Concrete building top surface crack detection method based on dual-light fusion

Abstract

The invention discloses a concrete building top surface crack detection method based on dual-light fusion, and the method comprises the steps: collecting dual-light fusion irradiation surface crack data of the top of a concrete building; the dual-light fusion irradiation surface crack data comprises different building structure deformation types of different materials at the top of the concrete building and crack distribution characteristic information of different materials at the top of the concrete building within a preset time before deformation of the building structure; according to the statistical condition of the crack distribution characteristic information, clustering the crack distribution characteristic information of different materials at the top of the concrete building to obtain the types of the crack distribution characteristic information at different environment temperatures at the top of the building and the detection conditions of the types of the crack distribution characteristic information; the method ensures that the concrete building top crack distribution characteristic information is accurately detected according to the building height.

Description

Technical Field

[0001] The present invention relates to the field of concrete building crack detection, and in particular to a concrete building top surface crack detection method based on dual-light fusion. Background Art

[0002] During the long-term use of concrete buildings, cracks on the top surface of the building are a common disease. The appearance of cracks will affect the safety and durability of the structure, so it is of great significance to detect them efficiently and accurately. At present, crack detection methods mainly include manual detection and traditional optical imaging detection. Manual detection relies on the experience of the inspectors, and there are problems of strong subjectivity and low efficiency. Especially for the detection of large-area buildings, manual detection is difficult to cover all areas. Although the traditional optical imaging detection method can provide crack image information, the image of a single light source is easily disturbed under different lighting conditions, resulting in blurred crack edges and insufficient contrast, affecting the accuracy and reliability of detection.

[0003] In addition, these methods lack specificity when dealing with crack characteristics of different materials and are difficult to adapt to the complex situations of different building surfaces. Summary of the invention

[0004] In order to solve the above technical problems, the present invention provides a method that utilizes multiple crack distribution feature information and graph neural network technology to realize crack detection and building structure deformation monitoring of the top state of concrete buildings at different ambient temperatures.

[0005] In an initial aspect, the present invention provides a method for detecting cracks on the top surface of a concrete building based on dual-light fusion, the method comprising: Collecting dual-light fusion irradiation surface crack data of the top of the concrete building, wherein the dual-light fusion irradiation surface crack data includes deformation types of different building structures of different materials on the top of the concrete building and distribution characteristic information of cracks of different materials on the top of the concrete building within a preset time before the building structure deformation occurs; The crack distribution characteristic information of different materials on the top of the concrete building includes the crack type, crack quantity, crack distribution density, crack length and width distribution, crack spacing, crack depth, crack direction and angle, and crack morphology of the top of the concrete building after acceptance; The crack type, crack number and crack distribution density are collected according to the building height, and the crack length and width distribution, crack spacing, crack depth, crack direction and angle and crack morphology are collected according to the preset building height classification; According to the statistics of crack distribution characteristic information, the crack distribution characteristic information of different materials on the top of the concrete building is clustered to obtain the types of crack distribution characteristic information at different ambient temperatures on the top of the building and the crack distribution characteristic information type detection conditions, wherein the types of crack distribution characteristic information at different ambient temperatures on the top of the building include crack type, crack number and crack distribution density, and the crack distribution characteristic information type detection conditions include crack length and width distribution, crack spacing, crack depth, crack direction and angle, and crack morphology; The types of crack distribution characteristic information at different ambient temperatures on the top of the building, the detection conditions for the types of crack distribution characteristic information, and the corresponding deformation types of the building structure with different materials on the top of the concrete building are uploaded to the graph neural network model to establish a physical crack distribution characteristic fluctuation detection algorithm and a chemical crack distribution characteristic fluctuation detection algorithm; The input of the physical crack distribution characteristic fluctuation detection algorithm is the statistical crack distribution characteristic information types of different ambient temperatures on the top of the building, and the output is the detection geometric characteristic parameters of the chemical crack distribution characteristic fluctuation detection algorithm; the input of the chemical crack distribution characteristic fluctuation detection algorithm is the crack distribution characteristic information type detection conditions and the determined detection geometric characteristic parameters, and the output is the building structure deformation type on the top of the concrete building; In practical applications, the types of crack distribution characteristic information at different ambient temperatures on the top of the concrete building are collected based on the building height, and input into the physical crack distribution characteristic fluctuation detection algorithm to obtain the detection geometric characteristic parameters of the chemical crack distribution characteristic fluctuation detection algorithm; Compare the detection geometric feature parameters of the chemical crack distribution characteristic fluctuation detection algorithm with the preset geometric feature size: When the size of the detected geometric feature parameter is smaller than the preset geometric feature size, there is no need to set the crack distribution feature information type detection condition, and the collection and analysis of the types of crack distribution feature information at different ambient temperatures on the top of the building shall continue; When the size of the detected geometric feature parameter is equal to or greater than the preset geometric feature size, the crack distribution feature information type detection condition is set, and the crack distribution feature information type detection condition and the detected geometric feature parameter are input into the chemical crack distribution feature fluctuation detection algorithm to obtain the building structure deformation type on the top of the concrete building.

[0006] Furthermore, the steps for establishing the physical crack distribution characteristic fluctuation detection algorithm are as follows: According to the surface crack data illuminated by dual-light fusion, the types of crack distribution characteristics of different materials on the top of the concrete building and the corresponding building structure deformation type data are sorted out; The crack distribution characteristics of different ambient temperatures on the top of the building are processed and features are proposed, and the raw data are converted into practical features for use in the graph neural network model; Using the graph neural network model, calculate the data according to the prepared ReLU function, use the selected algorithm to calculate the model ReLU function, and adjust the model parameters; After the ReLU function calculation is completed, the relevant parameters of the model are extracted as the output of the physical crack distribution characteristic fluctuation detection algorithm.

[0007] Furthermore, the steps for establishing the chemical crack distribution characteristic fluctuation detection algorithm are as follows: According to the surface crack data illuminated by dual-light fusion, the crack distribution characteristic information type detection conditions and corresponding building structure deformation type data on the top of the concrete building are sorted out; The graph neural network model is used for training, and the crack distribution feature information type detection conditions are used as the edges of the graph neural network, and the building structure deformation type parameters are used as the nodes of the graph neural network. Relationships are connected, and the ReLU function is calculated for the model. After the ReLU function is calculated, the chemical crack distribution characteristic fluctuation detection algorithm is used to predict the building structure deformation type parameters at the top of the concrete building; The graph neural network is used to perform dual-light fusion irradiation surface crack data, and the correlation coefficient between the building structure deformation type parameters and the actual building structure deformation type is calculated.

[0008] Furthermore, the data collection method of the crack distribution characteristic information type detection condition includes: Use the temperature monitoring device to monitor and collect the distribution of crack length and width on the top of the concrete building; monitor and collect the crack spacing on the top of the concrete building based on the building bearing capacity monitoring platform; monitor and collect the crack morphology on the top of the concrete building, and analyze the peak and fluctuation range of temperature changes; use the multimodal crack supervision platform to monitor and collect the crack depth signal generated on the top of the concrete building; use the crack image comparison interface to collect the crack direction and angle around the top of the concrete building, and count the impact of different crack directions and angles.

[0009] Furthermore, the preset influencing factors of the preset geometric feature size include the influence of crack size on detection sensitivity, the influence of crack morphology, the influence of environmental factors, the influence of material properties, the limitation of detection equipment resolution, the computational efficiency of the detection algorithm, the influence of crack development stage, noise interference, and the influence of detection area size.

[0010] Beneficial effects: The present invention proposes a method for detecting cracks on the top of a concrete building based on dual-light fusion. By collecting these data, the method can comprehensively understand the cracking situation on the top of the concrete building. The method uses a graph neural network model to establish a physical crack distribution feature fluctuation detection algorithm and a chemical crack distribution feature fluctuation detection algorithm. The dual-light fusion irradiation surface crack data and the corresponding building structure deformation type are illuminated by ReLU function calculation and model training. The model can identify and cluster different types of building structure deformation on the top of the concrete building. According to the comparison between the detection geometric feature parameters and the preset geometric feature size, it can be determined whether it is necessary to set the crack distribution feature information type detection condition, and the building structure deformation type on the top of the concrete building can be discovered in time. By integrating a variety of crack distribution feature information and the application of the graph neural network model, the adjustment accuracy of the building structure deformation on the top of the concrete building can be improved. Not only can the building structure deformation type be judged, but also the building structure deformation prediction and trend analysis can be performed according to different material data. The method uses a variety of crack distribution feature information and graph neural network technology to realize crack detection and building structure deformation detection of the top state of concrete buildings at different ambient temperatures, and ensures that the crack distribution feature information on the top of the concrete building is accurately detected according to the building height. BRIEF DESCRIPTION OF THE DRAWINGS

[0011] Figure 1 It is a flow chart of the method of the present invention. DETAILED DESCRIPTION

[0012] It should be noted that, in the absence of conflict, the embodiments in the present application and the features in the embodiments may be combined with each other. The present application will be further described in detail below in conjunction with the accompanying drawings and specific embodiments.

[0013] like Figure 1 As shown, a method for detecting cracks on the top surface of a concrete building based on dual-light fusion of the present invention specifically comprises the following steps: Step A1, collecting dual-light fusion irradiation surface crack data of the concrete building top, wherein the dual-light fusion irradiation surface crack data includes deformation types of different concrete building tops and different materials of building structure and distribution characteristic information of cracks of different materials on the concrete building top within a preset time before the building structure deformation occurs; In order to realize the state crack detection of the top of concrete buildings at different ambient temperatures, it is first necessary to collect the surface crack data of the top of concrete buildings illuminated by dual-light fusion. These data can provide information about the performance of the top of concrete buildings after acceptance, the deformation type of building structure and the distribution characteristics of related cracks, providing a basis for subsequent state evaluation and building structure deformation monitoring. The crack distribution characteristic information of different materials on the top of the concrete building includes: Crack type, number of cracks and crack distribution density: Collect the crack type, number of cracks and crack distribution density on the top of the concrete building, and analyze the fluctuations in the crack type, number of cracks and crack distribution density; the fluctuations in the crack type, number of cracks and crack distribution density can reflect the stability of the top of the concrete building. It is necessary to collect the crack type, number of cracks and crack distribution density parameters on the top of the concrete building according to the building height, and record their fluctuations.

[0014] Crack length and width distribution: Based on the temperature monitoring device, the crack length and width distribution on the top of the concrete building are monitored and collected, including the maximum value, average value and variance of the crack length and width distribution. The changes in the crack length and width distribution can reflect the cracking condition on the top of the concrete building and the deformation of the building structure.

[0015] Crack spacing: Use the building bearing capacity monitoring platform to monitor and collect changes in crack spacing on the top of concrete buildings, including crack spacing inside and outside the equipment. Abnormal changes in crack spacing can indicate structural deformations such as overload and poor heat dissipation on the top of the concrete building. Crack depth: Based on the multimodal crack monitoring platform, the crack depth signal generated on the top of the concrete building is monitored and collected, and analyzed. The crack depth signal can provide status information inside the top of the concrete building.

[0016] Crack direction and angle: Based on the crack image comparison interface, the crack direction and angle around the top of the concrete building are collected, and the changes in the impact of different crack directions and angles are counted, and the abnormal changes in the impact of different crack directions and angles are counted.

[0017] Crack morphology: Monitor and collect the crack morphology on the top of the concrete building, analyze the peak value and fluctuation range of temperature change, and the test results of crack morphology can provide information about insulation performance.

[0018] By collecting and recording the crack distribution characteristics of these different materials, an information database of the top of the concrete building is established for subsequent data analysis, correlation analysis and model ReLU function calculation; wherein, the crack type, number of cracks and crack distribution density are collected according to the building height, and the crack length and width distribution, crack spacing, crack depth, crack direction and angle and crack morphology are collected according to the preset building height classification; by collecting the dual-light fusion irradiation surface crack data of the top of the concrete building, it is possible to provide basic data on the performance of the top of the concrete building after acceptance and the type of building structure deformation, which provides an important basis for subsequent status evaluation, building structure deformation monitoring and performance optimization.

[0019] By analyzing the surface crack data of dual-light fusion illumination, the stability, cracking, insulation performance and potential building structure deformation of the top of the concrete building can be revealed; in addition to the fluctuation data of crack type, number of cracks and crack distribution density, through temperature monitoring devices, building bearing capacity monitoring platforms, crack depth detection, crack image comparison interface and crack morphology detection and other detection methods, the top of the concrete building is fully inspected for cracks, and the status information of the top of the concrete building is collected from different angles to improve the accuracy of building structure deformation detection and adjustment.

[0020] By collecting the surface crack data and crack distribution characteristic information of the top of the concrete building under dual-light fusion illumination, we can fully understand the performance of the top of the concrete building, the deformation type of the building structure and the crack distribution characteristic information, and provide basic data and basis for subsequent status assessment, building structure deformation monitoring and performance optimization.

[0021] Step A2: clustering the crack distribution characteristic information of different materials on the top of the concrete building according to the crack distribution characteristic information statistics to obtain the types of crack distribution characteristic information at different ambient temperatures on the top of the building and crack distribution characteristic information type detection conditions, wherein the types of crack distribution characteristic information at different ambient temperatures on the top of the building include crack type, crack number and crack distribution density, and the crack distribution characteristic information type detection conditions include crack length and width distribution, crack spacing, crack depth, crack direction and angle, and crack morphology; In this step, it is necessary to cluster the crack distribution characteristic information of different materials on the top of the concrete building according to the statistics of the crack distribution characteristic information, and divide the crack distribution characteristic information of different materials into two different information, namely, the types of crack distribution characteristic information of different ambient temperatures on the top of the building and the detection conditions of the crack distribution characteristic information types; The types of information on crack distribution characteristics at different ambient temperatures on the top of the building include crack type, crack quantity and crack distribution density. These data can be collected by monitoring and collecting the crack type, crack quantity and crack distribution density parameters on the top of the concrete building according to the building height; The crack distribution characteristic information type detection conditions include crack length and width distribution, crack spacing, crack depth, crack direction and angle, and crack morphology, which are used to monitor and collect the crack length and width distribution, crack spacing, crack depth, crack direction and angle, and crack morphology of the top of the concrete building after acceptance; The clustering basis for statistically analyzing the types of crack distribution characteristic information at different ambient temperatures on the top of the building and the detection conditions of crack distribution characteristic information types is as follows: Different types of building structure deformation correspond to different crack distribution characteristic information. By clustering the crack type, crack number and crack distribution density separately from other multiple characteristic data, different types of building structure deformation can be judged and analyzed more targetedly. The fluctuations of crack type, crack number and crack distribution density are more likely to be related to the building structure deformation under different ambient temperatures. These data can be used to quickly make a preliminary judgment on the building structure deformation. The crack distribution characteristic information type detection conditions involve more sensor collection and detection equipment, and the collection requires a certain amount of time and cost. Therefore, by presetting a preset geometric feature size, if the detection geometric feature parameters of the crack distribution characteristic information types at different ambient temperatures on the top of the building are smaller than the preset geometric feature size, it is possible to avoid frequently setting the crack distribution characteristic information type detection conditions, thereby improving the detection effect.

[0022] By dividing the crack distribution characteristic information into two pieces of information, different types of building structure deformation can be judged and analyzed more specifically; by presetting a preset geometric feature size, if the detection geometric feature parameters of the types of crack distribution characteristic information at different ambient temperatures on the top of the building are smaller than the preset geometric feature size, it is possible to avoid frequently setting the crack distribution characteristic information type detection conditions, thereby improving the detection effect; by clustering different types of crack distribution characteristic information separately, more accurate judgment and adjustment results can be obtained by comprehensively analyzing the correlation between different feature data.

[0023] Step A3, uploading the statistical crack distribution characteristic information types at different ambient temperatures on the top of the building, the crack distribution characteristic information type detection conditions and the corresponding deformation types of the concrete building top with different materials to the graph neural network model, and establishing a physical crack distribution characteristic fluctuation detection algorithm and a chemical crack distribution characteristic fluctuation detection algorithm; The steps to establish the physical crack distribution characteristic fluctuation detection algorithm are as follows: According to the surface crack data of the top of the concrete building illuminated by dual-light fusion in step A1, the types of crack distribution characteristics of different materials on the top of the concrete building and the corresponding building structure deformation type data are sorted out; The crack distribution characteristics of different ambient temperatures on the top of the building are processed and features are proposed, and the raw data are converted into practical features for use in the graph neural network model; According to the problem description, select a graph neural network model suitable for the scenario, use the selected algorithm to calculate the model ReLU function, and adjust the model parameters; After the ReLU function calculation is completed, the relevant parameters of the model are extracted as the output of the physical crack distribution characteristic fluctuation detection algorithm.

[0024] The steps to establish the chemical crack distribution characteristic fluctuation detection algorithm are as follows: According to the surface crack data of the top of the concrete building illuminated by dual-light fusion collected in step A1, the crack distribution characteristic information type detection conditions and the corresponding building structure deformation type data of the top of the concrete building are sorted out; The graph neural network model is used for training, and the crack distribution feature information type detection conditions are used as the edges of the graph neural network, and the building structure deformation type parameters are used as the nodes of the graph neural network. Relationships are connected, and the ReLU function is calculated for the model. After the ReLU function is calculated, the chemical crack distribution characteristic fluctuation detection algorithm is used to predict the structural deformation type of the top of the concrete building. The chemical crack distribution characteristic fluctuation detection algorithm is trained using a graph neural network model, and the expression is: Input feature matrix: ,in Indicates the number of nodes (such as crack feature points), Represents the characteristic dimension of each node (including crack distribution characteristics, detection conditions and building structure deformation type).

[0025] Adjacency Matrix: , which represents the connection relationship between nodes in the graph.

[0026] Degree Matrix: , defined as , Normalized adjacency matrix: , used to balance the influence of nodes.

[0027] The graph convolutional layer can be defined as: , in: Indicates The node feature matrix of the layer, . and Respectively The weight matrix and bias vector of the layer.

[0028] Assume that Layer graph convolution, the calculation of each layer is: , Through multi-layer graph convolution, neighborhood information can be gradually aggregated to extract deeper crack distribution features.

[0029] After the last layer of graph convolution, a fully connected layer is used to map the features and transform them into the target output dimension. Assume that the output is : , Among them, Softmax is used for multi-classification problems and can convert the output into a probability distribution.

[0030] Use the cross-extraction loss function for supervised learning: , in: Indicates The node in The true label of the class (0 or 1). Indicates the number of categories.

[0031] Combining the above parts, the overall model can be expressed as: , The model extracts features through multi-layer graph convolution, and then obtains the prediction results through a fully connected layer and Softmax activation.

[0032] The input of the physical crack distribution characteristic fluctuation detection algorithm is the statistical crack distribution characteristic information types of different ambient temperatures on the top of the building, and the output is the detection geometric characteristic parameters of the chemical crack distribution characteristic fluctuation detection algorithm; the input of the chemical crack distribution characteristic fluctuation detection algorithm is the crack distribution characteristic information type detection conditions and the determined detection geometric characteristic parameters, and the output is the building structure deformation type on the top of the concrete building; Among them, the physical cracks include: Temperature stress: concrete will expand or contract when the temperature changes, resulting in temperature stress, which in turn causes cracks. Shrinkage cracks: During the curing process, water evaporation or chemical reactions in concrete cause volume shrinkage, which may cause cracks. Load stress: When the static or dynamic loads (such as wind loads, snow loads, vehicle loads, etc.) borne by a building exceed its bearing capacity, cracks may occur. Structural deformation: Structural deformation caused by foundation settlement, earthquakes, etc. may cause cracks on the top of the concrete.

[0033] The chemical cracks include: Alkali-aggregate reaction: some aggregates react chemically with the alkali in cement, causing the concrete volume to expand and form cracks. Corrosive environment: in the environment of corrosive substances such as acids and salts, the components in the concrete may be affected, resulting in structural degradation and crack formation. Moisture intrusion: moisture penetrates into the concrete, which may cause steel bars to rust and cause cracks. Chemical erosion: concrete reacts with certain chemicals (such as chloride ions, sulfuric acid, etc.), which may cause the material strength to decrease and form cracks.

[0034] This step can obtain complete input information by sorting and processing the types of crack distribution characteristic information and building structure deformation type data of different materials on the top of the concrete building, so that the crack distribution characteristic fluctuation detection algorithm can more accurately judge the status of the equipment and the building structure deformation type; use the prepared ReLU function calculation data to perform model ReLU function calculation on the selected graph neural network model, and adjust the model parameters, which can improve the accuracy and performance of the model by continuously optimizing the model parameters and algorithm selection; by outputting the detection geometric feature parameters, it is possible to understand the importance of each feature to the judgment of the building structure deformation type, so as to better understand the relationship between the crack distribution feature fluctuation and the building structure deformation type; In summary, the crack distribution characteristic fluctuation detection algorithm can more accurately evaluate the state of the top of the concrete building and the type of building structure deformation, and provide effective decision support and guidance for eliminating building structure deformation for braking personnel.

[0035] Step A4, collecting the types of crack distribution characteristic information of the top of the concrete building at different ambient temperatures according to the building height, and inputting them into the physical crack distribution characteristic fluctuation detection algorithm to obtain the detection geometric characteristic parameters of the chemical crack distribution characteristic fluctuation detection algorithm; In this step, data is preprocessed before input, including data desensitization, sorting, and normalization, to ensure the quality and consistency of data input into the model; The types of crack distribution characteristics of the top of the building under different ambient temperatures collected according to the building height are used as input and passed to the physical crack distribution characteristic fluctuation detection algorithm for feature extraction and state evaluation, providing detection geometric characteristic parameters for the subsequent chemical crack distribution characteristic fluctuation detection algorithm; the detection geometric characteristic parameters reflect the current state and performance status of the top of the concrete building. By detecting the geometric characteristic parameters, a preliminary judgment and evaluation of the cracking situation on the top of the concrete building can be made; In this step, through preprocessing steps such as data desensitization, sorting and normalization, the quality and consistency of the data input into the model can be ensured, outliers, noise and inconsistent sampling rates in the data can be eliminated, and the accuracy and reliability of the monitoring algorithm can be improved; The detection geometric characteristic parameters obtained by the physical crack distribution characteristic fluctuation detection algorithm can reflect the current state and performance status of the concrete building top. These detection geometric characteristic parameters can be used for the preliminary judgment and evaluation of the concrete building top and provide important input for the subsequent chemical crack distribution characteristic fluctuation detection algorithm. By continuously detecting cracks and evaluating the status of the concrete building top, potential building structure deformation or abnormal conditions can be discovered in a timely manner, so that maintenance measures can be taken in advance to avoid the impact of equipment building structure deformation on the supply of different ambient temperatures. By analyzing and evaluating the status of the concrete building top based on the building height, the service life of the concrete building top can be increased.

[0036] Step A5: Compare the detection geometric feature parameters of the chemical crack distribution characteristic fluctuation detection algorithm with the preset geometric feature size: When the size of the detected geometric feature parameter is smaller than the preset geometric feature size, there is no need to set the crack distribution feature information type detection condition, and the collection and analysis of the types of crack distribution feature information at different ambient temperatures on the top of the building shall continue; Step A6: When the size of the detected geometric feature parameter is equal to or greater than the preset geometric feature size, the crack distribution feature information type detection condition is set, and the crack distribution feature information type detection condition and the detected geometric feature parameter are input into the chemical crack distribution feature fluctuation detection algorithm to obtain the building structure deformation type of the top of the concrete building.

[0037] The preset geometric feature size refers to a fixed value or range used to determine whether it is necessary to set the crack distribution feature information type detection condition. The preset geometric feature size is preset based on the following factors: Impact of crack size on detection sensitivity: The preset geometric feature size (such as crack length, width, and area) affects the sensitivity of the algorithm. Larger preset values ​​may ignore small cracks, while smaller preset values ​​may cause noise and non-crack features to be misdetected as cracks.

[0038] Impact of crack morphology: Cracks may have different geometric shapes (such as straight, curved or mesh-like). If the preset geometric features cannot match the actual crack morphology, it may lead to partial or complete failure of crack detection, especially for cracks with complex morphology.

[0039] Environmental factors: Environmental conditions (such as temperature and humidity) can cause concrete to expand or contract, thus affecting the size of the crack. If these factors are not considered in the geometric feature preset, the actual size of the crack may be deviated during detection, especially in an environment with large temperature fluctuations.

[0040] Influence of material properties: Different materials (such as reinforced concrete and ordinary concrete) have different characteristics when cracks propagate. If the preset geometric features do not take material properties into account, they may underestimate or overestimate the severity of the cracks, affecting the accuracy of detection.

[0041] Limitation of detection equipment resolution: The resolution of the equipment will affect the detectable size of the crack. If the preset crack geometry is too small and the equipment cannot distinguish it, it will lead to false detection or missed detection. Therefore, the preset geometry needs to match the equipment resolution.

[0042] Computational efficiency of the detection algorithm: Smaller geometric features increase the computational burden of the detection algorithm because the algorithm needs to process more tiny crack features. Conversely, larger preset geometric features may improve efficiency but affect detection accuracy, especially when there are more tiny cracks.

[0043] Impact of crack development stage: The geometric characteristics of cracks will change over time. If the preset values ​​fail to take into account the growth law of cracks, it may lead to inaccurate detection. For example, a small crack in the early stage may gradually expand into a large crack, so the preset geometric characteristics need to be adjusted according to the crack development stage.

[0044] Noise interference: If the preset crack geometry is too small, the detection algorithm may mistake surface stains, dust and other noise as cracks. Therefore, it is necessary to set the geometry size reasonably to reduce noise interference while ensuring accurate crack detection.

[0045] Impact of the size of the inspection area: In a larger inspection area, cracks may be unevenly distributed. If the preset geometric feature is too large, cracks in the edge area may be missed; if it is too small, the impact of noise will increase.

[0046] Furthermore, the presetting of the preset geometric feature size is a dynamic process, which needs to be adjusted and optimized according to the actual situation. By continuously observing and analyzing the system cracking situation, building structure deformation data and maintenance records, it is necessary to continuously optimize the preset geometric feature size to improve the accuracy and reliability of state crack detection.

[0047] When the size of the detected geometric feature parameter is smaller than the preset geometric feature size, the physical state evaluation result representing the top of the concrete building indicates that it is in a normal state. In this case, there is no need to set the crack distribution feature information type detection condition, and the collection and analysis of the initial force crack distribution feature information type can be continued to help monitor and collect the basic cracking conditions of the top of the concrete building; If the detected geometric feature parameter is equal to or greater than the preset geometric feature size, it indicates that the physical state assessment result of the top of the concrete building indicates the existence of potential building structure deformation or abnormality. In this case, it is necessary to set the crack distribution feature information type detection condition and input it into the chemical crack distribution feature fluctuation detection algorithm together with the physical detection geometric feature parameter to obtain the building structure deformation type of the top of the concrete building; The chemical crack distribution characteristic fluctuation detection algorithm will comprehensively consider the types of crack distribution characteristic information at different ambient temperatures on the top of the building, the detection conditions of the crack distribution characteristic information types, and the preset detection geometric characteristic parameters, and use algorithms such as graph neural networks to calculate and analyze the ReLU function, so as to determine the structural deformation type of the top of the concrete building; By comparing the detected geometric feature parameters with the preset geometric feature size, it is possible to quickly determine whether the top of the concrete building is in a normal state, thereby deciding whether it is necessary to set the crack distribution feature information type detection conditions. If the detected geometric feature parameters are smaller than the preset geometric feature size, additional data collection and analysis can be avoided, saving time and resources. When the detected geometric feature parameter is equal to or greater than the preset geometric feature size, the crack distribution feature information type detection condition is set, and it is input into the chemical crack distribution feature fluctuation detection algorithm together with the detected geometric feature parameter, which can provide accurate judgment of the building structure deformation type. By comprehensively considering multiple data information and preset parameters, and using algorithms such as graph neural networks to calculate and analyze the ReLU function, the building structure deformation type on the top of the concrete building can be accurately determined; The determination of the preset geometric feature size is based on multiple factors. By comprehensively considering these factors, a preset geometric feature size suitable for the top of a specific concrete building is formulated to improve the accuracy and reliability of state crack detection. By observing and analyzing the cracking conditions of the system, building structure deformation data and maintenance records, the preset geometric feature size can be continuously optimized to improve the accuracy and reliability of the monitoring algorithm.

[0048] The present invention provides a method for detecting cracks on the top surface of a concrete building based on dual-light fusion, which is implemented through different control modules, including: A dual-light fusion irradiation surface crack data acquisition module is used to collect dual-light fusion irradiation surface crack data of the top of a concrete building, wherein the dual-light fusion irradiation surface crack data includes deformation types of different building structures of different materials on the top of different concrete buildings and distribution characteristic information of cracks of different materials on the top of concrete buildings within a preset time before the deformation of the building structure occurs; The crack distribution characteristic information of different materials on the top of the concrete building includes the crack type, crack number, crack distribution density, crack length and width distribution, crack spacing, crack depth, crack direction and angle, and crack morphology of the top of the concrete building after acceptance; the crack type, crack number, and crack distribution density are collected according to the building height, and the crack length and width distribution, crack spacing, crack depth, crack direction and angle, and crack morphology are collected according to the preset building height classification; A crack distribution characteristic information statistics module is used to receive deformation types of building structures of different materials and crack distribution characteristic information of different materials, cluster the crack distribution characteristic information of different materials on the top of the concrete building based on the crack distribution characteristic information statistics, and obtain the types of crack distribution characteristic information at different ambient temperatures on the top of the building and crack distribution characteristic information type detection conditions; the types of crack distribution characteristic information at different ambient temperatures on the top of the building include crack types, crack numbers and crack distribution density, and the crack distribution characteristic information type detection conditions include crack length and width distribution, crack spacing, crack depth, crack direction and angle, and crack morphology; The graph neural network model module is used to receive and upload the statistical crack distribution feature information types and crack distribution feature information type detection conditions of different ambient temperatures on the top of the building to the graph neural network model, and establish a physical crack distribution feature fluctuation detection algorithm and a chemical crack distribution feature fluctuation detection algorithm; The input of the physical crack distribution characteristic fluctuation detection algorithm is the statistical crack distribution characteristic information types of different ambient temperatures on the top of the building, and the output is the detection geometric characteristic parameters of the chemical crack distribution characteristic fluctuation detection algorithm; the input of the chemical crack distribution characteristic fluctuation detection algorithm is the crack distribution characteristic information type detection conditions and the determined detection geometric characteristic parameters, and the output is the building structure deformation type on the top of the concrete building; The detection geometric characteristic parameter module is used to receive the physical crack distribution characteristic fluctuation detection algorithm and the chemical crack distribution characteristic fluctuation detection algorithm; collect and count the types of crack distribution characteristic information at different ambient temperatures on the top of the building according to the building height, and input them into the physical crack distribution characteristic fluctuation detection algorithm to obtain the detection geometric characteristic parameters of the chemical crack distribution characteristic fluctuation detection algorithm; The building structure deformation analysis module is used to receive the detection geometric feature parameters of the chemical crack distribution characteristic fluctuation detection algorithm, and compare the detection geometric feature parameters of the chemical crack distribution characteristic fluctuation detection algorithm with the preset geometric feature size: When the size of the detected geometric feature parameter is smaller than the preset geometric feature size, there is no need to set the crack distribution feature information type detection condition, and the collection and analysis of the types of crack distribution feature information at different ambient temperatures on the top of the building shall continue; When the size of the detected geometric feature parameter is equal to or greater than the preset geometric feature size, the crack distribution feature information type detection condition is set, and the crack distribution feature information type detection condition and the detected geometric feature parameter are input into the chemical crack distribution feature fluctuation detection algorithm to obtain the building structure deformation type on the top of the concrete building.

[0049] The system collects the deformation types of different building structures and the crack distribution characteristics of different materials on the top of concrete buildings, and performs information fusion and graph neural network models to achieve comprehensive crack detection of the top of concrete buildings. Compared with the traditional method of crack detection that only detects crack type, crack number and crack distribution density parameters, the system considers more crack distribution characteristic information and provides a more comprehensive and comprehensive status assessment. The system uses a combination of physical crack distribution characteristic fluctuation detection algorithm and chemical crack distribution characteristic fluctuation detection algorithm to evaluate the status of the concrete building top. The physical crack distribution characteristic fluctuation detection algorithm collects and statistics the crack distribution characteristic information of the building top at different ambient temperatures according to the building height; while the chemical crack distribution characteristic fluctuation detection algorithm comprehensively considers the deformation type of building structures of different materials and the crack distribution characteristic information of different materials for evaluation, which is more accurate and reliable. By comparing the detection geometric feature parameters of the chemical crack distribution characteristic fluctuation detection algorithm with the preset geometric feature size, the structural deformation type of the top of the concrete building can be determined. Since the system comprehensively considers the structural deformation types of different materials and the crack distribution characteristic information of different materials, and uses a hierarchical evaluation method, it can improve the accuracy of judging the structural deformation type and reduce the possibility of misjudgment and missed judgment. In summary, this method utilizes a variety of crack distribution characteristic information and graph neural network technology to realize crack detection and building structure deformation monitoring of the top state of concrete buildings at different ambient temperatures, ensuring that the crack distribution characteristic information on the top of the concrete building is accurately detected according to the building height.

[0050] The various variations and specific embodiments of the method for detecting state cracks on the top of concrete buildings at different ambient temperatures in the aforementioned embodiment 1 are also applicable to the system for detecting state cracks on the top of concrete buildings at different ambient temperatures in this embodiment. Through the aforementioned detailed description of the method for detecting state cracks on the top of concrete buildings at different ambient temperatures, those skilled in the art can clearly understand the implementation method of the system for detecting state cracks on the top of concrete buildings at different ambient temperatures in this embodiment. Therefore, for the sake of brevity of the specification, it will not be described in detail here.

[0051] Although embodiments of the present invention have been shown and described, it will be understood by those skilled in the art that various equivalent changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A method for detecting cracks on the top surface of a concrete building based on dual-light fusion, characterized in that: The method comprises: Step A1, collecting dual-light fusion irradiation surface crack data of the concrete building top, wherein the dual-light fusion irradiation surface crack data includes deformation types of different concrete building tops and different materials of building structure and distribution characteristic information of cracks of different materials on the concrete building top within a preset time before the building structure deformation occurs; Step A2: clustering the crack distribution characteristic information of different materials on the top of the concrete building according to the crack distribution characteristic information statistics, and obtaining the crack distribution characteristic information types and crack distribution characteristic information type detection conditions at different ambient temperatures on the top of the building; Step A3, uploading the statistical crack distribution characteristic information types at different ambient temperatures on the top of the building, the crack distribution characteristic information type detection conditions and the corresponding deformation types of the concrete building top with different materials to the graph neural network model, and establishing a physical crack distribution characteristic fluctuation detection algorithm and a chemical crack distribution characteristic fluctuation detection algorithm; Step A4, collecting the types of crack distribution characteristic information of the top of the concrete building at different ambient temperatures according to the building height, and inputting them into the physical crack distribution characteristic fluctuation detection algorithm to obtain the detection geometric characteristic parameters of the chemical crack distribution characteristic fluctuation detection algorithm; Step A5, comparing the detection geometric feature parameters of the chemical crack distribution characteristic fluctuation detection algorithm with the preset geometric feature size: when the detection geometric feature parameter size is smaller than the preset geometric feature size, there is no need to set the crack distribution characteristic information type detection condition, and continue to collect and analyze the types of crack distribution characteristic information at different ambient temperatures on the top of the building; Step A6: When the size of the detected geometric feature parameter is equal to or greater than the preset geometric feature size, the crack distribution feature information type detection condition is set, and the crack distribution feature information type detection condition and the detected geometric feature parameter are input into the chemical crack distribution feature fluctuation detection algorithm to obtain the building structure deformation type of the top of the concrete building; The chemical crack distribution characteristic fluctuation detection algorithm is trained using a graph neural network model, and the expression is: Input feature matrix: ,in Indicates the crack feature point, Represents the characteristic dimensions of each node, including crack distribution characteristics, detection conditions, and building structure deformation types; Adjacency Matrix: , represents the connection relationship between nodes in the graph; Degree Matrix: , defined as , Normalized adjacency matrix: , used to balance the influence of nodes; The graph convolutional layer is defined as: ， in: Indicates The node feature matrix of the layer, ; and Respectively The weight matrix and bias vector of the layer; Assume that Layer graph convolution, the calculation of each layer is: ， Through multi-layer graph convolution, neighborhood information is gradually aggregated to extract deeper crack distribution features; After the last layer of graph convolution, a fully connected layer is used to map the features and convert them into the target output dimension; assuming the output is : ， Among them, Softmax is used for multi-classification problems to convert the output into a probability distribution; Use the cross-extraction loss function for supervised learning: ， in: Indicates The node in The true label of the class is 0 or 1, Indicates the number of categories; The overall model is expressed as: ， The model extracts features through multi-layer graph convolution, and then obtains the prediction results through a fully connected layer and Softmax activation.

2. A method for detecting cracks on the top surface of a concrete building based on dual light fusion as claimed in claim 1, characterized in that: The crack distribution characteristic information of different materials on the top of the concrete building includes the crack type, crack number, crack distribution density, crack length and width distribution, crack spacing, crack depth, crack direction and angle, and crack morphology of the top of the concrete building after acceptance; the crack type, crack number and crack distribution density are collected according to the building height, and the crack length and width distribution, crack spacing, crack depth, crack direction and angle, and crack morphology are collected according to the preset building height classification.

3. A method for detecting cracks on the top surface of a concrete building based on dual light fusion as claimed in claim 1, characterized in that: The types of crack distribution characteristic information at different ambient temperatures on the top of the building include crack type, crack number and crack distribution density, and the detection conditions of the crack distribution characteristic information types include crack length and width distribution, crack spacing, crack depth, crack direction and angle and crack morphology.

4. A method for detecting cracks on the top surface of a concrete building based on dual light fusion as claimed in claim 1, characterized in that: The input of the chemical crack distribution characteristic fluctuation detection algorithm is the crack distribution characteristic information type detection condition and the determined detection geometric characteristic parameters, and the output is the building structure deformation type of the top of the concrete building.

5. A method for detecting cracks on the top surface of a concrete building based on dual light fusion as claimed in claim 1, characterized in that: The preset influencing factors of the preset geometric feature size include: The influence of crack size on detection sensitivity, the influence of crack morphology, the influence of environmental factors, the influence of material properties, the limitation of detection equipment resolution, the computational efficiency of detection algorithm, the influence of crack development stage, noise interference, and the influence of detection area size.

6. A method for detecting cracks on the top surface of a concrete building based on dual light fusion as claimed in claim 1, characterized in that: in, The physical cracks include temperature stress cracks, shrinkage cracks, load stress cracks, and structural deformation cracks; The chemical cracks include alkali-aggregate reaction cracks, corrosive environment cracks, moisture intrusion cracks, and chemical erosion cracks.

7. A method for detecting cracks on the top surface of a concrete building based on dual light fusion as claimed in claim 1, characterized in that: The steps for establishing the physical crack distribution characteristic fluctuation detection algorithm are as follows: According to the surface crack data illuminated by dual-light fusion, the types of crack distribution characteristics of different materials on the top of concrete buildings and the corresponding building structure deformation type data are sorted out; the types of crack distribution characteristics of different ambient temperatures on the top of the building are processed and features are proposed, and the original data are converted into practical features for use in the graph neural network model; Utilizing the graph neural network model, according to the prepared ReLU function calculation data, the selected algorithm is used to calculate the model ReLU function, and the model parameters are adjusted; after the ReLU function calculation is completed, the relevant parameters of the model are extracted as the output of the physical crack distribution characteristic fluctuation detection algorithm.

8. A method for detecting cracks on the top surface of a concrete building based on dual light fusion as claimed in claim 1, characterized in that: The steps for establishing the chemical crack distribution characteristic fluctuation detection algorithm are as follows: According to the surface crack data illuminated by dual-light fusion, the crack distribution characteristic information type detection conditions and corresponding building structure deformation type data on the top of the concrete building are sorted out; The graph neural network model is used for training, and the crack distribution feature information type detection conditions are used as the edges of the graph neural network, and the building structure deformation type parameters are used as the nodes of the graph neural network. Relationships are connected, and the ReLU function is calculated for the model. After the ReLU function is calculated, the chemical crack distribution characteristic fluctuation detection algorithm is used to predict the building structure deformation type parameters on the top of the concrete building. The graph neural network is used to perform dual-light fusion irradiation surface crack data to calculate the correlation coefficient between the building structure deformation type parameters and the actual building structure deformation type.

9. A method for detecting cracks on the top surface of a concrete building based on dual light fusion as claimed in claim 1, characterized in that: The data collection method for detecting conditions of crack distribution characteristic information types includes: using a temperature monitoring device to monitor and collect the distribution of crack length and width on the top of a concrete building; monitoring and collecting the crack spacing on the top of a concrete building based on a building bearing capacity monitoring platform; monitoring and collecting the crack morphology on the top of the concrete building, and analyzing the peak value and fluctuation range of temperature changes; using a multimodal crack monitoring platform to monitor and collect crack depth signals generated on the top of the concrete building; using a crack image comparison interface to collect crack directions and angles around the top of the concrete building, and statistically analyzing the effects of different crack directions and angles.

Images