similarity set of long-term environmental interference trend features and short-term environmental interference trend features; An interaction analysis matrix is constructed based on the feature interaction similarity set, and the short-time series environmental interference trend feature is enhanced using the interaction analysis matrix to obtain the environmental interference trend feature.

7. The method for detecting concrete cracks in combination with a temporal neural network according to claim 1, wherein: include: Build a causal-aware fuser; The causal perception fusion device is used to analyze the environmental interference trend characteristics, crack width-depth-length and crack extension direction to obtain the fused concrete crack detection result.

8. A concrete crack detection system combined with a time series neural network, characterized in that: The system is used to implement a concrete crack detection method combined with a temporal neural network according to any one of claims 1 to 7, and the system comprises A crack detection result acquisition module is used to use the detection component to perform real-time detection of concrete cracks in the target application scenario and obtain real-time concrete crack detection results, wherein the concrete crack detection results include crack width-depth-length and crack extension direction; A perception data sequence acquisition module is used to call a target scene environment clue template based on the target application scenario, extract environmental continuous perception data based on the target scene environment clue template, and obtain a scene environment clue perception data sequence; An environmental interference trend feature acquisition module is used to interactively analyze the environmental interference trend features of the scene environmental clue perception data sequence using a long-short time series neural network to obtain environmental interference trend features; The fused concrete crack detection result acquisition module is used to perform causal perception fusion on the crack width-depth-length and crack extension direction based on the environmental interference trend characteristics to obtain the fused concrete crack detection result.

Images